ICOLD - Twenty-Seventh International Congress On Large Dams - Vingt-Septième Congrès International Des Grands Barrages-CRC Press - Balkema (2022)

TWENTY-SEVENTH
INTERNATIONAL
CONGRESS
ON LARGE DAMS
VINGT-SEPTIÈME
CONGRÈS
INTERNATIONAL
DES GRANDS BARRAGES
27th May–3rd June 2022

MARSEILLE – FRANCE
AVERTISSEMENT – EXONERATION DE
RESPONSABILITE:
Tous droits réservés. Aucune partie de la présente publication ou des informations
contenues dans les présentes ne peut être reproduite, stockée dans un système de
récupération, ni transmise sous quelque forme ou par quelque moyen que ce soit,
électronique, mécanique, par photocopie, enregistrement ou autre, sans l’autorisation
préalable écrite de la CIGB.
Les informations, analyses et conclusions auxquelles cet ouvrage renvoie sont sous
la seule responsabilité de leur(s) auteur(s) respectif(s) cité(s).
Les informations, analyses et conclusions contenues dans cet ouvrage n’ont pas
force de Loi et ne doivent pas être considérées comme un substitut aux réglementations
officielles imposées par la Loi. Elles sont uniquement destinées à un public de
Professionnels Avertis, seuls aptes à en apprécier et à en déterminer la valeur et la portée
et à en appliquer avec précision les recommandations à chaque cas particulier.
Malgré tout le soin apporté à la rédaction de cet ouvrage, compte tenu de l’évolution
des techniques et de la science, nous ne pouvons en garantir l’exhaustivité.
Nous déclinons expressément toute responsabilité quant à l’interprétation et
l’application éventuelles (y compris les dommages éventuels en résultant ou liés) du
contenu de cet ouvrage.
En poursuivant la lecture de cet ouvrage, vous acceptez de façon expresse cette
condition.
NOTICE – DISCLAIMER:
All rights reserved. No part of this publication or the information contained herein may
be reproduced, stored in a retrieval system, or transmitted in any form or by any means,
electronic, mechanical, by photocopying, recording or otherwise, without written prior
permission from ICOLD.
The information, analyses and conclusions referred to herein are the sole responsibility
of the author(s) thereof.
The information, analyses and conclusions in this document have no legal force and
must not be considered as substituting for legally-enforceable official regulations. They
are intended for the use of experienced professionals who are alone equipped to judge
their pertinence and applicability and to apply accurately the recommendations to any
particular case.
This document has been drafted with the greatest care but, in view of the pace of
change in science and technology, we cannot guarantee that it covers all aspects of the
topics discussed.
We decline all responsibility whatsoever for how the information herein is interpreted
and used and will accept no liability for any loss or damage arising therefrom.
Do not read on unless you accept this disclaimer without reservation.
TRANSACTIONS
COMPTES RENDUS
Published by the
International Commission
On Large Dams
Publiés par
la Commission Internationale
des Grands Barrages
61, avenue Kléber, 75116 Paris, France
Tél.: 33 (0) 1 47 04 17 80 – Fax 33 (0) 1 53 75 18 22
http://www.icold-cigb.org
E-mail: contact@icold-cigb.org
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa
business
© 2022 ICOLD/CIGB, Paris, France
Typeset by MPS Limited, Chennai, India
Published by: CRC Press/Balkema

Schipholweg 107C, 2316 XC Leiden, The Netherlands
e-mail: enquiries@taylorandfrancis.com
www.routledge.com – www.taylorandfrancis.com
ISBN: 978-1-032-07848-9 (Hbk)

ISBN: 978-1-003-21180-8 (ebk)
DOI: 10.1201/9781003211808
TWENTY-SEVENTH CONGRESS ON LARGE DAMS
VINGT-SEPTIÈME CONGRÈS DES GRANDS BARRAGES
TRANSACTIONS COMPTES RENDUS
GENERAL TABLE OF CONTENTS TABLE GÉNÉRALE DES MATIÈRES
GENERAL SYNTHESIS / SYNTHÈSE GÉNÉRALE
Questions discussed at preceding Questions discutées aux précédents

Congresses Congrès
Questions for the 27th Congress Questions pour le 27ème Congrès
Papers and Communications Rapports et Communications
submitted présentés
1. Number per countries 1. Nombre par pays
2. Recap table per Congresses 2. Tableau récapitulatif par Congrès
3. Recap table of numbers per 3. Tableau récapitulatif des numéros
countries par pays
QUESTION 104
Wording of Question 104 Libellé de la Question 104

Table of Contents of Papers on Table des Matières des Rapports sur la
Question 104 Question 104
Papers on Question 104 Rapports sur la Question 104
General Report Question 104 Rapport Général Question 104
QUESTION 105

QUESTION 106

V
QUESTION 107

COMMUNICATIONS
Table of Contents of Communications Table des Matières des

Communications
Communications Communications
INNOVATION AWARDS / PRIX DE L’INNOVATION

Synthesis Synthèse
Table of Contents of Innovation Table des matières du Prix de
Awards l’Innovation
Reports Rapports
VI
GENERAL SYNTHESIS
TABLE OF CONTENTS
PAGE
Questions discussed at preceding Congresses VIII
Questions for the 27th Congress XVII
Recap table of numbers of Papers and Communications per countries XX
Recap table of Papers and Communications per Congresses XXI
Numbers of Papers and Communications per countries XXII
SYNTHÈSE GÉNÉRALE
TABLE DES MATIÈRES
PAGE
Questions discutées aux précédents Congrès VIII
Questions pour le 27ème Congrès XVII
Tableau récapitulatif du nombre de Rapports et Communications par pays XX
Tableau récapitulatif des Rapports et Communications par Congrès XXI
Numéros des Rapports et Communications par pays XXII
VII
QUESTIONS DISCUSSED QUESTIONS DISCUTÉES AUX
AT PRECEDING CONGRESSES PRÉCÉDENTS CONGRÉS
First Congress. Stockholm (Sweden), Premier Congrès. Stockholm (Suède),

1933: 1933 :
Question 1a - Deterioration by ageing of the Question 1a - Détérioration par
concrete of weight dams. vieillissement du béton des barrages-poids.
Question 1b - Influence of internal Question 1b - Questions relatives aux
temperature and distorsion of weight dams. influences de la température interne et à la
déformation des barrages-poids.
Question 2a - Research methods so as Question 2a - Méthodes de recherches
to ascertain whether a given materials is permettant de reconnaître si un matériau
suitable for being used in the construction donné est apte à être employé pour la
of an earthdam. construction d’un barrage en terre.
Question 2b - Study of physical laws Question 2b - Étude des lois physiques
governing infiltration of water through the commandant l’infiltration à travers un
dam and the subjacent soil barrage en terre et le sous-sol sous-jacent.
Second Congress. Washington (USA), Deuxième Congrès. Washington
1936: (États-Unis),1936 :
Question 3 - Special cements. Question 3 - Ciment spécial pour grands
barrages.
Question 4 - Design and waterproofing Question 4 - Constitution et étanchement
of shrinkage, contraction and expansion des joints de retrait, de contraction et
joints. dilatation
Question 5 - Study of facing of masonry and Question 5 - Étude des revêtements des
concrete dams. parements des barrages en maçonnerie ou
en béton.
Question 6 - Geotechnical studies of Question 6 - Études géotechniques des sols
foundations materials. de fondation.
Question 7 - Calculation of the stability of Question 7 - Calculs de stabilité des
earth dams. barrages en terre.
Third Congress. Stockholm (Sweden), Troisième Congrès. Stockholm (Suède),
1948: 1948 :
Question 8 - Uplift and resulting stresses in Question 8 - Exposé critique des mesures
dams. des sous-pressions et des contraintes en
résultant dans un barrage.
Question 9 - Methods and instruments for Question 9 - Méthodes de recherches et
measuring stresses and strains in earth and instruments pour mesurer les efforts et les
concrete dams. déformations dans les barrages en terre et
en béton.
Question 10 - Most recent dispositions to Question 10 - Les dispositions les plus
avoid piping. récentes pour éviter la formation de
renards.
VIII
Question 11 - Information obtained from Question 11 - Enseignements résultant de
the use of testing methods and of special l’utilisation des méthodes d’essai et de
cements in large dams. l’emploi des ciments spéciaux pour les
grands barrages.
Fourth Congress. New Delhi (India), Quatrième Congrès. New Delhi (Inde),
1951: 1951:
Question 12 - Methods for determining Question 12 - Méthodes pour déterminer
the maximum flood discharge that may be le débit de crue maximum qu’il est
expected at a dam and for which it should possible de prévoir pour un barrage et
be designed. Selection of type, capacity pour lequel le projet de barrage doit être
and general arrangements of temporary or établi. Choix du type et de la disposition
permanent outlets and spillways. générale des évacuateurs temporaires ou
permanents, et détermination de leur
capacité d’évacuation.
Question 13 - Design and construction of Question 13 - Projets de construction de
earth and rockfill dams with their core walls barrages en terre et en enrochements avec
and diaphragms. leurs écrans et diaphragmes.
Question 14 - Sedimentation in reservoirs Question 14 - Sédimentation dans les
and related problems réservoirs et problèmes connexes.
Question 15 - Concrete for large dams Question 15 - Béton pour grands barrages.
Fifth Congress. Paris (France), 1955: Cinquième Congrès. Paris (France), 1955:
Question 16 - Design and construction of Question 16 - Projets et construction de
dams on permeable soils and methods of barrages sur sols perméables et méthodes
foundation treatment. de traitement de la fondation.
Question 17 - Economics and safety of Question 17 - Bilan économique et sécurité
different types of concrete dams. des différents types de barrages en béton.
Question 18 - Settlement of dams due to Question 18 - Tassement des barrages
compressibility of the dams materials or of dû à la compressibilité des matériaux
the foundations soil, including earthquake constitutifs du barrage ou de la fondation,
problems y compris les questions liées aux
tremblements de terre.
Question 19 - The relation of the cements Question 19 - Effets du dosage en ciment
content of the concrete to performance in sur le comportement observé:
practice of:
a) gravity dams (interior and exterior); a) des barrages-poids (intérieur et
extérieur);
b) arch dams; b) des barrages-voûtes;
c) buttress dams c) des barrages à contreforts
and its influence on permeability and frost et influence de ce dosage sur la perméabilité
resistance. et la résistance au gel.
IX
Sixth Congress. New York (USA), 1958: Sixième Congrès. New York (USA), 1958:
Question 20 - Heightening of existing dams Question 20 - Surélévation de barrages
including methods of constructing new existants et méthodes pour la construction
dams in successive stages de nouveaux barrages par étapes
successives.
Question 21 - Observation of stresses Question 21 - Observations des contraintes
and deformations in dams and in their et des déformations dans les barrages,
foundations and abutments; and a dans leurs fondations et dans leurs appuis
comparison of these observations with latéraux. Comparaison de ces observations
computations and tests on small scale avec les calculs et les essais sur modèles
models. réduits.
Question 22 - Compaction methods and Question 22 - Méthode de compactage et
moisture content for materials used in the teneur en eau des matériaux employés dans
construction of earth core and supporting la construction du noyau en terre et du
fill for earth and rockfill dams. massif résistant des barrages en terre et en
enrochement.
Question 23 - Use of admixtures and Question 23 - Emploi des agents d’aération
pozzolanic materials in concrete for dams et des matériaux pouzzolaniques dans le
and the influence of the finer sand particles. béton destiné aux barrages et influence des
fines particules de sable.
Seventh Congress. Rome (Italy), 1961: Septième Congrès. Rome (Italie), 1961:
Question 24 - The selection, processing and Question 24 - Le choix, la préparation et
specification of aggregates for concrete for la spécification des agrégats dans le béton
large dams. pour grands barrages.
Question 25 - Underground work in Question 25 - Travaux souterrains se
connection with large dams. rapportant aux grands barrages.
Question 26 - Modern techniques of Question 26 - Techniques modernes
concrete dams for wide valleys and relatives aux barrages en béton pour larges
ancillary works. vallées et à leurs ouvrages accessoires.
Question 27 - Sealing of earth and rockfill Question 27 - Étanchement des barrages
dams with bitumen and other materials. en terre et enrochement par des produits
bitumineux et autres matières.
Eighth Congress. Edinburgh (Great Huitième Congrès. Edimbourg
Britain), 1964: (Grande-Bretagne), 1964:
Question 28 - Physical and mechanical Question 28 - Propriétés physiques et
properties of rock in situ, means of mécaniques du rocher in situ: procédés
determining these properties and improving permettant de déterminer ces propriétés
them, with special reference to the design et les améliorer, sous l’angle particulier
and construction of large dams. de l’établissement du projet et de la
construction des grands barrages.
X
Question 29 - Results and interpretation Question 29 - Résultats et interprétations
measurements made on large dams of all des mesures faites sur des barrages de toute
types, including earthquake observations. nature, y compris les observations sur les
tremblements de terre.
Question 30 - Design of concrete for large Question 30 - Études du béton des grands
dams of all types and influence of age on barrages de divers types et influence de
concrete properties. l’âge sur les propriétés du béton.
Question 31 - Design, methods of Question 31 - Étude, méthodes d’exécution
construction and performance of high et comportement des grands barrages en
rockfill dams (above or about 80 m). enrochement (de plus de, ou d’environ
80 m).
Ninth Congress. Istanbul (Turkey), Neuvième Congrès. Istamboul (Turquie),
1967: 1967:
Question 32 - The safety of dams from the Question 32 - Sécurité des barrages du
point of view of the foundations and the point de vue de la fondation et stabilité des
safety of reservoir banks. versants de la retenue.
Question 33 - Temporary and permanent Question 33 - Dispositions temporaires et
provisions for the control of flows. permanentes pour contrôler les apports et
le niveau de la retenue des barrages.
Question 34 - The behaviour and Question 34 - Comportement et
deterioration of dams. détérioration des barrages.
Question 35 - Dams in earthquake zones or Question 35 - Barrages dans des zones
other unfavourable situations. soumises aux tremblements de terre, ou
dans des situations exceptionnelles.
Tenth Congress. Montreal (Canada), Dixième Congrès. Montréal (Canada),
1970: 1970:
Question 36 - Recent developments in Question 36 - Progrès récents dans l’étude
the design and construction of earth and et la construction des barrages en terre et
rockfill dams. en enrochement.
Question 37 - Recent developments in Question 37 - Progrès récents dans
design and construction of dams and l’étude et la construction des barrages
reservoirs on deep alluvial, karstic, or other et des réservoirs situés sur des alluvions
unfavourable formations. profondes, sur des terrains karstiques ou
des terrains difficiles.
Question 38 - Supervision of dams and Question 38 - Contrôle des barrages et des
reservoirs in operation. réservoirs en exploitation.
Question 39 - Recent developments in the Question 39 - Progrès récents dans l’étude
design and construction of concrete dams. et la construction des barrages en béton.
XI
Eleventh Congress. Madrid (Spain), Onzième Congrès. Madrid (Espagne),
1973: 1973:
Question 40 - The consequences on the Question 40 - Conséquences de
environment of building dams. la construction des barrages sur
l’environnement.
Question 41 - Flow control and energy Question 41 - Contrôle des débits et de
control during construction and after la dissipation de l’énergie pendant la
completion. construction et après la mise en service.
Question 42 - Impervious elements and Question 42 - Dispositif d’étanchéité et
slope protection on earth and rockfill dams. protection des talus des barrages en terre
et des barrages en enrochement.
Question 43 - New ideas for more rapid and Question 43 - Idées nouvelles pour la
economic construction of concrete dams. construction plus rapide et plus économique
des barrages en béton.
Twelfth Congress. Mexico City (Mexico), Douzième Congrès. Mexico (Mexique),
1976: 1976:
Question 44 - Problems associated with Question 44 - Problèmes posés par les
special types of fill dams. barrages en remblai de type spécial
Question 45 - Leakage investigations and Question 45 - La recherche des fuites et le
drainage of dams and their foundations. drainage des barrages et de leur fondation
Question 46 - Preliminary planning of dam Question 46 - Études préliminaires
developments. d’aménagements de barrages.
Question 47 - The effects on dams and Question 47 - Les effets de quelques facteurs
reservoirs of some environmental factors. d’environnement sur les barrages et les
retenues.
Thirteenth Congress. New Delhi (India), Treizième Congrès. New Delhi (Inde),
1979: 1979:
Question 48 - Interface problems of dams. Question 48 - Problèmes de raccordement
dans les barrages.
Question 49 - Deterioration or failures of Question 49 - Détérioration ou ruptures de
dams. barrages.
Question 50 - Large capacity outlets and Question 50 - Vidanges et évacuateurs de
spillways. crue de grande capacité.
Question 51 - Seismicity and aseismic Question 51 - Résistance des barrages aux
design of dams. tremblements de terre
Fourteenth Congress. Rio de Janeiro Quatorzième Congrès. Rio de Janeiro,
(Brazil), 1982: (Brésil), 1982:
Question 52 - Safety of dams in operation. Question 52 - Sécurité des barrages en
service.
XII
Question 53 - Influence of geology and Question 53 - Influence de la géologie et
geotechnics on the design of dams. de la géotechnique sur la conception des
barrages.
Question 54 - Reservoir sedimentation and Question 54 - Alluvionnement des retenues
slope stability. Technical and environmental et stabilité de leurs versants. Conséquences
effects. techniques et effets sur l’environnement.
Question 55 - Materials and construction Question 55 - Matériaux et méthodes de
methods for embankment dams and construction des barrages et batardeaux en
cofferdams. remblai.
Fifteenth Congress. Lausanne Quinzième Congrès. Lausanne (Suisse),
(Switzerland), 1985: 1985:
Question 56 - Dam and foundation Question 56 - Auscultation des barrages et
monitoring. de leurs fondations.
Question 57 - Concrete dams - an old Question 57 : Barrages en béton - un
problem always present: cracking; a new problème ancien et toujours actuel: la
technology: rolled concrete (rollcrete). fissuration; une technique nouvelle: le
béton compacté au rouleau.
Question 58 - Foundation treatment of Question 58 - Traitement des fondations en
seepage. relation avec les infiltrations.
Question 59 - Rehabilitation of dam to Question 59 - Mesures pour renforcer la
ensure safety. sécurité des barrages en service.
Sixteenth Congress. San Francisco (USA), Seizième Congrès. San Francisco
1988: (États-Unis), 1988:
Question 60 - Reservoirs and the Question 60 - Retenues et environnement
environment - Experience in management - Expériences de gestion et de mesure
and monitoring. d’impact
Question 61 - Embankment dams: Question 61 - Barrages en remblai: organes
impervious elements other than clay cores. d’étanchéité autres que les noyaux en terre
Question 62 - New developments in the Question 62 - Progrès récents dans la
construction of concrete dams. construction des barrages en béton.
Question 63 - Design flood and operational Question 63 - Crue de projet et maîtrise des
flood control. crues après mise en service du barrage.
Seventeenth Congress. Vienna (Austria), Dix-septième Congrès. Vienne (Autriche),
1991: 1991:
Question 64 - Environmental issues in dam Question 64 - Les barrages et
projects. l’environnement.
Question 65 - Ageing of dams and remedial Question 65 - Vieillissement des barrages et
measures. méthodes de réparation.
Question 66 - Dams on difficult Question 66 - Barrages sur fondations
foundations. difficiles.
XIII
Question 67 - New developments for fill Question 67 – Progrès récents concernant
dams and fill cofferdams. les barrages et batardeaux en remblai
Eighteenth Congress. Durban (South Dix-huitième Congrès. Durban (Afrique
Africa), 1994: du Sud), 1994
Question 68 - Safety assessment and Question 68: Évaluation et renforcement de
improvement of existing dams. la sécurité des barrages en service.
Question 69 - Environmental experience Question 69 - Retenues en exploitation:
gained from reservoirs in operation. expérience dans le domaine de
l’environnement.
Question 70 - Staged construction, raising Question 70 - Construction par étapes,
or modification of dams. surélévation ou modification de barrages
Question 71 - Deterioration of spillways Question 71 - Détérioration des ouvrages
and outlet works.
Nineteenth Congress. Florence (Italy), Dix-neuvième Congrès. Florence (Italie),
1997: 1997:
Question 72 - Innovative financing of Question 72 - Méthodes nouvelles de
projects involving dams. financement des projets de barrages.
Question 73 - Special problems with Question 73 - Problèmes particuliers
earthfill dams. relatifs aux barrages en terre.
Question 74 - Performance of reservoirs. Question 74 - Comportement des retenues.
Question 75 - Incidents and failures of dams Question 75 - Accidents et ruptures de
barrages.
Twentieth Congress. Beijing (Chine), Vingtième Congrès. Beijing (Chine),
2000: 2000:
Question 76 – The use of risk analysis Question 76 – L’utilisation de l’analyse des
to support dams safety decisions and risques dans le processus de décision relatif
management. à la sûreté des barrages et à leur gestion.
Question 77 – Benefits and concerns about Question 77 – Bienfaits et préoccupations
dams. associés aux barrages.
Question 78 - Monitoring of dams and their Question 78 – Auscultation des barrages et
foundation. de leur fondation.
Question 79 - Gated spillways and other Question 79 – Évacuateurs de crue vannés,
controlled release facilities and dam safety. autres organes d’évacuation contrôlés et
sûreté des barrages
Twenty-First Congress. Montreal Vingtième-et-unième Congrès. Montréal
(Canada), 2003: (Canada), 2003:
Question 80 - Financing hydraulic projects Question 80 - Financement des projets
including dams. hydrauliques comprenant des barrages
XIV
Question 81 – Economic evaluation of Question 81 – Évaluation économique des
hydraulic projects including dams. aménagements hydrauliques comprenant
des barrages.
Question 82 – Ageing and rehabilitation Question 82 – Vieillissement et
of concrete and masonry dams and réhabilitation des ouvrages en béton
appurtenant works. et en maçonnerie: barrages et ouvrages
associés.
Question 83 - Seismic aspects of dams. Question 83 - Aspects sismiques relatifs aux
barrages.
Twenty Second Congress. Barcelona Vingt-deuxième Congrès. Barcelone
(Spain), 2006: (Espagne), 2006:
Question 84 - Technical solutions to reduce Question 84 – Solutions techniques
time and costs in dam design and favorisant la réduction des délais et des
construction. coûts de conception et de réalisation des
barrages.
Question 85 - Management of the Question 85 - Gestion des impacts en aval
downstream impacts of dam operation. des barrages en service.
Question 86 - Safety of earth- and rockfill Question 86 - Sécurité des barrages en terre
dams. et enrochement.
Question 87 - Flood and drought evaluation Question 87 - Évaluation et gestion des
and management. situations de crues, ou des étiages dus aux
sécheresses.
Twenty Third Congress. Brasilia Vingt-troisième Congrès. Brasilia
(Brazil), 2009: (Brésil), 2009:
Question 88 - Dams and hydropower. Question 88 - Barrages et hydroélectricité.
Question 89 - Management of siltation in Question 89 - Gestion de la sédimentation
existing and new reservoirs. des réservoirs existants ou nouveaux.
Question 90 - Upgrading of existing dams. Question 90 - Amélioration de barrages
existants.
Question 91 - Dam safety management. Question 91 - Sécurité des barrages.
Twenty Fourth Congress. Kyoto (Japan), Vingt quatrième Congrès.Kyoto (Japon),
2012: 2012:
Question 92 - Environmental friendly Question 92 - Techniques respectueuses
techniques for dams and reservoirs. de l’environnement pour les barrages et
réservoirs.
Question 93 – Safety. Question 93 – Sécurité.
Question 94 - Flood discharge. Question 94 - Évacuation des crues.
Question 95 - Ageing and upgrading. Question 95 - Vieillissement et
améliorations.
XV
Twenty Fifth Congress. Stavanger Vingt cinquième Congrès. Stavanger
(Norway), 2015: (Norvège), 2015:
Question 96 – Innovation in utilisation of Question 96 – Innovation dans l’utilisation
dams and reservoirs. des barrages et des réservoirs.
Question 97 – Spillway. Question 97 – Évacuateurs de crue.
Question 98 – Embankments and Question 98 – Barrage en remblai et
tailings dams. barrage de stériles.
Question 99 – Upgrading and Question 99 – Réhabilitation et réingénierie
re-engineering of existing dams. des barrages existants
Twenty Sixth Congress. Vienna Vingt sixième Congrès. Vienne (Autriche),
(Austria), 2018: 2018:
Question 100 - Reservoir sedimentation and Question 100 - Alluvionnement des
sustainable development réservoirs et développement durable
Question 101 - Safety and risk analysis Question 101 - Sécurité et analyse de risque
Question 102 - Geology and dams Question 102 - Géologie et barrages
Question 103 - Small dams and levees Question 103 - Petits barrages et digues
XVI
QUESTIONS QUESTIONS
FOR THE 27th CONGRESS POUR LE 27éme CONGRÈS
The 87th Annual Meeting of the La 87ème Réunion Annuelle de la

International Commission on Large Dams, Commission Internationale des Grands
held in Ottawa (Canada), in June 2019 Barrages, qui s’est tenue à Ottawa
selected the following four questions and (Canada) en Juin 2019, a décidé de choisir
question contents for the 27th International pour le 27ème Congrès International des
Congress on Large Dams. Grands Barrages les quatre questions
suivantes, accompagnées de commentaires
explicatifs.
QUESTION 104
CONCRETE DAM DESIGN INNOVATION DANS LA

INNOVATION AND CONCEPTION ET PERFORMANCE
PERFORMANCE DANS LES BARRAGES EN BÉTON
1. Innovations for arch dam analysis, 1. Innovations pour l’analyse, la

design and construction; including RCC conception et la construction des barrages
arch and arch-gravity dams. voûtes; y compris les barrages voûtes en
BCR et les barrages poids-voûtes.
2. Innovations for design, construction 2. Innovations relatives à la conception, au
materials and placement methods, flood choix des matériaux, aux méthodes de mise
management during construction and en place, au contrôle des crues pendant la
performance of concrete dams, including construction et performance des barrages
RCC and cemented material dams. en béton, y compris en BCR et en matériaux
cimentés.
3. Innovations for raising existing concrete 3. Innovations pour la surélévation des
dams. barrages en béton existants.
4. Innovations for extremely high concrete 4. Innovation pour les barrages en béton de
dams. hauteur extrême.
5. Operational performance of concrete 5. Performance opérationnelle des
dams during the life cycle, including under barrages en béton pendant l’ensemble
extreme conditions. du cycle de vie, y compris en conditions
extrêmes.
XVII
QUESTION 105
INCIDENTS AND ACCIDENTS INCIDENTS ET ACCIDENTS

CONCERNING DAMS CONCERNANT LES BARRAGES
1. Recent lessons from incidents and 1. Leçons récentes tirées des incidents et
accidents concerning dams during the life accidents de barrages durant tout le cycle
cycle, including during construction. de vie, y compris en construction.
2. Evaluation of the flows and flood, 2. Évaluation des débits et de l’inondation,
estimation and quantification of the estimation et quantification des
consequences, including social, economic conséquences en cas de rupture ou
and environmental aspects, in case of d’incidents, y compris du point de vue
failure or incidents social, économique et environnemental.
3. Emergency planning: regulation, 3. Plans d’urgence: réglementation,
organization, information of the population organisation, information des populations
and examples of implementation. et exemples de mise en œuvre.
4. Governance of the safety: definition 4. Gouvernance de la sécurité: définition
of the responsibilities, periodic reviews, des responsabilités, vérifications
implementation tests, organization of périodiques, essais de mise en œuvre,
lessons learned implementation. organisation du retour d’expérience.
QUESTION 106
SURVEILLANCE, SURVEILLANCE,
INSTRUMENTATION, INSTRUMENTATION,
MONITORING AND DATA AUSCULTATION ET ACQUISITION
ACQUISITION DES DONNÉES
1. Long term performance of existing 1. Performances à long terme des systèmes
surveillance systems including reliability de surveillance, y compris leur fiabilité et
and accuracy; importance of visual leur précision; importance des inspections
inspections. visuelles.
2. New technologies in dam and 2. Nouvelles technologies pour
foundation instrumentation and l’instrumentation et l’auscultation des
monitoring. barrages et des fondations.
3. Data acquisition and processing to 3. Acquisition et traitement de données
evaluate the behavior of dams, predict and pour évaluer le comportement des
identify incidents. barrages, prévoir et identifier les incidents.
4. Understanding and handling of large 4. Analyse et traitement de grandes
quantity of data, including artificial quantités d’informations, y compris
intelligence approach. l’approche par l’intelligence artificielle.
XVIII
QUESTION 107
DAMS AND CLIMATE CHANGE BARRAGES ET CHANGEMENT

CLIMATIQUE
1. Impacts of climatic change on existing 1. Impacts du changement climatique sur
dams and reservoirs and remedies; case les barrages et réservoirs existants et
studies and costs. adaptation correspondante : cas d’études et
coûts.
2. Impacts of climatic change on needs 2. Impacts du changement climatique sur
and designs of dams, reservoirs and levees les besoins et la conception de nouveaux
(water storage, floods mitigation, oceans barrages, réservoirs et digues (stockage
raising…). d’eau, contrôle des crues, montée des
océans…).
3. Favorable impacts of dams on climatic 3. Impacts favorables des barrages sur
change, including greenhouse gases le changement climatique, y compris
reduction by optimization of hydroelectric la réduction de gaz à effet de serre
production. Needs, potential and cost of par l’optimisation de la production
energy pumped storage. hydraulique. Besoins, potentiel et coût du
stockage d’énergie par pompage
4. Unfavorable impacts of dams and 4. Impacts défavorables des barrages et
reservoirs on climatic change: evaluation of réservoirs sur le changement climatique, y
greenhouse gases emissions by reservoirs compris bilan des émissions de gaz à effet
and dam construction. de serre par les réservoirs et la construction
des barrages.
XIX
PAPERS AND RAPPORTS ET
COMMUNICATIONS SUBMITTED COMMUNICATIONS PRÉSENTÉS
1) Number of Papers and Communications by alphabetical order of countries/

Nombre de Rapports et Communications par ordre alphabétique des pays
Countries Nbr of Papers Nbr of Comm. Total of Papers & Comm.

Argentina 1 1
Australia 1 1
Austria 1 1
Brazil 3 1 4
Bulgaria 1 1
Burkina Faso 1 1
Canada 14 14
China 15 15
Colombia 1 1
Czech. Rep. 1 1
France 24 3 27
Greece 1 1
India 9 9
Indonesia 13 1 14
Iran 2 2
Iraq 1 1
Italy 12 12
Japan 9 3 12
Laos 1 1
Macedonia 1 1
Morocco 3 3
Netherlands 2 2
Norway 3 3
Peru 2 2
Portugal 2 2
Romania 3 3
Slovakia 2 2
South Africa 4 4
Spain 5 5
Sweden 6 6
Switzerland 14 3 17
UK 4 4
USA 7 1 8
33 countries - Total 168 13 181
XX
2) Recap table per Congresses/Tableau récapitulatif par Congrès
Individuals General Number of
Papers Papers countries (*)
Congress Rapports Rapports de Nombre de
Congrés particuliers Communications Synthése Total pays (*)
10th - Montreal 204 14 13 231 34
(1970)
11th - Madrid 217 19 13 249 39
(1973)
12th - Mexico 190 21 12 223 38
(1976)
13th - Delhi 214 15 10 239 38
(1979)
14th - Rio 244 18 12 274 42
(1982)
15th - Lausanne 268 17 10 295 46
(1985)
16th - San 254 33 287 38
Francisco (1988)
17th - Vienna 275 19 294 39
(1991)
18th - Durban 233 28 261 41
(1994)
19th - Florence 212 32 244 44
(1997)
20th - Beijing 247 15 262 40
(2000)
21th - Montreal 203 8 211 32
(2003)
22nd - Barcelona 245 13 258 42
(2006)
23th - Brasilia 190 3 193 31
(2009)
24th - Kyoto 160 13 173 27
(2012)
25th - Stavanger 182 5 187 32
(2015)
26th - Vienna 220 7 227 36
(2018)
27th - Marseille 168 13 181 33
(2022)
XXI
3) Numbers of Papers and Communications per countries / Numéros des
Rapports et Communications par pays
Countries Q. 104 Q. 105 Q. 106 Q. 107 C. Total

Argentina 5 1
Australia 15 1
Austria 45 1
Brazil 43, 44 44 12 4
Bulgaria 25 1
Burkina Faso 50 1
Canada 19 23, 24, 25, 19, 20, 21 12, 13, 14 14
26, 27, 28,
46
China 27, 28, 29, 33, 34, 48, 32, 33 20, 21, 22 15
30, 31, 32 49
Colombia 36 1
Czech. Rep. 12 1
France 15, 16, 17, 17, 18, 19, 11, 12, 13, 6, 7, 8, 1, 2, 3 27
18, 40 20, 21, 22 14, 15, 16, 9, 10, 11
17
Greece 2 1
India 43, 44, 45 51 31, 32, 33, 9
34, 35
Indonesia 38, 39 39, 40, 41, 40, 42 25, 26, 27, 10 14
47 28, 29
Iran 1, 2 2
Iraq 13 1
Italy 3, 4, 5, 2, 3 1, 2, 3, 1 12
6, 7 4
Japan 21, 22 26, 27, 28, 17, 18 5, 6, 7 12
29, 30
Laos 1 1
Macedonia 14 1
Morocco 24 32 31 3
Netherlands 4 5 2
Norway 42 8, 45 3
Peru 8 31 2
Portugal 6 8 2
Romania 42 43 30 3
Slovakia 7 4 2
South Africa 23 29, 30 19 4
Spain 9, 10 6, 7 3 5
Sweden 20 22, 23, 24, 16 6
25
Switzerland 33, 34, 36 35, 36, 37, 34, 35, 36, 23, 24 4, 8, 9 17
38 37, 38
UK 13, 14 15, 16 4
USA 11 9, 10, 11, 5 11 8
12, 13
33 countries Total 41 Total 51 Total 40 Total 36 Total 13 Total 181
XXII
QUESTION 104
QUESTION 104
XXIII
QUESTION
104
CONCRETE DAMS DESIGN INNOVATION AND PERFORMANCE
a. Innovations for arch dam analysis, design and construction; including RCC
arch and arch- gravity dams.
b. Innovations for design, construction materials and placement methods,

flood management during construction and performance of concrete dams,
including RCC and cemented material dams.
c. Innovations for raising existing concrete dams.
d. Innovations for extremely high concrete dams.
e. Operational performance of concrete dams during the life cycle, including

under extreme conditions.
INNOVATIONS DANS LA CONCEPTION ET LA PERFORMANCE DES

BARRAGES EN BÉTON
a. Innovations pour l’analyse, la conception et la construction des
barrages-voûtes; y compris les barrages-voûtes en BCR et les barrages
poids-voûtes
b. Innovations relatives à la conception, au choix des matériaux, aux méthodes

de mise en place, au contrôle des crues pendant la construction et
performance des barrages en béton, y compris en BCR et en matériaux
cimentés.
c. Innovations pour la surélévation des barrages en béton existants.
d. Innovations pour les barrages en béton de hauteur extrême.
e. Performance opérationnelle des barrages en béton pendant l’ensemble du

cycle de vie, y compris en conditions extrêmes
XXIV
QUESTION
104
TABLE OF CONTENTS TABLE DES MATIÈRES
OF PAPERS DES RAPPORTS
R. 1 ALI NOORZAD, ALIREZA DANESHYAR, PAYAM SOTOUDEH,

MOHSEN GHAEMIAN (Iran)
Nonlinear response of foundation rock in seismic simulation of
concrete gravity dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
R. 2 ALI NOORZAD, PAYAM SOTOUDEH, ALIREZA DANESHYAR,
MOHSEN GHAEMIAN (Iran)
Numerical simulation of surface amplification due to
inclined harmonic excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
R. 3 FRANCESCO FORNARI, MICHELE CANCI, GABRIELLA
VASCHETTI, ALBERTO SCUERO, DANIELE CAZZUFFI (Italy)
Maintaining safe operation of concrete dams at high altitudes:
Lago Nero, Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
R. 4 MIGUEL MIRANDA, GABRIELLA VASCHETTI, ALBERTO
SCUERO (Italy)
Geomembranes in very high concrete dams: challenges and solutions . . . . . . 37
R. 5 SOPHIE CHARLOTTE KUHLMANN, ENRICO TITA, MARIA
VITTORIA VIGNOLI (Italy)
Barrage de Kruth-Wildenstein rehabilitation of the asphalt lining . . . . . . . . . . 54
R. 6 G. PIETRANGELI, G. PITTALIS, G. SIMONELLI, P. ZAFFARONI,
(Italy)
Stress and deformation analysis of the concrete faced rockfill saddle
dam at GERDP and design of peripheral joint system . . . . . . . . . . . . . . . . . . . . 70
R. 7 G. PIETRANGELI, A. BEZZI, P. MASTROFINI, A. MASCIOTTA (Italy)
Stress-strain characterization of RCC mixes at GERD project and
thermal - seismic dam behavior analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
R. 8 JOSÉ MENDIVIL RIVAS, MICHEL LINO, JOHN H. DUQUE,
JOHN E. YOUNG (Peru)
Asana river diversion dam design and construction . . . . . . . . . . . . . . . . . . . . . 109
R. 9 MANUEL G. DE MEMBRILLERA, MARGARITA PINTOS,
ALFONSO SANTA, ARY PAULO RODRIGUES, CLAUDIO MICHEL
NAHAS, FRANCISCO HOLANDA (Spain)
Singularities in the design of “Casupá dam” . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
R. 10 FRANCISCO ORTEGA (Spain)
Immersion vibrated RCC - innovation and performance case study:
The 103 m high Enciso dam in Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
XXV
R. 11 ANIL K. CHOPRA (United States of America)
Earthquake analysis of arch dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
R. 12 ONDØEJ HRAZDIRA, JIØÍ ŠVANCARA (Czech Republic)
Securing the Orlík dam against the impacts of extreme floods . . . . . . . . . . . 158
R. 13 M.R.H. DUNSTAN, K. TIREITO, J. FUKUWATARI, R.S. RUPRA
(United Kingdom)
Mwache RCC dam – the advantages of an early and extensive trial
mix programme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
R. 14 M.R.H. DUNSTAN (United Kingdom)
Speed of construction: the greatest advantage of RCC dams . . . . . . . . . . . . . 189
R. 15 EMMANUEL ROBBE, FREDERIC ANDRIAN, NICOLAS ULRICH,
CLAIRE JOUY (France)
Cisaillement à l’interface béton-rocher des barrages voûtes : retour
d’expérience de voûtes existantes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
R. 16 ABDELGHANI SI CHAIB, MATHIEU ROY, VINCENT BOINAY,
JEAN-CHRISTOPHE GIRARD, NICOLAS ULRICH, FRÉDÉRIC
ANDRIAN (France)
Conception et retour d’expérience sur les voûtes épaisses en BCR . . . . . . . 249
R. 17 FRÉDÉRIC ANDRIAN, EDOUARD MINE, QUENTIN BERCHER,
JEAN-LOUIS CERVETTI, GEOFFREY MATHIEU (France)
Retour d’expérience sur la conception et la technique de construction
de quelques barrages en BCR de grande hauteur . . . . . . . . . . . . . . . . . . . . . . . 270
R. 18 ETIENNE GRIMAL, PHILIPPE KOLMAYER, KATIA LALICHE,
CATHERINE CASTEIGTS, ROMAIN GIUNTI, CHRISTINE NORET
(France)
La modélisation du gonflement des bétons : une aide précieuse pour
la gestion ou la réhabilitation d’un ouvrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
R. 19 ALI RASEKH (Canada)
Estimating floor design spectra at the top of a dam using dam response
spectrum analysis results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
R. 20 RICHARD MALM, LISA BROBERG, JONAS ENZELL, JOHAN
BLOMDAHL, CARL-OSCAR NILSSON (Sweden)
Predicting the measured behavior and defining warning levels of a
concrete dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
R. 21 NORIKAZU YAMASHITA, TOSHIHISA KASE, SUBARU
TACHIHANA, SHINOBU MORIYAMA, HIDEO MORI (Japan)
Dam concrete automatic placement system – complete automation of
dam concrete materials supply, production, transport, and placement . . . . . 338
R. 22 YUSHI AOSAKA & SHOJI TSUTSUI, TATSUYA KAWATA &
BASSAM EGILAT (Japan)
Investigation, measures and controls against internal restraint cracks
due to thermal stress of RCC dam – practice at Nam Ngiep 1
hydropower project in Lao Pdr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
XXVI
R. 23 Q.H.W. SHAW (South Africa)
Developments in concrete dam engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
R. 24 ABDELILAH BOUKAIDI LAGHZAOUI, MOULAY LHASSAN
FAIK, AHMED TISSIR, AHMED ROJDAM (Maroc)
Adaptation de la formulation et de la mise en œuvre du BCR du
barrage Toudgha . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
R. 25 ANTON TZENKOV, DIMITAR KISLIAKOV, OGNYAN TODOROV
(Bulgaria)
Strengthening of Beli Iskar concrete gravity dam . . . . . . . . . . . . . . . . . . . . . . . 400
R. 26 REPORT MOVED TO Q.105
R. 27 WORKING GROUP OF CHINESE AUTHORS (China)
A case study of effective application of innovative management
on RCC dam construction at Kafue Gorge lower hydroelectric
project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
R. 28 WORKING GROUP OF CHINESE AUTHORS (China)
Application of Rotec belt conveyor equipment at Kafue Gorge lower
project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
R. 29 YUANGUANG LIU, QINGGUO ZHOU, YUANBO GAO, JUNJIE JIN
(China)
Application of RCC technical innovation at Kafue Gorge lower
hydropower project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
R. 30 JUNJIE JIN, QINGGUO ZHOU, SHUNCAI NING, YUANGUANG
LIU (China)
Successful application of RCC interlayer joint bonding technology with
surface long time exposed in dry and high temperature areas . . . . . . . . . . . . 440
R. 31 ZHONG ZHOU, JING ZHANG, LIJUN XUE, DEWEN CAI,
ZHONGXU LIU (China)
safety evaluation of Jinping I arch dam considering the left abutment
slope deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
R. 32 XUE LIJUN, DUAN SHAOHUI, ZHAO YONGGANG, WANG RUI,
SHEN MANBIN (China)
Foundation treatments of Jinping I arch dam . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
R. 33 MARC BALISSAT, JÉRÔME FILLIEZ, ANDRES FANKHAUSER
(Switzerland)
Decommissioning of an arch-gravity dam and measures taken for its
replacement by a double curvature arch dam . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
R. 34 MARCELO LEITE RIBEIRO, OLIVIER VALLOTTON,
ALEXANDRE WOHNLICH (Switzerland)
Two recent cases of arch dam raising, lessons learnt and innovation . . . . . . 482
R. 35 REPORT MOVED TO COMMUNICATIONS
R. 36 PATRICE DROZ, ALEXANDRE WOHNLICH (Switzerland)
Dam safety along dam lifetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
XXVII
R. 38 ANOM PRASETIO, ABDUL MUIS, PAMRIH PAMMU (Indonesia)
Monitoring performance of concrete dam after earthquake . . . . . . . . . . . . . . 524
R. 39 AHMAD SIDIK, DIMAZ I.V, STELLA M (Indonesia)
Application of modular precast concrete system on dam sidewall . . . . . . . . 544
R. 40 LUC DEROO, ETIENNE FROSSARD, FRANÇOIS LEMPERIERE
(France)
BCR-E & E-BCR : barrages hybrides BCR – enrochements. Principes
et conditions d’application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559
R. 42 S. BJØNNESS, N.RAKSTAD, V. KRATHE, T. KONOW (Norway)
Calibration of Fe-model with measured behavior of an existing concrete
arch dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
R. 43 Y.K. CHAUBEY, SANKHADIP CHOWDHURY, MUKESH BHORIA
(India)
Heightening of Dhanikhari concrete dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
R. 44 SANJEEV GUPTA, NAGARAJ, HARSH BHASKAR MEHTA &
S.P. BANSAL (India)
Operational performance and sediment handling by abrasion resistance
steel liner and rails in under-sluice spillways of Nathpa dam, India . . . . . . . 605
R. 45 S.P. BANSAL, RAKESH SEHGAL, REVATI RAMAN &
SURJEET SINGH (India)
Hydraulic design of barrage on Himalyan Bouldery river- with specific
reference to Sjvn’s Naitwar Mori Hep (60 mw), India . . . . . . . . . . . . . . . . . . . 624
General Report / Rapport Général Q. 104

RAFAEL IBÁÑEZ-DE-ALDECOA (Spain), General Reporter /
Rapporteur Général . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
XXVIII
PAPERS ON
QUESTION 104
RAPPORTS SUR LA
QUESTION 104
XXIX
Q.104 - R.1
COMMISSION INTERNATIONALE
DES GRANDES BARRAGES
-------
VINGT SEPTIÈME CONGRÈS
MARSEILLE, JUIN 2022
-------
NONLINEAR RESPONSE OF FOUNDATION ROCK IN SEISMIC SIMULATION

OF CONCRETE GRAVITY DAMS (∗)
Ali NOORZAD
VP, ICOLD & Faculty of Civil, Water and Environmental Engineering,
SHAHID BEHESHTI UNIVERSITY, Tehran, Iran
Alireza DANESHYAR
PhD, Civil Engineering Department,
SHARIF UNIVERSITY OF TECHNOLOGY, Tehran, Iran
Payam SOTOUDEH
PhD candidate, Civil Engineering Department,
Mohsen GHAEMIAN
Professor, Civil Engineering Department,
IRAN
SUMMARY
Seismic simulation of concrete gravity dams considering nonlinear response

of foundation rock has been addressed infrequently. However, it has been observed
that the dam-foundation interface is highly susceptible to cracking due to its geo-
metrical properties. Mathematically speaking, seismic waves induce singular stress
fields around sharp corners of the interface such that material integrity is locally
degraded, even under small excitations. This degradation can be extended to the
∗ Réponse non linéaire de la roche de fondation dans la simulation sismique des barrages-
poids en béton
DOI: 10.1201/9781003211808-1 1
Q.104 - R.1
foundation rock and affects structural behavior of the whole system. Thus, a sys-
tem of dam-foundation-reservoir considering nonlinear responses for both dam and
foundation, subjected to earthquake loading is analyzed here. Foundation mass is
included in the finite element model of the system in order to resemble wave propa-
gation effects within the half-space foundation medium. A similar problem with linear
response of foundation rock is also simulated for the sake of comparison.
RÉSUMÉ
La simulation sismique des barrages-poids en béton, tenant compte de la

réponse non linéaire de la roche de fondation, n’a été abordée que rarement. Toute-
fois, il a été observé que l’interface barrage-fondation est très sensible à la fissuration
en raison de ses propriétés géométriques. Mathématiquement parlant, les ondes
sismiques induisent des champs de contraintes singuliers autour des angles aigus
de l’interface de sorte que l’intégrité du matériau est localement dégradée, même
sous de petites excitations. Cette dégradation peut s’étendre à la roche de fondation
et affecte le comportement structurel de tout le système. C’est ainsi qu’un système
barrage-fondation-réservoir, considérant des réponses non linéaires pour le barrage
et la fondation, soumis à une charge sismique, est analysé ici. La masse de fonda-
tion est incluse dans le modèle par éléments finis du système afin de ressembler aux
effets de propagation des ondes dans le milieu de fondation en demi-espace. Un
problème similaire avec la réponse linéaire de la roche de fondation est également
simulé à des fins de comparaison.
1. INTRODUCTION
Earthquake analysis of concrete gravity dams is mostly performed under

the assumption of elastic foundation. However, simulations show that the dam-
foundation interface is highly prone to cracking. Assuming elastic response for
foundation prevents crack propagation within this medium and forces the crack to
propagate in a straight direction at the interface. This study aims to consider nonlin-
ear response of foundation to compare the resulted crack trajectory with the crack
profile of conventional approach, i.e. the elastic assumption for foundation response.
Nonlinear behavior of concrete has drawn ample attention from numer-

ous researchers. Considering concrete’s nonlinear behavior can be accomplished
through linear elastic fracture mechanics [1–3]. A number of other methods such
as nonlinear fracture mechanics [4–6], fictitious crack model [7], crack band model
[8], smeared crack approach [9], elastic [10] and elasto-plastic damage model [11]
are also presented by a number of researchers. Using smeared crack model, Bhat-
tacharjee and Leger [12] analyzed the cracking of a concrete dam. They found that
between coaxial rotating crack model and fixed crack with variable shear resistance,
the former produces more accurate results, due to its changing crack direction. Wepf
et al. [13] performed a seismic analysis on a concrete gravity dam using fictitious
2
Q.104 - R.1
crack model. Pekau et al. [2] and Pekau and Batta [3] work on utilizing linear fracture
mechanics in conjunction with boundary element method for crack analysis of con-
crete gravity dams is also of interest. Using discrete crack method Skrikerud and
Bachman [14] investigated the two dimensional cracking of a concrete gravity dam.
The crack growth direction was selected in line with plane of principal tensile stress
in their work. Cervera et al. [10] proposed a damage model with rate dependent
properties and different damage indices for compression and tension. Their results
indicated that damage distribution in the concrete dam body is highly sensitive to the
strain rate of concrete. Lee and Fenves [11] proposed a plastic damage formulation
and a constitutive model [15] for nonlinear seismic analysis of concrete structures
such as dams. To account for closing and reopening of crack joints, damage param-
eters for tension and compression differ in their model. Calayir and Karaton [16]
employed second order damage tensor to carry out the seismic analysis of a con-
crete dam. In a study of comparing methods, Pan et al. [17] compared three different
methods of nonlinear analysis of arch dams namely, cohesive cracks, isotropic dam-
age and Drucker-Prager plasticity model. Their results showed that damage models
are most reliable methods among the investigated methods of their study. By using
extended finite element method, Zhang et al. [18] analyzed a concrete gravity dam
under earthquake excitation.
Plastic damage models however, have received more attention due to their
superiority over other methods in aspects of correct reproduction of stiffness degra-
dation and permanent deformation. A correct damage plastic model should be
capable of considering various tensile or compressional behaviors of concrete,
reopening or closing of cracks in the concrete, the volumetric expansion under
compression and anisotropic behavior caused by the concrete damage.
In this paper, the novel damage-plastic method of Daneshyar and Ghaemian

[19] is employed to correctly replicate concrete’s nonlinear behavior. The capability
of the method to reproduce anisotropy, induced by damage, is so well-captured that
it enables the method to provide directional degenerations as well as permanent
deformations due to excessive concrete straining. In this method, by the use of
directional damage growth history, induced anisotropy is explained. Since two dif-
ferent states of damage for tension and compression are considered in this method,
different behavior of concrete in tension and compression is captured and reopening
or closing of cracks is also accounted for in the model.
2. CONTINUUM DAMAGE FORMULATION
Anisotropic damage-plastic model of Daneshyar and Ghaemian [19] is used

to resemble material degradation within the dam and foundation. Both stiffness
reduction due to micro-cracking, and permanent deformations resulted by slippage
of micro-crack faces are considered in a rate-dependent and direction-dependent
regime. Robustness of formulation in reproducing valid nonlinear response of con-
crete in seismic simulations is demonstrated in [20]. The model, which is formulated
3
Q.104 - R.1
based on effective surfaces of Kachanov [21], maps nominal stress tensor into the
effective space as
σ̄ij = Mijkl σkl (1)
where σ̄ij is stress tensor in the effective space, and Mijkl is the so-called mapping
tensor, defined as + + − −
Mijkl = Mijrs Prskl + Mijrs Prskl (2)
+ − +
with Mijkl and Mijkl are the tensile and compressive parts, respectively, and Pijkl
−
and Pijkl are stress projection tensors which are used to decompose total stress into
+
tensile and compressive parts. Positive projection tensor Pijkl and negative projection
−
tensor Pijkl are defined as

3
+ (r) (r) (r) (r)
Pijkl = H (σr )ni nj nk nl (3)
r=1
− +
Pijkl = Iijkl − Pijkl (4)
where Iijkl is the forth-order identity tensor as follows
1
Iijkl = δik δjl + δjk δil (5)
2
and δij is the Kronecker delta, such that

1 i=j
δij = (6)
0 i = j
+ −
Mapping tensors Mijkl and Mijkl are obtained by
± 1 ±
Mijkl = ψik δjl + ψjk± δil (7)
2
where ψij+ and ψij− are the inverse integrity tensors, respectively, which are resulted
by scalar degradations along different directions, such that

3
ψik± = ψ ± ni nj d (8)
2π
thus, an isotropic description of damage is constructed. Degradation along different

directions can be obtained by employing the empirical definition of Jirásek [22] as
follows −0.5
ψ ± = e0± + e± exp e± /e0± e0± (9)
where e± is the strain along the corresponding direction, and e0± is a material
constant.
Concrete behavior is susceptible to rate of loading, i.e. the fracture energy

and tensile strength are changed due to rate effects. Experimental observations of
4
Q.104 - R.1
Bischoff and Perry [23] and Brara and Klepaczko [24] affirm this response. This
behavior has prominent effect on concrete response under seismic condition, and
has to be considered in order to achieve acceptable results. To this end, viscous
model of Duvaut-Lions [25] is employed to enhance the formulation, such that
1
ε̄˙ ij = C̄ijkl (σ̄kl − σ̃kl )
vp
(10)
η
where η is the relaxation time in second, σ̃ij is the rate-independent stress tensor, σ̄ij
is the effective stress tensor, and C̄ijkl is the compliance matrix of material, defined as
−1
C̄ijkl = λ̄δij δkl + µ̄(δik δjl + δil δjk ) (11)
and λ̄ and µ̄ are the Lame constants of virgin material.
3. SEISMIC SIMULATION
Koyna concrete gravity dam is chosen for our investigation. Finite element
mesh of the body and near-field foundation are depicted in Fig. 1. Material proper-
ties are also presented in Table 1. The dam-foundation-reservoir system is excited
Fig. 1
Finite element mesh of Koyna gravity dam.
Maillage par éléments finis du barrage-poids de Koyna.
5
Q.104 - R.1
Table 1
Material properties of dam, foundation and reservoir
YOUNG’S BULK POISSON’S TENSILE E0 RELAXATION DAMPING
MODULUS MODULUS RATIO STRENGTH (−) TIME (%)
(GPA) (GPA) (−) (MPA) (MS)
dam 30.0 – 0.20 2.41 40 × 10−6 7 10
foundation 22.4 – 0.33 1.80 40 × 10−6 7 10
reservoir – 2.07 – – – – –
Fig. 2
Horizontal acceleration of Koyna earthquake.
Accélération horizontale du tremblement de terre de Koyna.
Fig. 3
Crack profile at t = 0 s (left), t = 4s (middle), and t = 8 s (right) for linear foundation.
Profil de fissure à t = 0 s (gauche), t = 4 s (milieu) et t = 8 s (droite) pour une
fondation linéaire.
using the first 10 seconds of koyna earthquake, which the horizontal acceleration
is presented in Fig. 2, under the massless foundation assumption. Two different
conditions are assumed for the system. First, the body of dam is considered to be
nonlinear, while linear elastic response is assumed for the foundation. Next, both
dam and foundation are considered to behave based on the presented damage
formulation.
6
Q.104 - R.1
Fig. 4
Crack profile at t = 0 s (left), t = 4 s (middle), and t = 8 s (right) for nonlinear
foundation.
Profil de fissure à t = 0 s (gauche), t = 4 s (milieu) et t = 8 s (droite) pour une
fondation non linéaire.
Crack profiles of the first case, i.e. the linear assumption for foundation rock,
in different stages of simulation are presented in Fig. 3. Results of modeling with
nonlinear assumption for foundation are also reported in Fig. 4. As it is evident in
the contours of modeling with linear response of foundation, lower crack is forced
to propagate along the dam-foundation interface. On the other hand, crack propa-
gates within the foundation rock in the second case and different trajectory, and also
different overall response for the whole system is obtained.
REFERENCES
[1] AYARI ML, SAOUMA VE. A fracture mechanics based seismic analysis of
concrete gravity dams using discrete cracks. Engineering Fracture Mechanics.
1990 Jan 1;35(1–3):587–98.
[2] PEKAU OA, CHUHAN Z, LINGMIN F. Seismic fracture analysis of concrete

gravity dams. Earthquake engineering & structural dynamics. 1991;20(4):
335–54.
[3] PEKAU OA, BATTA V. Seismic crack propagation analysis of concrete struc-
tures using boundary elements. International journal for numerical methods in
engineering. 1992 Nov 15;35(8):1547–64.
[4] EL-AIDI B, HALL JF. Non-linear earthquake response of concrete gravity

dams part 1: modelling. Earthquake engineering & structural dynamics. 1989
Aug;18(6):837–51.
7
Q.104 - R.1
[5] BHATTACHARJEE SS, LEGER P. Seismic cracking and energy dissipation in

concrete gravity dams. Earthquake Engineering & Structural Dynamics. 1993
Nov;22(11):991–1007.
[6] GUANGLUN W, PEKAU OA, CHUHAN Z, SHAOMIN W. Seismic fracture

analysis of concrete gravity dams based on nonlinear fracture mechanics.
Engineering Fracture Mechanics. 2000 Jan 1;65(1):67–87.
[7] HILLERBORG A, MODÉER M, PETERSSON PE. Analysis of crack forma-

tion and crack growth in concrete by means of fracture mechanics and finite
elements. Cement and concrete research. 1976 Nov 1;6(6):773–81.
[8] BAŽANT ZP, OH BH. Crack band theory for fracture of concrete. Matériaux et
construction. 1983 May 1;16(3):155–77.
[9] MIRZABOZORG H, GHAEMIAN M. Non-linear behavior of mass concrete in

three-dimensional problems using a smeared crack approach. Earthquake
engineering & structural dynamics. 2005 Mar;34(3):247–69.
[10] CERVERA M, OLIVER J, MANZOLI O. A rate-dependent isotropic damage

model for the seismic analysis of concrete dams. Earthquake engineering &
structural dynamics. 1996 Sep;25(9):987–1010.
[11] LEE J, FENVES GL. A plastic-damage concrete model for earthquake analy-
sis of dams. Earthquake engineering & structural dynamics. 1998 Sep;27(9):
937–56.
[12] BHATTACHARJEE SS, LÉGER P. Application of NLFM models to predict

cracking in concrete gravity dams. Journal of Structural Engineering. 1994
Apr;120(4):1255–71.
[13] WEPF DH, FELTRIN G, BACHMANN H. Influence of time-domain dam–

reservoir interaction on cracking of concrete gravity DAMS. Earthquake
engineering & structural dynamics. 1993 Jul;22(7):573–82.
[14] SKRIKERUD PE, BACHMANN H. Discrete crack modelling for dynamically

loaded, unreinforced concrete structures. Earthquake engineering & structural
dynamics. 1986 Mar;14(2):297–315.
[15] LEE J, FENVES GL. Plastic-damage model for cyclic loading of concrete
structures. Journal of engineering mechanics. 1998 Aug;124(8):892–900.
[16] CALAYIR Y, KARATON M. A continuum damage concrete model for earth-

quake analysis of concrete gravity dam–reservoir systems. Soil Dynamics and
Earthquake Engineering. 2005 Dec 1;25(11):857–69.
[17] PAN J, ZHANG C, XU Y, JIN F. A comparative study of the different procedures

for seismic cracking analysis of concrete dams. Soil Dynamics and Earthquake
Engineering. 2011 Nov 1;31(11):1594–606.
8
Q.104 - R.1
[18] ZHANG S, WANG G, YU X. Seismic cracking analysis of concrete gravity

dams with initial cracks using the extended finite element method. Engineering
Structures. 2013 Nov 1;56:528–43.
[19] DANESHYAR A, GHAEMIAN M. Coupling microplane-based damage and

continuum plasticity models for analysis of damage-induced anisotropy in plain
concrete. International Journal of Plasticity. 2017 Aug 1;95:216–50.
[20] DANESHYAR A, AND GHAEMIAN M. “Seismic analysis of arch dams using

anisotropic damage-plastic model for concrete with coupled adhesive-frictional
joints response.” Soil Dynamics and Earthquake Engineering 125 (2019):
105735.
[21] KACHANOV, L. M. “On creep rupture time.” Izv. Acad. Nauk SSSR, Otd. Techn.
Nauk 8 (1958): 26–31.
[22] JIRÁSEK, MILAN. “Comments on microplane theory.” Mechanics of quasi-

brittle materials and structures (1999): 55–77.
[23] BISCHOFF, P. H., AND S. H. PERRY. “Compressive behaviour of concrete at

high strain rates.” Materials and structures 24.6 (1991): 425–450.
[24] BRARA, A., AND J. R. KLEPACZKO. “Experimental characterization of

concrete in dynamic tension.” Mechanics of materials 38.3 (2006): 253–267.
[25] DUVAUT, GEORGES, AND JACQUES LOUIS LIONS. “Les inéquations en

mécanique et en physique.” (1972).
9
Q.104 - R.2
-------
-------
NUMERICAL SIMULATION OF SURFACE AMPLIFICATION DUE TO INCLINED

HARMONIC EXCITATION (∗)
Ali NOORZAD
VP, ICOLD & Faculty of Civil, Water and Environmental Engineering,
SHAHID BEHESHTI UNIVERSITY, Tehran
Payam SOTOUDEH
PhD candidate, Civil Engineering Department,
SHARIF UNIVERSITY OF TECHNOLOGY, Tehran
Alireza DANESHYAR
PhD, Civil Engineering Department,
Mohsen GHAEMIAN
Professor, Civil Engineering Department,
IRAN
SUMMARY
In seismic analysis of complex structures such as concrete dams, considering

wave propagation effects plays an influential role in correct estimation of stress levels
in the dam body. When the concrete dam is of arch type, these effects become even
more important. When excitation waves strike upon valley of the arch dam, they
reflect. These reflected waves construct with incoming excitation waves and cause
surface amplifications. In this study, a half-space with a cylindrical valley is analyzed
under inclined harmonic SH wave. To analyze the problem, a method, capable of
∗ Simulation numérique de l’amplification de surface due à une excitation harmonique
inclinée
10 DOI: 10.1201/9781003211808-2
Q.104 - R.2
considering wave propagation effects, is utilized. Multiple cases, with different angles
of incidence, are considered for investigation of surface amplification. The results
are also verified with analytical solutions indicating the accuracy of the implemented
method.
RÉSUMÉ
Dans l’analyse sismique de structures complexes telles que les barrages

en béton, la prise en compte des effets de propagation des vagues joue un rôle
déterminant dans l’estimation correcte des niveaux de contrainte dans le corps
du barrage. Lorsque le barrage en béton est de type voûte, ces effets devien-
nent encore plus importants. Lorsque les ondes d’excitation frappent la vallée d’un
barrage-voûte, elles se reflètent. Ces ondes réfléchies, combinées avec les ondes
d’excitation entrantes, provoquent des amplifications de surface. Cette étude anal-
yse un demi-espace avec une vallée cylindrique soumise à une onde de cisaillement
horizontale harmonique inclinée. Pour analyser le problème, une méthode, capable
de considérer les effets de propagation des ondes, est utilisée. Plusieurs cas, avec
différents angles d’incidence, sont considérés pour l’étude de l’amplification de sur-
face. Les résultats sont également vérifiés avec des solutions analytiques indiquant
la précision de la méthode mise en œuvre.
1. INTRODUCTION
Irregularities within massed medium have significant effects on amplification

of earthquake waves. This condition becomes more prominent in realistic inclined
propagation of earthquake waves. Despite the importance, this phenomenon is
mostly disregarded in seismic simulations, due to lack of a comprehensive method-
ology for dealing with irregular topologies and inclined wave propagation problem
simultaneously. In this paper, dynamic analysis of a semi-cylindrical canyon is per-
formed under propagating SH wave with various angles of incidence, using domain
reduction method. The results of the analysis are compared with analytical solutions
of the problem.
Assessing effects of topography on the responses obtained from the medium,

Mossessian and Dravinski [1-2] used indirect boundary integral method to obtain
amplification values of elastic waves due to the presence of a three dimensional
valley. Luco et al. [3] calculated the response of a uniform, infinite long canyon with
arbtrirary shape due to plane elastic waves. Zhang and Chopra [4] presented a
direct boundary element method to obtain the seismic response of an infinite long
prismatic valley which has arbitrary shape in a homogeneous half-space. Zhao and
Valliappan [5] studied the effects of canyon topography and geological condition on
ground motion due to P and SV earthquake waves. Sesma and Campillo [6] investi-
gated the effects of topography on propagation of incident P, SV and Rayleigh waves
11
Q.104 - R.2
using an integral representation. Sesma and Luzon [7] obtained the response of a
valley for incident P, S and Rayleigh waves. Paolucci [8] estimated the amplification
of earthquake motion due to steep geological irregularities using spectral element
method. Dravinski [9] used indirect boundary integral approach to study the scat-
tering of waves by an arbitrary-shaped basin in an elastic half-space. Tari nejad
et al. [10] studied the effects of topographic amplification of a canyon using a three
dimensional boundary element method.
2. DOMAIN REDUCTION METHOD
Domain reduction method (DRM) is a two-step formulation to obtain the

seismic response of any structure with any condition. This method is capable of
accounting for wave propagation effects in any environment. In the first step, free-
field response of the foundation is calculated. These forces are applied to a single
layer of elements in the second step to perform the full analysis. An unbounded
foundation with superstructure can be shown as Fig. 1.
In DRM, source of excitation is inside the model and therefore a simple bound-
ary condition on the sides of the foundation could effectively absorb the scattering
waves emanating toward these boundaries. Because of simplicity and easy imple-
mentation, Lysmer’s boundary condition is employed in this paper as the absorbing
boundary.
The problem at hand is represented in Fig. 2. + is the foundation boundary

and is a desired virtual boundary that effective seismic forces are applied on.
Introducing these boundaries results in two domains namely, and + and 3 sets
of nodal displacement by the names of ue, ub and ui. Subscripts e, b and i refer to
external, boundary and internal displacements.
Fig. 1
Truncated region containing superstructure.
Région tronquée contenant la superstructure.
12
Q.104 - R.2
Fig. 2
Domains of the problem.
Domaines du problème.
The domain is governed by Navier’s equation of motion. Partitioned finite

element equations of motion for each domain is represented as:

Mii Mib üi Cii
Cib üi Kii Kib ui 0
+ + = in (1)
Mbi
Mbb üb
Cbi
Cbb üb Kbi
Kbb ub Pb

+
Mbb
Mbe
+
üb
Cbb
+
Cbe
+
−Pb üb
Kbb
+
Kbe
+
ub
+ + = in +
Meb+ M + üe + C +
Ceb üe + K +
Keb ue Pe
ee ee ee
(2)
where M, C and K are mass, damping and stiffness matrices respectively. Adding
Eq. 1 and Eq. 2 gives the equation for the whole domain. To calculate effective
seismic forces, an auxiliary problem is solved in which the domain is replaced by
0 which has the same properties as + (Fig 3).
The equation of motion in + is represented as

+ + + + + +

Mbb
Mbe üb0
Cbb
Cbe üb0
Kbb
Kbe ub0 −Pb0
+ + + + + + + + = in +

Meb
Mee üe0
Ceb
Cee üe0
Keb
Kee ue0 Pe
(3)
Where superscript “0” refers to free-field value of the parameter obtained from
auxiliary problem (Fig. 3). Extracting Pe from Eq. 3, we have
+ + + + + +
Pe = Meb
0
üb + Mee
0
üe + Ceb
0
üb + Cee
0
üe + Keb
0
ub + Kee
0
ue (4)
13
Q.104 - R.2
Fig. 3
(a) Auxiliray problem (b) two sub domains.
(a) Problème secondaire (b) deux sous-domaines.
Consider following decomposition:
ue = ue0 + we (5)
where we is the relative displacement with respect to free-filed motion. By substi-

tuting Eq. 5 into summation of Eq. 1 and Eq. 2, new main equation of the system is
obtained as follow:
⎡ ⎤⎧ ⎫ ⎡ ⎤⎧ ⎫
Mii Mib 0 ⎪
⎨ üi ⎪ Cii
⎢ ⎥ ⎬ Cib 0 ⎨ üi ⎬
⎢M + M + ⎥ ⎢ + C + ⎥ ü
⎣ bi Mbb + Mbb be ⎦⎪ üb ⎪ + ⎣Cbi Cbb + Cbb

be ⎦⎩ b ⎭
+ + ⎩ ⎭ 0 +
Ceb +
Cee ẅe
0 M M ẅe
eb ee
⎡ ⎤⎧ ⎫ ⎧ ⎫
Kii Kib 0 ⎨ ui ⎬ ⎪
⎨ 0 ⎪
⎬
⎢ + + ⎥ + ü0 − C + ü0 − K + u0
+ ⎣Kbi + K
Kbb
Kbe ⎦ ub = −Mbe
bb ⎩ ⎭ ⎪ ⎩
e be e be e ⎪
⎭
0 Keb+
Kee
+ we + + +
Pe − M ü0 − C ü0 − K u0
ee e ee e ee e
(6)
Substitute Eq. 4 into Eq. 6 and the final form of effective seismic forces which
should be applied in the second step is obtained
⎧ eff ⎫ ⎧ ⎫
⎪
⎪ P ⎪
⎪ ⎪
⎪ 0 ⎪
⎪
⎨ i ⎬ ⎨ ⎬
+ + +
P eff = Pbeff = −Mbe ü0 − C ü0 − K u0 (7)
⎪
⎪ ⎪
⎪ ⎪
⎪
e be e be e ⎪
⎩ eff ⎭ ⎩ M + ü0 + C + ü0 + K + u0 ⎪
⎭
Pe eb b eb b eb b
As it can be observed from Eq. 7, subscript indices are either be or eb which

causes the matrices to have zero values except on one single layer of elements
bounded by b and e nodes.
To summarize, for the first step the auxiliary problem needs to be solved to
obtain the effective seismic forces on all nodes of elements bounded by and e
14
Q.104 - R.2
boundaries (Fig. 4). For the second step, effective forces of first step are applied at
their respective nodes. (Fig 5).
3. PLANE HARMONIC SH WAVE PROPAGATION
Consider a semi-cylindrical canyon within a half-space as shown in Fig. 6.

A cylindrical canyon with radius a has been removed from this half-space. The
half-space is isotropic, elastic and homogeny eous. The material properties of the
medium are represented by µ as elastic modulus and β as shear wave velocity.
Fig. 4
First step of the DRM model.
Première étape du modèle de données.
Fig. 5
Second step of the DRM model.
Deuxième étape du modèle de données.
15
Q.104 - R.2
Two different coordinate systems are utilized. Both Cartesian and cylindrical
coordinates have the same origin as shown in the figure. The excitation in the half-
space, ui, consists of an infinite train of plane SH waves with frequency ω. The
motion in the z-direction can be written as

x y
ui = exp iω t − + (8)
cx cy
For incident angle of γ , cx and cy are defined as phase velocities along x and
y axes and are calculated as follows
β
cx = (9)
sin(γ )
β
cy = (10)
cos(γ )
In the absence of a canyon the incident motion reflects from the free-surface.
Incident and reflected waves construct and the resulting free-field motion of the
half-space results as

x ωy
ui + ur = 2 exp iω t − cos (11)
cx cy
Using Fast Fourier Transform and by employing this frequency response of

the motion one can find the time history of response for any desired point in the
half-space.
By introducing the canyon, a third group of waves namely, scattering waves,

become present in the medium. These third group of waves are denoted by uR.
Fig. 6
A cylindrical canyon within a half-space.
Canyon cylindrique dans un demi-espace.
16
Q.104 - R.2
Representing Eq. 11 in cylindrical form, the response of the free-field half-space

can be rewritten as
∞

ui + ur = 2J0 (κr) + 4 (−1)n J2n (κr) cos 2nγ cos 2nθ
n=1
∞

− 4i (−1)n J2n+1 (κr) sin (2n + 1)γ sin (2n + 1)θ (12)
0
Where Ji(x) is the Bessel function of first kind and κ is the wave number which is the
result of dividing the frequency of motion by velocity of propagation wave. Solving
Navier’s equation of motion for the problem of a cylindrical valley in a half-space with
its proper boundary conditions, one can calculate uR as follows
∞
(2) (2)
uR = [an H2n (κr) cos 2nθ + bn H2n+1 (κr) sin (2n + 1)θ ] (13)
n=0
Where Hi is the Hankel function of second kind. Unknown coefficients an and bn for
n = 0 are obtained as
J1 (κa)
a0 = −2 (2)
(14)
H1 (κa)
κaJ0 (κa) − J1 (κa)
b0 = 4i sin γ (2) (2)
(15)
κaH0 (κa) − H1 (κa)
Fig. 7
Comparison on numerical analytical results for vertical propagation.
Comparaison des résultats analytiques et numériques pour la propagation
verticale.
17
Q.104 - R.2
Fig. 8
Comparison on numerical analytical results for incident angle of 30 degrees.
Comparaison des résultats analytiques et numériques pour un angle d’incidence
de 30 degrés.
And for larger values of n, we have
κaJ2n−1 (κa) − 2nJ2n (κa)

an = −4(−1)n cos 2nγ (2) (2)
(16)
κaH2n−1 (κa) − 2nH2n (κa)
κaJ2n (κa) − (2n + 1)J2n+1 (κa)

bn = 4i(−1)n sin (2n + 1)γ (2) (2)
(17)
κaH2n (κa) − (2n + 1)H2n+1 (κa)
Having the unknown coefficients, one can easily calculate the total response
of the medium (u) with cylindrical canyon by adding Eq. 12 and Eq. 13.
Surface amplitude, which is the normal value of the total motion, can be
calculated as follows
ū = [(Re(u))2 + (Im(u))2 ]0.5 (18)
To facilitate further comparisons, a dimensionless parameter known as

dimensionless frequency is introduced as
2a
η= (19)
λ
Where a is cylindrical canyon radius and λ is the wavelength of the excitation.
4. RESULTS AND CONCLUSION
Numerical results for different incident angles are compared next to analytical
data to validate the method. To this end, maximum amplitude of different locations
18
Q.104 - R.2
Fig. 9
de 60 degrés.
Fig. 10
de 90 degrés.
on the surface of the half-space is chosen for the comparison. First, numerical
and analytical results of vertical propagation are reported in Fig. 7. A very good
agreements between the curves is obtained.
Similar comparison for incident angle of 30 degrees is presented in Fig. 8.

Again, maximum amplitude of numerical model is in close agreement with analytical
data. It is also shown that more amplitude is reported for the left of canyon.
Maximum amplitudes of numerical and analytical models for the incident angle
of 60 degrees are also reported in Fig 9. Similar trend is also evident in this figure.
19
Q.104 - R.2
Finally, the comparison is made again for horizontal propagation, i.e. the inci-
dent angle of 90 degrees. Maximum amplitudes, which are presented in Fig. 10 are
in close agreements. In addition, amplitudes drop below 1 for the right side of the
canyon.
REFERENCES
[1] MOSSESSIAN, TOMI K., AND MARIJAN DRAVINSKI. “Amplification of elastic

waves by a three dimensional valley. Part 1: steady state response.” Earth-
quake engineering & structural dynamics 19.5 (1990): 667–680.
[2] MOSSESSIAN, TOMI K., AND MARIJAN DRAVINSKI. “Amplification of elastic

waves by a three dimensional valley. Part 2: transient response.” Earthquake
engineering & structural dynamics 19.5 (1990): 681–691.
[3] LUCO, J. E., H. L. WONG, AND F. C. P. DE BARROS. “Three-dimensional

response of a cylindrical canyon in a layered half-space.” Earthquake engi-
neering & structural dynamics 19.6 (1990): 799–817.
[4] ZHANG, LIPING, AND ANIL K. CHOPRA. “Three-dimensional analysis of spa-

tially varying ground motions around a uniform canyon in a homogeneous
half-space.” Earthquake engineering & structural dynamics 20.10 (1991):
911–926.
[5] ZHAO, CHONGBIN, AND S. VALLIAPPAN. “Seismic wave scattering effects

under different canyon topographic and geological conditions.” Soil Dynamics
and Earthquake Engineering 12.3 (1993): 129–143.
[6] SÁNCHEZ-SESMA, FRANCISCO J., AND MICHEL CAMPILLO. “Topographic

effects for incident P, SV and Rayleigh waves.” Tectonophysics 218.1-3 (1993):
113–125.
[7] SANCHEZ-SESMA, FRANCISCO J., AND FRANCISCO LUZON. “Seismic

response of three-dimensional alluvial valleys for incident P, S, and Rayleigh
waves.” Bulletin of the Seismological Society of America 85.1 (1995): 269–284.
[8] PAOLUCCI, ROBERTO. “Amplification of earthquake ground motion by

steep topographic irregularities.” Earthquake engineering & structural dynam-
ics 31.10 (2002): 1831–1853.
[9] DRAVINSKI, MARIJAN. “Scattering of elastic waves by a general anisotropic

basin. Part 2: a 3D model.” Earthquake engineering & structural dynamics 32.5
(2003): 653–670.
[10] TARI NEJAD R., MOHAMMAD T. AHMADI, AND NASER KHAJI. “Analysis of
topographic amplification effects on canyon sites using 3D boundary element
method.” (2007): 25–37.
20
Q.104 - R.3
-------
-------
MAINTAINING SAFE OPERATION OF CONCRETE DAMS AT HIGH

ALTITUDES: LAGO NERO, ITALY (∗)
Francesco FORNARI, Michele CANCI

ENEL Green Power
Gabriella VASCHETTI, Alberto SCUERO

Carpi Tech
Daniele CAZZUFFI
CESI
ITALY
SUMMARY
Lago Nero is a concrete gravity dam located in the Italian Alps. Like in many
dams at high altitude, the harsh mountain climate with frequent freeze/thaw cycles,
low temperatures, thick ice and snow, and the action of soft alpine waters, was
detrimental for concrete durability. Repeated repairs were carried out over the years
to stop leakage occurring at the upstream face and at foundations. Since such
repairs did not solve the problem ENEL Company, the owner of the dam, in 1979
undertook overall rehabilitation works at the upstream face and at the foundations.
Based on pioneer projects and on a previous successful experience on another

dam at high altitude owned by ENEL, an exposed PVC geomembrane was adopted
as new upstream water barrier, connected to a grout curtain injected from the new
plinth. For the first time in the world, a tensioning system developed by Carpi was
adopted to anchor the geomembrane to the dam face.
∗ Exploitation en sécurité de barrages en béton en altitude: LAGO NERO, Italie
DOI: 10.1201/9781003211808-3 21
Q.104 - R.3
The overall rehabilitation works were carried out in three campaigns, from
1979 to 1981. Since its installation, the exposed geomembrane system is being
monitored for leakage performance and for geomembrane’s ageing. In 2020, after
40 years’ service, the system is fully performing, and a long remaining service life
is predicted.
The paper describes the problems, the selected solutions, the works, and the
results of monitoring carried out in 40 years.
RÉSUMÉ
Lago Nero est un barrage-poids en béton situé dans les Alpes italiennes.
Comme dans de nombreux barrages en altitude, le rigoureux climat de montagne
avec de fréquents cycles de gel/dégel, des températures très basses, de grandes
épaisseurs de glace et neige, et l’action des eaux alpines douces, ont nui à la
durabilité du béton. Des réparations répétées ont été effectuées au fil du temps pour
arrêter les fuites sur le parement amont et au niveau des fondations. Ces réparations
n’ayant pas résolu le problème, ENEL, le Maître d’Ouvrage, a entrepris en 1979 des
travaux de réhabilitation générale du parement amont et des fondations.
Sur la base de projets pionniers et d’une précédente expérience positive

sur un autre barrage en altitude appartenant à ENEL, une géomembrane PVC
exposée a été adoptée comme nouvel écran d’étanchéité en amont, reliée à un
rideau d’injections réalisé à partir d’une nouvelle plinthe. Pour la première fois dans
le monde, un système de tensionnement développé par Carpi a été adopté pour
ancrer la géomembrane au parement du barrage.
Les travaux de réhabilitation globale ont été réalisés en trois campagnes,

de 1979 à 1981. Depuis son installation, le système de géomembrane exposée
est surveillé pour sa performance en termes de fuites et pour le vieillissement de
la géomembrane. En 2020, après 40 ans de service, le système est pleinement
performant et une longue durée de vie résiduelle est prévue.
Le document décrit les problèmes, les solutions retenues, les travaux et les
résultats du suivi réalisé pendant 40 ans.
1. INTRODUCTION
Lago Nero dam is a massive concrete gravity dam completed in 1929, used
for hydropower and now owned by Enel Green Power Company, a branch of Enel
Company, the Italian National Power Board. The dam is located in the Italian Alps
at high altitude, in a harsh climate with frequent freeze/thaw cycles, temperatures
from −24◦ C to +26◦ C, and ice thickness up to 70 cm. Such climate conditions, and
the action of soft alpine waters, are critical for concrete durability, and many dams
at high altitudes experience problems that often require remedial measures.
At Lago Nero dam, traditional repairs were made over the years to mitigate
water seepage occurring at foundations and at the upstream face, but the adopted
22
Q.104 - R.3
measures, including complete refacing of the dam with reinforced gunite, were not
effective in the long term. In 1975 Enel decided to plan extensive rehabilitation works
aiming to stop water infiltration through the upstream face and protect the concrete
from further deterioration, to restore watertightness of the foundation, and to improve
the static safety of the dam. The paper discusses the overall rehabilitation works,
with focus on the watertight liner system selected for the dam face, which for those
years was quite innovative, and gives information on the monitoring campaign that
Enel has been carrying out in the 40 years that have elapsed since 1980, to assess
the performance of the waterproofing system and the conditions of the exposed
geomembrane.
2. BACKGROUND
2.1. DAM DATA
Lago Nero dam is a massive concrete gravity dam 38 m high and 144.20 m
long at crest, at elevation 2025.15 m in the Italian Alps. The upstream face has a
total surface of 3,500 m2 , the reservoir formed by the dam has a capacity of 3,48
million m3 . The dam, built in the years 1924 to 1929, on a bedrock formed by quartz
porphyry free of schistosity and with moderate diaclasis, has an upstream peripheral
concrete block, vertical drains, and a drainage gallery along the bottom perimeter.
Fig. 1
Cross section of Lago Nero dam.
Coupe transversale du barrage de Lago Nero.
23
Q.104 - R.3
The main structure of the dam was formed by a concrete mixture consisting
of 250 kg of Portland cement per cubic meter of concrete mix, to which rocks were
added during pouring of concrete, in a ratio of 10% of the volume of the mix; for
the part of the dam upstream of the draining system, the mixture was made with
300 kg/m3 of concrete. The upstream and downstream faces were covered with a
reinforced gunite layer.
In the years in which the dam was built, the owner was a small private
hydropower producer. Preliminary testing of the concrete mix was not carried out,
and construction procedures were poor. These, together with environmental aggres-
sion, were contributing factors to leakage that started occurring, coming from (or
through) foundations and from (or through) the poor imperviousness of the upstream
facing.
2.2. ENVIRONMENTAL AGGRESSION AND FIRST REMEDIAL MEASURES
High altitude in the Alps implies high temperature excursions, and for several
months frequent and heavy snow falls, freeze/thaw cycles, and thick ice formation.
At Lago Nero the disintegrating action of freeze/thaw cycles on the concrete was
aggravated by the soft alpine waters, which with their low dissolved ion content
are aggressive on Portland cement-based concrete: the calcium hydroxide in the
hardened cement matrix leaches out as the water tries to establish an ion balance,
causing a breakdown and softening of the concrete matrix, which in turn normally
leads to a breakdown of the hydrated cement matrix and of the concrete. Coatings
with epoxy resin/Silane/Siloxane sealers were not used at that time. Despite many
Italian concrete dams at high elevation are equipped with an air bubble system to
keep the ice and snow away from the upstream facing and avoid an increase of thrust
in the top part of the dam, concrete deterioration cannot be avoided, and refacing
is often required. The degradation already occurring at the dam due to defects at
construction was accelerated by climate and soft water aggression.
Various repairs were undertaken to reduce water loss, including concrete

grouting in the bedrock, and restoration works of the reinforced gunite layer. Such
repair works were repeatedly carried out in the years from 1953 to 1962 in the
attempt to restore watertightness. Gunite facings, however, have poor resistance to
freeze/thaw cycles, and after a few years the reinforced gunite deteriorated again
and started detaching in the lower part of the dam. The remedial measures adopted
over the years were either insufficient, or ineffective in the long term, and water
seepage continued.
2.3. CRITICALITIES AND LONG-TERM REMEDIAL MEASURES SELECTED
The injections carried out until 1962 were intended to restore the mechanical
and physical characteristics of the concrete of the dam body. In 1967, low specific
24
Q.104 - R.3
weight and persistence of poor mechanical characteristics in the concrete were

ascertained, which resulted in the non-fulfilment of the stability checks for the dam,
and in water losses from the downstream face. The national civil works Authorities
imposed a limitation of the water level to 2020.00 (almost 4 meters below the normal
summer water level).
To remedy the situation and bring back the reservoir to its full capacity, in 1968
Enel, after carrying out coring and laboratory testing on the concrete, presented an
overall restoration project including: improvement of the stability of the dam by adding
weight at crest, restoration of the watertightness of the foundations by resuming the
execution of injections, and restoration of the watertightness of the upstream face by
a new facing that would stop water infiltration and protect the concrete from further
deterioration. The engineering of the project was carried out entirely in-house and
the final design was presented in 1975.
After approval of the designed measures by the competent authorities, in 1979

the dam was put out of operation to allow working in dry conditions. The alpine
climate at high altitudes leaves a short summer window for carrying out the works,
so rehabilitation was split in 3 separate campaigns of about 5 to 6 months each.
3. THE REHABILITATION WORKS
3.1. IMPROVING STABILITY, AND RESTORING WATERTIGHTNESS OF FOUNDATIONS
Additional weight was placed at the dam, in form of a new reinforced concrete
block raising the dam from elevation 2025.15/2027.25 to elevation 2031.50 in the
section between joints G2 and G4, for a total length of 56 m (Fig. 2).
Fig. 2
Upstream view of Lago Nero dam with the new reinforced concrete block at crest.
Vue amont du barrage de Lago Nero avec le nouveau bloc de béton armé au
couronnement.
25
Q.104 - R.3
The concrete raise, 4 m high at the upstream side and 6 m high at the down-
stream side, 5 m wide at the top and 3 m wide at the existing crest, guaranteed
compliance to the stability checks (in terms of concrete strength and of sliding resis-
tance) foreseen by the regulations in force at that time. In addition, a stability analysis
was carried out taking into account the maximum reservoir level under the thousand-
year flood, and excluding the lowering of the surface discharge sluice gates. The
results of this assessment were deemed acceptable by the National Dams Service
and by the Ministry of Public Works.
Inside the concrete raise a longitudinal gallery has been created at elevation
2025.00 (Fig. 3), along which two windows have been arranged that allow access
to the collimation instruments.
Fig. 3
Cross section of Lago Nero dam in correspondence of the new reinforced concrete
block at crest.
Coupe transversale du barrage de Lago Nero en correspondance du nouveau bloc
de béton armé au couronnement.
26
Q.104 - R.3
The downstream face was restored by demolishing the most degraded parts of
the existing reinforced gunite layer and by constructing a new gunite facing reinforced
with a metal mesh.
The spillway capacity was increased, to allow coping with the thousand-year
reference flood, reaching a reservoir level such as to guarantee the satisfaction of
the stability requirements even with the sluice gates closed.
The new reinforced concrete plinth embeds a steel plate with pre-welded studs
that can be seen in Fig. 4 at right and were used for the bottom anchorage of the
geomembrane liner, as detailed in the section that follows. Two series of injections
were performed: a first one, for a total of 40 holes between 10 and 30 m deep, and
a second one for a total of 40 holes between 12 and 40 m deep. The holes have
50-mm and 45-mm diameter and were injected with R325 pozzolanic cement.
Fig. 4
Where gunite was detaching local plastering was made. The bottom boundary of
the dam was cleaned down to the rock to prepare construction of the new
reinforced concrete plinth.
Là où le béton projeté se détachait, un plâtrage local a été fait. La limite inférieure
du barrage a été nettoyée jusqu’au rocher pour préparer la construction de la
nouvelle plinthe en béton armé.
3.2. THE UPSTREAM GEOMEMBRANE SYSTEM
3.2.1. Options considered for the upstream water barrier
The first option considered was a new layer of gunite reinforced with a metal
mesh, to be sprayed over the upstream face of the dam. This solution was quite
traditional, did not pose concerns in terms of execution, but entailed the complete
demolition and removal of the existing gunite facing, with resulting long time for
execution of works. From a technical standpoint, in the specific circumstances of
harsh climate and soft water aggression, as experience had shown, a new gunite
layer would not guarantee a long-term reliable performance.
27
Q.104 - R.3
The second option was a new facing with steel plates. This solution already
had some successful precedents, but entailed high costs and would be difficult to
execute in short times in such a climate. Additionally, steel facings needed frequent
maintenance, carried out by painting with special products requiring outage of the
dam. As a matter of fact, ENEL itself experienced expensive maintenance at its
174 m high Alpe Gera dam, at 2128 m altitude. Eventually, after the successful
experience at Lago Nero, in 1993 the Carpi geomembrane system was adopted to
waterproof the corroded painted steel plates in the bottom 90 m of Alpe Gera.
The third option was an exposed geomembrane system. Geomembranes had

first been used in new embankment dams in 1959, in covered position. Their use
as rehabilitation measure in exposed position was relatively new and quite limited:
in 1970 a polyisobutylene geomembrane had been used by ENEL in Italy at the
peripheral joint of Lago Verde 2 dam and at the central section of the upstream face
of Baitone dam, while the first reported use of polymeric geomembranes was at
Heimbach dam in Germany in 1974, and at Lago Miller dam in Italy in 1976, gravity
dams both lined with a polyvinylchloride (PVC) geomembrane. Lago Miller, at eleva-
tion 2170 m, was also property of ENEL, and the experience acquired there in terms
of installation, costs, efficiency, was useful for the selection process. Installation of
the geomembrane system had proven to be simple and not requiring long times,
demolishing the existing gunite would not be necessary since some local plastering
would be sufficient (Fig. 4 at middle), the cost of the system was lower as compared
to the other options; in the few years that had elapsed the geomembrane system
at Lago Miller had behaved well, testifying also great efficiency in waterproofing the
structural joints that is always a critical point of other repair systems. Lago Nero was
a relatively small project that however presented all the complexities of a large one,
so an ideal site where to validate a promising and fairly new technology in view of
possible future applications.
Based on the factors steering the selection, i.e. the simplicity and speed of
installation, the watertightness at the joints, the cost, the lack of routine maintenance,
and the possibility of easy and quick repair if needed (the underwater repair method
became available a little more than a decade later), the exposed geomembrane
system was the preferred option.
Restoration of watertightness was based on the concept of having a complete

water barrier from crest to foundations: a geomembrane barrier was constructed
from crest down to the heel of the dam, where it was watertight connected to a
new plinth from where reinforcement of the grout curtain was performed. The grout
curtain was reinforced by injecting cement mortar in the rocks after the excavation
works carried out along the foundation of the dam, before construction of the new
plinth.
28
Q.104 - R.3
3.2.2. Selection of the waterproofing liner
Several types of polymeric and elastomeric geomembranes were available on

the market, but since at the time the wide present literature on geomembranes’
properties, on real scale tri-dimensional testing, on precedents of geomembrane
systems in large dams, was still to come, the selection of the liner was based on the
descriptions of the manufacturers, which provided a first screening, on index test
performed in the ENEL-CRIS Geosynthetics Laboratory to verify the properties of
the candidate liners and assess their suitability for the project, and on a few small
projects where a geomembrane system had been installed.
The selected material was SIBELON® CNT 2800, a PVC geomembrane heat-
bonded during fabrication to a 200 g/m2 anti-puncture polyester geotextile, to form
what is called a geocomposite. Here below the characteristics of the geomembrane
material.
A second set of tests was performed by ENEL-CRIS to assess the change in

tensile properties due to high UV exposure, and possible chemical alterations due
to service, which were measured with an absolutely original analysis by infrared
spectrophotometer. The results of these tests as reported in literature [1], and the
feedback from the field in respect to flexibility, weldability and overall constructability,
validated the selection of the geomembrane liner.
The liner was supplied in rolled sheets 2.05 m wide, each sheet having a
length matching the height of the section where it would be installed and a selvedge
deprived of the geotextile, to allow easy reliable heat-welding at the overlapping of
adjacent sheets.
3.2.3. Fastening system
Exposed geomembrane liners must be fastened to the face of the dam to keep
them stable and avoid them being uplifted by wind and waves, and must be sealed
at peripheries to avoid water infiltration under the liner. The fastening and sealing
Table 1
Geomembrane properties
GENERAL PROPERTIES TEST METHOD SIBELON®
Thickness [mm] ASTM D 374-379 1.93
Tensile test D638M-81
Strength at break [MPa] 18.0
Elongation at break [%] 285
Moisture vapor permeability [g/m2 *24 h] ASTM D1653-72 3.2
Abrasion test [mm3 /1000 turns] ASTM D1944-78 60
29
Q.104 - R.3
systems of Lago Nero were especially conceived for the project, and tested on a
mock-up under a 2 MPa pressure.
The face fastening system developed, patented and used for the first time in
the world at Lago Nero was an innovative solution aiming not only to fasten the liner,
but also to avoid it sagging on the vertical face due to its own weight. This was
achieved with a system that since 1980 was then widely used all over the world,
in rehabilitation of all types of dams and in new construction of RCC dams, and
is discussed in ICOLD Bulletin 135 [2]. The system consists of two stainless-steel
shaped profiles connected by a three-components connection device, as shown in
Fig. 5. The profiles are placed at the vertical junctions of overlapping geocompos-
ite sheets, and by their geometry exert a slight horizontal tension on the sheets,
thus impeding the sagging that is detrimental for the longevity of the liner as it cre-
ates zones of localised higher stresses, especially detrimental in environments with
frequent accumulation of large snow packs. The geometry of the two profiles also
allows creating a rectangular hollow cavity that acts as a vertical drain placed at the
upstream face of the dam, as further described in the section on drainage.
Fig. 5
The tensioning system designed for Lago Nero dam (detail A in Fig. 8).
Le système de tensionnement conçu pour le barrage de Lago Nero (détail A en
Fig. 8).
Installation was carried out from small self-hoisting platforms secured at crest.
The sequence of installation of the various components is shown in Fig. 6: from
right to left, placement and fastening of the lower profiles with threaded anchor rods
embedded in resin hardening upon contact, placement of geocomposite sheets
overlapping upon the profiles and joined by heat-welding, placement of the upper
profiles, clamping of the two profiles by the connection device. The profiles are then
30
Q.104 - R.3
waterproofed by a SIBELON® geomembrane strip watertight welded at the perimeter

on the underlying geocomposite. At Lago Nero the spacing between profiles was
approximately 1.8 m, i.e. one profile for each sheet; in subsequent projects, slightly
modifying the configuration and dimensions of the profiles allowed doubling or tripling
the spacing between profiles, thus reducing times and costs.
Fig. 6
From right to left, the sequence of placement of the components of the system.
De droite à gauche, la séquence de mise en œuvre des composants du système.
The bottom seal, which must resist water in pressure and must be watertight
connected to the waterproofing of foundation, has a particular configuration that is
worth mentioning: a slot was excavated at the heel of the dam, to embed the top part
of the steel plate at plinth, and a U-shaped galvanised steel profile with pre-welded
studs was pre-welded to the steel plate. The geocomposite liner is punched on the
studs, and is compressed on the embedded steel plate and profile by a U-shaped
galvanised steel counter-profile. Anchor bolts tightened to a specific torque and a
semi-soft synthetic rubber gasket, as shown in Fig. 7, achieve the seal.
Fig. 7
Detail of the bottom seal at the heel of the dam. The tapered bottom of the vertical
tensioning profiles can also be seen (detail B in Fig. 8).
Détail de la fixation étanche en pied du barrage. L’extrémité effilée des profilés de
tensionnement verticaux est également visible (détail A en Fig. 8).
31
Q.104 - R.3
The compression exerted by the anchor bolts ensures that no voids are
present in between the various layers of the seal. The continuity of the barrier
geocomposite/bottom seal/steel plate/plinth/foundations is thus granted.
3.2.4. Drainage system
The drainage system has the purpose of avoiding uncontrolled back pressure
on the liner due to condensation or infiltration water, while at the same time providing
control of the performance of the liner by measurement of drained water. It consists
of the gap between the dam face and the liner allowed by the fastening system, of the
needle-punched nonwoven geotextile composing the liner, whose drainage capacity
was tested always at the ENEL-CRIS Geosynthetics Laboratory and discussed by
Cazzuffi and Puccio [3], by the vertical drains formed by the tensioning profiles, which
have lateral holes to let water flow by gravity to the bottom, of a porous drainage block
acting as bottom collector, and of 13 transverse pipes discharging into the gallery.
Fig. 8
Detail of the drainage collection and discharge system.
Détail du système de collecte et d’évacuation des eaux de drainage.
3.2.5. De-icing system
Laboratory tests had excluded the possibility of ice adhering to the smooth
PVC geomembrane. However, to avoid an increase of thrust in the top part of the
32
Q.104 - R.3
dam, and to prevent hat big ice blocks could form close to the dam and driven by
wind possibly cause damage to the geomembrane, a compressed air bubble sys-
tem, similar to those already mentioned for concrete dams at high altitudes, was
installed to create a water movement impeding the formation of ice in the proximity
of the upstream face. The system is composed by 13 pipes equipped with valves
manoeuvred from the peripheral gallery to regulate the air flow.
3.3. CONTROL OF PERFORMANCE
Overall restoration works ended in November 1981. When all works were com-
pleted, between 1981 and 1982 experimental fillings were carried out, and all data
related to the static behaviour of the dam and to the performance of the waterproof-
ing system were controlled. With reservoir level at 2012 m a.s.l., leakage had fallen
from 9 l/s (ante works) to 1.4 l/s (post works), including seepage from foundations
and (estimated to a lesser extent, as confirmed in year 1983) seepage from the new
upstream facing. Each time the reservoir was drained systematic inspections were
carried out on the geomembrane and on the fastening components. No damages
nor flaws were detected, all welds were in perfect conditions. The de-icing system
was successful in keeping the ice surface at a minimum distance of 15–20 m from
the face of the dam.
Fig. 9
At left, restoration works completed in 1981. At right, a winter inspection at empty
reservoir.
A gauche, travaux de réhabilitation terminés en 1981. A droite, une inspection
hivernale avec réservoir vide.
In June 1983 the reservoir was filled to full capacity. Daily readings of the
drained water were made. The maximum measured flow was 2.7 l/s, of which 0.39 l/s
(about 14%) attributable to the geomembrane system, possibly due to some residual
infiltration at the base of the dam. These results confirmed that the rehabilitation
works had been successful, and that a new reliable technology for rehabilitation of
the upstream watertightness of concrete gravity dams had been created.
33
Q.104 - R.3
4. MONITORING AND PERFORMANCE
In the 40 years that have elapsed since the completion of the waterproofing
system, Enel has been monitoring the performance of the dam and of the water-
proofing system, both as regards the aging of the geomembrane, and as regards
seepage.
Concerning the geomembrane aging, regular inspections were performed,

together with the sampling of the exposed liner and the testing of their properties, to
assess the aging process of the material. Sampling was executed in the years 1995,
1997 and 2010. Samples were taken both above and under the water level and in
different parts of the upstream face, with the aim of studying the different behaviour of
the same geomembrane in different conditions of exposure. In the determination of
the life expectancy of a geomembrane, it is important to identify the more critical
portion of the upstream face, as the first failure will affect negatively the whole
waterproofing system. Therefore, what presented here are the results of the tests
made on samples taken above the water level, as this is the area which suffers
more the exposure to atmospheric agents. The results obtained are thus referred to
the worst conditions for each geomembrane, which helped the owner to conduct a
precautionary analysis of the geomembrane durability.
All the samples taken from the dams’ upstream faces were tested at the
Geosynthetics Laboratory of CESI in Milano, Italy. Before the laboratory tests, the
samples were prepared by separating the geotextile layer from the geomembrane.
The results obtained on the sampled geomembrane show a constant small

decrease of the plasticisers content, while dimensional stability grows longitudi-
nally and declines transversally in the years. Mechanical parameters show that the
geomembrane gets a little bit stiffer over time, with a small growth of tensile strength
and a reduction of the correspondent strain.
With regard to the waterproofing properties, the results of the vapour trans-
mission test in general demonstrate the decrease of the permeability coefficient,
thus an improvement of the water-tightness of the geomembranes, as illustrated by
Cazzuffi & Gioffré [4].
In order to evaluate the residual life of the exposed geomembrane in Lago

Nero dam, the plasticisers content versus time has been analysed in detail (Fig.
10): according to the existing literature, this parameter is in fact considered the most
reliable for a proper evaluation of the aging of PVC geomembranes, see for example
Hsuan et al. [5]. In the case of the considered geomembrane, it was concluded that
a residual value of plasticizer of 70% on average of the sample virgin value allows
the geomembrane to guarantee a good performance on site.
34
Q.104 - R.3
In particular, for the case of the Lago Nero dam, this corresponds to a residual
life of 68 years, i.e. still 28 years from now (see Fig. 10).
Fig. 10
Evaluation of residual life of the exposed geomembrane in the Lago Nero dam,
based on the evolution of the plasticisers content vs. time.
Évaluation de la durée de vie résiduelle de la géomembrane exposée au barrage
du Lago Nero, basée sur l’évolution de la teneur en plastifiants en fonction du
temps.
Fig. 11
Lago Nero pictured 33 years after geomembrane installation. The de-icing system
keeping the ice away from the face of the dam can be seen at the bottom.
Lago Nero photographié 33 ans après l’installation de la géomembrane. Le
système de dégivrage éloignant la glace de la face du barrage est visible en bas.
35
Q.104 - R.3
Concerning seepage from the geomembrane system, the data reported in the
previous section have been substantially steady since 1983, confirming the excellent
performance of the system.
5. CONCLUSIONS
Lago Nero was planned as an innovative project to possibly validate a

SIBELON® geomembrane technology that could be applied to rehabilitation of con-
crete gravity dams. The results confirmed the expectations, and based on this
precedent ENEL waterproofed with Carpi geomembrane systems 5 more large grav-
ity, multiple arch and masonry dams, 7 reservoirs, 11 canals and hydraulic tunnels.
In 2020, 40 years after its installation, the exposed geomembrane system is perform-
ing successfully, without any routine maintenance, and together with the structural
works is ensuring the safe operation of an almost 100 years old dam at high altitude.
REFERENCES
[1] MONARI F. Waterproof Covering for the Upstream of the “Lago Nero” Dam.
Proceedings, International Conference on Geomembranes, 1984.
[2] ICOLD. Bulletin 135, Geomembrane Sealing Systems for Dams – Design
principles and review of experience, 2010.
[3] CAZZUFFI, D. and PUCCIO, M. The use of as drains. Laboratory tests and
design criteria. Proceedings, International Symposium on Soil and Rock
Improvement, 1982.
[4] CAZZUFFI, D., GIOFFRE’, D. Long-time behaviour of exposed geomembranes

used for the upstream face rehabilitation of concrete and masonry dams,
Proceedings, 11th International Conference on Geosynthetics, 2018.
[5] HSUAN, Y.G., SCHROEDER, H.F., ROWE, K., MUELLER, W., GREEN-
WOOD, J., CAZZUFFI, D., KOERNER, R.M. Long-term performance and
lifetime prediction of geosynthetics, Keynote Lecture, Proceedings, EuroGeo
4 – Fourth European Geosynthetics Conference, 2008.
36
Q.104 - R.4
-------
-------
GEOMEMBRANES IN VERY HIGH CONCRETE DAMS: CHALLENGES AND

SOLUTIONS (∗)
Miguel MIRANDA
Isagen
Gabriella VASCHETTI, Alberto SCUERO

Carpi Tech
ITALY
SUMMARY
Geomembrane systems have been used in dams since more than 50 years.
When very high dams are at stake, particular challenges are evidently faced. The
design of the geomembrane system must be conceived to grant full watertightness
in presence of high water pressure, to allow monitoring with adequate accuracy
separate sectors of the dam, which must be as numerous as possible, and to foresee
measures to remediate accidental damages even without dewatering the reservoir.
Their installation needs to be carried out in separate horizontal sections, with custom
designed equipment minimising installation time while granting protection against
possible damage by civil works being carried out concurrently, and under stringent
Construction Quality Assurance procedures implemented by a specialty contractor
having experience in projects of similar height and magnitude.
This paper addresses challenges and solutions by the case history of Miel I
188 m high RCC dam in Colombia, where an exposed geomembrane system was
installed as additional safety barrier on an RCC facing enriched with vibrated cement
grout. This project set a milestone in waterproofing of new RCC dams with exposed
geomembranes.
∗ Géomembranes sur des barrages en béton de très grande hauteur : défis et solutions
DOI: 10.1201/9781003211808-4 37
Q.104 - R.4
The paper illustrates the waterproofing and monitoring system, the installation
equipment and phases, gives data on performance, and presents some innovations
developed and adopted on several dams, following the experience gained in the Miel
I project.
RÉSUMÉ
Les systèmes d’étanchéité par géomembranes sont utilisés sur les barrages
depuis plus de 50 ans. Lorsqu’il s’agit de barrages de très grande hauteur, des défis
particuliers doivent évidemment être affrontés. Le système de géomembrane doit
être conçu pour garantir une étanchéité totale en présence de haute pression d’eau,
pour permettre la surveillance avec une précision adéquate de différents secteurs du
barrage, qui doivent être aussi nombreux que possible, et pour prévoir des mesures
de réparation des dommages accidentels sans même devoir vider le réservoir. Leur
installation doit être effectuée dans des sections horizontales séparées, avec des
équipements conçus sur mesure pour minimiser le temps d’installation tout en garan-
tissant une protection contre les dommages éventuels dus aux travaux de génie civil
exécutés simultanément, et selon des strictes procédures d’assurance qualité mises
en œuvre par un entrepreneur spécialisé ayant de l’expérience dans des projets de
hauteur et amplitude similaires.
Ce rapport aborde ces défis et solutions en se basant sur le cas du bar-

rage BCR de Miel I en Colombie, de 188 m de haut, sur lequel un système avec
géomembrane exposé a été installé comme barrière de sécurité supplémentaire sur
un parement amont en BCR enrichi de coulis de ciment vibré. Ce projet a posé un
jalon dans l’imperméabilisation des nouveaux barrages BCR avec géomembranes
exposées.
Le rapport illustre le système d’étanchéité et de surveillance, l’équipement et

les phases d’installation, donne des données sur les performances, et présente
quelques innovations développées et adoptées sur plusieurs barrages, suite à
l’expérience acquise sur le projet Miel I.
1. INTRODUCTION
Geomembranes have been used in the last 60 years to provide or restore

watertightness in all type of dams. ICOLD has dedicated three Bulletins to the use of
geomembranes in dams. Bulletin 38 (1981) considered geomembranes applicable
only to fill dams, with a mandatory cover layer, and recommended a maximum
height of 30 m and a surface of modest dimensions. Bulletin 78 (1991) extended the
application of geomembranes to rehabilitation of all types of dams and to new RCC
dams, considered a cover layer no more necessary, and did not pose any limitation
38
Q.104 - R.4
to the height of the dam. Bulletin 135 (2010) reports the use of geomembranes on
very high dams of all types (concrete, RCC, embankment), be it for new construction
of for rehabilitation, in the dry and underwater.
When very high dams are at stake, geomembrane systems evidently face
particular challenges. Their design must be conceived to grant full watertightness
in presence of high water pressure, to allow monitoring with adequate accuracy
separate sectors of the dam, which must be as numerous as possible, and to foresee
measures to remediate accidental damages even without dewatering the reservoir.
Their installation needs to be carried out in separate horizontal sections, with custom
designed equipment minimising installation time while granting protection against
possible damage by civil works being carried out concurrently, and under stringent
Construction Quality Assurance procedures implemented by a specialty contractor
having experience in projects of similar height and magnitude.
This paper addresses challenges and solutions by the case history of Miel I
188 m high RCC dam in Colombia, where an exposed geomembrane system was
installed as additional safety barrier on an RCC facing enriched with vibrated cement
grout. The paper illustrates the waterproofing and monitoring system, the installation
equipment and phases, gives data on performance, and presents some innovations
developed and adopted on several dams, following the experience gained in the Miel
I project.
2. GEOMEMBRANES IN DAMS
2.1. SOME HISTORY
A large amount of information is available, also through the three ICOLD bul-
letins, on the use of geomembranes in dams; the most recent bulletin [1] and its
chapters benefit of an extensive database formed with the contribution of owners,
designers, geomembrane scientists, and contractors.
The use of geomembranes in dams started in 1959, and the first applications
were in new construction of rockfill dams, where the geomembrane was used either
to waterproof a sloping core, or as upstream water barrier covered by concrete
slabs or interlocking concrete elements or granular material. The 1970s saw the
start of applications of geomembranes as repair measure to restore watertightness
of the upstream face of embankment dams and concrete dams. In some embank-
ment dams the geomembrane was covered, but as over the years the quality of
geomembranes improved and different anchorage systems were developed, the
use of exposed geomembranes mechanically anchored became more frequent. In
concrete dams, since the beginning geomembranes were always left exposed.
In the1980s, shortly after the advent of RCC dams, geomembranes were

adopted as upstream water barrier in new construction of such dams, in exposed
39
Q.104 - R.4
position or as watertight element of the prefabricated concrete panels used as per-

manent formworks for placement of the RCC. The use of geomembranes as repair
measure in RCC dams started in the year 2000 with an underwater installation to
waterproof a few cracks, and was later followed by projects where the geomembrane
was used to restore watertightness of the upstream face, or of leaking areas.
2.2. APPLICATION IN CONVENTIONAL CONCRETE DAMS AND IN RCC DAMS
In conventional concrete and RCC dams, the geomembrane is installed at the

upstream face. The exposed position, always adopted in rehabilitation projects, is
favoured also for new RCC dams, because it provides some technical advantages
as compared to the covered position. The main technical advantage is the possibility
of having a full face efficient drainage system underneath the waterproofing liner,
which at design stage allows reducing design uplift, and during service efficiently
avoids accumulation of water behind the waterproofing liner, collecting and removing
it continuously at the dam face, thus lowering saturation levels and pore pressures in
the dam, with beneficial effects on the stability safety factors, on AAR phenomena,
on appearance at the downstream face. Through the measurement and control
of the quantity of drained water, the face drainage system allows monitoring the
performance of the waterproofing system on a continuous basis. Additionally, the
exposed position allows controlling the integrity of the geomembrane at any time,
and repairing it in case of accidental damage.
In almost all projects on concrete and RCC dams, the waterproofing liner is
anchored to the upstream face of the dam with a well-known tensioning system
installed along vertical lines at regular spacing, and sealed at all peripheries by a
mechanical seal watertight against water in pressure at all submersible peripheries,
and watertight against water in feeble pressure (rain, snowmelt, waves) at crest.
2.3. HIGH DAMS
In rehabilitation as well as in new construction, geomembranes have been

used in several dams largely exceeding 100 meters of height, which although inferior
to the height records reached in the past and even more so in modern dams, are still
a considerable challenge for geomembrane barriers that are just a few millimetres
thick.
Some applications, e.g. at the 200 m high Kölnbreinsperre arch dam, were
addressing specific situations and are not within the scope of this paper, which
focuses on geomembranes as full-face water barriers. Chapter 1 has summarised
the challenges posed by high water heads. The chapter that follows discusses the
solutions conceived to overcome such challenges at Miel I RCC dam.
40
Q.104 - R.4
Fig. 1
RCC mixes.
Mélanges BCR.
3. MIEL I RCC DAM
3.1. DAM DATA
Miel I is a straight gravity dam constructed in a narrow gorge in Colombia, in a

medium-intensity seismic zone. With 188 metres of height, when it was completed
in 2002 it was the world’s highest RCC dam, surpassed only in 2009 by the Longtan
dam. The dam crest, at elevation 454 m, is 354 m long and the upstream face has
a total surface of 31,500 square meters. The RCC mix is of the low cementitious
content type, without pozzolan and without cooling of concrete. Four mixes were
designed, with a cement content of 85, 100, 125 and 150 kg/m3 .
41
Q.104 - R.4
Parent RCC is composed of mixes of 125 and 150 cement per cubic meter of
concrete, total RCC volume is 1,745,000 m3 . The RCC delivery system consisted of
eight conveyor segments, a tower crane and a crawler placer. Temperatures at site
ranged from 18 to 38◦ C, concrete placement was made at 26 to 32◦ C. Transverse
contraction joints are placed every 18.5 m and treatment against seepage consists
of a PVC waterstop embedded in the dam 0.2 m downstream of the upstream face.
The treatment of the foundation rocks consists in a consolidation blanket between
the drainage galleries and the plinth and in the construction of a deep grout curtain.
Grouting was carried out from galleries excavated at several elevations within the
abutments and also from galleries within the RCC body and from the plinth. The
design of the waterproofing system for the dam called for 20 m deep grout curtains,
complemented with a series of drainage holes both along the dam’s foundation and
its abutments, and within the dam’s body. The target was to reduce permeability
beneath the dam to values of 10−7 m/sec and water seepage pressures to less
than 1/3 of its original value in the vicinity of the drainage galleries and zero at the
dam toe. Water pressures will be monitored using piezometers installed along the
foundation and within the dam body.
In RCC dams, and even more so in those with a low cementitious content mix,
it is common to have an upstream barrier to prevent infiltration through lift joints or
zones of higher permeability of the RCC mass that may occur in the construction
process. The original design of Miel I foresaw an upstream face made of slip formed
reinforced concrete, to be built after completion of the RCC body. Construction of this
water barrier would have delayed of six months the impoundment, so the design was
changed to a configuration that would allow meeting the schedule and avoiding the
costs related to delayed hydropower production: instead of the reinforced concrete
wall a drained exposed polyvinylchloride (PVC) geomembrane system was adopted,
placed on a 0.4 m wide zone of grout enriched vibrated RCC. This double water
barrier was considered necessary in consideration of the height of the dam [2].
3.2. DESIGNING THE GEOMEMBRANE SYSTEM FOR HIGH WATER HEAD
Miel I was an innovative project in those times, and as a matter of fact the
concept and many components of its design still represent the state-of-the-art for
this type of dam, although, as it will be discussed, some developments have been
made in more recent projects. At the time, the precedents in RCC dams waterproofed
with a Carpi exposed geomembrane system accounted for Riou in France (1990,
20 m high), Conception and Nacaome in Honduras (respectively in 1991 and 1994,
and 70 and 55 m high), and Balambano in Indonesia (1999, 99.5 m high).
To grant full watertightness in presence of very high water heads, essential

to the design are the selection of an appropriate geomembrane liner, a fastening
system capable to resist the applied loads at full and empty reservoir, a peripheral
sealing system watertight connected to the waterproofing of foundations, and an
efficient drainage system to avoid water infiltration in the dam and to allow monitoring
42
Q.104 - R.4
with adequate accuracy separate sectors of the dam. Repair methods to remediate
accidental damages even without dewatering the reservoir are additional assets.
3.2.1. Waterproofing liner selection
When the waterproofing system of Miel I was designed, extensive scientific and
technical information and feedback from the field were available since many years.
All data, statistics, precedents and full-scale performance testing, among which
an extensive research program carried out by the US Army Corps of Engineers
on various types of geomembranes available in the market [3], indicated as most
suitable material a composite membrane (geocomposite) formed by a plasticised
SIBELON® PVC geomembrane heat-bonded during extrusion to an anti-puncture
geotextile. This material was selected for the project. The thickness of the geomem-
brane and of the bonded geotextile was chosen based on the characteristics of the
project. Theoretically, a 1 mm thick geomembrane is sufficient to provide watertight-
ness, but higher thickness is preferable when the surface on which the material will
be installed is irregular, or to grant longer durability under heavy UV exposure. At
Miel I, the steering factor was the pressure that would be exerted by the water, on a
surface that could present some irregularities. A thicker geomembrane was chosen
for the bottom 62 meters of the dam: from elevation 268 m to elevation 330 m the
waterproofing liner is SIBELON® CNT 4400, a 3.0 mm thick geomembrane lami-
nated to a 500 g/m2 nonwoven polypropylene geotextile, while from elevation 330 m
to elevation 450 m the waterproofing liner is SIBELON® CNT 3750, a 2.5 mm thick
geomembrane laminated to a 500 g/m2 nonwoven polypropylene geotextile.
Modern manufacturing technologies now allow producing SIBELON® geocom-

posites with geomembrane thickness of 4 mm and more, when exceptional durability
in presence of high UV radiation is required, bonded to 700 g/m2 geotextiles, to
enhance anti-puncture resistance.
3.2.2. Fastening system
An exposed geomembrane liner must be kept stable on the face of the dam
by a fastening system capable to sustain the compression exerted by the water of
the reservoir, and the uplift exerted by wind, waves and unbalanced back pressure
if the water level lowers and condensation or filtration water is present behind the
waterproofing liner.
The face fastening system selected for the project was the same used in the
RCC dams already mentioned, which had proven to be effective to secure the liner,
pre-tension it, and enhance drainage capability thanks to the vertical face drains
formed by the geometry of the fastening system itself. Fig. 2 shows the typical con-
figuration of the system: profile 3 and its attachment system 1 and 2 are embedded in
the face of the dam as it is being built, and connectors 4 attached to them. Adjacent
43
Q.104 - R.4
Fig. 2
Scheme of the tensioning system for face anchorage of the geocomposite (Carpi
patent).
Schéma du système de tensionnement pour l’ancrage du géocomposite (brevet
Carpi).
geocomposite sheets 5 are deployed and overlapped on it, heat joined along their
edges, and punched over connectors 4. Profile 6 is placed over the liner and tightly
clamped to profile 3 via connector 4, achieving a horizontal tension on the adjacent
liner sheets. Profile 6 is then waterproofed with a PVC geomembrane cover strip 7,
watertight heat-seamed to the waterproofing liner.
The highest water head experienced by this system was at Balambano 99.5 m
high RCC dam. The height of Miel required designing a more robust profile for the
part of the dam where the water head would be higher: in the lower 90 meters of the
dam, from elevation 268 m to 358 m, the design of profile 6 was modified to have a
central reinforcement, as shown in Fig. 2 at right. The fastening lines were placed
at 3.70 m spacing.
Starting at the middle of the 2000s, new configurations of the tensioning

system have been adopted to make the installation and quality control of the water-
proofing system as much as possible independent from the construction of the dam:
both profiles are now installed external to the face of the dam, so that embedment
of profile 3, and quality control on the accuracy of embedment, are no more nec-
essary. At the middle of the 2010s, a configuration not requiring waterproofing the
profiles with a geomembrane cover strip has also been developed and patented;
such configuration, which can provide some time savings in installation times in dry
conditions, allowed installing a tensioning system in underwater conditions.
Crucial to the behaviour of the system is the watertightness of the bottom

peripheral seal subject to high water heads, and of its connection to the foundation
treatment. At Miel I, a mechanical seal of the tie-down type, watertight clamping
together the geomembrane liner waterproofing the plinth and the geomembrane liner
44
Q.104 - R.4
Fig. 3
Watertight perimeter seal at plinth.
Fixation périmétrale étanche à la plinthe.
waterproofing the upstream face, was placed at the plinth from where consolidation
grouting was executed (Fig. 3). The watertightness of this type of seal had been
tested at that time in real scale under a water pressure of 2.5 MPa. For a recent
project in a high-pressure tunnel, the same type of seal has been tested and proved
watertight under a water pressure of 8.5 MPa (850 m).
The watertight seal was placed also at crest, so that the waterproofing system
can resist water in pressure, hence overtopping.
The lowest section is the one having the highest number of compartments: in
the central part there is one compartment for each line of vertical tensioning profiles,
i.e. one compartment every 3.70 meters, and one compartment for each abutment.
In the upper three sections, the central part of the dam has one compartment for each
monolith, and one compartment for each abutment. The purpose of making such
unusually numerous separate compartments was to facilitate locating any potential
malfunctioning of the system by monitoring the discharge flow of each compartment.
Should a failure occur in the geocomposite, water can be managed, i.e. captured
45
Q.104 - R.4
Fig. 4
Scheme of drainage compartments and installation sections.
Schéma des compartiments de drainage et des sections d’installation.
at the upstream face and discharged to the nearest gallery, where the increase
in measured flow will signal the location of a possible damage. The monitoring
procedure established by ISAGEN was based on the deviations of the measured
flows after the establishment of a fairly constant “normal” flow for each discharge
pipe and is available in recent literature [4, 5].
It must be mentioned that repair of the system can be easily and quickly per-
formed, if above water level, by welding a geomembrane patch or installing a new
sheet depending on the extent of the defect; in areas below water level, repair
can be performed by divers with geocomposite panels watertight fastened with
stainless-steel frames upon the defect.
3.3. INSTALLATION
One challenge of installation in very high dams is the need to plan it in separate
horizontal sections, since handling geocomposite sheet 2.10 m wide and weighing
4.40 kg/m2 (below elevation 330 on the case of Miel I) and 3.75 kg/m2 (above
elevation 330) having considerable lengths would be impossible or very difficult
to manage and install adequately and safely. A second challenge is the access
to the installation sections, which requires a bespoke installation equipment that
can be moved at different elevations. At Miel I, to further expedite construction of
the water barrier it was decided to install the geomembrane system concurrent
with construction of the dam body, and this required integrating in the installation
equipment a protection against possible damage by concurring civil works being
carried out above the section where the geomembrane was being installed. A third
challenge is related to planning the installation sections in function of the horizontal
drainage compartmentation, and in some cases (as in this project) also in function
of a possible change in the thickness of the geomembrane.
46
Q.104 - R.4
Fig. 5
From left to right, the galvanised steel profiles attached to the formworks, each
connected to a transverse drainage discharge pipe to the lowest gallery, the
profiles attached to the formworks are embedded in the GE-RCC, a
profile + connectors after embedment.
De gauche à droite, les profilés en acier galvanisé attachés aux coffrages, reliés
chacun à un tuyau transversal d’évacuation de drainage à la galerie la plus basse,
les profilés attachés aux coffrages sont enrobés dans le GE-RCC, un
profilé + connecteurs après avoir été enrobés.
At Miel I, installation was planned in three horizontal sections: a bottom sec-

tion from approximately elevation 358 m to the heel of the dam, executed with
the installation equipment placed at elevation 360m, an intermediate section from
approximately elevation 405 m to elevation 358 m, executed with the installation
equipment placed at elevation 407 m, and a top section from crest to elevation
405 m. The bottom section was divided into three sub-sections, due to the presence
of two horizontal drainage compartmentations and of a change in thickness of the
waterproofing liner. The top section was divided into two sub-sections, due to the
presence of a horizontal drainage compartmentation. So, in total there are 6 hori-
zontal sections, as shown in Fig. 4. After placement of the RCC and of the GE-RCC,
embedding the galvanised steel profiles and the drainage collection and discharge
system (Fig. 5), was completed in the lower part of the dam, the geocomposite
waterproofing the grouting plinth was installed, and the grouting plinth constructed
(Fig. 6 at left).
Fig. 6
At left, preparation for casting of the plinth; the waterproofing geocomposite can be
seen installed on the dam face. At right, placement of the waterproofing
geocomposite in the lowest section of the dam.
A gauche, préparation pour la coulée de la plinthe; le géocomposite d’étanchéité
peut être vu installé sur le parement du barrage. A droite, mise en œuvre du
géocomposite d’étanchéité dans la section la plus basse du barrage.
47
Q.104 - R.4
Installation of the geocomposite in the lower completed section of the dam

was then carried out, while placement of RCC lifts and of GE-RCC was continuing
above (Fig. 6 at right) thus without impacting on the placement of the RCC.
To allow executing geomembrane works completely independently from civil

works, the waterproofing contractor designed a railing system that could support the
travelling platforms from which all activities were carried out, and could be assembled
and disassembled at different elevations. The railing system was first secured to the
dam at approximately elevation 360 m, some 90 m above foundation (Fig. 7 at left),
and with the system in this position installation in the three bottom horizontal sections
was performed. In correspondence of the contraction joints, two layers of sacrificial
geocomposite, of the same type used for the waterproofing liner, provide support
avoiding intrusion of the liner in the active joint, as can be seen in Fig. 7 at left.
Fig. 7
At left, geocomposite installation almost completed in the two lower sections and
starting in the section above, with the railing system placed at elevation 360 m. At
right, geocomposite installation with the railing system placed at elevation 407 m,
and RCC placement ongoing above.
A gauche, l’installation du géocomposite est presque terminée dans les deux
sections inférieures et commence dans la section au-dessus, avec le système de
rails placé à 360 m. A droite, installation du géocomposite avec le système de rails
placé à 407 m, alors la mise en œuvre d du BCR est en cours au-dessus.
As construction of the dam proceeded, the railing system was disassembled

from elevation 360 m and reassembled at elevation 407 m, to perform installation in
the fourth horizontal section (Fig. 7 at right).
The horizontal connections between superimposed installation sections were

made by anchoring the geocomposite of the lower section with a flat stainless-steel
profile, overlapping and welding unto it the geocomposite of the upper section, and
then waterproofing the junction with a geomembrane cover strip. At elevations 405 m
and 330 m a horizontal band of drainage geonet was placed under the profile, not to
impede the free flow of water towards the bottom of the related compartment. Fig.
8 at left shows a horizontal profile, under it the stainless-steel tensioning profiles
installed on the geocomposite sheets, over it the galvanized profiles embedded in the
48
Q.104 - R.4
GE-RCC. At right, the stainless- steel profiles are waterproofed with geomembrane
cover strips.
Fig. 8
At left details of the transition of two superimposed geocomposite sections, at right
the tensioning profiles are waterproofed with geomembrane cover strips.
A gauche, détails de la transition entre deux sections superposées du
géocomposite, à droite les profilés de tensionnement sont imperméabilisés avec
des bandes de couverture en géomembrane.
A further challenge in very high dams is quality: good design is useless if instal-
lation is poor. It is essential that installation is carried out under stringent Construction
Quality Assurance procedures, implemented by a specialty waterproofing contrac-
tor having experience in projects of similar height, as specified also by Bulletin 135
[1]. At Miel I, in addition to Carpi internal quality control procedures, compliant with
ISO 9001 certification, ISAGEN appointed an independent body for execution of
third-party quality control of all steps essential to the final good performance of the
system.
The use of GE-RCC allowed achieving good compaction of the mix at the
upstream face, assuring a good finishing of the upstream concrete surface. Con-
struction of the dam started in April 2000 and ended in June 2002, for a total of 26
months including the waterproofing system, and seven months ahead of time.
The change in design and the installation procedure concurrent with construc-
tion of the dam allowed not only meeting the schedule, but being able to test the
turbines earlier than expected and while the dam was not yet completed (Fig. 9).
Savings of several tenths million US $ were possible because of reduced cement
quantity, faster completion, earlier power generation.
As reported by the designer of the dam [6], the effectiveness of the water-
proofing system was confirmed after impoundment: after one year of operation of
the reservoir, total seepage through the dam body was less than 2.5 l/s.
49
Q.104 - R.4
Fig. 9
Installation of the geocomposite in the two top horizontal sections, carried out from
travelling platforms suspended at crest, while finishing concrete works are ongoing
above.
Installation du géocomposite dans les deux sections horizontales supérieures,
réalisée à partir de plates-formes mobiles suspendues au couronnement, tandis
que les travaux de finition du béton sont en cours au-dessus.
4. PERFORMANCE AND MONITORING
Since 2002 ISAGEN has constantly been monitoring the dam’s and geocom-
posite system’s performance. The maintenance group is constantly looking for any
deviation from the “normal” flow that is considered a signal of anomaly. Bimonthly
membrane inspections are performed
In 2003, after reservoir’s impoundment, the recorded rate of leakage from the
whole geomembrane system was 3.89 l/s, registered in 2013 with a reservoir level of
446.47 m a.s.l. The average leakage registered since then has been of 2 l/s for the
geocomposite’s drainage system, and of 25 l/s from the abutments. Such levels are
below the design parameters 9.7 l/s from the geomembrane system and 30 l/s from
the abutments. The measured leakage has never surpassed the historical values
(with stable lecture through 13 years), or overcome the maximum allowed drainage
design values defined during the design stage.
Since an abnormal drainage flow does not necessarily mean there is a dam-
age in the liner or in its fastening system (as stated by ICOLD Bulletin 135 [1],
50
Q.104 - R.4
“…an abnormal rate of leakage would suggest a defect in the geomembrane, while
in reality the drainage system is mainly collecting water which may come from dif-
ferent sources other than a defect through the geomembrane. For instance, water
infiltrating through fissures and bypassing the perimeter seal, or infiltrating from
foundations, or from crest. Thus, it is possible that the amount of water at the dis-
charge point is high, while the dam body is totally dry”), in addition to the inflow
rate monitoring, ISAGEN performs an external visual inspection of the geomem-
brane. The inspection is executed every time that the reservoir level changes of
3 m. This visual inspection allows the maintenance group to detect any problem in
the geomembrane. The few small cuts detected occurred mainly during the rainy
season, when debris arrive to the dam and bring wood or other materials that can
produce the cuts.
In summer 2013 ISAGEN decided to perform a more comprehensive monitor-

ing campaign, in addition to the routine one, to assess the general conditions and
the weathering behaviour of the waterproofing geocomposite at Miel I. The changes
in properties [4] most relevant to the ageing behaviour of the geomembrane were
totally in line with standard behaviour and with the values that are statistically exhib-
ited by exposed SIBELON® CNT 3750 geocomposites installed in exposed positions
on other dams.
In 2016, ISAGEN carried out an inspection of the entire submerged face of

the dam to analyse the current state of the pipes and submerged structures of the
generation circuit. In this inspection it was concluded that the condition of the Carpi
membrane is very good. Available studies indicated that PVC has preserved its
original mechanical parameters, and leaks are low. This inspection was carried out
using a remotely operated vehicle (ROV).
Fig. 10
Lowering the ROV by means of a basket on the dam face.
Le ROV est descendu dans le réservoir au moyen d’un panier sur le parement du
barrage.
51
Q.104 - R.4
The last repair that was carried out was in September 2016, when a fissure
caused by the plant material that reaches this area was detected. The crack was
found at elevation 418.10, between construction joints 11 and 12 approximately, and
was repaired by ISAGEN by simple patching.
Fig. 11
Carpi membrane repair by patching.
Réparation de la membrane Carpi par rustine.
In multiple inspections carried out, little damage to the membrane has been
evidenced during the plant’s 18 years of operation (15 cuts or perforations in 14
years, with low flows to the galleries). This fact, together with the perfect state of
conservation and preservation of the mechanical properties of PVC, leads to infer the
scarce risks of structural damage to the SIBELON® membrane. From this derives
the good control of the pressures in the body of the dam, of the sub-pressures in its
foundation, and the preservation of the stability conditions of the dam. The arrival
of trunks and branches is observed to occur in restricted quantities and at very low
speeds. The low speed allows that the impacts rarely generate the punching or
shearing of the membrane. The approximate average frequency of 1.0 annual cuts
is considered very low. None of the cuts have been reported to have generated high
flows in the drainage system to the galleries.
The “floating booms” type structures to contain bulky floats such as logs would
be very expensive to implement, given the high draft and the length of the crest of
350 m. As there are no intermediate anchor points, the structural fit of these elements
is complicated. They would be very difficult to anchor in the facing without puncturing
the membrane. In the submerged inspection carried out in 2016, it was observed
that a high percentage of the trunks and branches are semi-submerged, so it is
estimated that a surface floating containment barrier would not give any guarantee,
and would only entail a reduction of part of the impacts.
5. CONCLUSIONS
Geomembrane systems in RCC dams provide several technical and financial

advantages. They are extremely effective in seepage control, do not affect construc-
tion sequence, provide substantial savings as compared to traditional water barriers,
52
Q.104 - R.4
and allow earlier impounding. They can be applied also as a repair measure in the
dry and underwater. They provide long lasting excellent performance with practically
no repairs nor maintenance.
It is considered that the use of the CARPI membrane was a good decision,
given the conditions of this dam: large reservoir, scarce volume of logs in avenues,
and low impact speed of logs with the facing. It also made it possible to advance
both the construction and the filling of the reservoir and, most probably, reduces the
interstitial pressures compared to a solution with a facing without a membrane.
REFERENCES
[1] ICOLD Bulletin 135. Geomembrane Sealing Systems for Dams – Design
principles and review of experience, 2010.
[2] MARULANDA, A., CASTRO, A., RUBIANO, N. R. Miel I: a 188 m high RCC
dam in Colombia. The International Journal on Hydropower & Dams, 2002.
[3] CHRISTENSEN J. C., MARCY M. A., SCUERO A. M. and VASCHETTI G. L.

A Conceptual Design for Underwater Installation of Geomembrane Systems
on Concrete Hydraulic Structures, Technical Report REMR-CS-50, U. S. Army
Corps of Engineers, 1995.
[4] JIMÉNEZ GARCIA, M., SCUERO, A., VASCHETTI, G., and MACHADO DO
VALE, J. Miel I 188 m high RCC dam: monitoring an exposed geomembrane
system after 11 years of service. Proceedings, ICOLD 83rd Annual Meeting
and 25th Congress, 2015.
[5] SCUERO, A., VASCHETTI, G., JIMÉNEZ GARCIA, M., and COWLAND, J.
Geomembranes on RCC dams: a case history after 13 years of service. Pro-
ceedings, 7th International Symposium on Roller Compacted Concrete (RCC)
Dams, 2015.
[6] MARULANDA, A., CASTRO, A., SILVA, J. Miel I dam: seepage control and
behaviour during impoundment. Proceedings, Fourth International Sympo-
sium on RCC.
53
Q.104 - R.5
-------
-------
BARRAGE DE KRUTH-WILDENSTEIN REHABILITATION OF THE ASPHALT

LINING (∗)
Sophie CHARLOTTE KUHLMANN (MSc)

CMC Wasserbauprojekte GmbH, Germany
Enrico TITA & Maria VITTORIA VIGNOLI

COOPERATIVA EDILE APPENNINO, Italy
ITALY
SUMMARY
The dam of Kruth-Wildenstein is an earth-fill dam with an asphalt face as

lining system including a drainage layer of concrete and its upstream inclination is
1:1,5. In 2016, a large crack and other defaults let to an exhaustive structural survey
conducted by CMC Wasserbauprojekte GmbH. Through a visual assessment and
7 cores, the significant parameters of the asphalt lining were evaluated and CMC
evaluated together with the owner the major possible rehabilitation methods: asphalt
concrete vs geomembrane. The major reason for the owner to opt for the asphalt
lining was the technical feasibility regarding the constructive particularities of the
dam and the possibilities for the connections to the flanks and the blanket.
CEA (Cooperativa Edile Appennino) inherits thirty years of dam water-

proofing experience and introduced cutting-edge tools to make works even more
efficient. In carrying out the work, CEA has implemented the use of state-of-
the-art instrumentation to perform and control the work, reaching absolute quality
standards.
First and foremost, laser scanner technology is used to guarantee the water-
proofing. By means of instrumentation specially calibrated it is possible, through
∗ Barrage de Kruth-Wildenstein – Réparation du masque bitumineux
54 DOI: 10.1201/9781003211808-5
Q.104 - R.5
interpolation of the data collected before and after the bitumen has been applied,
to have data with a millimetric margin of error on the quantity of material applied.
In this manner we reduce the need of cores picking which are critical points for the
waterproofing.
As far as the determination of voids is concerned, CEA uses the PQI (Pave-
ment Quality Indicator), a non-nuclear instrument that gives an indirect indication of
voids starting from the comparison with the data obtained from the calibration of the
instrument on a core of the same material.
On the safety side, CEA patented and developed an application for smartphone
(JOBSAFER) to ensure that every one of employees uses the personal protection
equipment. Thanks to detectors installed on the PPE, the app checks that every
user is wearing all that he needs to work safely, such as helmet, gloves, glasses. If
anything were left behind, the app would warn both the employee and the manager
in real time.
RÉSUMÉ
Le barrage de Kruth-Wildenstein est un barrage en terre avec un revêtement

d’asphalte comme système de revêtement comprenant une couche de drainage en
béton et son inclinaison en amont est de 1:1,5. En 2016, une grande fissure et
d’autres défaillances ont fait l’objet d’une étude structurelle exhaustive menée par
CMC Wasserbauprojekte GmbH. Grâce à une évaluation visuelle et à 7 carottes, les
paramètres significatifs du revêtement d’asphalte ont été évalués et CMC a évalué
avec le propriétaire les principales méthodes de réhabilitation possibles:béton
asphaltique vs géomembrane. La principale raison pour laquelle le propriétaire a
opté pour le revêtement en asphalte était la faisabilité technique concernant les par-
ticularités constructives du barrage et les possibilités de raccordement aux flancs
et à la couverture.
CEA (Cooperativa Edile Appennino) a trente ans d’expérience en matière

d’étanchéité des barrages et a introduit des outils de pointe pour rendre les travaux
encore plus efficaces. Pour la réalisation des travaux, CEA a mis en place l’utilisation
d’une instrumentation de pointe pour l’exécution et le contrôle des travaux, atteignant
ainsi des normes de qualité absolues.
La technologie du scanner laser est utilisée en premier lieu pour garantir

l’étanchéité. Grâce à une instrumentation spécialement calibrée, il est possible, par
interpolation des données recueillies avant et après l’application du bitume, d’avoir
des données avec une marge d’erreur millimétrique sur la quantité de matériau
appliquée. De cette manière, nous réduisons le besoin de prélever des carottes qui
sont des points critiques pour l’imperméabilisation. En ce qui concerne la détermi-
nation des vides, CEA utilise le PQI (Pavement Quality Indicator), un instrument non
55
Q.104 - R.5
nucléaire qui donne une indication indirecte des vides à partir de la comparaison
avec les données obtenues lors de l’étalonnage de l’instrument sur une carotte du
même matériau.
Côté sécurité, CEA a breveté et développé une application pour smartphone

(JOBSAFER) afin de s’assurer que chacun des employés utilise les équipements
de protection individuelle. Grâce à des détecteurs installés sur l’EPI, l’application
vérifie que chaque utilisateur porte tout ce dont il a besoin pour travailler en toute
sécurité, comme un casque, des gants, des lunettes. En cas d’oubli, l’application
avertit en temps réel l’employé et le responsable.
1. INTRODUCTION AND HISTORY OF THE CONSTRUCTION
The dam of Kruth-Wildenstein (Barrage de Kruth-Wildenstein) is situated

within the Vosges and owned as well as operated by the Département du Haut-Rhin,
through Rivières de Haute Alsace, company created to engage in the management
of rivers and dams and other different projects regarding flood protection.
Its main function is the regulation of the water of the Thur. With a retention
of 12 Mm3 , it can help regulate low water in summer and flooding in winter/spring.
Hence, the industry based downstream can rely on a steady water supply.
The dam was constructed as an earth-fill dam with upstream asphalt concrete
sealing in the early 1960s and began its operation in 1965. Initially, the asphalt
concrete lining was protected by an armed concrete layer. However, in the early
years of operation, a high level of leakage, as well as cracks and corrosion, forced
the operator to renew the protective layer. The armed concrete layer was replaced
by an asphalt concrete layer in 1974.
The dam has a height of 38,5m. Its upstream inclination is 1:2.5, whereas the
inclination of the upstream face is 1:1,5. Upstream, a grout curtain was injected until
the bedrock (approximately 15,0m depth) (Fig. 1). All around, a drainage gallery was
built of armed concrete, and tubes for leakage coming from the face were foreseen.
On the sides, the concrete protection layer was connected to the surrounding rock
by anchorage, and injections were executed as well.
A particularity of the dam is a drainage concrete underneath the bituminous

facing. Tubes were installed within the drainage to lead water into the gallery in order
to be collected. On top of the concrete layer, an asphalt concrete lining was built with
two layers of 3,5cm each. After 1974, a bituminous protective layer was installed –
with two layers of 5,0m thickness each. Additionally, there were several applications
of white colour (mostly synthetic material) in order to avoid heating and thus reduce
the ageing of the bitumen (Fig. 2).
56
Q.104 - R.5
Fig. 1
ower part of the asphalt face (blanket).
Partie inférieure de la face asphaltée (tapis).
Fig. 2
Section of the lining system after 1974.
Section du système de revêtement après 1974.
2. STRUCTURAL SURVEY (2016–2019)
After having identified major cracks in the protection layer and an increase of
leakage, the owner decided to do a structural survey in order to prepare a possible
rehabilitation. CMC (specialist in asphalt lining with exhaustive experience from Ger-
many) conducted the structural survey by doing a visual assessment of the surface
and having a qualified company drilling (and closing) cores. Despite a lowered water
57
Q.104 - R.5
level, the visual assessment was limited due to the white protective layer on top of
the asphalt face. Additionally, the surface only considered the protective layers and
not the dense layers. However, a number of cracks and especially a large crack
could be inspected. It was clear that the crack resulted from a gliding and affected
both protective layers. (Fig. 3)
Fig. 3
Large crack due to gliding | overview and detail
Grande fissure due au glissement | vue d’ensemble et détail.
With the help of indications, it was proven that the gliding process was pro-
cessing with time. That along, however, could not be responsible for the increase in
leakage as the dense layer wasn’t affected by the gliding or visible cracks.
Hence, 7 cores were taken under supervision of CMC and closed profes-
sionally afterwards. The assessment of the cores was executed individually for the
different layers and entailed the following analysis:
• Sieving line
• Content of bitumen
58
Q.104 - R.5
Fig. 4
Ageing of bitumen and voids per layer and core.
Vieillissement du bitume et des vides par couche et par carotte.
• Percentage of voids | core and Marshall specimen

• Ageing of bitumen | penetration and ring-and-ball
• Bonding between the layers
• Content of asbestos
The following results were drawn from the analysis:
• The percentage of voids of the dense layer was too high (>3,0 vol.-%) (Fig. 4)
• Strong ageing of the bitumen in all layers – no longer the required elasticity
(Fig. 4)
• No bonding between the protection layers and between protection layer and
dense layer resulting in a high vulnerability regarding frost (especially in
combination with the high void content)
• The void content of the drainage layer, however, was lower than expected.
• Asbestos content was confirmed in 5 out of 7 analysis (<0,008 m.-%)
Additionally, to the analysis of the material and consequently to the low void
content of the drainage concrete, the overall system was taken into consideration by
testing the functionality of the drainage system. For that reason, one of the cores was
taken in a vertical line with a drainage tube, and then water was introduced in that
core hole. Afterwards, the water could be detected within the gallery. Consequently,
a general functionality of the drainage system was confirmed.
59
Q.104 - R.5
3. INSTIGATION, REASONS FOR AND OBJECTIVES OF REHABILITATION
Based on the above results, a need for rehabilitation was identified. The con-
cept of rehabilitation needed to align with the specific circumstances and challenges
of the dam:
• Gliding/lack of layer-bonding and its possible link to the high inclination;

• Ageing of bitumen;
• High void content
Along with those challenges, the owner’s objective of a rehabilitation was

mainly a durable, dense, and leakage-free construction – for the sake of security
and usability. Within the envisaged durability of at least 40 years, the maintenance
should be minimized. Aside from technical parameters, the financial burden was to
be minimized as well.
Particular attention regarding the construction in Kruth was to be taken to the

connections within the flanks: this entails the connection between the gallery and
the dense layer as well as the gallery and the rock at the side – including anchors of
the old concrete protection layer in the rocks as well as the injections. Furthermore,
the drainage system was to be kept or replaced with a better one. The level of
controllability was not to be reduced. At the same time, the drainage concrete should
be used as a support and not be replaced.
Regarding a possible gliding, extra security should be attained by a static

calculation and a “security belt” comprising abutment support.
4. STUDY OF VARIANTS
To meet the above challenges, the owner needed an appropriate and optimal
solution. Hence, it was important to give a holistic overview of the possible solution –
not only including asphalt concrete. CMC Wasserbauprojekte GmbH and its French
partner conducted a study of variants, including asphalt concrete and membrane.
In consultation with the owner, possible solutions of armed concrete were excluded
from the start as a result of the negative experiences in Kruth.
The first steps of the analysis were to identify criteria and indicators for the
evaluation and identify possible solutions which were to be considered. The following
solutions were identified and taken into consideration – in full awareness of the
existence of multiple variations of these solutions:
Asphalt concrete
1) Full rehabilitation including new drainage system: milling of all existing asphalt
concrete and building a new double asphalt lining including connections to
60
Q.104 - R.5
the gallery (4 new layers: existing drainage layer, binder, lower dense layer,
drainage layer, dense upper layer)
2) Partly rehabilitation including new drainage system: milling only parts of the
existing asphalt concrete and building new drainage layer including connec-
tions to the gallery (2 new layers: existing asphalt dense layer, new drainage
layer, new dense layer)
3) Full rehabilitation maintaining the existing drainage system: milling of all exist-
ing asphalt concrete and building a new drainage layer as well as a new dense
layer (2 new layers: existing drainage concrete, new drainage layer, new dense
layer)
Geomembrane
4) Geomembrane without new drainage system: milling of all existing asphalt

concrete and putting a geomembrane on top
5) Geomembrane with new drainage system: milling only parts of the existing
asphalt concrete and putting a new asphalt drainage layer and a supporting
layer before putting a geomembrane as the sealing element
All these 5 solutions were evaluated along the following criteria:
Technical
– Durability and resistance to external impacts:

◦ UV, frost, mechanical etc.
◦ Rodents
◦ Seismic activity
◦ Durability considering joints within layers/geomembrane
◦ Vandalism
– Recycling/ecological
– Possibilities for connecting elements to crest and flanks (especially existing
anchorage, drainage gallery and blanket construction at the upstream bottom
of the dam)
– Stability
– Drainage system
◦ New/existing
◦ Level of controllability
Planning & period of construction
– Duration and planning

– Site installation
– Costs
Operation
– Maintenance (necessities and possibilities)

– Level of surveillance
– Possibility of intervention during operation
61
Q.104 - R.5
Along with those criteria, the asphalt concrete solutions came out stronger.
The main aspects were the possibilities of connecting elements without endangering
the existing structures. Furthermore, the resistance to outer impacts was evaluated
better for the asphalt concrete solutions. Another aspect that was considered of
high importance was the risk of further gliding or any other reaction of the existing
asphalt/drainage concrete underneath the geomembrane in terms of stability of the
lining system.
Among the asphalt solutions, the decisive aspect was the drainage system. In
order to reach the highest level of controllability – thereby increasing the security –
the owner decided for a new drainage system which could be controlled in segments
in order to allocate possible leakage as best as possible. The existing drainage
concrete was kept as a support for the new asphalt concrete layers – requesting for
a compensatory layer (binder) between those two very different materials (concrete
and asphalt concrete).
With regards to the stability of the system, the owner was convinced by a static
approach taking into consideration the self-sufficient stability of the asphalt concrete
system. Through a static calculation taking into account the weight of the layers per
1m2 and the bonding forces between the layers (based on the experience of other
constructions), it was proven that the bonding was sufficient to prevent gliding. This
conclusion can only be correct given that the execution of the works follows the
accepted standards and the interlocking, and tackifying effects can come into force
sufficiently.
Completely independent from the material of construction, the scope of reha-

bilitation needed to be defined. Generally, the lower part of a dam is less exposed
and hence longer lasting. In Kruth, however, the gliding showed that also lower
parts of the dam are affected. Especially because the end of the gliding wasn’t in
sight even after lowering the water level. The results of the bituminous ageing were
independently from the level of the core. Therefore, a partial rehabilitation wasn’t
favoured by the owner.
In order to secure a constant security and water supply for the downstream
industry, a cofferdam needed to be constructed before the rehabilitation of the dam
could start.
5. EXECUTION DESIGN
The execution design followed the objectives of the owner and envisaged a
segmented drainage system (Fig. 5):
The different segments were to be separated by so called bulkheads made of

dense layer material, and each segment is connected to the gallery with two tubes.
(Fig. 6).
62
Q.104 - R.5
Fig. 5
Drainage system with segments.
Système de drainage avec segments.
Fig. 6
Bulkhead and new construction | section.
Cloison et nouvelles constructions.
The abutment (as “safety belt” for stability) continues all the way around the
gallery and is connected with anchorages to the gallery concrete. The anchorage
of the concrete protection layer (dating from 1963/initial construction) was kept in
accordance with the owner. A connection was executed by a bituminous key between
the existing armed concrete and the new abutment. In that way, the new sealing
system is – in terms of surveillance – completely apart from the small remaining
part of the existing system dating from 1963.
Due to the actual position of the gallery, the space between the abutment
and the existing was only about 25cm. It needed manual placing and manual com-
paction. Also, the joints between the abutment and the new lining system needed
an innovative filling with bituminous material in order to reach the required quality
and save time.
After emptying the reservoir, the blanket at the upstream bottom of the dam was
visually assessed, and cores were taken in order to evaluate the quality. As a result,
the owner decided to rehabilitate this part of the dam as well. A binder layer was
used before putting a 6cm dense layer. The abutment was adjusted to the necessary
high and its position moved to have a connection to the grout curtain. To reinforce the
63
Q.104 - R.5
grout injections, further injections were executed as well. Consequently, the dense
horizon of the dam was rehabilitated completely.
Regarding the risk of flooding, the execution of one horizontal joint on a certain
level was included in the tender in order to keep that option. However, the lower
dense layer, was needed to be constructed on the entire surface of the dam. The
option remained open during construction until the last minute in order to be flexible
regarding meteorological circumstances.
6. EXHECUTION PHASE – INDUSTRY 4.0 LANDS ON DAM FACINGS
CEA (Cooperativa Edile Appennino) inherits thirty years of dam waterproofing

experience and introduced cutting-edge tools to make works even more efficient.
Working on sloping surfaces requires absolutely high-performance skills and

machinery, but alongside these, new technologies are an added value of the utmost
importance that allow to reach absolute quality standards.
In carrying out the work, CEA has implemented the use of state-of-the-art
instrumentation to perform and control the work. Even a historically old work, as
it seems paving asphalt (Fig. 7), can be modernized with the application of new
technologies.
Fig. 7
Asphant lining.
Revêtement en asphalte.
64
Q.104 - R.5
Fig. 8
Milled surface, detection with laser-scanner.
surface broyée, détection par scanner laser.
6.1. CRITICAL POINTS OF WATERPROOFING AND INNOVATIVE SOLUTIONS

INTRODUCED:
One of the fundamental objectives to ensure the impermeability of the asphalt

surface is to reduce the critical points, where waterproofing risks not being guaran-
teed; we refer in particular to cold joints and the cores carried out to verify the state
of the waterproofing itself.
The cores are in fact taken after the paving to check the thickness of the
layers and to check the voids that certify state of waterproofing. The asphalt could
be considered waterproof only if the voids are under the 3%. The cores must then
be closed making sure that there are no critical points in the seal between the
pre-existing layer and the new material.
The use of new technologies makes it possible to reduce cores picking and
therefore these critical points to a minimum.
First and foremost, laser scanner technology is used. (Fig. 8).
By means of instrumentation specially calibrated to be able to carry out the

surveys on inclined surfaces and to be able to perform the reading even in the case
of reflective surfaces such as those that occur after the bitumen has been applied, it
is possible, through interpolation of the data collected before and after the bitumen
65
Q.104 - R.5
has been applied, to have data with a millimetric margin of error on the quantity of
material applied.
Laser-scanner detection is carried out through the use of a drone that flying
over the dam allows to detect all the surface without the use of markers or other
applications. In this way the relief of the various layers takes place without affecting
the quality of the work performed. In this way, the need for carrots has been kept to
a minimum with the sole purpose of having a material feedback of what is already
available in digital format.
The drone is also used with thermo temperature camera (Fig. 9) to control the
asphalt during the laying. One of the fundamental aspects to obtain a waterproof
asphalt is to be able to roll the material when the temperature is still very high.
Fig. 9
Drone thermo camera
Drone à caméra thermique.
66
Q.104 - R.5
Fig. 10
PQI.
The use of the drone allows to have a continuous control of the temperature
and therefore to be able to verify, even remotely, that the rolling takes place within
the time required by the method-statement such that the expected result can be
obtained.
As far as the determination of voids is concerned, the use of the PQI (Pavement
Quality Indicator), a non-nuclear instrument that reduce the risk of using Troxler;
through this instrument it is possible to have an indirect indication of voids starting
from the comparison with the data obtained from the calibration of the instrument
on a core of the same material. (Fig. 10).
The PQI has several appealing features relative to the nuclear density gauge,
including its light weight and rapid operation in the field without the need for special
licensing, handling, or training. The density determined with the nuclear density
gauge generally exhibited a closer correlation with the density determined from
pavement cores, as denoted by significantly lower squared error of the regression
values in the regression analyses performed.
Therefore, two cores are taken during the face test, with which the thick-
ness check is carried out as described above and the calibration of the instrument,
reducing the need for more cores.
The closing of the cores is, as already mentioned, a critical issue. In fact,
you find yourself filling holes with new material and it is necessary to guarantee a
perfect sealing between the two materials. It is feasible to close the holes according
to technical specifications, however, they remain a slightly weaker part.
Also in this case machinery has been used from what is already available on
the market, implementing them with minor technical modifications. For example, the
pestles have been heated on the surface of the plate with resistances that allow to
67
Q.104 - R.5
keep the temperature high and constant, thus avoiding the use of open flames that
are not only unsafe but also have the defect of burning the bitumen and therefore of
ruining it.
The cold joints, as mentioned above, represent a further critical point in

the construction of waterproof surfaces. To make these points absolutely water-
proof CEA has implemented cutting-edge tools that reproduce what is usually done
manually. Machinery has been developed that rivets the joints both vertically and
horizontally. In this way it obtains the incommunicability and unpredictability of man-
ual labour which is subject to the freshness of the worker and his adaptability to
work.
After being hot-worked the joints are tested by vacuum test that certify the seal
of the point.
6.2. SAFETY
Technology and technical as well as technological implementation are indis-

pensable weapons to perform the works in a workmanlike manner, but in addition to
this it is particularly in the interest of the company the safety of workers on the site.
Safety on site is one of the most complex and important aspects when working
on slopes of this type. Working at inclinations of more than 33˚ represents a constant
danger for the operators of the vehicles and for the workers who cannot move unless
they are properly restrained.
Wearing personal protective equipment such as gloves, helmet, high-visibility

jacket, work shoes and, not least, for this kind of works harnesses to prevent the fall
is not only a legal necessity but also a guarantee for workers health protection.
For that reason, CEA patented and developed an application for smartphone
(JOBSAFER) to ensure that every one of employees uses the personal protection
equipment. Thanks to detectors installed on the PPE, the app checks that every
user is wearing all that he needs to work safely, such as helmet, gloves, glasses. If
anything were left behind, the app would warn both the employee and the manager
in real time. (Fig. 11).
An important contribution to the execution of the work was given by the on-
site availability of a permanent laboratory able to perform in real time all the tests
necessary to verify the quality of the material delivered.
The materials of the four different asphalt layers, a binder, a waterproof bottom
mat with standard bitumen, a drainage without added fibres, and an upper mat with
modified bitumen, were properly studied and defined before the start of the work
68
Q.104 - R.5
Fig. 11
Job Safer.
with a special campaign of studies aimed at having the products awarded for this
type of processing.
The availability of results in a very short time, thanks to the on-site laboratory,
allows to reduce the variables and to be able to intervene immediately, without too
much loss of time, for any modifications and corrections.
The COVID was one of the major challenges for the execution of this yard.
The contract was formalized in the first days of March 2020 and activities
began in April 2020 when the pandemic was at its peak in Europe and in particular
in France. The Upper Rhine area in particular was the most affected in the first
phase of the spread of the virus with peaks of infections in the Mulhouse area. The
challenge was posed from the first moment, start the activities with all the risks of the
case or postpone the departure with the risk of not ending up in the terms imposed
by the Client.
CEA has set up a first group of workers and technicians who moved imme-
diately to France, in compliance with anti-contagion rules as applicable to work on
site. Technicians and workers worked in the first phase with masks, with disinfectant
gel always available to cope with the emergency. In addition, an application was
developed with which every day all the staff were self-monitored, promptly report-
ing any anomaly. In addition, an innovative remote control system was installed via
video and thermal cameras which saved the site manager from numerous trips.
COVID prevention and asbestos prevention were the most important chal-
lenges in order to start work. The CEA workers, already trained to work with asbestos
in italy, followed specific courses and took the necessary certificates to work in
French territory. The use of special masks, decontaminating showers and air condi-
tioning systems have been developed and used throughout the milling phase of the
face.
69
Q.104 - R.6
-------
-------
STRESS AND DEFORMATION ANALYSIS OF THE CONCRETE FACED

ROCKFILL SADDLE DAM AT GERDP AND DESIGN OF PERIPHERAL
JOINT SYSTEM (∗)
G. PIETRANGELI, G. PITTALIS & G. SIMONELLI

STUDIO ING. G. PIETRANGELI s.r.l.
P. ZAFFARONI
MAPEI
ITALY
SUMMARY
The Grand Ethiopian Renaissance Dam (GERD) Project is located along the
Blue Nile (Abbai) almost at Ethiopian-Sudanese border, a few kilometres upstream
of Roseires Dam in Sudan and 700km NE of Addis Ababa.
The plant, with its 5’150MW of installed power and 15.7TWh of annual energy
production, is one the most important projects in the Ethiopian Government’s
commitment to meet the country’s present and future power requirement, once
completed GERD will feature the largest dam in Africa.
The whole project includes a roller compacted concrete (RCC) Main Dam
(175m high, 10.2Mm3 of RCC volume) and a concrete faced rockfill (CFRD) Saddle
Dam (65m high, 5km long, 15Mm3 of embankment volume).
∗ Analyse des contraintes et des déformations du barrage de col en enrochement du
barrage de Grande Renaissance avec noyau bitumineux et conception du système de joint

périphérique
70 DOI: 10.1201/9781003211808-6
Q.104 - R.6
At site scale the geological setting of the Saddle dam foundations is composed
of five main geological units characterized by metamorphism at different grades of
both volcanic, igneous and sedimentary rocks. In particular, the Saddle dam is
founded on rock in left and right abutments (for an overall length of 1.4km) and on
residual soils of variable depth in the central part (for an overall length of 3.6km).
The boundaries between these geological units are often characterized by sharp
and sub-vertical tectonic contacts giving rise to the weak zones that are continuous
in the upstream to downstream direction and are characterized by high permeability
and potentially erodible material with different deformability characteristics.
The waterproofing of the Saddle Dam is ensured by a reinforced concrete

facing poured on the underlying curbs and by the peripheral joint system connecting
the face slabs and the plinth/gallery constituted by an internal copper waterstop and
an external waterstop system. The design of both the above elements is largely
influenced by the deformations of the dam which has been analysed by means of
2D and 3D numerical models.
The analytical work has shown that the predicted deformations are in line with
those reported in technical literature. The numerical calculations included sensitivity
analysis on key parameters (such as stiffnesses of dam and foundation materials)
and allowed the identification of more conservative scenarios for the design of the
waterproofing system. In particular, the conceptual design of the external waterstop
system foresees two different materials with specific functions:
STRUCTURAL function is guaranteed by a mono-direction carbon-fibre

layer;
WATERPROOFING is guaranteed by a PVC plasticized geo-membrane.
The high stiffness carbon-fibre grid, once loaded, develops minimal elonga-
tions that can be followed by the PVC geomembrane without developing significant
tensions, thus preserving and ensuring its water tightness function.
The length over the joint of the carbon grid and PVC membrane, based on the
predicted joint movements, is supported by a profile constituted by folded polypropy-
lene geotextile. This material is forced into the joint opening by the reservoir pressure
and, in case of leakages, provides an extra line of defense by clogging the open-
ing and limiting the inflow of water (in nautical terminology it is known as “collision
mat”).
The peripheral joint system is well suited for accommodating the predicted
deformation and is also resilient to local defects. The detail of this system has been
developed through a specific campaign of testing and full-scale prototypes carried
out by Mapei in coordination with Studio Pietrangeli.
71
Q.104 - R.6
RÉSUMÉ
Le projet de Grande Renaissance Dam (GERD) se situe le long du Nil Bleu

(Abbai) à proximité de la frontière entre l’Ethiopie et le Soudan, à quelques kilomètres
à l’amont du Barrage de Roseires au Soudan et à 700 km au nord-est d’Addis Abeba.
Avec ses 5’150MW de puissance installée et sa production annuelle d’énergie

de 15,7TWh, cette centrale hydroélectrique représente un des projets les plus impor-
tants du Gouvernement de l’Ethiopie pour faire face aux besoins d’énergie présents
et futurs du pays. Une fois la construction terminée, le projet GERD inclura le barrage
le plus grand d’Afrique.
Dans sa globalité, le projet comprend un barrage principal en béton compacté

au rouleau (BCR) de 175m d’hauteur et 10,2Mm3 de volume et un barrage de col
en enrochement avec noyau bitumineux protégé par des dalles en béton (hauteur
du barrage 65m, longueur 5km, volume 15Mm3 ).
A l’échelle du site, la géologie de la fondation du barrage de col est composée

de cinq unités géologiques principales caractérisées par des différents degrés de
métamorphisme de roches volcaniques effusives et intrusives et roches sédimen-
taires. Les appuis gauche et droit du barrage de col sont fondés en roche (pour
une longueur totale de 1.4 km) tandis que la partie centrale est fondée sur des sols
résiduels de profondeur variable (pour une longueur totale de 3.6 km). Les limites
entre les différentes unités géologiques sont souvent caractérisées par des contacts
tectoniques aigus et sub-verticaux qui produisent des zones de faiblesse continues
dans la direction amont-aval et qui se distinguent par une perméabilité élevée et par
une forte tendance à l’érosion avec des différentes caractéristiques de déformabilité
des matériaux.
Le système d’étanchéité du barrage de col est composé d’un masque amont

en béton renforcé sur les bordures sous-jacentes et par un système périphérique
de joints reliant les dalles de face avec le socle / galerie d’inspection constitué
par un “ waterstop ” interne en cuivre et par un système “ waterstop ” externe. Le
dimensionnement de ces deux éléments est largement affecté par les déformations
du barrage qui ont été analysées moyennant des modèles numériques 2D et 3D.
Ce travail d’analyse a montré que les déformations estimées sont cohérentes

avec celles de la littérature technique. Les simulations numériques incluent ainsi
des analyses de sensitivité de paramètres clés (comme par exemple la rigidité de
la fondation du barrage et la rigidité des matériaux) et ont permis d’identifier des
scenarios plus conservatifs pour le projet du système d’étanchéité. En particulier, la
conception du système “ waterstop ” externe envisage l’utilisation de deux matériaux
différents avec des fonctions spécifiques:
• La fonction structurelle est assurée par une couche monodirectionnelle de

treillis en fibre de carbone.
72
Q.104 - R.6
• La fonction d’étanchéité est assurée par une géomembrane en PVC.
Les treillis en fibre de carbone à haute rigidité produisent des allonge-

ments très petits qui peuvent être supportés par la géomembrane en PVC sans
développer des forces de traction significatives qui pourraient affecter la fonction
d’imperméabilisation de la membrane. Le joint entre le treillis en fibre de carbone et
la membrane en PVC est protégé sur toute sa longueur par une couche constituée
par un géotextile en polypropylène plié. Le géotextile est forcé dans l’ouverture du
joint par la pression du bassin et, en cas de fuites, assure une protection addition-
nelle en bouchant le joint et en limitant l’écoulement de l’eau ( collision mat dans
la terminologie nautique).
Le système périphérique de joints est conçu pour reprendre les déformations

estimées et pour avoir une certaine résilience vis-à-vis de défauts localisés. Le projet
de détail du système a été développé au moyen d’une campagne d’investigation
spécifique et de prototypes à l’échelle réelle réalisés par Mapei avec l’appui de
Studio Pietrangeli.
1. INTRODUCTION
Blue Nile (Abbai) almost at Ethiopian-Sudanese border, few kilometres upstream
of Roseires Dam in Sudan and 700 km NE of Addis Ababa, in the Benishangul –
Gumaz region.
The plant, with its 5’150MW of installed power and 15.7TWh of annual energy
production, is one the most important projects in the Ethiopian Government’s
commitment to meet the country’s present and future power requirements.
(175m high, 10.2Mm3 of RCC volume) and a concrete faced rockfill (CFRD) Saddle
Dam (65m high, 5km long, 15Mm3 of embankment volume). The 5’150MW installed
power will be generated by 13 Francis turbines in two outdoor power-houses located
at the Main Dam toe on the right and left riverside. The project also includes a nine-
bay gated-spillway, an un-gated auxiliary spillway, an emergency spillway and two
middle outlets to control the reservoir impounding [1]. The dam is, to date, under
construction and when completed, GERD will feature the largest dam in Africa.
The Project is being implemented by the Ethiopian Electric Power company

(EEP), Webuild S.p.A (former Salini-Impregilo) is the EPC contractor and Studio
Pietrangeli the designer.
The paper describes the stress and deformation analysis of the Concrete
Faced Rockfill dam in relation to the movements generated at the peripheral joint
73
Q.104 - R.6
and at the joints between adjacent slabs. The conceptual design of the peripheral
joint system and, particularly, the external waterproofing system, is then presented
and discussed in the last chapter.
2. KEY CHARACTERISTICS OF THE SADDLE DAM
2.1. GEOLOGICAL SETTINGS
The Saddle Dam of GERd Project is in a contact zone where the meta-
sediment (low metamorphic) collided with the basement rocks (high metamorphic).
The compression stress is given by an orogenic cycle known as Mozambique Belt.
The boundaries between the gneissic and volcano-sedimentary sequences are typ-
ically of tectonic origin. The geomorphology indicates that between the two dam
shoulders, constituted of schist on the left side and metagranite on the right, the
foundation material derives from the highly decomposed base rock.
At site scale the geological setting of the Saddle dam foundations is composed
of five main geological units characterized by metamorphism at different grades of
both volcanic, igneous and sedimentary rocks. The identified units are stacked by
a pre-cambrian compression regime also witnessed by the presence of folds, faults
and boudinage structures.
Figure 1 outlines a schematic geological plan reporting the main geological

formations. In particular, the Saddle Dam is founded on rock (weathered, suffix “w”
and fresh) in left and right abutments, and on residual soils (i.e. decomposed rock,
suffix “d”) of variable depth in the central part.
From left to right bank the following five main geological units are present:
• Schist (SCH)
• Femic (FEM)
• Meta-gabbro (mG)
• Meta-basalt (mBa)
• Phyllite (PH)
• Marble (M)
• Metagranite (GR)
Minor quarzitic veins are found in the left and central portion of the foundation.
Hence, the dam is founded on residual soil for a length of 3.6km and on rock for the
remaining portion of 1.4km.
The boundaries between the above-mentioned geological units are often char-
acterized by sharp tectonic contacts giving rise to the weak zones that are continuous
74
Q.104 - R.6
Fig. 1
Saddle Dam. Schematic Geological Plan.
Barrage de col. Carte géologique schématique.
in the upstream to downstream direction and are characterized by high permeabil-

ity and potentially erodible material with different deformability characteristics. The
detailed geological mapping carried out along the plinth and the inspection gallery
footprint, following the progress of the excavation and concreting activities, outlined
a complex geological setting made of sub-vertical slivers of material at different
grades of weathering [2].
2.2. PROJECT LAYOUT
The Saddle Dam of the GERD Project is a rockfill dam with an upstream
concrete facing (CFRD). Main features are summarized hereinafter:
• crest length [m] 4’865

• maximum height (H) [m] 65
• U/S slope [h/v] 1.3:1
• D/S slope [h/v] 1.3:1
• crest width [m] 8
• Embankment volume [Mm3 ] 15
• Upstream face area [m2 ] 330’000
At full supply level the overall reservoir volume will be about 80’000 Mm3 ,
65’000 Mm3 of which will be stored above the foundation level of the Saddle Dam.
75
Q.104 - R.6
As described in the previous paragraph, the dam is founded on rock in

correspondence with the abutments and in residual soil in the central portion;
consequently 2 No. main typical sections have been designed:
• Section type A
For the portion founded on rock, approximately 1’400 m long, the dam is
provided with an upstream plinth and the seepage barrier is constituted by
grout injections in correspondence of the U/S toe plinth;
• Section type B2
For the portion founded on residual soil, approximately 3’600 m long, the
dam is provided with an upstream inspection gallery and plinth, and the
seepage barrier is constituted by composite cut-off (i.e. the combination of
plastic diaphragm and grout injections) conceived to address two different
requirements: permeability correction and erosion control.
This last requirement was generally met by deepening the diaphragm down to
a level of non-erodible rock or, in case of continuous at-depth potential erodible
material, down to a level where the corresponding seepage gradient at the U/S
toe is lower than the critical gradient [3].
Five transversal galleries allow access from downstream of the inspection

gallery.
As shown in Figure 2, the embankment is subdivided, from upstream to

downstream, in the following zones:
• Zone 1A: FINE BACKFILL

Fine-graded cohesionless silt and fine sand.
• Zone 1B: COARSE BACKFILL
Random material to provide protection to zone 1A.
• Zone 2A: PERIPHERAL JOINT FILTER
Transition zone consisting of crushed rock particles up to 36 mm. In the event
of damage of the peripheral sealing system, the filter zone 2A will prevent the
movement of silt size particles through the zone.
• Zone 2B: CUSHION
This zone provides support to the curbs and the face slab and consists of
crushed sand and gravel-sized particles up to 75 mm.
• Zone 3A_1: TRANSITION 2B > 3A_2
Transition zone 6 m wide between zone 2B and zone 3A_2, consisting of
selected quarry rock with maximum size of 500 mm.
• Zone 3A_2: TRANSITION 3A_1 > 3B
Transition zone between zone 3A_1 and zone 3B consisting of selected quarry
rock with maximum size of 500 mm.
• Zone 3B: MAIN ROCKFILL
Quarry rock particles up to 1’000 mm.
• FILTER RESIDUAL SOIL > EMBANKMENT (2B FILTER)
76
Q.104 - R.6
Fig. 2
Saddle Dam. Typical section on residual soil.
Barrage de col. Coupe type sur terrain meuble.
A filter is foreseen above the portion of residual soil foundation between the
inspection gallery and the top of the upstream trench. This filter consists of
crushed sand and gravel-sized particles up to 75 mm.
• TRANSITION RESIDUAL SOIL > EMBANKMENT
A transition layer is foreseen above the portion of residual soil between the top
of the upstream trench and the D/S toe.
• DOWNSTREAM FACE
Large blocks are dozed to the downstream face of the dam to protect the
downstream face and to realize a toe with a gentler slope of 1.6:1 (h/v).
2.3. CONCRETE FACE SLAB
The primary water barrier of the CFRD is constituted by the reinforced concrete
face slabs poured on the underlying support zones, represented schematically in
Figure 3.
The support zone includes:
• extruded concrete curb (0.4 m thickness),

• bond breaker (PVC sheets) between curbs and slabs to prevent bond and con-
sequently reduce tensions in the concrete slab induced by rockfill deformation
under reservoir water load.
77
Q.104 - R.6
Fig. 3
Zoning Detail at the face slab.
Détail du zonage de la dalle de face du barrage en remblai.
The polycentric shape of the Dam centreline has been discretized with seg-
ments 50m long therefore the dam U/S face results subdivided in a series of planar
elements with a trapezoidal shape.
To simplify and homogenize the geometry of the slabs each trapezoidal ele-
ment has been subdivided in No. 3 slabs (12.1m wide) and a variable “compensation”
slab with a trapezoidal shape. Locally the spacing of the slab joints has been adjusted
to better cope with the geometry of the plinth/gallery and of the geometry/quality of
the foundation.
Starter slabs have been locally introduced where the slope of the plinth is such
as to prevent the use of the slip-formworks from the peripheral joint.
Details of the reinforced concrete slab are summarized hereafter:
• Tk [m] 0,30 Thickness of the face slab

• Wb [m] 12.1 Width of the slab (predominant)
• C [-] C20D20 Concrete class
• R [-] 1 layer 16@15 Reinforcement bars (0.45% of
slab cross section area, both ver-
tical and horizontal directions)
• c [cm] 17 Reinforcement cover
3. STRESS AND DEFORMATION ANALYSIS
3.1. GENERAL
The main objectives of the stress and deformation analysis are:
• study of the deflection of the concrete slab under reservoir loading due to
the deformation of the underling rockfill body in the cross section-plane and
verification of the reinforcement bars (2D analysis);
78
Q.104 - R.6
• estimate of the peripheral joint movement under static and dynamic loadings
in the cross-section plane (2D analysis);
• identification of the zones of dam face in compression or in tension (3D
analysis).
The stress and deformation analyses described in this chapter are focused on
the last two points.
As far as the reinforced concrete verification is concerned, it is noted that the

slabs are supported by the underlying rockfill body and are not required to resist any
transversal loading. In general, at the scale of the dam, the rockfill deformations are
small leading to a large radius of curvature of the slab deformed-shape and negligible
bending stresses. Therefore, the design of the slabs is governed by the necessity
of maintaining their water-tightness under the deformations of the embankment,
ensuring their durability (and therefore adequate protection to the reinforcement).
Both the above requirements are satisfied by designing the reinforcement bars to
limit the concrete crack openings below an assigned value.
3.2. TWO-DIMENSIONAL ANALYSIS
The two-dimensional analyses have been carried out by means of finite ele-
ment modelling (Phase2 version 8 by Rocscience and Midas GTS NX) and include
both Static analysis and Dynamic analysis.
The analyses presented in this paragraph are relevant to section Type B2

at chainage ch . 3+300 (PH_d), which is the most critical section for the settlement
analysis of the portion of the Saddle Dam founded on residual soils (maximum height
of the dam and maximum thickness of decomposed soil).
As far as the dynamic analysis is concerned, two types of analysis have been
carried out:
• Modal analysis with Response Spectrum Analysis (RSA);

• Time History analysis (TH).
In a modal analysis with Response Spectrum, modal dynamic properties are

evaluated first and then the response of each individual mode using a Response
Spectrum. Effects of the different modes are combined using the Complete
Quadratic Combination rule, which is an approximate, probabilistic based rule and
introduces loss of the sign of the computed effects.
The modal analysis is performed on the same Finite Element model used for
the static analyses with masses of all components derived from material densities.
Boundary conditions are the same as for the static analysis. Additional masses for
water interaction have been derived based on Westergaard theory (as is normal
practice).
79
Q.104 - R.6
The time-history analysis has been used mainly for validation of the RSA and
therefore one single input signal has been used. The solution has been implemented
using a modal integration technique including the first seven modes and assuming a
modal damping equal to 10% of the critical damping. In this method, a discrete time
step integration is numerically carried out for each mode and the total response is
obtained by mode superposition. The technique exploits the computational advan-
tages of dealing with uncoupled equations and permits a reduction in the number
of relevant degrees of freedom (by not including higher modes above a threshold of
engineering interest) without compromising accuracy.
Deformability characteristics of dam and foundation materials (at section

3+300) used in the numerical model are summarized in the following figure.
Fig. 4
Deformability characteristics of dam and foundation materials.
Paramètres de déformabilité du barrage et des matériaux de fondation.
Results of the 2D static analysis for section 3+300 under reservoir load are
shown in the following figures where it is possible to observe that:
• maximum total displacement of concrete slab is equal to about 13cm and is

located at about one third of the embankment height
• displacement in correspondence of the gallery foundation is equal to about
7-8cm (3cm in horizontal direction, 6-7cm in vertical direction).
80
Q.104 - R.6
The relative movements between plinth and slab are equal to about:
• ∼ 1.0 cm bridging
• ∼ 0.5 cm sliding
substantially in line with the reference values reported in literature (see Figure 6).
Fig. 5
Section 3+300 (type B2), FSL, Total displacement contour map.
Section 3 + 300 (type B2), FSL, Carte de contour de déplacement total.
Fig. 6
Section 3+300 (type B2), FSL, Peripheral joint bridging.
Section 3 + 300 (type B2), FSL, Ouverture du joint périphérique.
81
Q.104 - R.6
Additional analysis has been carried varying both the elastic modulus of rockfill
and foundation to study the influence of these parameters on the slab deflection,
opening of peripheral joint and differential settlement. In particular, the following
conservative hypothesis have been analysed:
• Rigid Foundation
Elastic modulus of foundation is increased from 400MPa to 20’000MPa to
simulate a more rigid zone in the residual soil foundation (e.g. marble area)
and increase the relative movement between embankment and gallery (the
D/S face of the gallery is in fact poured directly against the upstream excavated
trench).
• Rockfill compaction anomaly
Elastic modulus of the rockfill underlying the concrete slab is reduced to
one third of the design value to simulate a local compaction anomaly in the
embankment.
Results, in terms of peripheral joint bridging, are reported in the following figure
where:
Fig. 7
Section 3+300 (type B2), Total displacement under SEE and SEE spectrum.
Déplacement total sous charge sismique SEE et spectre de réponse SEE.
• the RED dot corresponds to the reference calculation scenario

• the GREEN dot refers to the sensitivity analysis with rigid foundation and rockfill
anomaly
• the GREY dots correspond to reference projects reported in ICOLD bulletin
No. 141 [4]
The presence of several sub-vertical contacts nearly orthogonal to the dam

axis with sharp changes in the foundation moduli, may produce a differential sliding
of the slab-to-slab joints up to 8 cm.
82
Q.104 - R.6
The seismic displacements under SEE are reported graphically in the following
figure which shows displacement contours of the whole dam.
Same displacement calculations have been carried out also through direct
integration of Time Histories.
It is observed that:
• the two methods (RSA vs TH) are well consistent.

• total additional displacement due to earthquake (at perimetral joint) is in the
range of 12mm.
• relative additional displacement due to earthquake (at perimetral joint) is in the
range of a millimetre.
3.3. THREE-DIMENSIONAL ANALYSIS
The purpose of 3D analysis is to identify zones where the slabs are expected
to be horizontally in compression and zones where they are expected to be horizon-
tally in tension. Anti-spalling reinforcement should be provided at the joints where
significant compressions are foreseen.
Since the Saddle dam presents a very large valley shape factor (A / H2 > 50)
no compression zones are foreseen, except for localized area near the abutments
or in correspondence of geometrical irregularities of the foundation.
Three partial three-dimensional models have been generated using Midas

GTS NX software, to investigate the 3D effects at the dam surface, in relation to the
face slab behaviour (right abutment left abutment, central portion of the dam).
As for the two-dimensional model, boundary conditions assume fixed displace-

ments at the base of the volume modelled and horizontal restraint on the vertical
faces.
Gravity loading and water have been applied in stages as for the two-
dimensional analysis to isolate the strains at the slab face due to the reservoir loading
(which are the only component of interest in this study).
The following figure shows an isometric view of the total dam displacements
under reservoir loading (right bank).
The following diagram shows in isometric view the compressive strains mea-
sured along horizontal fibres of the dam face (Eyy). Zones where contours are not
displayed are tensile zones.
83
Q.104 - R.6
Fig. 8
Total displacement under reservoir loading (right bank model).
Déplacement total du barrage sous charge hydraulique du réservoir (modèle de
l’appui droit).
Fig. 9
Isometric view of Eyy compressive strains under reservoir loading.
Vue isométrique des déformations compressives d’Eyy sous charge hydraulique
du réservoir.
It is observed that compression is limited to the areas at the bottom of the

right bank and in correspondence of a foundation slope change located between
ch. 4+680 and 4+740.
The other portions of the dam face are almost neutral or under low tension.
Maximum compressive strains are estimated in the order of E-4 corresponding

approximately to a compressive stress of 2-3MPa and a closure of vertical joint of
about 1mm.
84
Q.104 - R.6
In view of the results of the 3D analysis no additional anti-spalling slab rein-

forcement has been designed. For the sake of safety, a compressive joint has been
adopted for the vertical slab joints that lie within the compressive zones.
In the other portions of the dam face (where tension is observed) the maximum
calculated opening of vertical joints is in the range of few millimetres.
4. PERIPHERAL JOINT SYSTEM
The peripheral joint system, connecting the concrete face slabs and the
plinth/gallery of the CFRD, must guarantee the water tightness of the dam against
reservoir load while allowing for the anticipated movements between plinth / gallery
and face slabs [4].
The perimeter joint of the Saddle Dam is constituted by a double system: an

internal copper waterstop and an external waterstop system.
The two systems have been designed so that they can accommodate, without
being damaged, the relative movements between face slabs and plinth / gallery,
with adequate margins to cover intrinsic inaccuracies of the numerical model and
possible unforeseen local weakness/defects of embankment and foundation.
The conceptual design of the external waterstop system, which is the object
of this paragraph, foresees two different elements with specific functions:
• Structural
The structural function (bearing of the water load over the joint opening) is
carried out by a mono-directional carbon-fibre layer.
This material includes an 85gr/m2 carbon-fibre warp yarn and an 85gr/m2
fibre-glass weft yarn (Mapenet Hybrid 85/100)
• Waterproofing
The waterproofing function is guaranteed by a polyvinyl-chloride plasticized
geo-membrane with a thickness of mm (Mapeplan WTS 40).
Both materials have been supplied by Mapei who carried out, in coordination
with Studio Pietrangeli, a testing program and 1:1 scale joint prototypes in their
laboratory in Milan and on site. These prototypes allowed the fine-tuning of the
complex 3D geometry of the joint, the optimization of the shape of materials and the
fastening details to the concrete structures, in particular at the crossing with vertical
joints.
Preliminary tests carried out at Mapei laboratory indicated that the adopted
PVC geomembrane shows a progressive deformation at constant load (Figure 11).
This behaviour becomes much more evident as the testing temperature increases
85
Q.104 - R.6
Fig. 10
Prototype of the external waterstop system (Mapei, Milan).
Prototype du système “waterstop” externe (Mapei, Milan).
Fig. 11
PVC geomembrane, Mapeplan WTS 40 (4mm), creep curve at 20◦ C.
Géomembrane en PVC, Mapeplan WTS 40 (4mm), courbe de fluage à 20◦ C.
and could lead to a progressive failure of the membrane, even if the initial elongation
is well below the ultimate one. It has been therefore necessary to introduce a material
with a structural function in order to keep substantially unloaded the membrane.
86
Q.104 - R.6
The adopted 85gr/m2 carbon fibre structural grid (Mapenet Hybrid 85/100)
has an ultimate elongation of about 2% (i.e. much lower than that of the PVC
geomembrane which is about 320%) and a modulus of 245GPa (i.e. much larger
than that of the PVC geomembrane). Consequently, the high stiffness material, once
loaded, develops minimal elongations that can be followed by the PVC geomem-
brane without developing significant tensions, thus preventing its water tightness
function.
The external waterstop system is fastened at its periphery to the concrete

structures by means of stainless-steel profiles placed over an epoxy resin bedding
and threaded bars chemically anchored to the concrete structures and fastened
in order to guarantee that the periphery of the external waterstop system is in
compression all along its length.
The length of carbon grid and PVC membrane, based on the predicted joint
movements, is accommodated by a profile constituted by folded polypropylene geo-
textile. Under the pressure of water this material is forced into the opening resulting
from the movements of the joints, supporting the membrane and giving, in case of
leakages, an extra-line of defense by clogging the opening and limiting the inflow of
water (in nautical terminology it is known as “collision mat”).
The carbon-fiber structural grid has been tested in the Department of Struc-
tures for Engineering of the University of Naples “Federico II”. The tests have been
carried out according to ASTM D 3039 “Standard Test Method for Tensile Properties
of Polymer Matrix Composite Materials” in one or double sheet configuration using
specimens made up of 2 or 4 threads.
Fig. 12
Summary of Carbon grid tensile test results, Mapenet Hybrid 85/100.
Résumé des résultats des tests de traction sur les treillis en fibre de carbone.
87
Q.104 - R.6
The tests were conducted in displacement control with a constant head-speed

of 2mm/min. Aluminum tabs were connected to the specimens ends in order to
assure appropriate and adequate stress distribution in the gripping area which was
kept constant during testing at 14MPa. Test results are summarized in figure 12.
A simple calculation carried out with Mariotte formulation indicates that the
carbon grid is capable of bearing the maximum hydrostatic load of 60 m over a 30 cm
wide opening, which is more than adequate for the requirements of the peripheral
joint of the Saddle dam.
5. CONCLUSIONS
This paper discusses the main design aspects pertaining to the waterproofing
of the Saddle Dam of the Grand Ethiopian Renaissance Dam Project.
Waterproofing is ensured by a reinforced concrete facing and by the peripheral

joint system connecting the face slabs and the plinth/gallery.
The design of both the above elements is largely influenced by the deformation
of the dam under water and seismic loading.
The analytical work carried out for the estimate of the possible displacements
(consisting of two- and three-dimensional modelling) has shown that they are in
line with those reported in technical literature. The numerical calculations included
sensitivity analysis on key parameters (such as stiffnesses of dam and founda-
tion materials) and allowed the identification of more conservative scenarios for the
design of the waterproofing system, which is discussed in the last part of this paper.
In particular, the conceptual design of the external waterstop system foresees

two different materials with specific functions:
• the STRUCTURAL function is guaranteed by a mono-direction carbon-fibre

layer (85gr/m2 );
• WATERPROOFING is guaranteed by a polyvinyl-chloride plasticized geo-
membrane with a thickness of 4mm.
The elevated stiffness carbon-fibre grid, once loaded, develops minimal elon-
gations that can be followed by the PVC geomembrane without developing significant
tensions, thus preserving and ensuring its water tightness function.
The length over the joint of the carbon grid and PVC membrane, based on the
predicted joint movements, is supported by a profile constituted by folded polypropy-
lene geotextile. This material is forced into the joint opening by the reservoir pressure
88
Q.104 - R.6
and, in case of leakages, provides an extra-line of defense by clogging the opening

and limiting the inflow of water (in nautical terminology it is known as “collision mat”).
As illustrated in the paper, the system is well suited to accommodate the

predicted deformation and is also resilient to local defects. The detail of this system
has been developed through a specific campaign of testing and full-scale prototypes.
REFERENCES
[1] Pietrangeli, G., Bezzi, A., Rossini, C., Masciotta, A., D’Alberti D., (2017).
Design of Gram Ethiopian Renaissance RCC Main Dam (H = 175M). Water
Storage & Hydropower Development for Africa.
[2] Pietrangeli, G., Cacciarini, A., Calabrese, M., (2013). Geo-Mechanical Char-
acterization of Residual Soils for Foundation Design of a Bituminous Faced
Rockfill Dam. Water Storage & Hydropower Development for Africa.
[3] Pietrangeli, G., Pittalis, G., Rinaldi, M., Cifra, R., (2019). Design and Con-
struction of Composite Cut-Off System at Grand Ethiopian Renaissance Dam,
Ethiopia. Proceedings of World Tunneling Congress.
[4] Icold (2010). Concrete Face Rockfill Dams. Bulletin 141. Paris: International
Commission on Large Dams.
89
Q.104 - R.7
-------
-------
STRESS-STRAIN CHARACTERISATION OF RCC MIXES AT GERD PROJECT

AND THERMAL – SEISMIC DAM BEHAVIOUR ANALYSES (∗)
G. PIETRANGELI & A. BEZZI

STUDIO ING. G. PIETRANGELI S.r.l.
P. MASTROFINI
WE BUILD S.p.A.
A. MASCIOTTA
STUDIO MASCIOTTA S.r.l.
ITALY
SUMMARY
Blue Nile (Abbai) near to the Ethiopian-Sudanese border, a few kilometres upstream
of Roseires Dam in Sudan and 700 km NE of Addis Ababa, in the Benishangul –
Gumaz region.
The project includes a roller compacted concrete (RCC) Main Dam, with a
maximum height of 175 m, a volume of 10.2 Mm3 and a length of about 2 km at crest
elevation (645 m a.s.l.).
Extensive mix designs and testing have been carried out in order to define the
specific RCC mixes for different areas.
Five different mixes have been used with guaranteed compressive strength fc
values ranging between 8 to 15 MPa.
∗ Caractérisation de la contrainte et de la déformation des mélanges de BCR du projet
Grande Renaissance et analyses thermiques et sismiques du comportement des barrages
90 DOI: 10.1201/9781003211808-7
Q.104 - R.7
The testing campaign for RCC carried at the GERD site laboratory, included
about 500 cores from different zones of the dam body, comparing compressive and
tensile strength with accompanying stress-strain curves and modulus values for both
horizontal and vertical samples of unjointed RCC cores.
These results of the extensive test campaign carried out at the GERD site
have been used by the designer to properly characterize the tensile strain capacity
of the mixes and to optimize the RCC zoning in the dam body, as follows:
• different moduli of elasticity have been used to estimate of stresses in the

structure for the different loading conditions (static, dynamic and thermal
loads);
• the lower tensile modulus, compared with its homologue compressive mod-
ulus, has been considered for the estimate of tensile strain capacity in the
thermal analysis;
• the higher horizontal direct tensile strength, compared with its homologue
vertical direct tensile strength has also been taken into account for the thermal
analysis;
• lower values of vertical direct tensile strength, compared with the homologue
horizontal values, have been considered for the dynamic analysis.
Finally, an advanced early warning temperature monitoring procedure has

been implemented, based on the tensile strain capacity calculated according to the
above described test results, to prevent the risk of cracks which could develop in the
dam body.
More than 1’000 days of monitoring for each thermocouple, placed in different
zones of the dam, have confirmed the assumption of the thermal analysis.
RÉSUMÉ
Le projet du Grand Ethiopian Renaissance Dam (GERD) se situe le long du

Nil Bleu (Abbai) à proximité de la frontière entre l’Ethiopie et le Soudan, à quelques
kilomètres à l’amont du Barrage de Roseires au Soudan et à 700 km au nord-est
d’Addis Abeba, dans la région de Benishangul – Gumaz.
Le projet comprend un barrage principal en béton compacté au rouleau (BCR)

de 175 m d’hauteur, 10,2 Mm3 de volume et une longueur d’environ 2 km.
Une conception pointue du béton et une campagne d’investigation poussée

ont été mises en œuvre afin de définir les mélanges de BCR pour les différentes
parties du barrage.
91
Q.104 - R.7
Cinq différents mélanges ont été utilisés avec une résistance à la compression
fc comprise entre 8 et 15 MPa.
Une teneur en ciment plus élevée est employée dans la partie amont du bar-
rage, afin de respecter les exigences de perméabilité et de résistance à la traction
vis-à-vis de charges sismiques extrêmes, et pour le socle aval, pour des exigences
de résistance à la compression.
Des mélanges de BCR avec une faible teneur en ciment sont utilisés dans
la partie centrale du barrage afin de limiter l’augmentation de la température et le
risque de fissuration.
La campagne d’essai du BCR réalisée au laboratoire de chantier du GERDp

a vu l’exécution d’environ 500 carottages à différents endroits du barrage afin
d’apprécier la résistance à la compression et à la traction directe ainsi que les
courbes contraintes-déformations et les modules d’élasticité d’échantillons de BCR
sans joints prélevés en direction horizontale et verticale.
Les résultats de la campagne d’essai réalisée au GERDp ont été utilisés par
le concepteur afin d’étudier la capacité de déformation sous charge de traction de
différents mélanges de BCR et pour optimiser le zonage du corps du barrage comme
suit :
Des différents modules d’élasticité ont été utilisés pour analyser les contraintes
dans l’ouvrage correspondant à plusieurs combinaisons de chargement (statiques,
dynamiques et thermiques);
Dans l’analyse thermique, un module de traction réduit, par rapport à son

homologue en compression, a été utilisé pour estimer la capacité de déformation
en traction;
Une résistance à la traction en direction horizontale plus élevée qu’en direction

verticale a été considérée dans l’analyse thermique;
Dans l’analyse dynamique, une valeur plus faible de résistance à la traction

en direction verticale a été utilisée.
Finalement, afin d’éviter le risque de fissuration du béton, une surveillance

de la variation de température et une procédure d’alarme ont été mises en œuvre
à partir des phases initiales de la construction en se basant sur le calcul de la
capacité de déformation en traction déterminée grâce aux résultats de la campagne
d’essai.
A présent, plus de 1000 jours de surveillance de chaque thermocouple placé

à différents endroits dans le barrage ont confirmé les conclusions de l’analyse
thermique.
92
Q.104 - R.7
1. INTRODUCTION
Blue Nile (Abbai) very close to the Ethiopian-Sudanese border, a few kilometres
upstream of the Roseires Dam in Sudan and 700 km NE of Addis Ababa, in the
Benishangul – Gumaz region.
The plant, with its 5’150MW of installed power and 15.7 TWh of annual
energy production, is one the most important projects in the Ethiopian Government’s
commitment to meet the country’s present and future power requirements.
(175 m high, 10.2 Mm3 of RCC volume) and a concrete faced rockfill (CFRD) Sad-
dle Dam (65 m high, 5 km long, 15 Mm3 of embankment volume). The 5’150 MW
installed power will be generated by thirteen Francis turbines in two outdoor power-
houses located at the Main Dam toe on the right and left riverside. The project also
includes a nine-bay gated-spillway, an un-gated auxiliary spillway, an emergency
spillway and two middle outlets to allow control of the reservoir impounding. To date
the dam is under construction; when completed, GERD will feature the largest dam
in Africa.
The Project is being implemented by the Ethiopian Electric Power company

(EEP), with Webuild S.p.A (formerly Salini-Impregilo) as EPC contractor and Studio
Pietrangeli as designer.
This paper describes the extensive campaign of tests carried out at GERD
Project on the RCC mixes to investigate the stress-strain curve showing a non-
linear behaviour, as well as the correlation between horizontal and vertical direct
tensile strength for the same portion of RCC and between tensile and compressive
moduli.
The results of these tests have been used for a proper evaluation of thermal
strain capacity, and to verify mixes with lower cementitious content and less heat of
hydration for areas of high stress.
The dynamic and thermal analyses are discussed along with the methodology
adopted for thermal monitoring and control measures.
2. KEY CHARACTERISTICS OF THE RCC DAM
2.1. GENERAL LAYOUT
The Main Dam is a roller compacted concrete gravity dam with a maximum
height of 175 m and a length of about 2 km at crest elevation (645 m a.s.l.). The two
main typical sections have the following characteristics:
93
Q.104 - R.7
• Overflow Section (stepped spillway)

The upstream face has a 0.14:1 (H:V) slope in the lower portion (below elev.
575 m a.s.l.) and vertical in the upper portion. The stepped downstream face
has an average slope ranging from 0.77:1 to 0.95:1 (H:V).
• Non-Overflow Section
The upstream face has a 0.10:1 (H:V) slope in the lower portion (below elev.
545 m a.s.l.) and vertical in the upper portion. The stepped downstream face
has an average slope of 0.77:1 (H:V).
The general layout of GERD Main Dam is illustrated in Fig. 1. The key
components are:
• a river diversion system, including four culverts for the dry season discharge up
to 2’700 m3 /s (December to June) and a temporary stepped spillway located
in the central part of the dam (see Fig. 2), for dam overtopping during the wet
season up to 14’700 m3 /s;
• a free-flow crest stepped spillway located on the overflow section of the main
dam which is auxiliary to the main service gated spillway, located on a saddle
area to the immediate left of the main dam;
• two steel-lined bottom outlets (6 m diameter), embedded in the dam body,
which allow the control of reservoir level and discharge during plant outage
periods;
• thirteen penstocks (8m diameter), embedded in the dam body. Two pen-
stocks at a lower elevation are dedicated to early generation during reservoir
impounding;
• two outdoor power houses located at the Main Dam toe on the right and left
river banks housing seven and six Francis turbine units respectively, totalling
5’150 MW installed capacity;
• one 500kV switchyard on the right bank.
Fig. 1
GERD hydroelectric project, Main Dam general layout plan view.
Centrale hydroélectrique du GERD, vue en plan du barrage principal.
94
Q.104 - R.7
Fig. 2
GERD hydroelectric project, Main Dam temporary stepped spillway.
Centrale hydroélectrique du GERDp, déversoir en gradins temporaire du barrage
principal.
The dam has eighty-five monolith blocks separated by cutting joints into the
freshly compacted RCC. The vertical contraction joints are equipped in the upstream
zone with double waterstops and control drainage. The contraction joint spacing
along the dam axis varies from 18 m to 27 m.
The joint spacing is mainly controlled by the dimensions of the concrete struc-
tures of electro-mechanical equipment (penstocks, culverts and bottom outlets)
crossing the dam body.
The dam is equipped with five main longitudinal galleries, every 30–40 m of
height, located close to the upstream face and sized in order to provide an efficient
drainage system aimed at mitigating and controlling uplift pressure.
3. RCC MIX ZONING
Figure 3 illustrates the overflow section of the dam with the zoning of RCC
mixes, including their mechanical characteristics and extent of systematic bedding
mix at lift joints [5].
Extensive mix designs and testing have been carried out to define the specific
RCC mixes for different areas of the dam as described in Chapter 3 below.
Five different mixes have been used with guaranteed compressive strength fc
values ranging between 8 to 15 MPa.
CEM II-BP 32,5N (Pozzolanic cement) is being used for the most of the dam
body (to date equal to about 8,5 Mm3 ).
95
Q.104 - R.7
Fig. 3
GERD overflow section (geometry, mechanical characteristics zoning and
extension of lift bedding mix).
Section de débordement du barrage du GERD (géométrie, caractéristiques
mécaniques, zonage, extension du “ bedding mix ”).
Cement contents vary, through the cross section of the dam, from 80 to
142 kg/m3 . No fly ash has been used given the presence of pozzolan in the cement
itself. A good proportion of sound granite and weathered granite has guaranteed
enough content of fine sand in the mix to prevent segregation.
Specific admixture Mapei has been used to retard the setting time and allow
hot and warm joint for the most of the joints placed.
A higher cement content is used in both the upstream part of the dam, to meet
the tensile strength under extreme seismic load and permeability requirements, and
the downstream toe, for compressive strength requirements.
Mixes with low cement content are used in the central zone of the dam in order
to control the temperature rise and the consequent risk of cracking.
The overall average of cement content used, taking into account the volume
for each different zone, is about 105 kg/m3 .
As shown in Fig. 3, a systematic bedding mix (continuous line hatch) is pre-

scribed only in the upstream portion of the dam, in order to meet the impermeability
requirements at lift joints, and for the first 3 m above the foundation.
In the remaining part of the dam the bedding mix (dashed line) is prescribed
only for the warm and cold joints with maturity higher than the initial setting time of
the mixes (8–12 hours).
96
Q.104 - R.7
Lift joint maturity monitoring and cohesion results have been used to divide
different types of joint condition.
The impervious upstream face uses grout enriched RCC (GE-RCC) with a
width varying from 40 to 60 cm.
In the overflow section the downstream steps are protected by conventional

concrete with Silica Fume; a GE-RCC transition is foreseen between the concrete
and the RCC mixes.
A GE-RCC layer, 40 cm wide, is instead used at the downstream slope of

non-overflow section to enhance the face appearance and durability.
4. DESCRIPTION OF RCC TESTING CAMPAIGN AT GERD
The testing campaign of tensile and stress-strain analyses described in this

paper has been performed at the Grand Ethiopian Renaissance Dam (GERD)
Project during the various steps of investigation.
The testing campaign for RCC included about 500 cores from different zones
of the dam body, comparing compressive and tensile strength with accompanying
stress-strain curves and modulus values for both horizontal and vertical samples of
unjointed RCC cores [7] [8].
In addition to testing on site, some cores were also sent to the Mapei laboratory
in Milan, Italy, for further verification and comparison.
Results obtained have shown that tensile modulus was significantly lower than
the compressive modulus as shown in Table 1.
Typical compressive and tensile stress-strain curves, plotted as a percentage

of ultimate strength and ultimate strain for the GERD project are shown in Figures 4
and 5 above.
Table 1
Compressive and tensile modulus at GERDp (avg values).
Modules de compression et de traction au GERDp (valeurs moyennes).
COMPRESSIVE MODULUS TENSILE MODULUS
Ultimate Modul Modul Modul Modul Ultimate Modul Modul Modul Modul
strength 25% 50% 75% 100% strength 25% 50% 75% 100%
MPa GPa GPa GPa GPa MPa GPa GPa GPa GPa
16.5 28 22 16 7 1.1 10.5 9.5 8.5 6.5
Gerdp
Radio tensile/Compressive 38% 43% 53% 93%
97
Q.104 - R.7
Fig. 4
RCC, Typical stress-strain curves for compression at GERD.
BRC, courbes typiques contraintes-déformations en compression au GERD.
Fig. 5
RCC, Typical stress-strain curves for tension at GERD.
BRC, courbes typiques contraintes-déformations en traction au GERD.
98
Q.104 - R.7
Fig. 6
Direct tensile strength machine used at GERD site laboratory.
Équipement d’essai de traction directe utilisé au laboratoire du GERD.
Figure 6 show the equipment used at GERD site laboratory for direct tensile
strength tests.
The properties of concrete for a horizontal core are typically thought to have
the same properties as for a vertical core.
However, this is not always true for RCC [3], especially for tensile strength [7]
[8]. The tests carried out at the GERD project in the site laboratory have shown a
tensile strength from the horizontal cores greater than the tensile strength from the
homologue vertical cores sampled in the same portion of RCC.
A specific investigation campaign, including about 100 adjacent pair cores,

has been made to establish the difference between properties in the horizontal and
vertical directions.
99
Q.104 - R.7
Equivalent couples of cores were obtained by first sawing blocks and later
coring the blocks in the two directions of the same lift. Tests were carried out in each
of the three main zones in the dam with cement contents of 130 to 142 kg/m3 (A),
80 to 85 kg/m3 (B), and 110 to 119 kg/m3 (C).
For each zone, the tests have shown a much greater RCC tensile strength and
strain capacity in the horizontal direction compared to the vertical direction.
These results were duly considered in the thermal cracking analysis, because
thermal stresses develop mainly along horizontal directions.
On the contrary, seismic analysis shows that major tensile stresses are along
a vertical direction and therefore the lower values of tensile RCC strength have been
accordingly considered for structural verifications. Table 2 summarizes the results
obtained.
The overall average horizontal tensile strength was a significant 172% greater
than the vertical tensile strength.
A close examination of the concrete, including microscopy and petrography

did not identify any issue such as trapped air or microscopic horizontal separations
under coarse aggregate particles. The overall air voids of the RCC was about 1-2%.
It should be highlighted that while compressive tests are not particularly diffi-
cult and can be done in every RCC project the direct tensile tests with meaningful
strain measurements require care, expertise and equipment and need to be care-
fully verified. This was also the reason for the repetition of some specific tests also
carried out in the Mapei laboratory in Milan after hundreds of cores and cylinders
had tested in GERD site laboratory.
Table 2
Correlation between vertical and horizontal properties for different zones
(mixes) at GERD.
Corrélation entre paramètres horizontaux et verticaux pour différentes zones
(mélanges) au GERD.
DT CORES RESULTS Zone A Zone C Zone B
IN MPa
DTVP Direct tensile 1.29 Radio 1.01 Radio 0.96 Radio
vertical parent V/H V/H V/H
1.25
DTVJ Direct tensile 1.27 0.95 0.97 0.76 0.84
vertical joint
DTHP Direct tensile 1.91 1.93 1.43
horizzontal parent 1.93 65% 1.85 52% 1.48 57%
DTSP Direct tensile by 1.96 1.83 1.5
splitting parent
100
Q.104 - R.7
5. THERMAL ANALYSIS AND TEMPERATURE CONTROL
Considering the RCC volume (greater than 10 Mm3 ) and production rates (up
to a maximum of 370’000 m3 /month) of GERD Main Dam, temperature control has
been one of the most important issues during construction of this project.
The main measures to control temperature rise in the dam consist in: pre-
cooling of materials (i.e. aggregates, water), utilization of cement with lower heat of
hydration, mixes with low cement content, appropriate construction schedule, solar
radiation protection by continuous curing.
The degree of pre-cooling (i.e. maximum allowable placing temperature of

RCC) is defined by an accurate thermal study. Transient thermal analysis has been
conducted by finite differences software developed by Studio Pietrangeli in order to
evaluate the temperature distribution histories in the dam [5]. SP’s in-house thermal
model considers the main parameters that influence thermal behaviour of an RCC
dam, including: time-dependent ambient conditions (fluctuation of air temperature
and solar radiation, heat transfer by convection from the external surface of RCC
lift); time variation of thermal properties of the RCC/Grout Enriched-RCC mixes and
production parameters (placing temperature, construction start, lift height and lift
placement rate).
The thermal properties of the mixes, preliminarily estimated by conventional

laboratory tests, are calibrated by a back analysis of the first RCC dam block partially
placed and used in the river diversion scheme. This block, measuring approximately
130 × 14 × 20 m(40 000 m3 ), is purposely equipped with thirty-one thermocouples.
The results of the transient analysis together with the thermo-mechanical prop-
erties and degree of restraint present in the different locations of the dam are used
to evaluate mass and surface cracking in the RCC mass and upstream face.
The non-linear behaviour and the related increased strain capacity also provide
major benefits with regard to thermal stresses and the associated maximum allowed
temperature.
During the progress of dam construction, continuous monitoring of the RCC

temperatures at GERD has been implemented, and the measurements were peri-
odically compared with the maximum allowable temperatures deduced from thermal
studies as a function of dam zone and mechanical characteristics of the RCC mixes.
Hundreds of thermocouples have been installed for this purpose in addition to

the basic design instrumentation, and supplemental instrumentation to be installed
for long term monitoring.
101
Q.104 - R.7
These temperature limits, together with an indication of prompt actions to be

taken in case of anomalous temperatures, are included in the temperature mon-
itoring and early warning procedure implemented during the GERD Main Dam
construction.
The handy tool of early warning procedure is represented by the plot of

Figure 7, which indicates, for each zone of the Dam (different RCC mixes and ele-
vation) the temperature limits and the relevant safety factors against mass cracking.
Specific procedures are defined for each warning area M2, M3 and M4.
An example of comparison between temperatures recorded in the dam body

and the threshold limit described above is illustrated in the following figures 8, 9
and 10 for each mixes type A, B and C, respectively.
Fig. 7
Temperature early warning zones.
Seuils d’avertissement pour les variations de température.
102
Q.104 - R.7
Fig. 8
Temperatures early warning graph (upstream zone A), in the zone A1 at 0-10% of
the elevation of the dam from the foundation.
Diagramme de la variation de température (zone A, à l’amont), dans la zone A1 à
0–10% de hauteur du barrage depuis la fondation.
Fig. 9
Temperatures early warning graph (central zone B), at 10-20% of the elevation of
the dam from the foundation.
Diagramme de la variation de température (zone B, centrale), à 10-20% de
hauteur du barrage depuis la fondation.
103
Q.104 - R.7
To date, with about 8.5 Mm3 of RCC placed, temperatures recorded in the
dam body are in line with calculation predictions, confirming the good match with
the values predicted by the calibrated model.
Fig. 10
Temperatures early warning graph (downstream zone C) at 20-40% of the
elevation of the dam from the foundation.
Diagramme de la variation de température (zone C, à l’aval), à 20-40% de hauteur
du barrage depuis la fondation.
6. DYNAMIC STABILITY ANALYSIS
The stability analyses of the RCC dam were performed following an increasing
degree complexity, as usual in engineering practice.
The basic geometry of the dam was preliminarily established by a simple Rigid
Body Analysis, verifying the overturning and the sliding stability in accordance with
international standards.
At Level One Design, FEM analyses were performed to define a more reliable
and complete stress distribution within the dam body, to optimize the shape of the
dam and to outline the RCC and bedding mix zoning.
The behavior of the dam under seismic loads was evaluated by means of
Response Spectrum Analysis for the calculation of maximum expected tensile and
104
Q.104 - R.7
compressive stresses under earthquake, and by means of Equivalent Lateral Force

method for the purposes of sliding stability analysis.
The dam behavior was assumed to be essentially two-dimensional, given that

the height of the dam is much smaller than the length of the crest and therefore
considering the 3D effects as being almost negligible.
The simplified Equivalent Lateral Force Method proved to be very reliable and
it allowed us to establish the final geometry of the dam, since all the investigated 2D
cross sections respected the requirement that the first vibration mode contributed
as much as 80% or more to the total seismic response of the structure.
At the Level Two Design stage, further and more accurate analyses were
performed to confirm the final dimensions of the dam and to study local effects with
the most refined procedures proposed by literature.
In order to ponder the possibility of cracks propagating from the upstream face
of the dam, a Non Linear Static Analysis with gap friction elements was performed
using Straus 7 finite element code, release 2.4.6., as shown in the figure 11.
Fig. 11
FEM model theoretical crack propagation in deformed mesh, with gap friction
elements (red line).
Propagation théorique des fissures par modélisation aux éléments finis avec
maillage déformé et éléments d’interface (ligne rouge).
The gap friction elements are linear elements located between two nodes;
a gap can open and close during the load process and it is active only when in
compression. When an element is active, it may have a stiffness contribution both
in axial and lateral directions.
The axial force is used to model the normal contact force, while the lateral
force is used to model the friction between the surfaces on which the nodes lie.
The calculations were performed with an iterative process supposing a crack

length and the consequent uplift distribution, checking that the computed crack
length was almost equal to the assumed value, until convergence was obtained.
105
Q.104 - R.7
This further analysis allowed us to establish the optimum position of the

drainage lines; for the blocks in which a crack could be expected, the drains were
moved downstream in order prevent the theoretical crack propagate to them.
The assessment of the seismic behavior was evaluated by means of linear

time history procedure. This type of analysis involves the direct integration of the
equations of motion and therefore it is the most powerful method available in literature
for evaluating the response of structures to earthquakes.
The performance evaluation and the assessment of damage level was formu-
lated based on magnitudes of Demand-Capacity Ratios (DCR), cumulative duration
of stress excursions beyond the tensile strength of the RCC [6].
Generally, tensile stresses should not exceed tensile strength of the mix. How-
ever, a certain number of stress excursions above the tensile strength is tolerated
for dynamic loadings.
The dam response to the earthquake was considered to be within the linear-
elastic range of behavior with little or no possibility of damage, if the computed stress
demand-capacity ratios was less than or equal to 1.
The dam would exhibit nonlinear response in the form of cracking of the RCC
and/or opening of construction joints if the estimated stress demand-capacity ratios
exceeded 1.
The level of nonlinear response or cracking considered produces low or mod-

erate damage if the demand-capacity ratios are less than 2.0, damage is limited
to 15 percent of the dam cross-sectional surface area and the cumulative dura-
tion of stress excursions beyond the tensile strength of the concrete falls below the
performance curve.
The following figure 12 shows an example of results of the analysis.
Fig. 12
Time history of vertical stresses under a SEE earthquake (left) and related graph
of cumulative inelastic duration (right).
Diagramme contraintes verticales-temps sous charge sismique SEE (gauche) et
période inélastique cumulatif (droit).
106
Q.104 - R.7
Given the main direction of stress due to earthquake loads, the lower values
of RCC tensile strength, corresponding to DTVJ vertical Joint Tensile strength have
been assumed.
The compressive and tensile modules have been assumed the same but have
been increased by 30 % for dynamic loads in accordance with the recommendations
of EM 110-2-6051, and Poisson dynamic rations assumed 30% smaller than those
used for static analysis (EM 1110-2-6053).
7. CONCLUSIONS
This paper describes the results of the campaign of tests carried out at GERD
Project on the RCC mixes showing the non-linear stress-strain behavior of RCC
mixes and anisotropy of tensile strength value. These results have been used by
the designer to properly characterize the tensile strain capacity of the mixes and to
optimize the RCC zoning in the dam body.
In particular, the following main aspects have been evaluated in detail:
• different moduli of elasticity have been used to estimate stresses in the

structure for various loading conditions (static, dynamic and thermal loads);
• correlation between tensile modulus and compressive modulus, has been
investigated and considered in the design. A lower tensile modulus, com-
pared with its homologue compressive modulus, has been considered for the
estimate of tensile strain capacity in the thermal analysis;
• higher horizontal direct tensile strength, compared with its homologue vertical
direct tensile strength has been taken into account. The horizontal stresses,
resulted, on average, 1,6 times greater than the vertical one, have been consid-
ered for the thermal analysis; the lower values of vertical direct tensile strength
have been considered in the dynamic analysis;
• an early warning temperature monitoring procedure has been implemented,
based on the tensile strain capacity calculated according to the above
described test results, to prevent the risk of cracks which could develop in
the dam body;
• more than 1000 days of monitoring for each thermocouple, placed in different
zones of the dam, have confirmed the assumption of the thermal analysis;
• the experience gained in GER Dam Project has been applied, by the authors,
also to optimize the design of other RCC dams.
ACKNOWLEDGEMENTS
The authors wish to thank the Salini-Impregilo laboratory staff, for their huge
commitment in testing, the site project manager Mr. Riccardo Marinai and the site
107
Q.104 - R.7
production manager Mr. A. Passarani for their full assistance. The authors wish to
thank the Mapei technical and laboratory staff. Finally, a special thanks must be
addressed to Ernest Schrader and Mauro Giovagnoli for their continuous support
and help during every phase of the design and evaluation of test results.
REFERENCES
[1] Schrader, E. K. and Rashed, A, 2002, Benefits of Non-Linear Stress-Strain

Properties and Membranes for RCC Dams, Proceedings of the International
Conference on Roller Compacted Concrete Dam Construction in the Middle
East, Jordan University of Science and Technology, Irbid, Jordan, pp.331–344.
[2] Schrader, E., 1999, Design, Construction, and Performance of Burton Gorge
Dam, Hydropower & Dams, Issue One, PP 63–71.
[3] Schrader, E. 1995, Strain, Cracking, and Failure Described by an Ulti-

mate Modulus, Proceedings of 2nd Canmet/Aci International Symposium on
Advances in Concrete Technology, Las Vegas, PP 27–43.
[4] C. Rossini, E. Schrader, 2013, “Gibe III Dam: Project Summary, Mixes, Proper-
ties, Thermal Issues and Cores”, Water Storage & Hydropower Development
for Africa.
[5] Pietrangeli, G., Bezzi, A., Rossini, C., Masciotta, A., D’Alberti D., (2017).
Design of Gram Ethiopian Renaissance RCC Main Dam (H = 175M). Water
Storage & Hydropower Development for Africa.
[6] Masciotta A., Bezzi A., Fontana C., Fiorani A.,October 2017, Evaluation of
Dam Performance Under Seismic Loads with Linear Time History Analysis,
Case Study Grand Ethiopian Renaissance RCC Main Dam.
[7] Giovagnoli, M., Mastrofini, P., Saccone, R., and Schrader, E., March, 2018,
RCC Tensile Strength and Non-Linear Stress Strain Behaviour at Gerdp
Ethiopia. How a Better Knowledge of Tensile Strength May Help in Bet-
ter Design of RCC Dams, Hydropower&Dam 2018 Conference, Danang,
Vietnam, Aqua-Media International, Wallington, Surrey, UK.
[8] Schrader E., Mastrofini P., Le Ngoc H., Saccone R., September 2018, Correct
Tensile vs Compressive Modulus & Stress-Strain of Concrete Horizzontal vs
Vertical Fast vs Slow Conventional vs RCC. Third International Dam World
Conference. Iguazu, Brazil.
108
Q.104 - R.8
-------
-------
ASANA RIVER DIVERSION DAM DESIGN AND CONSTRUCTION (∗)
José MENDIVIL RIVAS

Senior Engineer Manager,
ANGLO AMERICAN QUELLAVECO, Peru
Michel LINO
Consultant,
ISL INGÉNIERIE, France
John H. DUQUE
Technical Practice Leader,
STANTEC, United States of America
John E. YOUNG
Senior Principal Geotechnical Engineer,
STANTEC, Canada
PERU
SUMMARY
Anglo American’s Quellaveco Project, located in the Province of Moquegua,

in the southern part of Peru is one of the largest undeveloped copper reserves in
the world. Currently under construction, the Quellaveco project requires a large
infrastructure investment, including the Asana Dam. The dam was built immediately
upstream of the open pit to divert the Asana River and its small tributaries through
a tunnel that allows the river bypass the mine site maintaining the natural quality of
the water for the communities located downstream and to protect mining operations
against flooding.
∗ Conception et construction du barrage de dérivation de la Rivière Asana
DOI: 10.1201/9781003211808-8 109

Q.104 - R.8
The highly mineralized zone surrounding the mine resulted in very challenging
geological and geotechnical site conditions include a variable weathered granodior-
ite, an elevated alluvium deposit sandwiched into the rock mass at the left abutment,
numerous joints with clay infillings, and significant faults crossing the dam founda-
tion, ranging from 0.1 to 12 m wide. As the dam location was determined primarily by
the layout of the mine and not based on optimal foundation conditions, the structure
was designed to accommodate the geology and materials available for construction.
This report presents the best practical solution selected for the diversion dam
design, determined to be a symmetrical hardfill dam with a maximum height of 50
m, and built using 187000 m3 of hardfill mix material. The design was driven by the
highly variable foundation conditions, the use of plastic colluvial materials available
near to the site and the ability to construct the spillway over the body of the dam. This
paper also describes how construction challenges were mitigated and the lessons
learned.
RÉSUMÉ
Le projet Quellaveco d’Anglo American, situé dans la province de Moquegua,

dans le sud du Pérou, est l’une des plus grandes réserves de cuivre au monde. Il est
actuellement en construction et nécessite de lourds investissements en infrastruc-
tures, et en particulier la réalisation du barrage d’Asana. Le barrage a été construit
immédiatement en amont de la fosse à ciel ouvert de la mine afin de dériver la rivière
Asana et ses petits affluents à travers un tunnel qui permet à la rivière de contourner
le site minier afin de maintenir la qualité naturelle de l’eau pour les communautés
situées en aval et protéger les opérations minières contre les inondations.
La forte minéralisation des roches dans la zone entourant la mine a entraîné

des conditions géologiques et géotechniques très difficiles, notamment un substra-
tum de granodiorite altérée et variable, un dépôt d’alluvions pris en sandwich dans la
masse rocheuse à mi-hauteur de l’appui rive gauche, de nombreux joints avec des
remplissages d’argile et des failles importantes traversant la fondation du barrage,
allant de 0.1 à 12 m de large. L’implantation du barrage étant déterminée principale-
ment par la disposition de la mine et non par des conditions de fondation optimales,
la structure a été conçue pour s’adapter à la géologie et aux matériaux disponibles
pour la construction.
Cet article présente la solution optimale de conception du barrage de dériva-

tion, qui consiste en un barrage symétrique en remblai dur d’une hauteur maximale
de 50 m pour un volume de 187000 m3 de matériaux granulaires cimentés. La con-
ception a été déterminée par les conditions de fondation très variables, l’utilisation
de matériaux colluviaux contenant des fines plastiques disponibles à proximité du
site et la possibilité de construire le déversoir sur le corps du barrage. L’article décrit
également comment les défis de construction ont été relevés et les leçons apprises.
110
Q.104 - R.8
Fig. 1
Dam Location - Southern Peru.
Localisation du barrage - au sud du Pérou.
1. INTRODUCTION
The Quellaveco Open Pit Copper Mine Project is one of Peru’s largest mining
projects located 34 kilometers east of the town of Moquegua (Fig. 1). This green-
field project will be an open pit mine, located within the Asana valley at an altitude
of 3500 meters. It will process 127500 tonnes per day of ore to produce in aver-
age 225000 tonnes of fine copper and 5500 tonnes of molybdenum per year. The
estimated ore reserves stand at 1.3 billion tonnes with approximately 7.0 million of
tonnes of fine copper. Quellaveco is the first mine operation for Anglo American
in Peru and is expected to produce first copper concentrate in 2022. To allow the
construction of early works and the open pit mine excavation, the diversion of the
Asana river and its tributaries upstream the pit is achieved with the construction of a
hydraulic system. This system includes Asana dam, a diversion tunnel, and intake
structures and channels that convey Sarallenque and Millune tributaries to a vertical
shaft structure connected to the tunnel.
2. DAM MAIN FEATURES
The Asana dam is a flood control structure. Its main functions are to protect
the mine operation against a maximum flood of 300 m3 /s with a return period of
111
Q.104 - R.8
Fig. 2
Asana Dam – Cross Section.
Barrage Asana – Section Transversale.
1 Hardfill 1 Remblai dur
2 Upstream concrete face 2 Masque amont en béton armé
3 Stepped spillway 3 Evacuateur en escalier
4 Training wall 4 Mur bajoyer
5 Grout curtain 5 Ecran d’injection
6 Ground profile 6 Profil du terrain naturel
7 Plinth 7 Plinthe
8 Stilling basin 8 Bassin d’amortissement
1000 years, and to contain and divert the flood to a 7.8 km long tunnel located along
the right abutment designed to bypass the entire mine. Recognizing the challenging
geological conditions and the high seismicity of the site (0.36 PGA), a symmetrical
hardfill type dam was selected. The dam foundations consist of variable quality hard
rock that contains weak soil-like zones in the valley floor and left abutment. The
selected dam design distributes uniform stresses safely to the weak foundations.
The dam is 50 m high, 214 m long at the crest and has a volume of 187000 m3 .
Along the upstream toe of the dam a 3.0 m wide plinth connects to the 30 m deep
grout curtain.
The dam cross section (Fig. 2) consists of a symmetrical profile with a 6 m

wide crest and upstream and downstream slopes of 0.75:1.0 (H:V). A 120 m wide
convergent stepped spillway with capacity to pass a flood of 574 m3 /seg with a
return period of 10000 years, is integrated on the downstream face of the central
monoliths. The spillway transitions at the toe of the dam into a 60 m wide by 27 m
long stilling basin. The design criteria for Inflow Design Flood (IDF) and seismic
stability analyses meet or exceed the requirements of the internal Anglo American
Technical Standard [Ref 1].
112
Q.104 - R.8
3. SITE CONDITIONS
Asana dam site selection was driven mainly by the mine open pit location,
the hydraulic structures required to divert flow from Sarallenque and Millune creeks,
and the diversion tunnel alignment. These constraints did not leave much leeway to
adjust the dam axis location and find more suitable foundation conditions.
The dam site comprises a tectonized and variably altered granodiorite rock.
In the valley floor the rock is covered, by a 3 to 8 m thick blanket of sand and gravel
material with significant quantities of cobbles and boulders. The valley slopes are
mantled by 3 to 10 m of a stony colluvial overburden. Significant geological features
include a unique alluvial deposit layer at the left abutment that is covered by a tec-
tonically emplaced broken and sheared rock mass, and a number of inclined shear
and thrust faults exposed during the foundation excavation. The dam foundation
conditions are highly variable ranging from moderately weathered competent gran-
odiorite rock at the valley to the 10 to 15 m alluvium layer located approximately at
mid height of the left abutment, as shown on Fig. 3 and 4. All foundation excavations
were performed with a large backhoe.
Fig. 3
Asana Dam – Main Geological Features - Plan View.
Barrage Asana - Principales singularités géologiques – vue en plan.
1 Fault F03 1 Faille F03
2 Hardfill dam 2 Barrage en remblai dur
3 Plinth 3 Plinthe
5 Stilling basin 5 Bassin d’amortissement
6 Steeped spillway 6 Evacuateur en escalier
(dip dir/dip: 180/5 deg) (dip dir/dip: 180/5 deg)
8 Alluvium layer 8 Formation alluviale
113
Q.104 - R.8
Fig. 4
Asana Dam – Main Geological Features – Elevation.
Barrage Asana - Principales singularités géologiques - section transversale.
(dip dir/dip: 180/5 deg.) (dip dir/dip: 180/5 deg.)
5 Alluvium Layer Formation alluviale
6 Stepped spillway Evacuateur en escalier
3.1. AT VALLEY AND LEFT ABUTMENT
The dam’s highest monoliths located at the center of the valley are founded
at El. 3593 on a slightly to moderately weathered granodiorite rock (Grades II to
III as per ISRM classification). The rock structure is blocky to very blocky with a
Geological Strength Index of 45 to 75 with open joints and joints with clay infillings
of up to 1 cm thickness that required dental concrete and grouting during foundation
preparation. A steeply inclined fault F02, runs along the center of the valley and cuts
across the dam site, as shown on Fig. 3 and 4. This feature consists of an 8 to 12 m
wide soli-like gouge zone which required special dental excavation and backfilling.
From approximately El. 3619 m to El. 3627 m, the layer of fossil alluvium is
clearly visible in the excavation (Fig. 5 and 6). The alluvium material is stratified and
inclined in layers dipping 5o −10o into the hillside. The layers include sands, gravels,
cobbles and boulders (up to approx. 1 m diameter) within a weakly cemented sandy
matrix. In general, the alluvium is fairly well-graded, although there are a few isolated
areas where the alluvium appears poorly graded with clusters of gravel and cobbles.
The conditions above the alluvium layer (approx. El. 3627 m to dam crest
El. 3641 m) are moderately to highly weathered granodiorite (Grades III to IV as per
ISRM classification). In comparison to the rock conditions below the alluvial layer
(El 3619), the upper granodiorite is much more weathered and highly fractured with
a Geological Strength Index (GSI) on the order of 25 to 50.
114
Q.104 - R.8
Fig. 5
Left Abutment - General View.
Appui gauche - Vue d’ensemble.
Fig. 6
Left Abutment - Stratified Alluvium – Sand Matrix.
Appui gauche – Alluvions stratifiées - Matrice de sable.
3.2. RIGHT ABUTMENT
At the right abutment the foundation consists of a slightly to moderately weath-

ered granodiorite (Grades II to III as per ISRM classification). The rock structure is
115
Q.104 - R.8
Fig. 7
View of excavated right abutment surface. Note location of fault F-03.
Vue de la fouille - Appui rive droite. Noter l’emplacement de la faille F-03.
blocky to very blocky with a Geological Strength Index of 45 to 75. The rock mass pre-
sented fracture rock surfaces discolored by oxidation weathering and in some cases
with penetrative weathering that weakened the rock slightly, always with enough
strength to support the foundation loads. At the top portion of the abutment fault F-
03 a 0.8 to 1.0 m thick breccia was encountered, excavated, cleaned and backfilled
with conventional dental concrete (Fig. 7).
3.3. FAULT ZONES
Three significant faults were exposed during the dam foundation and stilling
basin excavations. At the left abutment a sub-horizontal fault F-01 dipping on the
upstream direction was exposed along the base of the dam, this fault had a variable
width between 10 and 20 cm infilled with gouge and breccia material. The fault
alignment was defined by two inclined planes 100◦ /16◦ and 180◦ /5◦ (dip dir./dip)
(Fig. 8). A dam stability analysis considering the orientation and low shear strength
116
Q.104 - R.8
Fig. 8
Lower left Abutment - Fault F-01.
Partie inférieure de l’appui rive gauche - Faille F-01.
of the infilled material was performed resulting in adequate safety factors above the
minimum established criteria. This fault was treated following standard practices.
However, extensive grouting to control seepage was performed at the upstream
intersection with the dam plinth.
Fault F-02 runs adjacent to the river channel and cuts across the plinth, dam
foundation and the plunge pool. This fault has a variable width zone from 8 to 12 m
sheared, altered and brecciated rock. The location and size of this feature make its
treatment critical to the dam stability and structural behavior. A detailed description
of the fault material and the corresponding treatment is presented in Fig. 9.
Fault F-03 cuts across the upper right abutment. The fault zone varies in width
from 0.8 to 1.4 m and consists of an altered fault breccia rock fragments with bands
and matrix filling of dark grey to buff colored, plastic clay gouge. The location of this
fault was not critical to the dam stability, and the standard treatment with localized
excavation, cleaning and backfilling with dental concrete was implemented on this
feature.
3.4. FAULT F-02 TREATMENT
This steeply inclined fault cuts across the dam site and has trends at approxi-
mate azimuth 125 degrees, the fault zone has a variable width varies between 8 and
12 m. The geological mapping of the dam site has traced the fault zone through the
plunge pool to the upstream dam foundation near the plinth foundation. It is likely
that this is the regional Asana Fault which runs along the Asana valley for many
kilometers.
Fault F-02 zone consist of altered breccia with clay and embedded rock blocks
(Fig. 9). The rock mass to either side of the fault zone is blocky and competent
granodiorite. Many of the included rock fragments within the fault zone are highly
altered, partially decomposed, pale gray monzonite type rock. This is possibly a
117
Q.104 - R.8
Fig. 9
Fault F02 - photo and geological sketch of upstream exposure of the fault.
Faille F02 - photo et esquisse géologique du exposition amont du la faille.
1 Blocky granodiorite 1 Granodiorite à blocs
2 Altered granodiorite blocks in a 2 Granodiorite altérée, blocs dans
matrix of decomposed, breccia une matrice de brèche décomposée
3 Sandy breccia and gouge with 3 Brèche sableuse et gouge avec des
scattered pieces of highly altered, morceaux épars de roche cristalline
pale gray monzonite crystalline rock de monzonite gris pâle fortement
altérée
4 Blackish grey clay gouge, firm, 4 Gouge argileuse gris noirâtre, ferme,
plastic, fissile with numerous plastique, fissile avec de nombreux
fragments of highly altered rock fragments de roche fortement altérée
5 Cobbles and boulders in a sand 5 Galets et blocs rocheux dans une
and gravel matrix matrice de sable et de gravier
6 Sand and gravel with frequent 6 Sable et gravier avec des galets
cobbles and occasional small fréquents et de petits rochers
boulders, distinct horizontal occasionnels, stratification
stratification, black staining on horizontale distincte, coloration
SE side of the blackish gray clay noire côté SE de la zone d’argile gris
zone. noirâtre.
dyke that either preexists fault development or was emplaced into the fault fissure
after initial rupture.
The location and characteristics of this fault zone adversely impacted the dam
foundation conditions. The weak and low modulus fault zone material made the dam
susceptible to differential settlement that could cause cracking on the dam structure
and its plinth. To mitigate this, the fault zone was strengthened by substituting the
weak infill material with a low strength concrete plug placed along the length of the
118
Q.104 - R.8
Fig. 10
Fault F-02 Treatment – Excavation scheme and concrete backfill.
Traitement de la faille F-02– Croquis de la fouille et du remblaiement en béton.
1 Concrete backfill 1 Remblaiement en béton
2 Fault zone F-02 2 Faille F-02
3 Sound bedrock 3 Roche saine
fault for the full footprint width of the dam downstream of the plinth. The weak fault
infill material was removed to a minimum depth of 4 m. Sound rock on either side
was sloped 1H:2V in order to allow for better bridging of the concrete plug placed
on the full width of the fault (Fig. 10).
Percolation tests (ASTM D5162) in the fault zone performed at the plinth area
indicated a permeability in the order of 8.8 × 10−4 cm/sec. The high permeability
and longitudinal configuration of the fault cutting through the plinth required special
remediation measures to control seepage. At the plinth area a deeper dental exca-
vation of the fault infilled material was performed by benching inside the fault and
reaching with a backhoe to a depth of 8m where broken weathered rock material
was encountered. The gouge and breccia fault material was removed and the exca-
vated zone was backfilled with concrete effectively creating a cut-off barrier against
seepage. The dam plinth was built on top of the backfill concrete from which three
rows of 30 m deep holes were drilled and grouted providing a continuation of the
dam grout curtain (Fig. 11).
4. UPSTREAM BLANKET AT LEFT ABUTMENT
To control seepage and prevent internal erosion of the alluvial layer crossing
the upper part of the left abutment, several alternatives were analyzed including
cut-off trenches, overlapping tunnels, secant piles and a surface blanket treatment.
119
Q.104 - R.8
Fig. 11
Fault zone F02 - Cut-off and grout curtain at dam plinth.
Faille F02 - Parafouille et ecran d’injection - Barrage plinthe.
1 Plinth 1 Plinthe
2 Grout curtain, L = 30 m 2 Ecran d’injection, L = 30 m
3 Grout curtain 3 Ecran d’injection
4 Concrete cut-off 4 Paroi en béton
Based on safety considerations, availability of equipment at site and economics the

blanket protection option was selected.
In order to decrease seepage and reduce the hydraulic gradient the upstream
blanket was extended 80 m along the abutment from the dam plinth. The length of
this treatment was determined based on the critical gradient (0.2) and estimated
exit gradient downstream after applying safety factor of 2.0 for normal operation
conditions and 1.5 for un-usual conditions. The upstream treatment consists of a
120
Q.104 - R.8
Fig. 12
Left Abutment – Blanket treatment under construction.
Appui gauche – Tapis amont en construction.
flexible PVC geomembrane covering the alluvial layer, a sand layer protecting the
membrane, rockfill and a hardfill berm required for the area slope stability during
operation. Downstream of the dam, the alluvial layer was protected with a filter
and transition material designed to prevent internal erosion of the alluvial material
(Fig. 12).
5. HARDFILL DAM CONSTRUCTION
Asana hardfill construction was based on conventional RCC practice. The

main hardfill body was constructed by placing and compacting horizontal 30 cm
thick layers. Requirements for tensile strength were not necessary as the dam will
be in compression under all load conditions. A low compressive strength of 5 MPa
at 90 days suitable for the foundation conditions was required for the hardfill mix.
The upstream face was constructed with a step configuration and covered by 40 cm
of conventional unreinforced concrete for durability and water tightness purposes.
To improve water tightness between hardfill layers, bedding mix was placed at the
lift joints over a 2.0 m wide strip next to the upstream face. A compressive strength
of 15 MPa at 90 days and a maximum aggregate size of 19 mm was specified for
the bedding mix.
The aggregates for hardfill production were obtained from the colluvium
deposits of quarry Q1 located 1.9 km upstream of the dam. The quarry material
was well graded from 15 cm cobbles to angular sands and gravels, and a fine con-
tent on the order of 10% - 12%. The production of aggregates was very efficient and
consisted of scalping the + 5 cm material and screening the material into two groups
121
Q.104 - R.8
of aggregate. Adjustments to the manufacturing process were required during con-

struction as the colluvium material presented a high fines content of 12% (passing
#200 mesh) with high plasticity IP=12 and LL=30%, affecting the water content and
consistency of the hardfill mix. This was resolved by reducing the fine content of the
colluvial material from 12% to 8%, and then blending it with either 25% of imported
sand or 30% of manufactured sand at site. Resulting on a fine content of 7%, a
plasticity IP=8% and a LL=25% for the combined aggregate. The imported sand
used in this process was acquired from an external quarry (Gaby quarry) located
close to Moquegua, the manufactured sand was the result of processing at site the
scalped + 75 mm material.
A workable and stable hardfill mix was achieved with the available aggregate
and a cement content of 120 kg/m3 , and a consistency of 35 seconds Vebe time.
The average compaction density of 2270 kg/m3 was obtained with 8 passes of a
19 tons compactor with a single drum.
The hardfill mixes consistently achieved the required strength of 5 MPa.

Another important characteristic of the hardfill mix selected is the low modulus in
the order of 7 GPa, ideal for the variable conditions encountered in the foundation,
as it allows for a higher deformation of the structure without cracking.
Fig. 13
Hardfill aggregates.
Granulats pour remblai dur.
122
Q.104 - R.8
6. DAM STABILITY
FEM 2D and 3D analysis were performed to evaluate the level of stresses and
stability safety factors of the dam at critical foundation conditions. The description,
materials properties and foundation geotechnical parameters used in the analysis
are summarized in Table 1.
Table 1
Asana Dam Analysis - Design Parameters.
Material Modul Poisson’s Density Peak Peak Apparent Residual
Elast. Ratio (Kg/m3 ) Cohesion Friction (φ) Cohesion Friction
(GPa) (kPa) (kPa) (φ)
Weathered 1.2 0.25 2500 930 53◦ 150 45◦
Granodiorite
Alluvial 0.07 0.20 1900 0 37◦ 0 36◦
Layer
Hardfill 6 0.20 2200 150 48◦ 0 45◦
The sliding stability analysis was performed for normal, unusual and extreme
load conditions with required safety factors of 2, 1.5 and 1.1, respectively. Normal
load conditions of Asana dam are not critical to the stability as the reservoir shall
remain only partially full during average conditions to allow for storage of the design
flood. Analysis results for the highest dam section (H = 50 m) under unusual con-
ditions when the dam is subjected to the design flood and full reservoir pressure,
indicate that the dam stayed in compression at its base and has an adequate sliding
safety factor of 2.8 (> 1.5) when using peak strengths, and 1.9 when using residual
strengths, on both cases meeting the stability criteria. The highest dam foundation
stresses computed at this section are in the order of 0.50 MPa.
Stability analysis of the dam at the upper left abutment where the alluvial
layer is located, resulted in a sliding safety factor of 2.9 (> 1.5) under unusual flood
conditions, and the dam stayed in compression exerting pressures on the alluvial
foundation in the order of 0.28 MPa.
Additional analysis was performed during the construction stage to verify the
dam stability at the fault zones encountered during excavation. A wedge foundation
analysis was performed for monoliths 5 and 6 located over the sub-horizontal fault
F01. This feature is infilled with low shear strength material and is a kinematically
valid dam sliding mechanism. The wedge analysis was performed for the full mono-
lith with the sub-horizontal fault F01 assuming a shear strength with a frictional angle
of 25◦ and zero cohesion. Results of this analysis indicate sliding safety factors for
the unusual (flood) load case of 1.7 (> 1.5) and for the extreme (MCE) load condition
of 1.5 (> 1.1), satisfying the stability safety criteria [Ref. 1]. Wedge stability analysis
performed on the foundation along fault F02 assuming a deeper failure plane at the
base of the concrete plug, resulted on safety factors of 1.6 (> 1.5) and 1.3 (> 1.1) for
123
Q.104 - R.8
Fig. 14
FEM Model - Settlement Analysis.
Modèle MEF - Analyse du tassement.
unusual and extreme (MCE) load conditions, respectively. The results of these anal-
ysis demonstrate the adaptability of hardfill symmetrical dams founded over variable
foundation conditions, obtaining for all foundation and loading conditions adequate
sliding stability, a dam-foundation contact in compression and transfer of low and
uniform stresses on the foundation.
7. DAM SETTLEMENT ANALYSIS
A 3D FEM analysis was performed in order to estimate total dam deformations

and differential settlements between monoliths. The 3D model (Fig. 14) used for this
analysis included all dam monoliths (13) separated by contraction joints spaced
between 14 m and 24 m, the dam foundation comprising the weathered rock at the
valley and right abutment, the alluvium layer at the left abutment and the weak fault
zone F-02 located at the river valley. The application of loads to the FE model was
done in stages, starting with the construction condition where the weight of the dam
was incrementally applied to the foundation followed by the operation loads under
flood conditions. Estimates of elastic modulus used in the analysis for the different
foundation materials and materials are presented in Table 2.
Table 2
Asana Dam Foundation - Estimated Elastic Modulus
Fossil Alluvial Material E = 150 MPa (Dry conditions – during construction)
E = 70 MPa (Wet conditions – during operation)
Rock (broken/weathered) E = 1200 MPa (Construction and Operation Conditions)
Rock (blocky/competent) E = 5000 MPa (Construction and Operation Conditions)
124
Q.104 - R.8
Fig. 15
Estimated Settlement - Dam axis.
Tassement estimé - le long de l’axe du barrage.
The maximum settlement occurred at monolith 3 over the alluvium material

of the left abutment, registering 35 mm for the construction stage and increasing to
47 mm for normal operation, with 75% of the total settlement occurring during the
construction phase. A total maximum differential settlement of 27 mm was computed
at contraction joint 3 between monoliths 2 and 3 (Fig. 15), determining that the
magnitude of this displacement would put in jeopardy the waterstop embedded in
the contraction joint. As a preventive measure, additional waterstops were installed
at the contraction joints. These waterstops consist of a membrane strip anchored to
the upstream face along the joint and installed after the dam construction reached the
crest elevation after most of settlements have taken place. The estimated differential
settlement at joint 3 after dam construction was completed was in the order of 11 mm
(Fig. 15).
Estimated maximum settlement at monoliths No 7 and No 8 over the fault

zone F02 were in the order of 33 mm. The effect of the differential settlement at
the contraction joints on this central monolith were largely mitigated by concrete
plug bridging the fault, which replaced the weak fault infilling material, and by relo-
cating dam contraction joints and placing a single monolith over the fault footprint,
improving the dam arch action over the width of the fault. The maximum differential
displacement of joints at the central valley was in the order of 4 mm.
125
Q.104 - R.8
Fig. 16
Asana Dam – Piezometer PCG2 Redings.
Barrage Asana - P iézomètre PCG2.
8. CONCLUSIONS
The construction of Asana dam was completed in November 2018. Since its
completion, the dam has been periodically monitored through visual inspections
and instrumentation readings. Major storms that savaged the southern region of
Peru in February 2019, filled the reservoir to El 3620.2, about half the dam height.
Instrumentation data collected including readings for uplift pressures, seepage and
dam deformations under this partial filling, are a good measure of the dam’s good
performance. Water pressures registered at piezometers located on both abutments
were below the action trigger levels (Fig. 16), foundation seepage measured at the
seepage collection box located over the right abutment was in the order of 1.8 L/min,
dam deformations were in the order of 5 mm (Fig. 17), and field inspections did not
reveal any major issues.
Asana dam is a good application of the benefits of concrete and rockfill dams
combined into a symmetrical hardfill type dam solution. The dam was built with
available materials at site and simple construction procedures, with a mix of sufficient
strength to support the spillway on the dam body and yet with a low modulus of
elasticity that enables the structure to adapt to variable foundation conditions. The
structural shape allowed for a uniform and low transfer of stresses to the weak
foundations with adequate safety factors, in a high seismic region.
126
Q.104 - R.8
Fig. 17
Asana Dam - Survey Readings.
Barrage Asana - Mesures de tassements.
Fig. 18
Asana Dam - Construction completed.
Barrage Asana - Construction terminée.
127
Q.104 - R.8
9. ACKNOWLEDGEMENTS
The authors would like to thank Anglo American for allowing to publish the infor-
mation about this challenging project, and to acknowledge the wider Anglo American
Quellaveco project team for the support provided during design and successful
construction of the Asana Dam.
REFERENCES
[1] ANGLO AMERICAN (2019). Anglo American Technical Standard AA TS

602 001 Mineral Residue Facilities and Water Management Structures Stan-
dard and Technical Specifications. Original Version released February 2014.
https://www.angloamerican.com.
[2] ICOLD. The gravity dam, a dam for the future. Bulletin 117. Paris, France,
2000.
[3] ICOLD. Roller-compacted concrete dams - State of the art and case histories.
Bulletin 126. Paris, France, 2003.
[4] MASON P.J., R.A.N HUGHES and J.D. MOLYNEU. The design and construc-
tion of a faced symmetrical hardfill dam, 2008.
[5] ZANIAR T. Structural Safety Studies of Kahir Dam in Iran, 2010.
128
Q.104 - R.9
-------
-------
SINGULARITIES IN THE DESIGN OF “CASUPÁ DAM” (∗)
Manuel G. de MEMBRILLERA
TYPSA, Expert Consultant Spain
Margarita PINTOS
OSE, Project Manager, Uruguay
Alfonso SANTA
TYPSA, Project Manager, Spain
Ary Paulo RODRIGUES

ENGECORPS-TYPSA, Deputy Project Manager, Brazil
Claudio Michel NAHAS

ENGECORPS-TYPSA, Geotechnical Expert, Brazil
Francisco HOLANDA
HOLANDA ENGENHARIA, RCC Dam Expert, Brazil
SPAIN
SUMMARY
The design of the Casupá Creek Dam, which is a tributary of the Santa Lucia
River, has been developed with the objective of ensuring a 100% water supply
guarantee for the city of Montevideo and its Metropolitan Area, having the year
2045 as time horizon. The reservoir will add 100 hm3 of storage capacity by means
of an RCC gravity dam that will complement Paso Severino reservoir. This paper
provides a general description of the dam and construction materials.
∗ Singularités dans la conception du barrage de Casupá
DOI: 10.1201/9781003211808-9 129

Q.104 - R.9
RÉSUMÉ
La conception du barrage du ruisseau Casupá, qui est un affluent de la rivière

Santa Lucia, a été développée dans le but de garantir à 100% une alimentation en
eau pour la ville de Montevideo et sa zone métropolitaine, avec l’année 2045 comme
horizon temporel. Le réservoir ajoutera 100 hm3 de capacité de stockage avec un
barrage-poids BCR qui complétera l’actuel de Paso Severino. Ce document fournit
une description générale du barrage et des matériaux de construction.
1. INTRODUCTION
Obras Sanitarias del Estado (OSE), is a state Agency that is responsible for the
supply of drinking water in Uruguay. Nowadays, the water supply to the Montevideo
Metropolitan Area currently relies on the Paso Severino dam and the reservoir at
the Canelón Grande Creek. In order to ensure the supply of the Montevideo and
its Metropolitan Area, OSE has proposed the construction of a new dam, located in
the Santa Lucía river basin. The purpose of the Casupá project is to improve this
system in terms of quality, reliability, and water security [3].
The basic problem is not a deficit of resources in the water basin, but rather
a lack of regulation. Over the last decades, different studies have been carried out
to solve the problem and ensure the water supply to the metropolitan system [5,6].
These studies concluded that a reservoir is needed to allow the regulation of the
upper area of the basin, storing natural waters in order to ensure moderate base
flows throughout the year and peak flows when natural flows begin to decrease.
In this way, the Casupá dam and reservoir was put forward as a solution to
the needs of the water supply system to the Montevideo metropolitan area many
years ago [4]. Current designs for Casupá [7] consider an RCC gravity dam with a
maximum height of about 31.50 m over the foundation and a crest length of 750 m.
The typical cross section of the dam has vertical slopes upstream and 0.8H:1V
downstream, whereas the dam body is divided into blocks by trans-verse joints
every 20 m. The total concrete volume of the dam body including the spillway is
117,681 m3 , of which 76,435 m3 correspond to RCC and the rest to conventional
vibrated concrete.
The dam creates a reservoir with a maximum useful volume of 100 hm3 at
the normal operation level of 101.0 masl (meters above sea level) and the crest
is located at elevation 106.0 masl, thus creating a 5 m freeboard. In addition, the
water basin yields and average of 380 hm3 /year, whereas the water supply demand
is estimated at about 21 hm3 /year, having the year 2045 as time horizon.
130
Q.104 - R.9
Fig. 1
Cross sections of Casupá dam.
Sections transversales du barrage de Casupá.
1 Roller Compacted Concrete 1 Béton compacté au rouleau
2 Enriched Vibrated Concrete 2 Béton vibré enrichi
3 Drainage Gallery 3 Galerie de drainage
4 Stop-log 4 Batardeau
5 Water Intake 5 Ouvrage de prise d’eau
6 Valve Chamber 6 Chambre des vannes
To allow the construction of the dam, the diversion for the Casupá Creek is
considered in two stages; a first provisional stage on the right bank, with a rock-
carved channel, and a second stage with a channel that goes through the dam body.
The temporary 1st stage is designed for a maximum flow of 215 m3 /s, corresponding
to a return period of 2 years, whereas the 2nd stage is prepared for a 10-year event
of 603 m3 /s.
2. LOCAL MATERIALS: FOUNDATION AND QUARRIES
The development of the basic and detailed design of the Casupá project has
included the execution of several geotechnical campaigns to minimize uncertainties
regarding the design of the dam body and the foundation treatments. The Project
area is part of the San José Belt, formed by a volcano-sedimentary association of
rock structures defined as “Paso Severino and associated plutons”. The dam site
presents numerous outcrops of quartzites and quartz-sandstones with medium to
fine grain, very well selected and very consolidated. Its grains are sub-rounded, of
high sphericity and they mainly consist of quartz (90%) and lithics (1%), whereas
the cement is siliceous, constituting 5% -10% of the total rock. In general, the rock
131
Q.104 - R.9
Fig. 2
Rock outcrops at the dam site and detail of the quartz sandstone texture.
Affleurements rocheux sur le site du barrage et détail de la texture du grès
quartzeux.
fracturing present in most outcrops is important, where surface fractures appear

closed, often with iron oxides, and the spacing of these fractures is around 0.5 m.
The general conditions of the rock mass are satisfactory after removal of the
surface layer (generally less than 4 m), and the value of the deformability modulus
for the rock mass is estimated at 4,000 MPa. However, these general conditions are
locally depleted at certain points, where it has been difficult to recover samples when
drilling boreholes or very weathered materials have been found. Specifically, there
are about 160 m along the axis where the rock is intensely fractured (RQD<40 and
RMR>21) and a previous treatment has been proposed by means of high energy
and low pressure grouting, to be executed inside of previous drills, in order to wash
and mix less competent materials with cement grout.
The geotechnical campaigns and laboratory tests were also extended to the
available quarries which will provide aggregates for the concrete mixes, made up of
granites, quartz-sandstones and quartz-sands. All of them are potentially reactive
with the soluble alkalis of cement.
3. CONSTRUCTION MATERIALS FOR THE DAM BODY
All the RCC in the dam body (76.500 m3 ) can be manufactured from crushed
rock coming from one of the quartz-sandstone quarries. Even though these rocks
are potentially reactive, a cement content of 85 to 100 kg per m3 of RCC has
been designed, and the equivalent alkalis Na2 O concentration is too low (less than
1 kg/m3 ) to trigger any significant expansive reaction within the hardened RCC mix.
132
Q.104 - R.9
The cementitious materials that will complement the Portland cement will take
advantage of the materials generated in the crushing process, more precisely of the
ones passing through #200 sieve. In addition to the fact that these materials make
up the filler and cement paste, a certain pozzolanic activity nature is expected in
those crushed materials and this will offset any potential expansive reaction in the
mix.
Conventional vibrated concrete for the dam will require the use of cements
that account for the potential alkali-aggregate reactions, and thus include pozzolan
materials such as fly ash, silica fume or metakaolin. Specifically, the logistics of the
Casupá region favors the use of fly ash, considering the operation of thermoelectric
plants in the border region with Brazil. Similarly, all mortars and grouting materials
used for the treatment of the foundation will have to use pozzolanic materials along
with cement in their mixes. Table 1 summarizes all types of concrete mixes designed
for Casupá dam.
It is worth mentioning that the RCC mix has been designed to provide a com-
pression strength greater than that indicated in Table 1, which is 7 MPa at 90 days.
This strength is higher than the maximum stresses yielded by the numerical model,
and it is usual for this type of dam. During the last 2 decades, and up to 110 m height,
we can find Río Jordão, Saco de Nova Olinda, and Tibagi dams in Brazil, Burton
Gorge dam in Australia, Urugua-í in Argentina, plus Capanda and Lauca dams in
Angola – all of them with 60 to 85 kg of cementitious content per m3 of RCC.
Table 1
Concrete Mixes for Casupá Dam.
CONCRETE MAX. AGGREGATE LOCATION COMPRESSIVE CONTROL
MIX SIZE (mm) STRENGTH (MPa) AGE (DAYS)
A 25 Structural concrete, pillars, 25 28
bridge deck, outlet works,
water intake
B 25 Spillway crest, spillway canal, 25 90
stilling basin, and temporary
diversion structures
C 50 Upstream and downstream 17.5 90
dam faces
D1 50 Dental concrete in foundation 15 90
– General areas
D2 25 Dental concrete in foundation 15 90
– Difficult access areas
E 4,8 Horizontal joint mortar bed- 15 90
ding
F 50 RCC 7 90
G 25 Spillway bridge – Deck, 30 28
slabs, and pillars
133
Q.104 - R.9
4. NUMERICAL MODELS
These materials and the geotechnical conditions in the foundation have also
determined the stability calculations and the cross-section eventually designed for
some blocks of Casupá dam. The design has followed international recommenda-
tions [1,2] and Spanish guidelines [8].
On the one hand, a 2D transient thermal model assesses the maximum tem-
peratures reached during the construction stage, plus the minimum and maximum
expected thermal gradients in the concrete are obtained, both for short and long-
term conditions. The cross-section of the dam is also analyzed using a static 2D
mechanical-structural model, from which the expected strains and stresses in the
various load scenarios are obtained. Safety coefficients against sliding of the dam
and safety factors for maximum stresses in the concrete have been obtained. This
model includes several sensitivity analyses considering the most relevant resistance
parameters for the stability of the dam at representative cross-sections. In addition,
a refined structural calculation has been developed for the central area of the dam,
where singular cross-sections are located, such as the spillway, outlet works and
the water intake. For this area, a static 3D mechanical-structural model has been
deemed necessary to verify the most critical hypotheses that were found with the
2D model.
Fig. 3
Input geometry of the 3D numerical model for the central blocks of Casupá dam.
Géométrie du modèle numérique 3D pour les blocs centraux du barrage de
Casupá.
134
Q.104 - R.9
As a complement to the preceding calculations, a 2D geotechnical model has

also been developed for the dam foundation, considering the typical cross-section.
With this model, the flow net and seepage in the foundation has been estimated for
several load scenarios, in conjunction with stability calculations in the foundation.
REFERENCES
[1] ICOLD. Bulletin on Computational procedures for dam engineering. Reliability

and applicability. Paris, 2001.
[2] ICOLD. Bulletin on Guidelines for use of numerical models in dam engineering.
Paris, 2013.
[3] ICOLD. The role of water storage. Bulletin on Multipurpose Water Storage –
Essential Elements and Emerging Trends. Paris, 2016.
[4] Organización de Estados Americanos (OEA) 1970. Cuenca del río Santa
Lucía – Desarrollo de los recursos hídricos. OSP, 1970.
[5] OSE. Plan Director de Agua y Saneamiento de la cuenca del río Santa Lucía
(SOGREAH-SAFEGE-CSI). Montevideo, 1999.
[6] OSE. Selección de alternativas para el aseguramiento del abastecimiento de

agua potable al sistema metropolitano (Aguasur). Montevideo, 2013.
[7] OSE. Proyecto Básico y Documentos de Licitación para la posterior construc-

ción de la Presa del Arroyo Casupá (TYPSA-ENGECORPS). Montevideo,
2020.
[8] SPANCOLD. Guía Técnica 2 – Criterios para proyectos de presas y sus obras
anejas. Segunda parte: Presas de fábrica. Madrid, 2003.
135
Q.104 - R.10
-------
-------
IMMERSION VIBRATED RCC - INNOVATION AND PERFORMANCE CASE

STUDY: THE 103 M HIGH ENCISO DAM IN SPAIN (∗)
Francisco ORTEGA
Director, FOSCE Consulting Engineers, Spain
SPAIN
SUMMARY
IV-RCC is the simplest and best quality method to form the faces of RCC
dams. Immersion vibration of a specially designed RCC mix with Loaded VeBe time
between 6 and 12 seconds is a practical approach that solves the traditional issues
of CVC faces in RCC dams. The same mix provides the required strength and
impermeability to the dam, including transverse and horizontal joints.
RÉSUMÉ
Le BCR vibré par immersion est la méthode la plus simple et la meilleure

pour former le parement des barrages de BCR. La vibration par immersion d’une
composition désignée de manière spécifique du BCR, qui résulte du temps VeBe
surchargé entre 6 et 12 secondes, fournit une approche pratique qui résout les
problèmes traditionnels du béton vibré conventionnel dans le parement des barrages
du BCR. Le même mélange apporte la résistance et l’étanchéité exigées, en incluant
les joints transversaux et horizontaux.
∗ BCR vibré par immersion – Innovation et performance – Cas du barrage de ENCISO
(Espagne) de 103 m de hauteur
136 DOI: 10.1201/9781003211808-10

Q.104 - R.10
1. INTRODUCTION
The design and construction of Roller-Compacted Concrete (RCC) dams has

clearly evolved over recent years towards the use of much more workable mixes.
The key objective is to minimize and control segregation in order to achieve the in-
situ properties that are required for a typical good-quality concrete dam, specially at
the lift joints, that are critical for the performance of the dam.
The fact that traditional RCC placed and compacted directly against a formed
face creates a relatively poor appearance and low durability surface has been also
a critical aspect in these dams. Therefore, a number of different techniques have
been developed along the past 35 years to form the faces of RCC dams. It must
be acknowledged that actually none of them fully met to date the requirements for
quality and/or simplicity.
In recent times, a combination of design principles and construction techniques

from both RCC and Conventional Vibrated Concrete (CVC) has improved the quality
and efficiency of RCC dams. The Immersion Vibrated RCC (IV-RCC) mix, halfway
between the High Cementitious RCC (HCRCC) and the traditional very stiff CVC for
dams, has successfully solved most of the critical issues encountered in early RCC
dams. The performance after first impoundment of the recently completed Enciso IV-
RCC dam in Spain has confirmed an extraordinary improvement in quality compared
with previous RCC dams.
2. IV-RCC DAMS
2.1. CONCEPT
The IV-RCC dam is an RCC dam built using an IV-RCC mix, or several IV-RCC
mixes should a dam zoning be considered in the design. The IV-RCC mix is a very
workable RCC mix. The consistency as measured in the Slump test is zero. The
consistency as measured in the Loaded VeBe test is between 6 and 12 seconds,
with a target typical range of 8 to 10 seconds.
It has been proven in a number of cases that specifically designed high-

workability RCC mixes, with a modified VeBe time below 12 seconds, can be
consolidated by immersion vibration with no other modification of the parent RCC
mix that is consolidated by external roller compaction. For RCC mixes with higher
VeBe times the consistency needs to be modified by adding water, grout, paste or
mortar in order to achieve the capability for immersion vibration. With IV-RCC mixes
this is no longer required and the same concrete can be consolidated by either
method.
137
Q.104 - R.10
2.2. KEY FEATURES OF THE IV-RCC MIX
The following parameters are specific of the IV-RCC mixes:
• Consistency of 6-12 seconds (Loaded VeBe test) and zero slump;

• Volume of total paste (cementitious materials + aggregate fines <75 microns
+ water + chemical admixtures + entrained air) above 20% (i.e. >200 l/m3 )
relative to the total volume of the mix;
• A continuous overall gradation of aggregates designed for minimum voids;
• Maximum size of aggregate between 40 and 50 mm;
• High proportion of non-plastic aggregate fines (more than 12% <75 microns
relative to the fine aggregate);
• Void content of the compacted fine aggregate (including fines) less than 0.30;
and
• A ratio of cementitious paste (total paste excluding aggregate fines) to mortar
(cementitious paste + fine aggregate including all fines) 8±2 percentage points
higher than the void ratio of the compacted fine aggregate.
The IV-RCC mix is a further development of the high-workability and super-

retarded RCC mixes that have been described in the new ICOLD Bulletin N.177
on RCC dams. As such most IV-RCC mixes to date are high strength, high fly-
ash content mixes, using mostly crushed aggregates, and have incorporated a
significant proportion of a set retarder admixture. However such features should
not be considered specific of IV-RCC mixes. In the future, it might be possi-
ble to design IV-RCC mixes for projects with less strength requirements, and
using other Supplementary Cementitious Material (SCM) when fly-ash is not
available.
Admixtures such as set retarder, water reducer or air-entraining agents have

been successfully tested and used in high-cementitious IV-RCC mixes [1]. However
the use of such admixtures is not a basic requirement for IV-RCC mixes and should
be limited to specific projects and construction methodologies that benefit from their
associated effects.
2.3. ADVANTAGES
Compared to traditional RCC the main advantage of IV-RCC is that the facing
can be built with the same mix, and that no other mix (i.e. grout used in grout-enriched
RCC (GEVR/GERCC), CVC facing or pre-cast concrete elements) is required. This
not only simplifies construction but it also provides a strong bonding and uniformity
between the facing and the core, developing both a similar strength, elastic and
thermal properties. Just one mix conforms the entire block in a similar way as it used
to be the case in traditional CVC dams, that have been extensively and successfully
138
Q.104 - R.10
built in the past. The only difference is that in IV-RCC the mix is consolidated either
by roller compaction in the core or by immersion vibration near the faces, whilst in
CVC it was all immersion vibrated.
In addition to the dam faces, in RCC dams there are several other loca-
tions where the rollers cannot operate. The contact between the RCC and the
rock abutments, or along previously built structures (or galleries) that are embed-
ded in the dam, are typical areas where the mix needs to be consolidated by
immersion vibration. A similar situation is found when embedding instruments
that need to be protected from large equipment. In all these cases, immersion
vibration of the same RCC mix provides an excellent design and construction
opportunity.
Additionally to the direct cost benefits that are derived from the elimination of
the production, transportation and placement of a separate mix at the dam faces,
by using the IV-RCC approach, the advantages of RCC are enhanced due to the
simplification of the construction methodology, which reverts positively in a higher
speed of construction. With speed of construction comes both, quality and economy.
The fastest the dam is built, the best bonding will be achieved between layers and
the lowest will be the indirect cost associated with the construction, apart from the
obvious benefits of a potential earlier operation of the dam.
2.4. FIRST INTERNATIONAL EXPERIENCES
IV-RCC is a relatively innovative approach for concrete dam construction

worldwide. The first IV-RCC was investigated back in 2004 in the trial sections
leading to the construction of Pirrís dam in Costa Rica [2]. Although later on
during dam construction (2009-2010), varying amounts of grout were added to
the RCC along the dam faces (GEVR), at times when production was running
in the low side of the specified range (VeBe time 5 to 15 seconds), it was
possible to achieve full consolidation by immersion vibration without adding any
grout.
The first fully completed IV-RCC dam is the 90-m high De Hoop Dam in South
Africa. The IV-RCC mix was first tested in April 2009 during construction of a trial
placement built as foundation of the permanent boat slipway upstream of the dam
wall [3]. Following a comprehensive coring and testing program, the mix was finally
approved for dam construction, and consolidated by roller compaction in the dam
core, and the same mix, immersion vibrated along the dam faces, against rock con-
tacts and embedded elements. Following this very successful concrete approach,
the Spring Grove Dam was built in South Africa in 2012 using an optimized IV-RCC
mix with only 50 kg/m3 of cement + 110 kg/m3 of fly-ash [4].
139
Q.104 - R.10
Fig. 1
View of the downstream face of Enciso dam during impoundment.
Vue du parement aval du barrage de Enciso pendant le remplissage.
3. CASE STUDY: ENCISO IV-RCC DAM (SPAIN)
3.1. GENERAL
The Enciso Dam is a 103 m high straight gravity dam with a total volume of
ca. 800.000 m3 . The dam is the main component of an irrigation and water supply
system for the Rioja region, in the north part of Spain. The Ebro River Basin Authority
(Confederación Hidrográfica del Ebro), as a subsidiary of the Ministry of Agriculture
and Environment, is the Owner of the project. Impoundment began in early 2019
and water level is now (summer 2020) at 75% of the total water head.
3.2. CONCRETE MATERIALS AND MIXES
Based on the successful experience of the first IV-RCC dams in South Africa,
and after a detailed analysis of the available materials, during the final mix design
stage it was proposed to modify the original dam design concept [5]. A traditional
RCC dam with a very rich cementitious mix (110 kg/m3 cement + 110 kg/m3 fly-
ash) for the dam core had been initially specified. The interfaces had been design
with an even richer CVC mix placed in the same layers as the RCC was raised.
140
Q.104 - R.10
Past experiences in the country and elsewhere have shown that such approach
generally leads to cracking of the "skin" CVC concrete and no bonding of the joints
between its layers. Regardless of the quality of the interior RCC, a CVC face could
be a main source of permeability in the RCC dam.
The raw material for aggregate has been finally extracted from a nearby
limestone quarry and produced by dry crushing in a well calibrated three-stages
processing plant. The production of crushed fine aggregate could be improved by
using vertical shaft impact (VSI) type crusher. The gradation and particle shape was
optimized to reduce the void content of the compacted fine aggregate from 0.32-0.38
in the early tests down to 0.26-0.27. At the beginning of construction, prior to having
the final set up of the crushing plant in place, a blend of crushed fine aggregate and
natural sand allowed achieving a void content of 0.28-0.30.
A retarder admixture was not part of the initial mix concept but has been also
incorporated in the RCC mix. The main aim was to increase the time for initial set
above 15-20 hours. This would provide much more flexibility to the placement in
large areas where the mix is kept fresh until the next layer is placed on top in "hot"
joint conditions (good joint bond without any treatment).
With the optimized aggregate and the use of the admixture, the first direct
consequence was the significant reduction of the water demand of the RCC mix
for a given VeBe time. This allowed not only to reduce the cement content to a
range between 54 and 60 kg/m3 , and the total cementitious material content down
to 180-200 kg/m3 , but also to design much more workable mixes with less water.
The two IV-RCC mixes that have been finally used in the construction of the dam
have a water/cementitious material ratio of only 0.47. This value is in fact lower than
those found in many traditional CVC dams that incorporated carefully graded sands
with much less fines. Despite this relatively very low water demand, the concrete
is very workable and cohesive. VeBe time is less than 10 seconds (typical 7 to
9), no segregation is observed when placed on the dam with the traditional RCC
equipment, and a significant amount of good quality paste is brought up to the
surface during compaction with very few passes of the roller.
3.3. IMMERSION VIBRATION OF THE RCC
Four different facing systems were comparatively tested in the Full-Scale Trial:
the originally specified CVC mix, grout-enriched RCC (GEVR), mortar-enriched
RCC (MEVR) and the IV-RCC. The surface finish was very good in any of the four
methods. Due to a better quality and uniformity, IV-RCC has been selected by the
Owner after visual inspection and testing of the horizontal and vertical cores cross-
ing the different types of facing and the contact edge with the RCC. The construction
of the dam has significantly benefit from this decision.
Due to its simplicity and good quality, the same methodology has been used
in any other area of the dam where the 10 t vibratory rollers could not operate. This
141
Q.104 - R.10
includes areas near rock abutments or against the diversion culvert, against formed
gallery faces, around waterstops and drains, or embedding instruments, cables or
pendulum shafts. Immersion vibration has been generally carried out by equipment
mounted with a gang of vibrators. Manual vibration was used in narrow spaces and
around waterstops. The vibrators are high-frequency 75-mm diameter, similar as
those used for consolidation of very stiff CVC mixes.
3.4. PERFORMANCE
Cores taken of the RCC/IV-RCC contact show no signs of discontinuity at all.

The horizontal joints between layers of the immersion vibrated zones have shown an
excellent interpenetration of the particles of either layer, as good as that achieved in
the roller-compacted area. "Hot" joints are not discernible in the cores that have been
extracted from different sections of the dam, regardless of whether the consolidation
method was immersion vibration or roller compaction.
The dam has been built in stages to avoid placement during the hottest and the
coldest times of the year. The most critical dam design load is that associated with
the surface thermal cracking during the first winter after each placement. In the first
winter season, the early tensile strength capacity of the concrete was exceeded in
the lowest part of two blocks due to the relatively high thermal gradients between the
surface and the core, thus inducing two small associated cracks. These have been
later on relatively easy sealed from the gallery. A larger longitudinal thermal crack
could be observed in the center of some blocks on the horizontal winter-exposed
surface. This was bridge by reinforcing to avoid further propagation prior to resuming
placement. Apart from these thermal cracks, the concrete has shown an excellent
performance.
During the current impoundment process seepage has been recorded at the
perimetric (foundation) gallery through foundation drains and through the lower part
of the transverse joints that are more widely open. It should be noted that no water-
stops are placed either horizontally on the foundation at the location of the transverse
joints, nor around the perimetric gallery closer to the foundation. Therefore any seep-
age coming from the foundation is collected in this gallery. Additional grouting of the
foundation carried out during the last months, especially in the right abutment, has
significantly improved the foundation impermeability. The IV-RCC however remains
absolutely dry. No seepage has been observed through the vertical drains drilled
between galleries or from the waterstop drains.
ACKNOWLEDGEMENT
The author is thankful to the Ebro River Basin Authority (Confederación

Hidrográfica del Ebro) for authorization to publish this article.
142
Q.104 - R.10
REFERENCES
[1] ORTEGA F., PAULETTO M., GARCIA J. Design, construction and evaluation
of a challenging RCC trial placement for Site C Project. Proceedings of the 8 th
International RCC Symposium, 2019.
[2] ARGUEDAS I., RIVERA A., ORTEGA F. Construction of Pirrís dam in Costa
Rica. Proceedings of the 6 th International RCC Symposium, 2012.
[3] WRIGHT H.J. Roller compacted concrete adventures at De Hoop dam.

Proceedings of the SANCOLD Annual Meeting, 2009.
[4] NYAKALE J., BADENHORST D., MOHALE N., TRÜMPELMANN M.,

ORTEGA F. RCC construction aspects and quality control of Spring Grove
dam. Proceedings of the 7 th International RCC Symposium, 2015.
[5] ALLENDE M., ORTEGA F., CAÑAS J.M., NAVARRETE J. Construction of

Enciso dam. Proceedings of the 7 th International RCC Symposium, 2015.
143
Q.104 - R.11
-------
-------
EARTHQUAKE ANALYSIS OF ARCH DAMS (∗)
Anil K. CHOPRA
Professor Emeritus, UNIVERSITY OF CALIFORNIA, Berkeley, California
UNITED STATES OF AMERICA
SUMMARY
This paper (1) presents an overview of the Direct Finite Element method
for earthquake analysis of dam–water–foundation systems considering the semi-
unbounded extent of the fluid and foundation domains, including all aspects of
dam–water–foundation interaction; (2) illustrates its implementation for Morrow Point
Dam and presents sample response results; and (3) discusses methods for selection
and scaling of ground motions for the dam site.
RÉSUMÉ
Ce document (1) présente une vue d’ensemble de la méthode des éléments

finis directs pour l’analyse sismique des systèmes barrage-eau-fondations en con-
sidérant l’étendue semi-illimitée des domaines des fluides et des fondations, y
compris tous les aspects de l’interaction barrage-eau-fondation; (2) illustre sa mise
en œuvre pour le barrage de Morrow Point et présente des résultats de réponse
d’échantillons; (3) discute des méthodes de sélection et de mise à l’échelle des
mouvements du sol pour le site du barrage.
∗ Analyse sismique des barrages-voûtes
144 DOI: 10.1201/9781003211808-11

Q.104 - R.11
1. INTRODUCTION
Research on earthquake analysis of arch dams including dam–water–

foundation interaction effects and recognizing the large extent of the impounded
water and foundation rock started in the 1980s. The earliest approach that was
developed for such rigorous analysis was the substructure method, which deter-
mines the response of idealized systems (shown in Figure 1) to free-field ground
motion specified at the interface between the dam and foundation; this is the motion
that would have existed in the absence of the dam and impounded water. The sub-
structure method permits different types of models for the three substructures—dam,
fluid domain, and foundation domain: finite-element model for the dam; and “contin-
uum” models for the fluid and foundation domains of semi-unbounded geometry. The
substructure concept permits modeling of the semi-unbounded fluid and foundation
domains without truncating them to finite size and specifying the earthquake exci-
tation directly at the dam-foundation interface. However, it was restricted to simple
geometry and to linear systems, hence, incapable of estimating potential damage
to the dam during intense ground motions.
Although the importance of the aforementioned factors had been established

by the mid-1990s, many engineers continued to ignore them because most com-
mercial finite-element analysis programs did not lend themselves to including these
factors. As a result, it became popular to ignore the mass and material damping
Fig. 1
Arch dam–water–foundation system.
Système de barrage-voûte–eau–fondation.
145
Q.104 - R.11
of rock as well as compressibility of water. This practice continued until not long
ago, although it grossly overestimated response. The one notable exception is the
Bureau of Reclamation, when they started a major program for seismic evaluation
of their dams and realized that they could not obtain credible results by such crude
analyses [1].
Much of the insight into the earthquake response of dams and the influencing
factors had been gleaned from response of linear systems obtained by the substruc-
ture method. Researchers have demonstrated conclusively that several factors must
be included in dynamic analysis of concrete dams subject to earthquake excitation:
(1) the semi-unbounded extent of the impounded water and foundation domains; (2)
dam-foundation interaction considering mass, flexibility, and damping of rock; and
(3) dam-water interaction considering compressibility of water and the sediments
that invariably deposit at the reservoir bottom.
The recently developed Direct Finite-Element Method (FEM) overcomes the

limitations of the substructure method. It can be implemented in commercial com-
puter programs used by the profession for earthquake analysis of arch dams and
is applicable to nonlinear systems; thus, it is able to estimate the extent of dam-
age during intense ground motions. The objectives of this paper are to outline (1)
the FEM and illustrate its implementation for an actual dam; and (2) methods for
selection and scaling of ground motions (GMs) for the dam site.
2. DIRECT FINITE-ELEMENT METHOD
Realistic modeling of dam–water–foundation systems requires a finite-element

(FE) model that includes all the previously identified factors and truncated fluid
and foundation domains with wave-absorbing boundaries [2] to simulate the semi-
unbounded size of these domains. The seismic input is specified by effective
earthquake forces applied directly to these boundaries [2, 3], or, alternatively, in a
single layer of elements interior of the boundaries [4, 5]. Utilizing the latter approach,
Basu [6] developed an advanced analysis procedure using Perfectly Matched Layer
(PML) boundaries [7] and the Effective Seismic Input method (ESI) [4] to specify
effective earthquake forces. However, ESI and PML methodologies require mod-
ification of the FE source code, and the procedure is currently available only in
LS-DYNA.
The recently developed Direct FEM uses viscous dampers to model wave-
absorbing boundaries, and a theory is presented to compute (in an auxiliary analysis)
the effective earthquake forces that are applied at these boundaries. These features
are available in almost every commercial code.
The Direct FEM overcomes the limitations of the standard FEM by introduc-
ing wave-absorbing (or non-reflecting) boundaries at two locations: (1) upstream
146
Q.104 - R.11
Fig. 2
Finite-element model of a dam–water-foundation system with wave-absorbing
boundaries.
Modèle par éléments finis d’un système de barrage-eau-fondation avec des limites
d’absorption des vagues.
end of the fluid domain to model its essentially semi-infinite length; and (2) the bot-
tom and side boundaries of the foundation domain to model its semi-unbounded
geometry; see Figure 2. The finite-element model of the fluid domain now includes
water compressibility and reservoir bottom sediments, and the finite-element model
of the foundation domain includes mass, stiffness, and material damping appropri-
ate for the rock; water–foundation interaction is also included. Thus, the untenable
assumptions of massless rock and incompressible water in the popular FEM are
eliminated.
3. EXAMPLE NONLINEAR RESPONSE HISTORY ANALYSIS
Implementation of the Direct FEM for nonlinear response history analysis

(RHA) of an actual dam is summarized in this section; a detailed presentation is
available in Lökke and Chopra [8, 9].
3.1. SYSTEM AND GROUND MOTION
Chosen for this example analysis is Morrow Point Dam, with appropriate prop-
erties of the dam–water-foundation system. The FE model shown in Figure 3 con-
tains 27,600 finite elements and 64,000 degrees-of-freedom; 4196 solid elements to
147
Q.104 - R.11
Fig. 3
Morrow Point Dam: (a) location of contraction joints; (b) FE model of dam; (c) FE
model of fluid domain; and (d) FE model of foundation domain.
Barrage de Morrow Point : (a) emplacement des joints de contraction; (b) modèle
d’éléments finis du barrage; (c) modèle d’éléments finis du domaine des fluides; et
(d) modèle d’éléments finis du domaine des fondations.
model the dam (with four elements through the thickness of the dam), 14,175 solid
elements to model the foundation domain; and 9200 acoustic elements to model
the fluid domain. The overall dimensions of the FE model are 700 m × 700m × 400 m
corresponding to approximately 5H × 5H × 3H, where H is the height of the dam.
Included in the model were several nonlinear mechanisms: cracking of

concrete, opening and closing of contraction joints, opening and closing of the dam–
foundation interface (but relative sliding at the interface is not permitted), tensile
cracking in concrete, and effects of shear keys.
Free-field ground motion at the horizontal surface of the foundation domain is

specified as the S69E, S21W, and vertical components of the Taft ground motion—
scaled by a factor of 2—applied in the stream, cross stream, and vertical directions,
148
Q.104 - R.11
respectively. From this specified motion, the effective earthquake forces to be applied
at the absorbing boundaries are computed, and they define the excitation for the
system.
3.2. COMPUTER IMPLEMENTATION
The Direct FEM is implemented in ABAQUS, a commercial FE program, using

a pre-processing script in MATLAB that interacts with the ABAQUS input to compute
and store the effective earthquake forces that define the excitation to this system.
The analysis procedure is organized in three phases: (1) initial static analysis sim-
ulates the sequence of construction of the dam and filling of the reservoir; (2)
deconvolution analysis of a one-dimensional foundation column to determine the
free-field motions that are used to compute the effective earthquake forces at the
bottom and side boundaries; (3) nonlinear dynamic analysis of the FE model sub-
jected to the effective earthquake forces determined in Step 2; the results of Step
1 provide the displacement and stress state of the system that becomes the initial
conditions for the dynamic analysis. Detailed explanation of implementation of the
Direct FEM in ABAQUS is available in Lökke and Chopra [8].
Note that Step 2 in the above procedure can be simplified by introducing “free-
field boundary elements” at the side boundaries of the foundation domain that are
processed in parallel with the main FE model as the analysis progresses in time
[10]. However, such special elements are currently not available in most commercial
FE codes used for analysis of concrete dams.
3.3. EARTHQUAKE RESPONSE RESULTS
Commercial FE programs provide results for any response quantity of engi-

neering interest. Examples of output are presented in Figures 4–8. Displacements
in the stream, cross stream, and vertical directions at the center of the crest of the
dam are plotted as functions of time in Figure 4. Envelope values of the maximum
(over time) displacements along the length of the crest of the dam are presented in
Figure 5. Displacements and accelerations at the crest of the dam are of interest in
checking the operability of appurtenant structures, such as mechanical equipment,
gates, and roadway bridges over any spillway.
Opening of two contraction joints at the crest level are plotted as a function
of time in Figure 6, and envelope values of openings of all joints are presented in
Figure 7 The maximum opening of any contraction joint is approximately 25 mm,
which is much less than the 150 mm depth of shear keys, implying that the shear
keys remained interlocked during this ground motion.
Opening of two contraction joints at the crest level are plotted as a function
of time in Figure 6, and envelope values of openings of all joints are presented in
Figure 7 The maximum opening of any contraction joint is approximately 25 mm,
149
Q.104 - R.11
Fig. 4
Displacement histories at center of dam crest in the stream, cross stream, and
vertical directions.
Historiques des déplacements au centre de la crête du barrage dans le sens du
courant, dans le sens transversal et vertical.
Fig. 5
Envelope values of upstream and downstream displacements along the dam crest.
Valeurs d’enveloppe des déplacements amont et aval le long de la crête du
barrage.
150
Q.104 - R.11
Fig. 6
Opening of contraction joints at two locations: the joint near the right abutment
where maximum joint opening occurs, and the joint between the center cantilevers.
Ouverture des articulations de contraction à deux endroits: articulation près du
pilier droit où l’overtuse d’articulation maximale se produit, et articulation entre les
porte- à-faux centraux.
Fig. 7
Envelope values of maximum contraction joint opening along the dam crest.
Valeurs d’enveloppe de l’ouverture maximale du joint de contraction le long de la
crête du barrage.
151
Q.104 - R.11
Fig. 8
Distribution of tensile damage on (a) upstream face; (b) downstream face; and (c)
section through center (crown) cantilever.
Répartition des dommages dus à la traction sur (a) la face amont; (b) la face aval;
et (c) la section à travers la console centrale (couronne).
which is much less than the 150 mm depth of shear keys, implying that the shear
keys remained interlocked during this ground motion.
The distribution of tensile damage over the two faces of the dam is presented
in Figure 8. Such damage is greater on the downstream face and along one side of
the dam–foundation interface [Figure 8(a) and (b)]. The dam is beginning to show
signs of a semi-circular crack pattern in the upper, central part of the dam, which
has been observed as a potential failure mode during model studies of arch dams
[10]. However, no single crack has formed through the thickness of the dam (Figure
8c) to develop such a failure mode.
4. GROUND MOTION SELECTION AND MODIFICATION
4.1. TARGET SPECTRUM
4.1.1. Single Horizontal Component
The goal of nonlinear RHA of a concrete dam is to estimate seismic demands

that may be imposed by unknown ground shaking at the site caused by future
earthquakes. Because the ground motion (GM) cannot be defined uniquely, we
are interested in the statistics of structural response to an ensemble of GMs. These
GMs should, in some sense, be consistent with the seismic hazard at the site, usu-
ally characterized by the uniform hazard spectrum (UHS). Although the UHS is a
candidate for the user-specified target spectrum, researchers have demonstrated
that the UHS is overconservative and is not representative of individual GM spectra
associated with an earthquake of magnitude equal to the mean value occurring at
the mean distance. Thus, the UHS is not an appropriate target for selecting GMs to
be used in dynamic analysis of concrete dams.
152
Q.104 - R.11
Fig. 9
Uniform hazard spectrum (UHS), conditional mean spectrum (CMS), and
CMS–UHS composite spectrum for the Pine Flat Dam site.
Spectre de risque uniforme (UHS), spectre moyen conditionnel (CMS) et spectre
composite CMS-UHS pour le site du barrage de Pine Flat.
Figure 9 shows the UHS with a 1% probability of exceedance in 100 years

for the Pine Flat Dam site in California (119.3◦ W and 36.8◦ N). This exceedance
probability corresponds to a return period of 9950 years; this is the mean time
between occurrences of the specified hazard, assuming that the exceedances follow
a Poisson random process. A return period of 10,000 years is often selected for
critical facilities such as major dams and nuclear power plants.
The conditional mean spectrum (CMS) has been developed by Baker [11] as
a target spectrum (TS) that overcomes the drawbacks of the UHS. The CMS is
constructed+ for a selected value of the conditioning period, denoted by T ∗ , where
the spectral acceleration is specified. Typically, T ∗ is selected as the fundamental
vibration period of the structure and A(T ∗ ) as the UHS value. Shown in Figure 9 is
the CMS for the Pine Flat Dam site and T ∗ = 0.5 sec. It has a (slight) hump near the
conditioning period of 0.5 sec where it matches the UHS and then drops off on both
sides. Shown to be representative of recorded GMs, the CMS is appropriate as a
TS for dynamic analysis of structures.
+ MATLAB implementation of a method to compute the CMS can be downloaded from:
http://www.stanford.edu/∼bakerjw/gm_selection.html.
153
Q.104 - R.11
The conditional mean spectrum (CMS) has been developed by Baker [11] as
a target spectrum (TS) that overcomes the drawbacks of the UHS. The CMS is
constructed+ for a selected value of the conditioning period, denoted by T ∗ , where
the spectral acceleration is specified. Typically, T ∗ is selected as the fundamental
vibration period of the structure and A(T ∗ ) as the UHS value. Shown in Figure 9 is
the CMS for the Pine Flat Dam site and T ∗ = 0.5 sec. It has a (slight) hump near the
conditioning period of 0.5 sec where it matches the UHS and then drops off on both
sides. Shown to be representative of recorded GMs, the CMS is appropriate as a
TS for dynamic analysis of structures.
Researchers have demonstrated several concerns with using the CMS as

the TS to select and modify GMs. First, the CMS conditioned on a single period,
T ∗ , almost always underestimates significantly the demand because the nonlinear
dynamics of a complex structure depends on many more features of the GM beside
spectral acceleration at a single period. Second, the demands from an intensity-
based assessment with a single CMS vary with the choice of T ∗ .
The most promising proposal for the target spectrum is a composite spec-
trum that combines features from both the CMS and the UHS [12]. The CMS–UHS
composite spectrum is identical to the UHS at periods between Tmin and Tmax —the
shortest and longest structural periods of interest—but less intense than the UHS
outside of this period range. At periods shorter than Tmin , it is identical to the CMS
with T ∗ = Tmin , whereas at periods longer than Tmax , it is identical to the CMS, with
T ∗ = Tmax ; see Figure 9.
4.1.2. Two Horizontal Components
Researchers have demonstrated that it is appropriate to define the target spec-

trum as a pair of identical spectra for the two horizontal components of GM [13]. This
is the CMS–UHS composite spectrum presented in Figure 9.
4.2. GROUND-MOTION SELECTION AND AMPLITUDE SCALING
The simplest goal of ground-motion selection and modification (GMSM) pro-

cedures is to estimate the median (or mean) values of seismic demands. For this
purpose, we want to select GMs whose response spectra are similar—in some
sense—to the target spectrum in amplitude and in shape. The number of recorded
GMs that satisfy both requirements simultaneously is often insufficient. For exam-
ple, the large majority of recorded GMs are weaker than the intensity represented
by the UHS and CMS in highly seismic regions. Furthermore, the response spectra
+ MATLAB implementation of a method to compute the CMS can be downloaded from:
http://www.stanford.edu/∼bakerjw/gm_selection.html.
154
Q.104 - R.11
Fig. 10
Response spectra for 11 scaled GMs selected for similarity with the CMS; the
period range for matching of GMs is 0.05–10 sec (not shown in the figure).
Spectres de réponse mis à l’échelle pour 11 mouvements de sol (GM)
sélectionnés pour leur similarité avec le CMS; la période de correspondance des
GM est de 0,05 à 10 secondes (non indiquée dans la figure).
for many of the records with the desired intensity may not be similar in shape to the
TS. Given this background, selection of GMs usually proceeds in two stages. First,
every record in the database is scaled to make its spectral amplitude(s) similar to
the target amplitude(s). Second, the scaled records whose response spectra are
similar in shape to the target spectrum over a specified period range are selected.
This period range for selecting GMs should include the vibration periods of all modes
that are significant in the response of the structure.
The first step of scaling GMs to an intensity similar to that of the TS can be
achieved in various ways, e.g., scale each GM so that the selected scale factor (SF)
minimizes the difference between the response spectrum of the scaled GM and TS
over the selected period range. A good metric for this “difference” is the sum of the
squared differences (SSD) between the two spectra over the selected period range.
To preserve the relative intensity of the two horizontal components in the original
recorded GM, both components are scaled by the same factor. Once all the GMs are
in a large database have been scaled appropriately, the SSD between the response
spectra pair for each GM and the TS pair is computed and the GMs with the smallest
SSD are selected.
155
Q.104 - R.11
Figure 10 shows the CMS (chosen, for example, as the TS) and the response
spectra for 11 GMs selected for similarity to the CMS. Presented for illustration, this
figure is restricted to one component of GM.
4.3. THREE COMPONENTS OF GROUND MOTION
The methodology to develop target spectra for two horizontal components

of GM, and then to select and scale GMs to match the TS outlined above, was
presented in Sections 4.1 and 4.2. This methodology has been extended to three
components of GM. Derived was a theory to develop the target spectrum for the
vertical component of the GM and the GMSM method was extended to consider
simultaneously the vertical component [14]. This methodology requires the user to
specify the relative importance of the horizontal and vertical components of GM in
the response of the structure.
Since horizontal GMs are usually much more important in the response of
concrete dams, a simpler approach is to first select these two components without
considering the vertical component. The target spectrum for the horizontal com-
ponents of the GM defined by Figure 9 remains valid, and a TS for the vertical
component is not required. The vertical component accompanying a selected record
is scaled by the same SF that was applied to the horizontal components
REFERENCES
[1] BUREAU OF RECLAMATION. Linear elastic dynamic structural analysis

including mass in the foundation for Hoover Dam, 1988, Technical Memo-
randum HVD-MDA-D8110-97-2, Denver, CO.
[2] WOLF J.P. Soil-Structure Interaction in Time Domain, 1988, Englewood Cliffs,
NJ: Prentice Hall.
[3] ZIENKIEWICZ O.C., BICANIC N., SHEN F.-Q. Earthquake input definition and
the transmitting boundary conditions. In: Advances in Computational Nonlinear
Mechanics, 1989. Springer-Verlag.
[4] BIELAK J., CHRISTIANO P. On the effective seismic input for non-linear soil-
structure interaction systems, 1984, Earthquake Engineering and Structural
Dynamics, 12(3), 107–119.
[5] BIELAK J., LOUKAKIS K., HISADA Y., YOSHIMURA C. Domain reduction
method for three-dimensional earthquake modeling in localized regions, Part
I: Theory, 2003, Bulletin of the Seismological Society of America, 93(2),
817–824.
156
Q.104 - R.11
[6] BASU U. Perfectly Matched Layers for Acoustic and Elastic Waves: The-
ory, Finite-Element Implementation and Application to Earthquake Analysis
of Dam-Water-Foundation Rock Systems, 2004, PhD thesis, Department of
Civil and Environmental Engineering, University of California, Berkeley, CA.
[7] BASU U., CHOPRA A.K. Perfectly matched layers for transient elastodynamics
of unbounded domains, 2004, International Journal for Numerical Methods in
Engineering, 59(8), 1039–1074.
[8] LØKKE A, CHOPRA A.K. A direct finite element method for nonlinear earth-
quake analysis of two-and three-dimensional semi-unbounded dam-water-
foundation rock systems, Report No. PEER 2019/02, Pacific Earthquake
Engineering Research Center, University of California, Berkeley, CA.
[9] LØKKE A, CHOPRA A.K. Direct finite element method for nonlinear earth-
quake analysis of concrete dams—Simplication, modeling, and practi-
cal application, Earthquake Engineering and Structural Dynamics, 48(7):
818–842.
[10] NIELSEN A.H. Absorbing boundary conditions for seismic analysis in

ABAQUS, 2006, Proceedings, 2006 ABAQUS Users’ Conference, Boston,
MA.
[11] BAKER J.W. Condition mean spectrum: Tool for ground-motion selection,
2011, ASCE, Journal of Structural Engineering, 137(3), https://doi.org/10.1061/
(ASCE)ST.1943-541X.0000215.
[12] KWONG N.S., CHOPRA A.K. A generalized conditional mean spectrum and
its application for intensity-based assessments of seismic demands, 2017,
Earthquake Spectra, 33(1) 123–143.
[13] KWONG N.S., CHOPRA A.K. Determining bidirectional ground motions for
nonlinear response history analysis of buildings at far-field sites, 2018,
Earthquake Spectra, 34(4) 1931–1954.
[14] KWONG N.S., CHOPRA A.K. Selecting, scaling, and orienting three com-
ponents of ground motions for intensity-based assessments at far-field sites,
2020, Earthquake Spectra, https://doi.org/10.1177/8755293019899954.
157
Q.104 - R.12
-------
-------
SECURING THE ORLÍK DAM AGAINST THE IMPACTS OF EXTREME

FLOODS (∗)
Ondřej HRAZDIRA
Project manager, Povodí Vltavy, STATE ENTERPRISE,
Jiří ŠVANCARA
Director, Department of Hydro Engineering and Hydropower, AQUATIS a. s.,
Czech Republic
CZECH REPUBLIC
SUMMARY
The Czech Republic was hit by several of extreme floods during the past
decades. The Orlík Dam on the Vltava river, the largest dam by reservoir volume not
only on the Vltava cascade but also in the Czech Republic, lies 91 km upstream of
the capital Prague. The flood which hit the Orlík Dam in September 2002 exceeded
the design flood. The Orlík Scheme was built in the 1950s and since then has been
reliably fulfilling its purposes. The scheme serves as a hydroelectric power plant,
ensures a minimal water flow in the Vltava river, contributes to the floods prevention
and is extensively used and is used for recreation as well. Although, like other major
Czech dam structures, the Orlík Dam withstood the extreme floods, new, much more
stringent regulations governing safety assessment of dam structures during floods
were adopted by the Czech authorities based on assessments of the situation at the
various dams. A safety improvement programme was then initiated on the major-
ity of Czech dams. This programme has been based on the requirement that the
dams should safely manage a reference flood determined by taking into account all
the potential consequences of a dam failure. A 10,000-year flood is taken as the
check flood for the Orlík Dam. The comprehensive preparatory work, computing
analyses and model experiments were performed, various technical concepts were
∗ Sécurisation le barrage d’Orlík contre les effets des crues
158 DOI: 10.1201/9781003211808-12

Q.104 - R.12
evaluated and the optimum concept was selected. The Orlík Dam refurbishment
project aimed at improving dam safety was prepared. It is based on the construction
of a supplementary high-capacity spillway and chute which, along with the exist-
ing spillways and outlets will make for a safe transfer of extreme floods. This paper
contains information on the development of the concept and preparatory works to
start up the dam refurbishment project. The dam will remain operable during the
building period, which puts very high demands on its safety. New structures will be
located at the right-hand side bank. Local conditions put very high requirements on
the contractor regarding construction site equipment and work organisation so the
schedule should be adhered to. The massive reinforced concrete structures are a
challenge, with high demands on their resistance and shape accuracy.
RÉSUMÉ
De nombreuses crues extrêmes ont touché la République tchèque ces

dernières décennies. L’ouvrage hydraulique d’Orlík sur la rivière Vltava se trouve
à 91 km en amont de la capitale de la République tchèque, Prague, et est le plus
grand barrage de la cascade de la Vltava et de Tchéquie. Lors des inondations
d’août 2002, l’ouvrage hydraulique d’Orlík a été touché par une crue plus impor-
tante que celle pour laquelle il avait été conçu. Le barrage a été édifié dans les
années cinquante du siècle dernier et, depuis lors, a rempli de manière fiable une
fonction hydroénergétique, assurant un débit minimal dans la Vltava, prenant part
à la protection contre les inondations et ayant une vaste utilisation récréative. Bien
que l’ouvrage hydraulique d’Orlík et les autres grands barrages aient résisté pendant
les crues extrêmes, de nouveaux règlements bien plus stricts pour l’évaluation de la
sécurité des ouvrages hydrauliques pendant les inondations ont été mis en place en
République tchèque sur la base d’une évaluation de la situation sur chaque ouvrage
hydraulique. Un programme de mesures d’amélioration de la sécurité, demandant
en substance que les barrages passent en toute sécurité la “ crue de contrôle ”
déterminée avec une appréciation des conséquences potentielles d’une avarie de
l’ouvrage hydraulique, a été ensuite lancé sur une grande partie des barrages en
République tchèque. La crue de contrôle est dans le cas de l’ouvrage une crue
décamillénale. Des diverses conceptions techniques ont été vérifiées sur la base
de travaux préparatoires complexes, d’analyses des calculs réalisés et d’études
modélisées et la reconstruction de l’ouvrage hydraulique d’Orlík a été préparée
après leur évaluation, afin d’améliorer sa sécurité. Sa nature est d’édifier un nou-
veau déversoir complémentaire à haute capacité et un couloir, assurant ensemble,
avec les ouvrages fonctionnels existants, un passage sécurisé des crues extrêmes.
Cet article apporte des informations sur l’évolution de la conception et du processus
de préparation de la reconstruction de l’ouvrage hydraulique d’Orlík. Au moment de
la naissance de l’article, l’ouvrage est avant son commencement. La construction
sera réalisée pendant que l’ouvrage hydraulique est en marche, avec des deman-
des élevées concernant sa sécurité. Les structures proposées sont placées dans
la pente de la rive droite, les conditions locales imposent au maître d’œuvre des
159
Q.104 - R.12
demandes exceptionnelles d’équipement du chantier, d’organisation du travail et de

respect du planning. La réalisation de structures massives en béton armé, avec de
hautes exigences concernant leur résistance et la précision dimensionnelle, est un
grand enjeu.
1. INTRODUCTION
The Orlík Dam is a part of the dam cascade on the Vltava River consisting of
9 dams in total. The Orlík Dam lies 91 km upstream of the capital Prague. The dam
is the largest by reservoir volume among the dams not only of the Vltava cascade
but also in the Czech Republic countrywide. Preparatory construction works were
commenced in autumn 1954 and the scheme was put into operation in 1962. The
Orlík Dam is a state-owned facility which is administered by the Povodí Vltavy [Vltava
River Basin] state enterprise.
During the past decades, the Czech Republic was hit by extreme floods, larger
than any flood ever recorded before. Although the most important dams withstood
the floods, the situation at each dam was evaluated and new, much more stringent
regulations applicable to dam safety assessment during floods have been adopted.
A safety improvement programme has been then initiated on the majority of Czech
dams. This programme is based on the requirement that the dams should safely
Fig. 1
Location of the Orlík Dam on the Vltava River
Situation le barrage d’Orlík sur la rivière Vltava.
160
Q.104 - R.12
manage a check flood determined by taking into account all the potential conse-
quences of a dam failure. In many cases, application of the new safety standards
gave rise to the requirement that the dams should manage much larger flow rates
than those for which they were initially designed.
This paper describes preparatory works for refurbishment the Orlík Dam (the
largest dam in the Czech Republic by reservoir volume) to increase its safety dur-
ing extreme floods. This article describes sets of analyses, technical concepts
considered and the current status of the project.
2. INFORMATION ON THE ORLÍK DAM
2.1. HISTORY OF THE ORLÍK DAM
The Orlík Dam forms a straight gravity concrete structure 450 m along the
crest with and height of over 80 m. The total volume of concrete embedded in the
dam is over 1 million cubic metres. The dam crest is used as a public road.
Water is let out through a crest by gated spillway consisting of 3 spans. Two
bottom outlets are located in the block under the spillway. A 364 MW hydroelectric
power plant equipped with 4 turbine sets is located downstream the dam, at the
left-hand bank. The Orlík peak hydroelectric power plant works in cooperation with
the Kamýk Dam located downstream of the Orlík, acting as compensating reservoir.
A system for the transfer of small sporting vessels consisting of a platform lift is
located at the right-hand bank. A system for vessels with shipload up to 300 tonnes
is designed as an inclined ship-lift. Only its building part has been completed so far
and its equipment with the technology is only underway.
Preparatory construction works started in 1954 and the dam construction pro-
cess itself was commenced in April 1957. The reservoir started to be filled in 1960
and the fourth, last power plant unit was put in operation in March 1962. Certain
land reclamation works lasted till 1966.
The construction of the dam gave rise to a reservoir 68 km long on the Vltava
River and affecting its tributaries – the Otava and Lužnice rivers. The reservoir area
is 27.3 km2 and its volume is 716.5 million m3 .
Currently, the Orlík Dam mainly serves for water accumulation to improve the
water flow on the Vltava and Labe downstream of it. They also partly protect the
downstream land against floods and serve generation of electric power. Those main
purposes are completed with uses such as recreation, water sports, fishery and
navigation in the reservoir, which are all important because of the vicinity of the
capital Prague.
161
Q.104 - R.12
Fig. 2
The Orlík Dam – view of the dam
Le barrage d’Orlík – vue sur la digue.
2.2. BASIC PARAMETERS OF THE ORLÍK DAM:
Water course: the Vltava River, river basin area: 12,106 km2
Mean long-term annual discharge: Qa : 83.5 m3 /s, discharge Q100 : 2,180 m3 /s
Dam: concrete gravity dam, straight, dam crest length: 450 m

Altitude of the road on the dam crest: 361.10 a.s.l.
Altitude of the spillway edge: 345.60 m a.s.l. (ref. 0.00 m)
Maximum reservoir water level: 351.20 m a.s.l. (+5,60 m)
Maximum retention level: 353.60 m a.s.l. (+8,00 m)
Altitude of the service dam crest: 354.60 m a.s.l. (+9,00 m)
Dam height above the valley bottom: 81.50 m
Two bottom outlets: maximum capacity of 371 m3 /s each. Diameter 4,000 mm,
control valve (Johnson) downstream, an upstream vertical slide gate valve.
Emergency spillways: 3 spans, radial gates, individual span width 15 m,

maximum total discharge: 2,183 m3 /s.
Power plant: Number of units: 4, installed power output: 364 MW, maximum
capacity: 4 × 150 m3 /s, operation range: 45.0 to 71.5 m.
2.3. THE ORLÍK DAM AND THE FLOOD IN AUGUST 2002
The August 2002 flood exceeded the design flood at the Orlík Dam. The water
level in the reservoir exceeded 355.17 m a.s.l. (+9.57 m above the spillway crest),
162
Q.104 - R.12
Fig. 3
The Orlík Dam during the August 2002 flood. Outflows from the venting openings
after the interior dam spaces had been filled with water.
Le barrage d’Orlík lors des inondations d’août 2002. Dégorgements des orifices
d’aération après remplissage des zones intérieures de la digue.
the discharge from the reservoir reached 3100 m3 /s. During the flood onset, the dam
met perfectly its purpose, owing to its accumulative capacity the scheme provided
time enough for securing and evacuation activities in the capital Prague and other
towns and villages downstream.
2.4. LESSON LEARNED FROM THE FLOODS AT THE ORLÍK DAM
The August 2002 flood was the largest event of this type during the dam history.
The reservoir inflow appreciably exceeded the 100-year flood value. The water level
in the reservoir exceeded the design level. As a result, the interior dam spaces were
flooded, the water draining system was thereby affected, with an adverse effect on
the system stability. Data for assessment of the dam behaviour were obtained from
the measuring system. The Orlík Dam withstood the flood reliably but the scheme
suffered considerable damage – this mainly concerns the substructures adjacent to
the dam. All damages were repaired during the years that followed. As part of the
remedial action, provisions against water ingress were made on the upstream side,
by using mobile flood barriers 1.1 m above the level of the service dam crest as far
as 355,70 m a.s.l., that is, higher than as matched the situation in 2002.
A programme of analyses and preparatory work for the future refurbishment

of the Orlík Dam was initiated based on the assessment of the 2002 flood event.
This programme was meant to comprehensively address the inadequacy of spillway
and encompassed structural and stability analyses of the dam including the bedrock
taking into account any potential load conditions of the dam.
163
Q.104 - R.12
3. CONCEPT OF THE ORLÍK DAM REFURBISHMENT PROJECT
3.1. CONCEPT DEVELOPMENT
As regards safety supervision in compliance with Czech legislation, the Orlík

Dam is among the dams with most strictly demands (Category I). New Czech
legislation requires that similar dam should withstand a 10,000-year check flood.
Considerable attention was paid to hydrological parameter analysis ([01] and [02]).
Subsequent analyses were made concering to a 10,000-year flood with a peak
discharge counting of 5,300 m3 /s and flood volume of 1,900 million m3 .
The safety of the Orlík Dam during floods was assessed in 2005 [03]. It was
found that even if the provisional remedial action to prevent water ingress into the
interior dam spaces made in 2002 is included, the Orlík Dam does not comply
with current applicable regulations or national technical standards. The initial Orlík
Dam stability data were reassessed in 2008 [04]. The stability of the dam blocks
during load matching the check flood was found to be poorer than as required by
the applicable standard.
A physical model of the dam body and spillways was set up in 2009 [05] with a
view to assessing the parameters of the discharge and extrapolating them to water
levels matching extreme floods, specifically a 10,000-year flood.
A study analysing existing options to improve the safety of the dam during
extreme floods was developed in 2010 [06]. This study included a set of variants to
increase the discharge capacity and safe passage of a 10,000-year flood wave. The
following variants were analysed:
– A new gated safety spillway with the same spillway crest altitude
– A new gated safety spillway with the same spillway span width
– Three new gated spillway spans
– Lowering the sills of the current spillways
– A new ungated safety spillway
– A new bypass tunnel
– New central outlets
– A new gated emergency spillway in combination with tunnels
– New central outlets in combination with tunnels
During its development, the study [06] was augmented with analysis of the
following provisions to improve stability:
– New injection screen from the injection tunnel

– New injection screen from a newly driven tunnel under the dam base
– Dam block anchoring to the bedrock from the inspection and injection galleries.
The archived design documents were digitised in 2012 [07] for use as a basis
for the subsequent preparatory work.
164
Q.104 - R.12
Comprehensive calculations using the finite element method (FEM) were

made during 2013–2015 [08] based on preceding engineering geological survey of
the bedrock and, in particular, of the contact between the dam body and the bedrock
in the dam base. The load conditions matching the variants in the feasibility study
that was being developed in parallel were stepwise refined. The structural analyses
led to key recommendations regarding the highest safe water level, requirements
for the performance of the draining system and suitability of the measures being
considered.
The selected prospective variants were detailed in the feasibility study devel-
oped in 2013 [09]. The variants were evaluated with respect to technical feasibility,
costs, operational aspects and safety aspects as well as to the impacts on the
scheme during the construction phase. Based on the dam operator’s requirement,
the study analysed the following selected proposals for discharge capacity increase
and stability improvement:
– New safety spillway spans in the dam body

– New safety spillway spans beyond the dam body, in the gated and ungated
spillway variants and also in the variant assuming a discharge through a driven
tunnel or chute on the surface
– Dam stability improvement via an injection screen carried out from a new
injection tunnel
– Dam stability improvement via a new injection screen constructed from existing
injection galleries
– Dam stability improvement by anchoring the dam blocks into the bedrock
3.2. SELECTION OF THE OPTIMUM VARIANT
Evaluation of the feasibility and other key aspects enabled the customer to
narrow the specification by including the optimum variant. Also, taking into account
the structural analysis results, the admissible water levels in the reservoir were
refined, as were the requirements for the capacities of the spillways based on the
development of the requirements for flood protection of the areas downstream of the
dam (in Prague in particular). For the ensuing preparation, it was required that the
water-management dam crest level during the reference 10,000-year flood should
not exceed the altitude of 354.60 m a.s.l. and the administratively determined highest
water level in the reservoir during a 1000-year flood should not exceed the altitude
of 353.60 m a.s.l.
The concepts of the 2 variants selected (see below) were refined in the ensuing
study of 2014 [10]:
– New gated emergency spillways in the dam body

– New gated emergency spillways with a chute beyond the dam body
165
Q.104 - R.12
Fig. 4
Current state and a variant with the new spillways in the dam body.
Etat initial et variante avec nouveaux déversoirs dans la digue.
3.2.1. New gated safety spillways in the dam body
This variant represented the construction of 3 additional gates of a type similar

to the existing ones. The dam blocks No. 26, 27 and 28 would be used for the new
spillways (i.e., located on the right of the original ones). The spillway sill altitude
would be 345.60 m a.s.l., gated profile 3 × 11 × 8.5 m (gated height the same as
the original ones). The spillway capacity would be 3 × 607ṁ3 /s (at a water level
of 354.60 m a.s.l. to manage a 10,000-year flood wave) and 3 × 504 m3 /s at the
administratively determined maximum level to manage a 1000-year flood wave.
The water would be led away from the gates through a newly built cascade
chute located between the path of the small boat lift and would end from the right
into the stilling pool.
Application of 3D models (CFM) resulted in stepwise modifications of the cas-

cade chute shape. The model identified complex flow conditions mainly consisting
in the stream directing onto the dividing wall between the water-management stilling
pool and the power plant stilling pool and stated that the effect on the performance
of the new and old facilities must be addressed. This variant implies a considerable
intervention into the dam structure and significant impacts on the dam operation
during the construction period, including closing the road on the dam crest.
3.2.2. New gated safety spillways with a chute beyond the dam
This variant represented the construction of 2 new spillways and chute located
at the right-hand side bank, occupying the current service areas. The initial design
included spillways with radial gates, sill at an altitude of 346,60 m a.s.l., the gated
profile of each span 19 m × 7.5 m. and spillway capacity 2 × 883 m3 /s (at a water
level altitude 354.60 m a.s.l. to manage a 10,000-year flood wave) and 2 × 705 m3 /s
at the administratively determined maximum water level (to manage a 1000-year
flood wave).
166
Q.104 - R.12
Fig. 5
Current state and variant with the new spillways with a chute beyond in the dam.
Etat initial et variante avec nouveaux déversoirs hors de la digue avec couloir.
A closed frame chute passing through the dam profile would be built to dis-
charge the flow away from the gates. The chute would be in follow-up part opened
in shape as a rectangular reinforced concrete trough, leading the flow downstream
behind the current stilling pool.
Application of the CFM model to assess and optimise the flow plus consider-
ations regarding the operational conditions resulted in the decision that the initial 2
gates would be replaced by 3 gates, which enables better shape the structure to the
local morphology. It also contributes to improving the flow condition downstream of
the spillway while adhering to sufficient capacity.
This variant, too, is associated with the issue of construction site protection
during the spillway construction phase and necessity to close the road on the dam
crest. The intervention, though, would be less pronounced than in the previous
variant.
3.3. FINAL CONCEPTION
The study [10] gave evidence that both variants considered were feasible and
neither was appreciably financially more advantageous than the other. At this stage,
the investor contacted a group of experts in the dam’s engineering and concrete
structures. In November 2015 a significant majority of the experts recommended
the variant with the new supplementary gated safety spillways beyond the dam with
a chute, and so this variant was selected as the final conception of the project.
Immediately after this decision was taken, the requisite starting procedures
were contracted. This included, in particular, detailed engineering-geological sur-
veys, geodesic surveys and inspections of the concrete structures. The most
important refinement was the design of 3 safety spillway bays, which enable the
167
Q.104 - R.12
spillway structure to be suitably sited in the confined space that is available between
the pair of boat lifts and the dam management building. The design was refined by
3D modelling (CFM) and subsequently verified and optimised by using the physical
model [11]. The physical model survey as completed in March 2017.
4. DESIGN PREPARATION
4.1. DOCUMENTATION FOR THE LAND-USE PLANNING PROCEDURE
Documentation for the Land-Use Planning Procedure (DLUPP), which was

completed in 2016 [12], encompassed all the existing supporting documents for
adapting the design to the specific conditions, specifically to the existing geological
conditions and technical state of the current structures. The documentation sta-
bilised the content of the design, including the main building structures (spillway,
chute), associated building structures (rebuilding of the bridge on the crest), and
other related partial parts.
Within the land-use planning procedure and based on the DLUPP, require-
ments were identified for the next stages of the geological survey, geodesic survey,
structural survey and other survey activities, now precisely targeting the specific
structures in the existing dam body, geology of the bank, etc. The conditions and
comments of the entities impacted by the construction were resolved within the
DLUPP. The land-use planning decision has been valid since September 2016.
Fig. 6
Embedding the new chute into the right bank area according to the DLUPP.
Intégration du nouveau couloir dans la zone de la rive droite selon les documents
d’autorisation d’implantation.
168
Q.104 - R.12
Fig. 7
Visualisation of the new supplementary spillway and chute at the right-side bank
according to the tender documentation.
Visualisation du nouveau déversoir complémentaire et du couloir dans la zone de
la rive droite sur la base du dossier d’appel d’offres.
4.2. DOCUMENTATION FOR THE BUILDING PERMIT
Documentation for the building permit [13] was completed in June 2018. It
details the technical design of the building structures and technological building
units. The building permit became effective in September 2018.
4.3. TENDER DOCUMENTATION
The tender documentation (TD) [14] was developed in 2019 as detail design
required for performing the construction works. The TD specifies the design solution
and detailed specifications for the related sub-parts together with requirements for
the construction process. The TD also contains lists of works and procedures. TD
specifies requirements for the organisation of the construction process and includes
a time schedule.
4.4. FINAL PARAMETERS OF THE MAIN STRUCTURES
The dam will be secured against the impacts of extreme floods by building a
supplementary high-capacity spillway and chute. The main structures – the spillway
and chute – are located at the right-side bank following the outputs of the prepara-
tory works and with the existing permits. The new 3-bay safety spillway is gated
169
Q.104 - R.12
Fig. 8
Final solution – visualisation.
Conception finale – visualisation.
with segment gates. The gated profile is 3 × 13.3 × 8.15 m. The fixed sill altitude is
346.45 m a.s.l. The throughput of the new spillway bays at the reference water level
altitude of 354.60 m a.s.l. is 3 × 588 m3 /s.
The entrance gated section is followed by the covered part of the chute at a lon-
gitudinal slope of approximately 5% – three reinforced concrete frames 13 m × 9 m,
with the lengths 53 m, 66 m and 72 m respectively. The adjacent open chute is
235 m long, 16 m wide and 6 m deep and consists of a reinforced concrete trough
at a longitudinal slope up to 40%.
4.5. SPECIFIC ISSUES OF THE CONSTRUCTION PROCESS
The construction process will be managed by the investor based on the

BIM methodology. This intent was taken into consideration by the designer when
developing the detail design.
The Orlík Dam must be in operation during the rebuilding works. The time
schedule identifies periods and limits of feasible water level decrease in the reservoir.
For safety reasons, the dam will be under more stringent supervision during the
construction period. Baseline measurements to obtain reference data of the affected
structures were made before commencing the construction works.
The construction site is located on a steep slope and puts extreme demands
on the contractor as regards site equipment and adherence to the time schedule.
The construction of the massive reinforced concrete structures over 200 m3 vol-
ume in each work step is a challenge. The total structural concrete volume exceeds
50 thousand m3 . A research project [15] focussing on the concrete mix technology,
concreting process, hydration heat evolution and volume changes was implemented
170
Q.104 - R.12
Fig. 9
Demonstration 3D (BIM) model of the structures of the spillway and chute.
Exemple de modèle spatial des structures du déversoir et du couloir.
to clarify the conditions of the feasibility of construction of the massive concrete struc-
tures. The requirements for the concrete quality and shape accuracy are stringent
concerning the extreme flow velocities in the chute.
5. CURRENT PROJECT STATUS
The tender procedure to select the construction contractor is currently (i.e.

at the time this paper is being prepared) underway. The construction works are
planned to start in mid-2021. Documents for the construction project management
and control are being prepared and a team of supporting experts is being assembled.
Funds from the Agriculture Ministry’s programme ‘Flood prevention IV’ will be used
by the investor as a contribution to partly fund the project.
6. CONCLUSIONS
The refurbishing project for the largest dam by the reservoir volume in the
Czech Republic is currently prepared for commencement. The preparation of the
requisite documents, the licensing procedures and the development of the design
documentation were time-consuming and the requirements for a perfect and thor-
ough preparation matched the specific situation and the special position of the Orlík
171
Q.104 - R.12
dam within the existing dams in the Czech Republic. The project investor (Povodí
Vltavy, state enterprise); the designer company (AQUATIS a.s., formerly Pöyry Envi-
ronment a.s.); the organisation responsible for technical safety and supervision
(Vodní díla TBD a.s.); research scientists and technologists, experts from techni-
cal universities and other experts – they are all confident that this unique project will
be completed within the next 4 years.
REFERENCES
[1] Hydrologická studie pro VD Orlík – Průběhy teoretických povodňových

vln + rozšíření studie2005 [Hydrological study for the Orlík Dam – Pat-
terns of the theoretical flood waves + extension of the 2005 study] Czech
Hydrometeorological Institute, 2005
[2] Posúdenie hydrologických veličín vodného diela Orlík – Záverečná správa

Slovenská technická universita, Stavebná fakulta, 2008 [Assessment of the
hydrological parameters of the Orlík Dam – final report. Slovak University of
Technology, Faculty of Civil Engineering, 2008].
[3] VD Orlík Posouzení bezpečnosti VD při povodních [Orlík Dam. Assessment

of safety during floods] Vodní Díla TBD a.s. 2005.
[4] Kontrolní přešetření stability vodního díla Orlík [Control re-evaluation of the
Orlík Dam stability]. Czech Technical University in Prague, 2008.
[5] Hydrotechnický výzkum VD Orlík [Hydrotechnical research of the Orlík Dam].

Czech Technical University in Prague, 2009.
[6] Vodní dílo Orlík. Zabezpečení VD před účinky velkých vod. Studie proveditel-
nosti [The Orlík Dam. Securing the structure against the impacts of extreme
floods. Feasibility study]. Czech Technical University in Prague, 2010, exten-
sion of the study 2012.
[7] Vodní dílo Orlík – digitalizace archivní výkresové documentation [The Orlík
Dam – digitisation of the archived drawings]. Pöyry Environment a.s., 2012.
[8] VD Orlík – Posouzení stability. Postupně etapy 1 až 4. [The Orlík Dam – stability
assessment. Stages 1 to 4]. Pöyry Environment a.s., 2013–2015.
[9] VD Orlík – zabezpečení VD před účinky velkých vod, Rozpracování vybraných

variant [The Orlík Dam. Securing the structure against the impacts of exteme
floods. Detailing of the selected variants], Pöyry Environment a.s., 2013.
[10] VD Orlík – zabezpečení VD před účinky velkých vod. Studie proveditelnosti

[The Orlík Dam. Securing the structure against the impacts of exteme floods.
Feasibility study]. Pöyry Environment a.s., 2014.
172
Q.104 - R.12
[11] VD Orlík, Fyzikální model nového bezpečnostního objektu, zabezpečení VD

před účinky velkých vod [The Orlík Dam. Physical model of the new safety
structure, securing the dam against the impacts of extreme floods]. University
of Technology in Brno, Faculty of Civil Engineering, Research Laboratory of
the Department of Water Structures, 2017.
[12] VD Orlík – zabezpečení VD před účinky velkých vod. Dokumentace pro územní
řízení [The Orlík Dam: securing the structure against the impacts of large
floods. Documentation for the land-use planning procedure]. AQUATIS a.s.,
Brno, 2016.
[13] VD Orlík – zabezpečení VD před účinky velkých vod. Dokumentace pro

stavební povolení [The Orlík Dam: securing the structure against the impacts of
large floods. Documentation for the building permission procedure]. AQUATIS
a.s., Brno, 2018.
[14] VD Orlík – zabezpečení VD před účinky velkých vod. Dokumentace pro

provádění stavby [The Orlík Dam: securing the structure against the impacts
of extreme floods. Detail design]. AQUATIS a.s., Brno, 2019.
[15] Individuální návrh technologie železobetonových konstrukcí ve vztahu k ČSN

EN 1992-1-1, VD Orlík – zabezpečení VD před účinky velkých vod [Individual
proposal for technology of reinforced concrete structures in relation to ČSN
EN 1992-1-1. The Orlík Dam: securing the structure against the impacts of
extreme floods]. Klokner’s Institute, Czech Technical University in Prague,
2019.
173
Q.104 - R.13
-------
-------
MWACHE RCC DAM – THE ADVANTAGES OF AN EARLY AND EXTENSIVE

TRIAL MIX PROGRAMME (∗)
M.R.H. DUNSTAN
MALCOLM DUNSTAN & ASSOCIATES, Great Britain
K. TIREITO
MINISTRY OF WATER AND SANITATION, Kenya
J. FUKUWATARI
NIPPON KOEI Co., Ltd., Japan
R.S. RUPRA
MANGAT I.B. PATEL (MIBP) Ltd., Kenya
UNITED KINGDOM
SUMMARY
Mwache dam will be the first RCC dam in Kenya and could be the first in East
Africa. It is to be constructed on the coast of the country in order to supply domestic
water for Mombasa and irrigation for Kwale County.
Early studies of the dam concluded that there were no suitable pozzolans
available in Kenya that could be used in an RCC for Mwache dam and a traditional
concrete dam was recommended. Nevertheless, during the detailed design of the
dam an extensive search was undertaken for suitable cementitious materials and it
was ascertained that there were several sources of natural pozzolan in Kenya and
the neighbouring Tanzania, two low-lime flyashes were being important into Kenya
∗ Barrage en BCR de Mwache – les avantages d’un programme d’essais complets et
précoces de composition
174 DOI: 10.1201/9781003211808-13

Q.104 - R.13
on a regular basis together with a GGBFS (Ground-Granulated Blast-Furnace Slag).

There were thus several potential options.
As the design age for the RCC is 365 days, an extensive Trial Mix Programme
was undertaken prior to the award of the construction Contract. It was soon ascer-
tained that there were no suitable manufactured retarders that could produce the
specified Initial and Final Setting Times for the RCC. In the event, retarders had to
be specifically designed for the particular cementitious materials. This took time that
would not have been available if such an early Trial Mix Programme had not been
undertaken.
It was found that some of the combinations of the potential cementitious mate-
rials were extremely efficient. An RCC containing 32.5 kg/m3 of Portland cement
and a low-lime flyash content of 187.5 kg/m3 and a further RCC containing a Port-
land cement content of 50 kg/m3 and a natural pozzolan content of 170 kg/m3 both
conformed to all the requirements of the RCC for Mwache dam.
RÉSUMÉ
Le barrage de Mwache sera le premier barrage du CCR au Kenya et pourrait

être le premier en Afrique de l’Est. Il doit être construit sur la côte du pays afin de
fournir de l’eau domestique pour Mombasa et l’irrigation pour le comté de Kwale.
Les premières études du barrage ont conclu qu’il n’y avait pas de pozzolanes
appropriées disponibles au Kenya qui pouvaient être utilisées dans le BCR du le
barrage de Mwache et qu’un barrage en béton traditionnel était recommandé. Néan-
moins, au cours de la conception détaillée du barrage, une recherche approfondie a
été entreprise pour trouver des matériaux cimentés appropriés et il a été établi qu’il
y avait plusieurs sources de pouzzolane naturelle au Kenya et en Tanzanie voisine,
deux mouches à faible teneur en chaux étaient régulièrement importantes au Kenya
avec un GGBFS (Ground-Granulated Blast-Furnace Slag). Il y aura donc plusieurs
options potentielles.
Comme l’âge de conception du BCR est de 365 jours, un vaste programme

de mélange d’essai a été entrepris avant l’attribution du contrat de construction. On
s’est vite assuré qu’il n’y avait pas de retardeurs fabriqués appropriés qui pouvaient
produire les heures de réglage initiale et finale spécifiées pour le BCR. En l’état,
les retardataires devaient être spécialement conçus pour les matériaux cimentés
particuliers. Cela a pris un temps qui n’aurait pas été disponible si un programme
de mixage d’essai aussi précoce n’avait pas été entrepris.
Il a été constaté que certaines des combinaisons des matériaux cimentés

potentiels étaient extrêmement efficaces. Un BCR contenant 32,5 kg/m3 de ciment
Portland et une teneur en mouches à faible teneur en lime de 187,5 kg/m3 et un autre
175
Q.104 - R.13
BCR contenant une teneur en ciment Portland de 50 kg/m3 et une teneur naturelle
en pozzolane de 170 kg/m3 étaient tous deux conformes à toutes les exigences du
BCR pour le barrage de Mwache.
1. INTRODUCTION
Mwache dam is part of a Multi-Purpose Project in Kwale County, 22 km west

of Mombasa in the Coastal Region of Kenya. The reservoir will have a capacity of
137M m3 and will provide potable water to the coastal region of Kenya and also
irrigation water for the Mombasa and Kwale Counties. The dam is due to start
construction in 2021 (after a delay of a year due to the COVID-19 pandemic). It will
be 84 m high, 525 m long and have an RCC volume of a little over 600 000 m3 ,
which it is planned will be placed in 15 months. It will be the first RCC dam in Kenya
and could be the first in East Africa.
An earlier study of the dam (by a Firm of Consulting Engineers with no RCC
experience), concluded that there were no locally-available pozzolans in Kenya that
could be used in the dam and thus Mwache should be a traditional concrete gravity
dam.
2. STAGES OF THE TRIAL MIX PROGRAMME
The main objective of the Trial Mix Programme for the Mwache RCC dam
was to provide sets of mixture proportions for the various combinations of locally-
available cementitious materials for the Specification of the Bidding Documents for
the construction of the Mwache Multi-Purpose Project. There were various Stages
to the Trial Mix Programme as follows:
1. A search for suitable locally-available cementitious materials;

2. A study of the alternative sources of aggregates;
3. The optimization of the gradation and the shape of the aggregates;
4. The optimization of the workability and water content of the RCCs;
5. The optimization of the dosages of at least two (and ideally three or four)
different retarders for each combination of cementitious materials;
6. The design of a matrix of mixes for testing in the Main Trial Mix Programme;
7. The Main Trial Mix Programme.
The Trial Mix Programme was undertaken at the ARM (Athi River Mining)
Cement Works laboratory at Athi River just outside of Nairobi. The great majority of
the work was undertaken by local staff who were trained in the procedures necessary
for such an extensive Trial Mix Programme.
176
Q.104 - R.13
3. MATERIALS
3.1. AGGREGATES
3.1.1. Sources of aggregate
Two potential sources of aggregate, both sandstones, are available for the
Mwache dam construction: a riverbed quarry some 10 km upstream of the dam site
within the reservoir area and four existing private quarries downstream of the dam
site. It was impossible to obtain aggregates for the Trial Mix Programme from the
upstream riverbed quarry and thus the aggregates were obtained from Beric Mining,
one of the private quarries located some 2.5 km downstream of the dam site. The
aggregates were delivered to the ARM Laboratory in several batches. After delivery,
the aggregates were separated into six sizes (40 to 28 mm, 28 to 20 mm, 20 to
14 mm, 14 to 10 mm, 10 to 6.3 mm, 6.3 to 5 mm and 5 to 0 mm) and the coarse
aggregates were tested for the Flakiness and Elongation Indices to BS812-105 and
the fine aggregate for the void ratio.
3.1.2. Shape, bulk densities and gradations of the aggregates
The shape of the delivered aggregates was rather poor despite the crushing
operations at the quarry utilizing a primary jaw crusher and a secondary horizon-
tal shaft impactor. In order to produce an aggregate for the Trial Mix Programme
that would be as close as possible to the strictly specified aggregates for the dam
construction, the coarse aggregates had to be sorted such that the sum of the Flak-
iness and Elongation Indices are approximately 15%. Removal of the badly-shaped
aggregates was done manually. About 70% of the delivered aggregates had to be
removed before the aggregates conformed to these strict requirements.
Sorting of the coarse aggregates yielded excellent results with the sum of
the Flakiness and Elongation Indices ranging from 13 to 16% and with an average
Loose and Compacted Bulk Densities (ASTM C29) of 1435 and 1575 kg/m3 respec-
tively. Unfortunately, the very poorly-shaped fine aggregate had a Compacted Bulk
Density of 1635 kg/m3 , giving a void ratio of 36.9% against the specified maxi-
mum of 30%. Nothing could be done to improve the shape of the fine aggregate
during the Trial Mix Programme but efforts will have to be made during construc-
tion to achieve the specified 30% voids ratio. Gradations of the individual sizes
of aggregates were undertaken and the aggregates combined into 40- to 20-mm,
20- to 10-mm, 10- to 5-mm and 5- to 0-mm sizes in accordance with the specified
ASTM C33 gradations.
177
Q.104 - R.13
3.1.3. Specific gravities and absorptions of the aggregates
Table 1 shows a comparison of the results of the Specific Gravity and Absorp-
tion tests on the sandstones within the vicinity of the dam. The Specific Gravities
of the sandstone from the downstream quarry used in the Trial Mix Programme
are exceptionally high for a sandstone and the absorptions are correspondingly
low. The potential upstream riverbed quarry also has very reasonable values. Also
shown in the Table are values of tests on typical sandstones from other quarries
near Mombasa.
Table 1
Specific Gravities and absorptions of the potential sandstones from near the
dam site.
DAM AV. SPECIFIC GRAVITIES ABSORPTION
OD SSD (%)
Upstream riverbed quarry 2.55 2.59 1.66
Downstream private quarries 2.63 2.66 1.30
Typical sandstone quarry near Mombasa 2.45 2.51 2.52
3.2. PORTLAND CEMENTS
Two CEM 1 42.5N Portland cements were utilized in the Trial Mix Programme.
These were sourced from two of the leading cement manufacturers in Kenya, from
the Bamburi Works (Bamburi Cement Ltd.) and from the Kaloleni Works (ARM
Cement Ltd.), both located within a radius of 30 km from the Mwache site. Their
suitability for use in the RCC was checked against the requirements of BS EN 197-
1 CEM 1 42.5N. Both cements have relatively high alkali contents of 0.54% for the
Bamburi cement and 0.73% for the Kaloleni cement against the specified maximum
of 0.60%. Nevertheless, the cements met all the other requirements of the British
Standard.
3.3. LOW-LIME FLYASHES
The Bamburi and ARM Cement Companies regularly import flyashes into
Kenya from the Mundra and the Krishnapatnam Thermal Power Stations in India.
The two flyashes were each considered as potential pozzolans for use in the Mwache
RCC. Their chemical and physical properties were checked against the requirements
of BS EN 450. A single source of flyash was specified rather than from any particular
supplier. In addition, the coal, from which the flyash is produced, must come from a
single source and the Thermal Power Plant must have sufficient capacity to supply
the flyash during dam construction. The Mundra and Krishnapatnam Thermal Plants
178
Q.104 - R.13
have capacities of 4620 MW and 1600 MW respectively, and thus the actual output
of the flyash should not be a problem. Nevertheless, bagging and transportation of
the flyash could be a limiting factor that must be considered. Following the results
of the Trial Mix Programme, it was decided to disregard the Krishnapatnam flyash
as a potential material for the RCC since its performance was inferior to that of the
Mundra flyash and there was no cost advantage.
3.4. NATURAL POZZOLANS
Two natural pozzolans were tried in the Trial Mix Programme – Lukenya natural
pozzolan from near Athi River in Kenya and Holili natural pozzolan from Tanzania.
Both are used in the manufacture of Portland Pozzolanic cements in Kenya. Although
the Holili borrow area is nearer to the dam site as compared to the Lukenya borrow
area, it has to cross the Tanzania – Kenya border through a less travelled route
making the costs of transport for both pozzolans to the dam approximately equal.
Both pozzolans will require milling to the specified fineness. Their chemical and
physical properties were tested in accordance to ASTM C618 Class N. The alkali
content of the Lukenya pozzolan is rather high at 7.33%, compared to the specified
maximum of 1.5%, while that of Holili is 0.73%.
3.5. GROUND-GRANULATED BLAST-FURNACE SLAG (GGBFS)
A Ground-Granulated Blast-Furnace Slag (GGBFS) was used in the Trial Mix

Programme. The Granulated Blast-Furnace Slag (GBFS) originated from the JSW
Steel Works at Bellary in Karnataka State in India and this was milled into a GGBFS
at the JSW Cement Works in Andhra Pradesh State, also in India, before shipping to
Kenya. As with the flyashes (see Section 3.3), this GGBFS was regularly imported
into Kenya. The chemical and physical properties of the GGBFS were compared
to the requirements of BS EN 15167-1 (and the earlier BS6699) and were found to
conform. The GGBFS demonstrated extremely high early-age strength development
making it impossible to retard the setting times of the concrete. The GGBFS was
thus disregarded as a potential material for the Mwache RCC.
3.6. RETARDERS
Retarders from BASF (MasterPozzolith 55R, Pozzolith 425, XP107N-2, Poly-

heed 789 and Masteroc HCA 10), Fosroc (Conplast R – ten different variants), Sika
(Plastiment TM25, Plastiretarder KE and Plastiroute 15KE) and Chryso (Tard CE
and Plasti RCC) were tried during the Trial Mix Programme. The aim was to have
two or three retarders, for each set of cementitious materials, available for testing in
the Full-Scale Trial.
179
Q.104 - R.13
4. PRELIMINARY TRIAL MIX PROGRAMME
4.1. OPTIMISATION OF THE GRADATION OF THE AGGREGATES
The different sizes of aggregates were combined to find the maximum density
and thus minimum voids. The large coarse aggregate (40 to 20 mm) was combined
with the medium coarse aggregate (20 to 10 mm) in various proportions and the
Loose Bulk Densities measured. Proportions of 40:60 (large:medium) were chosen
as they gave the maximum density. The two aggregates were then combined in
these ratios to produce a 40- to 10-mm aggregate. The 40- to 10-mm material was
then combined with the small coarse aggregate (10 to 5 mm) and proportions of
50:50 were chosen with a maximum combined Loose Bulk Density of 1575 kg/m3 .
This represents a solid ratio of 0.59 that is quite high and above the average for a
normal coarse aggregate that is usually within the range of 0.53 to 0.56. This high
solids ratio is due to the much-improved shape of the sorted aggregates leading to
excellent packing. The optimum coarse aggregate (20:30:50 (large:medium:small))
was combined with various proportions of the fine aggregate. Proportions of 62:38
(coarse:fine) were chosen with a maximum Compacted Bulk Density of 1920 kg/m3 ,
signifying a solid ratio of 0.74. Despite the poor void ratio of the fine aggregate, a
solids ratio of 0.74 is exceptionally high. The optimum combination of the various
aggregates was therefore found to be 12.4:18.6:31.0:38.0 (large:medium:small:fine
aggregates). However, during the initial mixes for assessing the optimum workability,
the optimum aggregate combination was adjusted to 16:19:25:40 in order to make
the RCC more cohesive.
4.2. OPTIMISATION OF THE WORKABILITY AND WATER CONTENT
The optimum workability of the RCC was determined by adjusting the

water content of an RCC with a ‘standard’ set of mixture proportions containing
80 + 140 kg/m3 (Portland cement + pozzolan i.e. flyash or natural pozzolan) and
50 + 170 kg/m3 for the GGBFS mixes. The optimum workability was determined by
visual inspection and by the ‘ball’ test at which point the Loaded VeBe time was taken
and an appropriate range of Loaded VeBe times set. The optimum range of Loaded
VeBe times was found to be 11 to 15 seconds (slightly higher than for most modern
RCCs for dams probably due to the less vigorous vibration of the VeBe apparatus
in use). Mixes with Loaded VeBe times slightly outside this range could still be used
but on site would require a change to the added water content. The water content
was further optimized using results from the Stage-5 (Retarder Programme) and
Stage-7 (Main Programme) mixes. A relatively low water demand (see Table 3) was
achieved mainly due to the very good shape of the coarse aggregate. This shows
the criticality of having a well-designed aggregate plant producing material with an
excellent shape conforming to the Specification.
180
Q.104 - R.13
4.3. OPTIMISATION OF THE RETARDER DOSAGE
4.3.1. Defined requirements for the setting times
The rationale for the use of retarders in RCCs is to retard the setting times of
the RCC so that when placed in a dam any layer will always be placed on ‘fresh’
RCC that has not set, thereby creating a true bond between the horizontal layers
resulting in a monolithic, impervious and durable structure with no horizontal joints.
It is planned to place one and half layers each day at Mwache with an exposure time
of circa 16 hours. The specified Initial Setting Time (IST) of the RCC of 21±3 hours
will ensure that any new layer of RCC should always be placed on fresh RCC. At the
optimum dosage, the Final Setting Time (FST) was required to be between 30 and
48 hours for an RCC containing a flyash and 36 to 60 hours for an RCC containing a
natural pozzolan because of the different properties of the two pozzolans. In addition,
the difference between the ISTs with dosages of 75% and 125% of the optimum must
be less than 15 hours.
An ideal retarder will reach a maximum IST and FST at a particular dosage
and if the dosage is increased still further, the setting times should decrease. When
the retarder is double dosed, the IST must not exceed 48 hours and the FST must
not exceed 96 hours. What is not required is a retarder that causes the setting times
to be excessive when the admixture is double dosed, something that has happened
at practically all RCC dams. The RCC that contains a retarder should also not bleed.
However, this can only be determined at full scale and thus two or three retarders
should be advanced to the Full-Scale Trial.
4.3.2. Results of the optimisation of the dosage of the retarders
During this Stage of the Trial Mix Programme over 60 mixes were tried using
several varieties of retarders from different manufactures. Initially, little success was
achieved in the optimization of the retarder dosage for the different combination
of cementitious materials despite trying several retarders at various dosages. For
instance, it was difficult to achieve the IST of 21±3 hours. A great majority of the
retarders tried with all the cementitious materials (except GGBFS mixes) demon-
strated instances of false sets. It was also impossible to retard the GGBFS mixes,
possibly due to the very high early-age strengths that were developed by the mixes.
These difficulties might be due to the high alkali contents of the cementitious materi-
als and more tests are necessary to confirm this hypothesis. It is hoped to finalize this
exercise during the on-site Trial Mix Programme and Full-Scale Trial. Nevertheless,
two retarders (each specifically designed for the Mwache cementitious materials)
were found that achieved the correct setting times for each of the chosen sets of
cementitious materials (see Table 3).
181
Q.104 - R.13
5. MAIN TRIAL MIX PROGRAMME
5.1. MIXTURE PROPORTIONS CHOSEN FOR THE MAIN TRIAL PROGRAMME
A total cementitious content of circa 220 kg/m3 was determined to be optimum

for the flyash, natural pozzolan and GGBFS mixes. Different sets of mixture propor-
tions were designed, each with their own water contents as determined during the
optimization of the workability and water contents. The ranges of Portland cement,
pozzolan and water contents tested are shown in Table 2.
Table 2
Mixture proportions chosen for the Main Trial Mix Programme.
TYPE OF POZZOLAN MIXTURE PROPORTIONS OF THE PASTE (KG/M3 )
PORTLAND POZZOLAN WATER
CEMENT
Flyash 30–70 150–190 105–108
Natural pozzolan 50–110 110–170 120–127
GGBFS 20–60 160–200 106–108
5.2. DEVELOPMENT OF STRENGTH WITH AGE OF THE RCCS CONTAINING FLYASH
Figures 1 and 2 show the development of cube compressive strength with age
up to an age of 365 days of the RCC containing various combinations of Mundra
flyash with Bamburi and Kaloleni Portland cements.
In order to achieve the design compressive strength of 20.9 MPa at 365 days,
the Bamburi/Mundra combination of cementitious material will only require a Bam-
buri Portland cement content of circa 35 kg/m3 and a Mundra flyash content of
185 kg/m3 . These results indicate an extremely efficient combination of cementi-
tious materials, certainly amongst the most efficient found to date in any of the 900
or so RCC dams that have been completed or that are under construction around the
World. With a Portland cement as low as 35 kg/m3 , the thermal conditions within the
structure will be almost ideal with commensurately low temperature rise. However,
a small change in the Portland cement content will make a significant difference to
the compressive strength, for instance, increasing the Portland cement content from
30 kg/m3 to 50 kg/m3 will more than double the compressive strength at 365 days.
This will require a very stringent Quality Control otherwise these efficiencies might
drop. It should be noted that these excellent results have been achieved by con-
centrating on producing aggregates with an excellent shape leading to a low water
demand.
The strengths achieved by the RCCs containing the Kaloleni/Mundra cemen-

titious materials are slightly lower than the Bamburi/Mundra combinations but are
182
Q.104 - R.13
Fig. 1 Development of cube Fig. 2 Development of cube

compressive strength with age of the compressive strength with age of
RCCs containing Bamburi Portland the RCCs containing Kaloleni
cement and Mundra flyash. Portland cement and Mundra flyash.
1. Cube compressive strength 1. Cube compressive strength
2. Age (days) 2. Age (days)
3. Design strength @ 365 days = 3. Design strength @ 365 days =
20.9 MPa 20.9 MPa
still very efficient. The optimum mixture proportions are likely to be circa 52.5 kg/m3
of Kaloleni Portland cement and 167.5 kg/m3 of Mundra flyash. Very good Quality
Control will still be necessary because of the very large differences in strength at
the design age for relatively small changes in Portland cement content.
5.3. DEVELOPMENT OF STRENGTH WITH AGE OF THE RCCS CONTAINING NATURAL

POZZOLAN
5.3.1. Bamburi portland cement with holili and lukenya natural pozzolans
up to an age of 365 days of the RCCs containing various combinations of Bamburi
Portland cement with the Holili and Lukenya natural pozzolans.
The Holili natural pozzolan has a relatively ‘flat’ strength development while the
Lukenya natural pozzolan has a more ‘increasing’ strength development. It seems
that the mixes containing the Lukenya natural pozzolan (in spite of theoretically
inferior chemical properties) have rather better long-term strength development than
those containing the Holili natural pozzolan. In order to achieve the design strength
of 20.9 MPa at the design age of 365 day, the optimum Bamburi Portland cement
content is circa 72.5 kg/m3 with a Holili natural pozzolan content of circa 147.5 kg/m3
and 65 kg/m3 with a Lukenya natural pozzolan content of 155 kg/m3 respectively.
The Holili natural pozzolan was eventually dropped as a potential material for the
183
Q.104 - R.13

compressive strength with age of compressive strength with age of
the RCCs containing Bamburi Portland the RCCs containing Bamburi Portland
cement and Holili natural pozzolan. cement and Lukenya natural pozzolan.
2. Age (days) 2. Cube compressive strength
3. Design strength @ 3. Design strength @ 365 days =
365 days = 20.9 MPa 20.9 MPa
Mwache dam construction since it does not seem to have any significant advantages
over the Lukenya natural pozzolan.
5.3.2. Kaloleni portland cement with holili and lukenya natural pozzolans
up to an age of 365 days of the RCCs containing various combinations of Kaloleni
Portland cement with the Holili and Lukenya natural pozzolans.
The strength development for the Kaloleni/Holili combination of cementitious

materials shows the optimum mixture proportions of circa 65 kg/m3 of Kaloleni Port-
land cement and 155 kg/m3 of Holili natural pozzolan are required to achieve the
design strength of 20.9 MPa at the age of 365 days. Unfortunately, it was impossi-
ble to obtain a readily-available suitable retarder for the Kaloleni/Holili cementitious
materials and thus this combination of cementitious material was dropped as a
potential material. The optimum mixture proportions for the Kaloleni Portland cement
and Lukenya natural pozzolan would be a cement content of circa 50 kg/m3 and a
Lukenya natural pozzolan of 170 kg/m3 . It is not believed that there is any RCC
dam in the World using natural pozzolan as a cementitious material having a design
strength of circa 20 MPa at a design age of 365 days that has a Portland cement
content as low as 50 kg/m3 .
184
Q.104 - R.13

the RCCs containing Bamburi Portland the RCCs containing Bamburi Portland
cement and Holili natural pozzolan. cement and Lukenya natural pozzolan.
2. Age (days) 2. Cube compressive strength
365 days = 20.9 MPa 20.9 MPa
5.4. DEVELOPMENT OF STRENGTH WITH AGE OF THE RCCS CONTAINING GGBFS
up to an age of 365 days of the RCC containing various combinations of Bamburi
and Kaloleni Portland cements and the Bellary GGBFS.
The strength developments with both Portland cements are extremely high
and indeed only one mix – that of Kaloleni/Bellary RCC with a Portland cement
content of 20 kg/m3 and a Bellary GGBFS content of 200 kg/m3 has a strength less
than 20 MPa at an age of 365 days. Because of these very high early strengths, it
was found to be impossible to retard these RCCs. Nevertheless, even if it had been
possible to retard the RCC, such a low Portland cement content would be less than a
practical content since it is unlikely to guarantee its even distribution in the concrete
mixer. For these reasons, Bellary GGBFS was dropped as a potential cementitious
material for the Mwache RCC.
5.5. OPTIMUM MIXTURE PROPORTIONS FOR THE VARIOUS COMBINATIONS OF

CEMENTITIOUS MATERIALS
Table 3 summarizes the optimum mixture proportions for each combination of

the cementitious materials that were tested in the Trial Mix Programme together with
a suitable retarder. The combinations that were not included in the Specification are
shown in grey.
185
Q.104 - R.13

the RCCs containing Bamburi Portland the RCCs containing Kaloleni Portland
cement and Bellary GGBFS. cement and Bellary GGBFS.
2. Age (days) 2. Age (days)
365 days = 20.9 MPa 20.9 MPa
Table 3
Optimum mixture proportions and retarders for the various combinations of
cementitious materials.
PORTLAND POZZOLAN MIXTURE PROPORTIONS RETARDER OPTIONS
CEMENT OF PAST (KG/M3 )
PORTLAND POZZOLAN WATER BASF Fosroc
CEMENT RETARDER DOSAGE RETARDER DOSAGE
Bamburi Mundra 35 185 105 Masteroc 1.20% Conplast 0.70%
flyash HCA10 CR1D
Bamburi Lukenya 65 155 125 Polyheed 1.50% Conplast 1.00%
natural 789 CR1C
pozzolan
Kaloleni Lukenya 50 170 127 Polyheed 1.50% Conplast 1.20%
natural 789 CR1C
pozzolan
Kaloleni Mundra 52.5 167.5 108 Masteroc 1.20% Conplast 0.825%
flyash HCA10 CR1D
Bamburi Holili 72.5 147.5 120 Polyheed 1.80% Conplast 1.00%
natural CR1C
pozzolan
Kaloleni Holili natural 65 155 123 – – – –
pozzolan
Bamburi Bellary 17.5 202.5 106 – – – –
GGBFS
Kaloleni Bellary 22.5 197.5 108 – – – –
GGBFS
186
Q.104 - R.13
5.6. DENSITIES OF THE RCC MIXES
The densities of the RCCs containing the various combinations of the cemen-
titious materials together with an estimated in-situ density of the RCC are shown in
Table 4. The design density of the dam is 2400 kg/m3 . The estimated in-situ den-
sities for both the Lukenya mixes are rather close to the design density and it is
possible that if the sandstone from the upstream quarry were to be used, the in-situ
density could be less than that required in the design unless the water demand is
reduced still further.
Table 4
Densities of the RCCs containing the various combinations of
cementitious materials.
TYPE OF CEMENTITIOUS DENSITY (KG/M3 )
POZZOLAN MATERIALS CUBES LOADED ESTIMATED
VeBe IN-SITU
Flyash Bamburi/Mundra 2441 2426 2435
Kaloleni/Mundra 2461 2433 2445
Natural pozzolan Bamburi/Holili 2471 2415 2445
Bamburi/Lukenya 2452 2387 2420
Kaloleni/Holili 2441 2413 2425
Kaloleni/Lukenya 2441 2384 2410
6. CONCLUSIONS
◦ The Mwache Trial Mix Programme concluded that there are four very efficient
sets of cementitious materials that could be suitable for use in the RCC for
Mwache Dam. The mixture proportions are exceptional and amongst the
most efficient, if not actually the most efficient, ever used in an RCC dam.
This is despite earlier studies stating that there are no available pozzolans
for the RCC dam construction in the country;
◦ The aggregates to be used in the dam will need to have a good shape in
accordance to the Specification, otherwise the efficiency of the mixture pro-
portions will drop and the in-situ RCC density might not meet the required
density;
◦ The choice of the quarry will have to be given careful consideration. The
Specific Gravities of the sandstone from the quarry in the reservoir area
upstream of the dam site are slightly lower than those of the sandstones
used in the Trial Mix Programme from the downstream private quarries. This
might result in lower in-situ densities;
187
Q.104 - R.13
◦ It was ascertained that specific retarders had to be designed in order to

obtain the specified Setting Times for the RCCs containing the particular
cementitious materials that are available for use in the RCC at Mwache;
◦ The early and extensive Trial Mix Programme allowed ranges of mixture
proportions and specific retarders to be included in the Bidding Documents
enabling the Contractors to bid on a reasonably-certain basis;
◦ With these mixture proportions, and with an equally carefully optimized con-
struction methodology, it should be possible to place the RCC in Mwache
Dam swiftly, resulting in an excellent quality of the dam structure, time
and cost savings together with the associated benefits of early Project
completion.
188
Q. 104 - R.14
-------
-------
SPEED OF CONSTRUCTION: THE GREATEST ADVANTAGE OF RCC DAMS (∗)
M.R.H. DUNSTAN
Principal, MALCOLM DUNSTAN & ASSOCIATES
UNITED KINGDOM
SUMMARY
There are now approximately 1000 (15-m+ high) RCC dams in the World, that
are either complete or under construction. There is therefore sufficient precedent
to allow the most economic RCC dam to be designed for practically any situation.
Unfortunately, very few (probably less than 5%) of all RCC dams are being con-
structed as rapidly as they could (and should) be. This seems to be primary due to
RCC dams moving away from the essential simplicity of the early RCC dams. To
quote Leonardo da Vinci “Simplicity is the ultimate sophistication”. It is strongly rec-
ommended that RCC dams should revert to the original concept of a simple design
and a simple construction methodology leading to rapid construction.
RÉSUMÉ
Il y a maintenant environ 1000 barrages en BCR au moins 15 m de haut dans

le monde, qui sont soit terminés, soit en construction. Il y a donc suffisamment
de retour d’expérience pour permettre de concevoir un barrage en BCR le plus
économique pour pratiquement n’importe quelle situation. Malheureusement, très
peu (probablement moins de 5%) d’entre eux sont construits aussi rapidement qu’ils
∗ La vitesse de construction – le principal avantage des barrages en BCR
DOI: 10.1201/9781003211808-14 189

Q. 104 - R.14
le pourraient (et devraient) l’être. Cela semble être principalement dû au fait que les
barrages en BCR s’éloignent de la simplicité essentielle des premiers barrages de ce
type. Pour citer Léonard de Vinci, “ la simplicité est la sophistication ultime ”. Il est
fortement recommandé que les barrages en BCR reviennent au concept original
d’une conception simple et d’une méthode de construction simple menant à une
construction rapide.
1. INTRODUCTION
At the end of 2019, there were approximately 1000 RCC dams (greater than
15 m in height) that had either been completed or were under construction. The
last time there was a Question on concrete dams at an ICOLD Congress was in
1985 in Lausanne. At that time there were only ten completed RCC dams. This
is a remarkable growth over a period of some 35 years for a new form of dam
construction.
The original concept for RCC dams was to keep the design and method of
construction simple, because with simplicity would come rapid construction and
with rapid construction would come quality (as the layers are placed more rapidly
and thus there is less need for joint treatment) and economy. Unfortunately, over
the past few decades, some modern RCC dams have become too complicated and
the advantages of the speed of construction has in some cases been lost. Given
the value of time, it is strongly recommended that RCC dams revert to the original
concept of simplicity and rapid construction.
2. RCC DAM DATABASE
A database containing details of all the RCC dams in the World has been
updated each year for more than 30 years[1] . A summary of the database is included
in the annual World Atlas of Hydropower & Dams[2] and a searchable version is
included on the website www.rccdams.com. Details of the 950+ RCC dams (with
heights greater than 15 m) that are presently complete or under construction can be
viewed and downloaded in various formats including a MapView showing the RCC
dams for which the coordinates are known. MapView uses Google Earth and can
cover any area from the whole World down to a single dam. Figure 1 shows most
of the RCC dams in Southwest China and Figure 2 a close-up of Longtan, the RCC
dam with the highest average rate of placement achieved at any RCC dam to date
(see Sections 5.1).
190
Q. 104 - R.14
Fig. 1
MapView of RCC dams in Southwest China.
Fig. 2
MapView of Longtan dam in Guangxi-Zhuang Autonomous Region.
191
Q. 104 - R.14
3. RCC DAMS AROUND THE WORLD
3.1. NUMBER OF RCC DAMS
At the end of 2019 there 936 known (NB during the 2020 annual update of
the database, neither China nor Brazil were able to supply their usual update, due
to the CIVID-19 pandemic, and therefore there are a number of RCC dams that are
missing from the database) RCC dams over 15 m in height (the ICOLD definition of a
large dam) that were either complete (854) or under construction (82). The numbers
of RCC dams of this size completed each year are shown in Figure 3 together with
a five-point moving average. However, a number of the major RCC-dam building
countries (e.g. China and to a lesser extent Turkey) only keep records of dams over
30 m high. Thus, for the majority of this paper, only RCC dams more than 30 m high
will be considered. There were 793 30-m+ high RCC dams at the end of 2019 of
which 717 were complete and 76 were under construction. Of the complete dams,
70 (9.8%) were RCC arch dams and 12 (1.7%) were RCC arch-gravity dams, i.e.
one in nine of all RCC dams were either arch or arch-gravity dams. The number
of 30-m+ RCC dams completed each year are shown in Figure 4 together with a
five-point moving average.
In both Figures 3 and 4 it can be seen that after an ‘experimental’ period

up to the mid 1980s, there was a rapid growth in the number of RCC dams being
constructed until the mid 1990s, after which there was a period of consolidation
before a further period of growth from the turn of the century. A decrease in the
number is shown from circa 2012, but this is probably due to the lack of reporting
of new RCC dams and it is suspected that the growth has flattened out to 35 to 40
Fig. 3 The number of 15-m+ high Fig. 4 The number of 30-m+ high
RCC dams completed each year. RCC dams completed each year
1. Number of RCC dams 1. Number of RCC dams completed
completed by year by year
192
Q. 104 - R.14
for 15-m+ RCC dams and 30 to 35 for 30-m+ RCC dams being constructed each
year.
3.2. LOCATION OF RCC DAMS AROUND THE WORLD
There are only 12 countries in the World that have more than ten RCC dams,
either complete or under construction, at the end of 2019. These countries are listed
in Table 1. It is clear that China has by far the greatest number of RCC dams; indeed,
somewhat over 30% of all RCC dams are in the country. The top three countries,
China, Turkey and Brazil, account for nearly 55% of all the RCC dams in the World.
It is of interest that of the 12 countries in Table 1, eight are the original countries
that first constructed RCC dams, i.e. China, Brazil, Japan, the USA, Mexico, Spain,
South Africa and Australia. The number of RCC dams in the 12 countries and the
Rest of the World are shown in Figure 5.
Table 1
Number of RCC dams in the top 12 RCC dam-building countries
COUNTRY (No ) (%) COUNTRY (No ) (%)
China 248 31.3 Vietnam 26 3.3
Turkey 95 12.0 Mexico 20 2.5
Brazil 88 11.1 Spain 18 2.3
Japan 61 7.7 Iran 15 1.9
USA 34 4.3 South Africa 13 1.6
Morocco 28 3.5 Australia 10 1.3
Fig. 5
Number of RCC dams in various countries around the World.
193
Q. 104 - R.14
Fig. 6
The number of RCC dams on each Continent.
The distribution of RCC dams on each Continent is shown in Figure 6. Just

less than three quarters of all RCC dams have been constructed in Asia and South
America, while the Rest of the World has only constructed just over a quarter of the
total number.
4. RATES OF PLACEMENT OF RCC IN DAMS
4.1. AVERAGE RATES OF PLACEMENT RELATIVE TO THE VOLUME OF RCC PLACED
Figure 7 shows the average rates of placement relative to the volume of

RCC placed for over 600 RCC dams. The ten RCC dams with the highest aver-
age rates of placement are named in the Figure (N.B. hard-fill dams are not named
because the Quality Control required for the placement of the RCC in a hard-fill
dam is not as high as that with an RCC dam and thus it is not possible to make a
direct comparison between the rates of placement). Only seven (less than 1% of
the total number of RCC dams) of the ten had average rates of placement in excess
of 100 000 m3 /month. Also shown in the Figure is a range of the average rates of
placement within which 97.5% of all rates fall. It can be seen that six of the seven
100 000-m3 /month dams fall outside of the range, three being above and three being
below. It is perhaps surprising that Upper Stillwater and Olivenhain, both of which
had volumes of just over 1M m3 , should have similar average rates of placement
as those of Tha Dan and the Gibe III that have volumes that are five or six times
194
Q. 104 - R.14
Fig. 7
Average rates of placement at RCC dams relative to the volume of RCC placed.
greater. Similarly, Beni Haroun, with a volume of 1.6 M m3 , has a similar average
rate of placement to Guanyinyan that has a volume four times greater.
The RCC at GERD (Grand Ethiopian Renaissance Dam), the largest RCC dam
so far constructed with an RCC volume of 10.1M m3 , is expected to be completed
next year (2021) but the average rate of placement is only anticipated to be circa
100 000 m3 /month (initially it had been hoped to average circa 285 000 m3 /month,
which was rather optimistic).
Figure 8 is a detail of Figure 7 and shows the average rates of placement of

those dams having a volume of RCC of less than 2M m3 . It is clear from Figure 8 that
the majority of the average rates of placement fall close to the bottom of the range,
particularly those with volumes of less than 0.5M m3 . Also, perhaps surprisingly,
the smaller hard-fill dams all seem to have rather, or very, poor average rates of
placement. Only 4% of the dams in Figure 8 have average rates of placement falling
within the upper 25% of the range. This is very disappointing and seems to show that
a significant proportion of all RCC dams are not being designed and/or constructed
as efficiently as they could (or should) be.
195
Q. 104 - R.14
Fig. 8
Detail of Figure 7 showing those dams with volumes of RCC of less than 2M m3 .
1. Average rate of placement (m3 /month) 2. Volume of RCC (Mm3 )
4.2. RCC DAMS WITH THE HIGHEST AVERAGE RATES OF PLACEMENT
The ten RCC dams with the highest average rates of placement (together with
three hard-fill dams) are shown in Table 2. The seven dams with average rates of
placement in excess of 100 000 m3 /month can be seen and in order to get into
the top ten an RCC dam has to have an average rate of placement in excess of
85 000 m3 /month. The average rate of placement is defined as the volume of RCC
divided by the time from the first placement of any RCC to the time that the last RCC
(or other concrete) is placed at the top of the dam (excluding any time lost due to
high monsoon rainfall or extremely cold winters).
In Table 3, a comparison has been made between the planned dates for the
placement of the RCC and the dates actually achieved. In one case, Laúca, the
RCC placement was very close to that originally planned and in five cases the RCC
was placed more rapidly than expected (see Section 5). However, in three cases the
RCC took longer, and in two of the three significantly longer, than originally planned.
There may have been a number of reasons for these delays, but significant monies
would been lost due to the delayed completion.
196
Q. 104 - R.14
Table 2
The ten (thirteen) RCC dams having the highest average rates of placement
DAM COUNTRY HEIGHT RCC PLACING MONTHLY MAXIMUM
(M) VOLUME TIME PLACEMENTS (M3 ) DAY
(M3 × 103 ) (MONTHS) AVERAGE PEAK (M3 )
Longtan (St.1) China 192 4623 32.4 142758 400755 18475
Gibe III Ethiopia 246 6200 46.5 133463 251000 18518
U. Stillwater USA 91 1125 9.0 125324 204430 8415
Tha Dan Thailand 95 4900 40.1 122266 201490 13280
Olivenhain USA 97 1070 8.8 121895 224675 12250
Beydag Turkey 96 2350 20.9 112566 165000
Guanyinyan China 159 6473 60.0 107883
Beni Haroun Algeria 118 1690 16.4 102860 175000 9100
Çetin Turkey 165 2400 24.0 100000
Taum Sauk USA 49 2448 25.0 95253 189470 11330
Nam Ngiep I Laos 167 2288 24.3 94156 188110 9141
Laúca Angola 132 2646 30.0 88200
Guangzhao China 201 2420 27.9 86738
Table 3
Planned and actual placement dates for the RCC dams in Table 2
DAM ACTUAL RCC PLACEMENT PLANNED
DATES ORIGINALLY RATE OF DIFFERENCE INSTALLED
PLANNED PLACEMENT IN TIME CAPACITY
START FINISH FINISH DATE (M3 /MONTH) (MONTH) (MW)
Longtan (St.1) 08.10.04 20.06.07 15.11.07 124110 4.9 6300
Gibe III 18.12.11 31.10.15 15.08.14 194216 −14.5 1870
U. Stillwater 29.09.85 10.08.87 30.09.87 105605 1.7 Irrigation
Tha Dan 01.03.01 02.07.04 15.07.03 172104 −11.6 Irrigation
Olivenhain 04.02.02 31.10.02 15.01.03 94336 2.5 Water
Guanyinyan 01.01.11 31.12.15 15.06.15 121087 −6.5 3000
Beni Haroun 18.10.98 07.07.00 31.12.00 75970 5.8 Irrigation
Nam Ngiep I 19.04.16 29.04.18 15.08.18 82068 3.6 290
Laúca 01.09.14 01.03.17 15.03.17 86914 0.5 2070
Guangzhao 01.02.06 31.05.08 31.12.07 105456 −5.0 1040
When considering the planned rates of placement in Table 3, it is of interest

that the rates for the majority of the dams were planned to be in the upper half of
the range shown in Figure 7. Even at Upper Stillwater and Olivenhain, the planned
rates were closer to the range although still above the upper limit. In two cases,
Longtan and Guanyinyan, the planned rates were both significantly below the lower
limit of the range, although by finishing early, the actual average rate of placement
at Longtan was slightly above the lower limit. Unfortunately, Guanyinyan fell even
further outside the range by finishing late (although the dates for the latter in Table 3
are not as clearly defined as the other dams).
197
Q. 104 - R.14
5. DETAILS OF SOME OF THE RCC DAMS WITH HIGH AVERAGE RATES

OF PLACEMENT
Each of the five RCC dams in Table 3 that were finished well ahead of schedule
have been studied in some detail in order to try to understand why the dams were
successful.
5.1. LONGTAN (STAGE 1) (CHINA)
Fig. 9
Longtan in China under construction.
The construction of Longtan was somewhat complicated by having four dif-

ferent Contractors involved. The left-abutment section (behind the intake structure)
was constructed by the Armed Police, whereas the main dam was constructed by
a Joint Venture (78G) of the 7th Construction Bureau, the 8th Construction Bureau
and Gezhouba. The dam was effectively split into four sections and each Contractor
worked relatively independent of the others with Gezhouba supplying the concrete
to the other Contractors.
Initially, the RCC was rather unworkable and GE-RCC was used for the form-
ing of the faces of the dam. However, minor changes were made to the gradations
of the aggregates and the workability was improved during construction. This led to
a more Contractor-friendly RCC and allowed the use of GEVR and the rate of place-
ment improved. The maximum monthly placement was 400 755 m3 of RCC (and
circa 150 000 m3 of CVC). With such a monthly placement it should have been pos-
sible to achieve an average rate of placement of 150 000 to 175 000 m3 /month
198
Q. 104 - R.14
and thus reduce the time for the placement of the RCC by further four to five
months.
Nevertheless, the average rate of placement of the RCC at Longtan is the

highest achieved to date and the RCC was finished a little under five months
ahead of schedule. The Powerhouse was correspondingly accelerated leading to
the generation of approximately $500M worth the power ahead of schedule.
5.2. UPPER STILLWATER (USA)
Fig. 10
Upper Stillwater in the USA under construction.
Upper Stillwater was constructed in the mid-1980s and was the first 1M-m3+
RCC dam. The dam has a TWL of 2495 m ASL and the temperature range on the
site is in excess of 70 deg C, meaning that the RCC could only be placed for some
five to five a half months of the year as during the winter the temperature drops to
−35 C. The RCC was placed over three construction seasons with seven-month
gaps between the seasons.
The dam was designed to be extremely simple to construct with the faces
were formed of slip-formed facing elements containing heavily air-entrained con-
crete – the dam has been in operation for approaching 40 years and there has been
199
Q. 104 - R.14
negligible deterioration of the faces of the dam in the extremely taxing freeze-thaw
environment. The maximum allowable placing temperature was 10 C and during the
summer the temperature could exceed 35 C – there could be a change in tempera-
ture of 30 deg C in 12 hours. The RCC was placed during two overnight 8-hour shifts
(18.00hrs to 02.00hrs and 02.00hrs to 10.00hrs) and the faces were slip-formed dur-
ing the day shift (10.00hrs to 18.00hrs). In spite of being one of the first RCC dams,
the average rate of placement was in excess of 125 000 m3 /month – this was the
highest average rate of placement at any RCC dam for 20 years from the completion
of the RCC in the summer of 1987 to the completion of the RCC at Longtan in the
summer of 2007.
5.3. OLIVENHAIN (USA)
Fig. 11
Olivenhain in the USA under construction.
Olivenhain dam is part of San Diego’s Emergency Storage Project and was
the first RCC dam in California. The Contract for the construction of the dam was
let with a very tight programme; just over 1M m3 of RCC having to be placed in
less than one year. The dam was placed over three valleys, the main central val-
ley, a small valley on the right abutment (that was used for a Full-Scale Trial) and
a large subsidiary valley on the left abutment. In order to speed the rate of place-
ment, a significant number of changes were made to the design of the RCC and the
200
Q. 104 - R.14
proposed construction methodology by the Contractor using a Value-Engineering

approach (with the Owner taking half of the savings and the Contractor the other
half). In the event, the 1.07M m3 of RCC was placed at an average rate of place-
ment of over 120 000 m3 /month and the Contract was completed two and a half
months early. After placing the RCC in the left abutment valley, the main placement
of just over 0.90M m3 was placed in a fraction under six months at an average rate
of placement of just over 150 000 m3 /month. This was in spite of the RCC place-
ment being limited to five-and-a-half-day weeks because of very strict environmental
restrictions.
5.4. BENI HAROUN (ALGERIA)
The initial Contract for the construction of Beni Haroun dam in Algeria was
for a central-core rockfill dam. After a few years, progress had been slow but the
excavation for the spillway on the right abutment and one diversion tunnel had been
completed and a second diversion tunnel half finished. Following a collapse of the
first diversion tunnel, the Contractor left the site and the Contract for the remaining
construction was awarded to another Contractor, who had constructed a number of
RCC dams. The latter proposed, and the Owner accepted, an alternative design
of an RCC dam in order to try to catch up with the overall programme. Due to the
Fig. 12
Beni Haroun in Algeria under construction.
201
Q. 104 - R.14
situation in Algeria at the time, all the cementitious materials (a Portland cement and
a low-lime flyash) had to be imported from Spain. Luckily the site was quite close to
the Mediterranean coast. A large integrated cofferdam (ICD) was constructed within
the body of the RCC dam, in order to speed up the overall programme and to avoid
the construction of a large separate cofferdam. The ICD also allowed the refinement
of the design of the RCC and the construction methodology. The RCC was placed
significantly more rapidly than expected and was finished nearly six months ahead
of the original completion date.
5.5. NAM NGIEP I (LAOS)
Fig. 13
Nam Ngiep I in Laos under construction.
When completed, Nam Ngiep 1 was the highest and largest RCC dam in
Laos. The foundation conditions at the dam site are challenging which required
a shear trench. Although the main Contractor had limited experience of RCC, he
sub-contracted the placement of the RCC to another Contractor that had already
constructed two large RCC dams in Vietnam (see Section 6.2) and who therefore
was well versed with the rapid construction of RCC dams. Nam Ngiep 1 is the only
dam in Table 2 to use the slope-layer method to place the RCC, all the other dams
used horizontal placement of retarded RCC so that fresh RCC could be placed on
fresh RCC.
After Nam Ngiep 1 was completed some three to four months ahead of the
initial programme, some additional grouting was necessary that limited the value of
the early completion.
202
Q. 104 - R.14
6. THE VALUE OF TIME
6.1. GHATGHAR PUMPED-STORAGE SCHEME (INDIA)
Fig. 14 The Lower Ghatghar dam under Fig. 15 The completed Lower
construction (showing the split-level Ghatghar dam (showing the perfectly
construction methodology). dry downstream face).
The Ghatghar Pumped-Storage Scheme contains three RCC dams; two, the
Saddle dam (9 m high) and the Upper dam (15 m high), retaining the Upper reservoir
and the Lower dam (86 m high) retaining the Lower reservoir. Figure 14 shows the
Lower dam under construction; the dam was split into two halves for the majority of
the height of the dam and the placement of the RCC alternated from the left side to
the right side (split-level construction). Figure 15 shows the completed dam and the
perfectly dry downstream face. The dams were the first RCC dams constructed in
India. In spite of being the first RCC dams in the country, at the time of completion in
June 2006, the Lower dam was the ninth fastest RCC completed dam. The dam is
100% RCC and the seepage at the Lower dam is zero, so much so that the galleries
have had to be painted to suppress the dust. The Project was completed well ahead
of schedule and a paper by GOMID, the Owner of the Project stated:
“Due to selection of RCC dam construction in lieu of conventional con-

crete or masonry dams, the Project has benefited by three years of early
commissioning thereby generation of 1350 Million units (MKWH) the cost
of which is 5200 million INR (Indian Rupees). This is equal to about half
the cost of project. This clearly tells the success story of RCC dams and
its role in hydro electric Projects.” [3]
6.2. SON LA AND LAI CHAU (VIETNAM)
The Son La HPP is the largest Hydropower Project in Vietnam and has an
installed capacity of 2400 MW. The dam is 139 m high and has an RCC volume of
2.677M m3 that was placed in 31.5 months (although there was a gap of five months
in the placement due to access requirements for the construction of the powerhouse
and the net time of the RCC placement was 26.5 months). The average rate of
placement (with the gross time of placement) was circa 85 000 m3 /month and the
203
Q. 104 - R.14
Fig. 16 The completed Son La dam. Fig. 17 The completed Lai Chau dam.
dam is still the 11th fastest RCC gravity dam so far completed. The Project was
so successful that all involved with the design and construction were awarded the
Contracts for the construction of Lai Chau dam that is immediately upstream of the
Son La reservoir. Figure 16 shows the completed Son La dam. A press release from
EVN, the Owner of the Son La HPP, stated:
“Giving general assessment on Son La HPP, the Government highly

appreciated all ministries, sectors, localities and entities involving in the
project with considerable efforts in implementing tasks and gaining posi-
tive results. The Son La HPP has been accepted and taken over and put
into operation three years prior to the approved schedule and up to now,
it has operated effectively and safety.
By the end of September 2018, Son La HPP has supplied 64.12 bil-
lion kWh for the national power system, playing an important role in
power supply for the sake of socio-economic development in the country;
the total contributory value to the State budget is 10.6 thousand billion
VND.” [4]
Lai Chau dam is 131 m high and has an RCC volume of 1.884M m3 that was
placed in 25.8 months. The average rate of placement was circa 72 900 m3 /month.
Figure 17 shows the completed Lai Chau dam. In the same Press Release from
EVN, the Owner of the Lai Chau HPP, stated:
“Lai Chau HPP, according to the report, has also completed all construc-
tions of project items. This project has exceeded its approved schedule at
Decree No. 40/2009/QH12 of the National Assembly one year in advance.
By the end of September 2018, the total power output of the plant sup-
plied to the National power system reaches 11.74 billion kWh. Up to
now, the total contributory value to the national budget is more than 2.36
thousand billion VND.” [4]
Between the Son La and Lai Chau HPPs some $600M worth of power was
generated ahead of schedule due to the efficient and rapid placement of the RCC.
7. CONCLUSIONS
◦ The original concept of RCC dams was to provide a simple method of con-
struction with RCC spread horizontally from one abutment to the other.
204
Q. 104 - R.14
This would lead (and has led) to rapid construction. Unfortunately, some
of this simplicity is being lost with some modern RCC dams and the majority
of RCC dams are not being constructed as rapidly as they could be.
◦ An ‘holistic’ view needs to be taken during the design and construction of
RCC dams as every aspect of the design and construction impacts on other
aspects. It is counter-productive to concentrate on any particular aspect
(e.g. aggregate production) and to try to find the minimum-cost solution as it
will almost certainly lead to a more expensive (and probably slower) overall
Project.
◦ Investment spent upfront, on the investigation of the design and construc-
tion of an RCC dam, will always lead to economic benefits. For example,
a very detailed study of the available materials and an extensive Trial Mix
Programme for the Mwache dam in Kenya, led to the development of an
RCC with a Portland cement content of 35 kg/m3 (and flyash content of
185 kg/m3 ), probably the most economic (and most Contractor-friendly) RCC
ever achieved[5] .
ACKNOWLEDGEMENTS
Grateful thanks must be given to all the engineers, who over the past several
years, have supplied data on their RCC dams so that the ICOLD RCC dam database
can be maintained as a reasonably accurate record. These data are accessed by
very many people using the Summary Table in the World Atlas of Hydropower &
Dams and also from the website www.rccdams.com.
REFERENCES
[1] DUNSTAN, M.R.H., Recent developments in roller-compacted concrete dam

construction, Handbook 1989, International Water Power & Dam Construction,
London, September 1989.
[2] DUNSTAN, M.R.H., The World’s RCC dams, World Atlas & Industry Guide,
Hydropower & Dams, London, September 2020.
[3] SHELKE, V.C. and JOSHI, J.R., The success story of Ghatghar RCC dams,
Proceedings of the Sixth International Symposium on Roller-Compacted
Concrete (RCC) Dams, October 2012, Zaragoza, Spain.
[4] EVN, Press Release: Government requests to terminate reports on Son La,
Lai Chau hydropower projects (in Vietnamese), 12 November 2018, Hanoi,
Vietnam.
[5] TIREITO, K. et al, Mwache – The most successful Trial Mix Programme for
an RCC dam ever? Proceedings of the Eighth International Symposium on
Roller-Compacted Concrete (RCC) Dams, November 2019, Kunming, China.
205
Q.104 - R.15
-------
-------
CISAILLEMENT À L’INTERFACE BÉTON-ROCHER DES BARRAGES

VOÛTES : RETOUR D’EXPÉRIENCE DE VOÛTES EXISTANTES (∗)
Emmanuel ROBBE
EDF
Frederic ANDRIAN & Nicolas ULRICH

Artelia
Claire JOUY
Tractebel
FRANCE
RÉSUMÉ
La France dispose d’un parc de barrages-voûtes importants (plus de 80

ouvrages) majoritairement réalisées durant les 2 décennies qui ont suivi la 2nde
guerre mondiale. Ces ouvrages ont été pour la plupart dimensionnés à partir de
méthodes de calcul évaluant uniquement les contraintes en parement. Les récentes
recommandations françaises proposent d’évaluer désormais également la prob-
lématique du cisaillement en pied des consoles dans le cadre de l’analyse du
comportement des voûtes. Une analyse comparative originale des cisaillements
calculés à partir de modélisation numérique sur une dizaine de voutes exploitées
est ainsi proposée dans ce papier. Elle met en évidence des cisaillements non
négligeables aux interfaces consoles-fondation, avec des cohésions mobilisées
pouvant dépasser le MPa ; valeur bien supérieure à la centaine de kPa qui est
habituellement considérée en France pour le calcul de stabilité au glissement
des barrages-poids par exemple. Les comportements mécaniques particuliers de
cinq ouvrages: Laouzas, les Cammazes, Pont du Roi, Gallens et St-Barthélémy,
∗ Shear at the concrete-rock contact of arch dams : feedback on existing dams
206 DOI: 10.1201/9781003211808-15

Q.104 - R.15
analysés par différents bureaux d’ingénierie sont ensuite présentés accompagnés

de solutions de confortement adaptées à chaque problématique.
SUMMARY
There are many arch dams in France (over 80), most of them built over the
two decades following World War II. The design of most of these dams is based on
an assessment of the stresses of the upstream and downstream faces. The 2018
French guidelines for analysing arch dam behaviour suggest assessing the shear
state at the base of the cantilevers. This paper proposes comparing the shear state
by means of numerical analyses of about ten arch dams in operation. The com-
parison highlights quite high shearing at the dam/foundation contact with mobilised
shear stress exceeding 1MPa. Such values are clearly higher than the 100 kPa
usually considered (in France) for calculating the stability of gravity dams for exam-
ple. The specific mechanical behaviour of five dams: Laouzas, Les Cammazes,
Le Pont-du-Roi, Galens and St-Barthélemy, analysed by several major consulting
companies, are subsequently presented together with the appropriate reinforcement
work, if any, adapted to each situation.
1. INTRODUCTION
La France dispose d’un parc de barrages-voûtes importants (plus de 80

ouvrages) majoritairement réalisées durant les 2 décennies qui ont suivi la 2nde
guerre mondiale. Ces ouvrages ont pour la plupart été dimensionnés à partir
de méthodes de calcul simplifiées (faisant l’hypothèse d’arcs indépendants ou
éventuellement en tenant compte d’un ajustement radial avec la console de clé)
avec une vérification se limitant généralement à l’évaluation des contraintes en
parement. En particulier, les méthodes de dimensionnement utilisées à l’époque
n’incluent pas l’analyse de l’état de cisaillement à l’interface béton/ rocher. L’évolution
du comportement de certains ouvrages a conduit les ingénieurs à s’interroger sur la
problématique du transfert des efforts de cisaillement à ces interfaces. Ces réflex-
ions et analyses ont été intégrées aux récentes recommandations françaises pour
l’analyse du comportement des voûtes [1], synthétisés dans le §2. Le cadre pro-
posé par ces recommandations permet dès lors de fournir un référentiel commun
de comparaison entre ouvrages : une analyse comparative originale des cisaille-
ments calculés sur une dizaine de voutes est ainsi proposée dans le §3. Enfin
les comportements particuliers de cinq ouvrages, analysés par différents bureaux
d’ingénierie, font l’objet du §4 : plusieurs solutions de confortement adaptées à
chaque problématique y sont en particulier détaillées.
207
Q.104 - R.15
2. RECOMMANDATIONS CFBR POUR L’ANALYSE DU COMPORTEMENT

DES VOUTES
En 2018, les recommandations [1] publiées par le comité français des bar-
rages et réservoirs (CFBR) proposent d’évaluer le comportement des voûtes au
regard de différents mécanismes de rupture issus de l’incidentologie et acciden-
tologie mondiale. Parmi ces mécanismes, deux cinématiques liées au cisaillement
à l’interface béton-rocher sont identifiées :
• La non-reprise des efforts en pied des plots centraux : correspondant à une

sollicitation des consoles centrales dans le cas de voûtes implantées en vallée
large, sous l’effet du chargement hydrostatique et de températures froides
(Fig. 1).
Fig. 1
Schéma de principe des sollicitations des consoles centrales.
Failure mechanism under the central cantilevers.
• Le glissement le long des appuis : dans le cas d’ouvrage soumis à des

pathologies de gonflement ou en situation thermique estivale (Fig. 2).
Les recommandations CFBR 2018 [1] proposent d’analyser ces mécanismes

en évaluant en particulier la reprise des efforts de cisaillement au contact avec
la fondation par la comparaison des efforts normaux et tangentiels au niveau de
l’interface console – fondation. Dans le cas de la problématique en pied des consoles
centrales, la stabilité de la fondation vis-à-vis des gradients hydrauliques générés
par le desserrement est également analysée en évaluant la propagation éventuelle
des sous-pressions vers l’aval.
Une analyse graduée est proposée, commençant par une approche élastique
linéaire avant d’introduire un joint avec différents types de comportement non-
linéaires (annulation des tractions, glissement de type Mohr-Coulomb). L’approche
élastique linéaire est nécessaire dans un premier temps afin d’identifier les zones
en traction et/ou en cisaillement excessifs. Cela permet également de quantifier ces
contraintes. Elle peut suffire si les contraintes sont jugées acceptables, c’est-à-dire
pour des contraintes de traction inférieures à 2 MPa en pied amont (dans le cadre
208
Q.104 - R.15
Fig. 2
Schéma de principe des sollicitations des plots de rives.
Failure mechanism at the side cantilevers.
de l’hypothèse de Navier de répartition des contraintes) et pour un dépassement

faible de la résistance au cisaillement, permettant le report d’une partie des efforts
sur les zones adjacentes.
Si ce n’est pas le cas, les mécanismes non-linéaires d’annulation des tractions

et/ou d’une loi de type Mohr-Coulomb pour le glissement doivent être utilisés. Le
choix de l’emploi de ces mécanismes est fait en fonction des résultats de la mod-
élisation élastique linéaire. L’acceptabilité du comportement de la voûte est alors
jugée selon plusieurs critères :
• Acceptabilité des adaptations calculées (report des efforts, ampleur des

déplacements irréversibles)
• Evaluation des marges de sécurité (méthode de c-phi réduction, rétro-
analyses)
Les recommandations actuelles [1] ne fixent pas de critère chiffré permettant

de juger de l’acceptabilité du cisaillement en pied de voûte ou au niveau des appuis
dans le cadre d’une approche élastique linéaire. En effet, ce critère n’a pas vraiment
fait l’objet d’une analyse lors de la conception des ouvrages français (seules les
contraintes dans la voûte étaient analysées avec des méthodes simplifiées : calculs
en arcs indépendants, éventuellement avec ajustement radial en clé et plus rarement
avec la méthode trial-load). Par conséquent, il a été décidé lors de la rédaction des
recommandations [1] de ne pas imposer de critère chiffré sur ce point dans l’attente
d’un retour d’expérience suffisant sur le parc d’ouvrages existants. Dans cet esprit,
le chapitre 3 évalue la cohésion mobilisée aux interfaces consoles-fondation d’un
ensemble d’ouvrages existants.
209
Q.104 - R.15
Le desserrement de la zone de contact en pied des consoles centrales peut

se traduire par l’ouverture de discontinuités diffuses en fondation, localisées à
l’interface béton-rocher, ou par des fissures localisées dans le béton, au-dessus
de l’encastrement dans la fondation. Ces situations ont des conséquences bien
différentes en termes de stabilité de l’ouvrage et de moyens d’action. Plusieurs
voûtes présentant ce type de pathologie ont été étudiées et certaines ont fait l’objet
de travaux de confortement. Dans le chapitre 4, les analyses de cinq ouvrages,
menées par trois bureaux d’études distincts, sont présentées.
3. COHÉSION MOBILISÉE ÀUX INTERFACES PLOTS-FONDATION

D’OUVRAGES EXISTANTS
Les recommandations du CFBR [1] proposent d’évaluer les mécanismes de

défaillance liés au cisaillement aux interfaces consoles-fondation en comparant
les efforts normaux et tangentiels à ces horizons et en évaluant un coefficient de
stabilité au glissement par plot. Ce coefficient dépend en particulier des caractéris-
tiques de résistance au cisaillement du contact béton-rocher : angle de frottement et
cohésion.
De telles analyses ont déjà été réalisées sur un certain nombre de barrages-
voûtes et partant du principe que pour la majorité d’entre eux, aucun déplacement
en pied n’a été observé, une évaluation de la résistance au cisaillement mobilisée
est proposée. Afin de comparer le cisaillement entre plusieurs ouvrages de taille et
géométrie différentes, le choix a été fait d’évaluer les cohésions mobilisées aux inter-
faces béton-rocher des consoles pour que les efforts résistants soient supérieurs
aux efforts de cisaillement en tenant compte de la sous-pression. Un angle de frot-
tement de 45˚ est choisi afin de disposer d’un critère de comparaison uniforme bien
que les irrégularités observées sur les fonds de fouilles permettent généralement de
considérer des valeurs plus importantes. Les contacts béton-rocher des plots sont
modélisés par des surfaces lisses. Plusieurs situations de calcul sont considérées
cumulant chargement hydrostatique à la cote normale d’exploitation et chargement
thermique (été ou hiver).
Au total, les résultats obtenus sur 12 barrages-voûtes, dont les hauteurs

sont comprises entre 20 et 150 m, sont présentés dans la suite de ce chapitre.
Ces résultats ont été obtenus à partir de calcul par éléments-finis ou différences-
finies. Les modèles utilisés utilisent une approche non-linéaire capable de relâcher
les contraintes de traction, souvent au droit d’un joint préalablement implanté.
En revanche, quelle que soit l’ouverture, la contrainte de cisaillement est inté-
gralement transmise dans le joint de manière à éviter tout glissement. Une telle
approche permet ainsi d’évaluer la résistance au cisaillement requise tout en tenant
210
Q.104 - R.15
Fig. 3
Cohésions mobilisées sous un barrage-voûte à RN (trait noir) combiné à une
situations estivale (rouge) ou hivernale (bleu) sans ou avec sous-pression (trait
continu ou pointillé).
Cohésion at the dam-foundation interface under normal reservoir load (black), in
summer (red) or winter (blue) with or without uplift (dotted or continuous line).
compte de l’ouverture normale du contact et de la propagation des sous-pressions

associées.
3.1. SOLLICITATIONS TYPIQUES EN CISAILLEMENT DES PLOTS DE

BARRAGES-VOÛTES
Les sollicitations en cisaillement des plots d’un barrage-voûte dépendent

directement de la situation (ou cas de charge) étudiée:
• Les sollicitations maximales des plots centraux ont lieu en hiver (cisaillement
dans la direction amont-aval),
• Les sollicitations maximales des plots latéraux (ou éventuellement des culées)
ont lieu en été (cisaillement dans la direction rive-à-rive).
La Fig. 3 présente pour un ouvrage d’une hauteur d’environ 30 m, la cohésion

mobilisée sous chaque plot en limitant l’angle de frottement à 45◦ . Cette cohésion
a été déterminée sans (trait plein) ou avec (trait pointillé) prise en compte des sous-
pressions (une analyse spécifique est proposée dans le paragraphe suivant).
211
Q.104 - R.15
L’exemple ci-dessus met bien en évidence une sollicitation maximale en

cisaillement des plots centraux en hiver et des plots latéraux en été. Les sollic-
itations obtenues en hiver sur les plots latéraux correspondent à des efforts de
cisaillement orientés vers le centre de la vallée sans signification physique réelle
(ouverture probable des joints de plots).
Sans prise en compte des sous-pressions, les cohésions mobilisées calculées

aux interfaces béton-rocher sont ici d’environ 0,3 MPa sur les plots centraux en hiver
et de 0,2 MPa sur les plots latéraux en été.
La prise en compte des sous-pressions conduit logiquement à une augmenta-

tion des cohésions mobilisées en partie centrale de la voûte (0,4 MPa sur l’exemple
présenté). Cet effet est limité sur les plots d’extrémité compte tenu de la faible
hauteur d’eau au niveau de ces plots. En partie centrale, la prise en compte des
sous-pressions maximise les cohésions mobilisées calculées.
3.2. ENSEMBLE DES RÉSULTATS
La Fig. 4 permet de visualiser l’évolution de la cohésion mobilisée sur les 12

barrages voûtes étudiés d’une rive à l’autre le long du contact béton-rocher. La figure
distingue les situations de calcul avec chargement hydrostatique seul ou combiné
aux chargements thermiques été et hiver.
En partie centrale, des cohésions non négligeables sont calculées, variant

entre 0.1 et 1.6 MPa (pour l’ouvrage de plus grande hauteur). Ces valeurs sont
bien supérieures à celles généralement retenues en France pour le calcul des
barrages-poids, dépassant rarement une centaine de kPa à l’interface béton-rocher
et interrogent sur les mécanismes de transmission des efforts de cisaillement en
pied des consoles. Trois hypothèses peuvent être formulées :
• De telles cohésions sont effectivement mobilisables à l’interface béton-rocher :

des essais de caractérisations expérimentales menées sur des échantil-
lons de différentes tailles mettent en évidence des cohésions de pics entre
0.5 et jusqu’à 2.5 MPa [3][4]. Les irrégularités parfois pluri-décimétriques
observées en fond de fouilles contribuent également vraisemblablement à des
caractéristiques de résistance cisaillement élevée par dilatance.
• L’encastrement des plots centraux dans la fondation peut permettre, lorsque
le rocher est de bonne qualité en pied, de mobiliser la butée rocheuse aval :
une partie des efforts de cisaillement est alors transmise au rocher via cet
encastrement. C’est dans cet esprit qu’a été analysé et renforcé le barrage du
Laouzas présenté dans le §4.1.
• Enfin dans de rares cas, une adaptation (glissement irréversible localisé au
niveau d’une fissure ou réparti sur plusieurs horizons) en pied d’ouvrage est
observée ou fortement suspectée : à l’interface béton-rocher (§4.5 pour le
212
Q.104 - R.15
Fig. 4
Evolution des cohésions mobilisées le long du contact béton-rocher pour un
ensemble d’ouvrage à RN, RN + été, RN + hiver.
Cohésion at the dam-foundation interface for several dams at normal water level
(NWL), NWL+ summer, NWL+ winter.
barrage de Saint Barthelemy) ou dans le béton au-dessus de l’encastrement

(§4.4 pour le barrage de Galens).
La prise en compte du chargement thermique induit une variation de la cohé-

sion mobilisée entre 0 et 0.4 MPa environ en partie centrale. Les variations les plus
213
Q.104 - R.15
Tableau 1
synthèse des résultats
summary of the results
12. BARRAGE 13. HAUTEUR 14. EPAISSEUR 15. LONGUEUR 16. RATIO 17. COHÉSION
(M) À LA BASE EN CRÊTE L/H MOBILISÉE
(M) MAX (M) À RN (MPA)
18. Galens 19. 23 20. 3 21. 109 22. 4.7 23. 0.3
24. Bage 25. 28 26. 7 27. 139 28. 5.0 29. 0.2
30. La Couze 31. 28 32. 6 33. 129 34. 4.6 35. 0.1
36. Saint- 37. 45 38. 6 39. 200 40. 4.4 41. 1.1
Barthelemy
42. Lanoux 43. 45.2 44. 5 45. 176 46. 3.9 47. 0.8
48. La Valette 49. 50 50. 9 51. 163 52. 3.3 53. 0.6
54. Laouzas 55. 52.6 56. 11 57. 295 58. 5.6 59. 0.3
60. Saint-Guérin 61. 68 62. 18 63. 250 64. 3.7 65. 0.7
66. Vouglans 67. 130 68. 22 69. 427 70. 3.3 71. 1.4
72. Roselend 73. 150 74. 20 75. 285 76. 1.9 77. 1.5
Fig. 5
ratio de l’effort de cisaillement en pied/charge hydrostatique sur la console
centrale.
ratio between the shear strength at the base and hydrostatic load at the central
cantilever.
importantes sont observées pour les ouvrages de petite hauteur dont la faible épais-
seur accroît la sensibilité aux sollicitations thermiques. A l’inverse pour des ouvrages
de grandes hauteurs et donc plus épais, l’influence des conditions thermiques est
moindre.
Le Tableau 1 synthétise les caractéristiques des ouvrages étudiés ainsi que

la cohésion maximale évaluée sous chargement hydrostatique.
La Fig. 5 positionne les ouvrages étudiés sur un graphique comparant l’effort

hydrostatique horizontal sur le profil de plus grande hauteur sur fondation (ρgH2 /2) et
l’effort de cisaillement en pied (estimé par les analyses numériques en multipliant la
214
Q.104 - R.15
cohésion calculée par l’épaisseur amont-aval en pied). Il est implicitement supposé

que pour les ouvrages étudiés, la résultante des efforts en pied d’ouvrage est proche
de l’horizontale : la majeure partie du cisaillement est alors repris par la cohésion
et non par frottement.
En moyenne pour les ouvrages analysés (Fig. 5), 1/3 de la charge hydro-
statique appliquée sur les consoles centrales est transmis sous la forme d’effort
pseudo-horizontal en pied de console. Ainsi, connaissant la hauteur ‘H’ du bar-
rage et l’épaisseur ‘e’ à la base des consoles centrales, il est possible d’évaluer
simplement l’ordre de grandeur des cohésions mobilisées en pied par la formule
suivante :
ρgH 2
c=
6e
Le ratio L/H (longueur/hauteur) qui modifie la répartition entre consoles et arcs,

et l’élancement de l’ouvrage qui joue sur l’épaisseur, accentueront ou réduiront la
cohésion mobilisée à partir de cette valeur moyenne.
4. COMPORTEMENT OBSERVÉS ET SOLUTIONS DE CONFORTEMENT
En complément des analyses globales relatives au cisaillement présentées

dans précédents paragraphes, ce chapitre présente le comportement de 5 barrages
qui ont fait l’objet d’analyses et parfois de renforcements liés à ces problématiques.
4.1. LAOUZAS
Le barrage de Laouzas est une voûte mince en béton à double courbure

(Fig. 6). Il a été construit entre 1961 et 1965. L’ouvrage, d’une hauteur sur fon-
dation de 51 m, est implanté dans une vallée large (rapport L/H voisin de 5). Le pied
amont des plots centraux a été doté, par le concepteur, d’un joint situé au niveau
de la galerie inférieure, permettant l’articulation de la voûte, associé à un ferraillage
important en pied afin de prévenir l’apparition de fissures dans cette zone.
Dès la mise en eau, d’importants débits sont collectés par la galerie de

drainage inférieure. Les pics de débit sont atteints en hiver alors que le niveau
du réservoir se situe dans la tranche des 5 m supérieurs. L’augmentation du débit
de fuites est corrélée au déplacement de la voûte vers l’aval. Le dysfonctionnement
du joint horizontal prévu par le concepteur au niveau des plots centraux est com-
pensé par l’ouverture du contact béton-rocher qui contribue à alimenter les drains
de fondation.
Afin de mieux maitriser les débits, de nouveaux drains ont graduellement

été réalisés en pied aval tandis que les drains en pied amont ont été scellés. La
215
Q.104 - R.15
Fig. 6
Vue aval du barrage de Laouzas et coupe d’un plot central.
Downstream view of Laouzas dam and central cantiliver.
propagation vers l’aval de la fissure affectant le contact béton/rocher et la diffu-

sion de la pression interstitielle se poursuivent jusqu’à proximité du pied aval en
février 2006 avec le cumul d’un état thermique froid de l’ouvrage avec une cote
d’exploitation haute. Dès lors, la cote d’exploitation est provisoirement limitée, le
dispositif d’auscultation étoffé, des études et de travaux de confortement vont se
succéder jusqu’en 2017.
Une modélisation par élément-finis du barrage a été réalisée avec des

éléments-joint à l’interface béton-rocher permettant de modéliser l’ouverture du
contact associé à un couplage hydromécanique permettant de représenter la prop-
agation des sous-pressions [2]. Cette étude montre que l’ouverture du contact
béton/rocher, avec pour conséquence la propagation des sous-pressions sous les
plots centraux jusqu’à proximité du pied aval de l’ouvrage, entraine une réorien-
tation de la résultante des efforts appliqués à la fondation qui devient proche de
l’horizontale (Fig. 7) et ne s’exerce plus que sur la tranche superficielle de rocher
constituant la butée aval de l’ouvrage. Afin que ce mode de fonctionnement soit
préservé, et bien que dans le cas du barrage de Laouzas, le cisaillement calculé
en pied des consoles en considérant les hypothèses proposées dans cet arti-
cle soit modéré (Fig. 4) du fait de l’épaisseur de la voûte en pied, il est apparu
préférable d’assurer le maintien de l’intégrité de la butée rocheuse sous l’action de
la propagation des sous-pressions vers l’aval susceptible d’ouvrir les discontinuités
subhorizontales affectant la fondation.
Afin de stabiliser la butée rocheuse, un dispositif combinant drainage et mise

en place de tirants d’ancrage confinant la tranche superficielle du rocher est pro-
gressivement mis en place au pied des 4 plots centraux (Fig. 8). Les tirants, espacés
216
Q.104 - R.15
Fig. 7
Transfert des efforts de la voûte à l’encastrement.
Forces from between the arch and the embedded rock.
Fig. 8
Dispositif de confortement en pied aval des 4 consoles centrales (vue générale et
vue en coupe de la poutre).
Reinforcement at toe of the 4 central cantilevers (downstream and sectional views).
de 3 m, sont tendus à 3000 kN, valeur correspondant à l’effort exercé par les
sous-pressions.
217
Q.104 - R.15
Fig. 9
Résultantes d’appui – plan horizontal: Plus Hautes Eaux + Chargement thermique
été.
Resulting forces – plan view : Probable Maximum Flood + Summer thermal
loading.
4.2. PONT DU ROI
Le barrage de Pont du Roi, situé sur la commune de TINTRY, dans le départe-

ment de la Saône et Loire, est un barrage de type voûte à double courbure en béton
construit en 1957. Pour une hauteur moyenne de 23 m en partie centrale, et dévelop-
pant une longueur de 196 m en crête, l’ouvrage est implanté en vallée large (L/H
proche de 8). Conçue selon un principe de double courbure, l’épaisseur de la voûte
varie entre 1,5 m en crête et 3,3 m à la base. Le contact béton/fondation, en partie
centrale, est enfoui sur environ 5 mètres de hauteur (sous le terrain naturel) selon
le principe Bioge (ou fouille Bioge).
Une analyse graduée a été effectuée, commençant par un calcul élastique

linéaire montrant que pour le cas de chargement à PHE, le pied amont est en
traction et le pied aval est comprimé. Les contraintes principales de compression
maximales sont atteintes, du fait de l’effet voûte, sur le tiers inférieur des rives. En
rives, les résultantes des efforts sur la fondation sont orientées vers l’aval et vers
les rives. En fond de vallée, elles sont relativement faibles et orientées vers le bas,
ce qui montre l’influence des arcs supérieurs qui supportent l’ouvrage.
L’ajout d’un joint de Mohr-Coulomb permettant l’ouverture au contact

voûte/fondation sur les rives et au contact entre la voûte et la fouille blindée (Bioge)
en partie centrale démontre l’ouverture du pied amont en période froide et pour
les périodes chaudes une orientation des contraintes de compression avec une
forte incidence par rapport à la surface d’appui de la voûte en haut des rives (voir
Fig. 9et Fig. 10). Cette orientation défavorable s’explique par le fait que les arcs
supérieurs de l’ouvrage se projettent davantage vers l’amont sous l’effet thermique
estival. L’ouverture en partie centrale est jugée acceptable tant que les efforts se
reportent correctement sur les rives en situation confortée.
218
Q.104 - R.15
Fig. 10
Résultantes d’appui – plan vertical : Plus Hautes Eaux + Chargement thermique
été.
Resulting forces – elevation plan : Probable Maximum Flood + Summer thermal
loading.
Fig. 11
Vues générales du modèle aux éléments finis du barrage de Pont du Roi.
Finite elements model of Pont du Roi dam – General views.
Il a été fait le choix de vérifier la stabilité des plots des rives du barrage
par un critère d’angle de frottement limite mobilisable à la surface de contact
voûte/fondation, en considérant les blocs RD et RG comme des blocs monolithiques.
En cohérence avec les recommandations CFBR pour les barrages poids, une cohé-
sion nulle est considérée. Pour les sous-pressions, on considère une répartition
triangulaire entre l’amont et l’aval. La résultante d’appui calculée est décomposée
en composante tangentielle et normale. L’angle de frottement mobilisé (ϕmob ) est
comparé à l’angle de frottement limite (ϕlim ). La stabilité est assurée pour ϕmob < ϕlim .
En l’absence de mesures et d’essais, l’angle de frottement limite a été déter-

miné par rétro-analyse. On considère que le barrage est stable sans marge de
sécurité (FS = 1) pour les cas de chargements quasi-permanents (RN+ETE,
RN+HIVER, etc…) qui se sont produits de manière répétée dans la vie de l’ouvrage
219
Q.104 - R.15
et pour lequel l’ouvrage est resté stable. L’angle de frottement limite est alors
assimilé à l’angle de frottement maximal mobilisé pour un cas de chargement
quasi-permanent : ϕlim = max(ϕmob (RN + ETE); ϕmob (RN + HIVER); . . . ).
L’angle de frottement théoriquement mobilisé est calculé sur l’ensemble de

la rive RG et RD considérées comme des blocs monolithiques par une analyse de
type corps rigide. Les résultats obtenus par les calculs donnent un angle limite de
stabilité de la voûte sur le rocher de ϕlim = 60◦ pour chacune des rives. Le calcul
du confortement a ensuite été réalisé en considérant cet angle de frottement afin
d’obtenir les réserves de sécurité suffisantes pour les configurations rares et excep-
tionnelles par l’ajout d’efforts verticaux apportés par la mise en place de tirants actifs
au droit des plots de rives.
4.3. LES CAMMAZES
Le barrage des Cammazes, situé sur les communes de Cammazes, Sorèze

et Saissac, à cheval sur les départements du Tarn et de l’Aude, est un barrage
de type voûte à double courbure en béton construit en 1958. Pour une hauteur
maximale de 70 m, et développant une longueur de 200 m en crête (+ 108 m de
culées), l’ouvrage est implanté dans une vallée fortement dissymétrique avec une
rive droite beaucoup plus douce que la rive gauche. Conçue selon un principe de
double courbure, l’épaisseur de la voûte varie entre 3 m en crête et 15 m à la base.
L’étude de dangers actualisée en 2014 a mis en exergue une dérive vers l’aval
de la clé de voûte du barrage, modérée (0,5 mm/an) mais dont l’origine n’est pas
clairement expliquée, ainsi qu’une faiblesse potentielle au niveau des culées. Une
étude de justification de la stabilité des culées a donc été menée, via la modélisation
aux éléments finis de l’ouvrage et l’étude de stabilité corps rigide des culées.
Un joint non-linéaire a été modélisé au contact béton/fondation. Ce joint

permet l’ouverture du contact si les tractions dépassent une valeur seuil (prise à
0,01 MPa dans le calcul) et le glissement du contact suivant une loi de frottement
de type Mohr-Coulomb. L’ouvrage est d’abord modélisé dans son ensemble afin
d’appréhender son comportement global et d’estimer les confortements nécessaires
pour chacune des culées. Ensuite, la stabilité des culées est évaluée grâce à un
calcul en corps rigide 3D prenant en compte la poussée de la voûte calculée dans
le modèle global.
Le comportement global de l’ouvrage est caractérisé par l’ouverture du contact

béton/rocher dans la partie centrale de la vallée. Cette ouverture, concentrée au
droit du joint non-linéaire dans le modèle, devrait se traduire dans la réalité par
un desserrement de la fondation en pied amont de l’ouvrage, hypothèse qui est
confortée par les mesures d’auscultation piézométrique en fond de vallée. Toutefois,
la configuration géologique du site, tout comme l’historique de son comportement,
sont rassurants pour ce qui est de la stabilité du pied aval de l’ouvrage.
220
Q.104 - R.15
Fig. 12
Vue générale du modèle aux éléments finis du barrage des Cammazes.
Finite elements model of Cammazes dam – General view.
Une étude de c-phi réduction a été menée afin d’évaluer l’ampleur des
déplacements irréversibles et l’impact sur la poussée de la voûte sur les culées.
L’analyse a montré qu’une dégradation des conditions de fondation, se traduisant
par un “ glissement ” des culées sur la fondation, n’entraînait pas de déséquili-
bre de la structure, celle-ci retrouvant un nouvel équilibre (système hyperstatique).
Les déplacements irréversibles sous les culées pour le cas RN sont millimétriques
lorsque l’on passe d’un angle de frottement de 70◦ à 30◦ . Les essais de cisaille-
ment sur les plans de schistosité du rocher de fondation montrent des angles de
frottement résiduels minimaux supérieurs à 30◦ , valeur qui n’est atteinte qu’après
des déplacements tangentiels supérieurs à 5mm.
Les calculs de stabilité en corps rigide valident la stabilité de culée rive droite
dans les conditions actuelles (culée confortée en 2004-2005 par ajout d’une masse
de remblai et drainage).
L’approche de Hoek, considérant un plan de glissement ayant un angle donné

avec les plans de schistosité, a été utilisée pour déterminer les paramètres du con-
tact de la culée rive gauche, puisque ce ne sont pas les discontinuités du rocher qui
vont piloter le comportement de la fondation rocheuse dans le plan étudié. Les résul-
tats obtenus donnent une loi de MohrCoulomb avec une cohésion de 210 kPa un
angle de frottement de 52◦ . Si l’analyse montre que la culée n’a pas glissé historique-
ment sous ses chargements usuels, elle démontre également que les réserves de
stabilité restent insuffisantes, même en acceptant un déplacement irréversible de
221
Q.104 - R.15
Fig. 13
Vue en élévation du barrage des Galens.
Downstream view of Galens arch dam.
la culée. Afin de disposer des réserves de stabilité conformément aux recomman-

dations du CFBR pour les barrages poids, un confortement par tirants actifs a été
retenu pour apporter une force additionnelle de 74 MN.
4.4. GALENS
Le barrage des Galens est une voûte mince construite en 1965 et implantée
dans une vallée large et fondée sur un rocher de type granite de bonne qualité.
La définition géométrique de la voûte est à simple courbure avec un parement aval
vertical et un parement amont légèrement incliné vers l’aval.
Compte tenu de la topographie de la vallée, la retenue est fermée par une

culée sur chacune des rives du barrage.
Les principales dimensions du barrage mettent en évidence une conception

relativement audacieuse. La hauteur maximale sur fondation est de 23 m, dont
19 m au-dessus du terrain naturel. La distance entre les culées est de 109 m, soit
un rapport L/H proche de 5. L’épaisseur varie de 1.5 m en crête à 3 m en pied, et le
rayon de courbure est de 55 m.
La vue en élévation du barrage montre un fond de vallée relativement plat.
Cette topographie, peu favorable à un barrage voûte, ainsi que la sensibilité

de l’ouvrage aux sollicitations thermiques, ont entrainé une fissuration traversante
de l’amont à l’aval. Cette fissuration en pied s’est développée dans une reprise de
bétonnage proche du contact entre le béton et le terrain naturel (Fig. 14).
Plusieurs paramètres permettent de mettre en évidence la sensibilité marquée

de ce barrage aux sollicitations thermiques hivernales. Les calculs montrent une
cohésion mobilisée allant jusqu’à 1 MPa (Fig. 15) lors de l’hiver centennal suivant
les hypothèses décrites dans cet article (contre 0.6 MPa en hiver normal et à peine
0.25 MPa sans sollicitations thermiques). Dans le cas de ce barrage, une telle valeur
222
Q.104 - R.15
Fig. 14
Fissuration en pied aval du barrage.
Crack at the toe of the dam.
Fig. 15
Cohésion mobilisée lors de l’hiver centennal.
Cohesion at the dam-foundation interface during a centennial winter.
de cohésion n’est pas jugée réaliste compte tenu de l’état de fissuration avéré du
pied et des cycles d’ouvertures/fermetures du contact.
Les calculs non-linéaires montrent un glissement supérieur au centimètre

et impactant un nombre significatif de consoles en cas d’occurrence de cet hiver
223
Q.104 - R.15
Fig. 16
Illustration du confortement du barrage des Galens.
Reinforcement of Galens dam.
centennal sans pour autant mettre en péril sa sécurité à court et moyen terme. Ce
comportement calculé n’est pas conforme aux critères des dernières recommanda-
tions du CFBR. La présence de fuites débitantes combinées au site d’implantation
du barrage concerné régulièrement par un gel marqué ont conduit au projet de
confortement.
Le dimensionnement du confortement tient compte de l’histoire de son charge-

ment jusqu’à la période actuelle en simulant plusieurs cycles saisonniers où
la fissure se propage progressivement vers les rives et le glissement potentiel
s’incrémente. En partant de cet état, une solution de confortement a pu être
dimensionnée afin de respecter les recommandations du CFBR. La solution de
confortement prévoit d’épaissir les consoles de rives (environ 3 consoles par rive
sur une totalité de 9 consoles, Fig. 16).
Dans cette configuration confortée, le glissement calculé de l’ordre de 3 mm

et le nombre de consoles montrant un glissement lors d’un hiver centennal est
significativement réduit. En revanche, les respirations saisonnières et hydrosta-
tiques de la console de clé sont faiblement modifiées, garantissant le maintien du
fonctionnement en voûte du barrage.
4.5. SAINT-BARTHÉLÉMY
Le barrage de Saint-Barthélémy (ou barrage du Gouët) se situe à proximité

de Saint-Brieuc dans les Côtes-d’Armor. Il s’agit d’une voûte à double courbure de
224
Q.104 - R.15
Fig. 17
Barrage de Saint-Barthélémy.
Downstream view of Saint-Barthelemy dam.
Fig. 18
Déplacements mesurés par l’élongamètre en pied de console de clé.
Measured displacement at the feet of the central cantilever.
45 m de haut sur ses fondations et de 200 m de longueur en crête. Son épaisseur

maximale en pied est de 6 m. Ces caractéristiques en font un barrage relativement
mince (e/H<0.2) construit en vallée large (L/H>3). Le barrage est fondé sur un massif
en gneiss de bonne qualité.
Lors des premières années de vie de l’ouvrage, un glissement de l’ordre de

quelques millimètres a été enregistré. Ce déplacement s’apparente à une phase
d’adaptation de l’ouvrage, typique de ce genre d’ouvrage, mais néanmoins ampli-
fiée par la largeur de la vallée et la minceur relative de la voûte. Cette adaptation
225
Q.104 - R.15
Fig. 19
Convergence géométrique de l’excavation en rive gauche, défavorable à un
glissement vers les rives.
Geometrical convergence of the excavation on the left abutement.
a également généré une décompression du pied de l’ouvrage dans les consoles

centrales.
La Fig. 18 montre les déplacements mesurés par l’élongamètre situé en pied

de console centrale et mesurant le déplacement relatif pratiquement horizontal entre
le rocher et l’ouvrage en béton. Un déplacement irréversible de quelques millimètres
qui s’atténue après une dizaine d’années est mesuré. Par ailleurs, la vidange par-
tielle de la retenue en 1983 a eu très peu d’effet sur les mesures de cet élongamètre,
confirmant le caractère irréversible du déplacement.
La modélisation de quelques situations historiques connues par l’ouvrage

(séisme, canicule, crue) a permis de retrouver les déplacements mesurés par le
dispositif d’auscultation en calant les paramètres de résistance à l’interface de fis-
suration. Ce calage a mis en évidence la nécessité de considérer des valeurs d’angle
de frottement et de cohésion importantes par rapport à ce qui est habituellement
utilisé par la profession. Ces valeurs s’expliquent d’une part par la géométrie du con-
tact qui converge vers les rives (Fig. 19) combinée à un blocage de pied en béton
de masse, et d’autre part par les aspérités macroscopiques qui ne sont pas repro-
duites par le modèle numérique (fissuration remontant à travers plusieurs reprises
de bétonnages).
Après avoir tenu compte de ces déplacements initiaux, la modélisation des

différentes situations règlementaires montre que l’ouvrage connait une respiration
saisonnière classique, accompagnée d’ouvertures et de refermetures de la fissura-
tion, faisant potentiellement varier les débits de fuite. Par ailleurs, l’étude a montré
que la stabilité de l’ouvrage est satisfaisante et respecte les critères fixés par les
226
Q.104 - R.15
recommandations en termes de contraintes et de déplacements, ne nécessitant

aucun confortement. Un renforcement du dispositif d’auscultation est en revanche
programmé à court terme pour permettre de bien confirmer le bon comportement
du barrage dans le temps.
5. CONCLUSION
Le principe de fonctionnement des barrages-voûtes répartissant la poussée

hydrostatique entre les arcs et les consoles a pour conséquence une sollicitation
en cisaillement dans la direction amont-aval des consoles centrales et dans la
direction rive à rive des arcs en partie supérieure. L’évaluation de mécanismes de
défaillance liés à ces cisaillements est désormais mise en avant par les nouvelles
recommandations françaises pour l’analyse du comportement des voûtes.
L’analyse comparative de plusieurs voutes existantes, majoritairement dimen-

sionnées par des méthodes de calcul simplifiées n’ayant pas évalué ces mécan-
ismes, met en évidence des cisaillements non négligeables aux interfaces consoles-
fondation, avec des cohésions mobilisées pouvant dépasser le MPa ; valeur bien
supérieure à la centaine de kPa qui est habituellement considérée pour le calcul
de stabilité au glissement des barrages-poids par exemple. Cet ordre de grandeur
peut d’ailleurs être aisément retrouvé à partir d’un raisonnement simple considérant
qu’en moyenne, 1/3 de la charge hydrostatique s’appliquant sur la console centrale
est transmise par cisaillement en pied de console.
Pour autant, pour la plupart des voutes analysées, il n’a pas été mesuré
de déplacement vers l’aval en pied confirmant la transmission d’efforts pseudo-
horizontaux à la fondation par 2 mécanismes : par cisaillement au niveau de
l’interface horizontale béton-rocher sous les consoles et par report d’une partie
des efforts à la butée rocheuse aval via l’interface verticale de la voûte encastrée
dans sa fondation. Pour le premier mécanisme, les essais de cisaillement de contact
béton-rocher [3] et [4] mettent en évidence que des cohésions supérieures au MPa
sont réalistes. Concernant le second mécanisme, la prise en compte de la butée
rocheuse aval dans les analyses (généralement négligée) confirme le rôle joué par
le rocher en reprenant une part importante des efforts et réduisant alors la sollicita-
tion en cisaillement de l’interface béton-rocher horizontale. Il faut alors vérifier qu’il
en a bien la capacité à long terme.
Enfin dans certains cas, des mécanismes d’adaptation ont pu être observés
(glissement au contact béton-rocher ou dans le béton juste au-dessus du terrain
naturel) mais il est généralement difficile de les quantifier précisément. Ces adap-
tations n’ont pas conduit à la défaillance des ouvrages : le renforcement de l’effet
227
Q.104 - R.15
voûte en pied reportant les efforts au niveau des appuis et permettant de trouver un
nouvel équilibre en pied.
Les analyses et réflexions originales menées sur cinq voûtes françaises sont
décrites dans ce papier : elles mettent en avant diverses solutions de conforte-
ment mises en œuvre afin de renforcer la sureté des ouvrages : renforcement de
la butée rocheuse aval, renforcement par ancrage des appuis de la voute, épais-
sissement des consoles de rives et suivi de la piézométrie en pied des consoles
centrales.
REFERENCES
[1] Recommandations provisoires pour la justification du comportement des

barrages-voûtes – CFBR – octobre 2018.
[2] E. Robbe – Laouzas Dam : behavior of an arch dam in a wide valley and
foundation stability improvement – USSD annual conference 2020 – Denver.
[3] Coubard G., Deveze G., Vergniault C. Amélioration de la caractérisation de

l’interface barrage-fondation. In Colloque Méthodes et techniques innovantes
dans la maintenance et la réhabilitation des barrages et des digues. Cham-
béry : CFBR – 27,28 novembre 2018. ISBN 979-10-96371-06-8. p. 85–95.
doi : 10.24346/cfbr_colloque2018_a24.
[4] Stone & Webster Engineering Corporation, “Uplift Pressures, Shear strength
and Tensile Strengths for Stability Analysis of Concrete Gravity Dams – Final
Report”, EPRI TR-100345, Project 2917-05, August 1992.
[5] Travaux de confortement du barrage du pont du roi, C. Bourgeois, A. Perche,

Symposium du CFBR du 28 janvier 2016 à Chambéry*
[6] Quel critère pour analyser les contraintes de cisaillement entre un barrage et
sa fondation ? Exemples mis en œuvre sur des projets de confortement, Jouy
C., Noret C., Molin X., Anthiniac P., Colloque CFBR, novembre 2019.
228
Q.104 - R.15
1. INTRODUCTION
There is a large number of arch dams in France (over 80), most of them
built over the two decades following World War II. The designs of most of these
dams were based on simplified calculation methods (assuming independent arches
or possibly taking a radial fit with the central cantilever into account) with verifica-
tion generally limited to assessing the upstream and downstream face stresses. In
particular, the design methods used at the time did not include an analysis of the
state of shear at the concrete/rock contact. The evolution of the behaviour of cer-
tain structures led engineers to question the issue of the transfer of shear stress
to these contacts. These reflections and analyses were integrated into the recent
French guidelines for analysing the behaviour of arch dams [1], summarised in §2.
The framework proposed in these guidelines now provides a common frame of refer-
ence for comparing structures: an original comparative analysis of shear calculated
on around ten arches is thus proposed in §3. Lastly, the special behaviour of five
structures, analysed by several major consulting companies, is the subject of §4:
several reinforcement solutions adapted to each issue are looked at in detail.
2. CFBR GUIDELINES FOR ANALYSING THE BEHAVIOUR OF ARCH DAMS
In 2018, the guidelines [1] published by the French Committee on dams and
reservoirs (CFBR – Comité français des barrages et réservoirs) proposed assessing
the behaviour of arch dams with regard to different failure mechanisms resulting from
global incidentology and accidentology. Amongst these mechanisms, two kinematics
linked to shear at the concrete-rock contact were identified:
• Failure to take stress at the toe of the central blocks: corresponding to stress
of the central cantilevers in the case of arch dams in wide valleys, under the
effect of the hydrostatic load and cold temperatures (Figure 1).
Fig. 1
Failure mechanism under the central cantilevers.
∗ Cisaillement à l’interface béton-rocher des barrages voûtes: retour d’expérience de voûtes
existantes
229
Q.104 - R.15
• Sliding along the abutments: in the case of a structure subject to swelling

pathologies or in thermal situations in summer (Figure 2).
Fig. 2
Failure mechanism at the side cantilevers.
The 2018 CFBR guidelines [1] propose analysing these mechanisms evaluat-
ing in particular the transfer of the shear stress to the contact with the foundation by
comparing the normal and tangential stress at the cantilever – foundation contact.
In the case of the problem at the base of the central cantilevers, the foundation’s sta-
bility relative to the hydraulic gradients generated by the loosening is also analysed,
evaluating the possible propagation of uplift downstream.
A graduated analysis is proposed, starting with a linear elastic approach before

introducing a joint with different types of non-linear behaviour (removal of tensions,
Mohr-Coulomb-type sliding). The linear elastic approach is necessary initially in
order to identify the zones with excessive tensile and/or shear stress. This also
allows the stresses to be quantified. It may be sufficient if the stresses are deemed
acceptable, i.e. for tensile stresses of less than 2 MPa at the upstream toe (in the
framework of Navier’s hypothesis of stress distribution) and for a low overrun of the
shear strength, allowing some of the stress to be transferred to adjacent areas.
If this is not the case, the non-linear mechanisms to remove the tensile stresses
and/or a Mohr-Coulomb-type sliding law must be used. The choice of which of these
mechanisms to use is made in accordance with the linear elastic modelling results.
The acceptability of the arch dam’s behaviour is then judged according to several
criteria:
• Acceptability of the calculated adaptations (transfer of stresses, extent of the

irreversible displacement)
• Evaluation of the safety margins (c-phi reduction method, retro-analysis)
230
Q.104 - R.15
The current guidelines [1] do not set any numerical criteria allowing the accept-
ability of the shear at the arch toe or at the abutments to be judged in the framework
of a linear elastic approach. Indeed, this criterion has not really been analysed for
the design of French structures (only the stresses in the arch were analysed using
simplified methods: calculations in independent arches, possibly with radial adjust-
ment at the crown and, less frequently, using the trial-load method). Consequently,
it was decided when drafting the guidelines [1] not to impose a numerical criterion
on this point before getting sufficient feedback on the existing structures. With this
in mind, Chapter 3 assesses the cohesion mobilised at the cantilever-foundation
contacts of a number of existing structures.
The loosening of the contact zone at the toe of the central cantilevers can
result in the opening of discontinuities spread through the foundation, located at the
concrete-rock contact, or by localised cracks in the concrete, above their embed-
ding in the foundation. These situations have quite different consequences in terms
of structure stability and methods of action. Several arch dams that present this
type of pathology have been studied and some have been reinforced. In chapter 4,
the analyses of five structures performed by three different engineering and design
companies are presented.
3. COHESION MOBILISED AT THE BLOCK-FOUNDATION CONTACTS OF

EXISTING STRUCTURES
The CFBR guidelines [1] propose that the failure mechanisms linked to shear
at the cantilever-foundation contacts be evaluated by comparing the normal and
tangential stresses in these zones and evaluating a sliding stability coefficient per
block. This coefficient depends in particular on the shear strength characteristics of
the concrete-rock contact: friction angle and cohesion.
These types of analyses have already been carried out on a certain number of
arch dams and based on the principle that no displacement has been observed at the
toe in most of them, an evaluation of the shear resistance mobilised is proposed.
In order to compare the shear between several structures of different sizes and
geometry, the choice was made to evaluate the cohesions mobilised at the concrete-
rock contacts of the cantilevers so that the resisting stresses are greater than the
shear stresses taking the uplift into consideration. A friction angle of 45◦ was chosen
to have a uniform criterion of comparison although the irregularities observed at the
base of the excavation generally allow higher values to be considered. The concrete-
rock contacts of the blocks were modelled with smooth surfaces. Several calculation
situations combining hydrostatic loading at the normal operating level and thermal
load (summer or winter) were considered.
In total, the results obtained on 12 arch dams of between 20 and 150 m

high, are presented in the rest of this chapter. These results were obtained using
231
Q.104 - R.15
finite-element or finite difference calculation. The models used use a non-linear

approach capable of releasing the tensile stresses, often at a previously implanted
joint. However, whatever the opening, the shear stress is fully transmitted into the
joint in such a way as to prevent any slipping. This type of approach therefore allows
the required shear strength to be evaluated whilst taking the normal opening of the
contact and the propagation of the associated uplift into account.
3.1. TYPICAL SHEAR STRESSES OF ARCH-DAM BLOCKS
The shear stresses of arch-dam blocks depend directly on the situation (or
load case) studied:
• The maximum stresses of the central blocks occur in winter (shear in an

upstream-downstream direction),
• The maximum stresses of the lateral blocks (or possibly the thrust blocks)
occur in summer (shear in a bank-to-bank direction).
Figure 3 presents, for a structure approximately 30 m in height, the cohesion

mobilised under each block by reducing the friction angle to 45◦ . This cohesion has
been determined without (solid line) or with (dotted line) consideration of uplift (a
specific analysis is proposed in the next paragraph).
Fig. 3
Cohesion at the dam-foundation contact under normal reservoir load (black), in
summer (red) or winter (blue) with or without uplift (dotted or solid line).
232
Q.104 - R.15
The example above clearly highlights maximum shear stress at the central
blocks in winter and at the lateral blocks in summer. The stresses obtained in winter
on the lateral blocks correspond to shear stresses going towards the valley center
without any real physical signification (probable opening of the block joints).
Without taking the uplift into account, the calculated cohesions at the concrete-
rock contacts are about 0.3 MPa on the central blocks in winter and 0.2 MPa on the
lateral blocks in summer.
Taking the uplift into account leads logically to an increase in cohesion in the
central section of the arch (0.4 MPa on the example presented). This effect is limited
on the end blocks considering the low water level at these blocks. In the central part,
taking the uplift into account maximises the cohesions calculated.
3.2. FULL RESULTS
Figure 4 allows the visualisation of the evolution of the cohesion on the 12

arch dams studied from one bank to the other along the concrete-rock contact. It
differentiates the calculation situations with just the hydrostatic loading or combined
with thermal loading in summer and winter.
In the central section, quite considerable cohesions are calculated, varying

from 0.1 to 1.6 MPa (for the highest structure). These values are much higher than
those usually used in France for calculating gravity dams, which rarely exceed a
hundred kPa at the concrete-rock contact and they pose questions on the shear
stress transfer mechanisms at the toe of the cantilevers. Three hypotheses can be
formulated:
• Such cohesions are effectively mobilised at the concrete-rock contact: experi-

mental characterisation tests carried out on samples of different sizes highlight
cohesions with peaks of between 0.5 and up to 2.5 MPa [3][4]. The sometimes
multi-decimetric irregularities observed at the base of the excavation are also
likely to contribute to high shear strength characteristics by dilatancy.
• The embedding of the central blocks in the foundation may allow the rock
abutment downstream to be mobilised when the rock is of a good quality at the
toe: part of the shear stress is then transferred to the rock via this embedding.
It was with this in mind that Laouzas dam, presented in §4.1, was analysed
and reinforced.
• Lastly, in rare cases, an adaptation (irreversible sliding located at a crack
or spread over several zones) at the structure’s toe is observed or strongly
suspected: at the concrete-rock contact (§4.5 for Saint Barthelemy dam) or in
the concrete above the embedding (§4.4 for Galens dam).
Taking the thermal load into account induces a variation in the cohesion of
between 0 and 0.4 MPa approximately in the central part. The largest variations
233
Q.104 - R.15
Fig. 4
Cohésion at the dam-foundation contact for several dams at normal water level
(NWL), NWL+ summer, NWL+ winter
are observed in low structures whose thinness increases sensitivity to thermal

stresses. Conversely, the influence of the thermal conditions is lesser for higher,
thicker structures.
Table 1 summarises the characteristics of the structures studied and the

estimated maximum cohesion under hydrostatic loading.
234
Q.104 - R.15
Table 1
Summary of the results
91. DAM 92. HEIGHT 93. THICKNESS 94. LENGTH 95. RATIO 96. MAX
(M) AT THE AT THE L/H AT COHESION
BASE (M) CREST (M) NWL (MPA)
97. Galens 98. 23 99. 3 100. 109 101. 4.7 102. 0.3
103. Bage 104. 28 105. 7 106. 139 107. 5.0 108. 0.2
109. La Couze 110. 28 111. 6 112. 129 113. 4.6 114. 0.1
115. Saint- 116. 45 117. 6 118. 200 119. 4.4 120. 1.1
Barthelemy
121. Lanoux 122. 45.2 123. 5 124. 176 125. 3.9 126. 0.8
127. La Valette 128. 50 129. 9 130. 163 131. 3.3 132. 0.6
133. Laouzas 134. 52.6 135. 11 136. 295 137. 5.6 138. 0.3
138. Saint-Guérin 140. 68 141. 18 142. 250 143. 3.7 144. 0.7
145. Vouglans 146. 130 147. 22 148. 427 149. 3.3 150. 1.4
151. Roselend 152. 150 153. 20 154. 285 155. 1.9 156. 1.5
Fig. 5
Ratio between the shear strength at the base and hydrostatic load at the central
cantilever.
Figure 5 positions the structures studied on a graph comparing the horizontal

hydrostatic stress at the structure’s highest point (ρgH 2 /2) and the shear stress at the
toe (estimated through numerical analyses by multiplying the calculated cohesion by
the upstream-downstream thickness at the toe). It is implicitly assumed that, for the
studied structures, the result of the stresses at the structure toe is close to horizontal:
most of the shear is then taken up by the cohesion, not by friction.
On average for the structures analysed (Figure 5), 1/3 of the hydrostatic load
applied to the central cantilevers is transferred in the form of pseudo-horizontal stress
at the cantilever toe. Therefore, knowing the height ‘H’ of the dam and the thickness
‘e’ at the base of the central cantilevers, it is possible to make a simple evaluation
235
Q.104 - R.15
of the order of magnitude of the cohesion at the toe using the following formula:
ρgH 2
c=
6e
The ratio L/H (length/height) which modifies the distribution between can-
tilevers and arches, and the length-to-height ratio which affects the thickness, will
accentuate or reduce the cohesion from this average value.
4. OBSERVED BEHAVIOUR AND REINFORCEMENT SOLUTIONS
In addition to the global analyses of shear presented in the paragraphs above,

this chapter presents the behaviour of 5 dams that have been analysed and some
of the reinforcement solutions linked to these issues.
4.1. LAOUZAS
Laouzas dam is a thin arch dam made of concrete with a double curvature
(Figure 6). It was built between 1961 and 1965. The 51 m high structure is situated
in a wide valley (L/H ratio close to 5). The designer put a joint at the level of the
lower gallery in the upstream toe of the central blocks, allowing the articulation of
Fig. 6
Downstream view of Laouzas dam and central cantilever.
236
Q.104 - R.15
the arch, and considerable reinforcement at the toe to prevent the appearance of
cracking in this zone.
From the initial filling, considerable flows were collected by the lower drainage
gallery. The peak flows occurred in winter when the reservoir level is in the top 5 m.
The increase in leakage correlates to displacement of the arch in the downstream
direction. The dysfunction of the horizontal joint included by the designer at the
central blocks is compensated by the opening of the concrete-rock contact that
contributes to supplying the foundation drains.
To control the flow better, new drains have gradually been built in the down-
stream toe, whilst the drains in the upstream toe have been sealed. The propagation
of the cracking affecting the concrete/rock contact in a downstream direction and the
spreading of the interstitial pressure continued to the vicinity of the downstream toe
in February 2006 with the cumulation of a cold thermal state of the structure and a
high operating level. From then on, the operating level was temporarily limited, mon-
itoring was increased, and a succession of studies and reinforcement work were
carried out through to 2017.
A finite-element model of the dam was created with the joint elements at the
concrete-rock contact making it possible to model the opening of the contact com-
bined with a hydromechanical coupling representing the spread of the uplift [2]. This
study shows that the opening of the concrete/rock contact, with the resulting spread
of uplift under the central blocks to the vicinity of the structure’s downstream toe,
gives rise to a reorientation of the resulting stresses applied at the foundation that
comes close to the horizontal (Figure 7) and is only exerted on the superficial part
of rock that constitutes the structure’s downstream abutment. To keep this operating
mode, and although in the case of Laouzas dam the shear calculated at the toe
of the cantilevers considering the hypotheses proposed in this article is moderate
Fig. 7
Transfer of stresses between the arch and the embedded rock.
237
Q.104 - R.15
Fig. 8
Reinforcement at the toe of the 4 central cantilevers (downstream and sectional
views).
(Figure 4) due to the thickness of the arch at the toe, it seemed preferable to pre-
serve the integrity of the rocky abutment under the action of the spread of the uplift
downstream that was likely to open the sub-horizontal discontinuities affecting the
foundation.
In order to stabilise the rocky abutment, a system combining drainage and the
installation of anchors confining the surface part of the rock was gradually put in
place at the toe of the 4 central blocks (Figure 8). The anchors, spaced 3m apart,
are stretched to 3000 kN, a value that corresponds to the stress exercised by the
uplift.
4.2. PONT DU ROI
Pont du Roi dam, in the commune of TINTRY in the Saône et Loire depart-
ment, is a double curvature arch dam made of concrete built in 1957. With an average
height of 23 m in the central section and 196 m long at the crest, the structure is in a
wide valley (L/H close to 8). Designed according to a double curvature principle, the
thickness of the arch varies from 1.5 m at the crest to 3.3 m at the base. The con-
crete/foundation contact in the central section is buried to a depth of around 5 metres
(under the natural terrain) according to the Bioge principle (or Bioge excavation).
238
Q.104 - R.15
A graduated analysis was carried out, starting with a linear elastic calculation
showing that for the PMF load case, the upstream toe is in traction and the down-
stream toe is compressed. The main maximum compression stresses occur, due to
the arch effect, on the lower third of the sides. On the sides, the resulting stresses
on the foundation go in a downstream direction and towards the sides. At the bottom
of the valley, they are relatively low and directed towards the bottom, showing the
influence of the upper arches that support the structure.
The addition of a Mohr-Coulomb joint allowing the opening to the

arch/foundation contact on the sides and the contact between the arch and the sup-
ported excavation (Bioge) in the central section shows the opening of the upstream
toe during cold periods and for warm periods an orientation of the compression
stresses with a strong effect on the bearing surface of the arch at the top of the
sides (see Figure 9 and Figure 10). This unfavorable orientation is explained by the
fact that the structure’s upper arches project further upstream under the summer
thermal effect. The opening in the central section is deemed acceptable as long as
the stresses are correctly transferred to the sides in a reinforced situation.
Fig. 9
Resulting forces – plan view: Probable Maximum Flood + Summer thermal loading.
Fig. 10
Resulting forces – elevation plan: Probable Maximum Flood + Summer thermal
loading.
239
Q.104 - R.15
Fig. 11
Finite element model of Pont du Roi dam – General views.
The choice was made to verify the stability of the blocks on the sides of the dam
using a criterion for the limit friction angle that can be mobilised at the arch/foundation
contact surface, considering the right and left bank blocks as monolithic blocks.
Consistent with the CFBR guidelines for gravity dams, zero cohesion is considered.
For the uplift, a triangular distribution between the upstream and downstream zones
is considered. The resulting force calculated is broken down into tangential and
normal components. The friction angle mobilised (ϕmob ) is compared to the limit
friction angle (ϕlim ). The stability is guaranteed for ϕmob < ϕlim .
In the absence of measurements and tests, the limit friction angle was
determined using retrograde analysis. The dam is considered stable without a
safety margin (FS = 1) for the quasi-permanent load cases (NWL + SUMMER,
NWL+WINTER, etc.) that occur repeatedly over the structure’s lifetime and under
which the structure has remained stable. The limit friction angle is then assimi-
lated to the maximum friction angle mobilised for a quasi-permanent load case:
ϕlim = max(ϕmob (RN + ETE); ϕmob = (RN + HIVER); . . . ).
The friction angle theoretically mobilised is calculated for the whole right bank
and left bank sides considered as monolithic blocks through a rigid body-type anal-
ysis. The results obtained from the calculations give a limit angle of stability for the
arch on the rock of for each of the sides. The reinforcement was then calculated
considering this friction angle to obtain the sufficient safety reserves for the rare
and exceptional configurations by adding vertical stresses by installing pre-stressed
anchors at the side blocks.
4.3. LES CAMMAZES
Les Cammazes dam, in the communes of Cammazes, Sorèze and Saissac,

on the borders of the Tarn and Aude departments, is a double curvature arch dam
240
Q.104 - R.15
Fig. 12
Finite element model of Cammazes dam – General view.
built in 1958. With a maximum height of 70 m and 200 m long at the crest (+ 108 m
of abutment blocks), the structure is in a highly skewed valley with a right bank that
is much gentler than the left bank. Designed on the principle of a double curvature,
the thickness of the arch varies from 3 m at the crest to 15 m at the base.
The danger study that was updated in 2014 highlighted a downstream drift of
the crown of the dam’s arch, moderate (0.5 mm/year) but the origin of which is not
clear, as well as a potential weakness at the abutment blocks. A study to justify the
stability of the abutment blocks was carried out, using a finite-element model of the
structure and a rigid body stability study on the abutment blocks.
A non-linear joint was modelled at the concrete/foundation contact. This joint

allows the contact to open if the tensile stresses exceed a threshold value (taken
at 0.01 MPa in the calculation) and the sliding of the contact according to a Mohr-
Coulomb type friction law. Initially, the whole structure was modelled in order to
understand its overall behaviour and estimate the reinforcement required for each
of the abutment blocks. Then, the stability of the abutment blocks was assessed
using a 3D rigid body calculation taking the thrust of the arch calculated in the
overall model into account.
The structure’s global behaviour is characterised by the opening of the con-

crete/rock contact in the central section of the valley. This opening, concentrated at
the non-linear joint in the model, should actually result in a loosening of the foun-
dation at the structure’s upstream toe, a hypothesis reinforced by the piezometric
monitoring measurements at the bottom of the valley. However, the site’s geologi-
cal configuration, like the history of its behaviour, are reassuring as concerns the
stability of the structure’s downstream toe.
241
Q.104 - R.15
A c-phi reduction study was carried out to evaluate the extent of the irreversible
displacement and the impact on the arch’s thrust on the abutment blocks. The anal-
ysis showed that a deterioration of the foundation conditions, resulting in a “slide”
of the abutment blocks on the foundation, did not unbalance the structure, since it
found a new balance (hyperstatic system). The irreversible displacement under the
abutment blocks in the NWL situation is millimetric at a friction angle of 70◦ to 30◦ .
The shear tests on the planes of schistosity in the foundation rock show minimum
residual friction angles of over 30◦ , a value that is only reached after tangential
displacements of over 5mm.
The rigid body stability calculations validate the stability of the right bank abut-
ment block in the current conditions (abutment block reinforced in 2004-2005 by
adding a fill mass and drainage).
Hoek’s approach considering a slide plane with a given angle with the planes
of schistosity, was used to determine the parameters of the contact of the left bank
abutment block, since it is not the rock’s discontinuities that will control the behaviour
of the rocky foundation in the studied plane. The results obtained give a Mohr-
Coulomb law with a cohesion of 210 kPa and a friction angle of 52◦ . If the analysis
shows that the abutment block has not slipped in the past under the usual loads,
it also demonstrates that the stability reserves remain insufficient, even accepting
an irreversible displacement of the abutment block. To obtain stability reserves in
line with the CFBR guidelines for gravity dams, reinforcement using pre-stressed
anchors was chosen to add 74 MN additional strength.
4.4. GALENS
Galens dam is a thin arch dam built in 1965 in a wide valley on good quality
granite-type rock. The geometric definition of the arch is a single curve with a vertical
downstream face and an upstream face that leans slightly downstream.
Given the valley’s topography, the reservoir is closed by an abutment on each

of the dam’s sides.
The dam’s main dimensions highlight its quite daring design. The maximum
height is 23 m, 19 m of which is above the natural terrain. The distance between the
abutment blocks is 109 m, i.e. a L/H ratio of close to 5. The thickness varies from
1.5 m at the crest to 3 m at the toe and the radius of curvature is 55 m.
The downstream view of the dam shows a relatively flat valley bottom.
This topography, which is not very favourable for an arch dam, and the struc-
ture’s sensitivity to thermal stresses, have resulted in upstream-downstream through
cracking. This cracking at the toe has developed in a construction joint close to the
contact between the concrete and the natural terrain (Figure 14).
242
Q.104 - R.15
Fig. 13
Downstream view of Galens arch dam.
Fig. 14
Crack at the dam toe.
Several parameters highlight the marked sensitivity of this dam to winter ther-
mal stresses. The calculations show a mobilised cohesion of up to 1 MPa (Figure 15)
during the 100-year winter according to the hypotheses given in this article (com-
pared to 0.6 MPa in a normal winter and barely 0.25 MPa without thermal stresses).
In the case of this dam, this kind of cohesion value is not deemed realistic con-
sidering the actual state of cracking at the toe and the contact’s opening/closing
cycles.
The non-linear calculations show sliding of more than a centimeter that would
impact a significant number of cantilevers in the case of the centennial winter occur-
ring without, however, endangering its safety in the short and medium term. This
calculated behaviour does not comply with the criteria of the CFBR’s latest guide-
lines. The presence of combined seepage flows at the dam site which is regularly
affected by severe frosts have led to the reinforcement project.
243
Q.104 - R.15
Fig. 15
Cohesion at the dam-foundation contact during a centennial winter.
Fig. 16
Reinforcement of Galens dam.
The design of the reinforcement takes the history of its loading through to
the current period into account by simulating several seasonal cycles where the
cracking spread gradually towards the sides and the potential for sliding increased.
On the basis of this condition, a reinforcement solution was designed that respects
the CFBR guidelines by thickening the side cantilevers (about 3 cantilevers on each
side out of a total of 9 cantilevers, Figure 16).
244
Q.104 - R.15
In this reinforced configuration, the calculated sliding of around 3 mm and

the number of cantilevers that would slide during a centennial winter is significantly
reduced. However, the seasonal and hydrostatic respiration of the central cantilever
is slightly modified, guaranteeing the dam’s continued arch operation.
4.5. SAINT-BARTHÉLÉMY
Saint-Barthélémy dam (or Gouët dam) is near Saint-Brieuc in the Côtes-

d’Armor department. It is a double curvature arch dam 45 m high and 200 m long at
the crest. Its maximum thickness at the toe is 6 m. Its characteristics make it quite
a thin dam (e/H<0.2) built in a wide valley (L/H>3). The dam’s foundations are on a
massif of good quality gneiss.
During the early years of the structure’s life, a slide of the order of a few
millimetres was recorded. This displacement was like an adaptation phase for the
structure and typical of this type of structure, but nevertheless amplified by the width
of the valley and the relative thinness of the arch. This adaptation also generated a
decompression of the structure’s toe in the central cantilevers.
Figure 18 shows the displacements measured by the extensometer situated

at the toe of the central cantilever and measuring the practically horizontal relative
displacement between the rock and the concrete structure. An irreversible displace-
ment of a few millimetres which diminishes after about ten years was measured.
Moreover, the partial emptying of the reservoir in 1983 had very little effect on the
measurements of this extensometer, confirming the irreversible character of the
displacement.
Fig. 17
Downstream view of Saint-Barthelemy dam.
245
Q.104 - R.15
Fig. 18
Measured displacement at the feet of the central cantilever.
Fig. 19
Geometrical convergence of the excavation on the left abutment.
The modelling of a few historical situations the structure has undergone (seism,
heatwave, flood) has made it possible to find the displacements measured by the
auscultation device by calibrating the resistance parameters at the cracking contact.
This calibration highlighted the need to consider high friction and cohesion values
relative to the ones usually used by the profession. These values can be explained
in part by the geometry of the contact which converges towards the sides (Figure 19)
combined with a mass concrete blockage at the toe, and also by the macroscopic
asperities that are not reproduced by the numerical model (cracking rising through
several construction joints).
After taking these initial displacements into account, the modelling of the dif-
ferent regulatory situations shows that the structure undergoes a classic seasonal
246
Q.104 - R.15
respiration, accompanied by openings and re-closings of the cracking, potentially

varying leakage rates. Moreover, the study showed that the structure’s stability is
satisfactory and respects the criteria set by the guidelines in terms of stresses and
displacement, not requiring any reinforcement. On the other hand, a reinforcement
of the monitoring system is programmed in the short term to confirm that the dam
is behaving correctly over time.
5. CONCLUSION
The functioning principle of arch dams distributing the hydrostatic thrust

between the arches and the cantilevers results in shear stress in an upstream-
downstream direction at the central cantilevers and in a side-to-side direction in the
upper part of the arches. The evaluation of failure mechanisms linked to this shear
is now highlighted by the new French guidelines for analysing the behaviour of the
arches.
A comparative analysis of several existing arch dams, mostly designed using

simplified calculation methods without evaluating these mechanisms, highlights con-
siderable shear at the cantilever-foundation contacts, with cohesions that can exceed
the MPa; a value that is much higher than the hundred kPa that is usually used to
calculate sliding stability in gravity dams, for example. This order of magnitude can
easily be found from a simple reasoning, moreover, considering that on average,
1/3 of the hydrostatic load bearing on the central cantilever is transferred by shear
at the cantilever toe.
However, for most of the arch dams analysed, no downstream displacement

at the toe has been measured, confirming the transmission of pseudo-horizontal
stresses to the foundation by 2 mechanisms: by shear at the horizontal concrete-
rock contact under the cantilevers and by transfer of part of the stress to the rock
abutment downstream via the vertical contact of the arch embedded in its founda-
tion. For the first mechanism, shear tests at the concrete-rock contact [3] and [4]
highlight that cohesions greater than the MPa are realistic. As concerns the sec-
ond mechanism, taking the rock abutment downstream into account in the analyses
(usually neglected) confirms the role played by the rock in taking a significant part
of the stress and thereby reducing the shear stress of the horizontal concrete-rock
contact. It then must be verified that it has the capacity to do so in the long term.
Finally, in certain cases, adaptation mechanisms have been observed (sliding

at the concrete-rock contact or in the concrete just above the natural terrain) but
it is generally difficult to quantify these exactly. These adaptations have not led to
structure failure: a reinforcement of the arch effect at the toe deferring the stresses
at the supports and allowing a new balance to be found at the toe.
247
Q.104 - R.15
The analyses and original reflections carried out on five French arch dams
are described in this document: they put forward various solutions implemented to
reinforce structure safety: reinforcement of the downstream rock abutment, rein-
forcement by anchoring the arch’s supports, thickening the side cantilevers and
piezometric monitoring at the toe of the central cantilevers.
REFERENCES
[1] Recommandations provisoires pour la justification du comportement des

barrages-voûtes – CFBR – octobre 2018.
[2] E. Robbe – Laouzas Dam : behavior of an arch dam in a wide valley and
foundation stability improvement – USSD annual conference 2020 – Denver.
[3] Coubard G., Deveze G., Vergniault C. Amélioration de la caractérisation de

l’interface barrage-fondation. In Colloque Méthodes et techniques innovantes
dans la maintenance et la réhabilitation des barrages et des digues. Cham-
béry : CFBR – 27,28 novembre 2018. ISBN 979-10-96371-06-8. p. 85–95.
doi: 10.24346/cfbr_colloque2018_a24
[4] Stone & Webster Engineering Corporation, “Uplift Pressures, Shear strength
and Tensile Strengths for Stability Analysis of Concrete Gravity Dams – Final
Report”, EPRI TR-100345, Project 2917-05, August 1992.
[5] Travaux de confortement du barrage du pont du roi, C. Bourgeois, A. Perche,

Symposium du CFBR du 28 janvier 2016 à Chambéry*
[6] Quel critère pour analyser les contraintes de cisaillement entre un barrage et
sa fondation ? Exemples mis en œuvre sur des projets de confortement, Jouy
C., Noret C., Molin X., Anthiniac P., Colloque CFBR, novembre 2019.
248
Q.104 - R.16
-------
-------
CONCEPTION ET RETOUR D’EXPÉRIENCE SUR LES VOÛTES ÉPAISSES EN

BCR(∗)
Abdelghani SI-CHAIB
Directeur de Projets, TRACTEBEL
Mathieu ROY
Ingénieur Spécialiste, ARTELIA
Vincent BOINAY
Directeur de Projets, TRACTEBEL
Jean-Christophe GIRARD
Ingénieur Principal, TRACTEBEL
Nicolas ULRICH
Ingénieur Spécialiste, ARTELIA
Frédéric ANDRIAN
Chef de Projets, ARTELIA
France
RÉSUMÉ
Cet rapport présente un retour d’expérience sur des barrages-voûtes épaisses

en BCR à travers les méthodes de dimensionnement et les dispositions particulières
mises en œuvre, notamment au droit des joints de dilatation. La mobilisation de
l’effet d’arc sous sollicitations thermique et sismique est abordée pour les phases
de conception, modélisation et construction.
∗ Design and feedback on thick RCC arched dams
DOI: 10.1201/9781003211808-16 249

Q.104 - R.16
La première partie de cet article met en lumière l’apport des modèles

numériques dans la compréhension du comportement thermomécanique du bar-
rage sur de longues périodes, et notamment au droit des joints de dilatation. Ces
éléments sont développés dans le cadre d’une étude de cas, mettant en évidence
certains aspects fondamentaux des modèles numériques
La deuxième partie de cet article porte une attention particulière au retour

d’expérience croisé entre modélisation et réalisation sur deux autres barrages
voûtes épaisses en BCR. Sont présentés et discutés : les apports et les limites
des modélisations numériques 3D non linéaires, la validation a posteriori des pro-
priétés mécaniques adoptées ainsi que les solutions techniques retenues, telles
que le clavage des joints et la mise en œuvre d’une ceinture parasismique.
SUMMARY
This report presents a feedback on thick RCC arched dams regarding design
calculation methods and specific implemented devices, especially the contraction
joints. The development of the arch effect under thermal and seismic loadings is
discussed, for the design, modeling and construction phases.
The first part of this paper highlights the benefits of numerical models to
understand the long-term thermo-mechanical behavior of the dam in general, and
especially in the contraction joints. These issues are discussed within the scope of
a case study, highlighting some critical aspects of the numerical models.
The second part focuses on a crossed feedback between numerical models

and construction, based on the experience of two thick RCC arched dams. The
benefits and limitations of 3D nonlinear numerical models are presented and dis-
cussed, the retrospective validation of the mechanical properties adopted as well as
the chosen construction provisions such as joint grouting system and seismic belt
are also discussed.
1. INTRODUCTION
Les barrages poids-voûtes, ou voûtes épaisses en Béton Compacté au

Rouleau (BCR) sont des ouvrages massifs dont la stabilité repose en partie sur
l’effet d’arc. Leur monolithisme doit donc être garanti, autant que nécessaire, dans
toutes les situations de chargement pour lesquelles l’effet voûte est requis. Les
déformations volumiques d’origine thermique (retrait consécutif à la dissipation de
la chaleur d’hydratation du ciment) ou liées au retrait endogène peuvent cependant
conduire à des contraintes de traction significatives, pouvant générer des fissures de
nature à mettre en cause le monolithisme. Les barrages-voûtes sont par conséquent
munis de joints de dilatation afin de localiser l’ouverture des fissures.
Contrairement aux barrages-voûtes en béton conventionnel, construits par

plots indépendants et pour lesquels la mise en œuvre de joints coffrés s’inscrit
250
Q.104 - R.16
de manière naturelle dans le planning de construction, les barrages en BCR sont

construits par couches horizontales ; l’exécution des joints nécessite alors une inter-
ruption du bétonnage et soulève des problématiques de coffrage et de mise en place
des équipements. Par ailleurs, dans le cas des voûtes épaisses, la dissipation de
la chaleur d’hydratation est un processus lent. A la fin de la construction et en
l’absence de dispositif de refroidissement, le barrage est encore chaud et les joints
sont fermés. Les éventuelles ouvertures n’apparaitront que quelques années après
la première mise en eau, et peuvent être réparties de manière inégale. Il est par
conséquent nécessaire d’anticiper le comportement thermomécanique du barrage
afin de prévoir les dispositions constructives nécessaires et suffisantes.
En situation sismique, l’effet d’arc se développe davantage que sous charge-

ments statiques et permet de mobiliser les appuis en rives, ce qui est bénéfique
dans le cas de sites fortement exposés aux séismes. Dans ce processus, les ouver-
tures en pied amont et au contact béton / rocher associées dans les cas extrêmes à
des mouvements non linéaires irréversibles (glissements en fondation ou entre plots
le long des joints de dilatation), doivent être évaluées, et leur compatibilité avec les
contraintes d’exploitation et de sécurité vérifiée. L’impact d’un monolithisme partiel
lié à un contrôle plus ou moins strict du retrait thermique et du clavage des joints est
un aspect important dans la conception et la justification de la stabilité de ce type
de barrage en BCR.
La première partie de cette publication traite des possibilités offertes par les
modèles numériques pour l’assistance à la conception des voûtes épaisses en BCR,
et plus particulièrement à l’étude des joints. La deuxième partie aborde les mod-
élisations numériques et les dispositions mises en œuvre pour la conception, dans
des contextes de sismicité intense et de géologie relativement difficile, des barrages-
voûtes de Tabellout, déjà construit, et de BT (nom de code du barrage) en cours de
construction.
2. APPROCHE CALCULATOIRE POUR LA CONCEPTION DES JOINTS DES

VOUTES EPAISSES EN BCR – CAS DU BARRAGE DE JANNEH
2.1. CONTEXTE
Les méthodes de dimensionnement des ouvrages évoluent de pair avec

l’évolution des techniques et des matériaux de construction. En parallèle, le
développement continu des capacités des logiciels de calcul permet désormais
d’accéder à des représentations détaillées de l’ensemble des paramètres utiles à
l’ingénieur. En plus d’anticiper le comportement de l’ouvrage et de vérifier sa sta-
bilité pour différentes situations, les méthodes actuelles basées sur des modèles
numériques permettent de réduire les coûts par optimisation de la géométrie, du
planning de travaux, des dispositions constructives ou encore en concourant à la
définition du matériau de construction le mieux adapté.
251
Q.104 - R.16
Dans le cas des structures massives en béton, la méthode de référence est la

procédure NISA (Nonlinear, Incremental Structural Analysis) [1]. La finalité de cette
procédure est de quantifier le potentiel de fissuration de la structure afin d’évaluer sa
stabilité. Cela comprend l’étude des déformations volumiques du béton (dilatation
ou retrait thermique, retrait endogène, fluage), pour lesquelles la procédure NISA
propose une méthode de modélisation numérique.
Schématiquement, la procédure NISA consiste à réaliser un modèle

numérique 2D ou 3D de l’ouvrage et à résoudre en alternance les transferts de
chaleur (calcul thermique) et la distribution des contraintes-déformations (calculs
mécaniques), en tenant compte des paramètres suivants :
• Paramètres liés aux matériaux : (a) la chaleur d’hydratation du béton; (b) les
caractéristiques mécaniques, comprenant les résistances à la traction et à la
compression, le module d’élasticité, le coefficient de dilatation thermique; (c)
le retrait endogène du béton; et (d) les propriétés thermiques parmi lesquelles
la chaleur spécifique et la conductivité thermique;
• Paramètres liés à l’environnement extérieur : le planning de construction et la
date de démarrage, la température de l’eau et de l’air, le rayonnement solaire
et la température de placement du béton.
Appliquée aux barrages-voûtes, cette méthodologie permet une analyse

détaillée des joints de dilatation, et notamment de leur degré d’ouverture en fonction
du temps. De cette analyse découlent des choix de conception aussi fondamentaux
que la répartition des joints, le dispositif d’injections, la mise en œuvre de clés de
cisaillement ou d’un système de refroidissement. Les études du barrage de Jan-
neh, décrites dans la suite, ont été réalisées sur la base d’une adaptation de cette
méthodologie aux spécificités du projet.
2.2. BRÈVE PRÉSENTATION DU BARRAGE DE JANNEH
Le barrage de Janneh est un barrage poids-voûte en BCR actuellement en

construction au Liban (Fig. 1). D’une hauteur de 157 m sur fondation, il a pour
vocation l’approvisionnement en eau du district de Jbeil, l’irrigation de la plaine de
Nahr Ibrahim et la production d’hydroélectricité (100 MW environ). Il est fondé sur
un rocher calcaire dolomitique de bonne qualité géotechnique.
Initialement conçu comme un barrage-poids, le barrage de Janneh a dû être

arqué pour résister au séisme SES. Par la suite, le profil a pu être aminci grâce à
une géométrie de fond de vallée plus étroite que prévue et suite aux résultats favor-
ables des modèles numériques, permettant ainsi d’économiser un volume de béton
significatif. Dans la conception actuelle, le barrage est muni de 5 joints de dilatation
verticaux traversants. Ces joints sont coffrés et munis de boites de cisaillement et
d’un dispositif d’injection.
252
Q.104 - R.16
Fig. 1
Coupe-type et vue en plan de principe.
Typical cross section and plane view.
2.3. APPROCHE DE CALCUL
Les modèles numériques ont accompagné les études du barrage de Janneh

depuis les phases préliminaires, en se complexifiant peu à peu au gré de l’avancée
de la conception. Le dernier modèle en date vise ainsi à représenter aussi fidèlement
que possible le comportement réel du barrage, depuis le début de la construction
jusqu’à 30 ans après la première mise en eau, en simulant l’évolution temporelle
des caractéristiques des matériaux, de l’environnement extérieur et des actions
mécaniques.
Ce modèle consiste en une reproduction maillée 3D du barrage et d’une par-

tie de la fondation (Fig. 2). Le barrage est divisé verticalement par les cinq joints
de dilatation et horizontalement par des couches de 3 mètres d’épaisseur implé-
mentées pour simuler la construction progressive du BCR. Les calculs sont réalisés
avec le logiciel FLAC3D, développé par Itasca, utilisant la méthode des différences
finies avec un schéma d’intégration explicite particulièrement adapté pour traiter les
problèmes tridimensionnels avec de fortes non-linéarités.
2.4. HYPOTHÈSES DE CALCUL
2.4.1. Composition du BCR
Les études de composition du BCR de Janneh, qui comprennent notamment

des essais en laboratoire et une planche d’essais (Full Scale Trial), étaient encore
en cours lors de la réalisation des calculs numériques. Les éléments présentés dans
cette publication s’appuient donc sur une estimation de la composition (Tableau 1) et
253
Q.104 - R.16
Fig. 2
Modèle du barrage de Janneh.
Model of the Janneh dam.
Table 1
Composition prévisionnelle du BCR
3. Liants 4. 160 kg/m3 (dont 104 kg/m3 de
matériau pouzzolanique)
5. Granulats 6. 2130 kg/m3 (dont 680 kg/m3 de sable)
7. Eau 8. 110 kg/m3
9. TOTAL 10. 2400 kg/m3
des caractéristiques prévisionnelles du BCR, sur la base des résultats préliminaires

disponibles à l’époque.
2.4.2. Retrait endogène et fluage
L’utilisation d’une formulation de BCR riche en liants permet d’atteindre une

ouvrabilité élevée (Vebe time prévisionnel de l’ordre de 15s). Cela favorise un bon
compactage qui se traduit par un réarrangement squelettique favorable des gran-
ulats dans le béton, avec pour conséquence une limitation du fluage. Ce fluage
intervient lorsque la dilatation thermique liée à la chaleur d’hydratation est gênée
par les conditions aux frontières. La simulation du retrait, tel que proposé dans la
procédure NISA, n’est donc pas nécessaire dans le cas du barrage de Janneh.
2.4.3. Hypothèses mécaniques
Le béton et le rocher de fondation suivent une loi de comportement élastique-

linéaire, dont les paramètres sont donnés dans le Tableau 2.
254
Q.104 - R.16
Table 2
Caractéristiques mécaniques et thermiques principales des matériaux
11. PROPRIÉTÉ 12. DIMENSION 13. BCR 14. FONDATION
15. Poids spécifique 16. kN/m3 17. 24 18. 28
19. Module d’Young 20. GPa 21. 15 22. 15
23. Coefficient de Poisson 24. – 25. 0.2 26. 0.25
27. Coefficient de dilatation 28. K−1 29. 9·10−6 30. 9·10−6
thermique
Le modèle comporte deux types de joints non-linéaires, au droit du contact

béton/rocher et au niveau des cinq joints de dilatation. Le contact béton/rocher suit
une loi de comportement de type frottement de Coulomb sans résistance à la trac-
tion. Il peut par conséquent s’ouvrir, glisser, et refermer des zones précédemment
ouvertes en cas d’alternance de la direction du chargement. Les joints de dilata-
tion n’ont pas de résistance à la traction et sont équipés de boîtes de cisaillement,
interdisant tout déplacement relatif radial entre deux plots adjacents.
2.4.4. Hypothèses thermiques
Dans le modèle, les matériaux suivent une loi de comportement thermique

isotrope. Les échanges thermiques suivants sont reproduits :
• Génération de chaleur par hydratation du ciment ;

• Conduction thermique dans le corps des matériaux;
• Convection thermique avec l’eau et l’air;
• Radiation solaire sous la forme d’une augmentation de la température de l’air
sur le parement aval plus exposé au soleil;
Les quantités de chaleur impliquées dans ces échanges dépendent de la tem-

pérature du milieu extérieur et des propriétés des matériaux. Les températures de
l’air ambiant et de l’eau de la rivière utilisées dans le modèle sont des moyennes
journalières des températures relevées sur site. Les propriétés thermiques des
matériaux (Tableau 2) sont déterminées sur la base d’essais spécifiques en labora-
toire ou bien sont des valeurs conservatives issues de la littérature. Des analyses de
sensibilité ont par ailleurs été effectuées afin d’éprouver la pertinence des valeurs
choisies.
Un des paramètres fondamentaux du calcul thermique est la chaleur

d’hydratation du ciment puisqu’elle détermine la montée en température du béton
(Fig. 3). Le ciment prévu pour la construction du barrage bénéficie d’une relative-
ment faible chaleur d’hydratation (125 J/g à long terme) grâce à la forte proportion
de matériaux pouzzolaniques.
La montée en température du béton dépend de la température du milieu

extérieur au moment où il est coulé. En hiver, l’air froid dissipe une large part de
255
Q.104 - R.16
Fig. 3
Evolution de la chaleur d’hydratation du ciment.
Cement hydration heat vs. time.
la chaleur d’hydratation du ciment et limite la montée en température du béton.

En été, c’est l’inverse. La durée de la construction du barrage étant de l’ordre de
deux années, un zonage thermique initial s’établit dans le corps du barrage entre
les zones plus froides coulées en hiver et celles plus chaudes coulées en été. Ce
zonage thermique tend à s’estomper avec les années, se transformant en zonage
de contraintes et de déplacements.
La date de début de construction est donc un élément déterminant dans le

comportement du barrage. Deux dates de commencement ont ainsi été testées par
le calcul, la première en hiver (1er février) et l’autre en été (1er août).
2.5. SÉQUENCE DE CALCUL
Comme évoqué précédemment, le modèle reproduit les trente premières

années de vie du barrage. La construction du barrage est simulée par une activation
successive des couches de BCR, en conformité avec le programme de travaux de
l’Entreprise. Les cycles annuels d’exploitation de la retenue sont pris en compte
sous la forme d’une variation du niveau d’eau amont.
De manière analogue à la procédure NISA, le couplage thermo-mécanique

utilisé consiste en une alternance entre les calculs thermiques, qui visent à déter-
miner la distribution de températures à une date donnée, et les calculs mécaniques
qui traduisent les variations de température en déformations puis en contraintes via
la loi de Hooke.
2.6. RÉSULTATS ET DÉFINITION DES DISPOSITIONS CONSTRUCTIVES
L’historique des températures montre qu’un état stationnaire est atteint après
une vingtaine d’années. La température maximale calculée est de l’ordre de 34◦ C
256
Q.104 - R.16
Fig. 4
Séquence de calcul (5 premières années)
Calculation sequence (first 5 years).
et le refroidissement maximal d’environ 23◦ C (valeur au cœur du barrage, dans

les zones bétonnées en été). Ces valeurs sont cohérentes avec les mesures de
températures effectuées sur le banc d’essai.
Un tel refroidissement conduirait à des contraintes thermiques de traction de

l’ordre de 3.1 MPa si le retrait thermique était empêché. Au contraire, la présence
des joints de dilatation permet un relâchement significatif de ces contraintes qui sont
converties en ouvertures. Après 30 ans, seules quelques zones majoritairement
situées à des singularités géométriques conservent des tractions supérieures à
1 MPa, qui est la résistance prévisionnelle à la traction verticale du béton. Afin de
relâcher les éventuelles contraintes de traction d’arc qui subsisteraient dans le corps
de l’ouvrage, chaque plot sera donc équipé côté amont d’un ou deux joints induits,
régulièrement espacés entre les joints principaux et équipés de waterstops.
Au droit des joints, la localisation des ouvertures dépend de la situation de

calcul (Niveau d’eau amont) mais également de la date de début de construction
(Fig. 5). Comme évoqué précédemment, les ouvertures les plus significatives sont
situées là où le retrait thermique est le plus important, c’est-à-dire dans les par-
ties coulées en été. A retenue normale, il n’y a aucun problème de fuites puisque
l’ouverture des joints est limitée par l’effet d’arc. A la cote minimale d’exploitation
cependant, les ouvertures sont plus larges et traversent le barrage en plusieurs
endroits. Des joints waterstops seront par conséquent mis en œuvre. En revanche,
la stabilité du barrage ne nécessite pas d’effet d’arc pour cette cote de retenue.
Corollairement, les résultats du calcul confirment que les injections de clavage

ne sont pas utiles à la fin de la construction puisque les joints sont fermés et qu’elles
ne peuvent/doivent être envisagées qu’après une dizaine d’années. Un disposi-
tif d’injection sera néanmoins mis en œuvre afin de conserver une possibilité de
diminuer les ouvertures au contact béton/rocher et par conséquent les débits de
fuite si nécessaire.
257
Q.104 - R.16
Fig. 5
Ouverture des joints verticaux.
Contraction joint opening.
Fig. 6
Modèle de calcul d’une voûte à double-courbure en BCR
Calculation model of a double-curvature RCC arch dam.
2.7. CONCLUSION ET PERSPECTIVES
Le modèle numérique du barrage de Janneh a permis de valider la concep-

tion et de préciser les dispositions constructives, notamment au droit des joints de
dilatation.
D’un point de vue plus général, on assiste aujourd’hui au développement de

BCR riches à faible potentiel de retrait et à ouvrabilité élevée, ce qui permet à la
pâte excédentaire de remonter en surface lors du compactage, formant ainsi un
mortier de liaison et assurant une étanchéité de bon niveau ainsi qu’une résistance
à la traction accrue entre les couches. Dans ces conditions, il devient désormais
envisageable de passer progressivement du poids-arqué, vers la voûte épaisse puis
vers une voûte pratiquement aussi optimisée qu’une voûte en béton conventionnel
258
Q.104 - R.16
vibré. A titre d’exemple, les études préliminaires du barrage supérieur d’une station
de pompage-turbinage à Dubaï (Fig. 6), ont confirmé la pertinence technique de
l’emploi de BCR dans une voûte à double courbure.
3. RETOUR D’EXPERIENCE SUR LA CONCEPTION ET LA

CONSTRUCTION DE DEUX GRANDS BARRAGE-VOUTES EN BCR
DANS UN CONTEXTE DE FORT ALEA SISMIQUE
3.1. PRÉSENTATION DES BARRAGES RÉALISÉS OU EN COURS DE CONSTRUCTION

ET SPÉCIFICITÉS DES SITES
3.1.1. Barrage de tabellout
Le barrage de Tabellout fait partie du projet de transfert Setif-Hodna situé dans

le nord de l’Algérie et destiné à l’alimentation en eau potable (20%) et à l’irrigation
(80%) de la région d’El Eulma. D’une hauteur de 121 m et d’une longueur en crête
de 360 m, ce barrage en BCR en voûte épaisse, situé à l’aval de la confluence de
l’oued Djendjene avec l’oued Raha dans la Wilaya de Jijel, créera une retenue avec
un volume total de 294 millions de m3 .
Au démarrage des études d’exécution en 2009, une étude spécifique de l’aléa

sismique a été réalisée [1]. Cette étude, confortée par un levé géologique [5] a
Fig. 7
Vue en relief d’un extrait de la feuille de Tamesguida.
3D View of an extract from the Tamesguida map.
259
Q.104 - R.16
révélé que le site du barrage de Tabellout est localisé à l’aplomb du chevauchement

de Petite Kabylie (Fig. 7) avec une probabilité non-négligeable d’une propagation de
rupture sismique à proximité immédiate du site du barrage. L’accélération maximale
au sol (Peak Ground Acceleration- PGA du SMP) est estimée à 0,70g. La mise en
eau du barrage de Tabellout a débuté en janvier 2019 et s’est déroulée avec succès.
A l’heure où nous écrivons ces lignes, la cote du plan d’eau a atteint 303 NGA ce qui
représente un remplissage de l’ordre de 93% de la hauteur. A ce stade et au vu des
données d’auscultation, le barrage montre un comportement général parfaitement
correct. Trois fuites localisées en rives, à travers des fissures au droit du contact BCR
et galeries BCV, sont apparues en phase de remplissage. Les débits restent stables,
avec une légère tendance à la baisse, et ne dépassent pas les 10 l/s en cumulé.
3.1.2. Barrage de bt
Le projet, en cours de construction en Asie du Sud-Est, est un barrage

BCR/BCV de 74 m de haut destiné à la production hydroélectrique. Le corps du
barrage est composé de 3 plots. La maîtrise des eaux en phase de crues ou de
vidange est assurée par trois pertuis de demi-fond et un évacuateur de surface,
implantés respectivement dans le plot central et le plot rive gauche.
Le site du barrage est localisé dans une zone caractérisée par une activité
sismo-tectonique intense, dont la source prépondérante est située à moins de 10 km.
L’évaluation de l’aléa sismique pour le projet conduit à considérer un séisme de base
d’exploitation SBE, avec une accélération maximale au sol (PGA) de 0.38g et un
séisme maximal probable SMP, de PGA estimée à 0,65g. La fondation du barrage
est rocheuse, composée de tufs volcaniques de qualité moyenne. La configuration
de la vallée permet de mobiliser un effet voûte et d’améliorer, à moindre coût, le
comportement du barrage sous sollicitation sismique.
3.2. CONCEPTION
3.2.1. Conception du barrage de tabellout
Au démarrage de la phase d’exécution, les données sismiques actualisées

ont conduit à modifier la géométrie de l’ouvrage, d’une forme rectiligne classique
d’un barrage poids en BCR, avec un fruit total de 0,9/1 à une nouvelle géométrie
arquée (Fig. 8). Ce choix a permis de limiter le volume du BCR et améliorer la tenue
du barrage sous les différents chargements notamment sismiques, en transmettant
les efforts sismiques à la fondation au moyen d’arcs plongeants.
Afin de justifier le comportement du barrage sous les différents cas de charge

auxquels il peut être soumis, une modélisation tridimensionnelle aux éléments finis a
260
Q.104 - R.16
Fig. 8
Ancienne forme du barrage (g) et nouvelle forme du barrage(d).
Old shape of the dam (l) and new shape of the dam (r).
Fig. 9
Modélisation COBEF (g) Effet d’arc (d).
COBEF Model (l) Arch effect (r).
été réalisée avec le logiciel COBEF (Fig. 9). La géométrie du barrage a été optimisée
pour les cas statiques et vérifiée pour les sollicitations dynamiques.
La fondation a été supposée sans masse et les déplacements sur son pour-
tour ont été bloqués. Des éléments joints avec un comportement non linéaire de
type Coulomb ont été utilisés pour modéliser le contact barrage fondation. Des
hypothèses prudentes ont été choisies pour définir le ratio entre le module du béton
du barrage Eb et du massif de fondation Er, avec un rapport Eb / Er égal à 2.
Pour matérialiser l’indépendance des plots du barrage les uns des autres
pendant la phase de construction de l’ouvrage, une loi de comportement orthotrope
du béton qui annule les contraintes tangentielles de la voûte a été utilisée unique-
ment pour l’application du poids propre des plots. Sous les autres cas de charges
les efforts sont transmis sur les rives par effet d’arc. Ces contraintes d’arc corre-
spondant au comportement du barrage après clavage, sont illustrées ci-après, avec
une coupe à mi-hauteur du barrage (Fig. 3). Elles ne sont pas réparties dans toute
l’épaisseur du barrage. C’est ce qu’André Coyne appelait arc actif, et qui correspond
aux zones où passent les contraintes d’arcs.
261
Q.104 - R.16
Fig. 10
Accélérations maximales en clef du barrage.
Maximum acceleration in the middle of the dam.
La surpression dynamique de l’eau a été prise en compte avec l’approche de

Westergaard. Suite au calcul linéaire transitoire avec un amortissement pris égal à
5%, l’enveloppe des accélérations maximales obtenues a été tracée. L’accélération
maximale en crête est importante et avoisine 4g (Fig. 10).
Sous l’effet des séismes, les oscillations du barrage font que les parements
sont alternativement en compression et en traction. Les tractions d’arcs se traduisent
par une ouverture des joints de construction, qui vont s’ouvrir et se refermer rapi-
dement pendant le séisme. L’élongation maximale en crête a permis d’optimiser
la localisation des joints coffrés qui seront liés en crête par une ceinture sismique
pour minimiser le décalage potentiel des déplacements amont-aval entre deux plots
adjacents, et favoriser la transmission des contraintes d’arc.
3.2.2. Conception du barrage de bt
Le profil initial du barrage poids avec un fruit total de 0.95, dimensionné pour
les cas statiques et les situations sismiques par des méthodes pseudo-statiques ou
spectrales simplifiées, a été conservé tandis que l’axe rive à rive a été arqué suivant
un rayon de 120 m environ. Dans une phase préliminaire de dimensionnement,
un modèle numérique de l’ouvrage et de sa fondation a été réalisé avec le logiciel
DIANA-FEA. Des éléments joints “ no tension ” ont été adoptés à l’interface barrage
fondation et dans les 2 joints de dilatation inter-plots (Fig. 11).
La fondation, dont le module est environ 6 fois inférieur à celui du béton, a

été modélisée par des éléments non pesants et bloqués en déplacement sur son
262
Q.104 - R.16
Fig. 11
Vue 3D du modèle avec maillage.
3D view of the model Meshing.
pourtour. Le phasage de construction selon le planning de l’Entreprise a été pris

en compte pour modéliser le chargement de poids propre par couches sur les 3
plots verticaux, ainsi que le retrait thermique (hypothèse de température moyenne
après prise du béton de 40◦ C et d’un refroidissement à long terme de 15◦ C) avec
possibilité d’ouverture des joints de dilatation. La mise en eau du réservoir est
traduite par l’application d’une pression hydrostatique du réservoir (70,5 mètres de
colonne d’eau sur le parement amont) et les sous-pressions associées. L’analyse
dynamique, comporte une analyse modale (sur le modèle linéarisé, éléments joints
bloqués) et une recomposition CQC des réponses spectrales de l’ouvrage aux
séismes SBE et SMP appliquées à l’ouvrage. L’accélération verticale est définie par
un ratio de l’accélération horizontale, variant de 2/3 à 1 [6]. L’interaction couplée de
l’ouvrage avec le réservoir est modélisée par des éléments fluide(incompressibles)
dans l’analyse modale. Les chargements sismiques combinés suivant la méthode de
Newmark [7] ajoutés aux charges statiques sont appliqués sur le modèle comportant
des non-linéarités (joints de clavage et contact béton rocher).
La Fig. 12, à gauche, présente les spectres de sol horizontaux et verticaux à

5%, ainsi que les fréquences propres des modes principaux dans les 3 directions. La
Fig. 12, à droite, montre que la réponse dynamique est principalement gouvernée
par le mode 1 (flexion “ simple ” des arcs et déformée en console de clef similaire
au profil des accélérations de la figure).
Le clavage des joints inter-plots (numériquement réalisés par la réinitialisation

des déplacements des nœuds en vis-à-vis) et l’effet du retrait thermique du béton
ont été étudiés selon les scénarios suivants:
• Clavage “ parfait ” : Le béton est refroidi à température ambiante moyenne

avant clavage.
• Clavage “ partiel ” : Le clavage est réalisé 6 mois après la fin de construction
du barrage, alors que son cœur est encore chaud.
• Sans clavage : Cas extrême où les joints ne sont pas clavés, et où le béton se
refroidit naturellement.
263
Q.104 - R.16
Fig. 12
Spectres d’accélération et modes propres (g) Profils des accélérations maximales
(combinaisons de Newmark) en clef de l’ouvrage, et du mode (1er )
prépondérant (d.).
Acceleration spectra and eigen modes (left) maximum acceleration (Newmark
combination) and main mode (1st ) shapes, along the vertical crown section of the
dam according to Newmark combined load (right).
La Fig. 13 montre qu’en absence de clavage et de contrôle thermique du

béton, les joints de construction restent ouverts sous le chargement hydrostatique
de la retenue normale ; mais que, sous sollicitation sismique extrême vers l’aval
(SMP), ils se referment au moins partiellement, permettant le développement d’arcs
actifs contribuant significativement à l’effet voûte, comme le montrerait l’examen des
contraintes et des déformées.
Cette première phase d’étude, reposant sur une méthode simplifiée, a permis
d’appréhender le comportement dynamique au séisme de l’ouvrage, de valider cer-
tains éléments de conception et d’orienter la définition des dispositions constructive
Fig. 13
Ouverture maximale du joint vertical rive gauche sous chargement hydrostatique
(g.) et avec application du SMP vers l’aval (dr.) avec 3 scénarios de clavage (de g.
à dr.) : “ parfait ”, “ partiel ” et sans clavage.
Maximal opening of the left bank vertical joint under hydrostatic loading (l.) And
with application of the SMP downstream (r.) with 3 joint grouting scenarii (from l.
to r.) “ perfect ”, “ partial ” and no grouting.
264
Q.104 - R.16
(joints, clavage et contrôle thermique de béton), qui seront à préciser et ajuster en

fonction de critères d’acceptabilité (à définir) des sollicitations et des déformations
non linéaires (ouverture des joints, décollement et glissement sur fondation, etc.)lors
d’une phase d’étude ultérieure à l’aide d’une modélisation plus fine et représentative
des comportements dynamiques et non linéaires.
3.3. DISPOSITIONS CONSTRUCTIVES
Les techniques de mise en place du BCR pour un barrage en voûte épaisse

sont globalement identiques à celles des barrages poids en BCR. Cependant, des
dispositions constructives particulières peuvent être adoptées pour gérer notam-
ment les joints entre plots, l’injection de ces derniers ainsi que l’interface entre le
corps en BCR et certaines parties en BCV. Diverses solutions constructives alter-
natives plus ou moins contraignantes ont été étudiées, pouvant tolérer un clavage “
imparfait ” et une certaine ouverture “ rémanente ” des joints après construction.
3.3.1. Joints coffrés
Dans le cas de BT, le modèle a mis en évidence les effets bénéfiques, mais
pas forcément indispensables, d’un contrôle thermique assurant au béton une tem-
pérature d’équilibre avant clavage, sur le développement des effets d’arc, et corréla-
tivement sur l’ouverture du contact de la fondation ainsi que des joints de dilatation.
La mise en place des solutions et méthodes “ classiques chinoises ” de joints BCR
réalisés par éléments en béton préfabriqué, assurant une surface de joint continu
ou discontinu (amorçage des joints de retrait) et pouvant intégrer des dispositifs
d’injection sont toujours à l’étude.
Pour le barrage de Tabellout, caractérisé par un BCR dosé avec 130 kg/m3 de
ciment et 140 kg/m3 de filler calcaire, six joints coffrés ont été disposés de manière
optimale, en prenant en compte les contraintes techniques et contractuelles du
projet. Ces joints distants de 50 m ont été associés à des joints non coffrés réalisés
avec une lame hydraulique localisés tous les 25 m soit à mi-distance de chaque plot.
Au droit de chaque joint coffré, des boites de cisaillement de 30 cm d’épaisseur et
réalisées avec des coffrages préformés en béton, ont été adoptées (Fig. 14). La
méthodologie consiste à utiliser un coffrage pour réaliser le premier bloc. Une fois
que ce premier bloc est réalisé (sur une hauteur d’une levée 1,20 m), il devient
lui-même coffrage pour la réalisation du deuxième bloc qui lui est contigu.
3.3.2. Système d’injection
Sur le barrage de Tabellout, chaque joint coffré, a été divisé en plusieurs

compartiments variant verticalement en fonction de la hauteur du joint et équipé
265
Q.104 - R.16
Fig. 14
Coffrage des boites de cisaillement.
Shear boxes and formworks.
Fig. 15
Système d’injection des joints.
Joints injection devices.
d’un dispositif d’injection. Les limites horizontales entre compartiments sont situées
au niveau des galeries de drainage. Ces compartiments comprennent des rampes
d’injection primaires et secondaires et des lignes de lavage (Fig. 15).
Des rainures verticales, permettant la diffusion du coulis au sein d’un même

compartiment ont été réalisées lors du coffrage du plot primaire. Pour chaque com-
partiment, l’injection se fait de bas en haut. Le coulis est collecté par la suite dans la
galerie de drainage, permettant ainsi de vérifier la bonne circulation du coulis dans
le joint.
266
Q.104 - R.16
Fig. 16
Ceinture parasismique.
Seismic belt.
3.3.3. Ceinture parasismique
Pour renforcer le comportement monolithique du barrage de Tabellout, une

ceinture parasismique, constituée de 64 barres d’acier HA 40 contenues dans une
section de BCV, est positionnée dans la partie haute du barrage (Fig. 16). Cette
ceinture parasismique aide notamment à limiter les déplacements relatifs entre les
plots sous l’action sismique. Au droit des joints de clavage, ces barres sont équipées
d’une gaine (similaire au goujon).
4. ENSEIGNEMENTS RETENUS
4.1. APPORTS DES MODÈLES NUMÉRIQUES
Dans les voûtes épaisses, l’équilibre thermique n’est généralement atteint

qu’après plusieurs années, les joints conservant un potentiel d’ouverture même
après l’éventuel clavage en fin de construction. Dans la mesure où la stabilité de ce
type de barrages dépend de la mobilisation de l’effet d’arc, il est essentiel d’anticiper
les conséquences d’un clavage imparfait sur le comportement de l’ouvrage en toutes
situations, et en particulier sous chargement sismique.
Les modèles numériques s’avèrent être des outils particulièrement perfor-

mants pour répondre à ces problématiques, dans la mesure où ils permettent de
simuler finement le comportement thermo-mécanique et dynamique des ouvrages.
Plus particulièrement, grâce aux modèles numériques, il est désormais possible
de quantifier l’évolution temporelle des déformations dans le volume du béton et
267
Q.104 - R.16
les ouvertures de joints qui en résultent, et par conséquent de se prononcer sur

la mobilisation de l’arc actif. Les modèles numériques permettent ainsi d’envisager
des économies significatives en matière de géométrie (Janneh, Tabellout), de définir
les dispositions constructives relatives aux joints de dilatation (Janneh, Tabellout,
barrage de BT) et de préciser les besoins à long terme (injections du barrage de
Janneh notamment). En situation sismique, l’analyse numérique permet enfin de
concevoir des dispositions appropriées (ceinture parasismique de Tabellout).
En cohérence avec le retour d’expérience décrit dans la littérature, les cal-

culs thermomécaniques du barrage de Janneh confirment l’importance des études
paramétriques en l’absence d’essais spécifiques. En particulier, il paraît essen-
tiel d’étudier en détail certains paramètres majeurs afin d’encadrer la réponse de
l’ouvrage et d’orienter sa conception : (i) module d’élasticité du barrage et de la fon-
dation, (ii) chaleur d’hydratation du ciment, (iii) coefficient de dilatation thermique,
(iv) température de placement du béton, dépendant notamment du planning de
construction.
De plus en plus, les modèles de calcul doivent être considérés comme des
outils évolutifs permettant d’accompagner un projet depuis les phases d’étude
préliminaires jusqu’à la fin de construction, et même au-delà. Les paramètres de
calcul peuvent en effet être recalés en fonction de l’évolution de la conception
et des conditions rencontrées sur site (planning de travaux effectif, températures
extérieures, niveau de retenue). En particulier, les calculs doivent nécessairement
être lancés après le choix définitif de la composition du béton.
4.2. DISPOSITIONS CONSTRUCTIVES MISES EN PLACE
Les enseignements retenus en phase de conception doivent se traduire par

la mise en place de dispositifs de construction qui permettent de concrétiser les
résultats des calculs. Cette étape qui correspond à la phase d’exécution nécessite
une réflexion qui lie l’aspect théorique de la modélisation, au retour d’expérience sur
d’autres projets ainsi qu’à la spécificité du projet aussi bien sur le volet technique
que contractuel. Pour une voûte épaisse en BCR, le retour d’expérience montre la
pertinence de l’alternance entre des joints coffrés munis de clés de cisaillement et
d’un dispositif d’injection, et de joints induits non coffrés. Ce système, mis en œuvre
sur le barrage de Tabellout, le sera également sur le barrage de Janneh.
5. REMERCIEMENTS
Les auteurs tiennent à remercier A. Yziquel (ARTELIA) pour sa participation

active au développement et à la validation de la méthode de calcul. Les auteurs de
TRACTEBEL tiennent à remercier J-L Cervetti et B. Tardieu pour leur contribution
268
Q.104 - R.16
tout au long de la phase d’étude et de réalisation du projet de Tabellout, ainsi que

l’équipe de la Direction technique et tous ceux qui ont participé de près ou de loin à
l’élaboration des études notamment X. Molin, M. Valadié, C. Bordas et une pensée
particulière à C. Honnert qui nous a quitté.
REFERENCES
[1] USACE ETL 1110-2-365 (1994). Nonlinear, Incremental, Structural Analysis

of Massive Concrete Structures.
[2] ROY M., ANDRIAN. F., MATHIEU G., YZIQUEL A. (2017). Thermo-mechanical
Calculations of the Janneh Dam. 14th International Benchmark Workshop on
Numerical Analysis of Dams, Stockholm, Sweden.
[3] ICOLD-CIGB. Bulletin 177. Béton Compacté au Rouleau.
[4] ANDRIAN F., MINE E., BERCHER Q., MATHIER G., CERVETTI J.L. ET
BERAUD M., Retour d’expérience sur la conception et la technique de con-
struction de quelques barrages en BCR de grande hauteur. (2021). 27eme
Congrès des Grands Barrages. Marseille. Q104.
[5] BES DE BERC S., WINTER TH. Evidence d’activité récente le long du système
de chevauchements de Petite Kabylie (Algérie). BRGM report for Coyne et
Bellier, Paris 2009.
[6] BOUGDAL R. Mission sur site du 06 au 13 Octobre 2009, Géologie, géo-

morphologie et caractéristiques géotechniques de deux sites de barrage à
Tabellout. France : Tractebel Engineering – Coyne et Bellier.
[7] U. S ARMY CORPS OF ENGINEERS Response Spectra and Seismic Analysis

for Concrete Hydraulic Structures. EM 1110-2-6050, 1999.
[8] NEWMARK N. M., HALL W. J. Earthquake spectra and design. Earthquake

Engineering Research Institute, June 1982.
[9] CERVETTI J-L., SICHAIB A., VALADIE M., RACHEDI R., Special features of
Tabellout RCC arched dam, HYDRO 2015, Bordeaux
[10] ROY M., MATHIEU G., ANDRIAN F., FRAY S., BOINAY V., GIRARD J-C., Jus-
tification des barrages poids-voutes en BCR – deux cas d’application, CFBR
2019, Chambéry
269
Q.104 - R.17
-------
-------
RETOUR D’EXPÉRIENCE SUR LA CONCEPTION ET LA TECHNIQUE DE

CONSTRUCTION DE QUELQUES BARRAGES EN BCR DE GRANDE
HAUTEUR(∗)
Frédéric ANDRIAN, Quentin BERCHER & Geoffrey MATHIEU

ARTELIA
Edouard MINE & Jean Louis CERVETTI

TRACTEBEL
FRANCE
RÉSUMÉ
L’obtention de paramètres de résistance acceptables entre les couches est

l’un des principaux enjeux de la construction des barrages en BCR. Ce matériau et
sa mise en œuvre continuent donc à connaître des évolutions pour faciliter l’atteinte
des critères propres à chaque projet de manière optimisée. Le présent article décrit
et analyse le retour d’expérience des bureaux d’études ARTELIA et TRACTEBEL
sur la base de plusieurs barrages en BCR étudiés, en cours de construction ou déjà
en service ces dernières années.
Les avantages qui peuvent être tirés de la mise en œuvre d’un BCR riche en
liant et à bonne ouvrabilité sont illustrés avec la réduction significative des discon-
tinuités inter-couches et le faible retrait du béton. Par conséquent, il est possible
d’espacer les joints transversaux et de limiter les difficultés techniques et les faib-
lesses y afférentes, aux conditions d’assurer un contrôle qualité accru des travaux
de mise en place.
En outre, les moyens de production ainsi que les méthodes de transports et

de mise en œuvre du BCR jouent un rôle prépondérant dans la rapidité de la con-
struction et peuvent être ajustés pour atteindre les cadences requises pour assurer
∗ Feedback on the design and on the construction methods used for a few high RCC dams
270 DOI: 10.1201/9781003211808-17

Q.104 - R.17
les critères de résistance et d’étanchéité associés au projet. Il peut exister plusieurs

combinaisons efficaces pour atteindre ces critères. La meilleure est probablement
celle pour laquelle l’Entrepreneur a déjà plusieurs retours d’expérience positifs.
Enfin, les retours d’expériences positifs relatifs au déversement des crues sur
le corps des ouvrages en BCR permettent de compléter l’observation des aspects
constructifs avantageux de ce type d’ouvrage.
SUMMARY
Achieving acceptable strength parameters between layers is one of the main

challenges in the construction of RCC dams. This material and its implementation
therefore continue to evolve in order to facilitate the achievement of the criteria
specific to each project in an optimized manner. This paper describes and analyzes
the feedback of the Engineering Consultants ARTELIA and TRACTEBEL on the
basis of several RCC dams studied, under construction or already in service in
recent years.
The benefits that can be derived from the use of high-cementitious content
and good workability RCC are illustrated with the significant reduction in interlayer
cold joints and the low shrinkage of the concrete. Consequently, it is possible to
increase the spacing of transverse joints and to reduce the technical difficulties and
weaknesses related to them, provided that appropriate quality control is applied for
the placement activities.
In addition, the means of production as well as the methods of transportation

and implementation of RCC play a major role in the speed of construction and can
be adjusted to achieve the rates required to ensure the resistance and waterproofing
criteria associated with the project. There may be several effective combinations to
achieve these criteria. The best one is probably the one for which the Contractor
already has several positive feedbacks.
Finally, the positive feedbacks related to the discharge of floods on the body of
RCC structures allow to complete the observation of the advantageous constructive
aspects of this type of structure.
1. INTRODUCTION
Le Béton Compacté au Rouleau (BCR) est devenu le matériau de référence

pour la construction des barrages-poids de ces quatre dernières décennies. Il est
efficace économiquement grâce à sa mise en place rapide par rapport au béton
271
Q.104 - R.17
conventionnel vibré. Traditionnellement, le BCR est un béton maigre et relativement

sec mis en œuvre par couches horizontales continues d’une rive à l’autre du barrage.
L’obtention de paramètres de résistance acceptables entre les couches est

l’un des principaux enjeux de la construction des barrages en BCR. Ce matériau
et sa mise en œuvre continuent donc à connaître des évolutions éventuellement
nécessaires pour faciliter l’atteinte des critères propres à chaque projet de manière
optimisée.
Le présent article décrit et analyse le retour d’expérience des bureaux d’études

ARTELIA et TRACTEBEL sur la base de plusieurs barrages en BCR étudiés, en
cours de construction ou déjà en service ces dernières années. Dans une première
partie, les formulations et le zonage éventuel, la méthode de mise en place et le
traitement des joints inter-couches ainsi que le choix du type de coffrage seront
traités. Par la suite, les moyens et cadences de production et de transport du BCR
seront partagés.
Enfin, les retours d’expériences relatifs au déversement des crues sur le corps
de l’ouvrage mais également au balancement de la rivière dans un chenal central
seront également discutés.
2. FORMULATION ET ZONAGE DU BCR
Contrairement aux ouvrages en béton armé dont la formulation du béton est

encadrée par des normes associées, le Concepteur et l’Entrepreneur disposent
d’une certaine liberté pour la formulation d’un BCR. En particulier, la teneur en
ciment ou en liant est très variable, généralement entre 60 et 220 kg/m3 .
Deux critères entrent généralement en jeu dans le choix de cette formulation :

la résistance structurelle et l’étanchéité. Si la résistance structurelle est uniquement
assurée par le corps du barrage, l’étanchéité peut être assurée par une membrane
externe appliquée contre le parement amont.
Dans les trois exemples décrits dans les prochains paragraphes, l’étanchéité
et la résistance structurelles sont assurées par le matériau du corps du barrage.
Le Tableau 1 résume les principales caractéristiques de ces barrages.
Pour l’ensemble de ces trois barrages, le Concepteur et/ou l’Entrepreneur ont

fait le choix de mettre en œuvre un BCR riche en liant avec une bonne ouvrabilité.
La richesse en liants permet généralement d’atteindre une résistance à la traction
plus élevée à long terme. L’ouvrabilité permet un meilleur compactage mais égale-
ment un meilleur collage grâce à l’existence d’une pâte excédentaire qui remonte
à la surface à la fin du compactage. Dans le cas de joints chauds (couche mise en
272
Q.104 - R.17
Tableau 1
Caractéristiques des trois cas d’études.
Main features of the three case studies
JANNEH NAM NGUM 4 HATTA
(LIBAN) (LAOS) (DUBAï)
Volume total de BCR (hm3 ) 1.4 0.23 0.18
Type Poids-arqué Poids Poids
Hauteur (m) 157 74 70
Longueur en crête (m) 300 228 210
Début Février 2020 Mars 2019 Octobre 2021
Durée prévisionnelle de (mois) 27 19 10
placement effectif
œuvre avant le début de la prise de la couche sous-jacente), cette pâte excéden-

taire remplace le mortier de liaison utilisé dans le cas des BCR maigres et plutôt
secs. Pour augmenter l’ouvrabilité, la proportion de fines dans la formulation est
augmentée et un retardateur de prise est utilisé. Le BCR est alors plus visqueux, la
viscosité s’opposant ici à la fluidité qui est obtenue en augmentant la quantité d’eau.
L’ouvrabilité du BCR de ces trois barrages est caractérisée par une valeur
relativement faible à l’essai VeBe modifié, entre 15 et 18 s. L’observation visuelle
du BCR à bonne ouvrabilité placé et compacté est assez caractéristique. Elle met
en évidence la pâte excédentaire qui masque les granulats et laisse apparaître les
traces de passage des rouleaux compacteurs de manière marquée.
Le barrage de Janneh est un barrage-poids arqué pour des raisons sismiques.

La sismicité du site est en effet élevée (SES ∼ 0.6g). L’optimisation de sa géométrie
implique la mobilisation de l’effet d’arc même en situations statiques [1]. C’est prob-
ablement l’un des plus hauts de ce type. Grâce à cette forme arquée, les contraintes
sismiques sont faibles en amont. En aval, la résistance verticale à la traction néces-
saire serait de 1.1 MPa. Cependant, une fissuration depuis l’aval n’est pas jugée
critique dans la mesure où les contraintes sont localisées et le plot concerné ne
bascule pas vers l’amont.
Le Concepteur a choisi une formulation plus riche sur une zone de 10 m

d’épaisseur située à l’amont. Des formulations moins riches ont été testées avec
la planche d’essai du BCR mais sans succès : les résultats des essais sont
hétérogènes. En effet, le non-respect des critères d’élongation des granulats a
empêché une bonne remontée de la pâte excédentaire et a généré par conséquent
une mauvaise liaison entre les couches. Il a donc été nécessaire d’améliorer
l’ouvrabilité du BCR en augmentant le dosage en pâte pour faciliter cette remontée.
Le Tableau 2 donne la formulation du BCR des différents barrages décrits.
Par ailleurs, un extensomètre à longue base a été installée dans cette planche
d’essai et a permis de mesurer une valeur très faible de retrait endogène et/ou
273
Q.104 - R.17
Tableau 2
Formulation des BCR des cas d’études.
RCC mixes for the case studies
JANNEH NAM NGUM 4 HATTA
ZONE AMONT ZONE AVAL
Ciment kg/m3 77 56 82 50
Matériau kg/m3 143(*) 104(*) 100(**) 110(**)
pouzzolanique
Eau l/m3 128 132 91 115
Granulats fins kg/m3 952 978 811 732
Granulats grossiers kg/m3 1181 1212 1447 1445
Retardateur de prise l/m3 2.2 1.6 - 1
(*) Laitier de haut-fourneau, (**) Cendres volantes
Fig. 1
Vue plongeante sur le BCR en cours de compactage sur le barrage de Nam Ngum
4 (à g.) et vue rapprochée sur le BCR compacté du barrage de Janneh (à dr.).
Bird-eye view of the RCC during compaction for Nam Ngum 4 dam (left) and close
view of the compacted RCC for Janneh dam (right).
éventuellement de fluage telle que montrée sur la Fig. 2. Les effets de la dilata-
tion thermique propre de la sonde sont retranchés des valeurs de déformation
mécanique sur cette figure. La valeur mesurée à 10 jours (∼10 µ) est très inférieure
aux valeurs habituellement rencontrées pour un béton conventionnel vibré ou un
BCR peu ouvrable (plusieurs centaines de microns). L’utilisation de laitier de haut
fourneau peut montrer un comportement contraire [2], mais ce n’est vraisemblable-
ment pas le cas ici : la bonne ouvrabilité du BCR semble permettre une réduction
significative des retraits. Si cette tendance est confirmée pendant la construction du
barrage, les joints de clavage constitueront en réalité un dispositif de correction à
utiliser en cas de fuites [3]. Il s’agit là d’une différence significative avec le cas des
barrages-voûtes en béton conventionnel vibré.
Le barrage de Nam Ngum 4 est construit pendant une durée de deux ans mal-
gré ses dimensions modestes pour éviter le placement du BCR pendant la saison
274
Q.104 - R.17
Fig. 2
Retrait au jeune âge et températures mesurés sur la planche d’essai du barrage
de Janneh.
Measured early-age shrinkage and temperature in the Full-Scale Trial RCC of
Janneh dam.
des pluies. Les calculs thermo-mécaniques réalisés par l’Entreprise considèrent un

barrage monolithique. En particulier, ils supposent l’absence de tout autre retrait
ou fluage hormis le retrait thermique dû à la différence entre la température de
placement du BCR et la température d’équilibre à long terme du barrage. Le site de
construction du barrage bénéficie d’un climat tempéré avec moins de 9˚ C d’écart
moyen saisonnier de température. La résistance à la traction horizontale dimen-
sionne donc le BCR du barrage, de l’ordre de 1.8 MPa. ARTELIA, assistant au Maître
d’Ouvrage, ne s’y est pas opposé compte tenu de retours d’expérience récents sur
les BCR riches en cendres volantes [1][4]. Par prudence, l’Entrepreneur a néan-
moins mis en œuvre des joints entre plots espacés d’environ 24 m dans le cas
du barrage de Nam Ngum 4. Les carottes de sondages réalisés dans la planche
d’essai montrent la structure squelettique du BCR comparée à celle du BCR enrichi
au coulis des bords. Pour le BCR, les granulats sont très rapprochés et sont beau-
coup plus en contact entre eux que ceux du BCR enrichis au coulis (Fig. 7). Cette
organisation squelettique est identifiée comme un facteur diminuant le retrait et le
fluage pour les BCR à bonne ouvrabilité [2].
Le barrage de Hatta sera construit dans un site à sismicité modérée (0.23 g)

avec un climat semi-désertique. Les calculs sismiques dimensionnent le BCR avec
une résistance à la traction inter-couches de 0.5 MPa. Ce barrage fait partie d’un
aménagement de pompage-turbinage pour lequel les critères d’étanchéité sont très
restrictifs. C’est la raison pour laquelle une formulation de béton riche en liants et
ouvrable a également été choisie pour bénéficier d’une pâte excédentaire dans le
BCR et assurer un très bon collage entre les couches dans le cas de joints chauds.
Dans ce sens, la formulation initialement prévue sans cendres volantes et incluant
275
Q.104 - R.17
Tableau 3
Classement des joints en fonction de la durée d’exposition – Barrage de Nam
Ngum 4.
Joint type classification with regards to exposure duration – Nam Ngum 4 dam
MOIS JOINT TIÈDE JOINT FROID
Novembre - Février 10h < t ≤ 24h >24h
Mars - Octobre 8h < t ≤ 16h >16h
Avril - Septembre 6h < t ≤ 12h >12h
une zone amont de BCR avec un traitement systématique des joints de levée a
été remplacée par une zone unique sans traitement systématique. Le critère de
résistance à la traction un peu plus faible que les deux autres barrages justifie ici un
dosage en ciment réduit. Néanmoins, une résistance à la compression de l’ordre de
30 MPa est probable à long terme avec cette formulation, compte tenu de la nature
des granulats. Les études et essais au laboratoire permettant de réduire le nombre
de joints du barrage sont actuellement en cours.
3. MÉTHODE DE MISE EN PLACE ET TRAITEMENT DES JOINTS

INTER-COUCHES
3.1. TRAITEMENT DES JOINTS HORIZONTAUX
Les procédés de construction des trois cas d’étude distinguent trois types de
joint de levée : chaud, tiède et froid. La distinction est similaire pour les barrages
de Janneh et de Hatta. Elle est basée sur le démarrage de la prise du BCR pour
les joints chauds et ensuite sur l’indice de maturité modifié pour les joints tièdes et
froids. Cet indice de maturité modifié (IMM) s’exprime comme suit :
IMM = Temps d exposition × (température moyenne d exposition + 12◦ C) (1)
Le démarrage de la prise dépend du dosage en retardateur de prise et peut

atteindre 24 heures lorsque le dosage en retardateur est élevé. Un joint est considéré
tiède lorsque IMM < 1000 et froid lorsque IMM ≥ 1000.
La distinction est plus empirique dans le cas du barrage de Nam Ngum 4. Un

joint est jugé chaud si sa durée d’exposition est inférieure à la durée de démarrage
de la prise retranchée d’une heure. Le joint est jugé tiède ou froid en fonction de la
saison, et en particulier de la température moyenne, tel que décrit sur le Tableau 3.
La méthode de traitement est identique pour les trois barrages (Fig. 3) :
276
Q.104 - R.17
Fig. 3
Traitement des joints tièdes : barrages de Nam Ngum 4 (à g.) et Janneh (à dr.).
Warm joint treatment: Nam Ngum 4 (left) and Janneh (right) dams.
• Nettoyage au camion aspirateur et à l’air basse pression des matériaux non

adhérents, de l’eau stagnante et toute autre matière étrangère sans détério-
ration de la surface du BCR avant coulage de la prochaine couche pour les
joints chauds ;
• En plus du traitement pour les joints chauds, nettoyage à l’air sous pression
et au camion aspirateur pour retirer le matériau superficiel contaminé sans
exposer les granulats puis mise en œuvre d’une couche fine (< 1 cm) de
coulis/mortier sur toute la surface avant mise en place immédiate de la couche
suivante pour les joints tièdes ;
• Décapage à la brosse métallique pour retirer le mortier de surface et exposer
les granulats puis nettoyage à l’eau sous pression (uniquement à la fin de la
prise), mise en œuvre d’une couche fine (< 1 cm) de coulis/mortier sur toute la
surface avant mise en place immédiate de la couche suivante pour les joints
froids.
Pour l’ensemble des joints, la surface doit être maintenue humide à l’aide de
brumisateurs.
3.2. MÉTHODE DE PLACEMENT
La mise en œuvre traditionnelle du BCR est le placement par couches horizon-

tales successives et continues d’une rive à l’autre de la vallée. C’est une méthode
simple à mettre en œuvre qui, une fois bien rodée, se suffit à un contrôle qualité
simplifié. En revanche, elle est assez peu flexible en cas d’existence d’ouvrages
annexes inclus dans le barrage. Ceux-ci doivent souvent être placés soit sur les
bords des ouvrages, générant souvent une surexcavation, soit au fond de la vallée
fondés sur un massif en béton conventionnel. Par ailleurs, lorsque la vallée est très
large, le traitement des joints inter-couches devient systématique puisque la prise
de la couche sous-jacente a souvent déjà démarré lorsque la couche supérieure
est mise en place. Le joint de levée devient alors tiède.
277
Q.104 - R.17
Fig. 4
Illustration de la méthode de placement horizontal non continue.
Illustration of the non-continuous horizontal placement method.
Une alternative qui se présente à cette méthode traditionnelle est la méthode

de placement non continue. Elle peut être horizontale (Fig. 4) ou inclinée. Le lecteur
pourra se référer à [5] pour les descriptions détaillées de ces méthodes. Elles ont
pour but d’assurer un meilleur collage entre les couches de BCR par rapport à
la méthode traditionnelle en plaçant une couche supérieure avant le démarrage
de la prise de la couche sous-jacente. La méthode inclinée a l’avantage d’avoir
essentiellement des joints chauds entre les levées mais inclut un grand nombre
de coins inférieurs où il y a risque d’écrasement des granulats au compactage. La
méthode horizontale a l’avantage de n’avoir qu’un nombre limité de coins inférieurs
mais implique un traitement de joint tiède pour chacune de ces couches inclinées.
Enfin, il est possible d’éviter les plans inclinés avec la méthode horizontale, au prix de
coffrage verticaux. L’utilisation de ces méthodes non continues permet d’envisager
l’existence de quelques ouvrages traversants dans le BCR sans que cela n’impacte
significativement la cadence de construction.
La construction du barrage de Janneh utilise les trois méthodes de place-

ment (Fig. 5) précédemment décrites. La méthode horizontale traditionnelle est
appliquée pour les 22 m inférieurs. La surface de chaque couche permet d’envisager
l’existence systématique de joints chauds dans cette partie, hors arrêt inopiné de
placement. La méthode inclinée est prévue pour les 40 m suivants avec une pente
de 15H/1V et une hauteur de levée de 2.4 m puisque la surface de levée aug-
mente. Les injections de consolidation seront réalisées sur la rive opposée à celle
de démarrage des couches inclinées. Pour les 75 m suivants, la méthode horizon-
tale non-continue est prévue. En effet, il existe deux joints injectables et coffrés dans
le barrage. Par ailleurs, le plot de rive droite sera construit en retard du fait de la
présence d’ouvrages annexes. Les deux plots de rive gauche seront construits avec
un décalage vertical de 1.20 m. Enfin, la crête du barrage sera construite avec la
méthode horizontale traditionnelle.
278
Q.104 - R.17
Fig. 5
Méthodes de placement appliquées dans le cas du barrage de Janneh.
Placement methods applied for Janneh dam.
Fig. 6
Photo du barrage de Nam Ngum 4 en cours de construction (Juillet 2020).
Picture of Nam Ngum 4 dam under construction (July 2020).
La construction du barrage de Nam Ngum 4 ne se situe pas sur le chemin

critique du planning de la construction de l’aménagement hydroélectrique. Par
ailleurs, la dérivation provisoire s’effectue par balancement de la rivière, en partic-
ulier pour faire passer la crue lors de la saison des pluies (Fig. 6). Par conséquent,
la construction du barrage se fait par plots coffrés.
Les surfaces des couches étant réduites, les joints de reprise sont essentielle-
ment chauds, sauf au passage des galeries. Les carottes de sondages réalisés sur
la planche d’essai (27 mètres linéaires au total) mettent en évidence 18 fractures
dont une seul au droit d’un joint de levée (Fig. 7).
Dans le cas du barrage de Hatta, la méthode horizontale non continue est

également choisie, incluant un joint incliné délimitant une surface maximale de
couche d’environ 1200 m2 . Le choix a été dicté par la bonne expérience de
l’Entrepreneur dans l’application de cette méthode mais également par le bon retour
d’expérience en termes d’étanchéité obtenue lorsqu’elle a été appliquée [6].
279
Q.104 - R.17
Fig. 7
Carottes de sondages issus de la planche d’essai du barrage de Nam Ngum 4 :
BCR enrichi au coulis (en haut) et BCR (en bas).
Cores from the Full-Scale Trial of Nam Ngum 4 dam: GERCC (top) and RCC
(bottom).
Fig. 8
Schéma du principe de mise en place du BCR par couches inclinées dans deux
directions, en fonction du niveau de la levée de bétonnage.
Schematic principle of RCC placing with sloped layers in two directions, in
accordance with the dam elevation.
Dans le cas des couches inclinées, bien que la pente de la couche puisse être
théoriquement modifiée aux différentes levées du barrage pour ajuster le volume de
béton à mettre en place, l’utilisation satisfaisante d’une inclinaison quasi systéma-
tique des couches de 10% a été observée au barrage de Souapiti, présenté dans
le Tableau 4.
En considérant une levée coffrée avec des panneaux verticaux de trois mètres,
la longueur de la couche inclinée est de trente mètres. Cette dimension permet de
guider la méthode de compactage de l’ouvrage, en considérant un compactage des
couches i) en partie basse du barrage, dans le sens aval-amont sur une largeur de
l’ordre de deux plots ; ii) en partie haute du barrage, dans le sens rive à rive lorsque
280
Q.104 - R.17
la levée se situe à environ cinquante mètres sous le niveau de la crête. La figure Fig.
8 illustre ce principe. En considérant un temps de prise, généralement conservatif,
de deux heures de la couche sous-jacente, le débit horaire requis pour la mise en
place du BCR est estimé à 200 m3 /h. Il apparaît bien calibré au regard des moyens
observés pour garantir la réalisation de joints chauds entre couches.
Enfin, une aide précieuse à la mise en place du BCR peut être apportée par
l’utilisation accrue des moyens de détection météorologique, en particulier les radars
de pluie qui fournissent une estimation du temps d’arrivée et de l’intensité de la pluie
au site. Pour les deux barrages de Gibe 3 et GERDP construit en Ethiopie, se référer
au Tableau 4, un radar de pluie a permis à l’Entrepreneur de maximiser les temps
de placement du BCR en conditions météorologiques incertaines et d’organiser au
mieux les opérations de fermeture de la couche et de redémarrage.
4. TYPE DE COFFRAGE
Un coffrage métallique est utilisé dans le cas des parements des barrages de
Nam Ngum 4 et de Janneh. Il s’agit d’un coffrage grimpant incluant 1 à 3 modules
de 2.4 m de haut chacun, soit jusqu’à 24 couches de BCR. Dans le cas du barrage
de Nam Ngum 4, pour les conditions de passage de la crue de la saison des pluies,
un chenal central a été aménagé. Par conséquent, trois faces du BCR sont coffrées.
Dans le cas du barrage de Janneh, les deux joints coffrés et injectables sont
construits avec un décalage vertical d’environ 1.20 m entre deux plots avec des
rampes d’accès reliant les deux plots. Les joints sont munis de boîtes de cisaillement
et de rampes d’injection étanches (Fig. 9). Aucune boîte de cisaillement n’est prévue
dans les joints transversaux du barrage de Nam Ngum 4 puisque chaque plot est
stable individuellement.
Fig. 9
Panneaux de coffrage du barrage de Nam Ngum 4 (haut) et principe des boîtes de
cisaillement des joints transversaux injectables du barrage de Janneh (bas).
Formwork layout of Nam Ngum 4 dam (top) and shear keys of the groutable
vertical transverse joints of Janneh dam (bottom).
281
Q.104 - R.17
Un choix différent a été réalisé par l’Entrepreneur du barrage de Hatta. Des

panneaux préfabriqués auto-stables en béton armé serviront de coffrage perdu
pour les parements amont et aval (Fig. 10). Avec une longueur de 3 m et une
hauteur de 1.20 m, ces panneaux sont conçus comme des ouvrages permanents
avec les critères de durabilité habituels relatifs à ces ouvrages (enrobage, classe
de béton, dispositions minimales d’armatures, …). Ce choix est guidé par un
retour d’expérience satisfaisant sur plusieurs autres barrages déjà construits par
l’Entrepreneur. Par ailleurs, il permet un contrôle qualité un peu moins strict pour
la mise en œuvre du BCR enrichi proche des parements du barrage. En effet, il
n’y a plus d’obligation de qualité de finition pour cette couche de BCR enrichi au
coulis dans ce cas. En revanche, l’utilisation de ce type de coffrage ne permet pas
la mise en œuvre de BCR à forte ouvrabilité. En effet, la verticalité des panneaux
de coffrage peut être difficile à respecter à cause d’un BCR dont la prise est longue.
Il s’agit là d’un point d’attention nécessitant un réglage fin de la formulation du BCR
dès la planche d’essai de ce barrage. Par ailleurs, il est supposé par avance que la
zone concernée par la base des panneaux préfabriqués n’est pas étanche. La par-
tie imperméable démarre immédiatement à l’aval de la base des panneaux et inclut
aussi bien le BCR enrichi au coulis que le BCR dont la formulation riche permet une
très bonne étanchéité.
Fig. 10
Vue rapprochée de la crête du barrage de Hatta (à g.) et illustration en perspective
des panneaux préfabriqués de coffrage (à dr.).
Close view of the crest of Hatta dam (left) and perspective illustration view of the
precast formwork (right).
5. PRODUCTION DU BETON ET CADENCES DE MISE EN PLACE DU BCR
Les divers retours d’expérience des moyens de production et des cadences

de mise en œuvre du BCR vont être illustrés en considérant essentiellement quatre
282
Q.104 - R.17
barrages, réalisés ces dix dernières années pour trois d’entre eux, et qui sortent
de l’ordinaire par leur fort volume combiné avec une hauteur importante. Leurs
caractéristiques sont récapitulées dans le Tableau 4. Les auteurs ont été ou sont
toujours directement impliqués dans la supervision de leur construction.
La figure Fig. 11 ci-après illustre les cadences observées avec l’échantillon

de barrages présenté plus haut.
Fig. 11
Cadences mensuelles observées de mise en place du BCR.
RCC monthly placement rates recorded.
Bien que les moyens employés soient divers d’un projet à l’autre, il n’est prob-
ablement pas surprenant de constater une augmentation progressive des cadences
de mise en œuvre du BCR, à mesure que le volume total du barrage croît. De multi-
ples facteurs viennent influer sur les cadences d’exécution, mais une taille d’ouvrage
importante offre certainement plus d’opportunités pour pallier aux aléas climatiques
et aux aléas de production, grâce à son vaste espace de travail disponible plus
longtemps.
En amont de la chaîne de production, la centrale à béton est dimensionnée

pour garantir la production des bétons sur une durée longue et continue, générale-
ment de l’ordre de 18h à 20h de fonctionnement par jour, avec la robustesse et la
redondance nécessaire pour produire régulièrement, généralement de l’ordre de
283
Q.104 - R.17
Tableau 4
Extrait des caractéristiques de la construction de quelques barrages BCR de
grande taille.
Sample construction features of some large RCC dams
SOUAPITI THA DAN GIBE 3 GERDP
(GUINÉE) (THAILANDE) (ETHIOPIE) (ETHIOPIE)
Volume total de BCR (hm3 ) 3,3 5,0 6,2 10,1
Volume BCR considéré (hm3 ) 3,0 (en construc- 3,6 (barrage -B) 6,2 7,6 (en con-
tion) struction)
Hauteur (m) 115 92 250 170
Longueur en crête (m) 1 200 2 720 610 1 790
Début Avril 2017 Mars 2001 Jan. 2012 Déc. 2013
Durée (mois) 33 27 46 42
Centrale à béton (m3 /h) 540 1 000 800 1 120
Centrale de (t/h) 1 700 1 700 1 200 1 200
concassage
Type de centrale à Malaxage par Malaxage con- Malaxage par Malaxage par
béton gâchée tinu gâchée gâchée
Type d’acheminement Camion Camion Bande trans- Bande
porteuse transporteuse
Type de mise en place Couches Couches hori- Couches hori- Couches hori-
inclinées zontales zontales zontales
Cadence maximale (m3 /mois) 180 000 200 000 250 000 330 000
mensuelle
Cadence moyenne (m3 /mois) 90 000 135 000 135 000 180 000
mensuelle
25 à 28 jours par mois. Pour faire face à l’exigence de la rapidité d’exécution, il

est important que la centrale ne constitue pas le maillon faible de la chaîne des
équipements de mise en œuvre des bétons.
En figure Fig.12 sont reportées les cadences mensuelles maximales réalisées

sur sites et les capacités de production nominale des centrales. Les cadences de
mise en place maximales représentent de l’ordre de 60% à 70% de la capacité de
production nominale du béton. Ce rendement global observé ne semble pas incrim-
iner une méthode particulière, compte tenu de la diversité des moyens observés.
Dans le cas du barrage de Tha Dan, bien que ce rendement soit plus faible (0,4)
et que la centrale à béton soit de type continue, contrairement aux autres projets,
le concassage des sables s’est trouvé être un facteur limitant la production des
bétons.
On relève, dans le cas de l’utilisation des couches inclinées à Souapiti, que le

rendement global de 0,7, supérieur aux autres projets, tend à confirmer la réputation
d’efficacité de cette méthode.
284
Q.104 - R.17
Fig. 12
Comparaison entre cadences maximales observées de mise en place du BCR et
capacités de production nominale du béton.
Comparison between RCC max placement rates recorded and batching plant
capacities.
Fig. 13
Vues illustrant l’utilisation d’une bande transporteuse dans le cas d’une vallée
étroite (à dr., Gibe 3) ou d’une vallée large (à g., GERDP).
Views picturing usage of conveyor belts in case of narrow valley (on the right, Gibe
3) and large valley (on the left, GERDP).
285
Q.104 - R.17
Fig. 14
Vues illustrant l’utilisation de goulottes (à g., Gibe 3) (à dr., barrage de Bui au
Ghana).
Views picturing usage of conveyance pipes (on the left, Gibe 3) (on the right, Bui
dam in Ghana).
6. TRANSPORT DU BETON
Le choix du mode de transport entre la centrale de production du béton et la

couche à compacter ne révèle pas de tendance vers un mode particulier. Il s’agit
fondamentalement d’un choix des Entrepreneurs.
Dans l’échantillon du Tableau 4, on observe que les deux modes de transport

sont utilisés, camion ou bande transporteuse, pour les trois barrages construit en
vallée large.
Dans le cas du site en vallée étroite (Gibe 3), l’acheminement du béton par
bande transporteuse a été retenu par l’Entrepreneur. Cette option a permis de
s’affranchir de forte difficulté d’accès liée au transport par camion. Elle a consti-
tué toutefois un élément délicat à mettre en œuvre, qui peut représenter un élément
limitant pour la cadence de mise en œuvre, tout du moins en partie inférieure de
l’ouvrage, lorsque la longueur totale de la bande demeurait importante.
Lorsque les rives sont très raides ou lorsqu’une marche d’un dénivelé impor-
tant est créée sur la levée du barrage, induite par exemple par le bétonnage des
ouvrages de restitution dans le cas de Gibe 3, l’utilisation satisfaisante de goulottes
a été observée à plusieurs reprises. On ajoute à cet effet l’exemple de la construc-
tion du barrage de Bui au Ghana (2010, 1 hm3 ), construit au moyen d’une goulotte
en vallée étroite.
286
Q.104 - R.17
Tableau 5
Observation de déversements en phase de construction de quelques barrages en
BCR.
Main features of spilled flow during construction of some RCC dams
THEUN HINBOUN (LAOS) BUI (GHANA) GERDP
(ETHIOPIE)
Hauteur de la lame d’eau (m) 5 5 12
Débit maximal (m3 /s) 2 500 1 100 6 500
Hauteur de chute (m) 15 35 30
Durée totale du déversement (mois) 0.3 2.5 3.5
Fig. 15
Vues du déversement sur le barrage de Theun Hinboun (à g.) et GERDP (à dr.).
Water overtopping RCC dams under construction, Theun Hinboun (left) and
GERDP (right).
7. CONTROLE DES CRUES EN PHASE DE CONSTRUCTION
Les barrages en BCR permettent le déversement en phase provisoire des

crues de chantier. Cette méthode est particulièrement adaptée pour les sites qui
présentent un fort contraste de débit entre les saisons sèches et humides.
Le déversement en phase de construction a été effectif sur les barrages

de GERDP et Bui, précédemment cités, ainsi qu’à Theun Hinboun au Laos. Les
conditions observées lors des déversements sont regroupées dans le Tableau 5
ci-dessous.
Les dispositions adoptées pour ces aménagements ont permis un déverse-

ment sur l’ouvrage en cours de construction dans des conditions satisfaisantes,
les dégâts dû à l’érosion sur les parements BCR sont restés minimes, ainsi que
ceux observés au pied aval des ouvrages. Les redémarrages du bétonnage sur les
surfaces de déversement n’ont pas posé de problème particulier.
287
Q.104 - R.17
De manière générale, des dispositions de protection des joints verticaux

entre plots doivent être adoptées, éventuellement des protections dans la zone
de restitution.
8. CONCLUSIONS
Un retour d’expérience sur la conception et la méthode de construction de

plusieurs BCR de grande hauteur en cours d’étude, de construction ou déjà con-
struits et suivis par deux bureaux d’études français est proposé dans cet article.
Les avantages qui peuvent être tirés de la mise en œuvre d’un BCR riche en liant
et à bonne ouvrabilité sont discutés. Lorsque combinée à une méthode de mise en
œuvre non continue d’une rive à l’autre, ce type de BCR permet d’atteindre des
critères de résistance et d’étanchéité proches voire plus exigeants qu’un barrage
en béton conventionnel vibré. En effet, cette approche permet de réduire significa-
tivement les discontinuités inter-couches et le retrait du béton. Par conséquent, il
est possible d’espacer les joints transversaux et de limiter les difficultés techniques
et les faiblesses y afférentes. Par ailleurs, il devient envisageable de considérer
quelques ouvrages traversants sans que cela n’impacte significativement le plan-
ning. Cette méthode est probablement plus intéressante lorsque la construction du
barrage n’est pas sur le chemin critique ou si des critères très exigeants s’appliquent
au projet. En revanche, un contrôle qualité accru doit être mis en œuvre puisqu’en
l’absence de mortier de liaison entre couches, il est primordial de s’assurer de
la bonne classification des joints et de la bonne mise en œuvre de leur prépara-
tion de manière continue. Par ailleurs, l’utilisation de coffrages auto-stables peut
nécessiter une adaptation de la formulation du BCR riche et ouvrable. Plusieurs
méthodes de mise en œuvre de BCR sont proposées et décrites dans cet article.
Les moyens, cadences et méthodes de production ainsi que les méthodes de trans-
ports sont analysées et comparées. En tout état de cause, ces facteurs doivent
être ajustés pour atteindre les critères de résistance et d’étanchéité associés au
projet. Il peut exister plusieurs combinaisons efficaces pour atteindre ces critères.
La meilleure est probablement celle pour laquelle l’Entrepreneur a déjà plusieurs
retours d’expérience positifs.
REFERENCES
[1] Andrian. F., Agresti P. and Mathieu G., Theme B Synthesis Report: Static
and seismic analysis of an arch-gravity dam, 14th International Benchmark
Workshop on Numerical Analysis of Dams, Sweden, September 2017.
288
Q.104 - R.17
[2] Shaw Q., The influence of early stress-relaxation creep on the design of large
RCC dams, 85th Annual Meeting of International Commission on Large Dams,
July 3–7, 2017, Prague.
[3] Si-Chaib A., Roy M., Boinay V., Girard J.-C., Ulrich N. and Andrian. F., Design
and Feedback on thick RCC arched dams, Q104, ICOLD CIGB Marseille,
2021.
[4] ICOLD-CIGB, Bulletin 117, Roller-Compacted Concrete Dams, 2019.
[5] Forbes B. A., RCC – New developments and innovations, 1st Brazilian RCC
Symposium, September 2008.
[6] Shaw Q., Quality Control Requirements for RCC Dams, DSI Quality Control
Symposium, Octobre 2017.
289
Q.104 - R.18
-------
-------
LA MODÉLISATION DU GONFLEMENT DES BÉTONS : UNE AIDE

PRÉCIEUSE POUR LA GESTION OU LA RÉHABILITATION D’UN OUVRAGE(∗ )
Etienne GRIMAL & Philippe KOLMAYER

EDF CIH
Katia LALICHE & Catherine CASTEIGTS

SCP
Romain GIUNTI
Demathieu Bard
Christine NORET
Tractebel Engineering, France
FRANCE
RÉSUMÉ
Dans le cadre des évolutions réglementaires, EDF CIH a engagé depuis plus
de 20 ans des recherches sur les pathologies des bétons. Ces recherches ont abouti
à la création d’un modèle numérique permettant d’évaluer le comportement des
structures atteintes. Ce modèle utilisé aujourd’hui de manière industrielle fait partie
intégrante de l’exploitation durable et sure d’un ouvrage présentant un gonflement.
Il permet de requalifier l’ouvrage mais également de concevoir de manière efficace
les programmes de maintenance du parc d’EDF Hydro.
De son côté, afin d’anticiper les futurs besoins en eau de la Région, la

Société du Canal de Provence a engagé une rénovation d’envergure du barrage
de Bimont, affecté par un phénomène de gonflement du béton. Le projet de rénova-
tion a été élaboré à partir des résultats d’une modélisation numérique 3D réalisée
∗ Modelling of concrete swelling: a great help for the operation or rehabilitation of a structure
290 DOI: 10.1201/9781003211808-18

Q.104 - R.18
par TRACTEBEL, et de deux campagnes de reconnaissances. De 2017 à 2019,

d’importants travaux de rénovation ont été engagés pour permettre la rehausse du
plan d’eau de 14 m, pour un gain de volume stocké d’environ 13 millions de m3 .
Après la vidange de la retenue imposée par les travaux, la remontée du plan

d’eau a débuté au printemps 2019. Elle s’effectue par paliers, avec un suivi attentif
du comportement du barrage tout au long du remplissage.
SUMMARY
In the context of regulatory changes, EDF CIH has been conducting research
on concrete pathologies for more than 20 years. This research has led to the devel-
opment of a numerical model for assessing the behaviour of affected structures.
This model, which is now used industrially, is an integral part of the sustainable and
safe operation of a structure with swelling. It enables the structure to be re-qualified,
but also to efficiently design the maintenance programs for EDF Hydro’s facilities.
For its part, in order to anticipate the regional future water needs, the Société
du Canal de Provence has undertaken a major renovation of the Bimont dam, which
experienced a phenomenon of swelling of the concrete. The renovation project
was developed based on the results of a 3D numerical modelling carried out by
TRACTEBEL and two investigations campaigns. From 2017 to 2019, major renova-
tion work was undertaken to allow the water level to be raised by 14 m, for a gain in
stored volume of approximately 13 million m3 .
After the reservoir was emptied because of the work, the impounding of the
reservoir began in the spring of 2019. It is being carried out in steps, with careful
monitoring of the dam’s behavior throughout the impoundment process.
1. INTRODUCTION
1.1. CONTEXTE
Près d’une structure hydraulique sur trois dans le monde est atteinte de
pathologies de gonflement interne du béton.
Pour les barrages, cela crée des endommagements qui peuvent compromettre
la stabilité globale de l’ouvrage. Pour les usines, cela peut entraîner des désordres
divers : désalignement des turbines, ovalisation des groupes, blocage des organes
de vantellerie…
Pour autant, il n’est pas toujours indispensable de conforter les ouvrages

atteints.
291
Q.104 - R.18
Les modèles numériques capables de décrire précisément l’état interne du

béton ainsi que son évolution peuvent être un outil efficace afin de continuer
l’exploitation en toute sureté des ouvrages. Malgré cela, dans certain cas, des con-
fortements s’avèrent nécessaires. Les modèles permettent alors de dimensionner
les projets de restauration les plus pertinents et optimiser le coût des travaux.
Dans ce contexte une méthodologie de gestion globale des ouvrages, basée

sur l’auscultation, les essais en laboratoire et la modélisation numérique a été
développée par l’ingénierie française.
1.2. MÉTHODOLOGIE GLOBALE
Cette méthodologie permettant d’assurer l’exploitation d’un ouvrage atteint

de gonflement dans de bonnes conditions de sureté et se base sur quatre phases
successives.
Les premiers signes de gonflement sont souvent détectés quelques années

après le début d’exploitation par une fissuration apparu dans un coursier ou un
bajoyer ou des déformations irréversibles (exhaussement de la crête par exem-
ple). Les mouvements sont ensuite auscultés afin d’en connaitre leur évolution puis
l’origine des déplacements irréversibles est imputée à un gonflement de l’ouvrage
en étudiant des prélèvements de béton en laboratoire.
Une seconde phase, dans la vie de l’aménagement, débute alors avec des
questionnements sur l’impact du gonflement sur la structure atteinte. Y aura-t-il une
évolution des sous pressions ? L’évacuateur de crue restera-t-il fonctionnel ? La
fissuration qui se développe a-t-elle un impact sur le comportement de l’ouvrage ?
Afin de répondre à ces interrogations, une troisième phase commence. Une

phase dans laquelle l’ouvrage est étudiée à l’aide de modèles numériques permet-
tant de reproduire le comportement du barrage. Les modélisations permettent à ce
stade d’identifier des mécanismes de défaillances, d’estimer un niveau de sureté.
Cette étape débouche sur une dernière étape permettant une gestion de
l’ouvrage à long terme, dans laquelle la nécessité de confortement doit être étudiée.
Dans le maintien en exploitation d’un ouvrage, la modélisation numérique

prend donc une place prépondérante.
2. L’APPORT DE LA MODELISATION DU GONFLEMENT DANS LA

GESTION D’UN OUVRAGE
La modélisation permet d’aider à la gestion d’un ouvrage en améliorant la

compréhension du comportement du barrage en vue d’en assurer l’exploitation en
toute sureté et ce sur le long terme.
292
Q.104 - R.18
Plusieurs types de modélisations sont mis en œuvre selon la complexité du

phénomène et l’importance de l’ouvrage. Elles peuvent être scindées en deux
parties, les modélisations par analogie thermique et les modélisations à l’aide de
modèle mécanique couplé avec les facteurs influençant le gonflement.
Les premiers sont généralement utilisés pour des ouvrages présentant des
gonflements très faibles car ils ont tendance à surestimer les contraintes. Une loi
d’atténuation du gonflement en fonction de la contrainte est alors utilisée pour éviter
ce phénomène. Les seconds reproduisent mieux la physique du phénomène mais
demandent des investigations in situ et une bonne connaissance des bétons. Ils
sont de ce fait utilisés pour des ouvrages présentant de forts gonflements.
EDF et le Laboratoire Matériaux et Durabilité des Constructions (LMDC)

de Toulouse ont développé un modèle numérique intégré dans le code aux élé-
ments finis Code_Aster, afin de simuler le gonflement interne dû à la réaction
alcali-granulats RAG [1] ou la réaction sulfatique interne [2].
Le modèle développé est basé sur un schéma rhéologique global qui consid-
ère le gonflement, le fluage et l’endommagement anisotropes de manière couplée.
Le modèle a été comparé à des expériences, pour vérifier sa robustesse [3][4]. Il
est désormais utilisé pour analyser le comportement des structures à long terme.
2.1. MODÈLE RHÉOLOGIQUE EDF/LMDC
Les réactions chimiques qui caractérisent les pathologies de gonflement

entraînent la création d’un volume de produit gonflant. Une fois le volume autour du
site de réaction comblé, une pression intraporeuse vient mettre en traction la pâte.
Cette mise en pression est maintenue dans le temps. L’évolution du gonflement met
alors la structure dans un état de contrainte pouvant entrainer la fissuration des
ouvrages. Il apparait donc qu’afin de bien reproduire le comportement de l’ouvrage,
les modélisations du fluage, de la fissuration et de l’évolution de la réaction chimique
sont nécessaires.
Le modèle poromécanique utilisé ici permet de prendre en compte la pression

de gel issue des réactions de gonflement interne, réaction d’alcali-granulats (RAG)
ou réaction sulfatique interne (RSI), en tenant compte des phénomènes de fluage
et d’endommagement. Dans un contexte poromécanique, la contrainte totale est
divisée en deux parties, la partie qui est appliquée sur le squelette solide (σij ’) et la

partie agissant sur la phase interstitielle ( bi Pi ) pondéré par un coefficient de Biot:

σij = σij − bi Pi (1)
La Fig. 1 présente le schéma rhéologique du modèle.
293
Q.104 - R.18
Fig. 1
Schéma rhéologique du modèle poro-mécanique.
Poromechanical rheological scheme.
2.2. MODÉLISATION DES RÉACTIONS DE GONFLEMENT INTERNE
Dans cet article, seule la partie concernant la réaction sulfatique interne est
décrite. Ce choix est porté par l’application sur l’ouvrage de Bimont. Une description
pour l’alcali réaction est disponible dans [3] et [4]. La formation d’ettringite différée
ou réaction sulfatique interne est induite par un ensemble de réactions chimiques
qui se produisent au cours de deux périodes de conditions thermiques différentes :
- une phase chaude durant la phase d’hydratation ou de cure où la tempéra-

ture dépasse une température seuil TSeuil,d pour permettre la dissolution des
monosulfoaluminates et de l’ettringite primaire
- une phase de refroidissement durant la vie de la structure, pendant laquelle le
produit de réaction (ettringite secondaire ou différée) va se former à partir des
ions aluminates et sulfates provenant des ions dissous lors de la première
phase.
Les lois gérant les réactions chimiques (création du volume d’ettringite) ne

sont pas décrites dans ce document mais peuvent être retrouvées dans [2].
Les grands principes de la modélisation chimique sont présentés par le

diagramme ci-dessous.
L’évolution du volume de produit expansif formé sur un intervalle de temps

dt est calculé en multipliant la variation de quantité d’ettringite différée calculée
294
Q.104 - R.18
Fig. 2
Principes de la modélisation chimique.
Chemical modeling principles.
précédemment par son volume molaire VAFT = 715 cm3 /mol [2].
∂∅g ∂E2
= VAFT (2)
∂t ∂t
Le produit expansif applique une pression interne sur la matrice solide et sert
de donnée d’entrée au modèle poromécanique.
En effet, la pression intraporeuse (Pg , équation 2) est estimée à partir du vol-

ume créé par la précipitation d’ettringite différée (φg ). Le produit expansif applique
une pression interne sur la matrice solide. Elle dépend principalement de la porosité
accessible (φgv ). Cette porosité étant initialement présente ou créée par les déforma-
tions évolutives (élastiques ε, fluage εcr et plastiques εp,g ), affectée par le coeficient
de Biot bg . R̃tI est la résistance à la traction. Ce paramètre permet de tenir compte
de la pression non efficace, c’est-à-dire la pression à atteindre pour micro-fissurer
le squelette solide dans un échantillon exempt de contraintes. Mg le module qui lie
le volume à la pression.
+
Pg
P g = M g φg − φgv + bg ε + ε
e cr
+ tr ε p,g
(3)
R̃tI
Ensuite, un critère de Rankine (fIasr , équation 3) est utilisé pour estimer

l’endommagement diffus du matériau (perte des propriétés mécaniques). Cet
endommagement évolue alors directement en fonction de l’évolution du gonflement.
Le critère est anisotrope et l’endommagement débute dès que la résistance à la trac-
tion est obtenue. Si un chargement de compression est appliqué dans une direction,
295
Q.104 - R.18
la fissuration diffuse est supprimée dans cette direction (si la contrainte effective est
de compression).
fIasr = Pg + min (σI ; 0) − R̃tI I ∈ [I , II , III ] (4)
L’endommagement du matériau dû au gonflement est alors estimé suivant

l’équation 4. ε k,ASR est la déformation caractéristique égale à 0,3% pour l’alcali
réaction.
p,g
ε
DIasr = p,g I k,g (5)
εI + ε
Le gonflement via sa loi d’évolution de la pression et de son endom-

magement et couplé avec un modèle d’endommagement anisotrope structurel
(ENDO_PORO_BETON et un modèle de fluage (FLUA_PORO_BETON). Ces
modèles sont détaillés dans [2, 5–7].
2.3. APPLICATION DU MODÈLE
Le modèle a été mis en œuvre sur des ouvrages en France tel le barrage
du Chambon (RAG), de Maury, de Castelnau (RSI) et aussi à l’étranger sur des
ouvrages du parc d’HydroQuébec (RAG). Nous présentons ici son application sur
un barrage atteint de RSI.
La première étape de la modélisation consiste à reproduire l’histoire thermique

du barrage afin d’estimer les zones de gonflements potentielles.
Cette histoire thermique est fortement influencée par la période de coulage

(temps chaud ou non) et également par la vitesse de la mise en place du béton.
Après avoir reproduit l’histoire thermique au jeune âge du barrage, grâce

aux phénomènes qu’il représente, le modèle est capable d’estimer l’évolution du
gonflement, les contraintes internes du barrage ainsi que la fissuration induite par
le gonflement. Par exemple sur un barrage atteint de RSI, après un calage des
Fig. 3
Températures maximales atteintes au centre de l’ouvrage (◦ C).
Maximum temperatures reached at the depth of the structure.
296
Q.104 - R.18
Fig. 4
Déplacements modélisés/auscultés – barrage RSI.
Modeling vs monitoring displacements - DEF.
Fig. 5
Fissuration modélisée [m/éléments].
Modeling of cracks.
paramètres, le modèle a été capable de reproduire l’auscultation et ce aussi bien

dans la direction verticale que dans les autres directions.
Il a permis de reproduire le comportement des ouvrages mais également

d’estimer l’évolution de leur fissuration et de leur état de contrainte interne. Une
confrontation avec des mesures in-situ aussi bien en contraintes qu’en faciès de
fissuration permet d’avoir une bonne confiance en la modélisation.
En particulier le modèle a donné une information sur la continuité de la fis-

suration en zone non visible. Sur notre ouvrage une fissuration est visible dans les
puits inter plots. Une possible continuité de cette fissuration était envisagée faisant
craindre à une instabilité du parement amont. La modélisation a bien retrouvé la
fissuration au niveau de drain mais a par contre estimé qu’il n’y avait pas continuité
de la fissuration à l’intérieur des plots.
297
Q.104 - R.18
Fig. 6
Fissuration relevée.
In situ cracking pattern.
Fig. 7
Résultantes au contact béton-rocher [MN].
Loading at Concrete-rocks interface.
Un comparatif entre les contraintes modélisées et les contraintes mesurées in-

situ par technique de surcarottage a également permis d’avoir une bonne confiance
en la représentativité du modèle. Les résultantes au contact béton-rocher ont été
calculées et une analyse du respect des critères de stabilité en vigueur a alors été
effectuée.
2.4. CONCLUSIONS PARTIELLES
La gestion de la sureté des ouvrages atteints de gonflement suit un processus

établi basé sur l’auscultation, les analyses en laboratoire et la modélisation. Dans
ce cadre, EDF/CIH en collaboration avec le LMDC Toulouse a développé un modèle
numérique capable de reproduire le comportement des ouvrages atteints de RSI. Ce
modèle intégré au code de calcul Code_Aster permet de prédire le comportement
passé ainsi qu’à long terme des ouvrages en couplant les phénomènes de retrait,
de fluage et d’endommagement à la RAG/RSI.
La bonne représentativité du modèle de comportement permet non seulement

d’améliorer la connaissance du comportement et la sureté des ouvrages atteints
mais également de définir des zones ou des travaux sont nécessaires. Dans ce
298
Q.104 - R.18
cas, il devient un outil indispensable pour concevoir, estimer l’efficacité et suivre les
restaurations envisagées.
La rénovation du barrage de Bimont présentée dans les prochains para-

graphes est l’exemple de l’utilisation d’un modèle numérique dans la gestion du
gonflement du barrage.
3. DE LA MODELISATION DU GONFLEMENT AUX TRAVAUX : LA

RENOVATION DU BARRAGE DE BIMONT
La modélisation du gonflement, couplée à des campagnes de reconnaissance

ciblées, s’est avérée indispensable pour la conception des travaux de rénovation du
barrage de Bimont.
3.1. CONTEXTE HISTORIQUE
Le barrage de Bimont, situé au pied du massif de la Sainte-Victoire, participe

à l’alimentation en eau de la ville de Marseille. Construit de 1947 à 1951 sur un
projet ACJB (André Coyne et Jean Bellier), le barrage a été mis en service en 1952,
et remis en concession à la Société du Canal de Provence et d’aménagement de
la région provençale lors de sa création en 1957. Il s’agit d’un barrage en béton de
type voûte à double courbure de 86,50 m de hauteur et dont l’épaisseur varie de 4
m en crête à 17 m en pied. Il délimite un réservoir de 25 hm3 sous la cote de retenue
normale d’exploitation 341 m NGF (10 m sous la crête). Compte-tenu des besoins
et des contraintes d’exploitation, la retenue est actuellement exploitée 13 m sous la
RN.
Dans les premières années, un réseau de fissures apparait progressivement

en partie basse des plots de la rive droite. Les fissures se développent suivant
des directions horizontales, verticales et également biaises, dont une fissure très
visible sur le parement aval du plot 3, ouverte de 13 mm. Le développement de la
fissuration s’accompagne de mouvements irréversibles de la rive droite du barrage,
notamment d’un soulèvement des plots, de leur basculement vers l’amont et d’un
déclavage des joints de construction.
Des investigations sont menées progressivement en parallèle d’un ajustement

du dispositif d’auscultation pour apprécier et comprendre le comportement évolutif
du barrage. Les études réalisées permettent d’attribuer l’origine de la fissuration
à un gonflement des bétons mis en place en rive droite concomitamment dans 4
zones bien identifiées, de 4 à 10 m de haut.
Entre 1965 et 1979, de nombreux essais en laboratoire sont réalisés sur des
échantillons du béton, qui concluent, avec l’état des connaissances de l’époque, à un
299
Q.104 - R.18
Fig. 8
Barrage de Bimont (Crédit photo SCP).
Bimont Dam.
phénomène d’alcali réaction. En 2003, des essais conduits par le LERM aboutissent
à la conclusion que le gonflement résulte d’une réaction sulfatique interne (RSI), et
que le potentiel de gonflement résiduel est négligeable. Ce constat conduit à lancer
les études de rénovation de l’ouvrage.
Le projet a été conçu en tenant compte des recommandations du CFBR et

en anticipant les futurs besoins en eau de la région : les travaux vont permettre
de rehausser la cote normale d’exploitation de l’ouvrage de 14 m, soit un gain de
volume stocké d’environ 13 millions de m3 .
Après la vidange de la retenue imposée par les travaux, la remontée du plan

d’eau a débuté au printemps 2019. Elle s’effectue par paliers, avec un suivi attentif
du comportement du barrage tout au long du remplissage.
3.2. LES ÉTUDES : APPORTS COMBINÉS DES RECONNAISSANCES ET DE LA

MODÉLISATION NUMÉRIQUE
Un des objectifs principaux des études de projet était de circonscrire la

zone à rénover de l’ouvrage en améliorant la connaissance des fissures internes,
pouvant influencer l’équilibre général de l’ouvrage. Une modélisation numérique
aux éléments finis et un programme de reconnaissances in-situ ont été lancés
simultanément.
300
Q.104 - R.18
3.2.1. Apports de la modélisation
Le modèle représente la voûte et sa fondation proche et reproduit les limites

des zones gonflantes. Les modules du massif de fondation et du barrage ont été
choisis de manière à reproduire les déplacements mesurés du barrage lors des
variations de la charge hydrostatique. Ils sont respectivement de 10 et 25 GPa.
Le gonflement a été modélisé à l’aide du module SCAS du logiciel Coyne et

Bellier COBEF élaboré à l’origine pour reproduire les effets de l’alcali réaction. Ce
module est inspiré de la loi de Charlwood [11], la contrainte limite est égale à 5 MPa,
et le gonflement libre 90 µm/m/an. Le gonflement, localisé et très fort, engendre des
contraintes de traction qui dépassent (en calcul linéaire) la résistance du béton. En
réalité, ces contraintes sont relâchées par l’ouverture des joints de construction et
du contact barrage-fondation ou la création de fissures. La loi NOTEN qui permet
de libérer toutes les contraintes de traction dans les éléments volumiques a été
utilisée. Cette loi a l’avantage de ne préjuger ni de l’emplacement ni de l’orientation
des fissures.
Le modèle permet non seulement de retrouver les déplacements irréversibles

de la structure, mesurés par le dispositif d’auscultation, mais aussi de reproduire
les fissures observées sur le barrage permettant ainsi l’analyse des mécanismes de
fissuration. La modélisation a de plus montré l’existence d’une fissuration interne de
feuilletage, parallèle aux parements, inconnue mais suspectée au début des études,
telle qu’illustrée par la Figure 10.
Fig. 9
Maillage du barrage et de sa fondation.
Dam and foundation mesh.
1 Joint entre deux plots 1 Joint between two blocks
2 Zones gonflantes 2 Swelling areas
3 Joint entre le barrage et la 3 Joint along the dam footprint
fondation
301
Q.104 - R.18
Fig. 10
Vue d’aval développée de la rive droite du barrage - Déformations plastiques
révélant la fissure de feuilletage.
Developed downstream view of the dam right bank – Plastic strains showing the
internal cracks area.
3.2.2. Apports des reconnaissances
Les objectifs des reconnaissances par sondages et imageries de paroi de

forage (OPTV) étaient essentiellement d’examiner l’ouverture des joints et des fis-
sures au sein du barrage. Les forages, horizontaux pour la plupart, ont été implantés
autour de chaque noyau de gonflement, en tenant compte des résultats du modèle
numérique. Le programme des investigations a été adapté à l’avancement. Au total
49 forages ont été réalisés autour des noyaux de gonflement (dont 46 inspectés par
caméra).
La fissuration interne de feuilletage est observée dans 28 forages avec des

ouvertures pouvant aller jusqu’à 20 mm. Il a ainsi pu être mis en évidence que le
réseau de fissures internes s’étend sur les sept plots de rive droite. La superficie
totale (incluant les zones gonflantes) est estimée à 2000 m2 (environ 20% de la
surface). Ces observations sont en cohérence avec les résultats du modèle.
3.2.3. Conclusions et définition des travaux
En plus de l’ouverture des joints inter-plots, les études ont donc mis en évi-
dence la présence de trois types de fissures internes qui rayonnent autour des
zones gonflantes : des fissures verticales dans le plan amont-aval, des fissures hor-
izontales/biaises et des fissures verticales situées vers la mi-épaisseur de la voûte,
parallèles aux parements. Ces dernières sont appelées “ fissures de feuilletage ”.
302
Q.104 - R.18
Fig. 11
Exemples de clichés pris par la caméra OPTV dans 2 forages : fissure ouverte et
joint déclavé. Examples of images in boreholes: crack and open joint.
Fig. 12
Schéma des plans de fissuration autour d’une zone gonflante.
Scheme of cracking planes around a swelling area.
1 Ouverture des joints de 1 Opening of construction joints
construction
2 Fissures de feuilletage 2 Internal cracks
3 Fissures horizontales 3 Horizontal cracks
4 Fissures verticales 4 Vertical cracks
5 Parement aval du noyau gonflant 5 Downstream face of the swelling
block
La combinaison de ces plans de fissures peut potentiellement prédécouper

des blocs de béton, instables sous l’effet de la pression hydrostatique ou sous l’effet
d’une pression d’injection. La définition des travaux d’injection des fissures internes
a donc intégré la nécessité de préalablement renforcer l’ouvrage afin d’éviter la
déstabilisation de ces blocs.
Les travaux comprennent un réseau dense de forages d’injection permettant

une diffusion du coulis depuis la fissure de feuilletage jusque dans les fissures
verticales et horizontales qu’elle recoupe. À l’issue des études, ont ainsi été définis :
303
Q.104 - R.18
Fig. 13
Implantation des forages d’ancrages (noir) et des forages d’injection (bleu).
Anchor drilling (black) and grouting (blue) locations.
- 140 ancrages passifs positionnés pour venir renforcer la structure préalable-

ment à l’injection des fissures.
- 260 forages d’injection dont la direction permet de venir recouper les fissures
de feuilletage.
- 130 forages venant recouper en biais les joints ouverts afin d’opérer un
clavage de la voûte à l’issue des travaux d’injection des fissures.
3.3. LES TRAVAUX D’INJECTION DES FISSURES
Les travaux d’injection des fissures, au cœur de la rénovation de l’ouvrage,

ont été délicats et innovants : la consistance du coulis et les pressions d’injection
devaient à la fois permettre de remplir des fissures de quelques millimètres, sans
toutefois provoquer un accroissement de leur ouverture ou de leur extension.
Les travaux ont été réalisés par l’entreprise Demathieu Bard Construction ([2]).
Etant donnée la difficulté d’accès et le nombre de forages à réaliser, le parti pris a
été de mettre en place un échafaudage de 2400 m2 , suspendu au parement aval,
et complété par une grue à tour.
3.3.1. Renforcement préalable à l’injection
Les forages pour la mise en place des ancrages passifs sont réalisés en
destructif à l’air. La longueur des barres varie de 4 à 20 m selon leur position sur
l’ouvrage. Les barres sont assemblées sur la crête de l’ouvrage avant leur mise en
place.
304
Q.104 - R.18
Fig. 14
Mise en place d’une barre d’ancrage au pied du barrage.
Installation of an anchor bar at the bottom of the dam.
Les barres sont scellées sur toute leur longueur ; il est complété par des
platines d’ancrage lorsque l’épaisseur du barrage est trop réduite ou lorsque la
fissure est trop proche d’un parement. Le coulis utilisé pour le scellement de ces
barres a des caractéristiques spécifiques puisque celui-ci ne doit en aucun cas
pénétrer dans les fissures avant l’injection de celles-ci. Il se caractérise donc par
une très grande thixotropie et un retrait compensé. Des essais préalables, dans des
tubes translucides notamment, ont permis de valider sa formulation.
3.3.2. Injection au coulis de ciment
La procédure consiste à injecter à 0,6 bars du bas vers le haut par niveaux de
2 m. La résurgence de coulis sur la ligne supérieure constitue un point d’arrêt. Un
volume maximum d’injection est également estimé sur la base de l’ouverture et de
l’extension des fissures repérées par les endoscopies.
La pression, très faible, combinée à des débits proches de 0 ne peut être main-
tenue par des presses “conventionnelles” sans pic. L’entreprise a donc proposé de
contrôler la pression par un positionnement du bac de reprise à une hauteur cor-
respondante à la pression maximum d’injection et par l’emploi des vannes d’entrée
au niveau des points d’injection. Cette méthode visait à éviter tout pic de pression
même en cas de défaillances électroniques.
3.4. AUTRES TRAVAUX LIÉS À LA PROBLÉMATIQUE DU GONFLEMENT
3.4.1. Pose d’un complexe d’étanchéité
Entre 2009 et 2013, dans le cadre des études de projet, de nouveaux essais
sur les bétons gonflants sont conduits par le laboratoire Matériaux et Durabilité
305
Q.104 - R.18
Fig. 15
Pose du complexe étanche sur le parement amont rive droite.
Installation of a watertight geomembrane on the upstream face of the right bank.
des Constructions (LMDC) de Toulouse. Le nouveau protocole met en évidence un

potentiel de gonflement résiduel des noyaux de béton atteints de RSI.
Ce constat conduit à intégrer dans le projet de rénovation la pose d’une

étanchéité visant à protéger la rive droite de la réhumidification du béton et empêcher
ainsi la réactivation de la RSI lors de la remontée du plan d’eau.
Les travaux d’étanchement ont consisté en la mise en place d’un géocompos-

ite sur le parement amont. Ce géocomposite est constitué d’une membrane PVC
souple thermosoudée sur un géotextile antipoinçonnant. Il est posé sur un géodrain
en PEHD qui fait office de drain. La géomembrane est fixée mécaniquement à
son support au moyen de profilés métalliques qui sont ancrés directement sur le
parement. La Société CARPI a réalisé ces travaux.
3.4.2. Tirants d’ancrage de la culée rive droite
Les deux culées de rive sont des ouvrages massifs, qui résistent aux actions
auxquelles elles sont soumises en mobilisant des forces de frottement au contact
de la fondation sous leurs poids propres. Les calculs de stabilité ont montré que,
essentiellement en raison des effets du gonflement et du chargement thermique
estival, les critères de sécurité n’étaient pas atteints pour la culée rive droite.
Celle-ci a été renforcée par l’entreprise SPIE Fondation au moyen de tirants

actifs afin d’augmenter artificiellement son poids. Quatre tirants verticaux ont été
mis en œuvre pour exercer une force verticale de 7,5 MN.
306
Q.104 - R.18
3.5. REMPLISSAGE ET PERSPECTIVES
Le remplissage de la retenue a atteint son premier palier et le barrage se

comporte de façon attendue. Les nouveaux pendules des culées RD et RG mon-
trent que celles-ci sont restées stables pendant toute l’opération. Le géocomposite
d’étanchéité fixé sur le parement amont du barrage est efficace, les débits col-
lectés à l’aval étant très faibles. Les opérations d’injection de fissures n’ont pas
modifié significativement le comportement de la voûte, mais le clavage est visible
sur les mesures de plusieurs appareils, en particulier les mesures des vinchons,
téléfissuromètres et pendules.
Dans le cadre du suivi post-travaux du barrage de Bimont, l’utilisation de mod-

èles de comportement dédiés au gonflement et couplant les différents phénomènes
évoluant dans les bétons gonflants (réactions de gonflements internes, retrait, flu-
age, endommagement) permettra à TRACTEBEL de suivre finement et de valider
le comportement macroscopique de l’ouvrage sur le long terme
4. CONCLUSION
Cet article présente l’apport de la modélisation dans la gestion des ouvrages

atteints de gonflement.
Dans un premier temps, la méthodologie de gestion a été présentée. Cette

méthodologie est basée sur trois axes : l’auscultation, les essais en laboratoire et
l’analyse de comportement. La modélisation de l’ouvrage prenant alors une place
prépondérante dans cette méthodologie.
Dans un second temps, la modélisation phénoménologique utilisée à EDF

CIH a été présentée. Cette modélisation utilise le cadre de la poromécanique et
prend en compte le couplage des différents phénomènes évoluant dans le béton
(retrait, fluage, endommagement, réactions de gonflements internes). TRACTEBEL
développe également ce type de modèle numérique. En plus de justifier le com-
portement de l’ouvrage dans son état actuel, les modèles permettent également de
prédire les déformations et contraintes à long terme d’ouvrages en béton atteints
de gonflement.
Enfin, un exemple de gestion d’ouvrage atteint a été présenté à travers

l’exemple du barrage de Bimont (SCP). Après une étude approfondie des mécan-
ismes et conséquences du gonflement (RSI), la rénovation suivante a été proposée :
- des travaux d’injection, après mise en place de barres d’ancrage horizon-

tales, afin de restaurer le monolithisme de la voûte,
- la mise en place de tirants précontraints verticaux afin de stabiliser la culée
rive droite soumise aux efforts provoqués par les chargements hydrostatique,
saisonniers et du gonflement,
307
Q.104 - R.18
- la pose d’une membrane d’étanchéité afin de limiter les apports d’eau et

d’exclure une reprise des gonflements.
REFERENCES
[1] GRIMAL E., “ Caractérisation007. Caractérisation des effets du gonflement

provoqué par la réaction alcali-silice sur le comportement mécanique d’une
structure en béton ”, PhD thesis, Université de Toulouse, Université Toulouse
III - Paul Sabatier, 2007.
[2] MORENON P., (2017), “ Modélisation des réactions de gonflement interne des
bétons avec prise en compte des couplages poro-mécaniques et chimiques ”,
PhD, Université de Toulouse, Université Toulouse III - Paul Sabatier), 2017.
[3] GRIMAL E., SELLIER A., PAPE Y. L. & BOURDAROT E., “ Creep, Shrinkage,
and Anisotropic Damage in Alkali-Aggregate Reaction Swelling Mechanism-
Part I: A Constitutive Model ”, Materials Journal, vol. 105, n˚ 3, p. 227-235,
2008.
[4] GRIMAL E., SELLIER A., PAPE Y. L. & BOURDAROT E., “ Creep, Shrinkage,
and Anisotropic Damage in Alkali-Aggregate Reaction Swelling Mechanism-
Part II: Identification of Model Parameters and Application ”, Materials Journal,
vol. 105, n˚ 3, p. 236-242., 2008 .
[5] GRIMAL E., MORENON P., SELLIER A., MULTON S., BOURDAROT E.,
“AAR and DEF Structural Effects Modelling”, Swelling Concrete in Dams and
Hydraulic Strcutures, WILEY, 2017.
[6] MORENON P., GRIMAL E., KOLMAYER P., SELLIER A., MULTON S., “ Mod-
elling, assesment and rehabilitation of a dam affected by an internal swelling
reaction ISR) ”, DAM Engineering, Volume XXIX issue 2, 2018.
[7] Code_Aster, “ Loi de comportement sur le fluage et l’endommagement du

béton (RSI/RAG) ”, www.code-aster.org, [R7.01.30].
[8] CASTEIGTS C., LALICHE K., NORET C., BOURGOUIN T. : Rénovation du

barrage de Bimont, 25th ICOLD Stavanger, 2015.
[9] NORET C., MOLIN X., CARRERE A., CASTEIGTS C., GRAWITZ B.,
LALICHE K. : Apports combinés des reconnaissances et de la modélisation
dans l’analyse du comportement du barrage voute de Bimont, 24th ICOLD
Kyoto, 2012.
[10] CHARLWOOD R.G., SOLYMAR Z.V., STEELE R.R., CURTIS D.D. : A review of
AAR in hydroelectric plants and dams, in: Proceedings of International Confer-
ence on Concrete Alkali-Aggregate Reactions in Hydraulic Plants and Dams,
Fredericton, NB. Canada, Canadian Electrical Association, 1992.
308
Q.104 - R.19
-------
-------
ESTIMATING FLOOR DESIGN SPECTRA AT THE TOP OF A DAM USING DAM

RESPONSE SPECTRUM ANALYSIS RESULTS(∗)
Ali RASEKH, PHD, P.ENG.

Sr Structural Engineer, KLOHN CRIPPEN BERGER, Vancouver
CANADA
SUMMARY
Floor Design Spectra (FDS) are developed to be used in the response spec-
trum analysis of the dam appurtenant structures such as gates, and hoists. They are
also used for selection of equipment for a prescribed level of shaking and deciding
about seismic isolation of some components. The comprehensive way of determin-
ing the FDS is through performing time-history analyses on a finite element model
of the dam using a set of seismic records. These seismic records are typically com-
patible with the target ground spectrum at the location of the dam. Time-histories of
the response at the location of the appurtenant structures and equipment are used
to compute floor design spectra for their analysis and design. A simplified solution
for estimating the floor spectra is presented in this study. In this solution an estimate
of the floor spectra is computed by properly combining the modal responses at the
location of the appurtenant, which are obtained from a typical modal analysis of the
dam. This simplified solution could be part of a progressive analysis and be used
in the initial design, assessment or equipment qualification. This paper presents
the simplified solution followed by a case study of this solution for an appurtenant
structure.
∗ Estimation des spectres de conception de plancher au sommet d’un barrage à l’aide des
résultats d’analyse spectrale de réponse du barrage
DOI: 10.1201/9781003211808-19 309

Q.104 - R.19
RÉSUMÉ
Les spectres de conception de plancher sont développés pour être utilisés

dans l’analyse du spectre de réponse des structures annexes faisant partie des
barrages tels que les vannes et treuils. Ils sont également utilisés pour la sélec-
tion des équipements pour un niveau défini de secousses séismiques et pour
décider l’isolement sismique de certains éléments. L’approche pour la détermina-
tion des spectres de plancher consiste à effectuer des analyses a partir de spectre
de réponses d’accélérogrammes historiques sur un modèle d’élément fini du bar-
rage. Ces relevés sismiques historiques sont généralement compatibles avec le
spectre du sol recherché à l’emplacement du barrage. Les spectres de réponse
historiques à l’emplacement des structures annexes et de l’équipement sélectionné
choisi sont utilisés pour calculer les spectres de conception de plancher pour leur
analyse et leur conception. Une solution simplifiée pour estimer les spectres de
plancher est présentée dans cette étude. Dans cette solution, une estimation des
spectres de plancher est calculée en combinant correctement les réponses modales
à l’emplacement de la structure en question, qui sont obtenues à partir d’une anal-
yse modale typique du barrage. Cet article présente la solution simplifiée suivie
d’une étude de cas de cette solution pour une structure annexe.
1. INTRODUCTION
Appurtenant structures, such as spillways, intakes, outlets, penstocks, gates,

control buildings, and their associated equipment, are essential components of a
dam and hydroelectric facility; failure of any of these structures could result in loss
of reservoir control after an earthquake (USSD, 2018). Water discharge systems,
such as gates, valves, and water conduits, may be required to perform their intended
functions after an earthquake, and the collapse or operation failure of the electrical
or mechanical equipment should be prevented.
Earthquake-resistant design of structures that are appurtenant to dams is an

important portion of the seismic structural design and upgrade of dams. However, the
input earthquake demand for an appurtenant structure or piece of equipment (both,
hereafter, referred to as ‘the appurtenant’) is different from the free-field earthquake
hazard used for the analysis of the dam. The characteristics and intensity of this
demand depends on the appurtenant location and support conditions, which, in
this paper, shall be described as the ‘interface of the appurtenant with the dam.’ In
fact, peak accelerations at the top of a dam can be as high as ten times the values
observed at its base (Sasaki, 2018).
The first step in a seismic evaluation of a dam and its appurtenant is an

assessment of the potential consequences of the dam failure; this assessment is
based on dam safety guidelines, such as those by the Canadian Dam Association
310
Q.104 - R.19
(CDA, 2007). The results from this assessment determine the dam classification and
the corresponding Annual Exceedance Probability (AEP) of the Maximum Design
Earthquake (MDE) used for seismic evaluation of the dam.
For the seismic evaluation of the appurtenant, one should also determine
whether it is a critical or non-critical structure. While critical structures should be
designed for the same MDE as the dam, non-critical structures can be designed for
a seismic event less than the MDE (USSD, 2018). Similarly, when there is redun-
dancy in the system, the appurtenant may not need to be designed for full functional
integrity after the MDE.
The second step of the seismic evaluation includes using the dam classifica-
tion to evaluate the seismic hazard at the site, which typically leads to developing
the target ground spectra of the MDE. Target ground spectra are often uniform haz-
ard response spectra, but deterministic or conditional mean spectra are also used
in different projects. Target spectra are typically developed for 2%, 5%, and 10%
damping. A corresponding set of input time-histories (in 3 directions) might also be
developed as part of the seismic hazard evaluations, based on the level of analysis
intended for the dam and the appurtenant.
The third step of the seismic evaluation involves determining the seismic analy-
sis methodology. The analysis level depends on the dam classification, its height and
complexity, and the intensity of the earthquake (CDA, 2007). The common practice
in the consulting industry is to start with simplified and quick solutions in the feasi-
bility stage of design, and increase the level of analysis sophistication throughout
the detailed design. This practice is sometimes called progressive analysis. These
analysis levels are described further in the following section.
The fourth step is to complete the appropriate level of seismic analysis and
evaluate the seismic performance of the dam and the appurtenant, based on the
analysis results.
2. SEISMIC ANALYSIS OF THE APPURTENANT
Seismic analysis of the dam and the appurtenant may be performed in one of
the following three levels of sophistication: Level 1 - pseudo-static analysis (seismic
coefficient); Level 2 - response-spectrum (linear modal) analysis; and Level 3 -
time-history analysis, which could be linear or non-linear (CDA, 2007).
Level 1 seismic analysis is dependent on estimating the maximum accelera-

tion at the interface of the appurtenant with the dam (henceforth referred to as ‘peak
in-structure acceleration’). To use Level 2 seismic analysis for the appurtenant, the
spectral accelerations at its interface with the dam (hereafter called ‘in-structure
311
Q.104 - R.19
response spectra’) should be estimated, which is equivalent to floor response spec-

tra for buildings. For Level 3, the time-history of acceleration at the interface should
be output and used in the analysis, design, evaluation, or retrofit of the appurtenant.
The methods described above assume that the mass of the appurtenant is
relatively small, and that no coupling or interaction of responses between the dam
and the appurtenant would occur.
The peak in-structure acceleration used in Level 1 could be estimated with

a response spectrum analysis of the dam (or pseudo static analysis, if the dam
is rigid). However, for seismic evaluation of the appurtenant, using the peak in-
structure acceleration at the interface does not account for the possible resonance
of the appurtenant to the dam primary modes.
Level 3 analysis is the most expensive method, both in preparation and imple-
mentation. This level of analysis can include nonlinearities, and would best account
for dam-foundation interaction and different damping effects, such as material and
radiation damping. Therefore, Level 3 is the ideal level of analysis for the detailed
design phase, or evaluation of a complex or tall dam.
2.1. RESPONSE SPECTRUM ANALYSIS (RSA) OF THE APPURTENANT
The RSA solution (USACE, 1999) can account for the appurtenant damping
and the characterization of uncertainty in the model, without intensive analysis, and
makes a good candidate for a mid-level analysis throughout the design.
In-structure response spectra at the appurtenant interface is typically devel-

oped from the results of a series of time-history analyses of the dam. This procedure
is followed by enveloping and widening of the seismic response results to develop
the Floor Design Spectra (FDS). The FDS can then be used to design and evaluate
appurtenant structures, select equipment for a prescribed level of shaking, or deter-
mine if a set of components need to be isolated in order to improve their ability to
withstand the imposed shaking (IEEE, 2005).
However, this solution may be considered more rigorous than what is expected
from a Level 2 analysis, as time-history analyses are costly and could not be com-
pleted in the feasibility phase if the set of input acceleration time-histories have not
already been developed. This level of dam analysis might be prohibitive for a quick
evaluation of the appurtenant or a parametric study of the effects that the dam and
foundation properties have on the appurtenant response.
An approximate solution is developed in the next section, based on a heuristic

approach to the RSA. This solution is intended to be part of a Level 2 seismic
analysis, and could be used for quick seismic evaluations. This solution could also
be used in validating the more sophisticated time-history solutions.
312
Q.104 - R.19
2.2. THE APPROXIMATE SOLUTION
The approximate solution provided in this paper is based on modifying the

classic RSA procedure to account for the possible resonance of the appurtenant to
the primary modes of vibration of the dam. In order to achieve this goal, the modal
analysis results for the dam-foundation-reservoir system are obtained using Finite
Element Analysis (FEA) of a complete or simplified model of the dam.
Modal acceleration of the i’th mode at the appurtenant interface could be

calculated by the following equation (Chopra, 2000):
q̈i = i φi (XP , YP , ZP ) Sa (Ti ) (1)
where i is the modal participation factor, ϕi (XP , YP , ZP ) is the mode shape at the
interface, and Sa (Ti ) is the spectral acceleration at the period of the i’th mode (Ti ).
Peak in-structure acceleration (PISA) at the interface could be obtained by combining
the modal acceleration of the dominant modes:

n

PISA = [q̈i ]2 (2)
i=1
n is the number of modes that are sufficient to achieve a total effective modal mass
more than 90% of the actual mass of the system in the direction of interest.
In order to obtain an approximate in-structure response spectra, the effect of

the steady-state response of the appurtenant to the interface acceleration would
be accounted for. Hence, the modal acceleration in Equation [1] is modified by the
response factor R (Chopra, 2000). The response factor for a period T is:
1
R(T ) = (3)
2 2
2
(1 − T
Ti + 2ζa TTi
where ζa is the damping ratio of the appurtenant. The in-structure spectral

acceleration (ISSa ) would be estimated to be:

n

ISSa (T ) = [R (T ) q̈i ]2 (4)
i=1
The above solution is approximate, and only accounts for the steady-state
response of the appurtenant. The approximate in-structure response spectrum
developed here could be used as an approximate FDS after widening the ordinates
by 15%, as recommended by the U.S Nuclear Regulatory Commission (NRC, 1978);
it could also be used for preliminary design and to validate time-history solutions.
313
Q.104 - R.19
3. CASE STUDY
Figure 1 shows a gated spillway structure which is studied here to establish the
procedure described in this paper. A model of the dam and foundation was created.
The foundation was modeled with a Young’s modulus of 2000 MPa and a Poisson’s
ratio of 0.29. The dam was modeled with a Young’s modulus of 19,600 MPa and a
Poisson’s ratio of 0.3. The foundation was modeled as massless, and the density
of the dam was assumed to be 2,180 kg/m3 . However, as the model created for
this study is a slice model with a thickness of unity, the material properties assigned
to the pier and trunnion are reduced by the ratio of the pier thickness to spillway
center-to-center length.
Fig. 1
Spillway-sluice Structure for Case Study.
Structure d’évacuateur de crues pour l’étude de base.
The hydrodynamic effects were accounted for by added mass on the upstream
face (Westergaard, 1933).
Figure 2 shows the FEA model of the same dam (including the foundation,
spillway, piers, and the trunnion block), created using ANSYS software (2013). Cen-
ter node of the trunnion block is the interface where the in-structure response spectra
is obtained using the simplified solution.
314
Q.104 - R.19
Fig. 2
FEA Model of the Dam.
Analyse MEF (model élément fini) du barrage.
Modal Analysis was performed on this model, and the first three primary fre-
quencies were obtained to be 4.97 Hz, 8.87 Hz, and 12.86 Hz; their total modal
masses accounted for more than 99% of the mass of the system. An RSA was
performed next and, as shown in Figure 3, the horizontal direction acceleration at
the trunnion location was obtained to be 7.15 m/sec2 (∼0.73g), which is the peak
in-structure acceleration at this interface, and could be used for initial design or
check of the gates. However, this value is an underestimate and does not account
for the proximity of the gate frequency (6.37 Hz) to the primary dam frequency, in
the upstream/downstream direction.
Fig. 3
Spectral Accelerations.
Accélération spectrale.
315
Q.104 - R.19
The approximate in-structure response spectrum of the dam was developed

per the solution provided in this paper. The spectrum is shown in Figure 4, which
shows that the proper response to be used in the preliminary design should be about
1.77g (more than twice the peak in-structure acceleration). This value was shown to
be in close agreement with the response obtained from a more rigorous time-history
analysis, and validated the results obtained from the more rigorous solution.
Fig. 4
Estimated In-structure Response Spectrum at the trunnion.
Spectre de plancher estimé au niveau de l’axe de support des vannes.
4. SUMMARY AND CONCLUSIONS
A simple solution was developed to be used alongside response spectrum

analysis for the appurtenant seismic evaluation. This solution was used to estimate
the approximate in-structure response spectrum at the trunnion of a gated spillway,
and was shown to improve the estimation of response compared with using the
peak in-structure acceleration at the trunnion location. The simplified solution is
intended to be used for preliminary design, sensitivity and parametric studies, and
as a verification tool to validate more rigorous time-history analysis.
ACKNOWLEDGEMENT
The author would like to acknowledge support and funding from Klohn Crippen
Berger Ltd. for presenting this study at ICOLD 2021 Annual Conference.
316
Q.104 - R.19
REFERENCES
[1] ANSYS (2013), Workbench User’s Guide.
[2] Canadian Dam Association (CDA) (2007), Dam Safety Guidelines.
[3] Chopra, A.K. (2000), Dynamics of structures, Pearson Education, USA.
[4] Institute of Electrical and Electronic Engineers (IEEE) (2005) Standard 693-
2005: IEEE Recommended Practice for Seismic Design of Substations.
[5] Sasaki, T., Ohmachi, T., and Matsumoto, N. (2018). Analysis on acceleration
data of dams collection by JCOLD. ICOLD Proceedings - Validation of Dynamic
Analyses of Dams and Their Equipment - Jean-Jacques Fry and Norihisa
Matsumoto, 1.
[6] US Army Corps of Engineers (USACE) (1999), Response Spectra and Seismic
Analysis of Concrete Hydraulic Structures, EM 1110-2-6050.
[7] U.S Nuclear Regulatory Commission. (NRC) (1978), Regulatory Guide

1.122. Development of Floor Response Spectra for Seismic Design of Floor
Supported Equipment or Components. Revision 1.
[8] US Society on Dams (USSD)(Draft-2018), Guidelines for Earthquake Design

and Evaluation of Structures Appurtenant to Dams.
[9] Westergaard, H.M. (1933), “Water pressures on dams during earthquakes.”

Transactions of ASCE 98, 418-433.
317
Q.104 - R.20
-------
-------
PREDICTING THE MEASURED BEHAVIOR AND DEFINING WARNING

LEVELS OF A CONCRETE DAM(∗ )
Richard MALM
Specialist, SWECO ENERGY AB
Lisa BROBERG & Jonas ENZELL

Structural Engineer, SWECO ENERGY AB
Johan BLOMDAHL
Project Manager, SWECO ENERGY AB
Carl-Oscar NILSSON
Dam Safety Manager, SYDKRAFT HYDROPOWER AB
SWEDEN
SUMMARY
Dam surveillance is important to ensure that high dam safety is maintained

throughout the service life of a dam. This is especially important as many dams
are approaching the end of their expected service life. Sensors should be installed
in strategic positions to give early indications if the dam behaviour changes and
early warnings of potential dam failures. In a new monitoring project, the commonly
used databased regression methods are not applicable to use to define warning
levels since they require sufficiently long recordings of the dam behaviour to train
the algorithms. In this paper, it is shown how numerical analyses based on the finite
element (FE) method can be used to plan the dam monitoring, such as placement of
sensors and type of sensors and how they can be used to predict the dam behaviour.
Measurement recordings of the first 1.5 years are also presented to validate the
numerical models. The results show that the FE models can be used with high
∗ Prédiction des observations du comportement d’un barrage en béton et définition de
niveaux d’alerte
318 DOI: 10.1201/9781003211808-20

Q.104 - R.20
accuracy for defining warning levels early in the project and be used for definition of
alert values, i.e. indications of potential failures. Accurate predictions are necessary
for supporting the owner in decision making and planning future maintenance.
RÉSUMÉ
Les mesures de surveillance sont importantes afin de garantir que la sécu-

rité des barrages soit maintenue pendant toute la durée de vie d’un ouvrage. Ceci
est particulièrement important alors que de nombreux barrages se rapprochent
de la limite de leur durée de service. Il est recommandé d’installer des capteurs
aux endroits stratégiques pour pouvoir donner des indications précoces en cas de
changement de comportement du barrage et pouvoir émettre des alertes en cas
de risque potentiel de rupture. Dans un nouveau projet de surveillance d’ouvrage,
les méthodes de régression couramment utilisées axées sur des bases de données
ne peuvent pas être utilisées pour définir les niveaux d’alerte, car elles nécessitent
que les enregistrements du comportement du barrage soient suffisamment longs
pour pouvoir entraîner les algorithmes. Dans cet article, il est montré comment les
analyses numériques basées sur les méthodes des éléments finis (EF) peuvent être
utilisées pour planifier la surveillance du barrage, comme le placement des capteurs
et le type de capteurs et comment ils peuvent être utilisés pour prédire son comporte-
ment. Des enregistrements réalisés durant les 1,5 premières années de mesures
sont également présentés et utilisés pour valider les modèles numériques. Les résul-
tats montrent que les modèles FE peuvent être utilisés avec une grande précision
pour définir les niveaux d’alerte au début du projet et être utilisés pour la définition
des valeurs d’alerte, c’est-à-dire des indications de ruptures potentielles. Des éval-
uations précises sont un élément crucial pour le Maître d’Ouvrage pour l’aide à la
décision ainsi que pour la planification des nécessaires pour aider le propriétaire à
prendre des décisions et à planifier les opérations de maintenance futures.
1. INTRODUCTION

throughout the whole service-life of dams. In many countries that early developed
hydropower, many of the dam structures are now approaching the end of their
expected service-life. Dam monitoring is important to ensure that the dam behaves
as expected during normal operation. Sensors should therefore be installed in
strategic positions to give early indications if the dam behaviour changes and early
warnings of potential dam failures. It is well known that the normal behaviour of
concrete dams is primarily governed by variations in ambient temperature and in
reservoir level. A common approach to define warning levels is by utilizing data-
based (statistical) models, such as regression models. With these, the expected
variation in the dams’ response can be determined in advance with high accu-
racy. The regression models are typically based on the assumption that the dam
319
Q.104 - R.20
behaviour can be described by the following effects; Hydrostatic–Seasonal-Time

(HST) or Hydrostatic-Temperature-Time (HTT), see for instance [1–3] among others.
In the recent years, data-based models based on machine learning techniques such
as Artificial Neural Networks (ANN), Support Vector Machines (SVM) and Boosted
Regression Trees (BRT) are becoming more and more widely used in the dam engi-
neering community, see [4–6]. All these methods have great possibilities to give
highly accurate predictions of the normal behaviour of a dam and to define suitable
warning levels. However, these methods need a sufficient amount of data, typically
a few years of recordings, in order to train and evaluate the models. In a new moni-
toring project, these methods are hence not applicable since no previous recordings
of the dam behaviour exist that can be used to train the algorithms. In these cases,
models based on finite element (FE) methods can instead be utilized to obtain a
best estimate of the future response.
Two Swedish concrete buttress dams have recently been updated with an
extensive monitoring program. In this paper, the numerical models utilized to predict
the dam behaviour during normal conditions and during potential failure modes are
presented. In concrete buttress dams, the monoliths act more or less independently
from each-other. This results in complications regarding designing dam surveillance,
since one monolith theoretically may go to failure while the others are unaffected.
Thereby, the monitoring system must be designed for this. A solution has been
developed where the monoliths are classified based on their risk and consequence
of failure and based on this, an instrumentation level has been designed, see [7]. As a
first step of the monitoring campaign of the dams Storfinnforsen and Ramsele, the six
of the most critical monoliths have been selected which are extensively instrumented.
This paper focus on how numerical models based on the finite element method have
been used to develop the monitoring program.
2. RAMSELE AND STORFINNFORSEN CONCRETE BUTTRESS DAMS
Ramsele (Fig. 1) and Storfinnforsen (Fig. 2), are thin concrete flat slab but-
tress dams and both were built in the 1950s. Both dams are located in Faxälven
river, where Ramsele is located 10 km downstream from Storfinnforsen. These
dams are two of the largest Swedish hydro power plants owned by Uniper / Sydkraft
Hydropower AB, with a total annual energy production of 1.4 TWh. Ramsele con-
sists of 49 and Storfinnforsen of 81 monoliths, where the tallest monoliths are slightly
above 40 m for both dams. Both dams are designed with an 8 m wide reinforced
concrete frontal slab having a thickness varying between 1.2 m at the crest to about
2.6 m at the foundation. For the top ten meters, the frontal slab is vertical and then
inclined (1:1.5) in the upstream direction for the bottom part to the foundation level.
The frontal slab is supported by a reinforced concrete buttress wall with a thickness
of 2 m. The width of the buttress wall is about 3.5 m at the crest and about 35 m at
the foundation for the largest monoliths.
320
Q.104 - R.20
Fig. 1
Ramsele concrete buttress dam. Photo: Christer Vredin.
Photo du barrage hydroélectrique de Ramsele. Photo: Christer Vredin.
Fig. 2
Storfinnforsen concrete buttress dam. Photo: Christer Vredin.
Photo du barrage hydroélectrique de Storfinnforsen. Photo: Christer Vredin.
2.1. OBSERVED CRACKS
In the 1960s, cracks were first observed on the downstream side of the frontal
slabs of both dams (type 1 in Fig. 3). These cracks were located in the inclined
lower part of the frontal slabs. In addition, these cracks had also propagated as
inclined cracks down through the buttress walls towards the foundation (type 2 in
321
Q.104 - R.20
Fig. 3
Sketch of observed cracks in the concrete buttress dams.
Schéma des fissures observées sur les barrages à contreforts.
Fig. 3). Besides these cracks, extensive freeze thaw damages were observed on
the upstream side of the vertical part of the frontal slab, [8].
Based on these cracks and damages in the frontal slabs, structural rehabili-
tation was performed during the 90s. During the structural rehabilitation program,
all damages on the upstream face on the frontal slabs were repaired. On the down-
stream side of the frontal slabs, the frost damaged concrete was removed with
the high-pressure water-jetting method and repaired with shotcrete. The areas with
extensive cracking, i.e. cracks of type (1) and (2) in Fig 3, were grouted. To prevent
further ice formation on the frontal plate, an insulating wall made of polystyrene-
insulated steel sheets was installed. The insulating wall was installed vertically, just
downstream of the inspection gangways, shown with blue color in Fig. 4. The area
between the frontal plate and the insulating wall was heated with warm air flow from
the underground power station [9].
The insulating wall thereby enclosed an area between the buttress walls and
the frontal slab resulting in that about half the monolith was in a climatically controlled
area. About 10 years after the performed rehabilitation program, further cracking
was observed and at this point, several cracks were observed in the buttress wall.
On all monoliths, an inclined crack had propagated from the inspection walkway
towards the frontal slab. Most of these cracks had propagated all the way up to the
frontal slab. In Fig. 3, these types of cracks are denoted as type (3). In addition
to these, it could also be observed that almost vertical, or slightly inclined, cracks
had propagated from the concrete rock interface towards the inspection walkway.
In a few cases, these cracks had propagated all the way up to the walkway. These
types of cracks are called type (4) in Fig. 3. On a few of monoliths, a crack could
also be observed near the dam heel, illustrated as type (5) in Fig. 3. In a few cases,
322
Q.104 - R.20
horizontal cracks can also be found on the downstream side of the buttress wall near
the elevation where the inclined front plate transcends to vertical. This type of crack
is denoted as type 6 in Fig. 3.
2.2. DAM STRENGTHENING AND MONITORING
The most critical monoliths are subjected to cracking to some extent, where
these cracks may result in potential alternative (internal) failure modes compared to
the traditional failure modes for global overturning and sliding. The risk of new poten-
tial internal failure modes resulted in a new extensive strengthening and service
life-extension program were conducted, [10]. Firstly, the monoliths were stabilized
with ground anchored VSL6-19 tendons that were drilled from the downstream part
of the buttress wall, near the crest, and through the buttress wall into the rock foun-
dation, see Fig. 4. The most extensively damaged monoliths were strengthened with
four tendons, each pre-stressed with a force of 3000 kN. In addition, up to four hori-
zontal VSL 6-12 tendons pre-stressed with a force of 2000 kN each, see Fig. 4, were
installed near the dam heel to prevent that toe crack causes movement of the point of
rotation. To further prevent thermal cracking caused by seasonal temperature vari-
ations, the old insulation (blue in Fig. 4) wall was demolished and a new was built
as shown (in red) in Fig. 4. Along with the new insulation wall, new heat exchangers
were installed to create even temperature distribution in the monoliths. The heat
system was designed to ensure that the temperature in the climate-controlled area
within the insulating wall should not decrease below +2◦ C during the winter. Fur-
thermore, the road bridge on top of the crest was widened to allow for heavy traffic
needed during the strengthening works.
3. NUMERICAL MODEL
In this section, the numerical model developed prior to monitoring is presented.

The numerical model was used to study potential failure modes, suitable type of
sensors and their placement. In addition, the numerical models were used to define
warning and alert values for some of the sensors.
3.1. GEOMETRY, MESH AND INTERACTIONS
To determine suitable type and placement of sensors, numerical models were

developed to simulate the expected behaviour of the monoliths under normal con-
ditions and during failure analyses. The six monoliths were chosen for extensive
monitoring were modelled using Abaqus ver. 2019 and Brigade Plus ver. 6.2. In Fig.
323
Q.104 - R.20
Fig. 4
Illustration of the dam with the old insulation wall (blue) and the new (red).
Illustration du barrage avec l’ancien mur d’isolation thermique (bleu) et le nouveau
(rouge).
5, the geometrical model of one of the largest monoliths of Storfinnforsen buttress

dam is shown. In this model, the concrete and rock parts are modelled with 3D solid
elements. All the reinforcement bars were included in the model as individual truss
elements and are assumed to be fully bonded to the concrete. The rock anchored
tendons and the horizontal tendons (crossing the toe crack) are included as beam
elements. The rock anchored tendons are modelled to be able to move inside of
cable ducts while the horizontal tendons are models as cement grouted. The ele-
ment length of the concrete dam, reinforcement and tendons is 0.25 m, while the
element length of the rock foundation varies from 0.25 m at the top to 6 m at the
base. In total, the models include about 1,000,000 degrees of freedom for each of
the studied monoliths.
Fig. 5
Illustration of the numerical model.
Illustration du modèle numérique.
324
Q.104 - R.20
In this paper, the actual crack propagation is not simulated. This have been
studied previously, see [8, 11], where it was shown that these cracks had origi-
nated from the seasonal temperature variations. Instead, in this paper the already
existing cracks are included directly in the model as weakened planes. The crack
mapping of the monolith was included as discrete cracks in the model, with zero
tensile resistance for crack openings, but with a coefficient of friction and able to
transmit compressive stresses. It should be noted that since the dam is reinforced,
the reinforcement bars crossing the cracks will act to prevent crack opening.
In the same way, the fractures detected in the rock mass were also included
as discrete cracks in the same manner. The exact extension of the rock fractures
is not known since it is only based on information from drill holes. Therefore, as a
conservative assumption, the extension of the rock fractures was assumed to have
their most unfavourable inclination.
3.2. MATERIAL PROPERTIES
The concrete and rock have been modelled as linear elastic materials since
these already have the pre-existing cracks and fractures included. The reinforcement
and tendons have been modelled as non-linear in order to be able to simulate the
dam failure. All material properties are presented in Table 1.
Table 1
Material properties used in the simulation.
CONCRETE ROCK REINFORCEMENT TENDONS
K25 (Ks 40) (VSL 6-19)
Density (kg/m3 ) 2300 2600 7800 7800
Elastic modulus (GPa) 25 60 200 195
Poisson’s ratio 0.2 0.2 0.3 0.3
Yield strength (MPa) – – 370 1640
Ultimate strength (MPa) 600 1840
Rupture strain 0,15 0,035
Thermal expansion (◦ C−1 ) 1e-5 1e-5 1e-5 1e-5
The coefficient of friction was assumed to be 0.62 for cracks in the concrete and
was assumed to be 0.73 for rock fractures. The coefficient of friction only accounts
for the base friction as a conservative measure, meaning that the effect from dilation
is neglected.
3.3. BOUNDARY CONDITIONS AND PERMANENT LOADS
The bottom part of the rock foundation is defined with boundary conditions
preventing displacements perpendicular to its surface. The part of the rock above
325
Q.104 - R.20
the bottom rock fracture are restrained to move in the cross-stream direction, while
deformation in the stream direction is counteracted by friction. If the fracture is
below the rock surface at the model edges, an assumed rock wedge in accordance
with Nicholson (1983), [12], is included were the restraining forces are included as
springs.
Both of these dams have large reservoirs, resulting in a the total variation in
water level over a typical year is only 0.2 m. Considering that the hydraulic head is
almost 40 meters, this small variation allows for considering the water level to be
constant over the year in the simulations. Thereby, the permanent loads for this dam
are; gravity load, pre-stressing of tendons, hydrostatic pressure and uplift pressure.
The water level is typically 1.5 m below the bottom of the overpassing bridge. In
case were rock fractures are considered in the rock mass, then a linear decaying
uplift pressure was also applied on both surfaces related to the fracture.
In the numerical model, the gravity load and pre-stressing is applied first to
ensure that friction is developed in the cracks and rock fractures before the horizontal
loads are applied. The effect from hydrostatic pressure and uplift pressure under
the frontal slab is applied in the second step. During these two steps, the loading is
causing the dam to deform and thereby several of the cracks may open.
3.4. VARIABLE LOADS
The main aspects that cause variation in behaviour of the studied dams is
seasonal variations in temperatures and influence of ice pressure from the frozen
top surface of the reservoir during winters. The ambient temperature variations are
well known, where daily historical data for the last 60 years (from 1961) are available
from a meteorological station located in the proximity of the dams. Daily tempera-
ture variations do not propagate far enough into these massive concrete structures
to influence the deformations to any greater extent, [13]. Hence, weekly average
temperatures were calculated, and based on this, the maximum, mean and mini-
mum for each month was collected as seen in Fig. 6. The warmest weekly average
temperature was +20◦ C recorded in July and the lowest of −21◦ C in January. In,
Fig. 6, the maximum curve represents a warm year and the minimum represents a
cold year. In order to reduce the number of simulations required, a combined year
with the lowest winter temperature and the warmest summer temperature assum-
ing a sinusoidal curve in-between these was used to determine suitable placement
and type of sensors. In the case when the numerical results are compared with the
experiments, the results from all three curves in Fig. 6 is used.
The water temperature was assumed to vary over the depth during summer,
where the surface water (top 10 m) was defined as +10◦ C with a temperature of
+4◦ C at a depth greater or equal of 10 m. During winter, the water temperature was
assumed to be 0◦ C over the whole depth. The temperature in the climate-controlled
326
Q.104 - R.20
Fig. 6
Ambient weekly temperature variations based on 60 years of statistics.
Variations hebdomadaires de la température ambiante basées sur 60 ans de
statistiques.
area enclosed by the new insulation wall and the frontal plate was defined to be
+2◦ C when the heaters are on and was calculated to be +12◦ C during summer.
The pressure from the frozen ice sheet on the reservoir is a great source of
uncertainty both regarding its size and variation. Measurements by [14] has shown
that it may have daily variation if there are medium to large daily variations in reservoir
level. In this case, the variations in water level is minor, hence, the ice load is likely
mainly influenced by the temperatures. In order to estimate the influence of the ice
pressure, two separate analyses were performed
1) Ice load was excluded

2) Maximum ice load (according to the design guideline, [15]) of 200 kN/m.
In the analysis with the maximum ice load, it will be assumed to fully active
from December to February. During November, the ice load increase from zero to
200 kN/m and during March the ice load decrease from 200 kN/m to 0.
327
Q.104 - R.20
3.5. FAILURE ANALYSIS
The failure analysis is assumed to start either during winter or during summer
conditions. In normal stability analyses, the temperature effects are not considered,
however in this case it is important since the failure mode may differ, and different
factors of safety may be obtained depending when the failure occurs. In the fail-
ure analyses, the load effect from hydrostatic water pressure, uplift and ice load is
increased until the monoliths reach failure. It is only the negative load effect that is
increased, i.e. the overturning loads. For the water pressure, for instance, the frontal
slab is inclined resulting in a restraining vertical force. Thereby, this means that only
the horizontal component of the water pressure load is increased. All overturning
loads are increased with the same rate, which means that the percentage of the
load increase at failure corresponds to the definition of global safety that is defined
for sliding and overturning failure modes.
In order to minimize the risk of missing potential failure modes, additional anal-
yses were also performed, where only one of the overturning loads were increased
at a time until a failure occurred. This does, however, not correspond to the current
definition of safety factor, hence it is not possible to determine a factor of safety for
these analyses. In many of the analyses, the failure mode was initiated and governed
by fractures in the rock. Therefore, analyses were also performed where these rock
fractures were inactivated in order to ensure a failure in the concrete monolith or at
the dam-foundation interface.
3.6. RESULTS
The results from the numerical analyses showed the expected variation in the
dam behaviour and also showed where and how different failure modes initiates and
develops. This is a load-controlled simulation, where the loads are increased contin-
uously until failure is reached. This means however that the progression of failure is
not possible to capture. Based on these analyses it was possible to identify suitable
type and placement of sensors for monitoring different potential failure modes. In
Fig. 7, two examples of potential failure modes are illustrated.
The analyses showed that one of the primary features to monitor was the crest
displacement of the monolith, which was also shown by [16]. The crest will start to
deform in relation to the rock foundation for the following potential failure modes;
• Internal failure modes caused by cracks in the monolith, where the downstream
part of the monolith fails
• Sliding failure in a crack or at the concrete-foundation interface
• Overturning failure
328
Q.104 - R.20
Fig. 7
Illustration of two potential failure modes, failure due to sliding in a rock fracture
(left) and internal failure of the monolith due to cracks (right).
Illustration de deux modes de rupture potentiels, rupture due au glissement dans
une fracture rocheuse (à gauche) et rupture interne du monolithe par fissuration (à
droite).
In the potential failure modes that are related to crack openings in the mono-
lith, it is of course also important to monitor these crack openings or crack sliding
displacements since they may indicate a deviation in normal behaviour significantly
earlier than the crest displacement.
For some of the potential failure modes, other features need to be monitored
since the crest displacement may not be able to give an imminent warning. Examples
of these failure modes are;
• Failure in rock fractures (sliding or overturning)

• Local failure of the frontal slab
• Internal failure caused by cracks in the monolith, where the upstream part of
the monolith fails
For the failures related to the rock fractures, monitoring these can be achieved
by using extensometers. If there are several rock fractures, then the analyses
showed that it is best if these individual fractures can be monitored individually,
since a deviation will be detected earlier and also that it can be used to track where
the failure mode is initiated.
In case the failure mode is related to local failure of the frontal slab, the analyses
showed that it is suitable to monitor the relative displacement of frontal slabs in two
adjacent monoliths in order to detect such a local failure.
329
Q.104 - R.20
The analyses also showed that the uplift pressure in the rock had a significant
influence on the global safety of the monoliths. Hence, it was deemed important to
monitor the uplift pressure. Another important aspect was that the toe crack could
potentially reduce the safety of the monolith significantly if the horizontal tendons
could not prevent the movement. If movement occurred in the toe cracks, then the
point of rotation for an overturning failure could move several meters upstream.
Hence, for these cracks it was deemed important to monitor both the crack opening
and crack sliding displacement.
In Fig. 8, calculated variation in crest displacement is shown for the combined

case with warm summer (+20◦ C) and cold winter (−21◦ C). Positive displacement
is defined to be in the downstream direction. The figure shows the results for both
an analysis without ice load (green) and an analysis with ice load (blue). As it can
be seen in the results, the expected variation in crest displacement is only almost 2
mm without consideration of ice loads and about 3.5 mm with ice load considered. It
should be noted that the seasonal variation in temperature results in small variations
in displacements due to the newly installed insulation wall that encloses the whole
downstream part of the dam. Thereby, the ice load constitutes a significant part of
Fig. 8
Calculated variation in crest displacement for a combined year with warm summer
and cold winter.
Simulation de la variation du déplacement de la crête pour une année combinant
été chaud et hiver froid.
330
Q.104 - R.20
the total crest displacement over a year if it is as large as specified in the design
guideline.
In Fig. 8, the calculated displacement in the failure analysis corresponding to

when yielding of the first reinforcement bar is initiated in the models are shown as
two horizontal lines, for winter and summer conditions respectively. It can be seen
in the figure, that about 2 – 3 mm of additional deformation would be required to the
crest before yielding occurs under both winter and summer conditions respectively.
Yielding of the reinforcement corresponds to permanent deformations and would
result in significant increase in crack openings. This is not considered to be accept-
able for normal conditions. Thereby, suitable warning levels should be defined lower
than these values. It should be noted that the calculated factor of safety when yield-
ing of the first reinforcement bar occurs is about 1.4, i.e. the overturning loads can
be increased 40 % before this occurs.
The crest displacement prior to the initiation of the failure is not shown in
this figure. As the maximum load capacity is reached a ductile failure progression
is obtained with significantly increasing displacements. This is governed primarily
by yielding of the reinforcement prior to rupture. The deformation prior to failure is
significantly larger and corresponds to more than 100 mm. At this deformation, it is
likely that significant increase in leakage may have occurred. The calculated factor
of safety when the internal failure in the monolith occurs is 2.8.
4. MEASUREMENTS
Based on the results from the numerical analyses, all six of the test mono-
liths were installed with several sensors to monitor their behaviour. All of these are
considered to be detectors of potential failure modes;
• Pendulum to monitor the crest displacement, (one hanging pendulum per

monolith).
• Crack width sensors, typically six per monolith, to monitor the crack openings
and crack sliding
• 3D displacement sensors that measured the relative displacement between
two adjacent frontal plates. Typically, six per monolith.
• Pore pressure measurements to monitor the uplift pressure, typically two
measurements between two adjacent monoliths.
• Short extensometers to monitor sliding at the concrete- foundation interface.
Typically, two per monolith
• Long extensometers to monitor displacements in the rock fractures, typically,
two per monolith
Besides these sensors, the leakage is also very important to measure. This
was of course not captured in the numerical model but was installed in addition to
331
Q.104 - R.20
the sensors indicated by the numerical analyses. The leakage is measured with
Thompson weirs on several locations of the dams.
In addition to these sensors, some other features were considered to be impor-

tant to monitor in order to gather information required to validate assumptions and
to use in future analyses. These types of sensors are called support instrumentation
and are for example,
• Temperatures in the concrete or in the air in the temperature-controlled area

enclosed by the insulation wall
• Temperature in the reservoir at different depths
• Relative humidity in the enclosed air
Fig. 9
Examples of installed sensors; hanging pendulum (top), 3D deformation sensor
between two frontal plates (bottom left) and 1D crack with sensors (bottom right).
Exemples de capteurs installés ; pendule suspendu (en haut), capteur de
déformation 3D entre deux plaques frontales (en bas à gauche) et extensomètres
(capteurs d’ouverture de fissure) (en bas à droite).
332
Q.104 - R.20
Some examples of installed sensors are shown in Fig. 9. In the top figure, the
hanging wire of the pendulum is shown together with the laser unit that records the
displacements. In the bottom left figure, an example of a 3D relative deformation
sensor is shown measuring the relative displacement between the frontal plats of
two adjacent monoliths. Since this sensor measures the relative difference, it should
be close to zero if the dam behaves as expected. In the figure, the sensors are
enclosed in a steel casing and with a roof to prevent moisture reaching the sensor.
In the bottom right corner, three 1D sensors are shown that monitor the crack width
opening of type 6 cracks (see Fig. 3) in the buttress wall of one of the monoliths.
These sensors are installed perpendicular to the crack width since they are intended
to capture the crack width opening.
The measurements have been recorded since February 2019 and thereby
recordings of up to 1.5 years are available so far. Initially, some problems have
occurred for few sensors with loss of recordings which resulted in incomplete signals
etc. During function control tests of the sensors, these initial problems could be
resolved. The recorded measurements have so far shown behaviour that is quite
close to the expected, with small variations in crest displacements and crack widths.
In Fig. 10, the raw data (unfiltered) of the measured crest displacement in one of
the largest monoliths of Storfinnforsen is shown. As it can be seen in the figure, the
annual variation in crest displacement is about 2 – 3 mm, i.e. in the same range as
the numerical analyses presented earlier.
Fig. 10
Measured crest displacement of monolith M43 of Storfinnforsen.
Mesure du déplacement en crête du monolithe M43 du barrage de Storfinnforsen.
333
Q.104 - R.20
The support instrumentation, especially regarding the temperatures have how-

ever shown that the previous assumed temperature variations presented in Section
3.4 did not fully correspond to the measured temperatures during 2019-2020. The
ambient temperature, especially for the winter of 2019/2020, was higher than recent
years. It was however still within the defined confidence intervals based on the
previous 60 years. The largest deviation was in the temperature in the temperature-
controlled area enclosed by the insulation wall. According to Section 3.4, this was
assumed to vary between 2◦ C (winter) and 12◦ C (summer). The measured temper-
ature showed however that the temperature was significantly higher, maybe caused
by the relatively warm winter. During the winter the temperature was about 8◦ C and
the temperature during the summer was typically 14◦ C as seen in Fig. 9, but may
vary within the enclosed area along the dam but also over the dam height.
Fig. 11
Example of temperature variation in ambient (outdoor) and in the
climate-controlled area (indoors). From [6].
Exemple de variation de température ambiante (extérieure) et dans la zone
climatisée (intérieure). Extrait de [6].
5. VERIFICATION OF THE NUMERICAL MODEL
In Fig. 12 the measured crest displacement (orange) is presented along with

the calculated displacement. In the figure, the calculated response was obtained
based on the three temperature curves previously presented in Fig. 6. The upper
bound thereby represent a “warm year” while the lower bound represent a “cold year”.
Fig. 12 presents the calculated results starting from a generic year from January to
December, while the measured results from February 2019 to December 2019 is
plotted.
334
Q.104 - R.20
Fig. 12
Comparison of measured and calculated displacements. From [6].
Comparaison entre les déplacements mesurés et simulés. Extrait de [6].
In Fig. 13 a similar comparison between the measured and calculated varia-

tion in crack width is presented. The result is here presented only for the average
temperatures.
Fig. 13
Comparison of measured and calculated crack width. From [6].
Comparaison entre la largeur de fissure mesurée et simulée. Extrait de [6].
6. CONCLUSIONS

especially as many dams are approaching their expected service life. It is important
that sensors are installed in strategic positions to give early indications if the dam
behaviour changes and early warnings of potential dam failures.
335
Q.104 - R.20
In this paper, a new instrumentation project has been presented where numer-
ical models based on the finite element method were used to plan the monitoring.
The use of numerical models is a great tool for determine placement of sensors
and type of sensors to use. The numerical models are also suitable to predict future
dam behaviour which is important for determine required sensitivity and accuracy
of the sensors. The results from the analyses are also compared with the measured
behaviour for the first 1.5 year. The results show that detailed FE models are good
for defining warning levels early in the project and that the simulations can predict
the dam behaviour with high accuracy. Later on, when sufficient amount of mea-
sured data is available data-based regression or machine learning models can be
utilized to define even more precise warning levels. It should be noted that FE mod-
els can also be used for definition of alert values, i.e. indications of potential failures,
in contrast to the data-based models since these cannot be used for extrapolation.
Accurate predictions are necessary for supporting the owner in decision making and
planning future maintenance.
REFERENCES
[1] WILLM, G.; BEAUJOINT, N.; Les méthodes de surveillance des barrages
au service de la production hydraulique d’Electricité de France, problèmes
anciens et solutions nouvelles (The methods of surveillance of dams to serve
hydraulic Production at Électricité de France: Old problems and new solu-
tions). In: 9th International Congress on Large Dams (ICOLD), Q34/R30,
pages 529–550, Istanbul, Turkey, 1967.
[2] CHOUINARD, L. E.; BENNETT, D. W.; FEKNOUS, N. Statistical Analy-

sis of Monitoring Data for Concrete Arch Dams. Journal of Performance
of Constructed Facilities, 9(4):286–301, 1995. DOI: 10.1061/(ASCE)0887-
3828(1995)9:4(286).
[3] LÉGER, P.; LECLERC, M. Hydrostatic, Temperature, Time-Displacement

Model for Concrete Dams. Journal of Engineering Mechanics, Vol. 133 (3),
March, 2007, pp. 267–277.
[4] SALAZAR, F.; MORÁN R.; TOLEDO, M.Á.; OÑATE, E. Data-Based Models of
the Prediction of Dam Behaviour: A Review and Some Methodological Con-
siderations. Archives of Computational Methods in Engineering, 24, 2017, pp.
1–21.
[5] HELLGREN, R., MALM, R., ANSELL A. Performance of data-based models

for early detection of damage in concrete dams. Structure and Infrastructure
Engineering, 2020. DOI: 10.1080/15732479.2020.1737146.
[6] ISAKSSON, N., & SANDSTRÖM, L. Behavior prediction of concrete dams.

MSc thesis, KTH Royal Institute of Technology, 2020.
336
Q.104 - R.20
[7] NORDSTRÖM E., MALM, R., BLOMDAHL, J., TORNBERG R., NILSSON, C-
O. Optimization of Dam Monitoring for long concrete buttress dams. In: ICOLD
25th Congress in Stavanger, Q99, 13–20 June, 2015
[8] MALM, R. & ANSELL, A. Cracking of Concrete Buttress Dam Due to Seasonal
Temperature Variation. ACI Structural Journal, 108(1), pp 13-22. 2011. DOI
10.14359/51664198.
[9] ERIKSSON, H., “Investigation and Rehabilitation of the Storfinnforsen

Dam,”ICOLD 18th International Congress on Large Dams, V. 1, Q68,
R19,1994, pp. 247–259.
[10] ISANDER, A., NILSSON, C-O., MALM, R. Evaluation and Structural Rehabil-
itation of Storfinnforsen Concrete Buttress Dam. In: CFBR Modernisation des
barrages, 4–5 December, Chambery, France, 2013.
[11] ANSELL, A. & MALM, R. Modelling of thermally induced cracking of a concrete

buttress dam. Nordic Concrete Research, vol. 38, pp. 69–88, 2008.
[12] NICHOLSON, G. A. Design of Gravity Dams on Rock Foundations: Sliding

Stability Assessment by Limit Equilibrium and Selection of Shear Strength
Parameters. Technical Report GL-83-13. Washington, DC: US Army Corps of
Engineers, 1983.
[13] MALM, R. Guideline for FE analyses of concrete dams. Energiforsk report

2016:270. Energiforsk, 2016.
[14] HELLGREN, R., MALM, R., FRANSSON, L., JOHANSSON, F., NORD-
STRÖM, E. & WESTBERG-WILDE, M. Measurement of ice pressure on a con-
crete dam with a prototype ice load panel. Cold Regions Science and Technol-
ogy, vol 170, 102923, 2020. https://doi.org/10.1016/j.coldregions.2019.102923
[15] RIDAS. Swedish Hydropower companies guidelines for dam safety, application
guideline 7.3 Concrete dams (In Swedish). Svensk energi, 2017.
[16] MALM, R., NORDSTRÖM E., NILSSON C-O., TORNBERG R. & BLOMDAHL
J. Analysis of potential failure modes and re-instrumentation of a concrete dam.
In: ICOLD Symposium - Appropriate Technology to Ensure Proper Develop-
ment, Operation and Maintenance of Dams in Developing Countries, 84th
ICOLD Meeting, Johannesburg, 15–20 May, 2016.
337
Q. 104 - R21
-------
-------
DAM CONCRETE AUTOMATIC PLACEMENT SYSTEM – COMPLETE

AUTOMATION OF DAM CONCRETE MATERIALS SUPPLY, PRODUCTION,
TRANSPORT, AND PLACEMENT (∗)
Norikazu YAMASHITA & Toshihisa KASE

Dam Department, Civil Engineering Technology Division, SHIMIZU Corporation
Subaru TACHIHANA
Civil Engineering Production Planning Department, International Branch, SHIMIZU
Corporation
Shinobu MORIYAMA & Hideo MORI

Construction Manager Civil Engineering Department, Tohoku Branch, SHIMIZU
Corporation
JAPAN
SUMMARY
The concrete placement of concrete gravity dams accounts for about 60% of
the construction cost and about 50% of the construction time. The operations for
the concrete of production, transport, and placement are repeated for several years.
The concrete placement also requires many experienced skilled workers for the
operation of large heavy machines and the compaction work. When rail rope cable
cranes are adopted for the concrete transport, the crane manipulation becomes a
complex 3-dimensional operation, so greater experience and skill than usual are
required. Therefore, we developed a "dam concrete automatic placement system"
that aiming to automate the series of processes of a material supply to the batching
plant, a concrete production, a concrete transport by a transfer car, and a transport
∗ Système de mise en place automatique du béton pour grands barrages - Automatisation
complète des opérations d’alimentation des matériaux, de malaxage, de transfert et de pose

du béton
338 DOI: 10.1201/9781003211808-21

Q. 104 - R21
and placement using a rail rope cable crane. This system integrates the automation
technology that has previously been applied to the batching plant and transfer car,
etc., with automation of the rail rope cable crane. This system was applied to about
40,000 m3 in the 213,000 m3 of total concrete using the rail rope cable crane on Yana-
gawa Dam (Morioka City, Iwate Prefecture), and the following effects were obtained
as a result of its introduction. First, the cycle time per placement was reduced by
about 20 seconds from 3 minutes 40 seconds to about 3 minutes 20 seconds, a 10%
reduction in time. The placement time for the approximately 40,000 m3 to which this
system was applied was reduced from 533 hours to 494 hours. Secondly, the num-
ber of workers required for placement in the dam was 2 crane operators, 6 batching
plant operators, and 7 for concrete placement, or a total of 15 workers. However, by
adopting this system it is expected that this conventional number of workers can be
reduced to 2/3, with 1 crane operator, 4 batching plant operators, and 5 concrete
placement workers, or a total of 10 workers. Thirdly, by automating operation of rail
rope cable cranes with this technology it is possible to automatically and rationally
transport the concrete bucket to the placement position, so the operation can be per-
formed with the same work efficiency as a proficient operator despite the shortage
of experienced operators, so this technology can help to respond to the shortage of
proficient operators.
RÉSUMÉ
La mise en place en béton du corps des barrages-poids en béton représente

environ 60% du coût de construction et environ 50% du temps de construction.
Les opérations pour le béton de production, de transport et de mise en place sont
répétées pendant plusieurs années. La mise en place du béton nécessite égale-
ment de nombreux ouvriers qualifiés expérimentés pour l’exploitation de grosses
machines lourdes et les travaux de compactage. Lorsque des grues à câble à câble
ferroviaire sont adoptées pour le transport du béton, la manipulation de la grue
devient une opération tridimensionnelle complexe, ce qui nécessite une expéri-
ence et une compétence plus grandes que d’habitude. Par conséquent, nous avons
développé un “ système de placement automatique du béton de barrage ” qui vise
à automatiser la série de processus d’un approvisionnement en matériau à la cen-
trale, une production de béton, un transport de béton par une voiture de transfert,
et un transport et un placement à l’aide d’un câble ferroviaire grue à câble. Ce sys-
tème intègre la technologie d’automatisation précédemment appliquée à l’usine de
dosage et à la voiture de transfert, etc., avec l’automatisation d’une grue à câble
sur rail. Ce système a été appliqué à environ 40000 m3 dans les 213000 m3 de
béton total à l’aide d’une grue à câble sur rail sur le barrage de Yanagawa (ville de
Morioka, préfecture d’Iwate), et les effets suivants ont été obtenus à la suite de son
introduction. Premièrement, le temps de cycle par placement a été réduit d’environ
20 secondes de 3 minutes 40 secondes à environ 3 minutes 20 secondes, soit une
réduction de 10% du temps. Le temps de mise en place des quelque 40 000 m3
auxquels ce système a été appliqué a été réduit de 533 heures à 494 heures.
339
Q. 104 - R21
Deuxièmement, le nombre de travailleurs requis pour la mise en place dans le

corps du barrage était de 2 grutiers, 6 opérateurs de centrales à béton et 7 pour
la mise en place du béton, soit un total de 15 travailleurs. Cependant, en adoptant
ce système, on s’attend à ce que ce nombre conventionnel de travailleurs puisse
être réduit à 2/3, avec 1 grutier, 4 opérateurs de centrale et 5 agents de mise en
place de béton, soit un total de 10 travailleurs. Troisièmement, en automatisant le
fonctionnement des grues à câble à câble avec cette technologie, il est possible de
transporter automatiquement et rationnellement le godet à béton vers la position de
placement, de sorte que l’opération puisse être effectuée avec la même efficacité
de travail qu’un opérateur compétent malgré la pénurie d’opérateurs expérimentés,
donc cette technologie peut aider à répondre à un manque d’opérateurs compétents.
1. INTRODUCTION
The productivity improvement by applying information technology is a recent

trend in the whole construction industry.
The concrete placement of concrete gravity dams accounts for about 60%
of the construction cost and about 50% of the construction time. The operations for
the concrete of production, transport, and placement are repeated for several years.
The concrete placement also requires many experienced skilled workers for the
operation of large heavy machines and the compaction work. When rail rope cable
complex 3-dimensional operation, so greater experience and skill than usual are
required.
It is expected that an automation of operations in the concrete placement will

have a great effect in increasing the efficiency of repetitive operations, raising pro-
ductivity, and reducing the demand for proficient skilled workers, which is a concern
at the present. Therefore, we developed a "dam concrete automatic placement sys-
tem" that aiming to automate the series of processes of a material supply to the
concrete production plant (hereafter referred to as "batching plant"), a concrete pro-
duction, a concrete transport by a transfer car, and a transport and placement using
a rail rope cable crane [1], [2]. This system was introduced in the Yanagawa Dam
(Morioka City, Iwate Prefecture), where the rail rope cable crane is used for concrete
placement, as a verification test to verify its effectiveness.
2. OVERVIEW OF THE CONSTRUCTION
The Yanagawa Dam is a concrete gravity dam currently being constructed on

the Yanagawa River of the class A river, the Kitakami River (Table 1, Fig. 1). It has a
height of 77.2 m, length along the crown of 249 m, and body volume of 228,500 m3 .
340
Q. 104 - R21
Table 1
Yanagawa Dam
OWNER: IWATE PREFECTURE
Construction name: Yanagawa Dam Construction Project
Location: Kawame, Morioka City, Iwate Prefecture
Construction period: December 10, 2014 to March 31, 2021 (Fig.1)
Contractor: Shimizu, Konoike, Hiranogumi Special Construction Joint Venture
Purpose: Flood control, River maintenance water
Conservation of the river environment
Water supply, Hydroelectric power generation
Details of construction: River diversion (Diversion tunnel): 285 m Dam
Foundation excavation: 168,000 m3
Dam construction: 228,500 m3
Foundation processing: whole set
Main materials: Aggregates for concrete 525,100 t, cement 47,500 t
Fig. 1
Yanagawa dam overall construction schedule.
Calendrier général de construction du barrage de yanagawa.
1 Article de construction 8 Construction divers
2 Préparation 9 Usine temporaire
3 Déviation de la rivière 10 Test d’inondation
4 Excavation de foundation 11 Mise en place du béton du
5 Construction du corps Corps du barrage
6 Traitement des fondations 12 Période de développement
7 Usine auxiliaire d’octobre 2017 à juin 2019
3. YANAGAWA DAM CONSTRUCTION PLANT
The construction plant in the Yanagawa Dam includes a concrete materials

storage plant with aggregate storage bins and a cement silo (Figs.2,3). The aggre-
gates storage plant (storage bins) includes one each for coarse aggregates G1
341
Q. 104 - R21
Fig. 2
Yanagawa Dam construction plant layout.
Plan de l’usine de construction du barrage de Yanagawa.
1 Aggregates storage plant, 1 Installation de stockage
Aggregate bins d’agrégats, bacs d’agrégats
2 500t cement silo 2 Silo à ciment 500t
3 2.5m3 batching plant 3 Centrale à béton de 2.5 m3
4 15.5t rail rope crane 4 Grue à câble sur rail 15.5 t
(150 mm-80 mm), G2 (80 mm-40 mm), G3 (40 mm-20 mm), and G4 (20 mm-5 mm),
and two for fine aggregates S (5 mm-).
A belt conveyor (width 750 mm) is used for transport from the aggregate
storage bins to the batching plant.
One 2.5 m3 forced biaxial mixer in the batching plant, one transfer car for
concrete transport, one 5 m3 concrete bucket, and one 15.5 t rail rope cable crane
are installed.
4. CONCRETE AUTOMATIC PLACEMENT SYSTEM
4.1. OVERVIEW
We developed the concrete automatic placement system using information

technology. This system works according to the placement plan of the concrete,
342
Q. 104 - R21
Fig. 3
Yanagawa Dam construction plant diagram.
Schéma de l’usine de construction du barrage de Yanagawa
1 Aggregate storage bins 1 Bacs de stockage d’agrégats
2 Batching plant, Cement silo 2 Centrale à béton, silo à ciment
3 Rail rope cable crane 3 Grue à câble sur rail
4 Transfer car, Concrete bucket 4 Voiture de transfert, godet en béton
containing the 3D coordinates of the placement position, the type of concrete mix-
ture, the quantity, etc. as shown in Fig.4. When this placement plan is input to the
concrete automatic placement system, the series of operations at each plant, such
as the materials supply plant (aggregates storage plant, cement silo), the batching
plant, the transfer car, and the rail rope cable crane, from production of the dam
concrete to transport and placement are repeated automatically.
343
Q. 104 - R21
Fig. 4
Concept of the automatic system.
Concept du système automatique.
1 Mixture 1 Mélange
2 Placement quantity 2 Quantité de placement
3 Placement position 3 Position de placement
4 Placement elevation 4 Élévation de l’emplacement
5 Automatic operation 5 Opération automatique
The locations of each placement position are described three-dimensionally

along with the corresponding concrete volume based on the CAD drawings of
the dam in the placement plan. The elevation for the concrete to be released is
determined based on the lift height.
At the batching plant, when the placement plan is entered into a control com-
puter, the concrete is produced to mixture and quantity in accordance with the
plan.
At the rail rope cable crane, the placement position and elevation (X, Y, Z) from
the placement plan is input to the control computer, and the concrete is automatically
transported to the position and elevation corresponding to the placement sequence.
344
Q. 104 - R21
4.2. AUTOMATION OF CONCRETE MATERIALS SUPPLY
Prior to the placement, the materials of the concrete such as fine aggregates,
coarse aggregates, cement, mixing water, etc. are automatically supplied from the
aggregate storage bins, cement silo, etc. to the top area of bins in the batching plant,
where materials are separately stored to the corresponding bins (Fig.5). A turn head
signal monitored by a sensor at the inlet of each aggregate storage bin control the
sorting of the aggregates in the top bins automatically.
A mechanism is constructed so that once any materials within the top head in
the batching plant is consumed as the placement progresses, a supply signal is emit-
ted from a residual quantity sensor within the bin, and the material is automatically
supplied from the aggregate storage bins, the cement silo, etc.
Fig. 5
Automation of aggregates supply.
Automatisation de l’approvisionnement en granulats.
1 Discharge of aggregate at the bottom of 1 Décharge d’agrégats au fond d’un bac
an aggregate storage bin de stockage d’agrégats
2 Transport of the aggregate by belt con- 2 Transport de l’agrégat par convoyeur à
veyor bande
3 Sorting of aggregates by a turn head 3 Tri des agrégats dans une tête tournante
345
Q. 104 - R21
4.3. AUTOMATION OF CONCRETE PRODUCTION
In the batching plant, the concrete is mixed automatically based on the

placement plan (Fig.6). The quantities of each concrete material are automatically
weighed according to the mixture type designated. The surface water content in
the fine aggregate is automatically measured by a non-contact water content meter,
and the quantities of the fine aggregate and the mixing water are adjusted. After
weighing of the materials is completed, the materials are automatically transported
in the mixer and mixed. Then the concrete produced is dropped into a hopper below
the mixer.
Fig. 6
Automation of aggregate supply to concrete production.
Automatisation de l’approvisionnement en granulats pour la production de béton.
1 Weighing aggregate in a weighing tank 1 Agrégat de pesage dans une cuve de
pesage
2 Concrete mixing in the mixer 2 Mélange de béton dans le malaxeur
4.4. AUTOMATION OF CONCRETE TRANSPORT AND PLACEMENT
After confirming that the quantity of concrete has been loaded in to the transfer
car in accordance with the placement plan by the weight sensor, the transfer car
heads towards the concrete bucket (Fig. 7). After confirming that the concrete bucket
has been returned to the bucket platform, the concrete is automatically dropped from
the transfer car into the bucket.
The bucket containing the concrete is automatically transported to the place-

ment position stated in the placement plan by the rail rope cable crane. The control
of the bucket position is performed by converting the 3D coordinates (X, Y, Z) into the
amount of winding of the wires of the cable crane. Also, a Global Navigation Satel-
lite System device is installed on the bucket, to double-check the position. After the
bucket has been moved to the placement position stated in the placement plan,
the bucket is lowered to a height from which the concrete can be placed, and the
concrete is discharged.
346
Q. 104 - R21
When the system was initially introduced, it was not possible to stabilize the
tension force in the main cables and track cables, so it was difficult to control the
crane position and about 5 m errors occurred. Therefore, a tension force mainte-
nance device was introduced, which brought about an improvement, and stable
maintenance of the tension was achieved. As a result, it became possible to control
the crane position within about 50 cm. This positional control improvement took four
months.
The oscillations associated with acceleration and deceleration when the cable
crane was moved laterally were minimized by constructing the control program by
reference to the operations of experienced crane operators. The rise of the bucket
due to the reaction when the concrete was released was likewise controlled by
constructing an algorithm to maintain the bucket height constant by reference to the
operations of experienced operators.
The system can be instantaneously switched from automatic to manual when

the operator moves the operation levers, to take the need to respond with the manual
operation in an emergency into consideration. Therefore, the manual response is
Fig. 7
Automation of concrete transport and placement.
Automatisation du transport et du placement du béton.
1 Loading into the transfer car 1 Chargement dans la voiture de
transfert
2 Transport of the concrete in the transfer 2 Transport du béton dans la voiture de
car transfert
3 Loading into the concrete bucket 3 Chargement dans le godet à béton
4 Concrete discharge and placement 4 Décharge et mise en place du béton
347
Q. 104 - R21
possible anytime in an emergency, so a high degree of safety is provided with the

system.
4.5. CONSTRUCTION OF AN INTEGRATED AUTOMATIC MANAGEMENT SYSTEM
4.5.1. Issues for Constructing an Integrated System
This system automates the series of operations from materials supply and
production to transport and placement, so it was necessary to integrate and control
all signals from various types of machine. The various types of plant for aggregates
and cement silo and transport, the batching plant, the transfer car, the concrete
bucket, and the rail rope cable crane were procured from specialist manufacturers,
the issue is the construction of an integrated system which manages various plants
sending out signals of analog or digital type.
4.5.2. Resolving the Issues Regarding Constructing an Integrated System
Integration software was designed from the start and a system integration com-
puter was introduced for integrated control of plant from multiple manufacturers and
having different types of signals. For this purpose the various types of signal were
assumed to be digital, and a analog/digital converter were installed on plant using
analog signals. In the development of the integration software it was necessary to
check the signals transmitted from the various machinery, and design the sequence
of signals during the automatic operation. Development of the integrated system
took six months. During the introduction of the system on site it was necessary to
modify and improve the sequence of receipt and transfer of signals, and commis-
sioning took two months. The integrated system was completed eight months after
start of development.
5. DAM CONCRETE OVERALL MANAGEMENT SCREEN
As part of the development of the dam concrete automatic placement system,

a "dam concrete overall management screen" for visualizing the status of concrete
placement was produced and introduced.
On this screen the status of progress is displayed in real time based on the
operational signals output from each plant and the produced placement plan (Fig. 8).
348
Q. 104 - R21
Fig. 8
Dam concrete overall management screen (normal screen).
Écran de gestion globale du béton du barrage (écran normal).
1 Start time 19:00 scheduled 1 Heure de début 19:00 achèvement
completion 05:00 prévu 05:00
2 Placement lift 65LF 2 Élévateur de placement 65LF
(272.8 m) (272.8m)
3 Placement speed 75.0 m3 /h 3 Vitesse de placement 75.0 m3/h
4 Placement block 13-14BL 4 Bloc de placement 13-14BL
5 Today’s placement quantity Present 5 Aujourd’hui quantité de placement
595.50 m3 Target 600.00 m3 Présent 595,50 m3 Objectif
600,00 m3
6 Top bins 6 Bacs supérieurs
7 Weighing tanks 7 Pesage des cuves
8 Mixer 8 Mixer
9 Hopper 9 Hopper
10 Mixture B:2.25m3 10 Mélange B: 2.25m3
11 Transfer car 11 Transfert voiture
12 Rolling train 12 Train roulant
The status of supply of the various aggregates and cement in the aggregate
storage bins and cement silo are displayed. Also, the temperature of the materials
measured by temperature sensors installed in the aggregate storage bins, cement
silo, and weighing tanks in the batching plant can be displayed on a separate screen
(Fig. 9). For the batching plant, the status of weighing materials before mixing, the
mix quantities, the mixing status of the concrete within the mixer, and the temperature
of each material and the mixing concrete temperature are displayed. Different colors
are allocated to each concrete mixture type as green for mixture A, red for mixture B,
blue for mixture C, orange for mixture D, gray for mixture H and yellow for mortar,
enabling that the concrete mixture during production can be visually recognized.
349
Q. 104 - R21
Fig. 9
Dam concrete overall management screen (temperature management screen).
Écran de gestion globale du béton du barrage (écran de gestion de la
température).
1 Start time 08:30 scheduled completion 1 Heure de début 08:30 achèvement
18:00 prévu 18:00
2 Placement lift 65LF (272.8 m)-67LF 2 Ascenseur de placement 65LF
(274.3m) (272.8 m)-67LF (274.3 m)
3 Placement speed 75.0 m3 /h 3 Vitesse de placement 75.0 m3 /h
4 Placement block 13-14BL 4 Bloc de placement 13-14BL
5 Today’s placement quantity Present 5 Quantité de placement du jour
1090.00 m3 Target 1110.00 m3 Présent 1090.00 m3 Objectif
1110.00 m3
6 Storage plant 6 Usine de stockage
7 Batching plant 7 Centrale à béton
8 Weighing tanks 8 Pesage des cuves
9 Mixer 9 Mixer
10 Hopper 10 Hopper
The transfer car moves toward the bucket while confirming the concrete loading
status. When the transfer car sends the loading information to the " dam concrete
overall management screen ", the transfer car on the screen also moves toward the
bucket. After arriving at the bucket platform, the transfer car unloads the concrete
and returns under the hopper of the batching plant.
The bucket position and the status of concrete placement are displayed using
the movements of the rail rope cable crane. The locations where the concrete place-
ment has been completed are colored on a 3D model of the dam, so the status of
placement progress can be easily confirmed visually.
350
Q. 104 - R21
This screen can be monitored using a tablet computer, enabling the concrete
mixture being produced or transported or the quantity placed to be easily confirmed
visually. Also, the status of placement is recorded, so it can be applied to the concrete
placement traceability management.
6. EFFECT OF INTRODUCTION OF THE DAM CONCRETE AUTOMATIC

PLACEMENT SYSTEM
At the Yanagawa Dam, this system was applied to about 40,000 m3 of the
dam concrete, the other 188,500 m3 was constructed by the traditional manual
operation from concrete production, transport, and placement. The following effects
were obtained as a result of its introduction.
6.1. SHORTENING OF CONCRETE PLACEMENT TIME
It is possible to automatically transport the concrete bucket to the unloading

position, and it is not necessary to fine-tune the unloading position above the dam,
so the cycle time is shortened.
Also, the placement direction at the top area of the dam usually changes from
the left-right bank direction to the upstream-downstream direction, so it would require
the time for the movement of a traveling trolley. However, as a result of introduction
of this technology the traveling trolley can be automatically adjusted to the next
placement position during the time from lowering of the concrete until arrival at the
bucket platform, so the cycle time can be reduced. As a result of introduction of
this system the cycle time per placement was reduced by about 20 seconds from 3
minutes 40 seconds to about 3 minutes 20 seconds, a 10% reduction in time. The
placement time for the approximately 40,000 m3 to which this system was applied
was reduced from 533 hours to 494 hours.
6.2. REDUCTION IN NUMBER OF WORKERS (LABOR-SAVING)
The number of workers required for placement in the dam was 2 crane oper-
ators, 6 batching plant operators, and 7 for concrete placement, or a total of 15
workers. However, by adopting this system it is expected that this conventional
number of workers can be reduced to 2/3, with 1 crane operator, 4 batching plant
operators, and 5 concrete placement workers, or a total of 10 workers.
351
Q. 104 - R21
6.3. RESOLUTION OF THE SHORTAGE OF PROFICIENT SKILLED LABOR
In recent years, a shortage of cable crane operators has become a major

issue, so securing staff is difficult. Also, the aging of operators is progressing, and
the majority of skilled personnel is over 60 years, and as they retire in the future
there are concerns that the shortage of personnel will further increase. Among the
various types of cable crane, the rail rope cable crane is the most difficult to operate,
and many years are required to become an experienced operator.
By automating operation of rail rope cable cranes with this technology it is

possible to automatically and rationally transport the concrete bucket to the place-
ment position, so the operation can be performed with the same work efficiency as
a skilled operator despite the shortage of experienced operators, so this technology
can help to respond to the shortage of proficient operators.
7. CONCLUSIONS
A dam concrete automatic placement system that automates the series of

operations from production of dam concrete to transport and placement has been
developed and introduced on the yanagawa dam. this system integrates the automa-
tion technology that has previously been applied to the batching plant and transfer
car, etc., with automation of the rail rope cable crane.
In this project the system was developed to deal with a rail rope cable crane,
but the system can be applied to other types of crane provided that the 3-dimensional
position of the crane can be determined from boom, arm, and wire information, etc.,
via Global Navigation Satellite System, etc. For the future we intend to increase the
number of applications of the technology and increase the reliability and versatility
of the system.
ACKNOWLEDGEMENTS
Finally, we would like to thank all those that were involved from the Yanagawa
Dam Construction Office, Civil Engineering Department of the Morioka City Wide-
area Development Bureau, Iwate Prefecture, all those in the cooperating companies
that worked together with us on developing this technology, and all those in the Yana-
gawa Dam Special Construction Joint Venture that cooperated on this development
work on the busy site.
352
Q. 104 - R21
REFERENCES
[1] N. YAMASHITA, S. TACHIHANA, S. MORIYAMA, H. MORI, T. KASE. Develop-

ment of a dam concrete automatic placement system. The Dam Digest, 2019,
No 899 (in Japanese).
[2] S. TACHIHANA, S. MORIYAMA, N. YAMASHITA, H. MORI, T. KASE. Dam con-

crete automatic placement system. Japan Society of Civil Engineers Annual
Meeting, 2019, VI-711 (in Japanese).
353
Q.104 - R.22
-------
-------
INVESTIGATION, MEASURES AND CONTROLS AGAINST INTERNAL

RESTRAINT CRACKS DUE TO THERMAL STRESS OF RCC DAM
PRACTICE AT NAM NGIEP 1 HYDROPOWER PROJECT IN LAO PDR (∗)
Yushi AOSAKA & Shoji TSUTSUI

Hydropower Civil Engineering & General Manager,
THE KANSAI ELECTRIC POWER CO.,INC
Tatsuya KAWATA & Bassam EGILAT

Deputy Manager & RCC Dam Construction Engineer
NAM NGIEP 1 POWER COMPANY & Freelance Consultant, Japan
JAPAN
SUMMARY
During construction of the Roller Compacted Concrete (RCC) dam at Nam

Ngiep 1 Hydropower Project (NNP1) in Lao PDR, cracks in the direction of the dam
axis emerged on the surface of the RCC placing on the left bank side where RCC
placing had been suspended in order to install penstocks that would be embedded
in the dam body. The NNP1 Power Company conducted a crack investigation and
thermal stress analysis by applying the observed ambient temperature and variation
in RCC strength and was found that the main cause was the difference in concrete
temperature between inner and surface portions of the RCC due to a sudden drop
in ambient temperature in the middle of December 2017 that was larger than what
occurs in an average year. This paper presents the process of the above investiga-
tion, related root cause analyses, dam stability check and effective measures taken
to control cracks in RCC.
∗ Enquête, mesures et contrôles des fissures de compression interne dues aux contraintes
thermiques d’un barrage en BCR – Pratique utilisée pour le projet hydroélectrique Nam
Ngiep 1 de la République Démocratique Populaire du Laos
354 DOI: 10.1201/9781003211808-22

Q.104 - R.22
RÉSUMÉ
Lors de la construction du barrage en béton compacté au rouleau (BCR) du

projet hydroélectrique Nam Ngiep 1 (NNP1) de la RDP du Laos, des fissures dans le
sens de l’axe du barrage sont apparues à la surface du BCR coulé sur la rive gauche,
coulage qui avait été suspendu afin d’installer des conduites forcées encastrées
dans le corps du barrage. NNP1 a mené une enquête sur les fissures et une analyse
des contraintes thermiques en appliquant la température ambiante observée et la
variation de la résistance du BCR. Il a été constaté que la cause principale était la
différence de température du béton entre les parties interne et de surface du BCR
due à une chute brutale de la température ambiante à la mi-décembre 2017 plus
importante que celle d’une année moyenne. Cet article présente le processus de
l’enquête ci-dessus, les analyses des causes profondes connexes, la détermination
de la stabilité du barrage et les mesures efficaces prises pour maîtriser les fissures
dans le BCR.
1. INTRODUCTION
1.1. SUMMARY OF PROJECT
The Nam Ngiep 1 Hydropower Project (NNP1) is a hydropower Independent

Power Producer financed by the Kansai Electric Power Co., Inc. of Japan, the Elec-
tricity Generating Authority of Thailand in the Kingdom of Thailand (Thailand) and
the Lao Holding State Enterprise in Lao People’s Democratic Republic (Laos). The
NNP1 comprises a main and a reregulation dams/powerhouses located on the Nam
Ngiep River in Laos, with the installed capacities of 272 MW and 18 MW, respec-
tively. The NNP1 Power Company (NNP1PC) concluded a Concession Agreement
with the Lao Government in August 2013 and certified a Commercial Operation Date
in September 2019. The NNP1PC will be transferred to the Lao Government after
a 27-year concession period.
1.2. CONSTRUCTION OF RCC DAM
The main dam is a gravity dam that is 167 m high and made of Roller
Compacted Concrete (RCC). An aggregate of RCC consists of sandstone and con-
glomerated sandstone. A quarry of the aggregate is located downstream of the main
dam, on the right bank of the Nam Ngiep River. A cement of Type I [1] and a fly ash of
Class C [2] were procured from TPI Polene Company and Mea Moh Power Plant in
Thailand, respectively. The foundation excavation of the dam and the RCC placing
were completed in March 2016 and April 2018, respectively.
355
Q.104 - R.22
1.3. EMERGENCE OF CRACKS
In the initial design stage of the dam before construction, the NNP1PC incor-
porated thermal stress analysis into the dam design and construction plan in order to
control the thermal stress of the concrete. Nevertheless, cracks in the direction of the
dam axis emerged on the surface of the RCC placing at EL.275.8 m on the left bank
side (surface of the RCC placing), and cracks in the vertical direction emerged on the
exposed surface of contraction joint No.12 on the right bank side (surface of CJ12),
while the RCC placing on the right bank side was suspended from October 2017 to
January 2018 in order to install the horizontal part of the penstocks as shown in Fig.
1. Though water was sprayed on the surface of the RCC placing and the surface of
CJ12 to promote curing, the cracks emerged due to a sudden larger-than-usual drop
in ambient temperature in the middle of December 2017 that pushed temperatures
below 15◦ C. This paper presents the process of the crack investigation, related root
cause analysis, dam stability check and effective measures taken to control cracks
in RCC.
Fig. 1
Crack locations in dam body.
Emplacements des fissures dans le corps du barrage.
1.4. MIX PROPORTION AND COMPRESSIVE STRENGTH OF RCC
The dam was divided into zones by mix proportion as shown in Fig. 2. The
design mix proportion varied the proportions of cement (C, kg/m3 ) and fly ash (F,
kg/m3 ) into C80F120, C70F130 to C70F105, and C60F100 to C60F60 [3], [4]. A
low cementitious mix proportion was adopted for the inner core and high elevation
portions of the dam body where a lower design strength was acceptable, in order
to reduce costs. Figure 2 shows which of the 21 measurement points were used in
this paper to measure the inner concrete temperature and concrete strain. The mix
proportion of the RCC with cracks on the surface of the RCC placing was C60F70
356
Q.104 - R.22
and its compressive strength at 7 and 28 days was 4.2 N/mm2 and 7.3 N/mm2 ,
respectively, and reached 12.3 N/mm2 after 96 days when the cracks emerged.
Fig. 2
Dam zoning by mix proportion (Cross-section).
Zonage du barrage par proportion de mélange (section transversale).
2. CRACK INVESTIGATION
2.1. EMERGENCE OF CRACKS
In investigations, 5 cracks in the direction of the dam axis were found on the
surface of the RCC placing, and 2 cracks in the vertical direction were found on the
surface of CJ12 in January 2018. At the time, the RCC placing on the right bank side
had reached an elevation of 314.8 m, which was 39 m higher than the left bank side
(the dam top is EL.322m) where the work on the surface of the RCC placing had
been suspended. During the work suspension on the left bank side, green-cutting
and water spray curing without laytency processing were conducted. Similarly, for
the surface of CJ12, a sprinkler house was installed on the top surface of the right
bank RCC placing, and water spray curing was conducted.
2.2. OBSERVATION OF CRACKS
Crack maps of the surface of the RCC placing and the surface of CJ12 are
shown in Figs. 3 and 4. Cracks on the surface of the RCC placing and the surface
of CJ12 were visually observed. In addition, at first 7 blocks locating cracks were
removed in order to observe the condition of the cracks inside the concrete (see
Fig. 5). By cutting out a section about 15 cm deep with a breaker and measuring
the width and depth of the crack with a crack scale. And then, 12 cores were bored
357
Q.104 - R.22
(see Fig. 6). Crack No.1 and 3 were found to be over 3 m deep, while crack No.3
was confirmed to be less than 7 m through coring observation. Crack No.6 was also
found to be over 3 m but less than 7 m in depth.
Fig. 3
Crack map (Surface of RCC placing).
Carte des fissures (surface de coulage du BCR).
Fig. 4
Crack map (Surface of CJ12).
Carte des fissures (surface du CJ12).
358
Q.104 - R.22
Fig. 5
Procedure of block removal observation of crack (Steps 1 to 4).
Procédure d’observation de l’enlèvement du bloc fissuré (étapes 1 à 4).
1 Confirmation of location of crack 1 Confirmation de l’emplacement de
la fissure
2 Cutting out a section about 15 cm 2 Découpe d’une section d’environ
deep 15 cm de profondeur
3 Observation of crack 3 Observation de la
fissure
4 Measurement of crack width by 4 Mesure de la largeur de fissure par
crack scale échelle de fissure
Fig. 6
State of block removal and coring observation of cracks.
État de l’enlèvement du bloc et observation du carottage des fissures.
Table 1
Summary of results of crack investigation
CRACK NO. LOCATION OF CUTTING OUT CRACK CRACK CRACK
CRACKS OBSERVA- DEPTH (M) LENGTH WIDTH (MM)
TION (M)
1 Surface of the 2 Nos. 3 m or more 65.9 0.85, 0.45, 1.7
RCC placing
2 ditto 1 Nos. Within surface 23.4 0.25
3 ditto 3 Nos. Over 3 m, less 82.1 0.35, 1.1. 0.9
than 7 m
4 ditto Not available Not available 12.3 Not available
5 ditto Not available 3 m or more 9.3 1
6 Surface of CJ12 1 Nos. Less than 7 m 7.3 1
7 ditto Not available Not available 5.9 m 0.4
359
Q.104 - R.22
3. THERMAL STRESS ANALYSIS
3.1. INNER CONCRETE TEMPERATURE AND AMBIENT TEMPERATURE
In the initial design stage before construction, the mean monthly ambient tem-
perature was input for the thermal stress analysis of the RCC placing process. The
inner concrete temperature was measured in upstream, central and downstream
points at EL.276.1 m after RCC placing started. The daily inner concrete tem-
perature, RCC placing temperature and ambient temperature from September to
December 2017 are shown in Fig. 7. The difference in ambient temperature was
29.3◦ C between a maximum of 38.8◦ C and a minimum of 9.5◦ C. On the other hand,
the difference in the RCC placing temperature was 14◦ C based on 27.3◦ C maximum
and 13.3◦ C minimum. The outer concrete temperature on both the upstream and
downstream face was lower than the inner concrete temperature because of their
exposure to the ambient temperature. It is assumed that the cracks occurred due to
a sudden particularly large drop in ambient temperature in the middle of December
2017 that pushed temperatures below 15◦ C. By the way, RCC placing temperature
was controlled to be less than 28◦ C.
Fig. 7
Change in inner concrete temperature (EL.276.1 m) and ambient temperature.
Changement de la température intérieure du béton (EL.276,1 m) et de la
température ambiante.
1 Ambient temperature 1 Température ambiante
2 RCC placing temperature 2 Température de mise en place du RCC
3 Inter concrete temperature EL.276.1m 3 Température inter béton EL.276.1m
(Upstream) (en amont)
4 Inner concrete temperature EL.276.1m 4 Température intérieure du béton
(Center) EL.276.1m (centre)
5 Inner concrete temperature EL.276.1m 5 Température intérieure du béton
(Downstream) EL.276.1m (en aval)
360
Q.104 - R.22
3.2. OUTLINE OF THERMAL STRESS ANALYSIS
Thermal stress analysis was conducted by monitoring the ambient temper-

ature in order to assess the cause of the cracks [5–9]. The 167 m high dam with
foundation bedrock is modeled with a 2-dimensional cross section composed of 6 m-
wide blocks and divided into 7 zones by mix proportion as shown in Fig. 8. Thermal
stress analysis was conducted sequentially according to the RCC placing and initial
impounding timeline. Figure 8 shows the whole cross-section of the dam body. The
applied software is ASTEA MACS Ver.9. The analytic conditions shown in Table 2
were determined as follows.
Fig. 8
Analytic model.
Modèle analytique.
• Basically, the condition of the concrete in terms of compressive strength,

adiabatic temperature rise and other parameters was extracted as much as
possible from the concrete test results, otherwise standard values from the
Standard Specifications for Concrete Structures [7–9] were applied.
• The effective elastic modulus was estimated by formula [9] using the compres-
sive strength test results.
• The condition of cured construction joints was based on water spray curing
with a heat conductivity (η) of 14 W/m2 degree.
• The reservoir water temperature was set to 25◦ C.
• The initial impounding was started in May 2018 and completed in Aug 2018.
361
Q.104 - R.22
Table 2
Analytic conditions of concrete
NO. ITEM UNIT VARIATION REMARKS
1 Condition of
construction
1) Lift height m 0.6 Construction
pian
2 Condition of concrete
1) Heat conductivity W/m · degreee 2.6 Refer to [7, 8]
2) Unit wight Kg/m3 2.300 Design value
3) Specific heat kJ/kg· degree 1.05 Refer to [7, 8]
4) Compressive strength N/mm2 fc = Test result Test result
5) Effective elastic N/mm2 Ee(t) = φ(t)2 4700Tc25 Refer to [7, 8]
modulus
0 < 1 < 3: φ(t) = 3
3 < 1 < 6: φ(t) = 0.135
(t − 3) + 0.73
5 < 1: φ(t) = 1.0
6) Adiabatic temperature degree T = K(1 − exp ( − αβ)) Test result
rise
7) Tenzile strength N/mm2 ft(t) = 0.12 ∗ fc(t) Test result
8) Poisson’s ratio 0.2 Refer to [7, 8]
9) Linear expansion µ/deg.in C 10.0 Refer to [7, 8]
coefficient
10) Placing temperature degree 25 ∼ 26 Monitored
value
Table 3
Case of ambient temperature input into thermal stress analysis of RCC dam in
tropical monsoon climate area
DAM COUNTRY CLIMATE CONDITION OF AMBIENT TEMPERATURE
NNP1 Laos Tropical monsoon Stage of design and construction
Meant monthly temperature
1 Vietnam Tropical monsoon (21.9-28.7 degree, Difference temperature: 6.8 degree)
2 Vietnam Tropical monsoon Mean daily temperature
(13-33 degree, Difference temperature: 20 degree)
3 Iran Arid, semi-arid Mean monthly temperature
(18-36 degree, Difference temperature: 18 degree)
As for the ambient temperature in the initial design stage, since Laos has a
tropical monsoon climate, the temperature varies little throughout the year and does
not drop notably in winter like in cold zones. For the NNP1, the mean monthly ambient
temperature (21.9 to 28.7◦ C) was input into analyses based on the condition of the
climate and data from neighboring projects, as shown in Table 3. Since the ambient
temperature was monitored at the site office near the dam site after construction
started, it was revealed that temperature at the site varied considerably over the
course of the year because the dam site is located at the entrance to a mountainous
area. The average, maximum and minimum ambient temperature of Village A from
2012 to 2019, which is located near the dam site, are shown in Fig. 9. The minimum
362
Q.104 - R.22
ambient temperature was 9.5◦ C in December 2017, which was 12◦ C lower than
the mean monthly minimum temperature of 21.5◦ C in December. In addition, the
minimum temperature in December dropped below 10◦ C twice in the past 8 years.
To verify the cracks by thermal stress analysis, the ambient temperature measured
3 times per day – at 6:00, 12:00 and 18:00 – was input. The compressive strength
of the RCC was extracted from the test results as shown in Fig. 10. When there
was no data on the long-term strength (over 182 days) of the RCC, the strength
was estimated using the strength of the C80F120 mix proportion projected out from
182 to 365 days. The tensile strength was set for safety sake to 12 to 15% of the
compressive strength based on split tensile strength tests.
Fig. 9
Mean, maximum and minimum monthly temperature of Village A (2012-2019).
Température mensuelle moyenne, maximale et minimale d’un village A
(2012–2019).
1 Maximum monthly temperature (mean) 1 Température mensuelle maximale
(moyenne)
2 Maximum monthly temperature 2 Température mensuelle maximale
(maximum) (maximum)
3 Maximum monthly temperature 3 Température mensuelle maximale
(minimum) (minimum)
4 Minimum monthly temperature (mean) 4 Température mensuelle minimale
(moyenne)
5 Minimum monthly temperature 5 Température mensuelle minimale
(maximum) (maximale)
6 Minimum monthly temperature 6 Température mensuelle minimale
(minimum) (minimum)
7 Mean monthly temperature (mean) 7 Température mensuelle moyenne
(moyenne)
363
Q.104 - R.22
Fig. 10
Variation in compressive strength.
Variation de la résistance à la compression.
Adiabatic temperature rise tests were conducted for each mix proportion at
the dam site [10–13] and corroborated by estimation formula Eq. [1]. The results of
the adiabatic temperature rise tests for each mix proportion are shown in Fig. 11.
Fig. 11
Results of adiabatic temperature rise tests for each mix proportion.
Résultats des tests d’échauffement adiabatique pour chaque mélange.
Q(t) = Q∞{1 − exp ( − rts)} (1)
Q(t): Temperature rise (degree)

Q∞: Ultimate adiabatic temperature rise (degree)
r: Coefficient of temperature rise speed
t: Age (day)
s: Coefficient of adiabatic temperature rise speed
364
Q.104 - R.22
3.3. RESULTS OF THERMAL STRESS ANALYSIS
The results of the thermal stress analysis described below focus on the RCC
lift surface of K47L (3m height) where cracks emerged on the surface of the RCC
placing during the suspension of RCC placing work.
• A map of the inner concrete temperature in the middle of December 2017 and
the variation in the concrete surface temperature (K47L) are shown in Figs.
12 and 13. In Fig. 13, the temperature variations are at the top, center and
bottom portions of a 0.6 m high section in the RCC placing surface. After RCC
placing, the inner concrete temperature rose to over 40◦ C, except in the upper
portion, and reached 35◦ C everywhere. However, the temperature at all areas
dropped in November to December 2017. The upper portion was particularly
sensitive to the daily temperature fluctuation and dropped below 20◦ C at its
lowest in the middle of December.
Fig. 12
Map of inner concrete temperature in mid-December 2017.
Carte de la température interne du béton à la mi-décembre 2017.
Fig. 13
Variation in concrete surface temperature at K47L.
Variation de la température de surface du béton à K47L.
• The crack index of the upper portion for the concrete surface (K47L) dropped
to 0.61 in the middle of December 2017 due greatly to the daily temperature
fluctuation at the surface of the RCC placing as shown in Fig 14.
365
Q.104 - R.22
Fig. 14
Variation in crack index of concrete surface at K47L.
Variation de l’indice de fissuration de la surface du béton à K47L.
• The tensile strength of the concrete surface (K47L) obtained from the analysis
and that from strength test are shown in Fig. 15. The maximum tensile stress
of the concrete in the upper and central portions increased to 2.5 N/mm2 in the
middle of December 2017, which is higher than the actual tensile strength of
1.5 N/mm2 . Also, the cracks that emerged on the concrete surface (CJ12) are
assumed to have formed by the same process as the cracks on the surface of
the RCC placing.
Fig. 15
Variation in tensile strength of concrete surface at K47L.
Variation de la résistance à la traction de la surface du béton à K47L.
3.4. VERIFICATION OF THERMAL STRESS ANALYSIS
The valuation of the inner concrete temperature in the upstream portion of

the dam body at EL.211.6 m is shown in Fig. 16. Both the analyzed and monitored
values are in good agreement. The monitored values were slightly lower than the
analyzed values from November 2017 because the ambient temperature dropped
366
Q.104 - R.22
Fig. 16
Analyzed and monitored concrete temperature and strain at EL.211.6 m.
Température du béton analysée et surveillée et surveillée à EL.211,6 m.
1 Analyzed concrete temperature 1 Analyse de la température du béton
2 Monitored concrete temperature 2 Surveillance de la température du
béton
3 Analyzed concrete strain 3 Analyse de la déformation du béton
4 Monitored concrete strain 4 Contrôle de la déformation du béton
more than what usually happens in a typical year, the data from which was taken
for the analysis. After the impounding in May 2018, the inner concrete temperature
in the dam body continued to decrease gradually in both sets of data, though the
measured values were slightly smaller than analyzed values. Because the RCC
placing temperature was set to 25 to 26◦ C, and then actually varied from 16 to
28◦ C, which was lower than the input data shown in Fig. 7, there is a difference
between the analyzed and monitored values.
Figure 16 also shows the variation in concrete strain monitored in the same
location as the concrete temperature. Both the analyzed and monitored values are
in good agreement like the variation in inner concrete temperature. However, the
analyzed and monitored values during and after impounding form May 2018 differ
slightly because thinkably the thermal stress analysis did not take into account the
dam’s behavior resulting from the effect of compressive stress by water-pressure
loading during impounding.
4. ROOT CAUSE ANALYSIS OF CRACK, STUDY OF DAM STABILITY AND

REMEDIAL WORKS
4.1. CASES OF CRACKS IN OTHER RCC DAMS IN SOUTHEAST ASIA REGION
By seeing other RCC dams in the Southeast Asia region, the differences in
high lift between dam blocks and long-term work suspension are acceptable after
367
Q.104 - R.22
considering climate conditions, dam type, construction methods and so on, as shown
in Table 4. In the case of long-term work suspension, countermeasures such as to
control RCC placing temperature and spray curing the RCC surface are taken. Even
if the thermal cracks emerged due to the extended work suspension, it did not lead
to problems such as deterioration of dam stability or leakage. The hearing results
of cases of RCC dam cracks and remedial works in the Southeast Asia region are
shown in Table 5. It was found that the cases of RCC dam cracks with a width of
about 2 mm, a depth of 5 to 20 m, and a maximum length of 400 m emerged in the
direction of the dam axis on the surface of the RCC placing for the same reason as
at the NNP1: the cracks emerged because of a difference in temperature between
the inner and surface portions of the concrete and the thermal stresses caused by
internal restraints during the suspension of the concrete placing for several months.
Also, the main remedial work that is taken is to install steel mesh over the cracks.
Table 4
Cases of maximum difference in lift between dam blocks and measures
DAM COUNTRY CLIMATE MAXIMUM DIFFERENCE MEASURE
OF LIFT
height between dam blocks
NNP1 Laos Tropical 46.2 m Control of placing temperature,
monsoon spray curing
1 Laos ditto Over 50 m ditto
2 Vietnam ditto 82 m Insulation foam, control of plac-
ing temperature, spray cutting
3 Vietnam ditto 50 m ditto
4 Jorden Warm- Over 50 m ditto
temperature
zone
5 Lebanon Mediterranean 63 m ditto
6 Malaysia Tropical 20 m ditto
7 Malaysia Temperate Over 30 m ditto
Table 5
Cases of occurrence of RCC dam cracks in Southeast Asia region
NO. COUNTRY CLIMATE DIRECTION CONDITION CASES OF CONDITION REMEDY
OF CRACK OF CRACK CRACK OF RCC WORKS
PLACING
1 Vietnam Tropical Dam axis W: 2 mm D: Thermal M: C60F120 Installation of the
monsoon 5m Stress T: less than steel mesh on
22 deg. crack
2 Vietnam Tropical Dam axis Thermal Installation of the
monsoon Stress steel mesh on
crack
3 Australia Temperate Dam axis W: 2 mm D: Thermal M: C80F105 Installation of the
Until dam Stress T: less than steel mesh on
foundation 20 deg. crack
4 India Savanna T: less than
28 deg.
[Remarks] W: Crack width, D: Crack depth, M: Mix proportion, T: Placing temperature
368
Q.104 - R.22
4.2. CAUSE OF CRACKS IN THE NNP1
The cause of the RCC cracks at the NNP1 was studied through crack inves-
tigation, thermal stress analysis and hearing from dam experts about the practices
at other RCC dam projects, as follows.
• The main causes were thermal expansion in mass concrete and thermal stress
due to internal restraints associated with non-uniform temperature drops at the
concrete surface and especially with the variation in the seasonal ambient tem-
perature. The spray curing was conducted but the temperature of the concrete
surface was greatly affected by the ambient temperature, which registered a
maximum difference of 29.3◦ C between a high of 38.8◦ C in early October 2017
and a low of 9.5◦ C in December 2017. On the other hand, the inner concrete
temperature rose to around 40◦ C and then gradually dropped. It is assumed
that cracks emerged due to this difference in temperature, that is to say, the
difference in the cooling rate on the concrete surface.
• In thermal stress analysis, this was evidenced by an excessive crack index for
RCC and tensile strength. The crack index dropped to 0.61 and the tensile
strength increased beyond the actual tensile strength of 1.5 N/mm2 , to 2.5
N/mm2 . This fact tells us that the site conditions promoted cracks near the
surface of the RCC placing.
• Although water spray curing was conducted, the surface of the RCC placing
was not completely covered with water, therefore cracks emerged due to drying
shrinkage of the concrete surface. In addition, when the water temperature for
curing dropped alongside the ambient temperature, it have possibly induced
cracking.
From the above assumption of the sudden drop of ambient temperature at

the site, it might be preferable to cover the surface of RCC placing with sheet for
insulation curing in addition to wet curing by water spraying and ponding.
4.3. CHECK OF DAM STABILITY
In light of the cracks, dam stability was studied by hearing from the Dam Safety
Review Panel (DSRP) of the NNP1 and other dam experts.
• Since the initial impounding, the dam has been monitored for displacement,
leakage from the dam body, uplift and so on. Both the analyzed and monitored
values for dam displacement and uplift are in good agreement. Also, leakage
from the drainage holes in the dam body is almost zero and leakage from
horizontal joints on the downstream surface of the dam body have not been
found [14]. In addition, the half-cut steel pipes installed along the crack in
surface of CJ12 have a role of measurement of leakage and so far no leakage
was observed. Based on the monitoring results at the moment, the necessary
369
Q.104 - R.22
requirements of the dam body have been achieved and the dam’s stability has
been secured. This is an integral part of the dam safety process to assess the
dam stability.
• The cracks on the surface of the RCC placing emerged due to a sudden drop
in ambient temperature. They may continue propagating to a certain depth,
but it is assumed that the cracking will stop at a point where the increase in
thermal stress and the concrete shrinkage due to the effect of the ambient
temperature reach an equilibrium. Even if the cracks grow, it is assumed that
they will close due to the water pressure from upstream after impounding. In
addition, it will not directly lead to water leakage through the dam because
the cracks are not in the upstream-downstream direction but in the dam axis
direction.
• The cracks on the surface of CJ12 are in the same direction as the verti-
cal construction joints of the columnar block method used to build the dam.
After impounding, the body will integrate under compressive stresses even
where there are vertical cracks, and stress in the dam body will be transmitted
accordingly, which means the dam is stable.
• However, it is important to continue dam monitoring, even urgent inspection
after occurrence of earthquake and check of dam stability in the long term [15].
4.4. CASES OF CRACKS AND REMEDIAL WORKS OF CONCRETE DAMS
Cases of cracks emerging in concrete dams in Japan due to internal restraints

have also been reported. Especially, so-called “autumn crack” tend to occur in early
winter, and it is said to be careful during the autumn cold season. However, remedial
works for cracks in concrete dams have not been frequently disclosed in Japan or
overseas, probably because they have been treated as minor events. According to
reference papers [15], remedial works are as follows.
(1) Remedial works for cracks on the surface of RCC placing

• A V-cut is shaped and filled with non-shrink mortar. And a half-cut
steel pipe (30 cm diameter) and bi-laterally arranged re-bar (D32) are
installed.
• A V-cut is shaped and a half-cut steel pipe and re-bar are installed.
• A V-cut is filled with micro cement.
(2) Remedial works for cracks on the surface of contraction joints
• A V-cut is shaped and filled with cement bond.
• The hole is drilled against the crack, and a pipe is installed and filled
with cement bond and colloidal cement.
Base on the above remedial works extracted from reference papers and hear-
ing from dam experts in Japan and overseas, The NNP1PC conducted remedial
works as follows. The re-bar was installed because there is concern that thermal
stress cracks may occur when new concrete is placed on the present RCC surface
due to external restraints.
370
Q.104 - R.22
(1) Remedial works for cracks on the surface of the RCC placing
• To avoid crack propagation, a V-cut was shaped and a half-steel pipe
and re-bar were installed, and then V-cut was filled with micro cement
(water cement ratio of W/C = 0.6) as shown in Figs. 17 and 18.
(2) Remedial works for cracks on the exposed surface of CJ12
• The cracks were filled with micro cement (W/C = 0.6). After that, a V-cut
was shaped along the cracks and filled with non-shrink mortar (W/C =
0.47).
Fig. 17
State of crack treatment in the NNP1
État du traitement des fissures du NNP1.
Fig. 18
Schematic diagram of remedial works for cracks in the NNP1.
Diagramme schématique des travaux de réparation des fissures dans le NNP1.
1 Placing of concrete 1 Pose de béton
2 Installation of distributing bar 2 Installation de la barre de
(D14mm@400mm) distribution (D14mm @ 400mm)
3 Installation of main re-bar 3 Installation de la barre d’armature
(D22mm@200mm, Length:2m) principale (D22mm @ 200mm,
longueur: 2m)
4 Injectionation of grout 4 Injection de coulis
5 Installation of half-cut steel pipe 5 Installation d’un tuyau en acier à
moitié coupé
6 V-cut, mortar 6 Coupe en V, mortier
7 Drilling (φ76mm) Install of steel 7 Forage (φ76mm) Installation du
pipe (φ50mm) tuyau en acier (φ50mm)
371
Q.104 - R.22
5. CONCLUSIONS
Regarding the cracks that emerged at the dam in the NNP1, the findings and
items to be studied further were obtained through a process of root cause analysis
and study of effective measures for controlling internal restraint cracks as follow.
The cracks emerged because of the thermal expansion in the mass concrete
and the thermal stress due to internal restraints associated with a non-uniform tem-
perature drop at the concrete surface. The thermal stress analysis proved that the
cracks near the RCC surface emerged because the tensile stress increased beyond
the actual tensile strength of 1.5 N/mm2 , to 2.5 N/mm2 .
Both the analyzed and monitored values for dam displacement and uplift are in
good agreement. Also, there have been almost zero leakage from the drainage holes
in the dam body and leakage from horizontal joints on downstream surface of the
dam body have not been confirmed. Based on the monitoring results, the necessary
requirements for the dam body have been achieved and the dam’s stability has been
secured.
The cracks may continue propagating to a certain depth, but it is assumed that
they will stop at a point where the increase in the thermal stress and the concrete
shrinkage due to the effect of the ambient temperature reach an equilibrium. Even if
the crack were to grow, it is assumed that they will close due to the water pressure
from upstream after impounding. The body will integrate under compressive stresses
even where there are vertical cracks, and stress in the dam body will be transmitted
accordingly, which means the dam is stable. In addition, it would not directly lead
to water leakage through the dam because the cracks are not in the upstream-
downstream direction but in the dam axis direction.
It might be preferable to cover the RCC surface with an insulated curing sheet
in addition to wet curing by water spraying and ponding when exposing the concrete
surface for a long time. By seeing other RCC dams in Southeast Asia region, the
difference in lift between dam blocks and long-term work suspension are acceptable
after considering climate conditions, dam type, construction methods and so on. In
the case of extended work suspension, countermeasures such as to control RCC
placing temperature and spray curing of the RCC surface are taken.
While the RCC placing temperature in the NNP1 varied from 16 to 28◦ C, the
maximum RCC placing temperature was set to 28◦ C in the thermal stress analyses.
In addition, the local minimum crack index was set to 1.3 to 1.6, therefore the initial
design gave some allowance for crack occurrence. However, the minimum ambient
temperature of 9.5◦ C in December 2017 was the lowest in years and lower than the
mean monthly minimum temperature of 12.3◦ C of the last 8 years. Laos has a tropical
monsoon climate and there is quite a difference in ambient temperature on a day
in the cool season when the temperature drops at the NNP1, since the dam site is
372
Q.104 - R.22
located in a mountainous area. Attention should be paid to the ambient temperature

input into the thermal stress analysis, when the concrete placing surface is exposed
to the ambient temperature for a long time. Namely, it is desirable to analyze the
thermal stress of RCC according to the conditions of the project site, such as by
applying the ambient temperature at every 12 hours rather than the mean monthly
temperature that is usually used in general design and obtained by referring to other
RCC dam projects in the Southeast Asia region. As a result, it will be possible
to obtain the actual concrete temperature and concrete strain, and improve the
accuracy of analyses. It could also be reflected in planning stage studies of RCC
placing method with temporary facilities.
From the above findings, it might be difficult to prevent extraordinary unfore-

seeable events caused by things such as climate change. However, even in such
cases, it is desirable to achieve reasonable solutions through investigation, anal-
yses based on actual monitoring data, and practices and effective measures for
controlling cracks in RCC implemented in other projects.
Attention should be paid to thermal stress cracking after impounding, it is desir-

able to consider the water level on the upstream side of the dam and the variation
in water temperature caused by impounding in thermal stress analyses, and to take
measures if necessary.
ACKNOWLEDGEMENTS
The initial impounding of the RCC dam in the NNP1 was completed and com-
mercial operation was started in 2019. Although no further cracks have been found
in the RCC dam since the start of power station operations, it is necessary to con-
tinue to attentively monitor the situation in the future. Finally, we would like to express
our sincere appreciation to Obayashi Corporation for constructing the dam, as well
as RCC expert Brian A. Forbes, Marco Conrad of Lender Technical Adviser, and
the members from DSRP for their generous cooperation during the execution of site
works.
REFERENCES
[1] ASTM International. C150 Standard Specification for Portland Cement, 2012.
[2] ASTM International. C618 Standard Specification for Coal Fly Ash and Raw
or Calcined Natural Pozzolan for Use in Concrete, 2012.
373
Q.104 - R.22
[3] TABUCHI T., MURAKAMI Y., ASAKAWA M., AOSAKA Y. Design mix proportion
and construction instance for RCC at Nam Ngiep 1 hydropower project in Laos,
Japan Concrete Institute, Vol.55, No.5, pp.397–402, 2007 (in Japanese).
[4] AOSAKA Y., SEOKA T, FORBES B., COCKCROFT J., MURAKAMI Y.,
ASAKAWA M. Trial mix and full-scale trial embankment for RCC dam at Nam
Ngiep 1 hydropower project, 85th Annual Meeting of International Commission
on Large Dams, 2017.
[5] MARCO C., BRUSIN M., MORRIS D., TRONG P. Thermal Stress Analysis for
the Lai Chau RCC Gravity Dam, 7th International Symposium on RCC Dams,
2015.
[6] CHEN H., Zhaoying W. Simulation Analysis on the Thermal Field and the
Stress Field for Gongguoqiao RCC Dam, 7th International Symposium on
RCC Dams, 2015.
[7] Japan Society of Civil Engineers. Standard Specifications for Concrete

Structures [Dam concrete], pp.1–17, 2002 (in Japanese).

Structures [Design], pp.302–318, 2012 (in Japanese).

Structures [Execution], pp.121–176, 2012 (in Japanese).
[10] ANSEN K., FORBES B. Thermal induced cracking performance of RCC Dams,
6th International Symposium on RCC Dams, 2012.
[11] GALANG A., TAM C., MAK S., HO D. Determination of adiabatic temper-
ature rise in concrete capabilities and performance of a new system, 28th
Conference on our World in Concrete & Structures, 2003.
[12] HACHIYO Consultant. Report of adiabatic temperature rise test of concrete,

2010 (in Japanese).
[13] Japan Dam Engineering Center. Detail technology of concrete dam (Revision 2
edition), pp. 252–263, 1992 (in Japanese).
[14] AOSAKA Y., TABUCHI T., NOHARA S., TSUTSUI S. Study of watertight-
ness and drainage hole in dam body of Roller Compacted Concrete on Nam
Ngiep 1 Hydropower Project in Laos, 8th International Symposium on Roller
Compacted Concrete (RCC) Dams, pp. 83–92, 2019.
[15] International Commission on Large Dams. Bulletin 166 Inspection of Dams

(Following earthquake guidelines), 2016.
374
Q.104 - R.23
-------
-------
DEVELOPMENTS IN CONCRETE DAM ENGINEERING (∗)
Q.H.W. SHAW
Director, ARQ CONSULTING ENGINEERS
SOUTH AFRICA
SUMMARY
When first introduced in the later years of the 20th century, Roller Compacted
Concrete (RCC) was treated simply as a different, and more rapid, method for the
construction of concrete dams. Since that time, developments in the understanding
of early RCC behaviour have highlighted the need for a different design approach,
or the consideration of additional factors in RCC dam design. Differences have been
established in the stress-relaxation creep behaviour of different RCC types during
the hydration process, which can be both less and greater than assumed in tra-
ditional practice for conventional mass concrete. Related issues consequential to
the horizontal construction of RCC dams, as opposed to the vertical construction of
conventional mass concrete dams, have also revealed dangers in the assumption
of 2-dimensional structural function, which has been a universal factor in the histor-
ical design of gravity dams. A revised design approach and an additional concrete
property, in the form of stress-relaxation creep have consequently been introduced
for the design of large RCC dams.
RÉSUMÉ
Lors de son introduction dans les dernières années du 20e siècle, le béton
compacté au rouleau (BCR) a été traité simplement comme une méthode différente,
et plus rapide, pour la construction de barrages en béton. Depuis lors, l’évolution
∗ Évolution dans l’ingénierie des barrages en béton
DOI: 10.1201/9781003211808-23 375

Q.104 - R.23
de la compréhension du comportement du béton compacté au rouleau a mis en

évidence la nécessité d’adopter une approche de conception différente ou de pren-
dre en compte des facteurs supplémentaires dans la conception des barrages en
béton compacté au rouleau. Des différences ont été établies dans le comporte-
ment de fluage par relaxation des contraintes des différents types de RCC au cours
du processus d’hydratation, qui peut être à la fois inférieur et supérieur à ce qui
est supposé dans la pratique traditionnelle pour le béton de masse convention-
nel. Les problèmes liés à la construction horizontale des barrages en béton armé,
par opposition à la construction verticale des barrages conventionnels en béton de
masse, ont également révélé des dangers dans l’hypothèse d’une fonction struc-
turelle bidimensionnelle, qui a été un facteur universel dans la conception historique
des barrage-poids. Une approche de conception révisée et une propriété supplé-
mentaire du béton, sous la forme d’un fluage de détente des contraintes, ont donc
été introduites pour la conception des grands barrages en béton armé.
1. BACKGROUND
Traditionally, concrete dams have been constructed vertically in a series of

monolithic blocks. Contraction joints have been included between the blocks in order
to accommodate concrete shrinkage associated with mass gradient thermal effects
during the hydration process and the width of the individual blocks was primarily
determined with the objective of avoiding intermediate cracking. In the case of grav-
ity dams, each individual monolith functions independently of those adjacent and
consequently, all that is required on the contraction joint is a seal against water leak-
age. In the case of an arch dam, structural continuity across the joints is required and
the contraction opening is filled with grout, once sufficient shrinkage has occurred.
In both cases, and before grouting in the case of an arch, the monoliths carry the
forces from their own mass vertically down into the foundation. In the case of a
conventional concrete gravity dam, each monolithic block must consequently be
individually stable on its foundation. In the case of an arch dam, the contraction
joint on the river side of a particular monolith must be grouted before the limit of its
stability on its foundation is reached.
As a consequence of the above behaviour, concrete gravity dams have his-

torically been designed as 2-dimensional structures, while particular assumptions
related to creep during the hydration process have allowed the development of a
temperature-based approach for the design of arch dams.
With the advent of Roller Compacted Concrete (RCC), often using high levels
of supplementary cementitious materials, two key factors have been introduced that
can together create significantly different early behaviour and related design require-
ments. Firstly, the levels of concrete shrinkage experienced during the hydration
process are not the same and can be substantially less than traditionally assumed
for conventional vibrated mass concrete (CVC) when high levels of fly ash are used.
376
Q.104 - R.23
Secondly, horizontal concrete placement implies that the mass of the dam will not
necessarily be carried directly, or vertically down into the foundation, particularly
when concrete shrinkage levels are low.
2. PROTOTYPE OBSERVATION & RESEARCH
As a consequence of lower cementitious materials contents than typical for

CVC dams, in the early days of RCC dam design and construction, it was expected
that induced joints could be located at wider spacings than applied for the formed
joints on CVC dams and that the final joint openings would also be less. In reality,
poorer performance, with intermediate cracking was often reported when significant
joint spacings were applied, while observation otherwise suggested that total con-
crete shrinkage during the hydration process was substantially lower in certain RCC
types than others. Although some low cementitious-content RCC mixes were found
to indicate total shrinkage exceeding that typically expected of CVC, other RCC mix
types were found to exhibit very low levels of total shrinkage.
Research simulating and comparing prototype performance indicated

behaviour on certain RCC arch-gravity dams where zero concrete shrinkage was
suggested [1]. In fact, an upstream movement during construction was measured
on RCC arch dams, implying the presence of horizontal compression through the
structure during construction, which substantially impacts the distribution of gravity
loading onto the dam foundation and within the structure itself.
3. INTRODUCTION
In this paper, the author describes the differences between the early behaviour
of RCC and CVC in large dams and how these influence dam design and necessitate
a slightly different design approach, taking cognisance of an additional property of
concrete in the form of stress-relaxation creep.
4. STRESS-RELAXATION CREEP
4.1. THE TRADITIONAL APPROACH TO STRESS-RELAXATION CREEP
The USACE’s EM 1110-2-2201 Arch Dam Design [2] defines a series of key
temperatures that are experienced in the concrete of the core of a dam structure
during the hydration process and subsequent natural cooling, as illustrated in Fig. 1.
377
Q.104 - R.23
Fig. 1
Exaggerated effect of upstream movement during construction
(as measured & modelled) [1].
Effet exagéré du mouvement en amont pendant la construction
(Tel que mesuré et modélisé)[1].
Fig. 2
Temperature-time cycle in the core of a concrete dam.
Cycle température-temps dans le noyau d’un barrage en béton.
1 Temperature (◦ C) 1 Température (◦ C)
2 Time 2 Temps
3 Summer 3 Été
4 Winter 4 Hiver
5 Hydration phase 5 Phase d’hydratation
6 Hydration heat dissipation 6 Phase de dissipation de la
phase chaleur d’hydratation
7 Seasonal equilibrium 7 Équilibre saisonnier
8 Placement temperature (T1) 8 Température de placement (T1)
9 Maximum hydration 9 Température d’hydratation
temperature (T2) maximale (T2)
10 “Zero stress” temperature 10 Température "zéro contrainte"
(T3) (T3)
11 Minimum equilibrium temperature 11 Température minimale
(T4) d’équilibre (T4)
12 Effective volume reduction 12 Réduction du volume effectif
(T3 - T1) (T3 - T1)
13 Long term temperature drop 13 Chute de température à long
(T3 - T4) terme (T3 - T4)
378
Q.104 - R.23
With the concrete placement temperature defined as T1, T2 represents the

maximum temperature reached during the hydration cycle and T4, the minimum
temperature experienced in winter, once the full hydration heat has been dissipation
from the structure. The T3 temperature is described as the “zero stress” temperature
and essentially represents the temperature at which the compression stresses asso-
ciated with constrained thermal expansion are relieved, with all subsequent cooling
and temperature drop resulting in tension in the concrete, or thermal shrinkage
below the volume, as originally placed.
In EM 1110-2-2201 [2], it is indicated that the difference between the T2

and T3 temperatures for a CVC arch dam will vary, depending on the lift height,
but will typically lie within the range of 3◦ to 5◦ F (or 1.6◦ to 2.8◦ C), while it is
further stated that T3 is usually equated to T2 for preliminary analysis. Equating
T3 to T2 implies that the concrete has effectively shrunk by an amount equiva-
lent to the thermal expansion of hard concrete that would have taken place with
a temperature increase from T1 to T2, or the full hydration temperature rise.
For typical mass concrete and RCC with adiabatic hydration temperature rises
of between 10 and 25◦ C, the equivalent shrinkage, or reduction in concrete vol-
ume that occurs during this process is consequently of the order of 100 to 250
microstrain.
Although the net effect has traditionally been treated simplistically through the
application of related temperature drop loads, the real cause is a series of complex,
plastic processes that will impact the core and the surfaces zones of a concrete dam
differently.
In a number of instances, the effective shrinkage of RCC during the hydra-

tion cycle has been modelled considering the temporal development of the concrete
elastic modulus. Applying this theory, early thermal expansion strain is translated
into a compression stress at the low elastic modulus of concrete of low maturity, while
the later temperature drop shrinkage is translated into a tensile stress at the elastic
modulus of the mature concrete. The difference between the tensile and compres-
sion stresses is subsequently translated back into a strain, which is equated to the
hydration cycle stress-relaxation creep. While this approach for RCC might typically
indicate a maximum shrinkage value of the order of 100 microstrain, it is in fact
flawed, as it does not recognize the non-linearity of the actual process that occurs,
which is dependent on the applicable level of containment stress. Applying this
approach for arch dams can significantly under-estimate the net thermal-structural
impact of the concrete hydration process, primarily as the analysis will imply con-
cave joint profiles, open in the core, but closed in the surface zones, which is not in
line with prototype observation.
As the primary process that results in the concrete volume change occurs as
compression stresses are developed and dissipated through creep in the restrained,
379
Q.104 - R.23
immature concrete, as it attempts to expand with rising temperature, the process is

most realistically defined as a stress-relaxation creep.
4.2. THE MECHANISMS OF STRESS-RELAXATION CREEP
Although the surface concrete of a large dam will not be subject to the full
adiabatic temperature rise, nor the same restraint as the core zone, the dominant
mechanism that incurs a shrinkage, or a reduction in the concrete volume during
the hydration process in large dams is stress-relaxation creep (SRC). SRC occurs
primarily during the first few day after placement, when the concrete is immature and
significant quantities of heat are evolved during through the process of cementitious
materials hydration.
While immature concrete within the mass pour of a large dam is being heated
by hydration of the cementitious materials, it is internally constrained against expan-
sion by the surrounding concrete that is experiencing the same temperature rise
and related expansion, as illustrated in Fig. 3. Consequently, the immature concrete
experiences compression stress.
As the hydration of the cementitious materials progresses, heat is evolved,

the concrete gains strength and the cementitious gel produced reduces in volume
Fig. 3
Illustration of internal constraint experienced in concrete during hydration
temperature rise.
Illustration de la contrainte interne subie par le béton lors de l’élévation de la
température d’hydratation.
1 Concrete as cast 1 Béton coulé

2 Concrete heated by hydration 2 Béton chauffé par hydratation
3 Concrete constrained from 3 Béton limité dans son expansion
expansion by surrounding a cause de béton environnant
concrete
380
Q.104 - R.23
compared to the volume of the component reactants (cementitious materials and

water). Consequently, three phenomena are evolving simultaneously; increasing
compressive stress due to physical restraint during temperature rise, shrinkage of
one component of the concrete and strength gain. Initially, the gel volume reduction
will cause shrinkage in the paste (chemical shrinkage) and consequently, the con-
crete, but eventually strength gain will prevent further shrinkage and the subsequent
volume reduction takes the form of the development of micro-pores, or voids in the
paste. Without anycontainment stress, the final shrinkage of the concrete is termed
autogenous shrinkage.
Fig. 4
Volume phases associated with cementitious hydration.
Phases volumétriques associées à l’hydratation du ciment.
1 Unhydrated cementitious materials 1 Matériaux cimentaires non

hydratés
2 Unhydrated water 2 Eau non hydratée
3 Hydration products 3 Produits d’hydratation
4 Voids generated by hydration 4 Vides générés par l’hydratation
5 Autogenous shrinkage 5 Rétrécissement autogène
6 Chemical shrinkage 6 Rétrécissement chimique
7 At casting 7 A la sélection
8 At initial set 8 A la préparation
9 After hardening 9 Après le durcissement
When containment compressive stresses are developed in mass concrete

during this process, the reduction in volume of the cementitious paste will create a
tendency for the overall system to collapse and to release stress through shrinkage,
effectively relaxing the containing stress in the weak, immature concrete. Once the
concrete has gained sufficient strength, it will hold the containing stress, without
381
Q.104 - R.23
creeping, or relaxation, as subsequent cementitious paste shrinkage simply results

in the development of micro-pores. Stress-relaxation creep consequently occurs
early during the hydration process, when heat development is at its highest and
the concrete strength is still low compared to the containment stresses developed
through thermal expansion.
4.3. STRESS-RELAXATION CREEP INFLUENCERS
The degree of stress-relaxation creep experienced in mass concrete is influ-

enced by the degree of restraint against thermal expansion, the applicable level of
autogenous shrinkage of the cementitious materials used, the paste content and
the nature of the aggregate skeletal structure of the concrete mix. The degree of
autogenous shrinkage is influenced by various aspects of the chemical composition
of the cement, but particularly by the amount and type of the supplementary cemen-
titious materials used. While certain blast furnace slags can significantly increase
autogenous shrinkage, fly ash is very effective in developing substantial reductions.
In mortar, Nawa & Tomoali reported that increasing fly ash content to 60% reduced
autogenous shrinkage by almost 80% [3]. Similarly, Bofang reported a reduction in
the 548 day autogenous shrinkage exceeding 90% when 40% of the cement was
replaced with fly ash [4].
Compared to CVC mixes, RCC mixes are lower in total paste content, while
a modern, high-workability RCC mix is specifically designed to minimize the voids
within the aggregate skeletal structure and the inherently high-energy compaction
process is very effective in creating a tight aggregate structure and moving paste to
the surface.
The final piece of the puzzle in respect of SRC is structural restraint. RCC
arch-gravity dams have been found to demonstrate significantly wider contraction
joint openings at the base of the structure than at the crest. In many cases, no joint
opening occurs over the central parts of the crest. In all of the applicable cases,
the arch-gravity configuration applied involved a single, constant upstream circular
radius, a vertical upstream face and a constant-slope downstream face. With such a
configuration, the bottom of the structure will be of maximum width in an upstream-
downstream direction and minimum length in a cross-canyon direction at the base,
where close to the restraint of the foundation. With height, the width will narrow
and the length and arch aperture will increase, implying a progressive decrease in
structural rigidity.
With the degree of SRC depending on autogenous shrinkage, RCC aggregate

skeletal structure, the hydration temperature rise and the level of restraint against
thermal expansion, a basic function will exist for each RCC mix, defining a rela-
tionship between SRC and applicable constraining stress, similar in principle to that
indicated for the purposes of illustration in Fig. 5.
382
Q.104 - R.23
Fig. 5
Illustrative relationship between SRC and compression stress.
Relation illustrative entre SRC et contrainte de compression.
1 Variation of SRC with compression 1 Variation de SRC avec contrainte

de com- stress pression
2 Compression stress (kPa) 2 Contrainte de compression (kPa)
3 Stress-relaxation creep 3 Fluage de relaxation des
(microstrain) contraintes (microdéf- formation)
5. THE INFLUENCE OF LOW SRC ON DAM BEHAVIOUR
5.1. HORIZONTAL ARCH-GRAVITY DAM CONSTRUCTION IN LOW SRC RCC
Conventional arch dam technology is based on the premise that monolithic

blocks are constructed vertically and suffer early stress-relaxation creep (SRC), or
shrinkage that is equivalent essentially to the full expansion associated with the
hydration temperature rise in mature concrete. In the case of an RCC arch-gravity
dam, constructed horizontally in a low stress-relaxation creep RCC, the reality is
in fact quite different. At the base, where restraint is significant and the section is
rigid, containment stresses will be at their highest and SRC levels will be higher.
With height, the progressively reducing levels of restraint and structural rigidity, in
tandem with the low levels of autogenous shrinkage, will allow the inherent thermal
expansion with temperature rise to occur, creating lower containment stresses, and
in turn less stress-relaxation creep. The result is seen, and has been measured,
on prototype dams, as a progressive movement of the structure with height in an
upstream direction [5]. This effect has been observed as an increase in compressive
383
Q.104 - R.23
strain on the downstream side and an opening of induced joints on the upstream
face of RCC arch-gravity dams.
Fig. 6
Illustration of downstream crest movement on arch-gravity structure with cooling.
Illustration du mouvement de la crête en aval sur une structure de gravité en arc
avec refroidissement.
1 Dam section with temperature 1 section de barrage dont la

raised by température est hydration
augmentée par l’hydratation
2 Dam section with hydration heat 2 Section de barrage avec chaleur
Dissipated d’hydratation dissipée
As an arch-gravity structure is being constructed in low SRC RCC, the reduced

restraint and SRC and the consequentially increased thermal expansion creates a
progressive expansion and upstream leaning effect in a dam with a curved axis
and structural modelling has demonstrated this to be effectively equivalent to a
lengthening of the dam crest of 30 to 50 microstrain on a structure of 70 to 100
m in height.
As the hydration heat is slowly dissipated from the body of the dam, with
heat retained for longer in the more massive sections, the crest of an arch-gravity
structure will move progressively downstream, increasing compression stresses in
the upper elevations of the arch, as illustrated in Fig. 6.
The above effect is the origin of the apparent observation of “zero SRC”, or T3
= T1 on a number of RCC arch-gravity dams constructed horizontally in fly ash-rich
RCC. In these cases, an actual RCC SRC of the order of 30 to 50 microstrain was
masked by an equivalent effective crest lengthening caused by thermal expansion of
the immature RCC and a consequential upstream leaning of the cantilevers during
construction.
384
Q.104 - R.23
6. THERMO-MECHANICAL ANALYSIS FOR Dam DESIGN
6.1. GENERAL
When a gravity dam is long in relation to its height and adjacent blocks are
of similar height, the level of SRC will be of little importance and the structure can
be designed, or analysed as 2-dimensional. For a horizontally-constructed RCC
gravity dam with a lower length/height ratio (< 6) [6], however, knowledge of the
applicable level of SRC is essential in developing a meaningful understanding of
the 3-dimensional structural function. Although curvature will reduce the restraint
against thermal expansion higher in an arch-gravity dam structure, restraint will also
decrease with height in a gravity dam, as the section becomes narrower and longer.
Whether an arch-gravity, or gravity structural configuration, an RCC dam con-

structed horizontally in a relatively narrow valley using a low SRC RCC will behave
as a 3-dimensional structure during construction. With heat retained within the core
and expansion due to low SRC, the upper part of the structure will effectively be lifted
during construction, resulting in self-weight load from new placement not being car-
ried directly, or vertically downwards, but being carried by a bridge formed in the
swollen lower structure, as illustrated in Fig. 7 and 8.
Fig. 7
Illustration of heat-swollen lower dam body of low SRC RCC.
Illustration d’un corps de barrage inférieur gonflé par la chaleur.
1 Heat swollen RCC experiencing 1 RCC gonflé par la chaleur
compression subissant stress des contraintes
de compression
The above thermo-mechanical effects are unlike those experienced in ver-

tically constructed CVC dams and can lead to quite different levels of SRC in
different parts of a dam; a quite different situation to the assumption of rela-
tively high, uniform SRC/shrinkage assumed in CVC design practice. In a sit-
uation of very low SRC, downwards settlement of the upper part of a gravity
dam in a relatively narrow valley could easily result in lateral bridging through
the dam structure. Lateral bridging in this manner during dam operation will
385
Q.104 - R.23
Fig. 8
Bridging causing gravity load to be carried laterally/indirectly to the foundation.
Pontage entraînant le transport de la charge gravitaire latéralement / indirectement
à la fondation.
1 Gravity load path 1 Trajectoire de la charge de gravité
imply that a gravity dam will not function as a 2-dimensional structure, as assumed
in all design guidelines and literature developed for concrete dam types to date.
A significant assumption for a concrete gravity dam is that the mass gravity
load of the constituent concrete of each block between formed, or induced joints
is transferred in 2-dimensions through the structure and into the foundation. The
entire premise of the critical shear stability calculations for a gravity dam, both in the
foundation and within the constituent concrete of the structure, is dependent on this
assumption.
The joints between monolithic blocks in a CVC dam are formed by robust
formwork, or shutters and consequently the surfaces produced are smooth and
flat. With these joints opening due to the contraction of the blocks on either side,
mechanical interlock, or shear transfer between adjacent monoliths is only created
when shear keys are constructed in the leader block. In the case of induced
Fig. 9
Potential bridging in gravity dam after hydration heat dissipated.
Pontage potentiel dans un barrage-poids après dissipation de la chaleur
d’hydratation.
1 Zone subject to cooling and 1 Zone sujette au refroidissement et
shrinkage au retrait
2 3-Dimensional bridging in gravity 2 Pontage tridimensionnel dans un
dam barrage-poids
386
Q.104 - R.23
joints in RCC, however, a tolerance of 50 mm is typically applied for both the

horizontal and vertical alignment and while the observance of this requirement is
not generally measured, the construction techniques applied imply that this figure
probably representative a typical average alignment accuracy. Consequently, the
surface of an induced contraction joint in an RCC dam will indicate a waviness, or
corrugations, both in the horizontal and vertical directions, which effectively implies
the presence of shear keys between adjacent blocks. When contraction on the joints
results in openings of no more than a few millimetres, shear keys that can transfer
shear forces between adjacent blocks, and further encourage the bridging effects
above, are consequently commonly created in RCC dams.
6.2. IMPLICATIONS
The effects and behaviour described imply that a full understanding of the
structural function of an RCC arch, arch-gravity, or gravity dam in a narrower val-
ley constructed in low SRC RCC can only be developed through an appropriate
thermo-mechanical analysis that takes cognisance of the actual variation of SRC
with compression stress level. Such an analysis currently requires an assumption of
an SRC function, such as illustrated in 5., reviewing the SRC experienced in each
zone of the dam structure on the basis of the temperature rise experienced during
hydration and the compressive stress experienced as a consequence of constraint
during thermal expansion. Only with such an analysis can the extent and influence
of the behaviour and effects described be quantified, allowing the designer to estab-
lish the true response of the structure under load and consequently, in the case
of a gravity dam, whether the fundamental assumption of 2-dimensional structural
function is preserved.
The related analyses are complex and must simulate the construction process,
reviewing the maximum temperatures and compression stresses experienced in the
various parts of the structure under thermal expansion over the period of 4 to 28
days after concrete placement.
7. CONCLUSIONS
Until recently, RCC dam design did not fully acknowledge the differences in
structural performance consequential to horizontal, as opposed to vertical con-
struction and the inherent inclusion of shear keys through the methods used to
construct induced (rather than formed) contraction joints. With the use of a low
stress-relaxation creep RCC accentuating the related effects, the impacts of differ-
ent levels of stress relaxation creep have only recently been considered in RCC dam
design.
387
Q.104 - R.23
These developments imply that there is another parameter to be considered

and optimised in RCC mix design and the associated testing requirements should not
be overlooked in any important materials and RCC investigation and development
programme.
Consequently, careful consideration, investigation and laboratory testing is

necessary during the RCC mix development process for dam structures where a
design sensitivity to the actual level of hydration-cycle stress-relaxation creep exists.
REFERENCES
[1] SHAW, Q.H.W. 2010. A new understanding of the early behaviour of RCC in
large dams. PhD Thesis. University of Pretoria. South Africa.
[2] USACE. 1994. Arch Dam Design. Engineering Manual. EM 1110-2-2201.

Department of the Army. United States Corps of Engineers. Washington, DC.
USA.
[3] Nawa, T & Tomoali, H. Autogenous Shrinkage of High-Performance Concrete.

Proceedings of the International Workshop on Microstructure and Durability
to Predict Service Life of Concrete Structures. Sapporo, Japan. 2004.
[4] Bafang, Z. Thermal stresses and temperature control of mass concrete. The
China Institute of Water Resources and Hydropower Research and Chinese
Academy of Engineering. Beijing, 2014.
[5] Shaw, Q.H.W. The influence of stress-relaxation creep on the design of large
RCC dams. Symposium – Knowledge-based Dam Engineering. 85th ICOLD
Meeting. Prague, Czech Republic. July 2017.
[6] ICOLD. Bulletin 177. RCC Dams. International Commission on Large Dams.
Paris, France. October 2018.
388
Q.104 - R.24
-------
-------
ADAPTATION DE LA FORMULATION ET DE LA MISE EN ŒUVRE DU BCR DU

BARRAGE TOUDGHA (∗)
Abdelilah BOUKAIDI LAGHZAOUI

DIRECTEUR DU PÔLE AMÉNAGEMENTS HYDRAULIQUES ET
HYDROAGRICOLES, Alkhibra
Moulay Lhassan FAIK

Chef d’Aménagement du barrage Toudgha, DIRECTION DES AMÉNAGEMENTS
HYDRAULIQUES
Ahmed TISSIR
Ingénieur résident sur le barrage Toudgha, Alhkibra
Ahmed ROJDAM
Chef de Laboratoire Extérieur sur le barrage Toudgha, LABORATOIRE PUBLIC
D’ETUDES D’ESSAIS
MAROC
RÉSUMÉ
Le barrage Toudgha sur l’oued du même nom est sous construction depuis
2016 et son achèvement est prévu en 2021. Ce barrage est construit dans le cadre
de protection de la vallée et des gorges emblématiques de Toudgha contre les
inondations et pour assurer les besoins en eau potable des populations avoisinantes
et d’irrigation des périmètres à l’aval.
Le barrage, situé à 30 km de la ville de Tinghir contrôle un bassin versant

de 330 km2 avec un apport moyen annuel de 13,1 hm3 . C’est un barrage de type
poids réalisé en BCR. Il a une hauteur maximale de 54,00 m sur terrain naturel et
de 67,50 m sur fondation. Le volume du BCR à mettre en place est de 213.000 m3 .
∗ Adaptation of the Toudgha’s dam RCC formula and placing
DOI: 10.1201/9781003211808-24 389

Q.104 - R.24
La carrière ouverte par l’Entreprise chargée des travaux pour les besoins en
matériaux de confection du BCR et béton conventionnel se situe à l’amont immé-
diat du barrage dans la série de calcaire dolomitique. Les essais de production du
sable fillérisé à partir de cette carrière n’ont pas été concluants. Ce qui a amené
l’Entrepreneur à s’approvisionner à partir de la carrière calcaire de Timkit située à
20 km du site ou le rocher est beaucoup plus tendre et plus facile à broyer.
Les essais d’étude du BCR ont concerné 7 formulations utilisant toutes le

sable fillérisé de Timkit dont 3 formulations ont fait l’objet de planches d’essais.
Des difficultés de mélange du sable fillérisé de Timkit et du sable 0/4 pro-

duit du concassage du calcaire dolomitique du site ont poussé l’Entrepreneur à
abandonner l’utilisation du sable 0/4 au profit du sable de Timkit en totalité. Afin de
respecter le fuseau granulométrique du cahier de charge, un sable traité par con-
cassage et lavage de classe 0/4 continue d’être utilisé. Cette adaptation a permis
une simplification de production et un gain substantiel de cadences.
Une deuxième adaptation de la formulation du BCR a été dictée par le souci

de rééquilibrer le bilan de production des matériaux de construction au niveau de
la station de traitement des matériaux. L’important excédent enregistré au niveau
de la production de la fraction sableuse 2/4 et le déficit constaté au niveau de la
production du sable 0/4 lavé du site justifiait ce choix.
Le barrage est doté d’une galerie de pied amont (GPA) de 3,00 m de largeur qui
sert à la réalisation des travaux de drainage et d’injection du voile d’étanchéité. La
faible pente de la rive droite était incompatible avec une mise en œuvre horizontale
des couches du BCR.
Des essais de mise œuvre du BCR en couches inclinées selon la pente du

fond de fouilles en rive droite furent réalisées. Des résultats probants ont été obtenus
et le procédé a été généralisé sur toute la rive droite.
Sur cette même rive, le parement aval du barrage en forme de gradins de 3,00
m de hauteur, nécessitant la mise en place de coffrage, ne pouvait pas être amorcé
depuis le pied aval, puisque l’espace disponible entre le coffrage du piédroit aval
de la GPA et celui des gradins ne permettait pas une évolution aisée des engins
d’étalage et de compactage du BCR. Le choix fut alors pris pour mettre en œuvre
le BCR en plein fouilles (sans construction de gradins), et ce jusqu’à atteinte du toit
de la GPA à partir duquel les couches du BCR ont été
SUMMARY
The Toudgha dam on the Toudgha River is under construction since 2016 and
its projected to be achieved in 2021. This dam’s objectives are the protection of
390
Q.104 - R.24
the iconic Toudgha Valley and canyons and to guarantee the water needs of the
surroundings.
The dam is located 30 km near the city of Tinghir and controls a 330 km2
watershed with an average annual intake of 13.1 hm3 . It is a weight type dam realized
in BCR, with a maximal height 54.00 m on a natural field of 67.50 m on foundation.
The BCR volume to set up is 213,000 m3 .
The quarry opened by the company in the charge of the construction is located
on the immediate upstream. It provides the BCR and concrete. But the fillerized
sand tests were not satisfactory so the company had to stock up from the lime-
stone quarry of Timkit 20 km nearby where the rocks are more tender and easier to
grind.
The study tests of BCR concerned 7 formulations that uses fillerized sand from
Timkit. 3 out of these have been subject to test boards.
Some difficulties concerning the mixing of fillerized Timkit sand and 0/4 sand
made the company to abandon the use of 0/4 sand replacing it with fillerized sand.
To respect the granulometric specifications, a sand treated by crushing and class
0/4 washing continues to be used. This adaptation led to a simplification of materials
production and a substantial gain of rhythm.
A second adaptation of the BCR formulation had been decided to rebalance the
production of construction materials at the materials processing station. An impor-
tant excess had been detected concerning the 2/4 sandy fraction production, a deficit
of 0/4 washed sand production had also been detected.
The dam has an upstream foot gallery (GPA) of 3.00 m width used in the
fulfillment of drainage and injection of the waterproofing veil. The right bank’ low
slope was incompatible with the horizontal implementation of the BCR layers.
Tests of BCR implementation in inclined layers in accordance with the

right bank’s excavation bottom’s gradient had been conducted. Conclusive results
have been obtained and the process has been generalized throughout the right
bank.
On this same bank, the dam’s downstream facing in the form of steps of 3.00 m
height need the setup of formwork, could not be initiated from the downstream foot
because there was not enough space between the GPA’s downstream pedestal
formwork and the layer’s formwork for the BCR’s display and compaction engines
evolution. A choice had been made, the BCR will be implemented in full excavation
(without construction of layers) until the BCR’s roof (from which the layers had been
achieved horizontally) is reached.
391
Q.104 - R.24
1. INTRODUCTION
Le barrage Toudgha sur l’oued du même nom est sous construction depuis
2016 et son achèvement est prévu en 2021. Ce barrage de capacité de retenue
normale de 33.2 hm3 est construit dans le cadre de la protection de la vallée et
des gorges emblématiques de Toudgha contre les inondations et pour assurer les
besoins en eau potable des populations avoisinantes et d’irrigation des périmètres
à l’aval.
Le barrage, situé à 30 km de la ville de Tinghir contrôle un bassin versant de

330 km2 avec un apport moyen annuel de 13,1 hm3 . La crue millénale de projet
et la crue décamillénale sont estimées respectivement à 1252 m3 /s et 1761 m3 /s.
C’est un barrage de type poids réalisé en BCR.
Le volume à mettre en place du BCR est de 213.000 m3, ce qui a nécessité

la recherche d’autres gites de filer autre que le calcaire dolomitique du site dont la
production du sable fillérisé n’a pas été concluante.
L’approvisionnement du sable fillérisé à partir de la carrière calcaire de Timkit

située à 20 km du site a permis d’obtenir une formulation adéquate du BCR tout
en cherchant à l’optimiser pour rester dans une marge de cout de production
acceptable.
Une adaptation du mode de construction du barrage a dû être introduite au

niveau de la rive droite présentant une faible pente.
2. CARACTERISTIQUES GENERALES DU PROJET
Le barrage Toudgha a une hauteur maximale de 54,00 m sur terrain naturel

et de 67,50 m sur fondation. Il a un parement amont vertical et un parement aval
de pente 0.9H/1V. De longueur en crête de 620 m, il est doté d’un évacuateur de
crues à seuil libre en partie centrale de 80 m de longueur pouvant transiter un débit
maximum de 796 m3 /s correspondant à la crue millénale et un débit de 1190 m3 /s
correspondant à la crue décamillénale (Fig. 1 et 2).
La Fondation du barrage est constituée de calcaire argileux compact en rive

gauche et de calcaire dolomitique en fond de vallée et en rive droite avec des inter-
lits marneux de faible épaisseur et de pendage apparent plutôt favorable de 7˚ vers
l’amont. Contrairement à la rive gauche, la rive droite présente une pente faible de
7,9%.
392
Fig. 1
Vue en plan générale du barrage.
General plan view of the dam.
1 Crête du barrage 6 Axe dérivation provisoire 11 Axe vidange de fond 16 Chambre prises AEP
Crest of the dam Temporary diversion axis Bottom outlet axis AEP outlets chamber
2 Parement aval 7 Chenal amont 12 Chambre des vannes de la vidange de fond 17 Galerie périmètre amont
Downstream face Upstream channel Bottom outlet gates chamber Upstream perimetral gallery
3 Joint amont/aval 8 Batardeau amont 13 Axe prise agricole 18 Route d’accès aux
Upstream/downstream joint Upstream cofferdam Agricultural outlet axis chambres des vannes
4 Joint amont 9 Chenal aval 14 Chambre prise agricole Access road to the bottom
Upstream joint Downstream channel Agricultural outlet chamber outlet gates chamber
5 Axe évacuateur de crues 10 Batardeau aval 15 Entonnements prises AEP
Spillway axis Downstream cofferdam AEP outlet entonnements
Q.104 - R.24
Fig. 2
Coupe transversale du plot déversant.
Transversal section of the overtopping block.
1 Alluvions grossières 6 Béton compacte au rouleau

Coarse alluviums (BCR) Rolled concrete
2 Calcaires dolomitiques 7 Béton de catégorie 25-31.5/27
Dolomitic limestones arme de l’EVC 25-31.5/27
3 Joint marno-argileux category of spillway reinforced
Marl-clay joints concrete
4 Batardeau amont en BCR 8 Pont sur EVC
Upstream cofferdam in rolled concrete Bridge over spillway
5 BCR avec traitement systématique des levées 9 Galerie périmètrale amont
Rolled concrete with systematic treatment of lifts Upstream perimetral
3. FORMULATION INITIALE DU BCR
La zone d’emprunt des matériaux de construction est une carrière située à

environ 1km à l’amont du barrage. Le rocher exploité est constitué de calcaire
dolomitique massif. L’entreprise a utilisé cette carrière pour produire les différents
granulats pour BCR. La cadence de production du sable fillérisé du broyeur à barres
installé au niveau de la station de traitement des matériaux était faible de l’ordre de
20 m3 par jour. Les essais de production industrielle donnaient lieu à un sable fil-
lérisé avec un taux de fines faible et ne permettant pas d’aboutir à un sable mélange
conforme aux exigences du cahier de charges.
394
Q.104 - R.24
L’approvisionnement d’enrochements de calcaire friable à partir d’une carrière

située à Timkit à 20 km du site du barrage fut décidé au vu des bons résultats
d’utilisation de ce matériau lors de la construction du barrage Timkit.
Le sable fillérisé obtenu par le concassage du calcaire de Timkit de classe 0/4

referme près de 65% de fines avec un indice de plasticité non mesurable. La valeur
de l’essai MBf est de 1.5 g/kg. Ce sable fillérisé répond donc à toutes les exigences
du cahier de charges.
Le sable 0/4 issu du concassage des enrochements calcaires du site du bar-

rage destiné à la confection du BCR a été utilisé initialement sous forme brute
mélangé au sable fillérisé afin d’optimiser l’utilisation de ce dernier dans la confec-
tion du sable S1 et sous forme lavé à la station de traitement des matériaux comme
sable S2.
Le sable brut 0/4 du site contient en moyenne 9.3% de fines dont la valeur
MBf est de 3.3 g/kg. Son équivalent de sable est en moyenne de 75%.
Le sable 0/4 du site traité par lavage contient en moyenne 2.2 % de fines et
un équivalent de sable de 80%.
Les graviers pour la confection du BCR ont été obtenus par concassage des
calcaires du site et criblage à la station de traitement des matériaux installée à
l’amont du barrage. Les contrôles granulométriques de ces agrégats montrent que
ces matériaux sont caractérisés par une propreté, un taux des fines et une dureté
conforme aux exigences du cahier de charges.
Le ciment utilisé pour la confection du BCR provient de la cimenterie de Mar-

rakech. Il est de la classe CPJ45 et présente des caractéristiques physiques et
mécaniques conformes aux normes marocaines en vigueur.
L’adjuvant Chrysoroute incorporé au BCR est un retardateur de prise. Les

résultats des analyses physiques réalisés sur ce produit montrent qu’il était conforme
à sa fiche technique.
Des essais d’étude et des planches d’essais sur BCR ont été conduits
au démarrage du chantier avec les formulations suivantes : 6/36/100, 6/38/100,
7/36/100 et 7/38/100.
La formulation retenue est 7/38/100 utilisant les différents sables produits dans
les proportions suivantes :
Sable S1 : Sable fillérisé de Timkit à hauteur de 20% mélangé au sable

concassé brut 0/4 du site à hauteur de 80% ;
Sable S2 : Sable 0/4 du site traité par lavage.
395
Q.104 - R.24
Le sable mélange pour BCR est recomposé à hauteur de 85% du sable S1 et

15% du sable S2.
Les performances mécaniques du BCR mis en œuvre avec cette formulation

étaient très satisfaisantes. La résistance à la compression obtenue est de 12.6 MPa
en moyenne à 28 jours et 16.9 MPa à 90 jours.
Cette formulation a été utilisée pour la confection du BCR mis en œuvre en

fond de vallée du barrage à hauteur de 40.000 m3 puis abandonnée en raison des
difficultés constatées lors des opérations de mélange du sable fillérisé de Timkit et
du sable concassé 0/4 du site.
4. PREMIERE ADAPTATION DE LA FORMULATION DU BCR
Afin de faciliter l’utilisation du sable fillérisé de Timkit, il a été décidé

d’utiliser ce matériau sans recomposition préalable avec le sable concassé brut
0/4 du site. La consommation du sable 0/4 lavé du site a cependant nettement
augmenté.
La formulation visée du BCR est restée inchangée (7/38/100), utilisant le sable

fillérisé de Timkit et le sable 0/4 lavé du site dans les proportions suivantes :
Sable S1 : Sable fillérisé de Timkit à hauteur de 30% du poids total de la

fraction sableuse rentrant la formulation du BCR ;
Sable S2 : Sable 0/4 du site traité par lavage à hauteur de 70% du poids total
de la fraction sableuse rentrant la formulation du BCR.

étaient aussi satisfaisantes que la première formulation.
Cette formulation a été utilisée dans la confection du BCR mis en œuvre dans
le barrage pour un volume total de 100.000 m3 durant la période juillet 20019-juillet
2020.
5. DEUXIEME ADAPTATION DE LA FORMULATION DU BCR
A partir de juillet 2020, la nécessité d’adapter à nouveau la formulation du

BCR, particulièrement en ce qui concerne la fraction sableuse, a été dictée par
le souci de rééquilibrer le bilan de production des matériaux de construction au
niveau de la station de traitement des matériaux. L’important excédent enregistré
396
Q.104 - R.24
au niveau de la production de la fraction sableuse 2/4 et le déficit constaté au niveau

de la production du sable 0/4 lavé du site justifiait ce choix.
Il s’agissait de concasser le matériau excédentaire 2/4 afin de produire un

sable 0/4. Ce sable étant très propre devait cependant être corrigé par un sable
d’apport 0/2 provenant de Timadriouine (région de Tinghir).
La formulation visée du BCR est restée inchangée (7/38/100) utilisant le

sable fillérisé de Timkit et le sable 0/4 recomposé à partir du sable 0/4 de con-
cassage du sable 2/4 et du sable de Timadriouine 0/2, selon les proportions
suivantes :
Sable S1 : Sable fillérisé de Timkit ;
Sable S2 : Sable 0/4 recomposé à partir du sable 0/4 de concassage du sable

2/4 à hauteur de 85% et du sable de Timadriouine 0/2 à hauteur de 15%.
Le sable mélange pour BCR est recomposé à hauteur de 30% du sable S1 et

70% du sable S2.

sont très satisfaisantes.
6. ADAPTATION DE LA METHODE DE MISE EN ŒUVRE DU BCR EN RIVE

DROITE
Le barrage Toudgha est doté d’une galerie de pied amont (GPA) de 3,00 m
de largeur qui sert à la réalisation des travaux de drainage et d’injection du voile
d’étanchéité. La faible pente de la rive droite (7.9%) était incompatible avec une
mise en œuvre horizontale des couches du BCR (Fig. 3 et 4).
Des essais de mise œuvre du BCR en couches inclinées selon la pente du

fond de fouilles en rive droite furent réalisées. Des résultats probants ont été obtenus
et le procédé a été généralisé sur toute la rive droite.
Sur cette même rive, le parement aval du barrage en forme de gradins de

3,00 m de hauteur, nécessitant la mise en place de coffrage, ne pouvait pas être
amorcé depuis le pied aval du barrage, puisque l’espace disponible entre le coffrage
du piédroit aval de la GPA et celui des gradins ne permettait pas une évolution aisée
des engins d’étalage et de compactage du BCR. Le choix fut alors pris pour mettre
en œuvre le BCR en plein fouilles (sans construction de gradins) et ce jusqu’à
atteinte du toit de la GPA à partir duquel les couches du BCR ont été réalisées
horizontalement (Photo 1).
397
Q.104 - R.24
Fig. 3
Mise en place du BCR en rive droite plein-fouille par couches inclinées Installation
of the rolled concrete on the right bank by inclined layers.
Fig. 4
Coupe B-B - Mise en place du BCR en rive droite en plein-fouille par couches
inclinées.
Section B-B - Installation of the rolled concrete on the right bank in full excavation
by inclined layers.
1 Galerie périmètrale amont 4 Béton compacte au
Upstream perimeter gallery rouleau (BCR)
2 Béton de catégorie 25-31.5/23 Rolled concrete
25-31.5/23 concrete category 5 Limite des couches inclinées de
3 BCR avec traitement systématique la 1ère phase du BCR
des levées boundary of the inclined layers of the
Rolled concrete with systematic 1st rolled concrete phase
treatment of lifts
398
Q.104 - R.24
Fig. 5
Mise en place du BCR en rive droite en plein-fouille par couches inclinées.
Installation of the rolled concrete on the right bank in full excavation by inclined
layers.
399
Q.104 - R.25
-------
-------
STRENGTHENING OF BELI ISKAR CONCRETE GRAVITY DAM (∗)
Anton TZENKOV
Senior Dam Engineer, STUCKY, Switzerland
Dimitar KISLIAKOV
Professor, DEPT. OF HYDRAULIC, IRRIGATION AND DRAINAGE
ENGINEERING, UACEG, Sofia, Bulgaria
Ognyan TODOROV
Owner and Senior Engineer, AQUAPRO ENGINEERING EOOD, Sofia, Bulgaria
BULGARIA
SUMMARY
The concrete gravity dam "Beli Iskar" in Bulgaria was built in the period 1936–
1945 in the Rila Mountain to supply drinking water to the capital Sofia. It provides a
typical example of the so-called alkali-silica reaction (ASR) of concrete caused by
the snow melt water in the past operating time. In 2002, the upstream surface was
covered with a geosynthetic sealing membrane as the first phase of the structural
rehabilitation of the dam, and the contact area to the rock foundation was grouted.
In the second phase of the planned rehabilitation activities, the seismic safety of the
dam is also to be brought up to date. Extensive dynamic investigations were carried
out in this century based on the available input information about the concrete dam,
however with the aim of assessing its load-bearing capacity and stability (incl. seis-
mic risk) in relation to the current regulatory requirements in Bulgaria. Based on the
results obtained, solutions for the structural safety of the dam were proposed. Con-
ceptual design for this purpose was also recently completed, in which two structural
design solutions were developed based on computations with design earthquake
excitations that had been determined for its site. In the present work, one of these
∗ Confortement du barrage-poidsde Beli Iskar
400 DOI: 10.1201/9781003211808-25

Q.104 - R.25
strengthening solutions is considered. The model, the excitation parameters and the
results are presented in detail. The latter clearly show that the proposed and pre-
sented here structural solution for dam strengthening completely fulfills its purpose,
which was checked for two specifically defined intensities of the earthquake impact.
RÉSUMÉ
Le barrage-poids de Beli Iskar en Bulgarie a été construit dans la période

1936-1945 dans la montagne de Rila pour alimenter en eau potable la capitale
Sofia. Il constitue un exemple typique de la réaction alcali-silice (RAS) du béton
qui a été provoquée par l’eau de fonte des neiges lors de la période d’exploitation
passée. En 2002, le parement amont a été recouvert d’une membrane d’étanchéité
géo-synthétique dans le cadre de la première phase de la réhabilitation structurelle
du barrage, et la zone de contact avec la fondation rocheuse a été injectée de
coulis. Dans la deuxième phase des réhabilitations prévues, la sécurité sismique
du barrage doit également être mise à jour. Des études dynamiques approfondies
ont été menées sur la base des informations disponibles sur le barrage-poids, dans
le but d’évaluer sa capacité portante et sa stabilité (y compris le risque sismique)
par rapport aux exigences réglementaires en vigueur en Bulgarie. Sur la base des
résultats obtenus, des solutions pour assurer la sécurité structurelle du barrage
ont été proposées. Un projet conceptuel à cet effet a également été récemment
achevé, dans lequel deux solutions structurelles ont été développées sur la base
de calculs avec les sollicitations sismiques qui avaient été déterminées pour le site.
Dans ce rapport, une de ces solutions de confortement est examinée. Le modèle,
les paramètres de sollicitation sismique et les résultats sont présentés en détail. Ces
derniers montrent clairement que la solution structurelle proposée et présentée ici
pour le confortement du barrage remplit complètement son objectif, qui a été vérifié
pour deux intensités d’impact sismique spécifiquement définies.
1. INTRODUCTION
After 70 years of operation under the harshest high mountain conditions, even
the best-designed and maintained concrete dam needs certain renovation activities
to ensure its operational reliability for the coming decades of its operational life.
Such measures are even more necessary if the dam is to meet the current regulatory
requirements in terms of dam safety. The Beli Iskar (Bulg.: “ ”) concrete
gravity dam in the Rila Mountain in Bulgaria has been in operation for 75 years, Fig. 1,
Fig. 2 [1], and if this time is counted from the start of construction – about 85 years.
This fact requires and argues the implementation of thorough rehabilitation activities
to ensure the load-bearing capacity and stability of the dam and thus also to extend
its reliable operating life.
401
Q.104 - R.25
Fig. 1
A look at the Beli Iskar dam (author’s photograph).
Un aperçudu barrage de Beli Iskar (photo de l’auteur).
Fig. 2
The Beli Iskar dam. Layout and longitudinal section [2].
Le barrage de Beli Iskar. Vue en plan et coupe longitudinale [2].
2. INITIAL DESIGN AND CONSTRUCTION. IDENTIFIED PROBLEMS
The purpose of the Beli Iskar dam is water supply for the city of Sofia. Its
construction began in 1936. The planning was carried out carefully, including a
contribution of the famous German structural engineer Prof. Dr.-Ing. A. Kleinlogel.
According to the original contract, the dam should be completed within four years.
In fact, during the construction period from 1936 to 1945, three different companies
were active on the construction site with different levels of experience in dam con-
struction, particularly in the high mountains. Of course, this had an adverse effect
on the quality of the placed concrete. During this time, the construction work was
interrupted from October to May due to the harsh winter conditions with very low tem-
peratures, but without corresponding conservation measures for the structural parts
402
Q.104 - R.25
that had already been built. Finally, the dam was completed and put into operation
by the municipality of Sofia between 1944 and 1945.
The main parameters of the dam are as follows:
Crest elevation: 1888.00 m a.s.l.
Crest width: 3.40 m
Crest length: 533.00 m
Maximum heightofthedam: 50.70 m
Total storagevolume: 15.3 mil. m3
Usablestoragevolume: 14.15 mil. m3
Catchmentarea: 28.1 km2
Slopes: upstream: 1:0.03, downstream: 1:0.683
The damconsistsof 35 structuralblockswith a widthof 7 m to 16 m, respectively.
The dam is located in the high mountain, at elevation above 1800 m a.s.l., in
an area with particularly difficult weather conditions - temperatures in winter down to
–35◦ C and extremely soft snow melt water, which leads to an intensely pronounced
alkali-silica reaction (ASR) in the long operating period, mainly because common
Portland cement was used at that time. In many places, the carbonized deposits are
already several centimeters thick. The process of concrete placement has also pre-
determined some other weak points in the dam structure. The concrete compaction
was only carried out by hand ramming, whereby the already month-old surfaces
were not additionally processed. This led to the formation of undesirable construc-
tion joints and cavities through which the seepage and atmospheric water could
work deeper and deeper into the dam body over the years. In addition, the following
further damages to the dam structure were found: damage to the block joints, leaks
everywhere in the inspection gallery with carbonized deposits, damage to the rein-
forced concrete structure of the machine room of the bottom outlet with intensively
progressing corrosion of the reinforcement. The most important causes that have
led to the deterioration of the concrete’s quality over the years are the alkali reaction
and the aging of the concrete. This finding was also confirmed by the concrete sam-
ples taken in 1961, 1982, 1999 and 2012-2013, whereby test specimens of different
ages were examined. Furthermore, the drainage boreholes in the dam underground
were damaged in the first 4–5 years after commissioning. From 1965 to 2002, new
drainage boreholes were drilled three times. The vertical drainage in the dam body
does not work already since 2002. This leads to the development of a seepage field
inside the dam structure which is inadmissible and additionally affects the stress
field in the dam body and its stability.
403
Q.104 - R.25
In addition to the purely physical reasons for the urgently needed dam reha-
bilitation, there is another important reason to carry out a structural refurbishment
of the dam. At the time of the construction, there were still no regulations in Bul-
garia regarding the consideration of seismic excitation. Nowadays it is well known
that Bulgaria is located in an earthquake prone area, and in the last few decades
national standardization in this area has also developed intensively (the last edition
of the national seismic codes is from 2012). The dam is now located in an intensity
zone IX (MSK scale) with maximum ground acceleration (PGA) 0.27 g defined for a
return period of 1000 years. Furthermore, the national codes prescribe that for dams
of the highest importance class to which this dam also belongs, a particular site-
specific earthquake hazard analysis is to be carried out to determine the relevant
input parameters. The latest seismic analysis of the dam carried out in 2004–2006 [3]
concluded that the stability of the dam is not guaranteed in the event of strong earth-
quake whose parameters were defined according to the national seismic code. In
the frame of the same study, several variants for structural strengthening of the dam
were developed, and the solution with strengthening structure on the downstream
side of the dam was then chosen for further development.
3. PREVIOUS STUDIES AND REHABILITATION WORKS
In the first phase of the subsequent rehabilitation activities, the reservoir was
emptied in 2002, and a PVC sealing geomembrane was installed on the upstream
side, Fig. 3 [1]. Repair works on the grout curtain and on the drainage system were
carried out, too.
In the second phase of the rehabilitation activities, the municipality of the city
of Sofia tendered and awarded the development of a conceptual design in all special
fields for the structural rehabilitation and upgrade of the dam and its appurtenant
facilities. As part of this design [2], the required solution was developed in two
variants, and all verifications based on related calculations were carried out.
In fact, conservative models and procedures have been used in all previous
investigations which all led to conclusions that conform to the code requirements,
moreover, they were on the safe side. This is also understandable due to the strategy
used for the design of dam strengthening. In this story, a question that remains open
is how the dynamic behavior of the dam would look under the determinantearthquake
excitations, i.e. SEE and OBE according to [4].
An attempt to answer the question about the seismic safety of the existing dam
was made in [1] considering a non-linear mechanism for crack initiation and propa-
gation. Two earthquake scenarios were used [5] with PGA values scaled according
to the recommendations of [4]. Additionally, and for purpose of comparison with
the requirements of the current Bulgarian seismic code (also valid for large dams
404
Q.104 - R.25
Fig. 3
The Beli Iskar dam. Upstream side with the spillway (author’s photograph).
Le barrage de Beli Iskar. Côté amont avec le déversoir (photo de l’auteur).
unlike Eurocode), the two used earthquake excitation scenarios were scaled to a
level corresponding to a return period of 1000 years.
On the other hand, the problem remains of the solution selection for structural
strengthening. As mentioned above, several structural solutions were developed
and analyzed. The most comprehensive summary of all known related studies and
design developments was presented in [3].
In the current paper, the solution chosen by the relevant authorities in 2018
is presented and analyzed in more detail. It is one of the two structural develop-
ments proposed by the conceptual design [2] consisting of a massive concrete
buttress placed on the downstream surface of the dam as an additional layer.
This concept also could serve some heightening of the dam if appropriate with
respect to the water resources management of the reservoir. In fact, the ideas of
all studied and proposed solutions for the “Beli Iskar” dam are not new. A thorough
and well-structured presentation of the design philosophy and the approaches for
strengthening of existing dams with an emphasis on their seismic safety is given
in [6]. All main aspects of this field are presented and discussed in detail. The
comprehensive report [7] also presents a special study of failure mechanisms and
scenarios resulting from the development of concrete deterioration of the gravity
dam body.
Case studies implementing such solutions are presented in [8], [9], and
[10]. Innovative and different strengthening solutions also are sometimes pro-
posed as for example the one presented in [11]. However, regarding the latter,
405
Q.104 - R.25
in order to be efficiently applied, it would require emptying the reservoir which,

in the present case, is not possible. In the field of arch dams, the recent project
[12] should also be mentioned as conceptually related to the one presented
here.
4. SEISMIC INPUT RECORDS AND EXCITATION SCENARIOS
As already mentioned above and in the case under consideration, for the
dams of the highest class of importancein Bulgaria, a special seismological study is
to be carried out in order to determine the relevant seismic input parameters for the
particular site. Within the scope of such a study carried out in 2004–2006 and used
according to [3], the peak ground acceleration (PGA) values and their associated
annual probabilities of exceedance (APE) and corresponding return periods for the
dam "Beli Iskar" formulated as mean values are as follows (the effect of the statistical
confidence interval was not investigated in this work):
• PGA = 0.305 g, APE = 0.1%, return period 1000 years; and

• PGA = 0.564 g, APE = 0.01%, return period 10000 years.
For the purposes of the design carried out in [2], as well within the frame-
work of the current study, relevant time histories of the ground acceleration also
were required for the dam site. Unfortunately, the strong motion devices built into
the dam could not provide any reliable information. For this reason, a special study
was carried out to determine such input records relatedparticularly to the dam [5].
In this study, two current hypocenters known for their activity in the region were ana-
lyzed, namely the Pernik and Krupnik faults, which have been known for decades.
Representative records of these two hypocenters made at the Rila Monastery were
processed accordingly and supplied for the purposes of the design [2]. This famous
monastery is only 15 km as the crow flies from the dam, and the signals there were
recorded in the three spatial directions on the bedrock, which best corresponds to
the local conditions at the dam site [5].
Taking into account the results obtained in [1], we decided to work here only
with the strong motion records from the Pernik fault since the computational model
of the existing dam experienced considerably heavier damages with these records
applied as earthquake excitation.As part of the present work, we further scaled these
two records of the horizontal E-W component (applied as upstream-downstream
one) and of the vertical component to the above-mentioned PGA values, Fig. 4.
The response spectrum of the original E-W component of the Pernik earthquake
for different damping values is presented in Fig. 5. Thus, we received a total of two
load cases - each set of records (US-DS and vertical) scaled to two PGA values,
respectively. Further, the so formulated load cases were applied to the computational
dam-foundation finite element models of both the existing and the strengthened dam,
respectively.
406
Q.104 - R.25
Fig. 4
Acceleration time histories Pernik scaled to PGA 0.564 g.
AccélérogrammesPernik à l’échelle PGA 0.564 g.
Fig. 5
Response spectrum of the original horizontal acceleration time history E-W
component Pernik (with PGA 40.851 cm/sec2 , i.e.0.0416 g)for 0, 2, 5, 10 and 20%
damping [5].
Spectre de réponse de la trace d’accélération horizontale d’origine du composant
E-W Pernik (avec PGA 40.851 cm/sec2 , soit 0.0416 g) pour un amortissement de
0, 2, 5, 10 et 20% [5].
407
Q.104 - R.25
5. SEISMIC VERIFICATION OF THE EXISTING DAM
The seismic verification of the current structure of the dam was carried out
on a two-dimensional finite-element method (FEM) model of the dam-foundation-
reservoir system. The analysis was performed in the time-domain by direct integra-
tion of the equations of dynamic equilibrium for the seismic excitation cases defined
above. The calculations were carried out by means of the computer program DIANA
[13] with the initial assumption of linear-elastic behaviour of the dam and its founda-
tion. The geometrical model was defined for the tallest block of the dam. The valve
chamber located at the toe of the block was neglected.
The two main concrete-type zones in the dam body were considered, namely
the zone of high-cement content (280 kg / m3 ) concrete on upstream side (Zone 1)
and the main zone consisting of mass concrete with amount of cement of 185 kg / m3
(Zone 2). The model of the foundation extends at three times the dam height in the
upstream, downstream, and downward directions. The foundation, which consists
of sound competent granite was considered as isotropic material, and its mass was
neglected. The geometry of the used mechanical model is shown in Fig. 6.
Fig. 6
2D geometrical model of the existing dam.
Modèle géométrique 2D du barrage existant.
The material parameters used in the analysis are defined on the basis of the
thorough and complete study carried out in [3]. Their values are shown in Table 1.
408
Q.104 - R.25
Table 1
Material parameters used in the seismic verification of the existing dam
Caractéristiques des matériaux utilisés dans la vérification sismique du
barrage existant
ELASTIC MODU- COMPRESSIVE TENSILE
DENSITY POISSON’S
MATERIAL LUS STRENGTH STRENGTH
RATIO
STATIC DYNAMIC STATIC DYNAMIC STATIC DYNAMIC
kg/m3 GPa GPa - MPa MPa MPa MPa
Rock 0 15 19 0.3 20 23 0 0
Zone 1 2 300 26 32 0.2 25 28 2.5∗ ; 3.6∗ ;
1.25∗∗ 1.80∗∗
Zone 2 2 300 23 28 0.2 20 23 2.0∗ ; 3.0∗ ;
1.00∗∗ 1.50∗∗
*Tensile strength of mass concrete.

**Tensile strength of construction joints.
The structural system (the dam-foundation system) model is meshed with sec-
ond order plane-strain elements of quadrilateral shape (8 nodes).The finite element
mesh, the boundary conditions, and the applied hydrostatic pressure are shown in
Figure 7.
Fig. 7
Finite element mesh of the dam-foundation system.
Maillage d’éléments finis du système barrage-fondation.
The investigated load combination takes into account the self-weight of the
dam, the hydrostatic pressure at Full Supply Level and a base load due to
earthquake.
It is noted that the hydrodynamic effect for the seismic case is represented by
means of the Westergaard approach and by employing a consistent mass matrix
that is added to the mass matrix of the structural system. Further, an eigenvalue
analysis is carried out in order to define the most important eigen modes of the
dam-reservoir-foundation system.The results of the computed eigen periods for the
cases of empty and full reservoir are given in Table 2.
409
Q.104 - R.25
Table 2
Existing dam eigenperiods and effective mass cumulative percentage
Périodes propres et pourcentage cumulé de la masse effective du barrage
existant
EMPTY RESERVOIR FULL RESERVOIR
MODE
T, SEC X, % Y, % T, SEC X, % Y, %
1 0.1629 59.5 2.2 0.2158 63.4 0.4
2 0.0937 64.8 95.3 0.0972 85.3 25.8
3 0.0759 93.3 98.3 0.0916 92.0 64.5
4 0.0458 98.9 98.7 0.0546 97.6 64.9
5 0.0291 99.2 99.0 0.0349 98.5 65.0
6 0.0278 99.7 99.7 0.0294 98.5 65.5
The dynamic analyses of the dam subjected to strong ground motions are
carried out using the method of Hughes, Hilbert and Taylor with -0.3. The viscous
damping of the system is represented by the Rayleigh approach, with proportionality
constants computed so as to give a modal damping ratio of 5% in the first and the
third vibration modes.
The envelops of the maximum compressive (S3) and tensile (S1) stress in case
of SEE (Pernik Earthquake of PGA = 0.564 g) for the existing dam are presented in
Fig. 8 and Fig. 9, respectively. The results show that the dynamic tensile strength of
the construction joints as well as of the mass concrete material would be exceeded
in extensive zones of both Zone 1 and Zone 2. This would entail initiation and
propagation of cracks that may endanger the capacity of the dam to withstand SEE
ground motions without uncontrolled release of water.
Fig. 8
Envelops of maximum compressive stresses (MPa) in case of SEE.
Enveloppes des contraintes de compression maximales (MPa) en cas de SEE.
410
Q.104 - R.25
Fig. 9
Envelops of maximum tensile stresses (MPa) in case of SEE.
Enveloppes des contraintes de traction maximales (MPa) en cas de SEE.
6. STRENGTHENING CONCEPT AND ANALYSIS OF THE

STRENGTHENED DAM
The strengthening concept is show in Figure 10. It involves the construction of

a concrete buttress placed on the downstream surface of the dam. The buttress shall
be an integral part of the dam body and the details of its construction and connecting
to the existing dam are to be provided during the final design and construction
drawings phase.
Fig. 10
2D geometrical model of the strengthened dam.
Modèle géométrique 2D du barrage renforcé.
411
Q.104 - R.25
The analysis of the strengthened dam is carried out following the same
approach as the one already described.The computed eigenperiodsfor the cases
of empty and full reservoir of the strengthened dam are given in Table 3.
Table 3
Strengthened dam eigenperiods and effective mass cumulative percentage
Périodes propres et pourcentage cumulé de la masse effective du barrage
renforcé
EMPTY RESERVOIR FULL RESERVOIR
MODE
T, SEC X, % Y, % T, SEC X, % Y, %
1 0.1604 67.6 3.4 0.1985 71.5 0.8
2 0.1005 73.9 95.7 0.1018 77.7 63.1
3 0.0740 96.8 98.8 0.0863 95.7 69.3
4 0.0423 99.3 99.1 0.0483 98.4 69.6
5 0.0290 99.4 99.7 0.0322 98.6 69.9
6 0.0264 99.7 99.7 0.0293 98.9 70.0
The envelops of the maximum compressive (S3) and tensile (S1) stress
for the strengthened dam in case of the investigated SEE (Pernik Earthquake of
PGA=0.564g) are presented in Fig. 11 and Fig. 12, respectively. The results show
that both the compressive and the tensile stresses remain lower than the respective
dynamic strength of the concrete material and, with regard to the tensile stresses,
the dynamic tensile strength of the construction joints. An exception is the heel of the
dam, where some local cracking may occur. Although a check of the global sliding
stability is to be carried out based on all the analyses carried out so far, one can
state that the global and the local stability of the dam is ensured.
Fig. 11
Strengthened dam envelops of maximum compressive stresses (MPa) in case of
SEE.
Barrage renforcé : enveloppes des contraintes de compression maximales (MPa)
en cas de SEE.
412
Q.104 - R.25
Fig. 12
Strengthened dam envelops of maximum tensile stresses (MPa) in case of SEE.
Barrage renforcé : enveloppes des contraintes de traction maximales (MPa) en cas
de SEE.
7. CONCLUSIONS
The presented dynamic analysis was intended as a preliminary study aimed

at obtaining a more clear view on the realistic dynamic response of the old Beli Iskar
dam to an earthquake excitation of SEE level. The obtained results showed, how-
ever, that the preliminary assumption of linear elastic material behaviour of the mass
concrete of the dam body delivered fully sufficient confidence about the efficiency of
the proposed strengthening solution. It was clearly shown that the implementation of
such structural strengthening decreases the tensile stresses down to an acceptable
magnitude. Only a small area with high tensile stresses causing possible subsequent
fracturing remained at the dam-foundation contact on the upstream side of the dam
in the case of a severe SEE earthquake. It should be emphasized that for earthquake
excitation of such intensity level this type of damage is acceptable since the overall
stability of the dam is not endangered. Thus, the efficiency of the particular investi-
gated strengthening solution was proven. As a task for future research, only particu-
lar studies on the parameters of the concrete buttress on the downstream side of the
dam remain to be performed for safely removing this small zone of high magnitude
tensile streses with fracture initiating potential, if economically appropriate.
REFERENCES
[1] KISLIAKOV, D. TODOROV, O., Nichtlineare dynamische Untersuchung einer

70 Jahre alten Gewichtsstaumauer in Bulgarien, Proc. 18. Gemeinschafts-
Symposium der Wasserbau-InstituteTU München, TU Graz und ETH Zürich:
413
Q.104 - R.25
“Wasserbau – mehr als Bauen im Wasser“, 29.Juni – 01.Juli 2016, TU

München, Wallgau / Oberbayern, Germany, 2016.
[2] TODOROV O., Conceptual design for the rehabilitation of the dam “Beli Iskar”
– Phase II (in Bulgarian), Egis Bulgaria EAD, 2015.
[3] PRIMA INVEST KONSULT EOOD. Review of the investigations, analyses and
design studies and update of the physical parameters of rock and concrete for
assessment of the conditions and elaboration of a proposal for the strength-
ening and ensuring of the long-term stability of “Beli Iskar” dam (in Bulgarian),
Sofia, 2013.
[4] Selecting seismic parameters for large dams, Guidelines, Bulletin 148, ICOLD-
CIGB, Paris, 2016.
[5] SIMEONOV S. (Head of the working team). Report on the processing and
analysis of the earthquake records of 22.05.2012 (seismicity zone Pernik) and
27.07.2013 (seismicity zone Krupnik), registered at bedrock in the vicinity of
the Rila monastery (in Bulgarian), Report NIGGG, Bulg. Academy of Sciences,
Sofia, 2015.
[6] WIELAND M., Safety AspectsofSustainableStorageDamsandEarthquake

Safety of Existing Dams, Engineering 2, 325–331, Elsevier, 2016.
[7] Selecting Analytic Tools for Concrete Dams to Address Key Events along
Potential Failure Mode Paths, FEMA P-1016, 2014.
[8] GUEST D., Tenterfield Creek Dam Safety Upgrade, Options study, Report No.
DC14010, NSW Public Works, Sydney, Australia, 2014.
[9] MELLAL A., Heightening of an existing gravity dam, Static and Dynamic
analyses, Proc. Numerics in Geotechnics and Structures, Lausanne, 2009.
[10] Advanced technologies to upgrade dams under operation, Water and Disaster
Management Bureau, Ministry of land, Infrastructure, Transport and Tourism,
Japan, 2013.
[11] ZHANG H., OHMACHI T., Seismic cracking and strengthening of concrete
gravity dams, Paper 1410, Proc. 12 th WCEE, Oakland, New Zealand, 2000.
[12] MÜLLER, O., WOHNLICH, A. Safety Enhancement and Strengthening of Les

Toules Arch Dam,Proceedings of the 8 th ICOLD European Club Symposium
Dam Safety-Sustainability in a Changing Environment, 22. – 23.September,
2010, Innsbruck, Austria.
[13] www.dianafea.com
414
Q.104 - R.27
-------
-------
A CASE STUDY OF EFFECTIVE APPLICATION OF INNOVATIVE

MANAGEMENT ON RCC DAM CONSTRUCTION AT KAFUE GORGE LOWER
HYDROELECTRIC PROJECT (∗)
CHINA
SUMMARY
Despite the Chinese enterprises, particularly those associated with infras-

tructure and engineering construction, have a long history since getting access to
African market, they suffer from inadequate working time and low skills level of
human resources in countries of operation, which causes constraints to effective
fulfillment of task and further development of these enterprises. This paper consid-
ers a RCC dam of a Zambian 750MW hydroelectric project under construction by a
large Chinese state-run engineering enterprise as a sample, and studies localiza-
tion management measures taken by the enterprise to increase attendance, relieve
communication obstacle, upgrade management level, improve skills of Zambian
employees. With these innovative localization management measures, the con-
straint mentioned above is substantially meditated and progress of construction is
further guaranteed.
RÉSUMÉ
Bien que les entreprises chinoises, en particulier celles associées à la con-

struction d’infrastructures et d’ingénierie, aient une longue histoire depuis leur accès
au marché africain, elles souffrent d’un temps de travail inadéquat et d’un faible
niveau de compétences des ressources humaines dans les pays d’opération, ce
qui entraîne des contraintes à la réalisation effective des tâches et le développe-
ment ultérieur de ces entreprises. Cet article prend en compte le barrage RCC
d’un projet hydroélectrique zambien de 750 MW en construction par une grande
entreprise d’ingénierie d’État chinoise comme échantillon, et étudie les mesures
∗ Une étude de cas sur l’application efficace d’une localisation de gestion novatrice sur la
construction d’un barrage en BCR au projet hydroélectrique inférieur de Kafue Gorge
DOI: 10.1201/9781003211808-26 415

Q.104 - R.27
de gestion de la localisation prises par l’entreprise pour augmenter la fréquentation,

soulager les obstacles à la communication, améliorer le niveau de gestion, améliorer
les compétences des employés zambiens. Avec ces mesures innovantes de ges-
tion de la localisation, la contrainte mentionnée ci-dessus est largement méditée et
l’avancement de la construction est en outre garanti.
1. BRIEF INTRODUCTION TO THE HOST COUNTRY
1.1. LOCATION AND HISTORY
Zambia, officially the Republic of Zambia, erstwhile Northern Rhodesia, is a

landlocked country in south-central Africa. Its neighbors are the Democratic Repub-
lic of the Congo to the north, Tanzania to the north-east, Malawi to the east,
Mozambique to the southeast, Zimbabwe and Botswana to the south, Namibia to
the southwest, and Angola to the west. The capital city is Lusaka, located in the
south-central part of Zambia.
Originally inhabited by Khoisan peoples, the region was affected by the Bantu
expansion of the thirteenth century. Following European explorers in the eigh-
teenth century, the British colonized the region into the British protectorates of
Barotseland-North-Western Rhodesia and North-Eastern Rhodesia towards the end
of the nineteenth century. These were merged in 1911 to form Northern Rhodesia.
For most of the colonial period, Zambia was governed by an administration appointed
from London with the advice of the British South Africa Company.
On 24th October 1964, Zambia became independent of the United Kingdom.
1.2. HUMAN RESOURCES AT NATIONAL LEVEL
The population of Zambia is approximately 17 million and is concentrated

mainly around Lusaka in the south and the Copperbelt Province to the northwest,
the core economic hubs of the country.
1.3. HUMAN RESOURCES AT THE DISTRICT WHERE THE PROJECT IS LOCATED
The Project is in Chikankata District, Southern Province, the Republic of

Zambia. about 68km from the capital of Lusaka and 50km from Mazabuka Town.
The district boarders with Mazabuka District on the south and south western side,
Chirundu District on the south eastern side, Siavonga District on south eastern side
416
Q.104 - R.27
and Kafue district on the north west and north eastern sides respectively. The district
covers an area of about 2,500 square kilometers.
Chikankata District has an estimated population of 50,909 people, which trans-

lates to 26,696 males and 30,213 females. The employment status in the district in
not so good due to lack of industries in the area, therefore the main source of income
is farming and charcoal burning.
The district though was blessed with the inception of the Kafue Gorge Lower
Hydroelectric Project in the late 2015 which came as a relief to the people of
Chikankata District regarding employment opportunities.
So far Sinohydro Corporation Limited is the main employer in the district with
a workforce of about 4,300 Zambian employees with different skill and abilities.
But since the beginning of the project is has been observed that Chikankata District
lacks skilled labor force in relation to heavy machinery operators like frontend loader
operators, bull dozer operators, excavator operators, drill rig machine operators,
mobile crane operators and even heavy vehicle drivers and as such the project is
forced to source such kind of labor from other districts around Zambia.
It is therefore, recommended that Chikankata residents should find a way

to acquire some skills in such areas to make themselves competitive in the labor
market and not to allow people from other district to surpass them in terms of getting
employment opportunities due to lack of necessary skills.
If this can be achieved then in future Chikankata residents can be more com-
petitive in the labor market. Chikankata District has potential to be exploited by
investors and new industries may mushroom.
1.3.1. Brief introduction to the project
The 750MW Kafue Gorge Lower Hydroelectric Project (herein referred to as

the Project) is located on the Kafue River, some 17.3km downstream of the 990MW
Kafue Gorge Upper Hydropower Station.
Permanent works of the Project include the Employer’s Township, access road
headrace tunnel, and a power generation plant.
The work to be performed under this Contract includes ( without limitation)

planning, design, engineering, procurement, construction, manufacture, transporta-
tion to site, installation, field testing, commissioning, testing and handing over to
the Employer of 750 MW (5*150 MW) hydroelectric project scheme, comprising
infrastructure, civil, hydro-mechanical and electro-mechanical works.
417
Q.104 - R.27
Design and construction of the power generation plant include the following
salient features:
Table 1
Features of the Project
S.N. DESCRIPTION OF FEATURE REMARK
1 Roller Compacted Concrete (RCC) gravity dam, with a maximum height of 139 m
2 Power Intake
3 Headrace Tunnel, with an approximate length of 4.4 km
4 Surge Shaff, with a diameter of 32 m
5 Five Power Tunnels, each with a diameter of 4.8 m
6 Sub-surface Powerhouse, with five (05) generating units having a total installed
capacity of 750MW
7 Tail-water pool
8 330kV Switchyard
The Project is an EPC project contracted to Sinohydro Corporation Limited.

Its total investment cost is USD 1.56 billion.
The construction period of the Project is 1549 days (51 months) and
commissioning of the first generating unit will be 1437 days (48 months) from
commencement of construction works.
The Project is the first large hydropower project that Zambia has undertaken
over the past 40 years. Construction of this project will mitigate power shortage
Zambia is experiencing now and will offer reliable power supply for the development
of the agricultural sector, the mining industry, and other sectors of the economy,
thereby improving the micro and macro economies of Zambia to a higher level.
Upon completion of the Project, the existing installed electric power capacity
of Zambia will increase by 38%. It will also have a positive effect in improving power
reliability of the Power Grid in Southern Africa.
Construction of the Project has greatly contributed to creation of job opportu-

nities. To date, it has provided more than 10,000 employment opportunities for local
communities.
1.4. RCC DAM
The site of the Project has been studied for two dam types as follows:
(i) A CFR Dam with conventional concrete face, a primary gated spillway sit-
uated separately on the left abutment side of the dam as well as a flood
release tunnel;
(ii) A RCC dam with a gated spillway
418
Q.104 - R.27
It has been determined that both dam types are feasible. The RCC option,
however, is preferred by the Employer and shall be the basis for preparation of the
bid. Bidders may additionally submit an alternative bid for a CFR dam.
The 130m-high RCC Dam has a crest width between 8m, a length of approxi-
mately 374.5m, and 83 million cubic meters roller compaction concrete to be placed.
2. APPLICATION OF INNOVATIVE LOCALIZATION MANAGEMENT ON RCC

DAM CONSTRUCTION
2.1. STAFF COMPOSITION
2.1.1. Ratio between Chinese and Zambian employees
At peak period of time for RCC construction, 50 Chinese and 932 Zambian
employees are provided, with the ratio between Chinese and Zambian approximately
0.05:1.
2.1.2. Profession of Chinese and Zambian employees
Table 2
Profession of Chinese Employees
PROFESSION NUMBER REMARKS
Section Head 1
Deputy Section Head 3
Division Head 3
Duty Engineer 5
Technician 3
Foreman (civil, mechanical, electrical, etc.) 35
Table 3
Profession of Zambian Employees
PROFESSION NUMBER REMARKS
Engineer 5
Foreman 23
Operator (Mobile Crane, Tower Crane, Roller, Loader, etc.) 49
Driver (Tripper Truck, Canter, etc.) 68
Skilled Worker (Welder, Electrician, Carpenter, etc.) 199
General Worker 535
Others 53
419
Q.104 - R.27
2.1.3. Constraints of Zambian human resources
When the Zambian employees, particularly those who have no experience

working on hydropower station, are employed and assigned to work at the RCC
Dam Section, skills level of them is generally not so high, which constitute a great
challenge and barricade for successful and smooth implementation of construction
activities.
More than half of them is from Southern Province which is mainly depending
on agriculture other than mines or industry. Those who are adequately skilled to
carry out their assignment only account for minority of the whole team and normally
they are from Copperbelt Province which has the most and skilled industrial workers
across Zambia. Even for these skilled ones, they still have difficulties facing different
construction activities, working environment, work sequence, Chinese foremen, and
types of machinery to use.
Apart from the generally low skills level, low attendance of Zambian employees
is another great challenge. Compared with Chinese employees who could nonstop
work all year round, Zambian employees seem to have different work ethics. The
Project makes payment to them on monthly basis, and they will take at least five
days off every time the salary is paid. Production during this period of payment time
would be greatly reduced, and quite a few construction resources such as Chinese
manpower, machinery and materials must be left idle.
In addition, having grown in a different culture and linguistic background, ways

of thinking and speaking of Zambian employees are quite unlike those of Chinese
ones, which lead to long-standing cooperation and communication obstacles, even
disputes and misunderstandings. This situation makes it really difficult for Chinese
and Zambian employees to fulfill their shared assignment in an effective way.
2.2. INNOVATIVE MEASURES FOR LOCALIZATION MANAGEMENT
2.2.1. Provide full attendance bonus to increase operational use time
To increase operational use time of Zambian employees, the Project has

worked out a form of bonus intended to encourage them to work more days.
Since the very beginning of the Project in January 2016, payment for Zam-
bian employees has been made in two different periods for some reasons, one for
28 days, and the other 35 days, and these two periods take turns, with exceptions
occasionally.
420
Q.104 - R.27
For the period of 28 days, the attendance bonus will be paid to an employee
on the condition that he or she has to work up to not less than 26 days; and for the
period of 35 days, not less than 33 days. Attendance bonus is in direct proportional
to the employee’s working days. Details could be seen in the table below.
Table 4
Breakdown of Attendance Bonus
PERIOD WORKING DAYS ATTENDANCE CALCULATION REMARKS
BONUS
26 Basic Salary for 1 Day 1*Rate*8 h
28 Days 27 Basic Salary for 2 Days 2*Rate*8 h
28 Basic Salary for 3 Days 3*Rate*8 h
33 Basic Salary for 1 Day 1*Rate*8 h
35 Days 34 Basic Salary for 2 Days 2*Rate*8 h
35 Basic Salary for 3 Days 3*Rate*8 h
2.2.2. Select Zambian foremen for mitigating communication constraints
Chinese foremen of different professions account for majority of the Chinese

employees engaged in RCC Dam construction and the main workforce to carry out
basic construction activities. However, almost all the Chinese foremen have not
gained a good command of English language, and similarly, almost all the Zambian
employees could not speak Chinese language. To improve linguistic skills of either
Chinese or Zambian employees in a short time is really difficult, let alone they all
face the fact that schedule for construction has been so tight that they rarely have
time to learn a language.
To mitigate communication constraints between the Chinese and Zambian

employees, a Zambian foreman mechanism is introduced. Currently, a total number
of 23 Zambian foremen are working on RCC Dam Section, accounting for approx-
imately 2.5% of its all Zambian employees. These foremen have been selected
through stringent and comprehensive procedures. They are assessed not only from
respective of skills, experience, but also from respective of cooperative spirit, atti-
tude, attendance, ability of leadership and solving problems within their own teams,
etc. To be a foreman is not a permanent appointment. Dynamic assessment and
evaluation is made to them periodically. Those who fail the periodic assessment and
evaluation, punishment measures such as warning or even relieving his of post may
be taken against him.
As an effective extension of mechanism of Chinese foremen, Zambian foremen

could act as a bridge between Chinese foremen and their Zambian fellow employ-
ees. They could partly take on responsibility of Chinese foremen for organizing and
supervising Zambian fellow employees in an assigned task.
Since application of Zambian foremen mechanism, obvious improvement in

communication constraints and construction efficiency has been noticed.
421
Q.104 - R.27
For those Zambian foremen who are able to fulfill their duty well as a foreman,
the Project pays them ten kwacha per day as foreman allowance, which has been
a great motivation to them.
2.2.3. Recruit well educated Zambian talents to improve construction

management
Standing as the largest scale infrastructure under construction in Zambia, KGL

inevitably requires adequate number of engineering talents to provide qualified tech-
nical guidance. For reasons of cost reduction, enhancing localization, and preparing
local talents for long term development in Zambia and Africa, the Project will hesitate
to bring too many Chinese counterparts. In such a circumstance, the Project is left
no choice but recruit the required talents from local communities. Since 2016, the
Project has been employing the graduates within Zambia. Currently, more than 80
Zambian graduates are working on the Project.
Majored in various subjects such as electrical engineering, mechanical

engineering, civil engineering, safety and environment engineering, etc., these grad-
uates have graduated both Zambian universities, such as University of Zambia, and
Copperbelt University, and foreign ones, such as those in Algeria, Russia and China.
To ensure qualification of the graduates, the Project has requested the engineering
graduates to show their membership certificate issued by Engineering Institution of
Zambia (EIZ for short) when they are looking for employment on the Project.
Considering the RCC Dam is mainly civil works, some Zambian civil and
safety engineers have been assigned for it, and these engineers are effective exten-
sion of Chinese engineering team. One of their advantages is that they could take
advantage of their linguistic skills and cultural background to communicate with the
Employer and Consultant’s staff that are from Zambia and other foreign countries
in a better way. Having practiced and learned from the Chinese counterparts, 2 of
them have been given a promotion both in position and salary.
2.2.4. Training opportunities are provided to the employees for improving skills
level
As aforesaid, skills level of Zambian employees is generally low, which could

not satisfy the demand of particular construction activities, such as metal fabrica-
tion, electrical, mechanical, etc. To remove this bottleneck and fulfill corporate social
responsibility, a skills training school has been built and operated by the Company
at its own cost. So far, it has witnessed successful graduation of 289 trainees from
five intakes. Currently, 61 trainees of the 6th intake are being trained. The trainees
are Zambian youths aged 20 to 30 who have no any kind of industrial skills before
they are recruited as a trainee. They are provided with 6 months intensified practical
and theoretical training for free, and employment opportunities in the Project upon
graduating successfully. Some trainees who successfully graduated from the train-
ing school have been sent to RCC Dam Section to work as welders, mechanics,
surveyors and or electricians.
422
Q.104 - R.27
In addition, some skilled heavy machinery operators are also in great short-
age, such as tower crane operator, mobile crane operator, etc. In this respective,
the Project provides some unskilled operators and even non-operators a chance to
be trained by skilled Chinese operators. When they are considered to be adequately
skilled, they will be advised to undertake examination or test by authorized certifi-
cate issuing institute for accreditation. Meanwhile, the Chinese operators who train
Zambian operator would be given a bonus as training allowance.
Through training offered by the training school and Chinese employees, the
overall skills levels of RCC Dam Section have been improving steadily. To date, skills
of Zambian employees can substantially meet some of the construction activities
requirements.
2.2.5. Group employees and stagger holidays to guarantee continuity of

construction
A great challenge for RCC Dam construction is the period of monthly holidays
of Zambian employees. It seems conventional that the Zambian employees would
take some days (5 days in general) off every time they get paid on monthly basis,
which causes discontinuity of construction of the dam while in most cases they are
expected to work continuously. Frequent and long time off by the Zambian employees
would cause serious idleness of materials and machinery and lead to accumulative
delay in the progress of works. To mitigate this situation, the company decided to
group the employees into Group-A and Group-B, whose payment days are separated
at an interval of 7 days. Group-A accounts for majority of employees while Group-
B is mainly responsible for carrying out some preparation works for the next block
concrete placement. Thus, the whole sequence will not have any disconnection and
efficiency could be evidently improved.
3. FURTHER EXPLORATION OF APPLICATION OF LOCALIZATION

MANAGEMENT IN CONSTRUCTION
Although Chinese enterprises have entered Africa market for a long time and
bilateral relations between China and Africa has been getting closer, various chal-
lenges still has to be addressed. Due to difference in history, culture, language,
social systems, policies, work ethics, etc., engineering construction enterprises
from China has a long way to go in Africa.
Some of the challenges include things such as local employees’ frequent long
time off, lack of self-discipline, low level of skills, slowness to learn, and linguistic
obstacles, shortage of materials or machinery needed, expensive transportation,
exceedingly strict labor laws, etc. could not be substantially addressed in a short
term. Chinese enterprises involved have to keep making efforts to search for more
space for further development in Africa.
423
Q.104 - R.27
With unremitting fulfillment of “The Belt and Road” initiative and increased
exchange and cooperation between China and Africa in various fields and sectors,
more opportunities would be available for Chinese enterprises to explore the African
market and expand their business in Africa.
As one of the key resources during this course, local employees are expected
to account for more of the whole workforce of a Chinese enterprise, and it’s surely a
trend that localization management would help a lot with survival and development
of Chinese enterprises in Africa.
REFERENCE
[1] https://en.wikipedia.org/wiki/Zambia
424
Q.104 - R.28
-------
-------
APPLICATION OF ROTEC BELT CONVEYOR EQUIPMENT AT KAFUE GORGE

LOWER PROJECT (∗)
CHINA
SUMMARY
Rotec belt conveyor equipment has the following characteristics: simple and
convenient operation, good adaptability to terrain, continuous production, high effi-
ciency, low cost, and can be used all year round. The speed of the belt conveyor
is fast, which can effectively reduce the temperature rise of the concrete and accel-
erate the paving speed of the concrete inside the dam. Crawler placer machine is
applied in the Kafue gorge lower project. The belt speed is 3.6 m/s, the maximum
concrete conveying capacity is 300m3 /h, the belt truss has a total length of 1559m,
the maximum length of single section is 582m, and there are 8 head sections. The
longest span of the on conveyor the bank slope is 57m, and the steepest slope is
27◦ (the maximum angle between the conveyor and the horizontal plane), the height
difference of the bank conveyor is up to 100m. During the operation of the belt con-
veyor, the belt different width of Rotec belt conveyor and the batching plant resulting
in too slow feed rate is solved by adjusting the feeding speed of the batching plant.
By increasing the generator, the problem that the long belt of the climbing slope con-
veyor cannot be restarted due to the system electrical quality problem is solved. At
the same time, the Rotec conveyor belt system layout has been improved to ensure
the normal pouring of the dam by means of dump truck transportation when a large
failure occurs before T4.
RÉSUMÉ
Le convoyeur à bande Rotec présente les caractéristiques suivantes : fonc-

tionnement simple et pratique, bonne au terrain, production continue, rendement
∗ Utilisation du convoyeur à bande Rotec pour le projet de Lower Kafue Gorge
DOI: 10.1201/9781003211808-27 425

Q.104 - R.28
élevé, faible coût et peut être utilisé toute l’année. La vitesse du convoyeur à bande
est rapide, ce qui peut réduire efficacement l’élévation de température du béton et
accélérer la vitesse de pose du béton à l’intérieur du barrage. La machine à placer
sur chenilles est appliquée dans le projet inférieur de la gorge de Kafue. La vitesse
de la bande est de 3,6 m / s, la capacité de transport de béton maximale est de
300 m3 /h, la poutre a une longueur totale de 1559 m, la longueur maximale d’une
seule section est de 582 m et il y a 8 sections de tête. La plus longue portée du
convoyeur sur la pente de la berge est de 57 m, et la pente la plus raide est de
27◦ (l’angle maximum entre le convoyeur et le plan horizontal), la différence de
hauteur du convoyeur de banque est jusqu’à 100 m. Pendant le fonctionnement du
convoyeur à bande, la largeur différente de la bande du convoyeur à bande Rotec
et de l’installation de dosage entraînant une vitesse d’alimentation trop lente est
résolue en ajustant la vitesse d’alimentation de l’installation de dosage. En aug-
mentant le générateur, le problème selon lequel la longue bande du convoyeur à
pente ascendante ne peut pas être redémarré en raison du problème de qualité
électrique du système est résolu. Dans le même temps, la disposition du système
de bande transporteuse Rotec a été améliorée pour assurer le déversement nor-
mal du barrage au moyen du transport par camion à benne basculante lorsqu’une
défaillance importante se produit avant T4.
1. PROJECT INTRODUCTION
Kafue Gorge Lower Hydroelectric Power Project (KGL Project) is located on

the Kafue River. The catchment area upstream the Dam Site measures 153,000Km2 .
The annual average inflow is 255m3 /s. The main purpose is for power genera-
tion. At FSL of 579.00m, the reservoir capacity is 83×106 m3 , and the active
storage is 61×106 m3 . The Dam foundation level is EL.450.5m. The main Dam
is RCC gravity-type, with a maximum height of 130.5m, Dam crest 8m in width
and 374.5m in length. The Spillway is equipped with three (3) radial gates with
overall flow width of 62m and located at approximately the mid-point of the Dam
length with weir crest at EL.561.00m. The Spillway maximum discharge capacity
is 7228m3 /s with ski-jump (Flip-Bucket) and downstream Plunge Pool for energy
dissipation. The Low Level (environmental compensation flow) Outlet structure with
inlet invert at El 523.0m, which is arranged on the right side of the Spillway, is
designed to ensure a downstream discharge from the Dam of not less than 11m3 /s
of environmental compensation flow under Reservoir Minimum Operating Level of
EL.530.00m.
The solution of the belt conveyor is actually a combination of the high-speed

transportation of the belt conveyor and the flexible turning of the crawler placer. The
belt conveyor is used to transport the concrete directly from the batch plant to the dam
surface to achieve high-intensity and rapid Pouring RCC. The conveyor belt machine
was first proposed and manufactured by the ROTEC company in the United States.
426
Q.104 - R.28
It was successfully applied in Wheatstone Dam, setting a world record of 72,000

cubic meters per month. The belt conveyor generally adopts the “one Batch Plant,
one machine, and one belt” method, that is, a dedicated batch plant to ensure mixing
and feeding; a belt conveyor supply line to ensure rapid concrete transportation; and
a dedicated crawler placer to ensure flexible distribution. The crawler placer places
the RCC with a telescopic arm. The up and down angle of the boom of the spreader
is generally + 20◦ ∼-15◦ , Feeding height is generally 5–8 meters. When the pouring
height is high, a special elephant nose tube is used at the hopper of the telescopic
arm to prevent concrete aggregate from separating. Both the belt conveyor and
crawler placer are equipped with climbing and jacking mechanisms in the dam for
automatic climbing after the dam is poured to a certain height.
The belt conveyor of this project is installed in sections: On October 15, 2018,
the first stage of the installation of the Rotec EL581 platform to EL494 was com-
pleted. lifting of Rotec construction is difficult, the truss span is large (the longest
span is 57m), the slope is steep (the angle between the truss and the horizontal
plane is 27◦ ), and the crane adjustment position is small. The tail of the next section
of the truss needs to pass through the nose of the previous section of the truss,
like a needle lead on a cliff. The second stage installation from the batch plant
to the EL581 platform was completed in January 2019. This installation scheme
not only guarantees the construction period of the dam, but also creates a dis-
charge point at the EL581 platform, so that when the belt conveyor at the batch
plant to T4 fails, Rotec belt conveyors can be repaired without affecting the dam
pouring.
The entire belt conveyor system is mainly composed of truss, rollers, driving
drums, motors, belt, crawler placer and electrical control system. Its mechanical
structure and electrical control system are relatively simple. When the equipment
fails, the fault point is easy to find, which is convenient for maintenance. The
RCC dam of this power station shall be poured by Rotec belt conveyor to trans-
port concrete into the dam. See Table 1 for belt conveyor parameters. The putting
into operation of the belt conveyor can replace the traditional way of entering
the warehouse of a car and speed up the speed of entering the warehouse of
concrete.
Table 1
Rotec system parameter table
TOTAL MAXIMUM THE MAXIMUM SPEED OF MAXIMUM
LENGTH
OF BELT SINGLE BELT MAXIMUM TRUSS BELT CONCRETE
CONVEYOR LENGTH MOTOR SLOPE CONVEYOR CONVEYING
(M) (M) POWER (DEGREE) (M/S) CAPACITY
(HP) (M3 /H)
1559 582 250 27 3.6 500
427
Q.104 - R.28
2. THE RUNNING MODE OF THE BELT CONVEYOR
The belt conveyor arranges the trusses according to the terrain of the site.
When the truss is on a relatively flat ground, arrange longer steel trusses and
high-power motors. When the truss is in a section with a large slope, a lightweight
aluminum alloy truss and a low-power motor are used. Material transfer between
trusses via hopper. The head is provided with a turning device, which can rotate the
trusses to a certain angle and change the angle between the trusses. Each head of
the belt conveyor requires at least two workers on duty. On the one hand, during the
operation of the equipment, failures can be discovered in time and the watchkeeper
can be notified in time. The supervisor decides whether and when to arrange main-
tenance. On the other hand, there is a control cabinet at each head. When a certain
section of the belt conveyor fails, the operator can press the emergency stop button
according to instructions. After pressing the emergency stop button, only the belt on
this section of truss and its previous belt will stop running, and the belt behind it will
run normally to prevent the concrete from staying on the belt.
The limiting factors for the quality of the material on the belt are the angle of
the rollers, the width of the belt, and the amount of material discharged from the
batch plant. How to balance the relationship between these three to make the belt
conveyor run normally and stably is the first problem to be overcome in this project.
At the same time, some faults often occur during the operation of the belt conveyor,
How to troubleshoot in a timely manner is also a key issue of the project. Such as: the
belt runs off, material spills and the subsequent belt damage; A virtual connection
or break of the control line causes a sudden stop of a certain section of the belt
conveyor or the entire system; Unstable system voltage and uneven distribution of
materials on the belt cause the motor current to increase, causing the control cabinet
to overcurrent protect and stopping the system; During the running process, before
the belt conveyor that can be automatically started with the concrete on the belt.
Later, the belt conveyor cannot start automatically with the concrete on the belt. The
downtime of the belt conveyor is limited by the RCC concrete hot joint and cold joint
formation time. Explained one by one below.
3. THE BELT CONVEYOR AND BATCH PLANT COOPERATE
The belt conveyor of this project is supplied by two batch plant. The gap
between the openings of the two batch plants is 10m. When the two batch plants
unload at the same time, it will cause the concrete on the belt to overlap, overload
the belt, and cause problems of leakage and spreading. Therefore, the project uses
two batch plants to feed the belt alternately. According to the speed of the belt and
the distance between the two mixing buildings, adjust the opening of the arc door of
the feeding opening of the two batch plants to adjust the feeding speed and feed-
ing time. By adjusting the feeding scheme, the belt transport capacity can reach
300 m3 /h.
428
Q.104 - R.28
Fig. 1
Feeding layout of batch plant.
Fig. 2
Layout of EL581 discharge platform.
4. IMPROVEMENT OF BELT CONVEYOR SYSTEM LAYOUT
Because the entire belt conveyor system operates on a single line, when a fault
occurs, the entire line cannot work properly. If the maintenance time is too long, it
will seriously affect the dam pouring quality. In response to this problem, the project
and the manufacturer negotiated a plan to set two 8-cubic blanking openings and a
section of transmission truss at the EL581 platform (see Fig .2), and the belt truss
was overlapped to the T4-T5 section of the ROTEC truss.
When the T4 section and the belt conveyor before it fails, you only need to
adjust the cabinet wiring to achieve the belt conveyor system segmented operation.
429
Q.104 - R.28
Fig. 3
Relation between pouring volume and elevation.
When repairing the front T4 belt, it will not affect the pouring of the dam, thus ensuring
the quality of the pouring of the dam.
5. ELECTRICITY LOAD
The total quantity of RCC is 1,290,000m3 , average of construction area of

RCC placement is around 10,000 and the area is around 4,000 at bottom of dam
while it is around 2,500 at top part. The maximum placement area reaches the
maximum in the middle of the dam body. The area is more than 10,000 from EL460
to EL535m and between 13,000 to 14,000 from EL470 to EL510m. Below Chart show
the relationship between RCC quantity of one layer (30cm thick) and the elevations.
Fig. 4
Layout of trusses in the dam.
430
Q.104 - R.28
Fig. 5
Hopper between trusses.
It can be seen from the Figure 3 that when the elevation of the dam is at
EL500, the pouring strength reaches the maximum, and the requirements for the
stable operation of the belt conveyor are also increased. At this time, it is another
peak period for the construction of the entire project, and the load of on-site electricity
consumption is also increasing. The capacity of the distribution transformer in the
field has been calculated which meet the construction requirements, however, the
belt conveyor has a problem that the long belt that climbs the slope cannot be started
with concrete on the belt due to the electrical quality of the system. By analyzing
the increase of the electric load in the whole field and observing the relationship
between the motor current and voltage when the belt conveyor starts with load, it
is found that when the motor starts, the voltage of the motor changes from 430V to
320V, and the voltage drops severely. In general, the voltage drop when the motor
starts should not be lower than 85% of the rated voltage. When the motor starts to
run, it is similar to a blocked rotor. There is no induced magnetic field in the rotor
that obstructs the alternating magnetic field of the stator, which results in a very
large stator current and a certain voltage drop on the line. The rotating power of the
motor requires insufficient reactive power to establish a magnetic field. The reactive
power is too much lost on the line. Both the transformer and the motor are devices
that generate and consume reactive power. During the transfer process, a part of
the reactive power is consumed in the form of heat, and the lost reactive power is
usually compensated at local. According to the above theory, it is necessary to add
a reactive power compensation cabinet to perform reactive power compensation
for the entire system, but due to the long procurement cycle, an initial increase in
generators is used to perform power compensation.
The reactive power loss is calculated as follows:
431
Q.104 - R.28
Reactive power loss of the motor:
U2
QM = Qm + Qσ = + I 2 Xσ
Xm
Note: Qm : the excitation power,
Qσ :the leakage reactance loss,
Xm : the excitation point
Xσ : the leakage reactance.

2
Io % Us %STN STL
QLT = Qo + QT = STN +
100 100 STN
Note: Qo excitation branch loss,
QT winding leakage reactance loss,
STN is the rated capacity of the transformer MV · A,
STL is the rated capacity of the transformer MV · A,
Io% no-load current is about 1% -2%, The loss in leakage reactance is

approximately 10% at full load.
Through calculation, the reactive power loss of the electric equipment in the
field was obtained, and the corresponding equipment was equipped to compensate,
and good results were obtained.
6. CORRECT THE BELT DEVIATION
Because the Rotec equipment used in this hydropower station is relatively

old, the truss strength is lower than the design value. At the same time, the truss
section in the dam is gradually skewed because the support columns cannot be
lifted vertically during jacking. During the operation of the truss transformation area,
the uneven feeding of the funnel will cause the belt to deviate and scrape the truss.
Each truss head hopper is a device for guiding and controlling the direction of the
material flow, and can also be used for dust prevention. At a certain elevation, the
angle between the trusses in the dam changes with the lifting of the dam body, and
the angle at which the material falls on the belt changes with the angle. For example,
432
Q.104 - R.28
below EL510, the truss in the T7-T10 dam rises with the lifting of the dam surface,
and its front T6 at this elevation does not need to be jacked up. The belt rotates at a
constant speed, and the parabola formed by the concrete material will hit one side
of the belt, resulting in uneven force on the belt, the belt deflects, and the material
spills. The first step of the solution is to change the refraction angle of the material at
the head so that it can land on the center of the next belt. Then a guide roller is added
to control belt on the long truss. The Belt deviation caused by uneven distribution of
materials can be controlled better.
For the problem that the falling point of the material cannot be in the center of
the belt, the method of modifying the structure inside the funnel at the corresponding
head is used to deal with it. A circular track is welded in the hopper, and a movable
tongue is arranged at the track to allow the material to impact the tongue to change
the feeding point to improve the belt deviation caused by the incorrect feeding point.
For the problem of incorrect belt itself, you need to adjust the belt. When
adjusting, when the amount of adjustment is small, the adjustment bolts at the head
or the tail need to be adjusted or some of the supporting rollers can be adjusted to
change the running direction of the belt. When the adjustment amount is large, a
special hydraulic jack can be used to adjust the belt by adjusting the drum at the tail
of the truss. Preventing belt misalignment is the most effective way to increase belt
life.
7. CONCLUSION
The application of the Rotec belt conveyor in the Kafue Gorge Lower project
was relatively successful, which ensured the pouring quality and construction period
requirements of the dam. Reasonable maintenance of the entire system is a prereq-
uisite for the stable operation of the entire system. In the process of operation and
maintenance, some experiences were summarized, which provided a rich practical
basis for the application of similar belt conveyor systems for dams in the future.
433
Q.104 - R.29
-------
-------
APPLICATION OF RCC TECHNICAL INNOVATION AT KAFUE GORGE LOWER

HYDRPOWER PROJECT (∗)
Yuanguang LIU, Qingguo ZHOU, Yuanbo GAO & Junjie JIN

SINOHYDRO BUREAU NO. 11 CO. LTD., Zheng Zhou City, Henan Province, China
CHINA
SUMMARY
For the RCC Dam of Kafue Gorge Lower Hydropower Station in Zambia,
full cross-section RCC with maximum aggregate diameter of 63mm is used. The
upstream impermeable area to be placed with RCC with maximum aggregate diam-
eter of 37.5mm is canceled, Rotec belt-conveyance system introduced, and capacity
of mixing and batching plant secured, to achieve rapid construction of the RCC dam.
With cement grout placed in interlayers of RCC, bonding strength, shear strength
and impermeability between layers could meet the design requirements. Imper-
meable GEVR is used to improve interlayer bonding strength and the reliability of
impermeableness. The fly-ash, stone powder and high efficient Naphthalene water
reducing agent are mixed with RCC so as to decrease unit consumption of cement,
and reduce the maximum temperature rise of hydration heat of concrete, which helps
with temperature control and crack prevention of concrete. In addition, the cooling
water pipes are not used, and the construction is more economical accordingly.
RÉSUMÉ
Pour le barrage en BCR de la station hydroélectrique de Lower Kafue Gorge

en Zambie, on utilise un BCR à granulométrie complète avec un diamètre maximal
d’agrégats de 63 mm. La zone imperméable amont à placer avec un BCR avec un
∗ Mise en œuvre d’une technique innovante de BCR pour le projet hydroélectrique de Lower
Kafue Gorge
434 DOI: 10.1201/9781003211808-28

Q.104 - R.29
diamètre d’agrégat maximal de 37,5 mm est supprimée, le système de convoyage

à bande Rotec est utilisé et la capacité de la centrale de mélange et de dosage est
sécurisée, afin de réaliser rapidement la construction du barrage. Avec un coulis
de ciment placé dans des couches intermédiaires de BCR, la force de liaison, la
résistance au cisaillement et l’imperméabilité entre les couches pourraient répon-
dre aux exigences de conception. Un BCR enrichi et imperméable est utilisé pour
améliorer la force de liaison inter-couches et la fiabilité de l’imperméabilité. Les cen-
dres volantes, la poudre de pierre et l’agent réducteur d’eau naphtalène à haute
efficacité sont mélangés avec le BCR afin de réduire le dosage en ciment et de
réduire l’élévation de température maximale de la chaleur d’hydratation du béton,
ce qui contribue au contrôle de la température et à la prévention des fissures du
béton. De plus, les conduites d’eau de refroidissement ne sont pas utilisées et la
construction est en conséquence plus économique.
1. PROJECT INTRODUCTION
Power Construction Corporation of China (hereinafter referred to as Pow-

erChina) started the geological exploration of the Project in 2010, and completed it
three years later. In 2012, the feasibility study report of the project was finalized and
approved by the Government of Zambia. On October 19, 2015 witnessed signing of
the EPC+F contract in between Zambia Electricity Supply Corporation Limited and
PowerChina in the name of Sinohydro Corporation Limited. Sinohydro Bureau 11
Co., Ltd. and Northwest Engineering Co., Ltd., were authroized by PowerChina to
form a consortium tojointly implement the contract whose amount totals $1.567 bil-
lion, out of which 15% funded by Zambian Government, and the rest of 85% funded
by Industrial and Commercial Bank of China (ICBC) and the Export-Import Bank of
China (EIBC)with guarantee by Ministry of Finance of Zambia. Norconsult serves
as the consulting company. The Project comprises of RCC Gravity Dam, Full-cross
Lined Headrace Tunnel, Powerhouse and Township.
Standing as a RCC Gravity Dam, the Dam has a design flood with recurrence
interval of 10,000 years. The normal water level is 579m and the total storage capac-
ity is 83 million m3. The Dam has a crest height of 581.00m, a maximum height of
139m, a crest length of 374.5m and a crest width of 8m. It is divided into 19 sections
which are as follows in sequence from the left bank to the right bank: the left bank
non-overflow section, the riverbed non-overflow section, the overflow section, and
the right bank non-overflow section. The overflow section is in the middle of the
riverbed, and the spillway has three radial gates with a size of 15 m × 19.9 m (width
× height). The upstream slope ratio of overflow weir is 1:0.2, and overflow weir is
a practical weir of WES type. The elevation of the Dam crest is El.561.00m. The
spillway has a total width of 62m (including the gate pier and overflow weir), with the
middle pier 5m wide and the side pier 3.5m wide. The joint is located at the center of
the overflow weir surface, and side walls are provided on both sides of the spillway.
435
Q.104 - R.29
2. SCIENTIFIC AND TECHNICAL RESEARCH BACKGROUND
KGL project adopted one type RCC mix with 12MPa @ 365d in full section
of the dam, so that simplifying the placement procedures and improving the site
placement rate. The technical requirement of RCC mix is shown in Table 2. The
concrete materials and final selection of RCC mix and its properties is introducing
as bellows.
2.1. KEY TECHNOLOGY IN RAPID AND CONTINOUS CONSTRUCTION
Site of the Dam is around 1.2km downstream the confluence of Kesya River
and Kafue River. The valley where the Dam site is located takes a shape of V, with
the mountain on left bank steep, and the excavation line high; the rock joints on right
bank develop along the slope, and are likely to collapse, imposing adverse effects
to excavation, and brings a tight schedule for the Dam construction; the rain season
lasts for three months with short-range torrential rain, making it essential to secure
continuous construction in rain season.
2.2. KEY TECHNOLOGY IN FULL-SECTION CONCRETE WITH MAXIMUM AGGREGATE

DIAMETER OF 63MM
Different with that of the concrete placed in the Dam, aggregate diameter of the
concrete placed in the upstream impermeable area increases difficulty in operation
of batching plant, and construction in the block, as well as entails section division
marking, etc., making it difficult to construct rapidly and efficiently.
2.3. TECHNICAL RESERCH IN PLACING INTERLAYERS CEMENT GROUT
Located in a continental climate, the Project has large temperature difference

between day and night, large temperature difference between different seasons,
and strong ultraviolet rays. Therefore, it is not appropriate to use the same setting
time to determine the status of the interlayer joint bounding in different external envi-
ronments. The use of cement mortar is greatly affected by the external temperature
while the use of tank trucks for transport of concrete brings cross-operation, and
reduces efficiency of construction, so it is difficult to achieve the purpose of rapid
construction.
436
Q.104 - R.29
2.4. TECHNOLOGY FOR GEVR
If cement grout is placed at the middle or on top of GEVR, it is difficult to

notice if it has been vibrated to the required extent, or if there is any missing vibra-
tion. If cement grout is placed at the bottom, however, it could guarantee quality of
construction, and security and durability of the grouted concrete placed around the
formworks upstream and downstream the Dam, and in other locations.
2.5. TECHNICAL RESEARCH IN MIXING FLY-ASH, STONE POWDER AND HIGH

EFFICIENT NAPHTHALENE WATER REDUCING AGENT
The external environment of the Dam is dry at high temperature, and control
of the adiabatic temperature rise of concrete is essential to construction quality and
shall meet the continuous and rapid rise of the Dam, and simplify the construction
procedure in the block as much as possible.
3. STUDY ON CHARACTERISTICS OF JOINT WITH DIFFERENT

MATURITY AND TREATMENT
3.1. CHARACTERISTICS OF THE DAM
(1) Criteria for joint Use of such innovated technologies as full-section concrete
with maximum aggregate diameter of 63mm, long-aged RCC, GEVR, criteria of
assessment and ways of treating, admixtures of various types, Rotec Conveyer belt
system, etc., has simplified the procedure of production, transport and placing of
RCC, improved placing speed, guaranteed quality and safety of construction, and
achieved the goal of rapid and continuous construction of the Dam.
(2) Consolidation grouting for the foundation of RCC Dam is carried out ran-
domly depending on class of bedrock so as to substantially reduce the quantity of
consolidation grouting.
(3) 'Shotcrete is placed at the slopes on banks of Plunge Pool.
3.2. TECHNICAL INNOVATION
3.2.1. Design innovation
(1) The impermeable area to be placed with the concrete with the maximum
aggregate diameter of 37.5mm is canceled. Permeability is accomplished by placing
437
Q.104 - R.29
the full-section RCC with the maximum aggregate diameter of 63mm. Use of the
RCC of the same maximum aggregate diameter greatly simplifies procedure of
construction and improve the efficiency.
(2) Depending on class of the rock exposed, the Dam foundation is optimized,
and the foundation in the riverbed rises by 8.5m while the foundation on both banks
rises to different extent as well. The above-mentioned optimization reduces the
maximum Dam height by 8.5m, i.e. from the design height of 139 m to 130.5m,
which saves 173,000 m3 of concrete, i.e. from the design quantity 0f 1,430,000 m3
to 1,257,000 m3 .
3.2.2. Construction technology innovation
(1) The crushing and batching palnt has a designed processing capacity of
950t/h and a production capacity of 800t/h. The production capacity of RCC is
640 m3 /h, and that of pre-cooled RCC is 370 m3 /h. The cooling system can meets
the minimum outlet temperature of RCC at 21.5 and Temperature in October is
high, with the average temperature of 28.9 and the average water temperature of
28. Cold water, primary air cooling, ice and other pre-cooling measures are adopted.
The designed cooling capacity (standard working condition) is 7,851KW.
(2) Rotec conveyer belt system is introduced to transport and place RCC in the
block, and ensures rapid rise and continuous construction of the Dam. The design
speed of the conveyor belt is 3.6 m/s, the maximum concrete conveying capacity is
500 m3 /h, the total length of the belt truss is 1,559 m, the maximum length of a single
section is 582m, and a total of 8 heads are set up to realize the rapid continuous
rise of the Dam.
(3) RCC with the maximum aggregate diameter of 63mm is used. By gather-
ing seepage control structural design and mix proportion design of the completed
RCC Dam both at home and abroad, and applying comparative analysis theory, the
upstream impermeable area to be placed with concrete with the maximum aggre-
gate diameter of 37.5mm is canceled and the same concrete with the maximum
aggregate diameter of 63mm is used in the seepage control structure for the full
section so as to achieve the goal of rapid construction of RCC Dam.
(4) Cement grout is placed in between the layers of the Dam. Cement grout
with a ratio between water and cement at 0.55-0.65 is placed in interlayers and cold
joints so that bonding strength, sheer strength and impermeability in between the
layers could meet the design requirement.
(5) Grouting and vibrating is done at the bottom of GEVR. Experimental study in
grouting is conducted through various ways such as surface grouting, hole grouting,
and bottom grouting, etc.By comparing The differences of mechanical properties and
anti-seepage properties of RCC with different grouting methods, the optimal grouting
method is obtained. The cement grout is placed at the bottom of concrete, which
438
Q.104 - R.29
ensures the uniformity of concrete and improves the inter-layer bonding strength
and inter-layer anti-seepage reliability.
(6) Use of fly-ash, stone powder and high efficient Naphthalene water reducing
agent reduces unit consumption of cement, decrease the maximum temperature
rise of hydration heat of concrete, which helps with temperature control and crack
prevention of concrete. In addition, the cooling water pipes are not placed, and the
construction is more economical accordingly.
4. TECHNICAL INNOVATION APPLICATION PROSPECT
RCC Dam-building technology has been widely popularized and applied in the
world due to its rapid construction speed, high mechanization degree, low investment
cost, and so on. It is one of the most competitive dam types in the dam industry today.
Based on the RCC dams PowerChina has completed or is constructing abroad

in recent years, the paper studies use of such innovated techniques as full-section
concrete with maximum aggregate diameter of 63mm, long-aged RCC, GEVR, crite-
ria of assessment and ways of treating, admixtures of various types, Rotec Conveyer
belt system, etc., and systematically refines and summarizes the new idea, new
technology, and new craft adopted in RCC dam construction all over the world. In
this way, RCC dam construction technology can be promoted to the direction of opti-
mized design, better quality, higher efficiency, and lower cost, providing important
reference for the design and construction of RCC dam in the future. In addition,
the application of these innovative technologies in RCC dam construction of Kafue
Gorge Lower Hydropower Station in Zambia achieves good economic and social
benefits and has a good prospect of promotion and application.
5. OPERATION OF THE PROJECT
Kafue Gorge Lower Hydropower Station is the largest single infrastructure

project in Zambia up to now. The construction of the Project will meet the domestic
electricity demand of Zambia in the next 5 to 10 years and guarantee stable electricity
supply for the development of mining industry and agriculture in Zambia, which is of
great strategic significance to the economic development of Zambia.
The construction of the Project will help create more employment opportuni-
ties, increase Zambia’s existing power supply capacity by 38% after generation, and
play a positive role in improving the power security of the southern African power
grid. Currently, the Project has provided more than 11,000 jobs.
439
Q.104 - R.30
-------
-------
SUCCESSFUL APPLICATION OF RCC INTERLAYER JOINT BONDING

TECHNOLOGY WITH SURFACE LONG TIME EXPOSED IN DRY AND HIGH
TEMPERATURE AREAS (∗)
Junjie JIN, Qingguo ZHOU, Shuncai NING & Yuanguang LIU
CHINA
SUMMARY
This rapport is depended on the Kafue Gorge Lower Hydropower Project in

Zambia. For the RCC dam construction, the Contractor conducted a comprehensive
study on the influences of interlayer joints bonding quality from RCC mix perfor-
mances. The contents is including: 1) The joint type judged by Modified Maturity
Factor (MMF) with joint surface long time exposed in the region with dry and high
temperature; 2) Technical measures of layer bonding quality control under condition
of joint surface long time exposed; 3) Effect of joints shear strength and imperme-
ability if the RCC Layer compacted existed in the condition of plastic and elasticity; 4)
Joints construction craftsmanship. The achievements and experiences of the Project
could be used for.
RÉSUMÉ
Ce rapport dépend du projet hydroélectrique de Lower Kafue Gorge en Zam-

bie. Pour la construction du barrage en BCR, l’entreprise a mené une étude
approfondie sur les influences de la qualité de collage des joints intercalaires à partir
des performances du mélange en BCR. Le rapport expose 1) le type de joint jugé par
le facteur de maturité modifié (MMF) avec une surface de joint longtemps exposée
dans la région à température sèche et élevée ; 2) les mesures techniques de con-
trôle de qualité de liaison de couche sous condition de surface de joint longtemps
∗ Application réussie de la technologie de liaison intercouche de BCR avec surface longue
durée exposée dans les zones sèches et à haute température
440 DOI: 10.1201/9781003211808-29

Q.104 - R.30
exposée ; 3) l’effet de la résistance au cisaillement et de l’imperméabilité des joints si

la couche en BCR compactée existait à l’état plastique ou élastique ; 4) la méthode
de construction de joints. Les réalisations et expériences du projet pourraient être
utilisées comme référence pour la construction de barrages similaires en BCR.
1. INTRODUCTION
Roller Compacted Concrete (RCC) dam construction technology has been

widely popularized and applied in the around world with advantage of its low cost
and short construction period. Since the first RCC dam in the world was built in Japan
in 1981, up to now there are nearly 1070 RCC dams (incomplete statistics) have
been built in the world. By the engineering practice and technological Improvement
in the past 40 years, the technology of RCC dam construction has made matures
gradually. However, there are still many problems to be further understood and
studied on RCC dam construction technology. This paper has carried out some
study on the characteristics of RCC interlayer joint and its quality control method in
the region with dry and high temperature climate.
Project description: The Zambia - Kafue Gorge Lower (750 MW) Hydroelectric
Project (herein referred to as the KGL Project) is located on the Kafue River, a
primary tributary of the Zambezi River, some 55km upstream from the confluence of
the Kafue River and the Zambezi River. KGL is located some 17.3km downstream
from the the Kafue Gorge Upper Hydropower Project (herein referred to as KGU)
and some 5.9km downstream from the KGU tail-water Outlet.
The catchment area upstream the Dam Site measures 153,000km2 . The
annual average inflow is 268.8m3 /s, the main purpose is for power generation. At
FSL of 579.00m, the reservoir capacity is 83 × 106 m3 , and the active storage is
61 × 106 m3 .
The main Dam is of RCC gravity-type, with a maximum height of 130.5m

tentatively, 374.5m in length. Dam crest is 8m wide and footprint width with120m.
The upstream slope of the dam is designed by 1:0.2 (V : H) and change to vertical
at EL530m, while the downstream slope is designed by 1:0.75 (V : H) and change
to vertical at EL569.5m. The dam is divided into 19 monoliths in which 7 monoliths
of non-overflow section at left bank and 4 monoliths of over-flow section(spillway)
and 8 monoliths of non-overflow section at right bank. Typical dam up-stream view
sees the Fig. 1.
In general, the Zambia weather is subtropical continental. Rainy season starts

in November and end in March, occasionally it rains in September and October. In
Zambia, there are four seasons: the weather is cold and dry in winter from April to
August with lowest temperature 7.5 ◦ ; the temperature rise and the weather becomes
hot and dry from September/ October which the temperature reaches at circa 40◦ C
according to site monitoring; Raining usually starts in November, it is humid and
stuffier at this time; the temperature start dropping from February to March, it is
humid and cool along occasionally raining. Meteorology data see Table 1.
441
Q.104 - R.30
Fig. 1
View at U/S of Dam.
Table 1
Meteorology Data of Kafue Gorge Lower Area
MONTH 1 2 3 4 5 6 7 8 9 10 11 12 YEAR AVER.
Average Rainfall 205 153 68 26 2 1 19 81 210 766
Average Temperature 26.4 26.1 25.7 24.5 22.1 20.1 20 22.5 26.1 28.5 28 26.8 24.7
Average Evaporation 149 133 160 166 163 144 158 195 232 259 202 159 2120
Ralative Humidity 78 80 77 68 63 62 57 50 43 44 56 73 62.6
Average wind velocity 1.6 1.6 1.8 2 2.1 2.2 2.5 2.7 3 2.9 2.5 2.1 2.2
Table 2
Performance requirements of RCC mix
COMPRESSIVE TENSILE AGE IMPERMEABILITY M.S.A
STRENGTH STRENGTH(MPA) (DAY) COEFFICIENT (MM)
(CYLINDER), MPA (CM/S)
INDIRECT DIRECT
12 ≥ 1.99 ≥ 1.14 365 ≤ 10−8 63
Raise of problems: As well known, RCC dam seepage is often occurred from
the interlayer joints, it is very important issue how to improve the quality of joint’s
bonding. During KGL dam RCC construction peak period, the maximum placement
block area is at circa 14000 m2 from the EL470 to EL.510. The single layer (30cm
thick) RCC maximum volume reaches at 4250m3 (see Fig. 2). Based on the compre-
hensive capacity of the batching plant production, transportation and site placement
progress etc., the interlayer joint surface maximum exposure time needs at least 16
hours which does not consider the condition such as the malfunction of batching
plant or other equipment. The joint exposure time could need within 20-24hours that
is considered all of emergency events situations.
In order to ensure the quality of interlayer joint bonding, the project technician
carried out some exploration and researches as follows: a) Optimizing of RCC mix
proportion. b) Study of characteristic of joints under different exposure time and
treatment. c) Establishing the control standard of Maturity degree at the placing site.
d) Site quality control system.
Based on the final evaluation of the joint quality by the core samples extraction
and water pressure test, the results meet the technical requirement and the quality
is satisfied. The main results and achievements applied KGL Project are introduced
as below.
442
Q.104 - R.30
Fig. 2
Single layer(30cm), RCC volume, m3 .
2. OPTIMIZING OF RCC MIX
KGL project adopted one type RCC mix with 12MPa @ 365d in full section
of the dam, so that simplifying the placement procedures and improving the site
placement rate. The technical requirement of RCC mix is shown in Table 2. The
concrete materials and final selection of RCC mix and its properties is introducing
as bellows.
2.1. PROPERTIES DO RAW MATERIALS
2.1.1. Cement
The Lafarge Cement in Zambia was chosen as the main supplier. The Lafarge
Cement Plant is relatively close to the site, and its output can reach 4000t/day. It can
be supplied steadily during the peak period of dam RCC construction. Type CEM-I
42.5N cement is selected for RCC mix design. The cement complies with BSEN
197-1. The test results are shown in Table 3.
443
Q.104 - R.30
Table 3
Cement test results
SETTING COMPRESSIVE
ITEM DENSITY SOUNDNESS FINENESS
TIME (MIN)
STRENGTH,
(G/CM3 ) (MM) (%)
(MPA)
initial final 2d 28d
Results 3.11 1.0 1.5 163 216 28.4 54.9
BS EN / ≤ 10 ≤ 10 ≥ 60 / ≥ 10 42.5≤f≤62.5
197-1
2.1.2. Fly ash
After the investigation, Mammba thermal power plant in Zambia is with

5×150 Mw units. The fly ash is collected by electrostatic collection system, and
with production of 2000 t/day fly ash qualified. It meets the maximum demand of
dam RCC construction peak period with 500 tons per day. The fly ash meets the
requirements of ASTM C618 class F. The test results are shown in Table 4.
Table 4
Fly ash test results
ITEM DENSITY SOUNDNESS LOSS ON FINENESS WATER ACTIVITY
(G/CM3 ) (MM) IGNITION (%) (0.045mm SIEVE), (%) REQUIREMENT (%) INDEX(%)
7d 28d 90d
Results 2.34 0.5 3.79 17.7 104 85.3 93.7 113
ASTM C618 ≥ 2.0 0.8 F,C: ≤6 N: ≤10 N, F, C: ≤34 F,C: ≤105 N: ≤115 ≥ 75 ≥ 75 /
2.1.3. Aggregates
The aggregate is made of artificial aggregate produced by on-site crushing

plant. Previous exploration confirms that there is no potential alkali aggregate reac-
tion in quarry rocks, and quality meet the technical requirements of ASTM C33,
EM1110-2-2006 and ACI 207.5R-2011. The test results of artificial sand are shown
in Table 5.
Table 5
Test results of artificial sand
ITEM DENSITY ABSORPTION FINENESS SOUNDNESS LIGHT ORGANIC CLAY < 0.075
(G/CM3 ) (%) MODULUS (%) MATERIALS (%) MATTER (%) MM
Results 2.61 2.2 2.73 5.8 0.1 0 0.3 12.8
EM1110-2-2006 >2.5 / 2.1-31 <12 <0.5 No allowed <3.0% /
The coarse aggregate is divided into three groups: 4.75-19 mm, 19-37.5 mm
and 37.5-63 mm. The technical properties of the coarse aggregate are in accordance
with ASTM C33 and EM1110-2-2006. The test results are shown in Table 6.
444
Q.104 - R.30
Table 6
Coarse aggregates test results
ITEM DENSITY ABSORPTION, LOS AGLES Soundness FLAT & CLAY LIGHT <75 µM
(G/CM3 ) % ABRASION, ELONGATION LUMP MATTER
4.75-19mm 2.62 0.8% 25%% 5.0% 3.7% 0.30% 0.08% 0.8%
19-37.5mm. 2.63 0.6% 27% 5.8% 2.5% 0.25% 0.08% 0.6%
37.5-63mm 2.64 0.3% 28% 7.0% 8.8% 0.20% 0.06% 0.3%
ASTM C33 >2.5 / <50% <12% <30% <3% <0.5% <1%
2.1.4. Admixtures
The admixtures used for RCC mix are type SN-2 super plasticizer and SN-GH
retarder produced by the Additive Factory of Bureau No.11 of PowerChina, China.
The admixtures performances meet the technical requirements of ASTM C494. The
test results are shown in Table 7.
Table 7
Admixture test results
ITEM DOSAGE % SETTING TIME (H: MIN.) RATE OF WATER RATIO OF STRENGTH, %
INITIAL FINAL REDUCTION % 3D 7D 28D
Reference concrete / 8:11 10:15 / 100 100 100
Reducer type SN-2 0.8 ≥ 1 : 30 ≥ 3 : 15 15.9 130 122 123
Test results 1.0 ≥ 5 : 15 ≥ 6 : 30 20.4 127 145 142
ASTM C494, Type F ≥ 1 : 00 ≥ 3 : 30 >12 >125 >115 >110
Retarder type SN-GH 0.15 ≥ 2 : 00 ≥ 1 : 50 102.2 102.1 100
Test results 0.3 ≥ 5 : 15 ≥ 6 : 15 98.9 106.3 100.9
0.6 ≥ 9 : 00 ≥ 10 : 45 93.3 114.7 100.8
ASTM C494, type B ≥ 5 : 00 ≥ 5 : 45 / >90 >90 >90
2.2. FINAL RCC MIX DESIGN AND ITS PERFORMANCES
2.2.1. Determined parameters of RCC mix
Table 8
Construction RCC Mix
RATIO FLYASH REDUCER RETARDER QUANTITIES OF MATERIALS
OF W/C % % % IN 1M3 , KG/M3
water cement Fly C+F sand Agg. Agg. Agg.
ash 4.75∼19 19∼37.5 37.5∼63
mm mm mm
0.71 65 1.0 0.3 122 60 112 172 752 403 539 406
445
Q.104 - R.30
2.2.2. Properties of fresh RCC mix
Table 9
Test result of fresh mix
DENSITY VEBE AIR INITIAL SETTING FINAL SETTING
(KG/M3 ) TIME(S) CONTENT % SETTING TIME(H:MIN) TIME(H:MIN)
2375 6-8 0.8 26:40 52:15
2.2.3. Mechanic properties of RCC mix
Table 10
Mechanic properties test results of RCC mix
COMPRESSIVE STRENGTH (D/MPA) DIRECT TENSILE,(D/MPA) INDIRECT TENSILE,(D/MPA)
7 28 56 90 180 365 28 56 90 180 365 28 56 90 180 365
5.6 10.8 12.3 12.8 15.1 17.3 0.7 1.06 1.14 1.33 1.44 0.61 1.27 1.29 1.78 1.80
Table 11
Developing rate of compressive strength
AGE 7D 28D 56D 90D 180D 365D
Rate,% 0.59 1.00 1.19 1.24 1.43 1.60
2.2.4. Deformation properties of RCC mix
Table 12
Test results of deformation properties of RCC mix
MODULUS OF ELASTICITY POISSON RATIO ULTIMATE TENSILE STRAIN
(GPA) VALUE, (10−6 )
7 28 56 90 180 365 28 56 90 180 365 28 56 90 180 365
5.6 12.2 13.5 14.1 17.8 18.8 0.11 0.12 0.13 0.16 0.2 60 68 74 81 88
446
Q.104 - R.30
2.3. THERMAL PROPERTIES OF RCC MIX
The thermal properties test was executed by Nanjing Hydraulic Research

Institute, China. The main test results are shown in tables 13.
Table 13
Thermal Performance Test Results of RCC mix
SPECIFIC HEAT, THERMAL CONDUCTIVITY, THERMAL CONDUCTIVITY COEFFICIENT OF LINEAR
KJ/(KG·◦ C KJ/(M·H·◦ C COEFFICIENT, M2 /H EXPANSION ×10−6 /◦ C
0.915 9.215 0.0038 0.84
3. STUDY ON CHARACTERISTICS OF JOINT WITH DIFFERENT

MATURITY AND TREATMENT
3.1. TEST JOINT MATURITY AND BONDING STRENGTH
Before the commence of dam RCC construction, the RCC trial section place-
ment was executed, the different exposure times of joint at 4, 10, 16, 22, 36, 48 and
60 hours was arranged. The trial section was separated into 3 zones with treated
separately mortar, grout and no any treatment. The test results of joint bonding
strength under different treatment and maturity are shown in table 14. The modified
maturity factor (MMF) calculation method is MMF= (average ambient temperature
+ 12◦ C) × joint exposure time (hours).
3.2. CRITERIA FOR JOINT TREATMENT
According to the bonding strength shown in the table 14, the longer the expo-
sure time of joint will make the lower the bonding strength. Beyond 22 hours of
exposure time, the strength of treated joint was higher than without treatment. For
hot joints, the joint bonding strength is greater than 60% of parent RCC within 16
hours. Based on the bonding strength results and the modified maturity factor value,
the conditions of Hot, Warm and Cold joints can be initially determined as follows:
1) when the maturity is less than 757 C. H, it is defined as “Hot” joint without any
treatment. 2) When the maturity is between 757–1200 C. H, it is defined as “Warm”
joint. 3) When the maturity is more than 1200 C. H, it is defined as “Cold “joint.
In order to ensure the joint bonding strength of joints, the control criteria were
defined for the dam RCC construction. The criteria for judging hot joints, warm joints
and cold joints are determined shown in table 15. The hot joint condition does not
need any treatment. If the joint maturity reach at warm joints condition to need
treatment with applying grout, after then the placement can be continued. If the joint
447
Q.104 - R.30
maturity reaches the condition the joint will be treated as cold joint, the placement
operation shall be stopped, the joint surface have to roughening by green-cutting
machine, clean up, and before restarting the next placement the joint shall be treated
with grout.
Table 14
Test results of Joint bonding strength with different treatment method
JOINT EXPOSED MMF (◦ C.H) JOINT BONDING STRENGTH WITH DIFFERENT TREATMENT METHOD (MPA.)
TIME(HRS) BY GROUT BY MORTAR WITHOUT TREATMENT
AVERAGE % BY AVERAGE % BY AVERAGE % BY
MPA PARENT MPA PARENT MPA PARENT
RCC RCC RCC
7/8 60 2064.9 0.9* 65.7 0.76* 55.5 0.50 36.5
6/7 48 1866.8 1.02* 74.5 0.80* 58.4 0.36 26.3
5/6 36 1330.1 1.10 60.3 0.73 53.3 0.44 40.1
4/5 22 757.1 1.02 58.6 1.25 71.8 0.64 36.8
3/4 16 589.7 1.34 77.0 1.28 72.0 1.04 59.8
2/3 10 300.1 1.47 84.5 1.33 75.0 1.28 73.6
1/2 4 178.5 1.43 82.2 1.55 88.0 1.44 82.8
*Treated by green-cut as well.
Table 15
Defined control standard of joints as to exposing time and MMF
Month Temperature of Defined control standard of joints as to exposing time and MMF
Ambient, ◦ C Hot joint Warm joint Cold joint
Exposed MMF, Exposed MMF, Exposed MMF,
◦
time, hrs H. C time, hrs ◦
H. C time, hrs H.◦ C
1 22.8 ≤ 16 557 16-28 556-974 >28 >974
2 22.5 ≤ 16 552 16-28 552-966 >28 >966
3 22.1 ≤ 16 546 16-28 546-955 >28 >955
4 20.9 ≤ 20 658 20-30 658-987 >30 >987
5 18.5 ≤ 22 671 22-32 671-976 >32 >976
6 16.5 ≤ 22 627 22-32 627-912 >32 >912
7 16.4 ≤ 22 625 22-32 625-909 >32 >909
8 18.9 ≤ 22 680 22-32 680-989 >32 >989
9 22.5 ≤ 20 690 20-30 690-1035 >30 >1035
10 24.9 ≤ 16 590 16-28 590-1033 >28 >1033
11 24.4 ≤ 16 582 16-28 582-1019 >28 >1019
12 23.2 > 16 563 16-28 563-986 > 28 > 986
448
Q.104 - R.30
4. EFFECT OF RCC LAYER COMPACTED EXISTED IN THE CONDITION

OF PLASTIC AND ELASTICITY
During RCC construction, as the existing RCC layer compacted is in the con-
dition of plastic and elasticity, some RCC dam is not allowed. For this issue have
conducted some study on the affection of the joint quality. The phenomenon is
occurred when the Vebe time is less the 5 seconds. During the RCC trial section
construction the test carried out in KGL project. Based on the core sample test
result, there is no any influence on the joint shear strength, impermeability, bonding
strength. Also the core sample compressive strength and density meets the require-
ment. So the compacted RCC in condition of plastic and elasticity is not limited in
the KGL project.
5. QUALITY CONTROL SYSTEM OF RCC CONSTRUCTION
In order to ensure the interlayer joints quality, the key control points are as
follows:
Temperature control measures, RCC temperature at batching plant: through

taking the measures of primary air cooling of coarse aggregate, adding cold water
and ice chips in mixer, RCC mixture temperature was controlled less than 20 ◦ at the
plant and less than 25◦ at placing site; sunshade used for RCC transport Truck and
belt convey system; Fogy system measures was adopted to reduce the temperature
placement site and to keep the joint surface in the SSD condition. Height of aggregate
pile stock is higher than 6m and set up shading measures for aggregates etc. the
partial test results of temperature control see Fig. 3
Fig. 3
Temperature control at site.
449
Q.104 - R.30
RCC mixture Vebe time, shall be adjusted according to the ambient temper-
ature, Vebe time controlled within 8±2s at placing site. The partial test records of
VeBe time see the Fig 4.
Fig. 4
VeBe time control.
The bedding grout mix quality control, ratio of water/cement within 0.6-0.65,
the flow value within 40±3s.
RCC Mixture segregation control: RCC discharging on RCC surface un-

compacted, so that the mixture is remixed when the spreading and leveling operation
by bulldozer to reduce the segregation.
Fig. 5
Density test at site.
Applying of grout, for the warm and cold joint shall be treated with grout, and
the distribution shall be uniformed on the joint surface.
6. EVALUATION OF INTERLAYER JOINT QUALITY CONTROL
During construction of the KGL RCC dam, for the hot joint the interlayer joint
exposed time reached at maximum 22hours, and the warm joint exposed time
450
Q.104 - R.30
reached at 30hours. Based on the extraction of core samples and water pressure
test results, the joints bonding quality and impermeability are satisfied. Picture 1.
Core sample
Table 16
Water Pressure Test Results
ITEM NO. 1 2 3 4 5 6 7 8 9 10 11
Depth From(m) 0.7 0.7 0.7 0.7 0.7 20 25 0.7 15 0.7 0.7
to(m) 10.1 10.7 25.7 28.5 32.8 20.0 29.9 5.0 20.0 22.5 19.7
Test Length m 9.4 9.9 25.0 27.8 32.1 5.0 4.9 4.3 5.0 21.8 18.9
Lugeon Average Value L/Min/m 0.17 0.31 0.29 0.28 0.29 0.00 0.00 0.01 0.11 0.31 0.41
7. CONCLUSION
By using the MMF criteria applied to the quality control for RCC interlayer
joint, and the practice application of the joint longer time exposed in KGL RCC dam
construction.in the region of dry and high temperature climate is successful, as well
as construction quality control is reasonable. With the long-time exposed makes
continuously RCC placement which improve the placement rate and progress. This
practice can be a reference for other projects that in hot and dry area, or the place-
ment intensity is high, or the batching plant capacity is lower than the requirement
or include these conditions together.
451
Q.104 - R.31
-------
-------
SAFETY EVALUATION OF JINPING I ARCH DAM CONSIDERING THE LEFT

ABUTMENT SLOPE DEFORMATION (∗)
Zhong ZHOU, Jing ZHANG, Lijun XUE, Dewen CAI & Zhongxu LIU
POWERCHINA CHENGDU ENGINEERING CO. LTD., Chengdu, Sichuan, China
CHINA
SUMMARY
The 305m-high double curvature concrete arch dam of Jinping I project, ranks
the highest in the world. It has undergone 6 times of the normal water level since
impoundment in October 2013. Based on monitoring data and analysis using 3D
nonlinear visco-elastic-plastic rheological model, this paper presents the prediction
and evaluation of the long-term deformation of the left abutment slope, and its effect
on the arch dam safety, as well as the overall stability analysis of arch dams and
foundation and provides reference for future practitioners.
RÉSUMÉ
Le barrage en béton à double courbure de 305 m de haut du projet Jinping I

se classe au premier rang mondial. Il a subi 6 fois le niveau normal de l’eau depuis
sa mise en eau en octobre 2013. Sur la base de la surveillance des données et de
l’analyse à l’aide d’un modèle rhéologique visco-élastique-plastique 3D, ce rapport
présente la prédiction et l’évaluation de la déformation à long terme de la pente de
l’appui rive gauche, et son effet sur la sécurité du barrage-voûte, ainsi que l’analyse
globale de stabilité des barrages et des fondations, et fournit une référence pour les
futurs praticiens.
∗ Évaluation de la sécurité du barrage-voûte de Jinping I compte tenu de la déformation de
l’appui rive gauche
452 DOI: 10.1201/9781003211808-30

Q.104 - R.31
Fig. 1
Jinping I Arch Dam.
1. INTRODUCTION
Situated in Liangshan Autonomous Prefecture of Yi minority, Sichuan province,

China, Jinping I hydropower station has a controlling reservoir in the lower reaches
of the Yalong River, the main tributary of the Yangtze River. The normal water level
of the 305m-high dam is at El.1885m. Its underground power plants provide an
installed capacity of 3600MW.
Located in a non-symmetrical deep V-shape canyon with slope of 50◦ ∼ 65◦

steep in left bank, 60◦ ∼ 85◦ in right bank, mainly composed of marble, partially of
sandstone in left bank above El.1800m.Jinping I dam has an extremely complicated
foundation. In sandstone, slope deformation, tension cracks, and stress release are
well developed with different intensity, including moderate, strong, and deep class,
corresponding to depth of 50m ∼ 90m, 100 ∼ 160m, and 200 ∼ 300m, varying with
elevation. Toppling-tension deformation in higher part and toppling deformation in
shallow area at lower part were well developed. Main faults in left abutment slope
are faults f2, f5, f8, f42-9, and lamprophyre dike X, obliquely crossing the slope from
top to bottom, with the strike of N60◦ ∼ 80◦ E, dipping angle of 70◦ ∼ 90◦ , and width
of 1.5m∼3m, resulting in slope failure modes of shear slip, block slip and wedge
shape damage, etc.
The stability of the left bank slope is mainly controlled by a large wedge block
(referred as “large block”) composed of the joint layer SL44−1 as the upstream
boundary, lamprophyre dike X as the trailing boundary, and the fault f42−9 as the
downstream and bottom slip plane. After comprehensive reinforcements of rock
anchors, rock cables and anti-slide concrete replacement tunnels, stability of the
excavation slope of 540m high in left bank has been assured. As the monitoring
453
Q.104 - R.31
Fig. 2
The left abutment slope during excavation of Jinping I Arch Dam.
data indicated, except the slope deformation at high elevation around the excava-
tion line has not yet converged, the slope including the “large block” as a whole is in
a stable state.
Excavation of Jinping I arch dam slope started in Sep. 2005 and completed
in Aug. 2009, then dam concrete placement completed in Dec. 2013, and then the
reservoir impounded to the normal water level 1880m at Aug.24th, 2014. Up to now,
undergone 6 times of high water level, the dam has been working well and the slope
is stable.
2. DEFORMATION CHARACTERISTICS OF THE LEFT BANK SLOPE
According to deformation data, the left bank slope could be divided into 6
deformation zones, as shown in Fig. 3:
Zone 1, Toppling deformation zone at high elevation around the excavation

line;
Zone 2, the deformation zone of upstream residual f 5 and f8 ;
Zone 3, the upstream abutment slope;
Zone 4, the abutment key slope;
Zone 5, the downstream abutment slope;
and Zone 6, the flood discharge atomization zone above plunge pool. Slope
deformation monitors were installed as follows:
454
Q.104 - R.31
Fig. 3
Slope Deformation Zoning in the left Bank of Jinping I Arch Dam.
2.1. SLOPE SHALLOW DEFORMATION
The shallow deformation of the left bank slope is mainly monitored by multi-
points extensometers of about 60m deep, the total displacements are between
−6.76 mm ∼ 27.20 mm so far. The incremental displacements were between
−1.37 mm and 26.42 mm during the construction stage, −2.87 mm ∼ 2.68 mm in
reservoir impounding to normal water level, and −0.61 mm ∼ 2.75 mm from the nor-
mal water level to the present. Slope shallow deformation is small, and by systematic
reinforcements including cables, the slope is stable.
2.2. SLOPE DEEP DEFORMATION
The slope deep deformation monitors have been installed in adit PD44 , PD42
and the drainage tunnels at El.1915m in the upstream abutment slope, as well as
the grouting tunnel at El.1885m, drainage tunnel at El.1829m and El.1785m in the
abutment slope, as shown in Fig. 4:
1) As observed in adit PD42 , PD44 and El. 1915m drainage tunnel, the deep
deformations of the upstream abutment slope have not yet converged, con-
tinue to deform outward, but the increase rate is small and relatively stable,
less than 0.6 mm / month.
2) As deformation monitored in tunnel at El.1855m, El.1829m, El.1785m, deep
deformation at El.1885m is stable, while that of at El.1820m and El.1785m
vary with the water level, specifically, decrease corresponding to water level
drop, and increase with water level going up.
455
Q.104 - R.31
Fig. 4
Deep deformation process line of Jinping I left abutment slope.
2.3. VALLEY DEFORMATION
Valley deformation monitors in the upstream abutment slope were installed in

tunnels alone river transvers direction. As monitored data shown, upstream valley
deformations are generally of shrinkage, and the deformation rates are compara-
tively large during the excavation period, then get smaller in following stage. The
downstream valley deformation is small and getting stable.
2.4. CONCLUSION ON THE LEFT BANK SLOPE DEFORMATION
Mutually verified by the monitored data of the shallow, deep and valley defor-
mation, it could be concluded the deformation in Zone 1 has not yet converged,
while Zone 2 is still in the adjustment period, and Zone 3 ∼ Zone 6 are in steady
state.
Deformations around the fault f42-9 mainly took place in its upper part. In
the area under fault f42-9, including the abutment slope and downstream abutment
slope, the deep deformation is also small and convergent there, suggesting that the
abutment slope and downstream abutment slope has been in a stable state, thus
the arch dam long-term safety is guaranteed.
No independent t deformation or the evidence of the overall slip trend of "large

block" was observed during in the initial water storage stage. The deformation of the
"large block" is mainly comprehensive effects of the water infiltrations into the rock
mass of class III2 -VI outside of fault f42−9 and the lamprophyre dikes X, rather than
that of the block slip. The left bank slope as a whole is in a stable state.
456
Q.104 - R.31
3. ANALYSIS OF THE INTERACTION BETWEEN THE SLOPE AND ARCH

DAM
According to the characteristics of slope deformation of the left bank, the slope
deformation will last for a certain period, therefore, it shall have three stages: the
excavation and unloading deformation stage, the adjustment stage while the effec-
tive stress adjustment and softening of rock mass take place after be immersed by
water during the initial water storage stage, and the stage of long-term deformation
convergence. In order to study the effect of left abutment slope deformations on the
safety of arch dam, the following procedures had been taken into account:
1) Analysis of geological structures and their rheological parameters, including:

Analysis of the mechanism of geo-mechanical deformation, study on the
influence of reservoir impoundment on slope rock mass and weak fracture
zone, rheological test of rock mass of class •2 and other weak interlayers,
analysis of rheological test results and determination of rheological model.
2) Long-term deformation prediction of the slope, including:
(1) Inversion of rheological parameters and long-term deformation pre-
diction. On the basis of slope monitored data and testing results,
rheological model and parameters of the rock mass in left bank were
determined by the displacement feedback analysis, therefore, the long-
term deformation trend and magnitude of the left bank slope could be
analyzed.
(2) Stability analysis of long-term deformation of the slope. The stability
of long-term deformed slope is analyzed by 3D Visco - elastic - plastic
finite element method.
3) Analysis of the effects of the long-term deformation on arch dam,
including:
(1) Establishment of FEM fine grid, and parameter inversion of the arch
dam and the foundation on the basis of incremental displacement during
reservoir impoundment.
(2) The applying method of loading of Long-term deformation.
According to the characteristics of the long-term deformation of the left bank
slope, and taking the displacement of the profile as the objective function,
the displacement loading could be employed on the FEM model boundary,
by adjusting displacement loading range and scale.
(3) The effects of long-term deformation of slope on arch dam. By the above-
mentioned load applying method, the stress and stability of the arch dam
considering the effects could be studied.
(4) Long-term deformation overload analysis and study on the bearing capacity
of arch dam.
Taking the predicted long-term deformation of the slope as the base load, then,
by magnifying the base load in FEM analysis, the overloading capacity of the left
bank slope could be studied.
457
Q.104 - R.31
Fig. 5
Three-parameter (H-K) visco-elastic-plastic model.
Table 1
Inversion results of three - dimensional rheological parameter
SAND SLATE OF SAND SLATE OF SAND SLATE OF Sand slate of FAULT
CLASS III1 CLASS III2 CLASS IV1 class VI2 F42−9
Viscoelastic Viscosity Viscoelastic Viscosity Viscoelastic Viscosity Viscoelastic Viscosity Viscoelastic Viscosity
modulus Coefficient modulus Coefficient modulus Coefficient modulus Coefficient modulus Coefficient
(GPa) (×108 (GPa) (×108 (GPa) (×108 (GPa) (×108 (GPa) (×108
GPa·s) GPa·s) GPa·s) GPa·s) GPa·s)
13.4 10.0 7.6 7.0 5.0 4.0 1.2 1.0 0.7 0.8
4. LONG-TERM DEFORMATION PREDICTION AND SLOPE STABILITY

ANALYSIS
4.1. INVERSION ANALYSIS OF RHEOLOGICAL PARAMETERS
According to the results of creep tests of rock mass, a three-parameter (HK)

model of stable rheological model is adopted for the left bank slope sandstone and
marble, and Moore-Coulomb rule as the plastic yield criterion. The combination of
the two is shown in Fig. 5.
On the basis of the monitored data of displacement across river in each stage
of the slope, the rheological parameters are obtained by inversion analysis.
4.2. LONG-TERM DEFORMATION PREDICTION OF THE LEFT BANK SLOPE
During the span from the slope excavation completion in August 2009 to
September 2014, the dam base abutment deformed 0 ∼ 22.1mm outside, specif-
ically, 0 ∼ 16mm below El.1770m, 16 ∼ 22.1mm between El.1770m ∼ El.1885m.
Since this deformation happened before the completion of dam concrete placement
and joints grouting, this slope deformation had little effect on the arch dam.
By the 3D Visco-elastic-plastic FEM, the long-term slope deformation is pre-

dicted to be between 0 mm ∼ 21.5mm in the span from April 2014 to September
2034, that is, 0 mm ∼12.3mm below El.1770m, 12.3 mm ∼21.5mm between
El.1770m ∼ 1885m.
In general, the subsequent long-term deformation is small, may not has

significant effect on the dam overall stability, but requires further evaluation.
458
Q.104 - R.31
Fig. 6
Deformation increment of left abutment slope bellow El.1885.0m.
Fig. 7
The yield zone of left abutment slopet by 3D FEM.
4.3. LONG-TERM SLOPE STABILITY ANALYSIS
As FEM analysis predicted, after going through 300 months of rheological

deformation since the slope excavation completion in August 2009, most of the
slope will be in compressive state, only in some small surface areas are small ten-
sile stress, and will be further successfully alleviated by the anchor reinforcement.
Although there is also a small tensile stress zone near the fault f42-9 and the deep-
buried fissures, where the stress will also be improved by the anti-shear concrete
replacement tunnels.
As FEM analysis shown, there are some plastic yielding zones below
El.1880m, and got larger after reservoir impoundment, due to the boundaries of
"large block", namely, joint layer SL44-1, lamprophyre X and fault f42-9 are mostly
higher above the normal water level El.1880m, they have not the overall yield as
shown in Fig. 7. So the ‘large block’ as a whole is stable.
459
Q.104 - R.31
5. ANALYSIS ON THE INFLUENCE OF SLOPE LONG-TERM

DEFORMATION ON ARCH DAM
5.1. STUDY ON THE LOAD APPLYING METHOD OF THE SLOPE LONG-TERM

DEFORMATION
As for arch dam structure, Long-term deformation of the left bank slope could
be regarded as an additional external displacement load, therefore, the load applying
method of slope long-term deformation is a key research topic.
By use of 3D elasto-plastic and Visco-elasto-plastic FEM by TFine program

and Flac3D program, combined with the inversion analysis of monitoring data for
mechanical parameters of dam foundation, several loads applying approaches of
slope long-term deformation had been put forward.
Taking the slope deformation of several profiles as objective function, by using

of multiple linear fitting method, and increasing the weight of the displacement target
near the abutment slope, the least squares optimization based on the elasto-plastic
calculation were carried out, and the optimal load applying method of slope long-term
deformation is obtained.
Taking into account the effects of different displacement distribution scope,

magnitude and patterns, e.g. triangles, trapezoids, as well as other factors, the
optimization for load applying had been implemented, and the inverted triangular
Polyline distribution pattern was proved to be reasonable, as shown in Fig. 8.
Fig. 8
Dragram of load applying method of long-term deformation of left abutment slope.
5.2. DAM DEFORMATION AND STRESS ANALYSIS
1) Deformation
Under normal water level of El.1880m, and the long-term deformation of the
slope, the maximum increment of the displacement across river is 16.6mm at the left
460
Q.104 - R.31
arch dam crest toward the river bed, while that along river is 9.6mm near the crest of
crown cantilever toward upstream, accounting 13.8% to the maximum displacement
by water load alone.
Similarly, under dead water level of El.1800m, and the long-term deformation of
the slope, the maximum increment of the displacement across river reached 16.1mm
at the left arch dam crest toward the river bed; while that along river 9.6mm near
the crest of crown cantilever toward upstream, accounting to 29% of the maximum
displacement of water load alone. The maximum incremental displacement along
river in Left arch end at crest elevation is 2.5mm toward downstream.
2) Stress
Under the normal water level, the long-term deformation of the slope make
the maximum value of the main tensile stress decreased from 2.93MPa to 2.47MPa,
and the maximum value of the downstream main compressive stress decreased
from –21.88MPa to –21.39MPa, but the upstream tensile stress of the arch is about
0.70MPa and 1.25 MPa increased to 0.88 MPa and 1.39 MPa.
Similarly, under the dead water level, the main compressive stress and the
main compressive stress are decreased, but the upstream tensile stress region is
extended from the abutment to crown cantilever, and the downstream compressive
stress is concentrated in the left of crest arch.
Fig. 9
The upstream dam displacement.
Fig. 10
Principal stress on the dam face under the normal water level and long-term
deformation of slope (0.1Mpa).
461
Q.104 - R.31
3) Summary of the influence of left abutment slope deformation on dam

structure
(1) The dam stress was mainly sensitive to the deformation of the abutment
slope below dam crest, and less sensitive to that above. Given that slope
deformation mainly took place in the slope above the crest, it has no
significant effect on dam structure.
(2) The high stress area of the dam caused by the deformation of the slope was
not coincident with that under the normal load combination. The long-term
deformation of the slope has little effects on the dam maximum principal
tensile stress and compressive stress.
(3) The long-term deformation of the left bank slope were assumed to pro-
vide a thrust to the arch dam which is equivalent to a slight increase in
the stiffness of the middle and upper dam foundation, inhibited the asym-
metry of foundation in two banks. Therefore, the dam safety has not been
undermined.
6. SAFETY EVALUATION OF ARCH DAM UNDER SLOPE LONG-TERM

DEFORMATION
6.1. SLOPE DEFORMATION OVERLOAD ANALYSIS
Considering the complexity of the deformation mechanism of the left bank

slope and the reliability of the long-term deformation prediction, in order to further
analyze the safety of the arch dam under the deformation of the left bank slope, the
deformation overloading analysis was carried out.
Taking the predicted long-term deformation of the slope during May,2014 to

September,2034, as the base load, then, by magnifying the base load in FEM anal-
ysis, the overloading capacity of the left bank slope could be studied by the following
three indexes:
1) The crack-initiation displacement load α1: cracks occurred in dam (normally,

in dam heel or the left dam abutment, or other parts prone to cracking);
2) The nonlinear deformation displacement load α2, indicates the load make
the dam start large nonlinear deformation. It could be mainly observed by
the turning point of dam displacement, yield area expanding, or plastic
energy increment;
3) The ultimate bearing capacity displacement load α3, indicates the bearing
capacity of the dam, and could be found by the connecting of all the yield
areas of the dam in FEM analysis.
462
Q.104 - R.31
As the overloading analysis shown, the crack-initiation displacement load α1 =

3, corresponding to under normal water level and the dead water level are 60.21mm
and 59.28 mm respectively; The nonlinear deformation displacement load α2 = 8;
The ultimate bearing capacity displacement load α3 = 12. It shows that Jinping I
Arch Dam has high safety for long-term deformation of left bank slope. Under the
long-term deformation predicted according to the monitoring data, Jinping I arch
dam is safe.
6.2. SUMMARY OF THE DEEP DEFORMATION OF THE LEFT ABUTMENT SLOPE
As the monitoring data shown, the deep deformation of left abutment slope is
small, see in Fig. 4. The process of deep deformation also shown that, deformation
at EL.1885m has been stable; while at EL.1829m and EL.1785m were affected by
the water level changes, that is, abutment slope at these elevations has a trend of
move outside during the water level drop, while inside during water level increases.
It suggests that the deep deformation of the abutment slope of the left bank is
basically stable and its influence on the arch dam is small. The predicted long-term
deformation mainly take place in the upstream abutment slope, the deep deformation
of the dam foundation slope is small, it will have little effect on the safety of arch
dam.
6.3. ANALYSIS OF ARCH CHORD LENGTH CHANGE
As monitoring data shown in Fig. 11, the length decrease amount of dam chord
totally reached 4.15mm ∼ 15.36mm by March 21, 2016, while the largest chord
shortening was at El.1730m. In general, in the early stages of reservoir impound-
ment, due to the effects by the excavation unloading, effective stress adjustment in
slope and the softening of rock mass immersed in water, dam chord length had a
small contraction. Hereafter, the chord length was elongated and compressed with
the water level up and down, respectively, exhibiting a good correlation with the
water level.
Fig. 11
Pass line of arch chord length Jinping I Dam.
463
Q.104 - R.31
7. CONCLUSIONS
1) The deformation of the left bank slope will have 3 deformation stages,
namely, the deformation stage of slope excavation and unloading, the defor-
mation adjustment stage when the effective stress adjustment and softening
of rock mass immersed under water take place, and the stage of long-term
deformation convergence. Until now, the slope is in the second stage, and
"large block" is in a steady state.
2) As the long-term deformation prediction and safety analysis suggested, the
converge of left bank slope deformation will last for a long time till 2034; the
long-term deformation will have little effect on the arch dam, and the slope
is stable.
3) As the over-load analysis of long-term deformation shown, the arch dam has
strong overload capacity to slope deformation. Equally important, deforma-
tion of the dam foundation is small, thus the slope is stable, and the length
change of dam chord is consistent with water level in like manner. In a word,
the arch dam is proved to be safe under the long-term deformation of the
slope.
4) The effect of long-term deformation of the slope of the arch dam is com-
plex, still requires comprehensive studies in a complex model, taking into
account the concrete creep, dam temperature rise, slope and foundation
Visco deformation and other factors, and even multi-field coupling effects,
to further evaluate the effect of the long-term deformation on arch dam safety.
REFERENCES
[1] Yang Qiang, Liu Yaoru, Cheng Li, Zhou Zhong, Xue Lijun. 2015. Safety Study
on Long-term Deformation of Left Bank Slope and the Mutual Influence of Arch
Dam of Jinping I Hydropower Station, Tsinghua University Research Report.
[2] Zhou Chuangbing, Jiang Qinghui, Wei Wei, 2015. Left bank slope long-term
stability analysis and prediction’, Wuhan University Research Report.
[3] Zhou Zhong, Zhang Jing, Xue Lijun, 2017. The Left Abutment Slope Defor-
mation and its effect on the safety of Jinping I Arch Dam, Yangtze River, No. 2,
Vol. 603.
464
Q.104 - R.32
-------
-------
FOUNDATION TREATEMENTS OF JINPING I ARCH DAM (∗)
Xue LIJUN1 , Duan SHAOHUI2 , Zhao YONGGANG1 , Wang RUI2 & Shen MANBIN2
1 POWERCHINA CHENGDU ENGINEERING CO. Ltd., Chengdu, China
2 YALONG RIVER HYDROPOWER DEVELOPMENT COMPANY, Ltd., Chengdu,
China
CHINA
SUMMARY
Jinping-I high arch dam is the highest arch dam (305 m) in the world. As
located in a complex tectonic area in southwest China, the project has the most
complicated geological conditions among ultra-high arch dams, characterized by
deep canyon, high steep slopes, high in-situ stress, high water head and deep stress
relaxation etc. In the dam site, well-developed faults and weak rock, lamprophyre
dikes, deep-buried fissures and intensively stress relaxation zones in left abutment
led to 10∼13 times in difference between deformation modulus of the two banks
(see Fig. 1), giving rise to a great challenge in dam design. By the comprehensive
foundation treatments including a concrete cushion with height of 155 m and volume
of 560,000 m3 , 5 layers of grouting gallery, 720,000 m long consolidation grouting,
3 layers of anti-shear gallery, and systematic concrete replacements of fault zone
and lamprophyre dikes with volume of 146,000 m3 , the dam foundation has been
systematically treated, ensure the highest dam been successfully constructed and
operated.
RÉSUMÉ
Le barrage-voûte de Jinping-I est le plus haut barrage d’arche (305 m) au

monde. Situé dans une zone tectonique complexe dans le sud-ouest de la Chine, le
∗ Traitement des fondations du barrage-voûte de Jinping I
DOI: 10.1201/9781003211808-31 465

Q.104 - R.32
projet présente les conditions géologiques les plus complexes parmi les barrages-
voûtes ultra-hautes, caractérisées par une vallée profonde, des pentes abruptes,
des contraintes élevées, une hauteur de chute importante, etc. Sur le site du
barrage, des failles bien développées et des roches faibles, des digues lampro-
phyres, des fissures enfouies en profondeur et des zones de resulding intensivement
stressées dans l’appui gauche ont entraîné 10 à 13 fois de différence entre les
modules de déformation des deux rives, ce qui donne lieu à un grand défi dans la
conception des barrages. Par les traitements complets de la fondation comprenant
un tapis en béton avec une hauteur de 155m et un volume de 560 000 m3 , 5 couches
de galerie de coulis, 720.000 m de long coulis de glyndeon, 3 couches de la galerie
anti-cisaillement, et des changements en béton du système de la zone de faille et
de digues lamprophyre avec un volume de 146.000 m3 , la fondation du barrage a
été traitée système, assurer le plus haut le barrage a été construit et exploité avec
succès.
1. INTRODUCTION
The arch dam of Jinping I Hydropower Station is 305m high, and the total water
thrust of the arch dam is nearly 12 million tons at the normal storage level. The geo-
logical conditions of the resistance body on the left bank of the dam site area are
extremely complex, mainly including f5, f8 faults, lamprophyre veins, f2 faults, inter-
layer compression dislocation zone, deep cracks, and low-wave speed tensile and
loose rock masses. These geological defects will have a greater impact on the choice
of arch dam shape, the stress behavior of the dam body, the anti-sliding stability and
deformation stability of the arch abutment, and the seepage control of the foundation.
Therefore, Comprehensive foundation treatments are required to these geological
defects reasonably based on the topography and geological characteristics of the
dam site combined with the force characteristics of the arch dam.
Fig. 1
Jinping I Hydropower Station.
466
Q.104 - R.32
Fig. 2
Dam-foundation Interface.
2. FOUNDATION TREATMENT DESIGN
According to location and size of the geological conditions, the effects of dif-
ferent defects on the deformation and stability of the arch dam has been analysed
numerically, as well as systematic treatments been evaluated and determined to
make sure the arch dam foundation meet the requirements of strength, deformation
stability and anti-seepage, and ensure and improve Treatment measures include:
Concrete cushions, gravity piers, concrete replacement grid, consolidation grouting
and curtain grouting.
Fig. 3
Numerical model of Jinping I arch dam.
As Foundation treatment effects analysis shows, The concrete cushion

can prominently enhance stiffness of the foundation between El.1750m to 1850m,
with 30% reduction to the displacement along the river near the 1830m elevation.
Secondly, The abutment consolidation grouting can reduce the displacement by
13.3% to 21.3% displacement between El.1710m to 1830m. Thirdly, the concrete
grid replacement of fault f5. The deep-buried fissures is far from the arch and the
f14 fault is on the right bank, which has little effects on the displacement near the
left bank foundation.
467
Q.104 - R.32
Fig. 4
Foundation displacement along &cross the river under different measures.
Fig. 5
Foundation displacement along the river before and after foundation reinforcement.
By numerical analysis and geomechanical model tests, the technology of

deformation stabilization control focusing on the stiffness matching between dam
and foundation was proposed, the largest comprehensive foundation treatments
project has been put forward and conducted including a concrete cushion with height
of 155m and volume of 560,000 m3 , 5 layers of grouting gallery, 720,000m long con-
solidation grouting, 3 layers of anti-shear gallery, and systematic concrete replace-
ments of fault zone and lamprophyre dikes with volume of 146,000 m3 (see Fig. 6∼7)
Fig. 6
Foundation treatments layout in the upper part of the Jinping I Dam left abutment.
468
Q.104 - R.32
3. DAM SAFETY EVALUATION
The results of three-dimensional non-linear numerical analysis and geome-

chanical tests show after the reinforcement treatment, the integrity and uniformity of
the rock mass of the arch abutment are improved, and the deformation resistance of
the dam foundation and the resistance body rock mass is improved, and the Jinping
I arch dam has a strong overload capacity, the foundation treatment measures are
effective, and the arch dam foundation system is overall safe.
It has undergone 7 times of the normal water level, since October 2013.The
Seepage is very small, the displacement of Jinping I Arch Dam is basically elastic
along with the water level (see Fig. 7 and Fig. 8)
Fig. 7
Seepage of Jinping I Arch Dam.
Fig. 8
Displacement of Jinping I Arch Dam.
469
Q.104 - R.32
4. CONCLUSION
The Jinping I dam foundation treatment measures are reasonable. the numer-
ical analysis of dam foundation treatments in dam design is useful. The Jinping I
dam works well and is safe.
REFERENCES
[1] Xue Lijun, Huang Zhipeng, Finite Element Analysis of Foundation Treatment
Effect of Jinping I Arch Dam, Hydropower Station Design, Issue 4, 2005.
470
Q.104 - R.33
-------
-------
DECOMMISSIONING OF AN ARCH-GRAVITY DAM AND MEASURES TAKEN

FOR ITS REPLACEMENT BY A DOUBLE CURVATURE ARCH DAM (∗)
Marc BALISSAT
Dam Safety Expert, STUCKY Ltd
Jérôme FILLIEZ
Dam Project Manager, STUCKY Ltd
Andres FANKHAUSER
Head of Civil Works, KRAFTWERKE OBERHASLI LTD, Switzerland
SWITZERLAND
SUMMARY
The 110 m high Spitallamm arch-gravity dam built in the early 1930’s was at
his time one of the highest concrete dams in Europe. The massive curved structure
was erected in single blocks separated by breaches to allow for dissipation of the
heat of hydration of the concrete. The use of gutters for bringing concrete to the
pouring locations led to adopt a very fluid mix with high water content. After fifteen
years of operation the downstream dam face showed damages due to concrete
freezing and the upstream dam face presented some leaking spots. A campaign
of grouting was undertaken that did only partially resolve the problems. Ten years
later a severe increase of seepage through a construction joint 12 m below the dam
crest led to additional waterproofing measures at the upstream dam face. Additional
instruments were installed, and it became evident that the upper dam arch was mov-
ing independently from the remaining structure. A heightening project that included
∗ Mise hors service d’un barrage poids-voûte et mesures prises pour son remplacement par
une voûte à double courbure
DOI: 10.1201/9781003211808-32 471

Q.104 - R.33
removal of the critical dam zones was objected by nature conservation groups and
could not be enforced. The unusual condition of the dam led finally to the decision of
building a new double curvature arch dam as main retaining structure immediately
downstream of the existing one. Special precautions had to be taken to realize this
task.
RÉSUMÉ
Le barrage poids-voûte de Spitallamm, qui a été édifié au début des années

1930, était à cette époque un des plus hauts ouvrages en béton d’Europe. Le barrage
a été construit en blocs de béton isolés, séparés par des brèches, afin de permettre
la dissipation de la chaleur de prise du béton. Le transport du béton par gouttières
jusqu’aux emplacements de mise en place a nécessité l’emploi de mélanges très
fluides avec une teneur en eau élevée. Après quinze années d’exploitation la face
aval du barrage présentait des dommages dus au gel et le parement amont était
le siège de fuites localisées. Une campagne d’injections n’a permis de résoudre
que partiellement ces défauts. Dix ans plus tard une augmentation importante des
fuites au niveau d’une reprise de bétonnage 12 m sous le couronnement a néces-
sité des mesures additionnelles sur le parement amont. La zone concernée a été
munie de nouveaux instruments de mesure. Il s’est alors avéré que l’arc supérieur
se déformait indépendamment du reste de l’ouvrage. Un projet de surélévation, qui
prévoyait d’éliminer les zones critiques, a fait l’objet d’oppositions par des groupes
environnementaux et n’a pas pu être réalisé. L’état particulier du barrage a finale-
ment conduit à la décision de construire une nouvelle voûte à double courbure,
comme principale structure de retenue, directement à l’aval du barrage existant.
Des précautions particulières ont été prises pour réaliser cette tâche.
1. INTRODUCTION
The 114 m high Spitallamm arch-gravity dam, together with the Seeeuferegg
gravity dam, form the Grimsel reservoir with a live storage of 100 mio m3 at an
altitude between 1’800 and 1’900 masl in the Bernese Alps (see Fig. 1). The dam
was erected from 1926 to 1932. It was at its time one of the highest concrete dam in
Europe. The dam is set in a narrow gorge cut mainly in Aare granite and granodiorite,
both rocks of high strength and low deformability. Although a design with a straight
gravity dam was first considered the designers eventually opted for a curved form,
allowing to direct part of the forces on the abutments and therefore selecting a
steeper downstream face (1V:0.5 H instead of 1V:0.7 or 0.8 H). The upstream face
was also slightly inclined (1V:0.1 H) to have a more favorable direction of the water
pressure forces.
472
Q.104 - R.33
Fig. 1
The Spitallamm Dam on the Aare River.
Barrage de Spitallamm sur la rivière de l’Aar.
The total volume of the dam amounts to 340’000 m3 with a crest length
of 258 m. Due to the excellent foundation conditions excavation depth was mini-
mal. Concrete aggregates were obtained from the alluvium in the reservoir basin.
They are exclusively composed of hard igneous rock. In accordance with contem-
porary practice concrete was brought to the pouring locations by gutters. This led
to adopt a very fluid mix with a water/cement ratio of 0.83. The cement content
was in general 190 kg/m3 with an increase to 300 kg/m3 at the upstream face for
a better wear resistance and water tightness. The cross section was subdivided in
3 zones: upstream face, trough and main volume/downstream face. The dam has
been erected in single blocks separated by breaches (see Fig. 2) to allow for dissi-
pation of the setting heat of the concrete. Furthermore the thirteen radial joints and
partial joints in the upper part of the dam were carefully filled under pressure with
grout to obtain the required arching effect from abutment to abutment.
2. AGEING PROBLEMS
After fifteen years of operation the downstream dam face was showing already
damages due to concrete freezing and the upstream dam face presented some
leaking spots. A campaign of grouting was undertaken that did only partially resolve
473
Q.104 - R.33
Fig. 2
The dam during construction (Note the gutters used for carrying the concrete).
Barrage en cours de construction (notez les gouttières utilisées pour le transport
du béton).
Fig. 3
Schematic view of the upstream dam face showing the local grout takes.
Vue schématique de l’amont du barrage illustrant les absorptions locales de coulis.
the problems. A total of 97 m3 of grout was injected behind the upstream face over a
surface of roughly 8’000 m2 (see Fig. 3). This corresponds to an average volumetric
displacement of 1.2 cm per m2 . The resulting but unexpected effect was a separation
of the upstream face from the trough and the formation about 12 m under the dam
474
Q.104 - R.33
Fig. 4
Cross section with the grouted area and the location of the crack).
Coupe transversale avec la zone d’injection de coulis et la localisation de la fissure.
crest of a nearly horizontal crack extending to the downstream face (see Fig. 4).
The level of this crack corresponds to a former concreting joint over the Winter
1930/31.
At the downstream face single places damaged by frost were regularly repaired
with concrete and mortar. Ten years later a severe increase of seepage through the
upstream face led to additional waterproofing measures. The seepage discharge
could be substantially reduced. In the course of time additional instruments were
installed to better monitor the behavior of the upper part of the dam. It became then
obvious that the upper arch was no longer behaving as the rest of the dam, and that it
was continuously moving in upstream direction by a cumulative rate of 1.5 to 2.0 mm
per year at the crown cantilever.
Inclinometer tubes installed on both sides of and at the crown cantilever in

1996 allowed to closely follow the movements along the afore mentioned crack.
Thus a total displacement of 25 mm towards upstream could be registered over a
period of 19 years. The inclinometer readings are complemented by measures with
a sliding micrometer to document vertical displacements also.
475
Q.104 - R.33
Fig. 5
Displacements (2011–2015) registered in the inclinometer tubes at the crown
cantilever.
Déplacements (2011–2015) enregistrée dans les tubes inclinométriques à la clé.
Levelling of the dam crest confirmed that the crest was the seat of a swelling
process with a constant rate of increase of 1.4 mm per year. Several hypothe-
ses were formulated to explain this behavior. However analyses showed that the
ambient temperature could not to be the main trigger for the upstream move-
ments of the upper arch. Testing of dam concrete samples did not reveal either
significant signs of swelling through alkali aggregate reaction (AAR). A further
attempt to identify tectonic displacements at the dam flanks by cross-valley geode-
tic measurements did also not reveal movements that could affect the dam slender
upper part.
In the late 1990s plans to heighten the dam by 22 m and increase substantially
the live reservoir volume were made, that called for a removal of all parts of the dam
476
Q.104 - R.33
which were considered problematic, i.e. essentially the upstream face and the trough
concrete. The heightening consisted in building a new double curvature arch dam
directly set up on the sound mass concrete.
The dam heightening plans had to be eventually postponed due to the opposi-
tion by environmentalist groups. It became then necessary to improve the situation
of the Spitallamm Dam that was slowly deteriorating with time and could in a near
future no longer match the safety requirements for a large dam impounding a major
amount of water.
3. NEW PLANS
At the initiative of the Owner it was decided to build an entirely new structure
downstream of the existing one. This concept was presenting several advantages:
• construction of the new dam would not interfere with the operation of the
existing reservoir
• the existing dam could be left in place with a bypass system to equilibrate the
water levels at the reservoir and in the open space between both dams.
Considering on one hand the limitation of space due to the presence of a

storage cavern on the right bank, but on the other hand also the excellent foundation
conditions, it was clear that a double curvature arch dam would be the best, if not
the only feasible solution.
The dam crest was set up at the same level as the existing dam with the
possibility of heightening the new dam at a later stage by 22 m. The geometry
was defined by sets of parabolas both horizontally and vertically. The detailed
layout was adjusted several times to eventually end out with following main
dimensions:
• height from lowest foundation level: 113 m (136 m after heightening)

• crest length: 209 m
(279 m for the heightened dam)
• radius at dam axis: 153 m
• crest width: 8.1 m (4.0 m for the heightened dam)
• thickness at base 19.8 m
• volume of the dam 212’500 m3
(245’500 m3 for the heightened dam)
3D stress analyses were conducted for both the basic configuration and
for the heightened dam to ascertain that no strengthening would be required,
should such heightening be carried out at a later stage. Therefore, relatively thick
477
Q.104 - R.33
Fig. 6
Layout of the new dam downstream of the existing structure.
Vue d’ensemble du nouveau barrage à l’aval de la structure existante.
arches were adopted for the basic configuration with the main static compressive
stress not exceeding 6 MPa. After heightening the same stress would increase
to 9 MPa.
According to the Swiss dam safety regulations a dynamic analysis of the dam
was also conducted applying three nondependent synthetic accelerograms to the
structure. The max. horizontal tensile stress was found to reach 3 MPa, but it cannot
be considered as critical as the vertical block joints will open during strong shaking
and the stress will be relieved.
The general layout and the main cross section of the new dam are presented
in Fig. 6 and Fig. 7
478
Q.104 - R.33
Fig. 7
Main cross section of the new dam.
Section principale du nouveau barrage.
Foundation excavation started in 2019 and is foreseen to be concluded in 2020.

As the operation of the reservoir is continuing without restriction a set of additional
measures have been taken to guarantee the safety of the existing dam and of the
ongoing works. It encompasses more frequent and automatic readings of plumb
lines, uplift pressures and seepage discharges, as well as tachymetric readings of
an increased array of measuring points at the downstream face of the existing dam.
A continuous geological survey of the excavation works allows to check the presence
and the condition of discontinuities. Rock anchoring is prescribed where the local
stability of the excavation walls proves to be marginal or insufficient. Extended zones
with problematic conditions have not been encountered so far, neither are expected
to be found until termination of the excavation works.
Should an unexpected behavior at the existing dam and its foundation be

observed during the excavation works it will be possible to rapidly lower the reser-
voir level by discharging water through the bottom outlet at the left bank. A new
by-pass gallery connecting the reservoir and the space left between both dams
shall be excavated through the existing dam in Winter 2025. Thus an equilibrium
479
Q.104 - R.33
of water levels on both sides of the existing dam will be achieved and constantly
maintained.
Fig. 8
Ongoing excavation at the left abutment just next to the existing dam (as of Aug.
20, 2020).
Excavation en cours de l’appui gauche juste à l’aval du barrage existant
(20.08.2020).
Concreting of the new dam is due to start in Spring 2021 and to last until Fall
2024. The existing dam, although decommissioned, will remain just behind the new
dam, since a removal operation proved to be unpractical, as it would impair the
operation of the reservoir over a longer period. Monitoring of the existing dam will
continue but be limited to the observation of displacements at the dam crest. Once
the largely submerged structure will have lost its primary water retaining function,
significant deformations are not expected to occur.
480
Q.104 - R.33
4. CONCLUSION
The case of the Spitallamm Dam illustrates the replacement of an older

structure by building a new one immediately downstream, while continuing the unre-
stricted operation of the reservoir. As excavation for the new dam proceeds a careful
attention shall be given to the existing structure and to its foundation to ascertain that
no critical uplift is taking place nor unusual displacements are occurring, that could
impair the stability and the strength of the existing dam. Thanks to a very compre-
hensive monitoring program it was possible at Spitallamm to carefully control these
aspects during the excavation works. Observation shall continue for the coming
years (2021–2024) during which the new dam will be concreted. Full impounding of
the new dam once the bypass gallery will be operational (2025) will also requires a
careful monitoring of the structure and its foundation.
481
Q.104 - R.34
-------
-------
TWO RECENT CASES OF ARCH DAM RAISING,

LESSONS LEARNT AND INNOVATION (∗)
Marcelo LEITE RIBEIRO, Olivier VALLOTTON & Alexandre WOHNLICH

STUCKY SA
SWITZERLAND
SUMMARY
In the last decade, the authors had the unique opportunity to engineer the
raising of two arch dams in two different countries with completely different contexts
and preexisting conditions. One is Vieux Emosson dam in the Swiss Alps, 55 m high
(before raising), which was built in the early 1950s, whilst the second is Cambambe
arch dam in Angola, 72 m high and built in the early 1960s. Both were raised by
approximately 25 m between 2010 and 2020.
The hurdles to be overcome to succeed the challenge of raising an arch dam

are many, some similar for both dams, some very site specific. The article provides
interesting comparisons and insights on technical details which had to be solved,
such as the definition of the contact surface between the old initial dam and the
new concrete, the geometrical definition of the double curvature shape and the
management of the river flow during the raising works.
The paper shares the lessons learned on both cases and gives recommen-
dations on the critical design and construction issues to be addressed to complete
successfully such a challenging project as an arch dam raising.
RÉSUMÉ
Au cours de la dernière décennie, les auteurs ont eu l’opportunité unique de

concevoir la surélévation de deux barrages-voûtes dans deux pays différents avec
∗ Deux cas récents de surélévation de barrages-voûtes, enseignements et innovation
482 DOI: 10.1201/9781003211808-33

Q.104 - R.34
des contextes et des conditions préexistantes complètement différents. Le premier

est le barrage de Vieux Emosson dans les Alpes suisses, de 55 m de haut (avant la
surélévation), qui a été construit au début des années 1950, tandis que le second
est le barrage de Cambambe en Angola, haut de 72 m et construit début des années
1960. Les deux ont été surélevés d’environ 25 m entre les années 2010 et 2020.
Les obstacles à surmonter pour réussir le défi de surélever un barrage-voûte

sont nombreux, certains similaires pour les deux barrages en question, certains très
spécifiques au site. L’article fournit des comparaisons et des aperçus intéressants
sur des détails techniques qui ont dû être résolus, tels que la définition de la sur-
face de contact entre l’ancien et le nouveau béton, la définition géométrique de la
forme à double courbure et la gestion du débit de la rivière pendant les travaux de
surélévation.
L’article partage les enseignements gagnés sur les deux cas et donne des
recommandations sur les problèmes critiques de conception et de construction à
résoudre pour mener à bien un projet aussi difficile et audacieux que la surélévation
d’un barrage-voûte.
1. INTRODUCTION
Over the past decade (2010-2020), the authors had the unique opportunity to
be deeply involved in the design and site supervision of the raising of two different
arch dams: one is Cambambe arch dam in Angola, whilst the second is Vieux
Emosson arch dam in Switzerland. Both projects differ in many aspects, but they
have similar heights and raising heights. Although so different in weather and thermal
conditions, river management during the works, and flood range, both dams can be
interestingly compared, and general trends and conclusions can be drawn from such
experiences. The article presents and discusses both cases from the perspective
of selected technical topics which turned out to play a critical role in the design and
construction process.
2. THE RAISING PROJECTS
2.1. CAMBAMBE DAM AND HYDROPOWER SCHEME
2.1.1. Original Arch Dam
The Cambambe arch dam is located in Angola and was built between 1959
and 1963 (commissioned) on the Cuanza River, 10 km southeast of the town of
Dondo and 180 km from the capital Luanda.
483
Q.104 - R.34
Fig. 1
Hydrological data of Cambambe site (left) and measured water levels in the
reservoir (masl) from 1998 to 2007 (right).
Données hydrologiques du site de Cambambe (à gauche) et niveaux d’eau
mesurés dans le réservoir (msm) de 1998 à 2007 (à droite).
The original arch dam was 72 m high, with a crest length of 250 m. It is a double
curvature arch dam, designed according to a geometric definition of the three-circle
arch. The dam and its reservoir have an energy vocation; The original powerplant
had an installed capacity of 4 × 45MW = 180 MW.
The catchment area is 115’500 km2 . The average discharge is around

750 m3 /s, significantly more than the original installed discharge of the Cambambe I
powerplant (4 × 65 m3 /s, i.e total 260 m3 /s). Therefore, the overflow spillway at the
centre of the dam operated continuously from the end of the construction (Fig. 1).
From its conception the dam was designed to be raised and a second power
station was also planned in parallel to the existing power station. However, for various
historical and political reasons, this work to heighten and enlarge the development
never saw the light of day before the beginning of the 2010s.
2.1.2. Dam Raising
The dam raising works concluded in 2019 are based on the initial design project
(Fig. 2). These works took place in parallel with the rehabilitation and transformation
of the existing 50-year-old plant and the construction of a new open air powerplant
with a capacity of 4 × 175 MW = 700 MW. After the raising works, the Cambambe
scheme is currently rated at 700 MW+4 × 65 MW (increase in the power of the
existing plant thanks to its rehabilitation and the elevation of the reservoir by around
30 m), for an upgraded total of 960 MW. This represents more than five times the
original installed capacity in operation since its construction in the 1960s.
The flood evacuation system is designed for a project flood with a return period
of 10’000 years, estimated at 9’000 m3 /s. The raised dam is characterized by the
484
Q.104 - R.34
Fig. 2
Original design (left) and the design executed between 2012 and 2019 (right).
Conception originale (à gauche) et exécutée entre 2012 et 2019 (à droite).
Fig. 3
Plan view and cross-section showing the central orifice spillway.
Vue en plan et coupe montrant le déversoir de l’orifice central.
incorporation of five orifice spillways with a capacity of 5 x 900 m3 /s, i.e. 4’500 m3 /s
(Fig. 3) in its central blocks. At design flood, another 4’500 m3 /s flow through a lateral
spillway constructed on the left bank, composed of two gated passes, each 19.5 m
wide.
Fig. 4 presents the view from downstream (left) and the bird’s eye view (right)
of the Cambambe dam after the conclusion of the works.
485
Q.104 - R.34
Fig. 4
View from downstream (left) and bird’s eye view (right) of the Cambambe dam
after the conclusion of the works.
Vue aval (gauche) et vue amont (droite) du barrage de Cambambe après la
finalisation des travaux.
2.2. VIEUX EMOSSON DAM
2.2.1. Original Arch-Gravity Dam
The original Vieux Emosson dam is a 55 m high, single curvature arch-gravity

dam which was built in the 1950s in the Swiss Alps (Canton Valais) at an elevation
of 2’200 masl.
The catchment area is rather small (4.5 km2 ). Precipitations are more or less
constant throughout the year and reach up to approximately 1’700 mm. In winter, it
is not rare that the cumulated snow height on the ground reaches 5 m. The average
yearly flow from the catchment area is around 0.4 m3 /s. In winter, the flow is quasi
nil, whilst in summer it increases to approx. 1.2 m3 /s. Flash floods are generally
smaller than 10 m3 /s.
The average site temperature throughout the year is around 1˚ C. In wintertime,

the air temperature can be as low as -20˚ C while in summer it rarely reaches 20◦ C.
2.2.2. Dam Raising
The raised Vieux Emosson dam forms the upper reservoir of the Nant de
Drance pumped storage scheme, while the lower reservoir is formed by the existing
Emosson dam. The new pumped storage scheme, with a capacity of 900 MW is
being commissioned in 2020.
486
Q.104 - R.34
Fig. 5
Vieux Emosson dam raising concept.
Concept de surélévation du barrage de Vieux Emosson.
Economic studies have shown that it would be interesting to increase the

original upper reservoir capacity in order to give more operational flexibility to the
pumped storage scheme. With the storage level being raised by 20 m, the storage
volume of the upper reservoir is more than doubled, going from 12.1 to 25.3 hm3 .
When considering the water conveyed from the natural and neighboring catch-
ment areas, more than two years would be necessary to fill up the increased upper
reservoir. On the other hand, only some pumping hours from the Nant de Drance
scheme (flow larger than 300 m3 /s) would be enough to fill the reservoir.
With the dam raising, the original single curvature arch dam is converted into a
double curvature arch dam (parabolic shape). Mainly for geometrical constraints at
the contact between the original and the heightened dam, it was found compulsory
to demolish the upper part of the original dam by 10 m on the upstream face and 20
m on the downstream face, thus forming a 10 m high step in the heightened dam
section (Fig. 5). When considering the demolition and adding a 1.5 m freeboard, the
final downstream height of the raising amounts to 21.5+20 = 41.5 m.
Due to the particular valley shape on the right bank, the heightened arch dam
crest is approximately 15 m higher than the rock foundation. Therefore, there is no
abutment to accommodate the dam thrust forces. To fix this issue, the last dam
blocks on the right bank are designed as a gravity structure with ungrouted vertical
487
Q.104 - R.34
Fig. 6
Downstream elevation of the raised dam. To be noted the right bank featured with
a gravity-type structure (blue area).
Vue aval du barrage surélevé. A noter la rive droite avec une structure de type
gravitaire (zone bleue).
dam joints so that the thrust forces are compelled to plunge into the underlying rock
foundation (Fig. 6).
To close the contour lines of the heightened reservoir, it was necessary to build
a wing wall on the right bank. This gravity-type structure is 15 m high and accom-
modates an ungated free flow spillway. In accordance with the Swiss regulations on
dam safety, the design capacity of this spillway is equal to the maximal pumping flow
value (336 m3 /s), while the 10’000-year return period hydrological flood would have
requested a 47 m3 /s flow capacity only. The plan view of the raised Vieux Emosson
dam is illustrated on Fig. 7.
2.3. MAIN PROJECTS FEATURES
The main features of both Cambambe and Vieux Emosson dam projects are
given in Table 1 below.
3. CHALLENGES
3.1. DAM GEOMETRY
3.1.1. Cambambe Dam
The Cambambe dam is a double curvature arch dam. The geometry of the
heightened Cambambe dam body follows the same shape as per the original project
(Fig. 8), i.e. the three-circle geometric definition.
488
Q.104 - R.34
Fig. 7
Plan view of the raised Vieux Emosson dam.
Vue en plan du barrage Vieux Emosson surélevé.
Table 1
Main Projects Features.
Caractéristiques Principales des Projets
DESCRIPTION UNIT CAMBAMBE DAM VIEUX EMOSSON DAM
Dam Type (Before/After Double curvature arch Single curvature arch-gravity
works) dam/Double curvature arch dam/Double curvature arch
dam dam
Construction (Original/ 1959–1963/2012–2018 1952–1956/2012–2015
Raising)
Dam Height (Before/After m 72/92 55/76.5
works)
Additional Concrete m3 64’000 65’000
volume
Crest elevation masl 112/132 2’205/2’226.5
(before/after works)
Concrete lift height m 2.0 1.5
Annual air temperature ◦C 21/29 –20/20
(min/max)
Reservoir volume hm3 4.5/96 12.1/25.3
(before/after works)
Catchment area km2 115’500 4.5
Average annual discharge m3 /s 750 0.4
Capacity of the flood m3 /s 9’000 336 (largest pumping flow)
evacuation system
Main geological features Siltstone, sandstone and Gneiss
conglomerate
Additional foundation Yes, but only at the banks Local treatment of the existing
treatment grout curtain/Complete grout
curtain under the new dam
portions
489
Q.104 - R.34
Fig. 8
Geometric characterization of the dam body following the original design.
Caractérisation géométrique du corps du barrage suivant la conception originale.
Some adjustments to the original raised dam layout were however necessary,
especially around the orifice spillways at the central blocks and the inclusion of the
lateral spillway, which was not considered in the original design. The inclusion of the
lateral spillway was a major change in the layout, as the original project foresaw, in
addition to the thrust block, an additional arch at the left bank to close the contour
lines and distribute the efforts at the banks, therefore avoiding having them almost
parallel to the contour lines (Fig. 9).
3.1.2. Vieux Emosson Dam
Vieux Emosson dam was not conceived to be raised. The early stage alterna-
tive studies demonstrated quickly that raising the arch-gravity dam while keeping the
arch-gravity shape and corresponding behaviour would lead to a high concrete con-
sumption. In addition, such solutions would request large contact surfaces (joints)
between the old and new concrete, which remain critical during construction and
difficult to handle on the long run (Fig. 10).
The solution retained for the dam raising works consisted therefore in trans-
forming the original arch-gravity dam into a double curvature arch dam (see Fig. 5).
The geometric definition of the arch is adjusted with parabolas.
490
Q.104 - R.34
Fig. 9
Schematic thrust forces from the dam at the left foundation.
Forces schématiques du barrage à la fondation en rive gauche.
Fig. 10
Alternative study for the Vieux Emosson dam raising.
Etude de variantes pour la surélévation du barrage de Vieux Emosson.
Although such solution requested to demolish the upper dam part for geometric
constraints due to the contact between the original and the heightened dam, this
solution was the most economic when considering the reduced contact surface and
the reasonable concrete volume compared to other raising alternatives.
491
Q.104 - R.34
Fig. 11
Top of the mass concrete blocks of the original dam (∼112 masl).
Surface des blocs du barrage original (∼ 112 m d’altitude).
3.2. PREPARATORY WORKS
3.2.1. Cambambe Dam
The main preparatory works for starting the raising of the dam was the demo-
lition of the finishing works at the crest to reach the original mass concrete blocks. It
was finally a relatively easy work (Fig. 11). There were some 0.5 m steps between
the adjacent mass blocks.
The demolition of the guide walls (green zone in Fig. 11) was a little more com-
plicated. A relatively hard cyclopean concrete about 10 m high had to be carefully
removed in order to avoid blocks falling into the reservoir or on the plunge pool.
Before and during the execution of the works, a detailed survey of all existing
galleries around the original dam was conducted. The main objective was to identify
and to plug any future connection between the reservoir and the dam after the raising
of the reservoir.
The preparatory works in the case of Vieux Emosson dam raising were rel-
atively more complicated. It mainly concerned the demolition of the upper part of
the dam, which was necessary to transform the original arch-gravity dam into the
planned double curvature arch dam. The geometric constraints imposed by the
parabola shape of the heightened dam made it necessary to demolish the original
dam by 10 m on the upstream face, respectively 20 m on the downstream face (see
492
Q.104 - R.34
Fig. 12
Tentative demolition works of the original dam concrete by blasting.
Essais de démolition du barrage par explosifs.
Fig. 5). Having a stepped contact surface allowed to reduce the concrete volume
to be demolished as well as the volume of new concrete to be cast for the raising
works.
First, the demolition of the old dam upper part was planned to be performed
by blasting (Fig. 12), however under very strict allowable shaking conditions (peak
particle velocity) to be complied with in order to reduce as much as possible damages
to the remaining part of the dam which will accommodate the raising.
After several different attempts to master the vibration development through

smooth blasting, controlled blasting, reduced volumes, etc. it came out that such
demolition method could not be further implemented. In addition to vibration in
excess, the method itself was found to be dangerous for the staff and material, more-
over it was very difficult to manage the disposal of blasted concrete rubbles. Finally,
the progression rate was also very low causing delays to the general construction
schedule.
The blasting method was given up and substituted using two planers which
allowed faster progress and above all better results in terms of vibration reduction
and reduced damages to the remaining structure. The use of two planers (Fig. 13),
each one working on half of the dam section, allowed reaching a demolition output
at 400 m3 /day.
3.3. CONTACT BETWEEN OLD CONCRETE AND NEW CONCRETE
3.3.1. Cambambe Dam
The contact between the old and new concretes, with different ages and moduli
of elasticity always represents a critical point in the raised structure. There is not
much to do, except mitigate the risks that the formation of cracks could lead to
leakage at the contact of both concretes and potential safety issues. For that reason,
three main actions have been taken in the case of Cambambe dam in order to
guarantee adequate horizontal and vertical contacts as shown in Fig. 14:
493
Q.104 - R.34
Fig. 13
Demolition of the original dam upper part by using planer machines.
Démolition par rabotage de la partie supérieure du barrage original.
Fig. 14
Water stop at the top of the original crest (left) and detail of the design of the
contact between the old and new concrete (right).
Joint d’étancheité installé dans le couronnement du barrage original (à gauche) et
détail du contact entre l’ancien et le nouveau béton (à droite).
Installation of the water stops at the original crest. A small trench with rough
surfaces has been opened close to the upstream part of the crest. The objective
is to guarantee the watertightness in case of formation of cracks between the two
concretes.
Construction of a drainage gallery at the surface of the old crest in order to

follow up, monitor and collect the potential leakages
Installation of two vertical extensometers in order to monitor the behaviour of

the contact.
As for Cambambe, the contact between the original dam and the raising is
critical and was given great design effort. Namely during the operation of the raised
dam this contact surface is subject to high hydraulic pressure in the range of 3 to
494
Q.104 - R.34
Fig. 15
Installation for pressurized water jet to roughen the contact surface.
Installation pour le jet à haute pression et état de la surface de contact verticale.
4 bars. In order as much as possible to guarantee a monolithic behaviour of both

concretes of very different ages (60 years difference), several constructive measures
were taken.
The contact shape is a compromise between geometric constraints to trans-

form the dam from an arch-gravity shape to a double curvature shape, and the
economic interest to keep as much concrete from the original dam as possible.
It was also defined by constructive measures, imposing shaking and vibration

stringent restrictions during demolition works of the old concrete. Such restrictions
could be complied with thanks to the use of planers instead of blasting, allowing
very limited damages to the original concrete accommodating the dam raising.
All contact surfaces were treated with high pressure water jet in order to clean
them and make them rough, to ensure a proper bounding between both old and new
concretes (Fig. 15).
The new concrete layers in direct contact with the old concrete were made
of a special concrete mix with relative smaller aggregates than for a conventional
concrete layer (0-32 mm instead of usual 0-63 mm).
Steel bars were drilled and anchored into the old concrete to guarantee
additional bounding between both concretes (Fig. 16).
Finally, a dam inspection gallery was accommodated at the salient angle

of the preserved original concrete (see Fig. 5). Such layout offers several useful
advantages. For example, it reduces the wedge effect of the old concrete into the
new concrete due to the fact that the old concrete is harder that the new one. The
position of this gallery also allows for good inspection conditions of the contact
between both concretes and a privileged access to specific monitoring devices at
concrete contact.
495
Q.104 - R.34
Fig. 16
Steel bars to ensure proper bounding between both the old and the new concrete.
Barres de liaison pour assurer le contact entre l’ancien et le nouveau béton.
3.4. WEATHER AND HYDROLOGICAL CONDITIONS
3.4.1. Cambambe Dam
While the construction of the arch blocks and the thrust blocks of the Cam-
bambe dam have been relatively easy as it was completely independent of the
weather conditions, the construction of the central blocks (Fig. 17) has been quite
challenging.
The diversion of the river flow was done through the following structures: (i)
Bottom outlet (Qmax = 650 m3 /s), (ii) Cambambe 1 HPP (Qmax = 230 m3 /s, (iii)
desilting orifice of the Cambambe 1 HPP intake (100 m3 /s), (iv) the lateral spillway
(starting operating water levels above 115 masl) and (v) the central blocks (variable
discharge depending on the progress of the construction).
The critical point for the construction is that the blocks are not completely safe
(stable) before the joints are grouted. Depending on the upstream water level, they
may become unstable.
The water level has therefore been maintained as much as possible at 100 masl
(Fig. 18) and therefore the central spillways start overflowing when the discharge
exceeded around 1’000 m3 /s, which can be discharged through the bottom outlet,
the desilting orifice and the Cambambe I HPP.
The discharge of the Cuanza River is mainly regulated by the Capanda

Reservoir, located around 140 km upstream of Cambambe. As there are no major
tributaries joining the river between these two dams, there is no significant increase
between the outflow in Capanda and the inflow in Cambambe.
496
Q.104 - R.34
Fig. 17
Central block with an orifice spillway (left) and cross-section along the block joint
(right) showing the three defined joint grout panels (red, blue and green).
Bloc central avec orifice (à gauche) et section à travers un joint central (à droite)
montrant les trois compartiments d’injection (rouge, bleu et vert).
Fig. 18
Upstream view of the Cambambe dam with the water level at 100 masl.
Vue amont du barrage de Cambambe avec le niveau d’eau à 100 msm.
The discharge of the Cuanza River is reduced during the filling up of the
Capanda Reservoir. Even if a major flood occurs, it will not flow downstream. There-
fore, during the construction of the central blocks, a close follow-up of the filling up
of the reservoir with some inflow scenarios was necessary so any schedule of con-
struction could be adjusted accordingly. Fig. 19 illustrates the Capanda Reservoir
curve with the scenarios chosen in beginning of November 2014 (top image) and at
the end of January 2015 (bottom image).
A detailed action plan has been put in place in case of progressive increase of
the discharge of the Cuanza River and a sudden increase of the reservoir level during
the works (emergency shut down of the Cambambe I HPP for instance). Fortunately,
497
Q.104 - R.34
Fig. 19
Follow-up of the filling up of the Capanda Reservoir under different inflow
scenarios.
Suivi du remplissage du réservoir de Capanda selon différents scénarios de débit
du fleuve.
there were no major events during the construction of the central blocks which could
have jeopardized the safety of the job site and the progress of the works.
In the case of Vieux Emosson Dam, the conditions to raise the dam are com-
pletely different from those prevailing in Cambambe. The reservoir was maintained
empty during the raising works. The natural river flow from the catchment area is so
small that it can be passed through the bottom outlet of the dam. Therefore, strictly
speaking, there was no water management plan to be set up in the specific case.
However, the weather conditions impacted on the raising works in a different

way in Vieux Emosson. Due to the very severe winter conditions existing in the
Swiss Alps at 2’200 masl, with very strong and frequent snow falls (Fig. 20) and
high avalanche risks on site and on the access roads around the dam site, the
job site had to be scheduled over 6 months a year, from April to October. In the
end of October, the job site had to be stopped and the materials and construction
equipment protected and stored from the winter conditions. For safety and logistic
reasons, two long access tunnels (total length approx. 9.5 km) constructed for the
Nant de Drance project from elevation 1’100 masl, guarantee an access to the dam
site at any time of the year.
498
Q.104 - R.34
Fig. 20
Downstream view of the Vieux Emosson Dam under a relative thick snow cover.
Vue aval du barrage de Vieux Emosson sous une importante couverture de neige.
Fig. 21
Upstream view of Cambambe raised dam with the joints to be grouted (red zone).
Vue amont du barrage surélévé de Cambambe avec les joints devant être clavés
(zone rouge).
3.5. DAM JOINT GROUTING
3.5.1. Cambambe Dam
For the raising of the Cambambe dam, it has been considered that the existing
part of the dam, i.e, below elevation ∼112 masl for the arch dam and thrust blocks
and below elevation ∼102 masl for the central part of the dam has already been
grouted (Fig. 21).
In the framework of the raising of Cambambe Dam, three elevations have

been determined as the upper limits of the grouting compartments, i.e. elevations
∼107.0 masl (central orifice blocks only), 119.85 masl and 131.00 masl. These
elevations guarantee the stability of the individual blocks against tilting and sliding
during floods before the joint grouting works.
499
Q.104 - R.34
Fig. 22
Typical dam joint showing the primary and secondary systems for joint grouting.
Joint type du barrage indiquant les systèmes de clavage primaire et secondaire.
As illustrated in Fig. 22, the compartments for the dam joint grouting in Cam-
bambe are limited by water stops. Each compartment is constituted of two systems:
(i) Primary grouting system, formed by lower and upper grooves installed in the
leader blocks and (ii) Secondary grouting system; constituted by manchette tubes
(2” diameter) implemented exactly in the joints. For the raising of Cambambe dam,
the manchettes tubes cross the horizontal water-stops which limit the compartments.
The joints of the central orifice blocks are in the axis of the piers between the
orifices. Different from the other blocks of the dam, the grouting system of these
blocks is divided into 3 compartments (see Fig. 17).
The grouting of the dam joints has been done in 3 stages: (i) primary grouting
of the compartment 0 in June 2015, (ii) primary grouting of the compartments I and
II of the whole raised dam in Mai/June 2016 and (iii) secondary grouting of the whole
raised dam joints in August/September 2019. The average cement consumption for
all joints was around 7.5 kg/m2 in the primary grouting campaign and 1 kg/m2 in the
secondary grouting campaign. These values were considered adequate for the dam
and in line with the results of other dams.
It is important to note that the grouting of the joints was performed with the
reservoir close to the elevation 102 masl, i.e the full supply level of the original dam.
In 2016 the reservoir was filled for the first time and between 2016 and 2019, the
raised dam behaved according to expectations.
500
Q.104 - R.34
Fig. 23
Details of the joint grouting compartments of a typical joint.
Détail des compartiments d’injection d’un joint type.
All vertical joints of the original dam were extended in the raised dam. The
watertightness of the joints was achieved by means of PVC waterstops, and the
connection with the old watertightness system made of cooper strips was particularly
difficult to design. Two grouting compartments were designed for the raised dam,
see Fig. 23. The joint grouting was performed from both dam galleries. The grouting
system had to be accommodated as possible between the many circular shear keys
formed in the joint plane for seismic behaviour improvement.
4. CONCLUSIONS AND LESSONS LEARNED
Both Cambambe and Vieux Emosson arch dams were raised between 2010
and 2020. As it appears from the description and comparison of both projects pre-
sented in this article, both dams see very different site, weather and hydrological
501
Q.104 - R.34
conditions, which made adapting the dam raising design and works to the specific,
local conditions compulsory.
Site conditions include the original dam to be raised, its geometry, shape and
existing concrete and foundation conditions. Weather conditions have impact on the
concrete mix design and the specific concrete curing (pre- and post-cooling), but
also on the dam site access and construction schedule. Hydrological conditions play
a role on the water management in the reservoir, the raised arch dam joint grouting
operation and the first impounding of the reservoir.
All above aspects and characteristics of the project interfere with each other,
making each project unique and to be designed and planned in its own way. There-
fore, very strong understanding, deep knowledge and wide experience of arch dam
technology are a must, spanning from the most theoretical issues of arch dam
behaviour and stress pattern to the practical construction issues such as concrete
mix design, concrete cooling, or joint grouting technology.
Generally speaking, arch dams are structures which can readily be raised
thanks to their 3D behaviour and high degree of redundant constraints, therefore
usually making such raising projects economically interesting. However, at the
same time arch dams are also the most challenging dam type in terms of struc-
tural behaviour, design and construction technologies to be implemented. Therefore,
raising such slender structures should be carefully conceived, designed, planned
and managed. Experience is key in this specific area of dam and hydropower
engineering.
502
Q.104 - R.36
-------
-------
DAM SAFETY ALONG DAM LIFETIME (∗)
Patrice DROZ
Technical director, STUCKY SA
Alexandre WOHNLICH
Head of dam department, STUCKY SA
SWITZERLAND
SUMMARY
Dam safety is not a mere concept to be applied only during the design phase
of dams. Of course, during such initial phase, safety is one of the key elements to
take into consideration. Many countries have developed a general regulatory frame-
work, most of the time based on recommendations provided by the International
Commission on Large Dams (ICOLD), adapted to the local specific conditions and
experience.
Dam safety issues do not stop at the end of the construction and the first
impounding of the reservoir, although the first impounding being one of the most
important phases regarding safety. Then a regular assessment of the safety con-
ditions of the dam is essential for assuring safety all along the life of the dam.
Surveillance and regular testing of safety equipment are of outmost importance.
Each dam being unique, a specific strategy of monitoring must be developed during
the design and adapted later to the effective dam behavior and its particularities.
Along the years, the dam may develop pathologies which may endanger its
safety. Therefore, the surveillance procedures must be adapted, in terms of instru-
mentation and frequency of measures. Only a regular assessment of the dam safety
can detect such an evolution before the situation worsens and gets out of control.
Recent examples of development of alkali-aggregate reaction effects in dams are
∗ Sécurité des barrages tout au long de leur vie
DOI: 10.1201/9781003211808-34 503

Q.104 - R.36
given, showing how the monitoring system has been adapted to the changing con-
ditions of the dams. Adaptation is also necessary when rehabilitation works are
needed.
Major modifications to the dam structure such as strengthening or heightening,

impose evolution of the monitoring system.
Other aspects impacting the safety evolution along the lifetime of dams are
addressed based on series of typical examples: climate change, geological issues,
technological evolution, legal framework evolution, erosion and scouring, etc.
In fine, the application to natural dams of the concept of surveillance and dam
safety continuous assessment is described, based on an outstanding case.
RÉSUMÉ
La sécurité des barrages n’est pas un simple concept à appliquer uniquement

durant la phase de conception des barrages. Bien sûr, durant ces phases initiales, la
sécurité est un des éléments clé à considérer. Beaucoup de pays ont développé un
cadre général réglementaire, la plupart du temps basé sur les recommandations de
la Commission Internationale des Grands Barrages (CIGB), adaptés aux spécificités
locales et à l’expérience.
La problématique de la sécurité des barrages ne s’arrête pas à la fin de

la construction ou lors du premier remplissage, bien que cette étape soit une
phase des plus importantes en termes de sécurité. Par la suite, une évaluation
régulière des conditions de sécurité du barrage est essentielle tout au long de sa vie.
L’auscultation et la vérification régulière du bon fonctionnement des équipements
de sécurité sont de la plus haute importance. Chaque barrage étant unique, une
stratégie d’auscultation propre à chacun doit être développée pendant la conception
et adaptée ensuite au comportement effectif de l’ouvrage et à ses particularités.
Au long des ans, le barrage peut développer des pathologies qui peuvent
mettre en danger sa sécurité. Ainsi, les procédures d’auscultations doivent être
adaptées, en termes d’instrumentation et de fréquence des mesures. Seule une
évaluation régulière de la sécurité permet de détecter certaines évolutions avant que
la situation n’empire et devienne hors contrôle. De récents exemples de développe-
ment de réaction alcali-agrégats dans des barrages illustrent comment le système
d’auscultation a été adapté. De telles adaptations sont également nécessaires
quand des travaux de réhabilitations sont engagés.
Des modifications majeures au barrage telles que des renforcements ou des

surélévations imposent également une évolution du système d’auscultation.
504
Q.104 - R.36
D’autres aspects impactant la sécurité des barrages au long de leur vie sont
également mentionnés et illustrés par des exemples: changements climatiques,
géologie, évolution technique, cadre règlementaire, érosion et affouillements, etc.
In fine, l’application du concept de surveillance et d’évaluation régulière de

la sécurité des barrages est étendu aux barrages naturels par le biais d’un cas
particulièrement remarquable.
1. INTRODUCTION
Dams are among the largest constructed structures made for ensuring water
availability for agriculture, domestic and industrial usages, for ensuring power and
navigation and for protecting land from flood disasters.
Dam safety concepts apply during the design of the dam and its appurtenant
structures: safety is one of the key elements to take into consideration. Many coun-
tries have developed a general regulatory framework, most of the time based on
recommendations provided by the International Committee on Large Dams (ICOLD)
adapted to the local specific conditions and experience. Hence, flood design may
differ from one country to another, depending on the hydrological context and on the
level of knowledge and confidence of the available hydrological data.
But once constructed, after the first impounding, are dams safe forever? What
are the challenges to ensure safety all along the lifetime of the dam until its eventual
decommissioning? Past events, errors, rehabilitation works, and dramatic failures
are sources to help the designer, the supervisory authority, the dam owner and
operator to improve the safety of dams all along their lifetime.
2. SAFETY FOREVER
In 1995, the ICOLD published a bulletin summarizing statistics analysis of past

failure of dams [1]. More recently, a detailed study analysed the failure of several
dams taking into account the age of the dam [2]: a database of dam safety incidents
and failures was developed with an initial focus on determining how old dams were
when they suffered a failure or safety related incident. The database had records of
more than 4000 incidents and failures.
In this study, as illustrated in Fig. 1, it appears that, among the dams for which
the age at failure could be identified, 31% of the failures or incidents occurred during
the 5 first years after construction.
505
Q.104 - R.36
Fig. 1
Number of failure and incidents with respect to the age of the dams [2].
Nombre de ruptures ou d’incidents en fonction de l’âge des barrages [2].
Of course, the study does not consider that dam design and technology
improved over years as well as the legal framework, taking advantages of these
past failures to improve safety of the structures. Nevertheless, the first years of
operation of a dam appear to be the most dangerous phase. Beyond 5 years of
operation, significant percentage show delayed safety problem even after 50 years
of operation. In other words, a lengthy apparent successful dam operation does not
preclude failure over the remaining lifetime.
A conceptual representation (Fig. 2) has been proposed to explain the above

statistical results [3] For well-designed dams surviving the first years of operation, a
“global” safety factor (SF) well over 1 can characterize the safety of the dam. With
time, due to various degradation causes which are described later, the safety factor
tends naturally to decrease.
The first 5 years may be compared to infant illnesses for which a attention
must be devoted in terms of surveillance. Then the adult period with no major trou-
bles leads to the aging time during which an acute and regular attention is again
requested. The danger appears to be the transition between the “normal” behavior
years and the beginning of the decay, when trends of abnormal behavior lead to a
degradation of safety. This transition is critical for the dam since:
• the transition time is not known

• the aging factors are unknown
• as well as the severity of the aging process
506
Q.104 - R.36
Fig. 2
Conceptual degradation of the “global” safety factor with time [3].
Décroissance conceptuelle du facteur de sécurité “global” [3].
Of course, this conceptual vision may be perturbed by external events such

as a major flood or an earthquake during which the safety factor may decrease
suddenly. Alternatively, rehabilitation measures or lowering the operating reservoir
level would increase the “global” safety factor. This conceptual approach must be
understood for the dam as a whole system, integrating the structure itself but also
the appurtenant structures, the foundations, the hydromechanical equipment and
even the reservoir banks.
3. DESIGN AND CONSTRUCTION
Design standards and investigations means are constantly improving. Interna-

tional organizations play a major role in this respect, and the regulatory frameworks,
when existing are also evolving, taking also into account the specificity of the country:
hydrological regime and measurement history, seismic conditions, etc.
Nevertheless, errors in design, partial or poor investigations or poor con-

struction surveillance may impact the operation and the safety of the dam or its
appurtenant structures all along its lifetime. For instance, due to limited geological
investigations at several dam projects the reservoir could never be fully operational
due to leakage. In some other cases, continuous grouting works are required to
ensure the safety of the dam such as Mosul Dam in Irak.
An inappropriate design or execution flaws may stay latent for years until an
external event triggers the failure. Errors in the geological investigations and in
the design of the Oroville spillway chute were pointed out by the Forensic Team
507
Q.104 - R.36
Fig. 3
Failure of the Oroville spillway chute; regressive erosion of the emergency
spillway [5].
Destruction du coursier de l’évacuateur de crue du barrage d’Oroville; érosion
régressive de l’évacuateur de secours [5].
nominated after the incident of February 2017 [4]. The foundations of the chute as
well as the drainage system adopted are the root causes of the event. In addition,
the surveillance and the analysis of the early signs of degradation and deterioration
have been neglected leading to the destruction of the spillway chute. Furthermore,
the emergency spillway was not fully designed for preventing regressive erosion
(Fig. 3).
During the construction phase, the river diversion system and its safety must
be the object of particular attention as illustrated by the incident of the Ituango Dam in
Colombia in Spring 2018: the diversion tunnel was blocked by a landslide leading to
an early uncontrolled impounding as the dam and its spillway were not achieved yet.
4. EVOLUTION OF EXTERNAL CONDITIONS
4.1. HYDROLOGY
During the design period, safety floods and extreme floods scenarios have
been selected. Poor or erroneous data, misinterpretation or appearance of excep-
tional events due to climate change may lead to damages and collapse of the dam.
During the construction of the Karun IV dam in Iran, the upstream cofferdam,
a 40 m high clay core embankment, collapsed after overtopping in 2005 due to a
flood corresponding to a 500 years return period (the diversion system had been
designed for a 20 years return period flood). After reconstruction, the same event
appeared 2 years later destroying again part of the cofferdam (Fig. 4) and damaging
the construction site.
508
Q.104 - R.36
Fig. 4
General view of the Karun IV second cofferdam overtopping.
Vue générale de la surverse du second batardeau du barrage de Karun IV.
4.2. SEISMIC EVENTS
The literature is full of examples of heavy damages to dams and even collapses
due to earthquake. Many safety guidelines impose nowadays to perform a specific
seismic site evaluation for new dams or for assessing the safety of existing dams.
5. AGING
Along the years, the dam may develop particular pathologies which may
endanger its safety. Therefore, the surveillance procedures must be adapted, in
terms of instrumentations and frequency. Only a regular assessment of the safety of
the dam can detect such an evolution before the situation degenerates and enters
into a dangerous zone.
5.1. ALKALI AGGREGATE REACTION
The Salanfe dam is a 52 m high concrete gravity dam located in the Swiss
Alps. The dam was built in 1952. Since early 1970’s, monitoring of the behavior
of the dam based on pendulums started to present an irremediable deformation
towards upstream (Fig. 5, Fig. 6 and Fig. 7), in particular at the most pronounced
angle between the left wing and the central part. After in-depth studies along the
years, investigations proved that the dam was affected by alkali-aggregate reaction
(AAR) inducing swelling of the concrete mass [6].
509
Q.104 - R.36
Fig. 5
General view of Salanfe dam.
Vue Générale du barrage de Salanfe.
Fig. 6
Pendulum displacement of block 13 versus time.
Déplacement du pendule du bloc 13 en fonction du temps.
Extensive inspections and specific extension of the monitoring system have

been decided as well as a restriction in reservoir operation before the rehabilitation
works took place in 2012 and 2013. The rehabilitation works consisted in the exe-
cution of 22 vertical 11 mm slot cuts, maximum 22 m high, in order to release the
stresses in the structure. Specific numerical modelling, integrating the AAR effects
in the concrete constitutive law was used to define the strategy of intervention (depth
of the slots, location, order of execution) [7, 8].
510
Q.104 - R.36
Fig. 7
Geodetic displacements along the crest since 1993 (one square represents 50 m
or 5 mm).
Déplacement mesuré par géodésie sur le couronnement depuis 1993
(un carré représente 50 m ou 5 mm).
The monitoring system, in particular the pendulums, was modified in prevision

of the works in order to follow more accurately the movements of the dam during
the sawing of the slots. The pendulums have been modified in such a way to give
information on displacements along the shaft of each one [6].
Since the beginning of operation of the Inga 1 dam in 1972 and Inga 2 dam
in 1982 in DR of Congo (Fig. 8), the two structures, located close to the mouth
of the Congo River, presented a continuous and regular displacement towards
downstream.
Fig. 8
General views of Inga 1 and Inga 2.
Vues Générales des barrages de Inga 1 et Inga 2.
511
Q.104 - R.36
In the framework of the rehabilitation of the monitoring system of the dams, a

satellite data analysis has been performed in order to have a better understanding of
the movements of the dam and identify possible movements of the foundation. The
SAR-interferometry technology was used [10] . As the technique allows the analysis
of objects like buildings, dams etc., it is possible to check also for geologic causes
of deformation such as settlements or landslides (Fig. 9).
Fig. 9
Displacements of points along the line of sight of the satellite (mm/year).
Déplacements de points le long de la ligne de visée du satellite (mm/an).
The results of the SAR-interferometry contributed to the identification of the

causes of the deformations, pointing out not only horizontal displacements that
had been identified with the conventional survey measurements, but also upwards
vertical movements.
In-depth analysis of concrete samples has been performed showing the devel-
opment of Alkali-Aggregate Reaction. The improved monitoring system takes into
account the development of this pathology which must be followed regularly although
presently and in the near future, it does not present any safety issues.
In some specific cases, the swelling of the concrete may also lead to difficulties
and even blockage of the hydromechanical equipment leading to dramatic safety
issues if the spillways gates cannot be operated. The evolution of the AAR in the
Kariba dam, located at the border of Zambia and Zimbabwe, leads to rehabilitation
works which are undertaken at present.
512
Q.104 - R.36
5.2. FOUNDATION
Not only the evolution of the dam structures must be followed carefully but also
the foundations. Slow modifications may appear in time due to dissolution or creep
effect under the constant stresses induced by the dam and the reservoir. This was
the case for instance for Les Toules dam in the Swiss Alps [9, 11].
The dam is 86 m high and the crest across the valley is 460 m long. The
concrete volume is 235’000 m3. The ratio between crest length and dam height is
5.35, which is far beyond common values for double curvature arch dams. The dam
presents a particular design, with a very slender shape, a high vertical curvature
toward downstream and no shear keys.
The dam foundation lies on gneiss and mica-schist rocks, forming alternate
subvertical strips almost parallel to the valley. The gneiss appears to have better
mechanical properties and to be more rigid than the mica-schist; thus the dam rests
on an alternate series of rock formations which have variable properties and stiff-
ness, especially on the left bank in the vicinity of dam blocs 5 to 7, where small
irreversible displacements toward downstream were observed since the first reser-
voir impounding. This displacement amounted to 4 mm after 30 years. A local
stability analysis was performed and showed, first of all, that the required factor
of safety was barely ensured in the static case and, secondly, in case of a seismic
event, the factor of safety dropped dramatically. The constant increase with time of
this displacement as well as the stability analysis performed and an update of the
legal framework regarding seismic resistance called for remedial works of the dam
as presented in the figure below, undertaken in the years 2008 to 2011 (Figs. 10
and 11).
Fig. 10
Downstream face of Les Toules dam with strengthening solution.
Parement aval du barrage des Toules avec les solutions de renforcement.
513
Q.104 - R.36
Fig. 11
Cross-sections of Les Toules dam with strengthening solution Left: Bloc 21 with
abutment thickening; Center: Shear keys at the crest; Right: Bloc 6 with abutment
thickening and a 10 m deepening into the rock foundation.
Coupes verticales du barrage des Toules avec les solutions de renforcement.
Gauche: bloc 21 épaissis; Centre: clés de cisaillement en crête; Droite: bloc 6
épaissis et approfondissement de la fondation de 10m.
Fig. 12
General view of the Nurek dam with the two spillway tailrace tunnels in operation
(photo Wikipedia).
Vue Générale du barrage de Nurek montrant les deux tunnels d’évacuation des
crues en activité.
Evolution of foundation of the dam itself is not the only geological processes
which may affect the safety of the dam. The surface spillway and low-level spillway
tailrace tunnels of the Nurek dam are located on the left abutment of the 300 m high
embankment dam in Tajikistan (Fig. 12).
514
Q.104 - R.36
Tectonic faults cross the tailrace tunnels which are significantly deteriorated
by the geological movement. With time, the safety of the tunnels is compromised,
and remedial works are foreseen in the near future (Fig. 13).
Fig. 13
Degradation of the Nurek dam surface spillway tailrace tunnel due to the crossing
of an active fault.
Endommagement du tunnel de l’évacuateur de crue de surface du barrage de
Nurek provoqué par une faille active le traversant.
Scouring downstream the dam because of frequent use of the spillway may
be a source of concern for the safety of the structure. This is the case for the Kariba
dam at the border between Zambia and Zimbabwe.
The Kariba dam (Fig. 14) is a 128 m high arch dam constructed in the late
1950’s. The operation of the surface spillways induced a major scouring at the
Fig. 14
General view of the Kariba dam.
Vue Générale du barrage de Kariba.
515
Q.104 - R.36
foot of the dam, digging a 90 m deep scouring hole some tens of meters down-
stream from the toe of the dam. This situation endangers the safety of the dam and
leads to presently on-going rehabilitation works. In parallel, rehabilitations due to
the development of AAR are undertaken.
The plunge pool rehabilitation aims at re-profiling the shape of the pool in
order to prevent further regressive erosion. Laboratory tests (Fig. 15) have been
undertaken in order to optimize the shape of the pool in order to avoid further
degradation. The reshaping of the plunge pool is presently underway.
Fig. 15
Spillway operating at Kariba; Model test at EPFL, Lausanne which enabled to
define the optimal shape of the plunge pool.
Evacuateur de crue de Kariba en fonctionnement ; Modèle physique à l’EPFL,
Lausanne, qui a permis de définir la forme optimale du bassin amortisseur.
5.3. OTHER ASPECTS OF FOUNDATION AGING
Another aspect of aging of the foundations is related to the deterioration of the

grout curtain. In certain geological circumstances, dissolution of the gout curtain
may occur, leading to a leakage increase and possible increase of uplift pressure.
A chemical analysis of the leakage water compared to the quality of the reservoir
water may give information on a potential leaching of the grout curtain as well as the
analysis of the evolution of the leakage discharges with respect to the reservoir level.
Of course, creep deformations may lead to cracking in concrete dams, or

arching effect in embankment. Adequate monitoring is therefore necessary.
5.4. AGING OF MONITORING SYSTEM
Safety of dams relies on three major aspects:
• design and execution

• surveillance and monitoring
• emergency preparedness plan
516
Q.104 - R.36
In order to ensure the safety of the dam along the years, it is essential to main-
tain the monitoring system which serves as a base for surveillance. Maintenance
of the monitoring system becomes as important as the structure itself for catching
and following up any trend or deviation of parameters from the range it was foreseen
during the design phase.
The pictures of Fig. 16 give examples of poor or absence of maintenance of

the monitoring system of a large dam.
Fig. 16
Foundation of a geodetic pillar maintained by cables; undersized gauging station;
out of order manometers.
Fondations d’un pilier géodésique de référence; seuil de jaugeage
sous-dimensionné; manomètres inutilisables.
Beside maintenance, the monitoring system must be adapted, if needed, to the

behavior of the dam. Only in-depth analysis of the monitoring data, cross-checking
of the evolution of various parameters and critical analysis will lead to a reliable
safety assessment.
6. MAJOR MODIFICATIONS
6.1. STRUCTURAL MODIFICATIONS
Major modifications to the dam structure impose a full revision and updating
of the safety of the dam and an appropriate evolution of the monitoring system. This
is particularly the case for dam heightening. The Vieux Emosson is located in the
Swiss Alps, close to the French border. The dam was built in the years 1952-1956. It
is located upstream the Emosson dam which was built later, in the years 1969-1974.
The original Vieux-Emosson is an arch-gravity dam shown on Fig. 17
517
Q.104 - R.36
Fig. 17
Vieux-Emosson and Emosson reservoir.
Réservoirs de Vieux-Emosson et Emosson.
The original dam (Fig. 18) had the following characteristics: 55 m high, a 170 m
long crest, a downstream face slope from 10:4 at the center to 10:8 close to the
abutments and 62’000 m3 of concrete.
The proximity of the two Emosson reservoirs enabled the development of a

pump-storage scheme called Nant de Drance. During the first stages of studies,
a 600 MW pump storage plant was considered and no modifications to Vieux-
Emosson were necessary but further studies showed the advantages of increasing
the installed capacity to 900 MW. As a result, the volume of the upper reservoir had
to be increased from 11.2 to 24.6 Mm3 . Therefore, the dam had to be heightened
by 21.5 m [12].
Fig. 18
Vieux-Emosson dam before and after heightening.
Le barrage de Vieux-Emosson avant et après surélévation.
518
Q.104 - R.36
Since the heightening of the dam had to be founded on the original dam, the
resulting geometry was not only dictated by the shape of the valley but also by the
shape of the original dam. After adjustment, at the end of the process, the dam type
was changed from an arch gravity to a double curvature arch dam (Fig. 18).
Two horizontal galleries were built, one at the interfaces between original and
new concrete and one in the new concrete (Fig. 19). These galleries can be accessed
from the tunnel network of the pump-storage complex.
Fig. 19
Cross sections showing the construction phasing of the heightening.
Coupes montrant les phases de construction de la surélévation.
The execution of the works started with the demolition of the upper part of the
original dam in order to receive the double arch curvature shape on the existing arch
gravity part.
The cleanness and roughness of the contact surfaces was very important. A
first concrete (0-32 mm) was first poured and then mass concrete (0-63 mm). In
order to assure a perfect bonding, shear dowels every 1.5 to 1 m were also installed
at the interface and shear keys in the vertical joints.
A horizontal gallery has been managed at the edge of the step for reducing
wedge effect. Furthermore, this gallery enables to monitor the inner behavior of the
contact and makes possible future interventions if needed.
Regarding monitoring, the original monitoring system of the dam has been fully
revised and extended to meet the requirements of the new dam. New pendulums
and inverted pendulums have been installed and extended. Thermometers in the
519
Q.104 - R.36
new concrete and rockmeters in the foundations of the closing gravity elements have
been placed.
As a conclusion, the modification of the Vieux Emosson dam was not only in
the heightening of the dam but led also to the modification of the type of the dam,
from an arch gravity to a double curvature arch dam.
6.2. OPERATIONAL MODIFICATIONS
Most of the dams built in the Swiss Alps have been built to store water during
spring and summer while the snow cover and the glaciers melt and to produce
energy in late autumn and winter. But the demand in electricity evolved during the
last 10 years and due in particular to the photovoltaic and solar energy produced
in Western Europe. The mode of operation of the reservoirs has changed implying
among others modifications of the exposure of the upstream face of concrete dams
to solar radiation and temperature. This modification of the thermal regime of the
structure has to be taken into account in the analysis of the dam behaviour during the
regular safety assessment analysis. In some cases, this modification is the cause
of cracking, also due to the slight increase in temperature observed as a result of
climate change.
In addition, human errors can also be the sources of dramatic consequences,

as illustrated by the Taum Sauk upper reservoir collapse in 2005 in the USA, or by
inappropriate operation or maintenance of hydromechanical equipment.
7. OTHER ASPECTS
Many other aspects may impact the safety evolution along the lifetime of a
dam.
Climate change may impact the dam safety through for instance, temperature
increase and cracking or hydrological extreme events. In many regions, the melting
of glaciers let uncovered soft erodible material which was never exposed to precipita-
tion and run-off. This results in an increase of sediment depositing into the reservoir.
Sediment deposit may block the bottom outlet and impose additional destabilizing
forces.
In the same order of idea, the evolution of climate change and of land use
planning may lead to hydrological modifications. This implies to review regularly the
hydrological conditions in order to check the capacity of the spillway of the dams.
520
Q.104 - R.36
Technological improvements followed by legal framework evolution lead to

more realistic evaluation of the safety of dams. This was the case with the progresses
made in earthquake engineering and dynamic analysis. And based on this evolution,
all the Swiss dams had to be reassessed recently in order to match with the modern
tools of analysis [13]. Some dams had to be reinforced after this reassessment.
8. NATURAL DAMS
What is true for man-made and designed dams is also true for natural dams,
in particular because the evolution of their safety status is unknown.
Lake Sarez in Tajikistan was formed in 1911 when a massive earthquake-

triggered landslide buried the village of Usoy under a 650 m high obstruction which
dammed the Murghab River (Fig. 20). The resulting 60 km long lake containing over
17 km3 of water is located behind the so-called Usoy natural dam in the Pamir range,
the highest dam in the world.
Although the safety of the lake has been studied over many years, significant
gaps and inconsistencies existed in the available data. Analyzing the data in the
light of a risk analysis approach was the first action in the framework of the Lake
Sarez Risk Mitigation Project (LSRMP) [14].
Fig. 20
Views of Lake Sarez, from upstream and from downstream - the white arrows
show the springs flowing from the downstream face of the natural embankment.
Vues du Lac Sarez depuis l’amont et depuis l’aval - les flèches blanches indiquent
les sources situées sur le parement aval de la digue naturelle.
The design and the implementation of a monitoring system were based on the
results of the risk analysis. The aim was to follow up the long-term evolution of the
521
Q.104 - R.36
natural dam and its surroundings and to couple the monitoring to an early warning
system (Fig. 21) installed in 17 villages downstream in order to mitigate the risk.
Fig. 21
General layout of Monitoring and Early Warning Systems; dam house with weather
station.
Disposition générale du système de surveillance et d’alerte; bâtiment du barrage
et station météorologique.
At present, the yearly inflows are almost equivalent to the yearly outflows.
Hence, it is very important to follow up the evolution of the lake level with respect to
the spring discharge downstream in order to detect any modification in the normal
leakage regime through the dam.
9. CONCLUSION
Dam safety is not a concept to be applied only during the design phase of
the dam. Of course, during design, safety is one of the key elements to take into
consideration.
Dam safety issues do not stop at the end of the construction and the first
impounding of the reservoir, although the first impounding being one of the most
important phases regarding safety. A regular and continuous assessment of the
safety condition of the dam is essential for assuring safety all along the life of the
dam. Surveillance and regular testing of safety equipment is of outmost importance.
Each dam being unique a specific strategy of monitoring must be developed during
the design and adapted to the behavior and the particularities the dam may develop.
522
Q.104 - R.36
REFERENCES
[1] Dam Failures – Statistical Analysis, Bulletin 99, ICOLD, 1995.
[2] P. J. REGAN, “An Examination of Dam Failure vs Age of Dams”, Hydroworld,

vol 29/4, US Federal Energy Regulatory Commission (FERC), 2010.
[3] PH. MÉAN, P. DROZ, “Improving dam safety with a surveillance self-
assessment toolkit”, Hydropower & Dams, Issue 3, 2019.
[4] FERC - DWR (2017), Independent Forensic Review Team for Oroville Spillway,
January 2018, presented during the 2018 ICOLD Congress, Vienna, 2018.
[5] CALIFORNIA DEPARTMENT OF WATER RESOURCE/Wikipedia.
[6] P. DROZ, O. VALLOTTON, T. MENOUILLARD, R. LEROY, “Monitoring of an

AAR affected dam”, ICOLD, Q99-R33, Stavanger, June 2015.
[7] P. DROZ, J.-F. SEIGNOL, R. LEROY, L. BOLDEA, “Strategy for the reha-
bilitation of Salanfe dam”, 14éme Conference Internationale sur la réaction
alkali-agrégats (ICAAR), Austin, Texas, Mai 2012.
[8] P. DROZ, O. VALLOTTON, T. MENOUILLARD, R. LEROY, “Slot cutting an

AAR-affected dam: case study of the Salanfe dam, Hydropower & Dams, Issue
5, 2013.
[9] O. MÜLLER, A. WOHNLICH, “Safety Enhancement and Strengthening of Les

Toules Arch Dam”, ICOLD European Club Symposium, Innsbruck, 2010.
[10] M. MARES, P. DROZ, “Deformationsmonitoring der Staudämme Inga 1 und

Inga 2 mittels satellitengestützter SAR-Interferometrie”, 15. Geokinematischer
Tag, Freiberg, Germany, 2014.
[11] A. WOHNLICH, O. MÜLLER, “Strengthening of Les Toules arch dam”, ICOLD,

Lucerne, 2010.
[12] O. VALLOTTON, “Surélévation du barrage de Vieux-Emosson”, Q99-R22,

ICOLD, Stavanger, 2015.
[13] R. PANDURI, P. DROZ & AL., “Ongoing seismic safety assessment of Swiss
dams”, Q93, ICOLD, Kyoto, 2012.
[14] P. DROZ, L. SPACIC-GRIL, “Lake Sarez risk Mitigation Project: a Global Risk
Approach”, Q86-R75, ICOLD, Barcelona, 2006.
523
Q.104 - R.38
-------
-------
MONITORING PERFORMANCE OF CONCRETE DAM AFTER EARTHQUAKE (∗)
Anom PRASETIO
Manager of Civil & Hydro Surveillance, PT. VALE INDONESIA, Sorowako
Abdul MUIS
Senior Civil Engineer, PT. VALE INDONESIA, Sorowako
INDONESIA
SUMMARY
The territory of the Republic of Indonesia is at the confluence of 2 main earth-

quake lanes, namely the Circum Pacific earthquake route and the Transasiatic Alpine
earthquake route. Sorowako itself is an area prone to earthquakes due to the pres-
ence of the Matano fault and the Palu koro fault which can have an impact on the
failure of building structures in the surrounding area. The last earthquake was quite
strong, measuring 4.9 SR on September 21 2020 at 18:26:32 WITA with an earth-
quake center 5 km southeast of East Luwu with a depth of 10 km and the distance
from the epicenter to the Balambano and Karebbe Dam is 4 km. To minimize the
risk of dam failure, intensive monitoring is necessary anywhere through data and
visualization and in real time. From the results of the measurement, we can see that
the data and visual information are obtained in real time and comprehensive condi-
tions, which can be accounted for scientifically as well as by manual measurement.
Rapid response after an earthquake can be followed up immediately if abnormal
conditions are found.
∗ Surveillance des performances d’un barrage en béton après un tremblement de terre
524 DOI: 10.1201/9781003211808-35

Q.104 - R.38
RÉSUMÉ
Le territoire de la République d’Indonésie est à la confluence de 2 principales

lignes sismiques, à savoir la ceinture sismique Circum Pacific et la ceinture transasi-
atique alpin. Sorowako elle-même est une zone sujette aux tremblements de terre en
raison de la présence de la faille Matano et de la faille Palu koro qui peuvent avoir un
impact sur la défaillance des structures des bâtiments dans les environs. Le dernier
tremblement de terre a été assez fort, (4,9 sur l’échelle de Richter) le 21 septembre
2020 à 18:26:32 WITA avec un centre de tremblement de terre à 5 km au sud-est
de Luwu oriental avec une profondeur de 10 km et la distance de l’épicentre au
barrage de Balambano et Karebbe est 4 km. Afin de minimiser le risque de rupture
de barrage, une surveillance intensive est nécessaire partout grâce aux données et
à la visualisation et en temps réel. À partir des résultats de la mesure, nous pouvons
voir que les données et les informations visuelles sont obtenues en temps réel et
dans des conditions complètes, qui peuvent être expliquées scientifiquement ainsi
que par une mesure manuelle. Une réponse rapide après un tremblement de terre
peut être suivie immédiatement si des conditions anormales sont trouvées.
1. BACKGROUND
PT Vale Indonesia Tbk, formerly known as PT Inco, is a mineral mining com-

pany that produces nickel as its main product. PT Vale Indonesia Indonesia operates
in East Luwu Regency, South Sulawesi. In this location there are The Larona River
Basin which cover an area of 2,477 km2 , including the three cascading lakes:
Matano, Mahalona and Towuti Lakes, is a strategic watershed which acts as the
water resource for three hydroelectric power plants which supply electricity needs
for PT Vale Indonesia Tbk’s nickel mining operation and its surrounding areas1. The
total installed power in the Hydroelectrics is 420 MW, consisting of a 3 × 60 MW
from the Larona power plant, 2 × 65 MW from the Balambano power plant and a
2 × 55 MW from the Karebbe power plant, all of which are in cascade systems.
The three hydropower plants (Fig. 1) operate by utilizing the water resources
from the mentioned cascading Lakes in the Larona River Basin channelled through
the Larona River to the plants’ turbines. The drop of the Lakes water volume will
affect the optimal operation of the hydropower to efficiently support PT Vale Indone-
sia’s nickel ore smelting operations in Sorowako. Lower than average rainfall has
significantly impacted the Matano and Towuti Lake’s water. Minimum elevation limit
of Towuti Lake for optimal operation of the Hydroelectrics is 318.00 meters. The drop
of the water level in the Towuti Lake will disrupt the capacity of the three hydropower
to serve the need of nickel production processes. In turn, PT Vale Indonesia will need
to operate alternative electricity sources, like diesel generators and other sources
which drive greater costs.
The Balambano Dam was built on the Larona river, which is located in
Balambano Village, Nuha District, East Luwu Regency, South Sulawesi Province.
The location of the Balambano Dam with the main road Sorowako leading to the
525
Q.104 - R.38
Fig. 1
Dam location.
Emplacement du barrage.
port of Balantang. The Larona River originates from 3 (three) natural lakes, namely
Matano Lake, Mahalona Lake and Towuti Lake which are the inlet of the Larona River.
Along the Larona river, series dams have been constructed in the following order:
a. The Batubesi Dam was operated starting in 1978

b. The Balambano Dam was operated from 2000
c. Karebbe Dam which was operated starting in 2012
The Balambano Dam is a Concrete Gravity Dam that uses Roller Compacted
Concrete material, so the Balambano dam is an RCC Dam, while the RCC material
used is "Lean Mix RCC". Karebbe Dam was built in the Larona River segment, about
two 2 km from the upstream of Karrebe Village, Laskap Village, Malili District, East
Luwu Regency, South Sulawesi Province. The Karebbe Dam is a series of cascades
of two previously constructed dams, the Batubesi Dam and the Balambano Dam
(Fig. 2). The Balambano Dam is located about 5 km from the upstream of the
Karebbe Dam, while the Batubesi Dam is located about 15 km from the upstream
of the Balambano Dam or about 20 km from the Karebbe Dam. The Karebbe Dam
was originally designed as Roller Compacted Concrete (RCC) with a spillway in
the middle of the dam. The power plant will be built within the dam body under
the spillway. The 66 m high Karebbe Dam is measured from the deepest base
foundation. For power generation, two (2) turbine generators are installed with an
installed capacity of 2 x 65 MW. The two series dams were built and managed by
PT. Vale Indonesia with a single function as a power plant used for the operation of
a nickel smelting plant.
PT. Vale Indonesia has a hazard map depicting the impacted area, when the
dam collapses due to earthquakes, floods, landslides and so on. Various attempts to
find out that the dam performance is still good and according to the dam safety pro-
cedures, including operation and maintenance of the dam instrumentation must be
carried out according to the procedure, inviting consultants to check the actual con-
dition of the dam and reporting the results to the dam safety commission, operation,
526
Q.104 - R.38
Fig. 2
Cascade system of three lakes and three dams along Larona River.
Système de cascade de trois lacs et de trois barrages le long de la rivière Larona.
Fig. 3
PT. VI hydropower dam failure mind mapping.
PT. VI Cartographie mentale des défaillances de barrages hydroélectriques.
dissemination of emergency action plans including their simulation involving the

government and the community around the dam through a mind mapping (Fig. 3).
2. INTRODUCTION
The territory of the Republic of Indonesia is in a collision of three plates in the

earth’s crust, namely the Eurasian plate, the Pacific plate, and the India-Australia
plate and if we look geologically, the territory of the Republic of Indonesia is at
the confluence of 2 main earthquake lanes, namely the Circum Pacific earthquake
route and the Transasiatic Alpine earthquake route. By looking at this condition, the
527
Q.104 - R.38
movements of the earth’s surface originating from volcanic activity and earthquakes
are very common occurrences. Sorowako itself is an area prone to earthquakes
due to the presence of the Matano fault and the Palu koro fault which can have
an impact on the failure of building structures in the surrounding area. The last
earthquake was quite strong, measuring 4.9 SR on September 21 2020 at 18:26:32
WITA with an earthquake center 5 km southeast of East Luwu with a depth of 10 km
and the distance from the epicenter to the Balambano and Karebbe Dam is 4 km .
This requires intensive monitoring so that the function of the infrastructure can be
maintained.
Fig. 4
Earthquake 4.9 Richter Scale.
Échelle de Richter du séisme 4.9.
The need for a fast and routine dam performance monitoring system will help
the dam owner know the specific condition of the dam and the provincial govern-
ment as the owner of the area. Visual monitoring, manual instrumentation data
retrieval and real time data retrieval using three-dimensional and one-dimensional
sensors for both static and dynamic movements for dam deformations. To ensure
the condition of the equipment remains fit for operation, the hydro utilities team as
the manager of the Larona, Balambano and Karebbe hydropower plants monitored
observations and collected data on instrumentation equipment, especially on dams
and other supporting buildings in the entire Dam area and its surroundings. This
report provides a brief overview of these conditions represented by data taken after
the earthquake and depicted in graphs. From the observational data there is no
instrumentation that has increased significantly in the Dam area.
The purpose of this report is monitoring or monitoring so that disaster preven-

tion can be anticipated due to deformation due to horizontal and vertical movement
of dams such as earthquakes, landslides, leaks, and so on, so that anticipation
of dam failure can be done immediately as a way of mitigating or evacuating the
community in downstream Regulation of the Minister of Public Works and Public
Housing of the Republic of Indonesia Number 27 / PRT / M / 2015 [1] concerning
Dams CHAPTER VII Documentation and Information Article 158:
528
Q.104 - R.38
Dam managers must submit periodic reports regarding information on the

condition of the dam and its reservoir to the relevant agencies
Information on the condition of the dam and its reservoir as referred to in

paragraph (1), includes:
• Structural and operational behavior

• The results of the reading of the instrument and its interpretation, the results
of the inspection and safety evaluation
• Change or rehabilitation
• Dam safety-related incidents and extraordinary events
• The water condition of the allocation reservoir includes water
Report Submission:
• If the earthquake that occurs is below <4 SR, a report will be sent to the Lutim
Regency Government
• If the earthquake that occurs is> 4 SR then a report will be sent to BKB Jakarta,
BBWS Pompengan Jeneberang and the Lutim Regency Government
3. SURVEILLANCE
Surveillance activities as a frontline team in monitoring all concrete dam instru-

mentation are expected to respond and analyze in the event of an earthquake caused
by shifting faults in the Sorowako area and its surroundings. The formation of this
team is in accordance with Government Regulation No. 72 / PRT 1997 concern-
ing Dams [2] which consists of routine monitoring, periodic inspections, dam safety
inspections and special inspections related to dangerous conditions. Equipment
testing will be carried out during the inspection period.
3.1. ROUTINE MONITORING
The purpose of this monitoring is to detect early hazardous conditions that

threaten the safety of the dam which is carried out every week for 3 days.
Monitoring activities include maintenance in accordance with checklists, reading
instrumentation tools, plotting results and making reports to the dam owner.
3.2. PERIODIC CHECKS
Ensure routine monitoring work is appropriate and can detect hazardous con-
ditions as early as possible including: checking stored data, checking monitoring
529
Q.104 - R.38
manual procedures and emergency action plans, inspecting all dam facilities, testing
all equipment, making reports to dam owners.
3.3. DAM SAFETY CHECKS
Inspections are carried out annually for dam instrumentation and 5 years for
the entire dam related facilities including: hydrology, hydromechanical, geology and
others, where the report will be presented to the Dam Safety Center.
3.4. SPECIAL CONDITIONS INSPECTION
Dam inspection after disasters such as floods, earthquakes and others that
threaten the safety of the dam.
4. CONCRETE DAM INSTRUMENTATION
Concrete dam at PT. Vale has instrumentation to monitor dam safety as well
as equipment that can be observed 24 hours continuously in the control room area,
so that when an earthquake occurs, the first information conveyed is the earthquake
value received by the dam from the epicenter.
4.1. BALAMBANO DAM
The Balambano Dam is located seven kilometers downstream from the

Batubesi Dam. The Balambano Dam is a Gravity Dam Roller Compacted Concrete
(RCC) dam. The peak elevation of the Balambano Dam is 167.00 meters above sea
level. This dam is equipped with a free threshold type spillway and chute channel
with three radial doors with a size of each door being 15m x 8m. Mercu overflow
is at elevation. 151.50. The catchment area of the Balambano Dam is 2749.7 km2 .
This dam is equipped with an Ogee type spillway.
4.1.1. Main dam
• Type : Roller Compacted Concrete (RCC)

• Peak elevation : EL. 167.00 m
• Parapet peak elevation : EL. 168.10 m
• Peak length : 360.00 m
• Peak width : 6.00 m
530
Q.104 - R.38
• Height above foundation : 100,00 m

• Volume of concrete : 546,071 m3
• Upstream slope : Upright
• Downstream slope : 1 : 0.8
• Membrane peak : EL. 139.34 m
• Membrane floor : EL. 76.00 m
• Upper inspection tunnel bottom : EL. 115.00 m
• Lower inspection tunnel bottom : EL. 82.00 m
4.1.2. Instrumentation
The instrumentation installed at the Balambano Dam is as follows:
Table 1
Balambano Dam Instrumentation
INSTRUMENTS PARAMETERS MONITORED TOTAL REMARKS
Pneumatic piezometer Pore water pressure in the dam foundation 14 a. 8 unit in lower gallery
b. 6 unit in upper gallery
Standpipe piezometer Pore water pressure at the foot of the dam 10 a. 5 unit in the right bank
downstream b. 5 unit in the left bank
Box drain Seepage from the layer upstream of the 6 a. 2 unit in lower gallery
dam b. 4 unit in upper gallery
Vertical drain hole Seepage through foundation and dam 77 a. 28 unit lower gallery in left bank
body b. 19 unit lower gallery in right bank
c. 18 unit upper gallery in left bank
d. 12 unit upper gallery in right bank
Weir Current from box drain, drain hole and 8 a. 2 unit lower gallery
leakage around gallery wall b. 1 unit access lower gallery
c. 5 unit in upper gallery
Crest Monument Deflection and movement top dam to flat 18 Top dam
and vertical side
Joint meter Different deflection between monolith and 88 a. 8 x 4 unit stud top dam
monolith others b. 2 x 4 unit stud in lower gallery wall
c. 2 x 4 unit stud in lower gallery floor
d. 2 x 4 unit stud in upper gallery wall
e. 5 x 4 unit stud in upper gallery floor
Crack meter Movement in downstream of dam 12 a. 7 unit downstream left bank

b. 5 unit downstream right bank
Inclinometer Slope stability 5 Switchyard and tower
Strong Motion Accelerograph Earthquake acceleration 2 Top dam and powerhouse
Robotic total station Deformation dam 1 Top dam
4.1.3. Balambano dam instrument threshold value
The parameters monitored by the Balambano Dam instrumentation each have

a threshold value or warning value and / or an alarm value, as mentioned below.
531
Q.104 - R.38
a. Pneumatic piezometer
The tip elevation and piezometric elevation alarm limits for each pneumatic
piezometer are shown in table 2 below.
Table 2
Pneumatic Piezometer Alarm
NO. PIEZOMETER ELEVATION OF TIP (M) ELEVATION DESIGN PNEUMATIC
PIEZOMETER (M)
PP1 65,484 106,5
PP2 74,084 106,5
PP3 71,411 99,0
PP4 76,111 99,0
PP5 66,895 92,0
PP6 74,395 92,0
PP7 69,394 85,0
PP8 77,294 85,0
PP9 105,982 128,5
PP10 114,482 128,5
PP11 106,139 126,0
PP12 116,139 126,0
PP13 104,919 120,5
PP14 113,419 120,5
b. Standpipe piezometer
The tip elevations and the piezometric elevation alarm limits defined for each
standpipe piezometer are as shown in table 3 below.
Table 3
Standpipe piezometer Alarm
NO. Elevation of Tip Elevation Design Standpipe
STANDPIPE (M) Piezometer (m)
SP1 144,60 156,68
SP1A 126,57 136,57
SP2 120,10 129,83
SP2A 95,46 105,46
SP3 82,10 93,66
SP4 75,10 90,07
SP5 75,60 96,17
SP6 107,10 119,78
SP7 130,60 146,56
SP8 148,60 157,42
532
Q.104 - R.38
c. Drain box
The drain box functions to collect and measure water seepage from the Carpi
waterproof membrane installed on the upper surface of the dam. There are 6 (six)
box drains named BSD1 to BSD6. The flow alarm limit for each drain box is 5 liters
/ second.
d. Vertical drain hole
The drain hole serves to capture seepage through the dam body and the
foundation of the upper gallery and lower gallery. The numbering of drain holes in
the upper and lower galleries starts from the As Spillway (center of the dam) towards
the right abutment and the left abutment. The flow alarm limit for each drain hole is
5 liters / second.
e. Gallery drain weir
The total seepage from the drain box, drain hole and leaks around the gallery
walls in the upper left and right galleries and the lower left and right galleries is
measured by flat weir measuring instruments installed in the drain gallery, namely
W1, W2, W4, W5 , W6, W7 and W8. The limit for warning flow and alarm flow for
each drain weir is shown in table 4 below.
Table 4
Weir Alarm
NO. WEIR PERINGATAN ALIRAN ALARM ALIRAN
(LITER/DET) (LITER/DET)
W1 10 12
W2 10 12
W3 10 12
W4 25 30
W5 25 30
W6 25 30
W7 10 12
W8 25 30
f. Crest monument
Crest monuments or sliding stakes are installed at the top of the dam to deter-
mine the amount of deflection and settlement that occurs at the top of the dam.
Limits for warnings and alarms for deflections and subsidies on crest monuments
are shown in table 5 below.
533
Q.104 - R.38
Table 5
Deflection Alarms and Derivation Alarms
DAM MONOLITH CREST par DEFLECTION DEFLECTION DERIVATION DERIVATION
JOINT MONUMENT WARNING WARNING WARNING WARNING
1 (MM) 2 (MM) 1 (MM) 2 (MM)
1 (Abutment) CM1 +2 +4 +1 +2
1/ CM2,CM3 +2 +4 +1 +2
2
2/3 CM4,CM5 +3 +5 +1 +3
3/ CM6,CM7 +7 +9 +2 +4
3
4/5 CM8,CM9 +8 +10 +2 +4
5/6 CM10,CM11 +10 +12 +3 +5
6/7 CM12,CM13 +10 +12 +3 +5
7/8 CM14,CM15 +6 +8 +2 +4
8/9 CM16,CM17 +3 +5 +1 +3
9 (Abutment) CM18 +2 +4 +1 +2
g. Joint meter
Joint meter is installed at the top of the dam to measure the difference in
deflection between one monolith and the other. The difference in deflection is also
monitored in the gallery at the meeting between the 2 monoliths and the "joint
meter stud" installed on the wall and on the gallery floor. In design, the maximum
displacement is set at 20 mm.
4.2. KAREBBE DAM
The Karebbe Dam is located six kilometers downstream from the Balambano
Dam. Karebbe Dam is a Combined Concrete dam (RCC-LCVC). The peak elevation
of the Karebbe Dam is 79.50 meters above sea level, with the elevation of the river
bed is 9 meters above sea level. This dam is equipped with 3 spillways, namely 1
Service Spillway and 2 Auxiliary Spillways. Service Spillway type Ogee is equipped
with a door with a threshold elevation of 60.75 meters and a width of 8.00 meters.
Doors for Service Spillway use radial doors with a height of 18.68 m; 8.00 m wide.
Auxiliary Spillways are located to the left and right of the Service Spillway. This
spillway is an overflow with the Free Threshold type with a threshold elevation at El.
79.50 meters The catchment area from the Balambano Dam outlet to the Karebbe
Dam is 53 km2 .
4.2.1. Main dam
• Type : Combined (RCC-LCVC)

• Peak elevation : + 79.50 m (minimum)
• Riverbed elevation : + 9.0 m
• The deepest base of the foundation : + 5.50 m (plinth)
534
Q.104 - R.38
• The top length of the dam : 202.00 m

• Top dam width : 6.00 m (minimum)
• Height from riverbed : 66.90 m
• Dam height : 74.00 m
4.2.2. Instrumentation
The instrumentation installed at the Karebbe Dam is as follows:
Table 6
Karebbe Dam Instrumentation
INSTRUMENTS PARAMETERS MONITORED TOTAL REMARKS
Abutment Monument (AM1 Dam deflection 2 Abutment left and right
& AM2)
Crest monument Dam deformation 18 Top dam
Monolith Joint Pergerakan sambungan antar monolith 76 a. 9 x 4 unit stud in topdam
Movement b. 10 x 4 unit stud in the wall and gallery floor
Inclinometer Slope stability 4 a. 1 unit in left bank
b. 1 unit in right bank
c. 2 unit in dambody
Crackmeter Movement in downstream of dam 14 a. 7 unit in left bank
b. 4 unit in right bank
c. 3 unit in Power House
Upward & Downward Drain Seepage through foundation and dam body 170 a. 53 unit in Gallery 1 (Downward)
Hole b. 73 unit in Gallery 1 (Upward)
c. 8 unit in Gallery 4 (Upward & Downward)
d. 23 unit in Gallery 5 (Downward)
e. 13 unit in Gallery 5 (Upward)
Box Section Drain Seepage from upstream water proof 9 Gallery
Weirs Current from box drain, drain hole and leak- 6 Gallery
age around gallery wall
Standpipe Piezometer Pore pressure in downstream 7 Stepdam
Vibrating Wire Pore pressure in foundation 14 a. 2 bh di sandaran kanan
Piezometer b. 4 bh di sandaran kiri
c. 7 bh di bendungan
Strong Motion Earthquake 2 Dam site and Power House
Accelerograph
Robotic total station Deformation dam 1 Top dam
4.2.3. The Karebbe dam instrumentation threshold value
The parameters monitored by the Karebbe Dam instrumentation each have a

threshold value or warning value and / or an alarm value, as mentioned below.
a. Crest monument
Crest monuments or sliding stakes are installed at the top of the dam to deter-
mine the amount of deflection and settlement that occurs at the top of the dam. Limits
535
Q.104 - R.38
of warning (warning) and alarms for deflections and descent on crest monuments
are as shown in Table 7 below.
Table 7
Crest Monument Threshold
DAM MONOLITH JOINT CREST MONUMENT HIGH DAM ESTIMATED ESTIMATED VERTICAL
AT JOINT (M) HORIZONTAL MOVEMENT (SETTLEMENT)
MOVEMENT (MM) (MM)
Right Abutment to Monolith CM 1 0 0 0
9
Monolith 8B to Monolith 9 CM 2, CM 3 2 0 0
Monolith 7 to Monolith 8A CM 4, CM 5 15 3 1
Monolith 6 to Monolith 7 CM 6, CM 7 36 8 3
Monolith 5 to Monolith 6 CM 8, CM 9 47.6 11 4
Monolith 1 to Monolith 2 CM 16, CM 17 28 6 2
Left Abutment CM 18 0 0 0
b. Vibrating piezometer and pneumatic piezometer standpipe
The tip elevation and piezometric elevation alarm limits for each piezometer
are shown in Table 8 below.
Table 8
Vibratic Pneumatic and Standpipe Piezometer Threshold
VIBRATIC PNEUMATIC TIP ELEVATION PIESOMETRIC PIESOMETRIC
PIESOMETER AND (M) ELEVATION (M) PRESSURE (M)
STANDPIPE PIEZOMETER
VWP1-2 (upper) 37.282 51.083 13.801
VWP1-1 (lower) 28.882 51.083 22.201
VWP2-2 (upper) 36.747 43.914 7.167
VWP2-1 (lower) 28.347 43.914 15.567
VWP3-2 (upper) 4.420 28.654 24.234
VWP3-1 (lower) -3.980 28.654 32.634
VWP4-2 (upper) 0.930 24.987 24.057
VWP4-1 (lower) -7.470 24.987 32.457
VWP5-2 (upper) -0.100 19.578 19.678
VWP5-1 (lower) -8.500 19.578 28.078
VWP6-2 (upper) 15.879 43.916 28.037
VWP6-1 (lower) 7.749 43.916 36.167
VWP7-2 (upper) 16.964 27.756 10.792
VWP7-1 (lower) 8.564 27.756 19.192
SP1 42.186 52.936 10.750
SP2 25.985 36.135 10.150
SP3 6.278 21.328 15.050
SP4 15.671 27.021 11.350
SP5 27.739 38.139 10.400
SP6 34.399 45.349 10.950
SP7 55.368 65.218 9.850
536
Q.104 - R.38
5. METHODOLOGY AND DATA ANALYSIS
The method of monitoring work when an earthquake occurs in the form of

observation and data collection as well as data analysis and follow-up carried out
by the Hydro Surveillance Team of PT. Vale [3,4,5] as follows.
5.1. ONLINE OBSERVATION
Earthquakes that occur in the Sorowako area or adjacent to the dam must
be known quickly with a fast observation tool, there are 2 websites that can be
monitored from housing if an earthquake occurs at night or occurs outside working
hours [3,4].
5.1.1. Earthquake value
The geotechnology department in collaboration with IT provides a website,

where the magnitude of the earthquake that occurs will be informed by BMKG [6]
including the distance to the epicenter. Calculation of earthquake distance, design
value of dam earthquake and earthquake alarm value, so that follow-up is carried
out immediately referring to the emergency action plan document [4] that has been
approved by the Regional Head of East Luwu Regency. Following are the results of
data collection on the website which states that the distance of the Balambano Dam
from the epicenter is 4.75 km with an earthquake value received by 0.1329 g which
is still below the design value of 0.6 g, while the distance of the Karebbe Dam from
the epicenter is 4.34 km. The earthquake value received was 0.1357 g which was
still below the design value of 0.45 g (Fig.5), so that the two dams were still safe.
Fig. 5
Earthquake value at dam.
Valeur sismique au barrage.
537
Q.104 - R.38
5.1.2. Robotic total station
Monitoring the need for a dam deformation monitoring system in real time
using three-dimensional and one-dimensional sensors for both static and dynamic
movements is very important. As a pilot step, an activity to implement dam deforma-
tion monitoring technology based on active stations GPS / GNSS - CORS (Global
Positioning System / Global Navigation Satellite System - Continuously Operating
Reference Station) [5]. The use of Geodetic Monitoring Systems (Geomos) can be
considered as a step forward in a dam monitoring system based on the latest tech-
nology, which is based on speed, accuracy, reliability and effectiveness with very
little human error in dam measurement and monitoring activities.
Fig. 6
Robotic and prism location (left) also robotic monitoring system (right).
Emplacement robotique et prisme (à gauche) et système de surveillance robotique
(à droite).
There are several types of monitoring of movement in the dam using robotic,
including:
a. Longitudinal movement
The movements in the direction of the landslide are upstream and down-
stream
b. Transverse movement
Movements that are perpendicular to longitudinal movements
c. Height movement
Up and down movement of the dam
Fig. 7
Longitudinal (left) and Transversal (right) movement.
Mouvement longitudinal (à gauche) et transversal (à droite).
538
Q.104 - R.38
Fig. 8
Height movement.
Mouvement de hauteur.
5.1.3. Nivel
Monitoring the alignment of the concrete dam is very necessary to ensure the
vertical line from the top to the dam foundation, therefore at the time of installing
Fig. 9
Nivel.
Nivel.
Fig. 10
Nivel movement to X axis (left) and Y axis (right).
Nivel mouvement à l’axe X (à gauche) et de l’axe Y (à droite).
539
Q.104 - R.38
Table 9
Standpipe Piezometer Monitoring After Earthquake [3]
SP APPROX. AUG 7-SEP 22-SEP WARNING REMARKS
BOTTOM (AFTER SP
ELEVATION EARTHQUAKE) ELEVATION
(M) (M)
READ. (m) ELEV. READ. (m) ELEV. READ. ELEV.
(m) (m) (m) (m)
SP-1 144.6 5.11 151.99 5.18 151.92 5.14 151.96 156.68 Normal
SP-1A 126.57 8.04 129 8.31 128.73 7.56 129.48 136.57 Normal
SP-2 120.1 5.65 124.59 6.12 124.12 6.52 123.72 129.83 Normal
SP-2A 95.46 9.46 96.45 9.39 96.52 9.8 96.11 105.46 Normal
SP-3 82.1 6.16 87.84 6.45 87.55 6.75 87.25 93.66 Normal
SP-4 75.1 7.81 82.58 7.98 82.41 7.92 82.47 90.07 Normal
SP-5 75.6 9.84 86.76 10.53 86.07 10.72 85.88 96.17 Normal
SP-6 107.1 11.78 108.34 12 108.12 12 108.12 119.78 Normal
SP-7 130.6 8.01 138.97 8.47 138.51 8.17 138.81 146.56 Normal
SP-8 148.6 3.1 154.6 3.8 153.9 3.95 153.75 157.42 Normal
the robotic a level equipment is added which functions to measure the alignment.
These tools are installed in the upper and lower galleries.
5.1.4. Online Data Analysis
Based on the measurement results of the automatic robotic online survey

measuring instrument, there are no conditions that indicate abnormal movements
after the earthquake. All survey prisms in the dam or in the slope area remain stable
or normal. Nivel is used to observe the straightness of the dam from above (upper)
to lower (lower) under normal conditions.
5.2. MANUAL OBSERVATION AND MEASUREMENT
5.2.1. Dam instrumentation
Measurement of instrumentation equipment such as standpipe piezometer,

inclinometer, weir and others is still done manually by measuring data directly in
the field by the surveillance team. The measured data results will be compared with
the design parameter values. If the value is below the design, it can be said that the
dam is still safe or normal as the measurement results below
540
Q.104 - R.38
Fig. 11
Standpipe piezometer graph.
Graphique du piézomètre de la colonne.
Fig. 12
Strong Motion Accelerograph.
Accélérographe de mouvement puissant.
Based on the results of data observations on the piezometer standpipe instru-

mentation on the body of the dam, it shows that all the elevation readings of the
piezometer standpipe are still in normal conditions. Furthermore, for instrumenta-
tion measurements are taken and plotted according to the data to see the conditions
before and after the earthquake occurred.
5.2.2. Strong motion accelerograph
This equipment is installed in the dam area to find out the effect of the earth-
quake on the dam which is shown in the graphic image if there is and the information
there is no effect at all after the earthquake [3].
541
Q.104 - R.38
5.3. VISUAL CHECK
In accordance with statutory regulations and reporting procedures, in addi-

tion to reporting the condition of the instrumentation both online and manually,
the dam manager is also required to visually inspect the condition of the dam and
its facilities according to a standard format and complete with photos and date of
collection.
5.4. MANAGEMENT REPORTING SYSTEM
Every month the management holds a meeting to find out the safety conditions
of the dam based on its operation and maintenance system. The dam management
unit, especially surveillance reports on the condition of the dam in the form of a dash-
board, key performance indicator and closes with an executive summary consisting
of positive, attention and negative.
6. CONCLUSION
Dam safety is aimed at protecting the dam from dam failure and protecting life,
property, and public infrastructure in areas that are affected by potential hazards
due to dam failure. In order to minimize the risk of dam failure, intensive monitoring
is necessary anywhere through data and visualization and in real time. From the
results of the measurement, we can see that the data and visual information are
obtained in real time and comprehensive conditions, which can be accounted for
scientifically as well as by manual measurement. Rapid response after an earth-
quake can be followed up immediately if abnormal conditions are found, therefore
PT. Vale Indonesia has a program for installing instrumentation online through a
communication network connected to the PLTA control room.
Implementation of routine flood warning trials is carried out by the PTVI Hydro
Team every 6 months to ensure the public’s response to hazards that threaten safety
due to dams experiencing earthquakes, floods and others, while RTD simulations
have been carried out twice with coordinated by PTVI in collaboration with local
government and communities. Henceforth, the RTD simulation will be carried out
according to the schedule initiated and coordinated by the agency appointed by the
local government.
542
Q.104 - R.38
ACKNOWLEDGEMENT
This paper is dedicated to management of PT. Vale Indonesia Tbk such as:
Mr. Vinicius Mendez as Chief Operating Officer (COO), Mr. Abu Ashar as Director of
Environment & Energy (EE) Department, Mr. Pamrih Pammu as General Manager of
Utilities Operation; and Surveillance Team as our respectful counterpart in operation
and maintenance of hydro dam, and all colleagues.
REFERENCES
[1] MINISTER OF PUBLIC WORKS AND HOUSING. Minister’s Regulation Num-

ber 27 / PRT / M / 2015 concerning Dams. 2015.
[2] Government Regulation No. 72 / PRT 1997 concerning Dams. 1997.
[3] PT. VI SURVEILLANCE. Dam inspection report after earthquake. Sorowako,

Sept 2020.
[4] PT. VI SURVEILLANCE, Emergency action plan document Larona river series.
Sorowako, 2016.
[5] SNC. LAVALIN. KLOHN CRIPPEN, PACIFIC RIM, OM & surveillance manual
of Balambano Dam, Sorowako, 1998
[6] BMKG. Meteorology, Climatology, and Geophysical Agency. https://www.bmkg.

go.id/gempabumi/gempabumi-terkini.bmkg. Accessed September 2020.
543
Q.104 - R.39
-------
-------
APPLICATION OF MODULAR PRECAST CONCRETE SYSTEM ON DAM

SIDEWALL (∗)
Ahmad SIDIK
Site Engineering Manager of the MARGATIGA DAM PROJECT, PT. Waskita Karya
Dimaz Irja VIRATAMA & Stella MONICA

Site Engineering Officer of the MARGATIGA DAM PROJECT, PT. Waskita Karya
INDONESIA
SUMMARY
The construction sector is rapidly developing along with technology develop-

ments, in seeking innovations to support the implementation of construction work
which are expected to be able to simplify and accelerate work and increase efficiency
and effectiveness in carrying out construction work, both physical and administrative.
Modular concrete system is an innovation and improvisation from conventional con-
crete systems that has been widely used. Conventional systems that are currently
widely used have several shortcomings, including longer work times and quality
results that are not necessarily as planned. The existence of concrete modular sys-
tems or precast concrete is a form of solution to overcome these problems, because
it can shorten work time and has a guaranteed quality because it is produced and
created directly in the batching plant or factory. This concrete modular system will be
applied to the dam sidewall building which functions as a retaining foundation for the
soil. With careful planning and other considerations, it is hoped that this work can
be carried out more efficiently and can be completed in a shorter time and in accor-
dance with the quality of the plan, so that it can provide good benefits for service
users and service providers.
∗ Utilisation d’un système modulaire préfabriqué en béton pour des parois latérales
544 DOI: 10.1201/9781003211808-36

Q.104 - R.39
RÉSUMÉ
Le secteur de la construction se développe rapidement parallèlement aux

développements technologiques, à la recherche d’innovations pour soutenir la mise
en œuvre des travaux de construction et qui devraient être en mesure de simpli-
fier et d’accélérer les travaux et d’accroître l’efficience et l’efficacité des travaux
de construction, à la fois physiques et administratifs. Le système de béton mod-
ulaire est une innovation et une amélioration par rapport au béton conventionnel.
Le système conventionnel qui est actuellement largement utilisé présente plusieurs
lacunes, notamment des temps de travail plus longs et des résultats de qualité
qui ne sont pas nécessairement optimaux. Un système modulaire en béton ou en
béton préfabriqué est une solution pour surmonter ces problèmes ; il peut raccourcir
le temps de travail et a une qualité garantie car il est produit et créé directement
dans le plan de dosage ou l’usine. Ce système sera appliqué au bâtiment de paroi
latérale du barrage qui sert de fondation de soutènement pour le sol. Avec, notam-
ment, une planification minutieuse, on espère que ce travail pourra être effectué
plus efficacement et achevé dans un délai plus court et conformément à la qualité
attendue.
1. INTRODUCTION
The Sekampung River is located in Lampung Province, and it has the potential
to meet various needs, including irrigation , water supply and electricity. The next
water resource development plan in Way Sekampung is construction of two dams,
namely Way Sekampung Dam and Margatiga Dam. The construction of these two
dams is a unitary use of the Way Sekampung river with cascade characteristics.
Cascade dams are types of dams that are built in series along the river with a
certain distance or in a terraced.
The Margatiga dam is a composite type dam with a combination of a concrete

dam type (concrete gravity dam) and rock filled with an upright core. This dam has
MYC (Multi Years Contract) type with a fairly long duration of work within December
2017 to November 2021, or about 4 years of work construction time. During the
construction process of this project, many problems were found and caused the
delay of construction delays which directly led to time reduction of its completion.
Which results in reduced implementation time. To overcome this problem, innovation
or improvement is needed to increase productivity and work efficiency without having
to reduce specifications so that the project completion time is still managed. The
implementation of innovation that can be used is applying the modular concrete
system for sidewall work.
The key to the successful implementation of a modular system depends on

the management aspect. Several factors from the aspects of proper management
545
Q.104 - R.39
are: (a) technology, (b) materials, (c) human resources, (d) planning, (e) logistics, (f)
production, (g) transportation and distribution, (h) installation and repair. By apply-
ing a modular system in this project, it is hoped that it can provide effectiveness,
efficiency and quality according to specifications so that the Margatiga dam project
can be completed on time and does not experience delays [1].
The study of the application of modular systems in concrete sidewall of dams

is intended to provide an overview of the innovative implementation of construction
on concrete works and to provide alternatives for time acceleration while maintaining
quality and achieving construction targets to shorten the implementation time while
maintaining quality and to achieve implementation targets.
2. MATERIALS AND METHODS
The modular system used is a construction implementation method that uti-

lizes a material or component that is manufactured inside or outside the work area
(project), but needs to be put together in the place where it should be. The use of
a modular system has some advantages when compared to conventional methods,
in the aspect of constructability (ease in implementing), positive aspect of the work
schedule, fewer workers needed, quality aspect, productivity, and testing [2].
2.1. MODULAR SYSTEM CONCEPT APPROACH
Prefabricated concrete is the industrialization of concrete construction meth-

ods where components are mass produced and assembled in buildings with the help
of cranes and lifting equipment. The advantages of using prefabricated concrete
compared to conventional methods are as follows:
a) Good quality control, because the production process is in the factory and
can be measured well
b) The implementation is short, because the implementation of the lower struc-
ture coincides with the production of components in the factory and the
implementation of the superstructure simultaneously with the architectural
finishing work
c) Costs are more economical because there is a reduction in the use of molds,
shorten construction time and high labor productivity in the field
d) Environmentally friendly, all casting work is carried out in the fabrication
place so that there are no remains of casting in the field [3]
For example in the construction of buildings, a cost comparison chart of several

different systems in units of square meters per floor of the building, is shown in Fig. 1.
546
Q.104 - R.39
Fig. 1
Comparative cost industrialised system for high, medium and low-rise building [4].
Système industrialisé à coût comparatif pour les bâtiments de grande, moyenne et
faible hauteur [4].
Advantages of the use of precast concrete technology can be seen early, the cost of
which is required every square meter floor of the building is smaller than the in-situ
concrete building system, especially at a high level [4].
2.2. MODULAR SYSTEM APPLICATION FACTOR
2.2.1. Planning
Planning is one of the aspects of the proper management. Planning is very

important because in the modular system there are so many aspects to prepare
compared to the conventional method. In the planning process, it is necessary
to consider, predict and also control various processes during implementation.
Planning begins with the conceptual stage until completion of the implementation.
Planning is a very critical activity stage because the modular system is not easy to
adapt to many changes that can occur at any time. On the other hand, the flexibility
level of this method is low and not as flexible as conventional systems. Therefore, the
design of the modular components must be approved in advance to avoid possible
changes.
2.2.2. Resources
This stage begins with selecting service users who have competences in their
fields such as planning consultants and precast concrete company. In general, both
consultants, suppliers and contractors must have a good understanding, sufficient
experience in carrying out work with this modular system. Most of the material and
component production processes will be carried out in the factory, there are some
547
Q.104 - R.39
Table 1
The characteristic differences between factory workers and the open field workers
NO FACTORY WORKERS OPEN FIELD WORKERS
1 Not affected by weather Affected by weather
2 Workers are easier to adapt to technological Workers are harder to adapt to technology
changes changes
3 Relatively unchanged work patterns Always have different work pattern and location
4 Constant productivity Unconstant productivity
5 Relatively low number of workers Relatively high number of workers
differences between those working in the factory environment and those working
in the open field. The work performed In the factory is repetitive work to minimize
the possibility of failure caused by workers. Table 1 explained some characteristic
differences between factory workers and the open field workers [5].
By directing a construction project into industrialization, it is possible to

increase efficiency, reduce construction costs and time to implement it.
2.3. PRODUCTION
The important thing in the production factor is to determine priorities, which one
is manufactured first, so that coordination between the fabricator and the installer
is needed. The benefits of precast concrete fabrication are fewer workers needed,
fast production process and improvement of product quality.
Compared with the conventional method, things that stand out in the produc-
tion of precast concrete is the use of machinery in a factory to produce precast
concrete components. In addition to requiring less labor, use of the machine will
reduce errors caused by "human factors" that will produce a more uniform product
quality.
2.4. TRANSPORTATION AND HANDLING
Modular system planning carried out at the beginning of the project is related to
the dimensions and weight of each module that has been planned. Precast concrete
manufacturers are generally not only responsible for the production but also respon-
sible for the transportation or even installation of precast concrete components. In
general, producers have their own mode of transportation to transport their products
to work locations.
Precast concrete components are usually supported on two pedestals to avoid

creating stress which occur during the transportation process to the project site.
548
Q.104 - R.39
For installation purposes, a two-point lift system is used if the precast concrete
components are double T, inverted T, L beam, hollow-core slab [6]. It is necessary
to check the carrying capacity and the maximum weight of the road to be traversed.
3. RESULT AND DISCUSSION
3.1. DAM SIDEWALL
Sidewall building or retaining wall construction is one type of civil construction

that has function to withstand the active lateral pressure forces of soil or water.
Therefore a retaining wall construction must be planned and designed to be secure
against the forces that can potentially cause structural failure. Basically, the retaining
wall has several functions, such as:
a) Restrain active lateral forces (active lateral force soil) which can potentially
cause lateral soil collapse, for example landslides.
b) restrain lateral force water which can potentially cause lateral collapse due
to high water pressure.
c) Preventing the occurrence of lateral seepage infiltration caused by a high
groundwater level. In this case it also functions in the dewatering process,
by cutting the water flow (flow net) on the ground (cut off) [7].
Gravity retaining wall is a type of retaining wall used in the Margatiga Dam,
as can be seen in Fig. 2. This type is widely used to restrain lateral soil pressure
on landfills and on steep sloping cliffs. The constituent materials used in this type of
construction are usually masonry or reinforced concrete.
Fig. 2
Gravity Retaining Wall [7].
Gravity mur de soutènement [7].
549
Q.104 - R.39
3.2. SIDEWALL STRUCTURE LAYOUT
Sidewall building at Margatiga Dam is planned on the right and left between
spillway and retrieval building (Fig. 3). One example of a building that is planned to
use a modular concrete system is wingwall structure and the work plan can be seen
in Fig. 4.
Fig. 3
3D View of the dam building.
Vue 3D du bâtiment du barrage.
Fig. 4
Concrete Wingwall Structure Work Plan.
Plan de travail du bâtiment en béton mur d’aile.
550
Q.104 - R.39
The work plan of the right concrete wingwall structure is divided into 10
segments with a length of 12.2 meters for each segment. Based on Fig. 5, the
cross-section of the wingwall has a structure height of 12 meters, which is divided
into a foundation height of 3 meters and a wall height of 9 meters, it is planned to
cast into 5 stages of casting (foundation, segment 1 to segment 4). The utilization
of conventional methods is estimated and requires a longer implementation time
than a modular system. So that acceleration of the construction completion time is
planned using a modular precast concrete system.
Fig. 5
Cross section of the concrete wingwall structure.
Coupe transversale du bâtiment de l’aile.
3.3. SPECIFYING MODULAR DIMENSIONS
The determination in dividing the modular dimensions must be done properly

and precisely, the modular segment must also be adapted to the existing design
cross-section. For the weight of each modular segment, it must be adjusted to
availability of conveyances that will be used later. The determination of the modular
dimensions is in accordance with the previously approved design drawings on which
dam sidewall stability is already considered.
3.3.1. Modular type
The segment division is made for each modular segment based on the cross
section of the structure. The modular segmentation is divided into several types
551
Q.104 - R.39
as shown in Fig. 6. Segment details can be seen in Table 2, which consist of 11

concrete modular segmentations, with the volume and weight of each segment.
Fig. 6
Modular Segmentation.
Division du segment modulaire.
Table 2
Modular Concrete Segmentations
NO TYPE VOLUME (M3) WEIGHT (TON) LENGTH (M) QTY PER M2 (PCS)
1 A1 1.565 3.76 1.5 3
2 A2-1 1.82 4.37 1.5 1
3 A2-2 2.04 4.90 1.5 1
4 A2-3 2.27 5.45 1.5 1
5 B 1.95 4.68 1.0 4
6 C1 1.53 3.67 1.5 4
7 C2 1.60 3.84 1.5 4
8 D1 0.99 2.38 1.5 3
9 D2 1.08 2.59 1.5 3
10 E1 1.57 3.77 1.5 1
11 E2 1.34 3.22 1.5 1
3.3.2. Joints between modular
The process of unification of precast concrete structure components into a

monolith structure is very important in applying precast concrete technology. The
connection between the precast components not only serves as a drain load but
must be able to effectively integrate these components, so that the overall structure
can behave monolithic [5].
552
Q.104 - R.39
The method used to unify precast concrete components is divided into two
ways, using dry and wet connections [5], using connections dry and wet connections.
In this sidewall job, In this sidewall work, wet joints are chosen to be perfect for site
condition. Wet joint is a method of joining precast concrete components in which
the joint can only function effectively after a certain period of time. Included in this
type is in-situ concrete joints. Concrete quality of wet joints are managed to have
the same as the modular concrete, which is K-225 (fc’ 18.68 Mpa). Not only using
wet concrete, there are also male-female joints between horizontal joints and 30 mm
bolt anchors with grouting for vertical directions. For details on the connection can
be seen in Fig. 7.
Fig. 7
The detail joints of vertical direction.
Les joints de détail de direction vertical.
Wet joints using a half-slab system, that means there are some parts using
precast method and the other parts using in-situ cast method. There is a cutting
reinforcement in the concrete modular segment which functions as a connection
between the modular segment and the in-situ concrete during casting. In addition,
the cuttings also function as shrinkage reinforcement in in-situ concrete. For details
on the modular segments with reinforcement see Fig 8.
Fig. 8
Details of Modular Concrete Reinforcement
Les joints de détail de direction vertical.
553
Q.104 - R.39
3.4. METHOD OF THE IMPLEMENTATION
This sub-chapter will discuss what matters are related to road access that will
be made at the work site, the location for placing the concrete (stockyard) and also
how the ability of the lifting equipment is used.
3.4.1. Site working layout
Work area planning must be well planned, by evaluating certain conditions

that can support efficiency in work so as to increase productivity. The layout of the
work area at the Margatiga Dam can be seen in Fig. 9.
Fig. 9
Working area of the Margatiga Dam.
La zone de travail du barrage de margatiga.
3.4.2. Lifting equipment specifications and lifting plan
There are many things that need to be considered in carrying out a work that
requires heavy equipment, such as the type, capacity and load that will be lifted later.
554
Q.104 - R.39
This work is planned to use a 50 ton Crawler Crane . This type was chosen because
the heavy equipment is easy to find at the work / project site, apart from having good
mobility for difficult areas and also having good lifting coverage. The weight of the
lifted load must be in accordance with the reach of the boom crane used. Crane lifting
distance planning can be seen in Fig. 10. The furthest distance to lift is 14 m, then
next is knowing the maximum load that can be lifted with a distance of 14 m and a
boom length of about 18 m. To find out, it can be seen in Table 3. Based on the loading
chart, it is obtained with a lifting distance of 14 m, which can lift a maximum load
weighing 7200 kg or 7.2 tonnes. In calculation from the heaviest planned modular
segment with 5.4 tonnes (segment A2-3), safety factor obtained is 1.3.
Fig. 10
Crawler crane 50 ton.
Grue sur chenilles 50 tonnes.
Table 3
Loading chart crawler crane 50 ton
555
Q.104 - R.39
3.5. Evaluation of construction time
Based on the evaluation of the work implementation time plan, there is a

difference in implementation time between using the conventional method and using
the modular system method, it can be seen in Table 4 and Table 5.
Table 4
Wingwall Structure Schedule Plan Using Conventional Methods
Table 5
Wingwall Structure Schedule Plan Using Modular System
556
Q.104 - R.39
Based on the table plan schedule for wingwall work implementation, it is found
that the time to complete the work using the modular system is 1 month faster than
the implementation using the conventional methods. In the implementation of the
wingwall work using the conventional method, it is planned to use 4 teams with each
team consisting of 25 workers or a total 100 workers, and using a crane as a tool for
lifting weight and with a casting target of 250 m3 / day. Meanwhile, the implemen-
tation of the modular system is planned to use 2 cranes and 2 casting teams with
5 workers or a total of ten workers. Based on this analysis, the implementation of
wingwall work with the modular system is 1 month faster and the number of workers
required is less, and it is estimated that the implementation costs are 20% more
efficient than using conventional methods.
4. CONCLUSIONS
Based on the analysis above, it can be concluded that using a modular system
in sidewall concrete work at the Margatiga Dam has several advantages compared
to conventional methods, including:
1. Can increase the efficiency of work implementation, because it does not

require formwork and scaffolding during implementation
2. Can increase the productivity of implementation so as to get accelerated
work
3. Quality of work is more guaranteed
4. Easy to implement
5. Requires fewer workers than conventional methods
The modular system can save implementation time so that it can accelerate
the work with hope that the construction of the Margatiga Dam can be completed
according to the contract timeline.
ACKNOWLEDGMENTS
For the completion of this article, the authors would like to thank many people
whose names may not all be enumerated. Their contributions are sincerely appreci-
ated and gratefully acknowledged. However, the authors would like to express deep
appreciation particularly to Dr. Eng. Evi Nur Cahya, ST., MT. who have provided
suggestions and input, the Margatiga Dam Project Team, and PT. Waskita Karya
(Persero) Tbk which has supported the authors.
557
Q.104 - R.39
REFERENCES
[1] ERVIANTO, W. I. Potential Use of Modular Systems in Construction Projects

(Potensi Penggunaan Sistem Modular Pada Proyek Konstruksi), Research
Report UAJY. 2008.
[2] HESLER, WARREN, E. Modular Design - Where It Fits, October, pp.120. 1990.
[3] JUWANA, J. S. The High-Rise Building System Guide (Panduan Sistem

Bangunan Tinggi), Erlangga, Jakarta. 2005.
[4] SEELEY, I.H. Building Economics, McMillan. 1972.
[5] KONCS, T. Manual of Precast Concrete Construction, Berlin. 1979.
[6] SHEPPARD, P. Plant-Cast Precast and Prestressed Concrete, McGraw-Hill

Book Company, New York. 1989.
[7] BUDI, G.S. The Shallow Foundation (Pondasi Dangkal), Andi Publisher, 2011.
558
Q.104 - R.40
-------
-------
BCR-E & E-BCR : BARRAGES HYBRIDES BCR – ENROCHEMENTS.

PRINCIPES ET CONDITIONS D’APPLICATION(∗)
Luc DEROO
Président, ISL Ingénierie
Etienne FROSSARD
Consultant
François LEMPERIERE
FRANCE
RÉSUMÉ
Les barrages en BCR et les barrages en enrochements constituent aujourd’hui

une part très importante des constructions neuves. Les barrages en BCR présen-
tent l’avantage de la structure rigide, qui peut résister aux déversements, y compris
pendant les travaux. Les barrages en enrochements présentent l’avantage de pou-
voir être installés sur des fondations de qualité moyenne, et d’être essentiellement
constitués d’un matériau – les enrochements – souvent très économique.
Ce rapport présente le principe de barrages hybrides, associant BCR et

enrochements. Deux options sont envisagées, avec le BCR à l’amont (BCR-E)
et avec le BCR à l’aval (E-BCR). Le rapport montre que cela conduit à des per-
formances intéressantes : des économies, une sécurité accrue, et une meilleure
adaptabilité aux conditions locales (fondation, matériaux).
Les avantages en termes de coûts sont quantifiés à partir de prix unitaires

constatés pour les deux types de barrage (BCR, enrochements), et en considérant
les incidences de la construction mixte. Avec les hypothèses moyennes considérées
∗ RCC-R & R-RCC dams: Hybrid RCC–Rockfill dams. Principles and conditions of
applications
DOI: 10.1201/9781003211808-37 559

Q.104 - R.40
dans cet article, les conceptions BCR-E et E-BCR sont 10 à 20% plus économiques.
De plus, elles sont également meilleures en termes de bilan carbone (mais moins
bonnes que le barrage CFRD pour ce qui concerne la consommation de sable). Bien
entendu, ces conclusions ne sont que des moyennes, et la comparaison dépend
des conditions topographiques, géologiques et hydrologiques de chaque site.
La sécurité de ces ouvrages est très bonne, pour deux raisons fondamentales.
La première raison est que le comportement mixte diminue beaucoup les sollicita-
tions sur la fondation : l’angle de frottement mobilisé sous le BCR est de l’ordre de
20◦ avec les géométries évaluées ici. La seconde raison est que la sécurité en crue
extrême, en service ou pendant le chantier, est améliorée par rapport aux options
CFRD (qui résiste mal à la submersion) et BCR (dont la réserve de stabilité ultime
peut être épuisée par la mise en traction au pied amont).
Les conceptions E-BCR et BCR-E peuvent être envisagées dans de nombreux

sites.
De ces deux approches, la plus prometteuse est certainement la conception

BCR-E. Elles paraissent notamment bien adaptées dans les cas suivants. La con-
ception BCR-E est très efficace économiquement, en particulier dans les cas où
le barrage est développé sur de grands linéaires. La conception E-BCR peut être
intéressante dans les sites où on souhaite faire porter l’évacuateur par le barrage
(vallée relativement étroite ou débits de crue élevés).
Et ces deux conceptions devraient être examinées systématiquement dans les

cas où les crues sont fortes et rapides, en raison de la sécurité intrinsèque qu’elles
apportent – par exemple les barrages écrêteurs de crue.
SUMMARY
RCC dams and CFRD are nowadays a very important part of new construction.
The advantage of RCC dams is that they have a rigid structure that can withstand
spillage, even during construction. CFRD have the advantage that they can be
installed on medium-quality foundations and are essentially made of a material –
rockfill - which is often very economical.
This report presents the principle of hybrid dams, combining RCC and rockfill.
Two options are considered, one with RCC upstream (RCC-R) and the other with
RCC downstream (R-RCC) and an upstream face. The report shows that this leads
to interesting performances: cost savings, enhanced safety, and better adaptability
to local conditions (foundation, materials).
With the average assumptions considered in this paper, the RCC-R and R-
RCC designs are 10 to 20% more economical than the traditional RCC and CFR
560
Q.104 - R.40
dams. Moreover, they are also better in terms of carbon balance (but less good than
the CFRD dam in terms of sand consumption). Of course, these conclusions are
only averages, and the comparison depends on the topographical, geological and
hydrological conditions of each site and project.
The safety of these structures is very good, for two fundamental reasons. The
first reason is that the hybrid design greatly reduces the stresses on the foundation:
the angle of friction mobilized under the RCC is of the order of 20◦ with the geome-
tries evaluated here. The second reason is that safety in extreme flood conditions,
in service or during construction, is improved compared to CFRD (which is not very
resistant to submersion) and to RCC dams (whose ultimate stability reserve can be
exhausted by cracking at the upstream foot).
R-RCC and RCC-R dams can be considered in many sites, including those
with poor bedrock.
Of these two approaches, the most promising is certainly the RCC-R option.
They appear to be well adapted in the following cases. RCC-R design is very eco-
nomically efficient, especially in projects with great crest length. R-RCC design
might be interesting in sites where it is desired to install the spillway on top of the
dam (relatively narrow valley and/or high flood flows) but the bedrock is of poor
or uncertain quality and no well-suited to RCC options. And both hybrid designs
should be systematically examined in cases of rivers with high and rapid floods,
because of the intrinsic safety they provide – for example in cases of flood mitigation
dams.
561
Q.104 - R.40
Fig. 1
Illustration du barrage BCR-E, avec une géométrie très simple (à gauche) ou plus
travaillée (à droite).
Sketch of a RCC-R dam, with a simple geometry (left) or a more refined design
(right).
1. PRINCIPE
1.1. BARRAGE BCR-E
Le barrage BCR-E est un barrage mixte construit en BCR et en enrochements,

les deux parties du corps du barrage étant montées simultanément, avec le BCR
à l’amont. Dans la montée des levées, la face aval du BCR n’est pas coffrée : la
séparation entre BCR et enrochements est donc constituée de redans.
Le BCR amont forme l’étanchéité. Dans ce rapport, les calculs sont présentés
avec un talus aval du BCR penté à 0,4H/1V ; des pentes plus raides peuvent être
envisagées. Les enrochements sont montés en même temps que le BCR. Un zon-
age du remblai aval peut être envisagé, pour optimiser les coûts et le fonctionnement
mécanique.
La géométrie la plus simple, illustrée à gauche sur la Fig. 1 ci-dessus, peut

souvent être plus économique que les solutions concurrentes de barrage en BCR
et de barrage en remblai. Une géométrie plus travaillée, telle que figurée à droite
sur la Fig. 1 est parfois pertinente : un BCR s’épanouissant à la base améliore les
contraintes au contact avec la fondation, des cotes de fondation différentes pour
BCR et remblai permettent des économies de matériau, et il peut être parfois même
opportun d’envisager une crête du remblai moins haute que la crête du BCR. Toute
la gamme des BCR est envisageable (du “ high paste ” au “ remblai dur ”), et la
géométrie adaptée en conséquence. Une géomembrane peut être envisagée en
parement amont pour régler la question de la fissuration et de l’étanchéité. Un
zonage du matériau de remblai est parfois utile pour faciliter la construction et/ou
optimiser le fonctionnement mécanique à l’interface ; il est généralement possible
de s’en passer.
562
Q.104 - R.40
Fig. 2
Illustration du barrage E-BCR, et deux options pour la géométrie.
Sketch of the R-RCC dam, with two options for the RCC geometry.
1.2. BARRAGE E-BCR
Le barrage E-BCR est un barrage mixte construit en BCR et en enrochements,

les deux parties du corps du barrage étant montées simultanément, avec le BCR à
l’aval.
L’étanchéité est assurée par un masque amont (typiquement en béton armé,

ou par un dispositif d’étanchéité à géomembrane). Un zonage du remblai est néces-
saire, comme pour les barrages CFRD, pour fournir le support au masque, et assurer
une perméabilité croissante de l’amont vers l’aval. Le BCR est conçu de sorte à être
très perméable, par exemple en y ménageant des galeries de drainage. Sous cer-
taines conditions, discutées plus loin dans ce rapport, il peut porter un évacuateur
de crues.
Les deux géométries présentées ci-dessus ont des comportements très

différents. Celle de droite, à parement aval vertical, diminue la sollicitation au
glissement du barrage BCR, par augmentation de l’effort vertical stabilisateur, mais
conduit à soulever le pied amont du BCR et augmente la sollicitation de la capacité
portante du rocher au pied aval. Il s’agit d’un comportement qui s’éloigne sensible-
ment du comportement classique d’un barrage en BCR et qui n’ a pas été étudiée
plus avant dans ce rapport.
2. LE RETOUR D’EXPÉRIENCE ET RÉFLEXIONS ANTÉRIEURES
Construire des barrages mixtes n’est pas une idée nouvelle, même si dans le
passé les matériaux de construction n’étaient ni le BCR ni les enrochements, mais
plutôt la maçonnerie et les remblais.
Ainsi, le barrage de Saint-Ferreol, le plus ancien des grands barrages français

(1675) est un barrage mixte. Par ailleurs, depuis plus de 100 ans, des barrages en
563
Q.104 - R.40
maçonnerie, dont la stabilité était jugée précaire, ont été transformés en barrages
mixtes maçonnerie – remblai. C’est par exemple le cas du barrage d’Alfeld, en
1895. Ces barrages sont illustrés par la Fig. 3. Ce type de confortement est toujours
régulièrement pratiqué, et illustré par la Fig. 4 dans le cas du barrage du Vérut, en
France conforté en 2005.
Plus généralement, l’idée d’associer le béton et le remblai pour former de

nouveaux barrages économiques et sûrs n’est pas nouvelle : les bulletins 83 et 117
de la CIGB en présentent plusieurs illustrations de principe, dont certaines sont
reproduites en Fig. 5.
En Chine, environ 200 petits barrages ont été construits de manière

économique, entre 1960 et 1975, dans la province de Zhejiang (Zhaogushe). Ces
barrages déversants, de conception peu orthodoxe, ne contenaient pratiquement
pas de ciment : la maçonnerie était en pierres sèches.
3. UN INTÉRÊT RENOUVELÉ POUR LES BARRAGES MIXTES
Les barrages BCR-E et E-BCR constituent une solution attractive. En effet, ils
répondent bien à toute une série d’enjeux actuels : diminution des coûts, adaptabilité
Fig. 3
Barrage de Saint-Ferreol © Voies Navigable de France et barrage d’Alfeld
© Conseil Départemental du Haut-Rhin.
Saint-Ferreol dam and Alfeld dam.
Fig. 4
Confortement du barrage de Vérut © M. Hoonakker [1].
Strengthening of the Vérut dam.
564
Q.104 - R.40
Fig. 5
Images extraites des bulletins 83 (à gauche) et 117 (à droite).
Illustrations from bulletins 83 (left) and 117 (right).
aux matériaux disponibles, maîtrise des aléas géologiques et géotechniques,

sécurité améliorée, empreinte environnementale réduite.
3.1. DIMINUTION DES COÛTS ET ADAPTABILITÉ AUX CONDITIONS DE SITE
Les sections courantes de barrage BCR-E et E-BCR sont économiques. Pour

un barrage de 50 m de hauteur de retenue, avec des longueurs en crête importantes
(la comparaison ne fonctionne pas pour les vallées étroites), et avec des profils BCR-
E et E-BCR sans optimisation, les ordres de grandeur sont ceux exprimés par le
Tableau 1.
Table 1
Comparaison des coûts typiques pour quatre types de barrages (hors ouvrages
hydrauliques)
SECTION COURANTE (SANS LES BCR CFRD BCR-E E-BCR
OUVRAGES HYDRAULIQUES ET EN
FONDATION SUR BON ROCHER)
Hauteur sur fondation (m) 50 50 50 50
Largeur en crête (m) 6 6 6 6
Pente talus amont 1.4H/1V 1.4H/1V
Pente talus aval 0.8H/1V 1.4H/1V 1.4H/1V
Pente talus intermédiaire 0.4H/1V 0.4H/1V
Volume BCR (m3 /m) 1020 510 510
Volume Enrochements (m3 /m) 3800 1400 1900
Surface masque amont (m2 /m) 86 86
Prix unitaire BCR (/m3 ) 100 120 80
Prix unitaire enrochements (/m3 ) 20 20 20 20
Prix unitaire masque (/m2 ) 300 300
Prix / ml section courante (k/m) 102 k 102 k 89 k 105 k
Ainsi, les coûts sont globalement analogues pour la section courante, à con-
ditions équivalentes de fondation ; avec une performance environ 10% meilleure
565
Q.104 - R.40
pour BCR-E. Cependant, trois autres facteurs vont dans le sens d’un coût global
plus faible pour les barrages BCR-E et E-BCR :
- Intégration des ouvrages hydrauliques : pour les deux solutions de barrage

hybrides, il est possible d’opérer une transition vers un profil standard BCR
là où on place les ouvrages hydrauliques : en vallée, là où les ouvrages
hydrauliques sont projetés, le profil mixte peut être progressivement rem-
placé par un profil BCR standard. Et, dans le cas E-BCR, il est possible de
placer l’évacuateur par-dessus la section courante
- Fondation moins profonde : les calculs présentés au §4 montrent que les sol-
licitations de cisaillement sur la fondation sont sensiblement plus faibles et
plus favorables que celles apportées par le barrage BCR : frottement mobil-
isé 20–22◦ contre plus de 35◦ pour le BCR. Il est donc possible d’accepter
une fondation de moins bonne qualité – tant qu’elle n’est pas trop compress-
ible. De plus, cette conception diminue l’impact des surprises géologiques
pendant les travaux, car le barrage est moins sensible à la qualité de la
fondation.
- Adaptabilité aux matériaux disponibles : la conception des barrages hybrides
peut être adaptée aux matériaux disponibles, à condition d’adapter la
géométrie en conséquence. BCR moins épais s’il est “ high paste ” ou plus
épais s’il est “ hardfill ”. Zonage du remblai selon la qualité des enrochements
ou des remblais disponibles. Possibilité de réaliser une partie du remblai avec
du matériau tout venant.
Il est impossible de proposer une comparaison universelle entre les différentes

options, car cela dépend beaucoup des conditions de site : topographie, géolo-
gie, hydrologie. Un ordre de grandeur est cependant donné ici, à titre d’illustration,
en utilisant des ratios moyens∗ .
Table 2
Comparaison des coûts typiques pour quatre types de barrages (y compris
ouvrages hydrauliques et surcoût pour la fondation du BCR)
SECTION COURANTE (AVEC BCR CFRD BCR-E E-BCR
OUVRAGES HYDRAULIQUES)
sans les ouvrages hydrauliques (k/m) 102 102 89 105
avec les ouvrages hydrauliques (k/m) 147 153 116 131
Cette analyse, qui est à reprendre au cas par cas selon les sites, montre
que les options de type BCR-E et E-BCR peuvent être plus économiques que les
solutions classiques.
∗ On postule que les ouvrages hydrauliques correspondent à une certaine fraction du coût
du corps du barrage : 20% pour BCR, 50% pour CFRD, 30% pour BCR-E et 25% pour
E-BCR. Et que le barrage en BCR nécessite des travaux plus importants d’excavation et de
traitement des fondations, avec un surcoût de 20%. Il est entendu que ces ratios diffèrent
fortement d’un projet à l’autre, et qu’ils ne sont utilisés ici qu’à titre d’illustration.
566
Q.104 - R.40
Pour les barrages en vallée étroite, le coût des installations de chantier

séparées (BCR / Enrochements) pénalise les solutions BCR-E et E-BCR.
3.2. SÉCURITÉ AMÉLIORÉE
La sécurité des barrages hybrides BCR-E et E-BCR est discutée dans cette
section, sous trois angles : la sécurité intrinsèque liée au principe de conception,
la sécurité vis-à-vis des crues extrêmes, la sécurité au séisme, et la maîtrise des
percolations.
3.2.1. Sécurité intrinsèque
Les barrages E-BCR et BCR-E sont intrinsèquement sûrs, en raison des

mécanismes de “ poussée – butée ”.
En effet, la stabilité de ces barrages est calculée en considérant que les rem-
blais (amont ou aval) exercent une sollicitation moyenne, proche de l’hypothèse des
“ terres au repos ”. Cela revient à dire que, en dimensionnement, on considère
que le barrage en BCR ne se déplace pas. Dans le cas où le dimensionnement
est mis en défaut, il y a amorce de déplacement du barrage en BCR, par excès de
sollicitation ; alors le massif d’enrochements se met en état de “ butée ” (BCR-E) ou
de “ poussée ” (E-BCR). Cela modifie substantiellement les efforts exercés par le
remblai, dans un sens favorable. L’inconvénient est le déplacement du barrage en
BCR, et l’éventuelle fissuration associée ; mais il n’y a pas libération dangereuse
de l’eau de la retenue. On obtient donc un système où la résistance est “ ductile ”
plutôt que “ fragile ” : passé le pic de résistance (à l’interface BCR-fondation), il n’y
a pas rupture brutale, mais des déplacements et des désordres qui pourront être
réparés.
Les calculs présentés au §4 permettent de quantifier cette sécurité intrin-

sèque : avec une cote de retenue à la crête du barrage, et un rocher de qualité
moyenne classique (phi=45◦ ), il subsiste un coefficient de sécurité de l’ordre de 3
avant rupture.
3.2.2. Sécurité en crue extrême et vis-à-vis des crues fortes de chantier
Par “ crue extrême ” on entend ici une crue de très grande amplitude,
éventuellement plus grande que la crue de dimensionnement de l’évacuateur, et
pouvant donc amener le niveau de la retenue au-delà des PHE voire au-delà de la
cote de crête.
567
Q.104 - R.40
Le barrage BCR-E est sûr pour les deux raisons suivantes. (1) : tant que la
crue ne dépasse pas la cote de crête, les conditions de stabilité du barrage sont
très peu modifiées. (2) : en cas de surverse limitée, le BCR tient la cote de crête et
limite donc les débits de surverse.
Le barrage E-BCR est sûr pour les mêmes raisons, et peut même supporter un
déversement avec des lames d’eau importantes, à condition que soient examinés
avec attention deux points de conception clés, en particulier si l’évacuateur est
placé sur le BCR : il faut garantir contre la formation d’une fosse d’érosion en pied
de barrage et il est nécessaire de limiter les infiltrations d’eau en crête, entre masque
et barrage en BCR (raccord déformable étanche).
Par “ crue forte de chantier ” on entend ici une crue se produisant pendant la
construction, et suffisamment forte pour saturer la dérivation provisoire, déborder
au-dessus des batardeaux et submerger le barrage en cours de construction.
Dans le cas des barrages hybrides, le BCR et le remblai d’enrochements sont

montés en parallèle. Le barrage BCR-E en cours de construction est alors très sûr :
en cas de submersion, même importante, la stabilité est assurée, même si une
partie des enrochements est emportée.
Le barrage E-BCR est plus exposé, car le masque amont n’est pas construit.
Mais le barrage est également sûr tant que le barrage n’est pas submergé, à condi-
tion que le BCR soit, globalement, plus perméable que le barrage en enrochements.
Ce qui est possible en y incluant des galeries amont-aval. Alors, le zonage des
enrochements peut assurer un rabattement suffisant de la ligne d’eau dans le rem-
blai ; et le barrage en BCR peut empêcher l’érosion régressive des enrochements.
Le comportement en cas de submersion complète reste à étudier.
3.2.3. Sécurité en cas de séisme fort
Ce rapport ne présente pas de calculs au séisme, et il s’agit d’un sujet devant

être examiné par des études complémentaires.
3.2.4. Maîtrise des percolations et des sous-pressions
Pour le barrage BCR-E, ce qui importe est de maîtriser l’ampleur des perco-
lations, et maîtriser les gradients hydrauliques au pied aval du barrage BCR, au
raccordement avec le remblai d’enrochements. Les calculs présentés au §4 mon-
trent que des tractions se développent au pied amont du barrage en BCR. Pour cela,
le voile d’étanchéité en fondation ne devrait pas être placé trop près du pied amont,
et plutôt à mi épaisseur du BCR, et un massif filtrant devrait être interposé entre les
enrochements et la fondation à l’aval du barrage en BCR.
568
Q.104 - R.40
Cela peut conduire à une augmentation des sous-pressions sous cette partie
du barrage. Cependant, cette augmentation peut être tolérée : la stabilité du barrage
BCR-E n’est pas menacée y compris dans le cas où la sous-pression maximale
(égale à la cote de retenue) se développe sous toute la section du barrage BCR.
Dans le cas du barrage E-BCR, le comportement normal est analogue à celui

des barrages CFRD : le masque assure l’étanchéité et les fuites provoquées par
les défauts éventuels du masque sont facilement drainées par les enrochements.
Le système doit être conçu de sorte que des fuites éventuellement importantes,
voire une rupture du masque, ne mettent pas en péril le BCR. La situation la plus
dangereuse est rencontrée en phase travaux, avec le risque d’un remplissage acci-
dentel tant que le masque n’est pas construit. Une conception adaptée consiste à
prévoir un zonage des enrochements et des galeries traversant le BCR, pour obtenir
une perméabilité croissante depuis l’amont vers l’aval.
3.3. EMPREINTE ENVIRONNEMENTALE RÉDUITE
L’empreinte environnementale est évaluée ici de manière simplifiée en exam-

inant deux paramètres significatifs: le bilan carbone et les besoins en sable. Pour le
calcul du bilan CO2, l’approche classique consiste à utiliser les facteurs d’émission :
c’est-à-dire la quantité de GES (gaz à effet de serre) associée à la production et
la mise en œuvre d’une unité d’œuvre, quantité de GES traduite en termes de
tonnes équivalentes de CO2 . Pour le calcul, les facteurs d’émission suivants ont été
retenus : 70 kg équivalent CO2 par m3 de BCR (augmenté de 20% pour BCR-E
et diminué de 20% pour E-BCR), 10 kg équivalent CO2 par m3 pour les enroche-
ments et 435 kg équivalent CO2 par m3 pour le béton armé du masque† . Pour la
consommation de sable, les ratios classiques sont utilisés.
Pour le bilan environnemental, il est nécessaire d’inclure l’impact de

l’évacuateur de crues et des ouvrages hydrauliques, car le béton armé a un coût
carbone élevé. A cet effet, un calcul simplifié d’ordre de grandeur est effectué‡ . Le
Tableau 3 donne les résultats de la comparaison.
† Ces facteurs d’émission ont été calculés à partir de bases de données publiques françaises,
et les ordres de grandeur suivants : ciment 600 kgCO2 /t; agrégats : 5 kgCO2 /t ; acier :
3100 kgCO2 /t ; BCR à 100 kg de ciment par m3 ; béton armé à 300 kg de ciment par m3 et
80 kg d’acier par m3 . Seuls les coûts CO2 des fournitures ont été pris en compte, et pas les
coûts CO2 de transport et mise en œuvre.
‡ On postule que les ouvrages hydrauliques correspondent à une certaine fraction du coût
du corps du barrage : 20% pour BCR, 50% pour CFRD, 30% pour BCR-E et 25% pour
E-BCR. Et que ce coût correspond essentiellement à des travaux de béton armé, dont on
connaît le bilan CO2 et la teneur en sable. Le tableau intègre un “ surcoût ” de 20% pour
excavations et traitement des fondations du barrage en BCR, plus exigeant à cet égard.
569
Q.104 - R.40
Table 3
Empreinte environnementale pour quatre types de barrages
SECTION COURANTE BCR CFRD BCR-E E-BCR
Bilan CO2 (t eq CO2 / m) 72 60 50 76
Consommation sable (t / m) 820 230 480 540
En incluant les ouvrages hydrauliques BCR CFRD BCR-E E-BCR
(exemple en ordre de grandeur)
Bilan CO2 (t eq CO2 / m) 110 11 85 100
Consommation sable (t / m) 1000 300 500 600
Ces résultats doivent bien entendu être ré-évalués pour chaque site et chaque
projet. Ils montrent les bonnes performances CO2 de l’option BCR-E et ils mettent
en évidence l’importance des ouvrages hydrauliques.
4. SIMULATIONS NUMÉRIQUES
4.1. DESCRIPTION DE LA MODÉLISATION
Les simulations numériques portent sur un barrage de hauteur 50 m, avec

un chargement hydrostatique jusqu’à la crête. Pour ces calculs, la géométrie du
barrage est définie comme suit : largeur en crête 6 m, talus du BCR à 0,4H/1V et
talus des enrochements à 1,5H/1V.
Le tableau ci-dessous précise les paramètres des matériaux intégrés dans la

modélisation :
Table 4
Paramètres géotechniques pour la modélisation
SUBSTRATUM BCR (*) Enrochements
Poids volumique (kN/m3 ) 25 23 21
Loi de comportement Mohr-Coulomb Mohr-Coulomb Hardening Soil Model (HSM)
Module (GPa) 5 10 01
Coefficient de Poisson 0,2 02 02
Résistance à la traction Rt (MPa) 0.3
Cohésion effective C’ (kPa) 200 500 1
Angle de frottement Phi’ (◦ ) 45 45 40
Angle de dilatance Psi (◦ ) 5
(*) avec anisotropie : la résistance mentionnée est selon les joints horizontaux.
A l’interface entre le BCR et les enrochements, la cohésion effective est réduite

à 0, et l’angle de frottement à 40◦ .
570
Q.104 - R.40
Les pressions d’eau en situation normale d’exploitation sont prises en compte

de la manière suivante. Pour BCR-E, les pressions sont calculées par le modèle,
sans considérer le drainage ; en particulier, et de manière prudente, les sous-
pressions à l’interface entre BCR et fondation décroissent linéairement depuis la
pleine pression au pied amont jusqu’à zéro au pied aval. Pour E-BCR, le drainage
est considéré comme étant parfait : enrochements et BCR restent secs.
4.2. RÉSULTATS DE LA MODÉLISATION, BARRAGE BCR-E
Les résultats du calcul illustrent le comportement du barrage BCR-E, en soulig-

nant les différences de sollicitation entre retenue vide et retenue pleine. Très peu de
modifications dans les sollicitations sur les enrochements ; en revanche, les con-
traintes de compression dans le BCR, en pied de barrage, sont fortes notamment en
fin de construction, et des tractions se développent au pied amont à retenue pleine.
Cela peut conduire à une plastification (ou rupture en traction) de la fondation si elle
n’est pas suffisamment résistante.
Fig. 6
Résultats des calculs de stabilité pour BCR-E, en situation normale d’exploitation.
Stability simulation results for RCC-E, for normal operation conditions.
571
Q.104 - R.40
Cependant, la sécurité du barrage n’est pas menacée, car, comme le montrent

les modèles, le développement de zones plastifiées ou de rupture en traction dans
la fondation reste local ; à l’échelle globale, le coefficient de sécurité du barrage
demeure fort, comme l’illustrent les résultats ci-dessous.
Fig. 7
Mécanismes potentiels de rupture et coefficients de sécurité pour BCR-E sous
sollicitations normales d’exploitation.
Possible failure modes and associated safety factors for RCC-E in normal
operation conditions.
Seuls les modes (1) et (2) conduisent à la ruine de l’ouvrage. Dans ces deux
modes, le coefficient de sécurité est bien supérieur aux exigences classiques,
et cela grâce au comportement mixte. Dans le mode (1), la mise en butée des
enrochements empêche la rupture en glissement dans le BCR ou à l’interface BCR-
fondation. Dans le mode (2), le poids apporté par les enrochements joue pleinement
son rôle stabilisateur.
Dans le mode (3), la hauteur de la recharge est diminuée en raison du glisse-

ment. Mais elle continue à contribuer à la résistance du barrage, comme le montrent
ces calculs simplifiés:
572
Q.104 - R.40
Fig. 8
Résultats des calculs de stabilité pour BCR-E, avec remblai partiellement emporté.
Stability simulation results for RCC-E, with partial rockfill only.
En définitive, le barrage BCR-E apparaît sensiblement plus sûr qu’un barrage-

poids au profil classique à 0,8H/1V. Cependant, il sollicite davantage la résistance
locale de la fondation, aux pieds amont et aval du BCR, et de manière sensiblement
différente à retenue pleine ou vide. Cela peut être un point de vigilance (par effet
de fatigue sous cycles) pour les retenues qui sont amenées à subir des variations
importantes et fréquentes de la cote de retenue
4.3. RÉSULTATS DE LA MODÉLISATION, BARRAGE E-BCR
Les résultats du calcul illustrent le comportement du barrage E-BCR, en soulig-

nant les différences de sollicitation entre retenue vide et retenue pleine. Sont illustrés
les directions des contraintes principales, et les contraintes à l’interface barrage-
fondation pour le BCR. Le résultat principal est qu’il y a très peu de différences
entre les situations de retenue vide et de retenue pleine. Les contraintes dans le
BCR restent modestes, moins de 2 MPa.
573
Q.104 - R.40
Fig. 9
Résultats des calculs pour E-BCR, en situation normale d’exploitation.
Simulation results for E-RCC, for normal operation conditions.
Fig. 10
Mécanismes potentiels de rupture et coefficients de sécurité pour E-BCR sous
sollicitations normales d’exploitation.
Possible failure modes and associated safety factors for E-RCC in normal
operation conditions.
574
Q.104 - R.40
4.4. INTERPRÉTATION DES RÉSULTATS DE MODÉLISATION
Les modèles illustrent le comportement très sûr en situation normale

d’exploitation. Le cisaillement mobilisé, à l’interface rocher – barrage en BCR
n’excède pas 20◦ à 22◦ , pour une cote de retenue très haute (au niveau de la
crête du barrage). Et la rupture complète n’intervient que dans le cas d’une divi-
sion par 3 des caractéristiques mécaniques de la fondation par rapport à celles du
rocher standard. Cela montre que les barrages hybrides BCR-E ou E-BCR sont peu
exigeants en ce qui concernent les conditions de fondation – tant que la fondation
n’est pas trop compressible§ .
Dans le cas BCR-E, la sollicitation du BCR s’inverse entre la phase de fin de

travaux et la phase de mise en eau, ce qui s’observe notamment sur le diagramme
des contraintes dans le BCR. A retenue haute, il peut y avoir développement de
tractions en face amont du barrage BCR, si le barrage BCR est conçu avec un
talus aval plus raide que 0,4H/1V. Cela conduit à devoir placer le voile d’injection
un peu en retrait à l’intérieur du barrage. Dans le cas de BCR-E, il est utile de
prévoir un BCR capable de résister à quelques centaines de kPa de contraintes de
traction. Il importe également de s’assurer du bon comportement de la fondation,
qui est localement fortement sollicitée ; en particulier si la retenue connaît des cycles
fréquents et importants de marnage.
Dans le cas E-BCR, la sollicitation du barrage en BCR varie peu avec le mar-
nage de la retenue. La fondation reste entièrement comprimée dans toutes les
configurations. Les contraintes dans le BCR restent très modérées, et il n’y a pas
besoin de mobiliser de résistance à la traction : les spécifications sur le matériau
BCR peuvent être assouplies.
L’interaction entre les enrochements et le BCR joue de manière différente

selon que les enrochements se trouvent à l’amont ou à l’aval. A l’amont, la poussée
des enrochements est déstabilisatrice pour le BCR ; à l’aval, la poussée est stabil-
isatrice. Cela peut être pris en compte pour la conception. Lorsque l’enrochement
est à l’aval du BCR (cas BCR-E), le compactage doit être modéré, afin de limiter
le frottement interne, pour maximiser la poussée stabilisatrice du remblai. Lorsque
l’enrochement est à l’amont du BCR (cas E-BCR), le compactage doit être intense,
afin de maximiser le frottement interne, pour limiter la poussée du remblai qui est ici
déstabilisatrice ; de plus cela va aussi dans le bon sens pour limiter les mouvements
sous le masque ou la géomembrane à la mise en charge de l’ouvrage.
La résistance à des sollicitations accidentelles fortes (séisme important, crue

de chanter venant submerger le barrage E-BCR) restent à examiner.
§Lorsque la fondation est érodable, il faut également maîtriser le risque d’érosion interne,
notamment par l’interposition de filtres adaptés.
575
Q.104 - R.40
REFERENCES
[1] HOONAKKER, M. Les vieux barrages-poids en maçonnerie, Journée CFBR

Roanne 28/01/2017.
[2] CIGB, Evolution des barrages – influence des coûts, Bulletin 83, 1992.
[3] CIGB, Le barrage poids un barrage d’avenir, Bulletin 117, 2000.
[4] Chinese National Committee on Large Dams, Dam Construction by the

Chinese People, 1979.
576
Q.104 - R.42
-------
-------
CALIBRATION OF FE-MODEL WITH MEASURED BEHAVIOR OF AN

EXISTING CONCRETE ARCH DAM (∗)
S. BJØNNESS, N. RAKSTAD, V. KRATHE & T. KONOW

Dr.techn. OLAV OLSEN
NORWAY
SUMMARY
This report describes the general process involved in using data from instru-
mentation in order to verify the capacity of an existing concrete arch dam. The data
from the instrumentation is used as a basis for load definition and as a calibra-
tion/verification of a global finite element (FE) model. Temperature loading is very
important when assessing the capacity of arch dams. The instrumentation of dams
proves to be an essential contribution in order to establish an FE-model that repre-
sent the actual behavior of the structure. A non-linear analysis is required to obtain a
realistic model. In this paper it has been shown that it is possible to construct longer
time series from short time series using both physics and machine learning on mea-
surements from instrumentation. It is often difficult to document sufficient safety for
old structures, and thereby demonstrate that they fulfill today’s safety standards.
Better knowledge of the structural behaviour based on instrumentation and corre-
sponding calibration of the design model can then be the only alternative to prove
that the structure meets the current design and safety requirements
∗ Calibrage de modèles d’éléments finis à l’aide de l’instrumentation et analyse de la capacité
de barrage-voûtes
DOI: 10.1201/9781003211808-38 577

Q.104 - R.42
RÉSUMÉ
Ce rapport décrit le processus général impliqué dans l’utilisation des données

de l’instrumentation afin de vérifier la capacité d’un barrage-voûte existant. Les
données de l’instrumentation sont utilisées comme base pour la définition de la
charge et comme étalonnage/vérification d’un modèle global aux éléments finis
(EF). Le chargement de la température est très important lors de l’évaluation de
la capacité des barrages-voûtes. L’instrumentation des barrages s’avère être une
contribution essentielle afin d’établir un modèle qui représente le comportement réel
de la structure. Une analyse non linéaire est nécessaire pour obtenir un modèle
réaliste. Ce rapport démontre qu’il est possible de construire des séries de temps
plus longues à partir de séries de temps courtes en utilisant à la fois la physique
et l’apprentissage automatique sur les mesures de l’instrumentation. Il est souvent
difficile de démontrer un niveau de sécurité suffisante pour les anciennes structures
et de démontrer ainsi qu’elles remplissent les normes de sécurité d’aujourd’hui.
Une meilleure connaissance du comportement structurel basé sur l’instrumentation
et l’étalonnage correspondant du modèle de conception peut alors être la seule
alternative pour prouver que la structure répond aux exigences actuelles en matière
de conception et de sécurité.
1. INTRODUCTION
This paper describes the general process involved in using data from instru-
mentation in order to verify the capacity of an existing concrete arch dam. The
data from the instrumentation is used as a basis for load definition and as a
calibration/verification of a global finite element (FE) model.
The geometry of the arch dam is shown in Fig. 1. The geometry of the model
is based on original drawings and a recent laser scan of the entire dam, verifying
the geometry.
During the first year of use leakages were observed towards the foundation of
the arch. This was sealed off with the injection of grout the following year. Seventy
years later some of the same leakage was observed. This was sealed again a few
years ago with new grouting underneath the foundation. Simultaneously, the work
of instrumentation, and analysis of the dam was initiated in order to improve the
documentation of the dam.
When looking at the deformation and temperature data, it was clear that the
dam was highly influenced by temperature load. When emptying and filling the dam,
the change in structural response was small compared to the change caused by the
varying temperature.
578
Q.104 - R.42
Fig. 1
Overview of FE-model and placement of instrumentation.
The temperature both in the water and behind the isolation wall for this dam
is constant across the height of the dam unlike most dams which has a varying
temperature distribution over the upstream face.
The sensitivity for temperature load and local marginal capacity induced a
need for precise analysis and the use of measured data. External temperature
data, together with both thermodynamic analysis and machine learning, was used
to estimate a longer time series from the relatively short data that was available.
The non-linear behavior of the material and the foundation can be found
based on an iterative analysis of the global model. This analysis is performed using
ShellDesign, which is a design tool for reinforced concrete, owned and developed
by Dr.techn. Olav Olsen.
2. INSTRUMENTATION
The dam is equipped with instruments for measurements of deformation in

three directions in four vertical sections (named D1 to D4), measurements of tem-
perature in the water reservoir (T1) and behind the insolation wall (T2). In addition,
there are four pressure gauges that are installed underneath the foundation down-
stream of the arch. Reference is made to Fig. 1 for description of instrumentation
locations. Deformation and temperature data have been collected since 2017 and
the pressure gauges since 2019.
Fig. 2 shows the results from a time period of 3 months in which the reser-
voir was emptied and filled, as shown in the continuous red line. As shown in the
graph a modest temperature change of about 50 C has a similar effect on horizontal
deformation as emptying the reservoir.
579
Q.104 - R.42
Fig. 2
Data from instrumentation during emptying (1) and filling (2) of the reservoir 2019.
Most dams typically have a vertical variation of temperature, which is not the
case for this dam. The reservoir is small with an upstream powerplant operating
almost non-stop throughout the year. This results in a constant water temperature
of about 10 C during the winter period and no variation in temperature vertically in
the reservoir. Behind the insulation wall on the downstream face of the arch, there
is also a relatively uniform temperature distribution, both horizontal and vertically.
The instrumentation serves two main purposes in the analysis described in

the following; 1) give input to the temperature loading and 2) calibrate the stiffness
of the model. The problem with the former was the relatively short time period of
each data series and the absence of a measurement of the air temperature.
2.1. AIR TEMPERATURE
Time series of the temperature at the dam site was available for the period
1955 to 1992. Because this series only had one measurement, registered at the
morning of each day, this gave a good indication of the minimum temperature but
did not give a good representation of the maximum temperatures. It turned out that
the old series correlated well with temperature measurements in lower parts of the
valley in which the dam is located.
These had measurements three times a day from 1954 onwards which made
it possible both to compare the time series to the old one from the dam site and to
use the series in combination with the instrumentation on the dam. The last 50-years
was used for the temperature load. The resulting curve showed a total temperature
difference just short of 700 C (50-year return period) which corresponds well with the
temperature given in Eurocode [1].
580
Q.104 - R.42
2.2. EXTENDING THE TIMESERIES
Two independent methods were used for extending the time series; physics
and machine learning with a deep neural net. For the water temperature, the data
from the instrumentation was used to scale the temperature during the summer
according to how warm the summer period where (the summation of days multiplied
with the temperature) to make a 50-year time series.
For the temperature behind the insulation wall, a one-dimensional diffusion

equation is solved with the air and water temperatures as input [4]. The results
where then calibrated and verified against the instrumentation data.
The deep neural net was trained on the data from 2017 onwards. Based on the
resulting model the temperatures in the water and behind the insulation wall were
estimated from the air temperature in the period 1970-2020. The results from the
two methods are compared to the instrument data in Fig. 3 and Fig. 4. The figures
present the last ten years of the series. The difference between the two methods
are similar for the rest of the 50 years.
Fig. 3
Comparison of temperature behind the insulation wall (last 10 years).
Fig. 4
Comparison of water temperature (last 10 years).
581
Q.104 - R.42
3. TEMPERATURE ANALYSIS
The complexity of the geometry and the importance of the temperature load
lead to the use of thermodynamic analysis. Either steady-state or a transient analysis
is possible. For a steady-state analysis, the temperature of the structure surface must
be estimated as the use of the extreme air temperature would be very conservative.
For that reason, a transient thermodynamic analysis was run for the entire
series of 50 years to find the temperature distribution over time in the concrete. The
temperature series (air, water/foundation and behind the insulation wall) were cou-
pled with its corresponding surface on the FE-model. The thermodynamic analysis
was done using Abaqus.
The figure below shows the extreme temperature averaged over three, seven,
14 and 21 days. The transient analysis showed that the actual distribution over time
could be compared with a steady-state analysis with the temperature averaged over
7-14 days.
Fig. 5
Tmin and Tmax when averaging over a longer time period (50-years return periods).
4. NON-LINEAR ANALYSIS AND CALIBRATION
The program ShellDesign is used in order to analyze the non-linear behaviour

of the dam. The program is based on the extensive development and project experi-
ence accumulated through a large number of both land-based and offshore concrete
projects over the years
582
Q.104 - R.42
The foundation is modeled with special elements that has no tensile capacity,
which means that there can be uplift of the structure. The compressive stiffness
of the elements represents the stiffness of the rock beneath the foundation. Rein-
forced concrete cross sections also behaves in a non-linear way and reinforced
structures are able to distribute forces from areas with high utilization to areas with
lower utilization. The workflow of this iterative process in ShellDesign is shown in
Fig. 6.
Fig. 6
Overview of global non-linear analysis in ShellDesign.
The model in Abaqus and ShellDesign had corresponding element mesh, and
the resulting temperature obtained from Abaqus (a small set of extreme cases) was
used as loads for the non-linear structural analysis in ShellDesign.
From the available data several time periods were selected of varying length.
We compared the relative displacements from the start until the end of the period
between the non-linear FE-analysis and the deformations recorded by the instru-
mentation. The two results are compared against each other for one time period in
Fig. 7.
583
Q.104 - R.42
Fig. 7
Comparison of change in deformations over time between FE-model (dotted lines)
and measured behaviour from instrumentation (continues lines).
In the same manner, as a neural net is trained, the calibration of the FE-model
must be calibrated on a subset of the total number of time periods. In that way, the
chance of overfitting is reduced, and the remaining time periods will serve as a test
of the calibration quality.
The resulting model with calibrated stiffness of the concrete and foundation
is used to calculate the design capacity of the concrete vault. It is important to
note that this stiffness corresponds to the serviceability limit state (SLS). For the
ultimate (ULS) and accidental (ALS) limit states, larger deformations and thus a
softer behavior will occur.
The curves (FE model versus measurements) for the two outermost sections
(INC1 and INC4) do not correspond as well as those for the two sections in the
middle. INC1 are located right next to the massive gravity section of the dam. This
part is defined with linear cross-sectional stiffness and does thus behave in a stiffer
manner in the analysis. Another contribution to the difference can be the recent
injection on the upstream side in this area.
INC4 is located against a vertical steep abutment. This part is not visible on
the laser scan and the FE-model geometry of this area is based on drawings alone.
Other effects that would contribute to local differences between the model and the
instrumentations are the possible presence of rock anchors in some parts, varying
stiffness due to the concrete mix, concrete work or the temperature at with each
part were casted.
584
Q.104 - R.42
5. DESIGN AND CAPACITY
The capacity of the concrete arch is checked in accordance with Eurocode 2

[2]. There is no specific standard for dams in the Eurocode system, and especially
the empiric shear formulation is not primarily made for the design of massive dams.
For that reason, the capacity is also verified using the "Modified Compres-
sion Field Theory" (MCFT) developed at the University of Toronto [3]. This method
combines both in-plane forces and transverse shear forces in the non-linear sec-
tional response. Tension stiffening is included through average stresses by adding
post-cracking tensile stresses to the stress-strain curve of the concrete. Modelling
tension stiffening by averaged stresses requires the inclusion of a “crack check”.
The crack check ensures that no concrete tensile stresses are necessary to obtain
equilibrium at the location of a fully developed crack.
The two methods gave more or less similar results, indicating that the struc-
tural capacity of the vault does not fulfill the requirements. Carbon fiber can be an
attractive solution to increase the capacity. The difference of the methods was that
the MCFT approach unified sectional response and gave a modest requirement of
strengthening in all directions, while Eurocode indicated a quite high increase in one
direction because of the way transverse shear forces are handled.
Generally, the utilization of the capacity is low both for concrete and reinforce-
ment, except the vertical reinforcement on the downstream side shown in Fig. 8. The
capacity for the other reinforcement layers and shear reinforcement is only highly
utilized towards the boundary of the arch. The utilization of the shear capacity is
shown in Fig. 9.
Fig. 8
Utilization in ULS of vertical reinforcement downstream (Eurocode 2 [2]). Blue
color represents no utilization, red full utilization of capacity.
585
Q.104 - R.42
Fig. 9
Utilization in ULS with respect to shear (Eurocode 2 [2]) Blue color represent no
utilization, red full utilization of capacity.
6. FINAL REMARKS
Temperature loading is very important when assessing the capacity of arch

dams. The instrumentation of dams proves to be an essential contribution in
order to establish an FE-model that represent the actual behavior of the struc-
ture. A non-linear analysis is also required to obtain a realistic model. In this
paper it has been shown that it is possible to construct longer time series from
short time series using both physics and machine learning on measurements from
instrumentation.
It is often difficult to document sufficient safety for old structures, and thereby
demonstrate that they fulfill today’s safety standards. Better knowledge of the struc-
tural behaviour based on instrumentation and corresponding calibration of the design
model can then be the only alternative to prove that the structure meets the current
design and safety requirements.
REFERENCES
[1] NS-EN 1991-1-1-5, Eurocode 1, Part 1-5: Thermal actions.
[2] NS-EN 1992-1, Eurocode 2, Design of concrete structures, Part 1-1: General
rules and rules for buildings.
[3] E. C. Bentz, “Sectional Analysis of Reinforced Concrete Members,” University

of Toronto, 2000.
586
Q.104 - R.42
[4] H. P. Langtangen, “The 1D diffusion equation,” Center for Biomedical Com-

puting, 2015. [Internett]. Available: http://hplgit.github.io/num-methods-for-
PDEs/doc/pub/diffu/sphinx/._main_diffu001.html. [Nov 2019].
[5] Tensorflow, https://www.tensorflow.org/
[6] ShellDesign, Dr.techn. Olav Olsen.
587
Q.104 - R.43
-------
-------
HEIGHTENING OF DHANIKHARI CONCRETE DAM (∗)
Y.K. CHAUBEY
Director Technical,
NHPC LTD, Sec-33, Faridabad, Haryana
Sankhadip CHOWDHURY
D & E Division,
NHPC LTD, Sec-33, Faridabad, Haryana
Mukesh BHORIA
Kiru HEP, CVPP(P)L, Chenab Nagar Sector-II, Kishtwar, J&K
INDIA
SUMMARY
Aimed at mitigating the shortage of potable water requirements of South

Andaman area of Anadaman & Nicobar Island (India), the scheme to augment the
live storage capacity of existing Dhanikari reservoir by 3.23 mcm through raising
the height of dam by 5 meters was an important project executed by NHPC Ltd.
The 132 m long existing dam across the Dhanikhari nallah with dam top EL 62.5 m,
consisted of eight non-overflow blocks and two spillway bays on a central overflow
block of ±32 m height above deepest foundation, impounding a reservoir with gross
storage capacity of 4.61 mcm at FRL 61 m. The entire water supply requirement of
Port Blair is met from this reservoir.
Detail study of augmentation of the existing storage capacity of Dhanikhari

reservoir, included evaluation of heightening of dam by two methods i.e. Prestressed
Anchors and Backing Concrete methods. Backing concrete method was adopted
∗ Surélévation du barrage en béton de Dhanikhari
588 DOI: 10.1201/9781003211808-39

Q.104 - R.43
due to its techno economic advantage. Several issues in this method were addressed
such as effect of permanently built-in stress on new concrete, effect of shrinkage in
new concrete on old concrete, restraining effect of old concrete on new concrete,
bonding between old & new concrete, enhancement of spillway capacity and seismic
retrofitting, etc.
While heightening the dam, the existing overflow section was partially demol-
ished, to construct a new spillway with augmented spillway capacity of 225 cumecs.
The heightening was completed on December 2014.
RÉSUMÉ
Le projet qui visait à augmenter de 3,23 millions de mètres cubes la capacité de

stockage utilisable du réservoir existant de Dhanikari en augmentant la hauteur du
barrage de 5 mètres et dont l’objectif était d’atténuer la pénurie d’eau potable dans
la zone sud d’Andaman sur l’île d’Andaman et de Nicobar (Inde), était un projet
important exécuté par NHPC Ltd. Le barrage actuel de 132 m de long construit
sur le Dhanikhari nallah a une cote de 62,5 m au sommet. Il est composé de huit
blocs sans débordement et de deux baies de déversoir sur un bloc de débordement
central qui se trouve à ± 32 m de haut au-dessus de la fondation la plus profonde. La
capacité de stockage brute du réservoir est donc de 4,61 millions de mètres cubes
à niveau de réservoir plein à 61 m. Ce réservoir fournit l’ensemble des besoins
d’approvisionnement en eau de Port Blair.
Une étude détaillée de l’augmentation de la capacité de stockage du réser-

voir existant de Dhanikhari comprenait l’évaluation de la surélévation du barrage
par deux méthodes, à savoir la méthode utilisant des tirants d’ancrage–s précon-
traints et celle utilisant des renforts de béton. La méthode de renfort du béton a
été adoptée en raison de son avantage technico-économique. Plusieurs problèmes
ont ainsi été résolus à l’aide de cette méthode tels que l’effet de la contrainte per-
manente existant–e dans le nouveau béton, l’effet du retrait du nouveau béton sur
l’ancien béton, l’effet de retenue de l’ancien béton sur le nouveau béton, l’adhésion
entre l’ancien et le nouveau béton, l’amélioration de la capacité du déversoir et la
réhabilitation séismique, etc.
Lors de l’élévation du barrage, la section de déversoir existante a été partielle-

ment démolie, pour construire un nouveau déversoir ayant une capacité augmentée
de 225 m3 /s. La surélévation a été achevée en décembre 2014.
1. INTRODUCTION
Port Blair is the capital town of A&N islands, India. Dhanikhari dam is con-
structed across Dhanikhari nallah in the year 1973 by APWD to cater the drinking
water needs of the Port Blair town. The existing Dhanikhari water supply scheme
589
Q.104 - R.43
comprised a 32.25 m high (from deepest foundation), 132 m long concrete gravity
dam constructed across the Dhanikhari nallah with dam top level at EL 62.5 m, and
a storage capacity of 4.61 M cum at an FRL of EL 61 m. In order to augment the
water supply to cater the rising demand levels of future, after detailed study [1], it
was found that by raising the height of existing Dhanikhari Concrete Dam by 5 m
present live storage capacity can be increased by nearly eighty percent of existing
storage capacity. The additional storage was suitable to cater the projected drinking
water needs for the forthcoming 15 years.
Two methods of dam heightening i.e. Pre-stressed Anchorage method and

Backing Concrete Method was explored. Due to construction constraints and non-
availability of indigenous specialized agencies the former method has not been
pursued. During the detailed study the existing condition of dam was found good
and suitable for increasing the height. Subsequently following was deliberated,
• Various design parameters of the dam and rock parameters of foundation and
abutments were verified.
• Hydrological data of Dhanikhari Nallah and its catchment was examined to
arrive at the design flood for increased height of dam.
• The geological and geotechnical studies of dam area have been carried out to
ascertain the strata underlying the dam and rock parameters of dam foundation
and abutments.
• The reservoir geology has also been studied.
• The seismicity of the region has been studied and earthquake parameters of
the Dhanikhari dam have been evaluated.
• Construction material survey and testing have been carried out for identifica-
tion and suitability of construction material.
• A preliminary environmental study has been conducted to evaluate the effects
on flora and fauna and other socio-economic aspects due to raising in height
of dam.
Fig. 1 Fig. 2
Downstream face of Dam Downstream face of Dam
before Heightening after Heightening
Face aval du barrage avant Face aval du barrage après
la surélévation. la surélévation.
590
Q.104 - R.43
2. DEMAND ASSESSMENT
The Dhanikhari dam was conceived and constructed for a live storage capac-
ity of 3.24 mcm at its full reservoir level of EL 58.50 m. Since the demand of
water increased, the storage level was raised by 2.0 m by providing two no. of
7.9 × 2.2 m vertical gate and thus enhancing the live storage capacity to 4.37 mcm.
But the incumbent increase in population of Port Blair town has concerned the
A & N administration on future water demand. In this regard demand assess-
ment up to AD 2010 has been carried by APWD, which indicated a deficit of
23.69 lgpd which is equivalent to an additional storage requirement of 2.31 mcm
for 215 days dry period. Thus, dam heightening of Dhanikari dam 5 m was
envisaged.
3. FIELD WORK
3.1. GENERAL
Following field investigations were carried out to fulfill the following objectives:
• Determination of the area due to additional submergence.

• Evaluation of the existing condition of the dam.
• Investigation for ascertaining rock parameter of the foundation of the existing
dam.
• Evaluation of strength and design parameters.
• Preliminary assessment of Environmental Impact due to additional submer-
gence.
• Construction Material Survey
3.2. TOPOGRAPHICAL SURVEY
Topographical survey was carried out in the entire reservoir area during the
investigation up to EL 70 m to identify any low pockets i.e. having elevation less than
EL 66 m in the reservoir rim.
591
Q.104 - R.43
3.3. INVESTIGATION ON EXISTING DAM
Investigation on existing dam has been carried out to ascertain various con-
crete and rock parameters. Drilling through foundation gallery for collecting rock and
concrete samples of the foundation were carried out at both Overflow blocks and
Non-Overflow blocks at different elevations. Concrete core from upstream & down-
stream face of the dam were collected for ascertaining the compressive strength
of the present concrete. Pullout test was conducted inside the foundation gallery to
evaluate the bond strength between the concrete and rock of existing dam. Following
laboratory tests were conducted on concrete and rock samples,
• Modulus of Elasticity of concrete core,

• Poisson’s ratio of concrete core,
• Shear strength parameters of interface between concrete & rock,
• Compressive strength of rock,
• Compressive strength of foundation concrete sample,
• Tensile strength at concrete and rock joint,
• Direct shear strength test on old concrete and new concrete.
Instrumentation in the form of 4 Nos. Vibrating type electrical piezometers were

installed in the foundation to record the uplift parameter of the existing dam. One
Vibrating wire type electrical thermometer was also placed in the downstream face
of right non overflow block to record the temperature of the face of dam body and its
behavior with respect to changing ambient temperature, which is a vital parameter
in finalizing the suitability of backing concrete method.
It was seen from the plots that the piezometric head downstream of grout
curtain is about 1/10th to 1/8th of the corresponding head in reservoir. This uplift
is considered very small compared to the design value considered in the stability
analysis. This low uplift is considered due to the ingress of fine silt in the rock joints
and a blanket on the floor of the reservoir.
3.4. CONSTRUCTION MATERIAL SURVEY
The designated quarry for production of coarse & fine aggregates for use in
South Andaman is located in carbines Cove Area, situated at about 16 km from the
Dam site. Development of a quarry near Dhanikhari Dam was very remote possibility
due to MOEF restriction & Hon’ble Supreme Court order. For the raising of height of
the existing dam, sufficient quantity, around 2.5 lakh cum of material for the produc-
tion / processing of coarse and fine aggregate is available from the rock quarry. The
physical test results of the quarry material indicate that the material is suitable for the
production of concrete for wearing as well as non-wearing surface. The petrographic
analysis indicated aggregates are Trachy Basalt, which is a volcanic igneous rock.
592
Q.104 - R.43
4. HYDROLOGY
The hydrological analysis has been carried out with a view to,
• Determine the additional reservoir capacity for a raise of 5 m in dam height.

• Review the spillway design flood.
Reservoir Elevation Area-Capacity curve has been prepared based on the

contour maps received from site in the scale of 1:2000 at 2 m contour interval. The
dam envisages a gross storage of 4.61 M cum at maximum water level of 61 m and
the difference between maximum water level and minimum water level is 14.5 m.
Hence as per IS 11223:1985, the dam is to be designated for Standard project Flood
(SPF). SPF has been found by convoluting the 1-day Standard Project Storm (SPS)
equal to 40.34 cm on the unit hydrographs computed using Clark’s, Synder’s and
Traingular methods. Base flow of 5.0 cumec and infiltration loss at a constant rate
of 2.0 mm/hr is taken. Using these parameters, the various values of 1-day SPF are
obtained. The value of design flood, i.e. 1-day SPF, recommended is 225 cumec
based on the developed design flood hydrographs. The 1-day SPF hydrograph is
then routed through the reservoir by keeping the impinging level equal to FRL i.e.
EL 66 m. After routing, a maximum level of 66.17 m and maximum outflow of 193.93
cumec was obtained.
5. ENGINEERING GEOLOGY
The bedrock constituting the Dam site comprises of mainly buff to grey, fine
to medium grained, massive sandstone with minor bands of well bedded flaky, very
fine grained sandstone or siltstone. The strike of the stratification is ENE-WSW with
dips of 45 to 67o generally towards the left abutment. There is a gradual swing in
the bedding direction from due south in the left bank to SE in the right banks d/s of
the bend. Surface indications of a minor local fault parallel to the dam axis with fault
plane dipping steeply d/s, have been observed 56 m d/s of the Dam axis.
Eight nos. bore holes, 3-5 m deep in the dam foundation, and about 2 m into
bedrock has been done in the drainage gallery which generally intercepted grayish
green fine grained sandstone exhibiting core recovery from 62-90% and RQD values
ranging from 50-80%.
Engineering properties viz. density, crushing strength and triaxial shear

strength were determined by laboratory tests. Laboratory test revealed UCS values
of sandstone ranging from 23 MPa to 67 MPa whereas a single sample of shale has
indicated UCS value of 55.8 MPa.
The project area has been shown in zone V of the seismic zoning map of India.
The maximum value of peak ground acceleration (PGA) worked out to be 0.51g. The
593
Q.104 - R.43
report on the site-specific earthquake parameters prepared by the Department of

Earthquake Engineering, IIT Roorkee.
6. DESIGN AND ENGINEERING ASPECTS
6.1. THE EXISTING DAM
The existing dam consists of eight non-overflow blocks and one overflow block.
The design flood of spillway was considered as 141.5 Cumecs.
6.2. SELECTION OF METHOD OF HEIGHTENING
Two possible methods were identified to achieve the objective of raising the
dam. These methods are:
• Use of permanent pre-stressed anchors,

• Adding additional concrete on downstream face of dam to achieve a cross-
section corresponding to increased height.
Due to construction constraints and non-availability of indigenous specialized

agencies the former method was not pursued.
6.3. FINALLY SELECTED METHOD – BACKING CONCRETE METHOD
Backing concrete method has been adopted with a considerable degree of suc-
cess in heightening and strengthening of gravity dams worldwide. However following
principal issues were deliberated during selection of this method,
• The new concrete placement, if done while the old structure is under load,
deformed and in a state of stress, the placing of new concrete on the down-
stream face will tend to retain and permanently build-in these stresses. To
minimize this affect, the reservoir is kept preferably at MDDL during place-
ment of new concrete so that the unfavorable stresses due to high water level
are not locked in.
• The shrinkage of the new concrete induces un-favorable pre-stressing in the
old concrete on one hand and restraining effect of old concrete causes stresses
in new concrete on the other. To minimize this effect various measures such as
using low heat cement and controlling the rise of temperature of new concrete
by artificial cooling.
• Method to ensure essential monolithic character of the completed structure.
594
Q.104 - R.43
6.4. ADOPTED SECTION
In the existing arrangement, top of the dam was at EL 62.5 m with 5 m top
width. While heightening the dam, the level of the dam top has been raised by 3.5 m
at EL 66.00 m which corresponds to the maximum water level after raising. The
dam top is kept at EL 67.5 m by a thick mass concrete parapet to account for free
board of 1.5 m which is considered sufficient. The existing overflow section was to
be demolished up-to EL 56.00 m and replaced with new concrete having modified
spillway profile. 3 nos. opening size 3.6 m (w) X 3.6 m (h) are provided in the overflow
block which discharge into the common chute. 3 nos. vertical gates operated from
EL 66.00 m are provided to regulate the discharge from spillway.
Fig. 3 Fig. 4
View of foundation under Dam Heightening under progress
extended Dam toe Surélévation du barrage en cours
Vue de la fondation sous le pied
étendu du barrage
6.5. INTERFACE OF OLD CONCRETE AND NEW CONCRETE
Test had been got done to obtain the shear strength of old concrete / new
concrete interface and following values of shear strength at interface were obtained.
• Using M-25 as new concrete cohesion = 14.5kg/cm2 , ϕ = 42◦

• Using M-15 as new concrete cohesion = 8.6kg/cm2 , ϕ = 47◦
Shear stress of the order of 2 kg/cm2 was calculated using elementary beam theory.
6.6. STABILITY & EARTHQUAKE ANALYSIS
The overflow and non-overflow portions of dam were designed as gravity

sections and were checked for stability (IS 6512-1984) for the following loading
conditions:
595
Q.104 - R.43
• Load combination A (construction condition) – Dam completed but no water

in reservoir and no tail water.
• Load combination B (Normal operating condition) i.e. Full reservoir elevation,
normal day weather tail water, normal uplift (drains operative).
• Load combination C (Flood discharge condition) maximum flood pool reser-
voir elevation, tail water at flood pool elevation, normal uplift (drains
operative)
• Load combination D – combination A with earthquake
• Load combination E – combination B with earthquake but no ice.
• Load combination F – combination C, but with extreme uplift (drains inopera-
tive)
• Load combination G – combination E, but with extreme uplift (drains inopera-
tive)
Stresses, forces and moments due to static forces i.e. weight of dam, reservoir
hydrostatic force, uplift force (both normal uplift and extreme uplift) were calculated
for A, B, C, F load combinations.
The tensile stresses in the dam in earthquake (DBE conditions) were found
significantly low compared to its strength. For the dam site, Earthquake Engg.
Department, IIT Roorkee, has recommended 0.51g and 0.255g as peak ground
acceleration (PGA) corresponding to MCE and DBE conditions.
It was found that Dhanikhari dam in its raised form incorporates all necessary
design details for an earthquake resistant design and excellent behaviour of this
dam after raising should be expected during earthquakes in future.
6.7. SPILLWAY AND ENERGY DISSIPATION SYSTEM
The hydraulic design of Spillway and Energy Dissipation System was firmed
up through hydraulic model studies
Spillway had its overflow profile conforming, as nearly as possible, to the profile
of the lower nappe of a ventilated jet of water issuing through an orifice. Spillways
openings were provided below MWL for considerations of increasing and regulating
storage of flood discharge.
The design flood for the spillway worked out as 225 Cumecs. The discharging
capacity of 2 Bays of sizes 4.5m(W ) × 2.5m(H ), corresponding to reservoir at EL
66.00 M, was computed as 192 cumecs.
The existing stilling basin length was increased by 9 m and the height of the
side increased by 1.5 m to accommodate the spillway discharge.
596
Q.104 - R.43
6.8. TEMPERATURE STUDIES FOR MASS CONCRETE
Peak temperature rises in mass concrete, needed to be controlled so that

minimal thermal stresses were induced in the old concrete due to placement of new
concrete. The average ambient temperature, for not allowing the temperature of
concrete to rise above the stable temperature of concrete, was 27◦ C.
Many measures were taken to restrain the net temperature rise to acceptable
levels like placement temperature kept as 12◦ C, using Portland – Pozzolana cement.
Also, maximum aggregate size was fixed as 80 mm to minimize segregation problem.
A study was conducted through FEM with idealized construction schedule

wherein concrete upto El. 55 m was done in season one and concrete above El.
55 m was done in second season. A lift height of 1.5 m has been considered in the
model with a gap of 72 hours in between two lifts.
Following inputs were considered in FEM analysis
Material properties for thermal analysis
• Unit weight (ρ): 2400kg/m3 (concrete) and 2500kg/m3 (rock)

• Specific heat (C): 0.85kJ/kg/◦ C (concrete and rock)
• Conductivity (k): 7.5kJ/m/h/◦ C (concrete corresponding to aggregate type
Trachy Basalt)
• Conductivity (k): 12kJ/m/h/◦ C (rocking corresponding to foundation rockmass
type Sandstone)
Cementitious materials of 210kg/m3 has been considered with following

proportions
• Cement: 157.5kg/m3
• Flyash: 52.5kg/m3 (it is assumed that heat of hydration of flyash is ca. 50 %
of that of cement and that the time development curves of cement and fly ash
are identical, in reality fly ash leads to a retardation of the heat of hydration
process)
• Equivalent cement content (cc ): 157.5 kg/m3 (cement) plus 50% of 52.5kg/m3
flyash (this corresponds to an equivalent cement content of 26.25kg/m3 ) =
183.75kg/m3
Maximum adiabatic temperature rise computed by using equation (1) below

according to IS 14591 [2] gives To = 28.3◦ C (considering heat of hydration of cement
at 28 days qc = 314 kJ/kg).
cc qc
To = (1)
ρC
597
Q.104 - R.43
The adiabatic temperature rise curve, T(t), is determined using equation (2)
T (t) = To (1 − e−mt ) (2)
where m = 0.025/h
Rate of heat of hydration of mass concrete is determined through equation
dT (t) cc qc dT (t)
h(t) = ρC = (3)
dt To dt
The heat loss is modeled through a convective boundary. The Convection

coefficient (hc ) between, concrete and air and rock and air are determined using
equation (4) suggested in ETL 1110-2-542 [3].
hc = 5.622 + 1.086vinW /m2 /◦ C (4)
◦
hc = 954kJ/m2 / C/day; where wind speed v = 5 km/h has been considered
Initial steady state temperature distribution (Fig. 5) was established for old
dam section considering ambient air and water temperature as 27◦ C and 22◦ C
respectively.
A stage construction analysis was carried out to estimate the temperature rise
in backing concrete and at the interface with old dam section.
As can be seen from the results of the analysis below (Fig. 6 to 8), the maximum
temperature attained by concrete within the lift of backing concrete is approximately
32.9◦ C and the maximum temperature at the interface of old concrete is about
28◦ C. The differential of 5◦ C between ambient temperature and maximum temper-
ature attained by concrete within the lift of backing concrete and raise of interface
temperature by 1◦ C is found acceptable.
598
Q.104 - R.43
Fig. 5 Fig. 6
Estimated initial steady state Estimated temperature distribution
temperature distribution after 3 days of casting of first lift
Distribution estimée de la Distribution estimée de la
température d’équilibre initiale température 3 jours après la coulée de
la première couche de béton
Fig. 7 Fig. 8
Estimated temperature distribution Estimated temperature distribution
after 3 days of casting of fourth lift after 3 days of casting of concrete up
Distribution estimée de la to El. 55.00 m in first season
température 3 jours après la coulée Distribution estimée de la
de la quatrième couche de béton température 3 jours après la coulée
de la couche de béton à la cote 55,00
m en première saison
Concrete placement temperature was maintained by washing of aggregate,

spraying chilled water for a reasonable time before un-loading into bins of batching
plant, using ice flakes during mixing.
599
Q.104 - R.43
Fig. 9 Fig. 10
Case-I; Third principal stress Case-I; Vertical stress
Cas-I; Troisième contrainte principale Cas-I; Contrainte verticale
Fig. 11 Fig. 12
Case-II; Third principal stress Case-II; Vertical stress
Cas-II; Troisième contrainte principale Cas-II; Contrainte verticale
6.9. RESERVOIR MANAGEMENT DURING CONSTRUCTION
In an ideal situation, placement of concrete on the downstream face of an

existing dam should be carried out when the dam is completely empty. However, in
the present case the existing water supply arrangements cannot be allowed to be
seriously disrupted during placement of concrete. Thus, concrete placement was
carried out during the months when the reservoir level is low.
According to Robert B. Jansen [4], a drawdown of 20% of total head provides

75% of the relieving effect that can be obtained with a complete reservoir draining.
Based upon practical considerations, it was proposed to carry out the concrete
placement during the period when the reservoir level is at or below EL 52.00M
600
Q.104 - R.43
(i.e. 8m less than existing full reservoir level of 60 M). The proposal was analyzed
through FEM. The stresses were observed for two cases of analysis, Case-I: The
backing of concrete was done with water level at El. 52.00 m (Fig. 9. & Fig.10.) and
Case-II: The backing of concrete was done with water level at EL 61.00 m (FRL,
Fig. 11. & Fig.12.). The dam foundation was modeled as fixed. In both the cases
the reservoir was brought down to El. 45.00 m (MDDL) after placement of backing
concrete and the stresses along lift line was noted. It was observed that in Case –I
there was no tensile stress on the downstream portion of backing concrete whereas
in Case-II the tensile stresses were notice in downstream face. The results of the
analysis is shown above in Fig. 9 to 12 (-ve is tension).
Therefore, in order to maintain a biaxial compression state in the whole dam

body reservoir levels shown in Table 1 were recommended for adoption during
construction.
Table 1
Recommended reservoir
levels during construction
Month Reservoir level (M)
March El. 52.00
April El. 50.00
May El. 48
June El. 50
6.10. SHRINKAGE ANALYSIS
The shrinkage of the new concrete induces un-favorable pre-stressing in the

old concrete on one hand and restraining effect of old concrete causes stresses in
new concrete on the other hand.
The total shrinkage of concrete depends upon the constituents of concrete,

size of the member and environmental conditions. For a given humidity and temper-
ature, the total shrinkage of concrete is most influenced by the total amount of water
present in the concrete at the time of mixing and, to a lesser extent, by the cement
content.
The total shrinkage strain is composed of two components, the autogenous

shrinkage strain and the drying shrinkage strain.
The autogenous shrinkage strain develops during hardening of concrete; the

major part develops in the early days after casting. Autogenous shrinkage is a linear
601
Q.104 - R.43
function of concrete strength. It should be considered specifically when new concrete

is cast against hardened concrete.
Hence the temporal effect of autogenous shrinkage strain was modeled

with stage construction model of dam as done for thermal analysis. The autoge-
nous shrinkage strain parameters were taken as per equation (5) according to IS
1343 [5]
εca (t) = βas (t)εca (5)
Where
βas (t) = 1 − e−0.2t 0.5 andεca = 0.0003; t is time in days
The maximum tensile stress only due to shrinkage was about 1.8 mpa (tensile,
Fig. 13) at the toe of the backing concrete. Therefore, a mass concrete of higher
compressive strength envisaged at foundation grade. However, the maximum stress
due to shrinkage at interface of old and new concrete was restricted to about 0.4
mpa which was found acceptable.
6.11. FOUNDATION TREATMENT
At the time of construction of existing Dhanikhari concrete dam, foundation

treatment like consolidation grouting, contact grouting and drilling of drainage holes
were carried out. Also installed piezometers indicates piezometric head down-
stream of grout curtain of about 1/10th to 1/8th of the corresponding head in
reservoir, which is very small compared to the design value considered in the stability
analysis.
Seepage monitoring was continued in year 2002 corresponding to different

reservoir level. The seepage from body of dam through formed drains was 10 lpm
corresponding to reservoir level 54.15 m which increases to 45 lpm corresponding to
reservoir level of EL 60.4 m. So increased seepage corresponding to raised reservoir
level after dam heightening to a level of 66 m was not difficult to handle.
Due to excellent performance of the dam both from uplift and seepage points
of view, no special treatments were foreseen due to raising of dam and subse-
quent rise in reservoir level from present FRL of EL 61.00 m to reservoir level of
EL 66.00 m.
However, at both banks, the grout curtain provided previously ends at EL ±

56.00 m. This curtain was required to be extended on both banks from EL 56.00 m to
EL 66.00 m. This was achieved by providing only 10 m deep holes from the surface
at 3m c/c to achieve permeability of the order of 5 lugeons. In addition, the previously
602
Q.104 - R.43
drilled drainage holes in the gallery, which had been subsequently plugged, were
re-drilled up to a depth of 15 m. Also, in the foundation of all the blocks weak pockets
were excavated and filled back with concrete.
Fig. 13
Principal stress vectors due to shrinkage at 10th year.
Principaux vecteurs de contraintes dus au retrait de béton à la 10 è année.
7. CONCLUSION
Several issues viz. effect of permanently built-in stress on new concrete, effect
of shrinkage in new concrete on old concrete, restraining effect of old concrete
on new concrete, bonding between old & new concrete, enhancement of spillway
capacity and seismic retrofitting, etc. were addressed during the design and engi-
neering for raising the height of Dhanikari dam. The effect of permanently built in
stress were addressed through lowering the reservoir during placement of the back-
ing concrete. The temperature gradient of the backing concrete was with ambient
air temperature was restricted to 5◦ C by keeping the placement temperature at
12◦ C and the effect of shrinkage has addressed by using higher grade concrete at
foundation grade.
603
Q.104 - R.43
REFERENCES
[1] Dhanikhari Concrete Dam Heightening For APWD, Port Blair Detailed Study
Report (Unpublished).
[2] BIS IS 14591: Temperature Control of Mass Concrete for Dams – Guidelines.
[3] ETL 1110-2-542 Thermal Studies of Mass Concrete Structures.
[4] Advanced Dam Engineering for Design, Construction and Re-habilitation.
[5] BIS IS 1343: Prestressed Concrete – Code of Practice.
604
Q.104 - R.44
-------
-------
OPERATIONAL PERFORMANCE AND SEDIMENT HANDLING BY

ABRASION RESISTANCE STEEL LINER AND RAILS IN UNDER-SLUICE
SPILLWAYS OF NATHPA DAM, INDIA (∗)
Sanjeev GUPTA, NAGARAJ, Harsh BHASKAR MEHTA & S.P. BANSAL

SJVN LIMITED, Shimla
INDIA
SUMMARY
In India, earlier the spillways in Himalayan regions followed conventional

design viz high ogee spillways in dams. However, considering the experience of
silting of reservoirs and damages of underwater parts of turbine, the trend has
changed to design run-of- the river schemes which utilizes the river flow as it comes
without large storage. Dams of this type of schemes are using under-sluices for
flood disposal and reservoir flushing spillways. Nathpa dam is among the first dams
in Himalayan regions which adopted these large sized under- sluice spillways (low
level spillways). These under- sluices spillways have to pass high concentration of
suspended sediment laden water along with bed load comprising of cobbles and
boulders. Due to high sediment load, the abrasion resistance material is essential
since exceptional durability is required. This paper focuses on the operational per-
formance of abrasion resistance steel liner and steel rails in under-sluice spillways
of Nathpa Dam.
∗ Performance opérationnelle et gestion des sédiments par un revêtement en acier résistant
à l’abrasion dans les évacuateur de crues du barrage de Nathpa, Inde
DOI: 10.1201/9781003211808-40 605

Q.104 - R.44
RÉSUMÉ
En Inde, les déversoirs des barrages himalayens suivaient la conception

conventionnelle avec des évacuateurs à seuil libre. Toutefois, compte tenu de
l’expérience de l’envasement des réservoirs et des dommages causés sur les tur-
bines, la tendance est de concevoir des barrages au fil de l’eau sans stockage. Les
barrages de ce type de régimes utilisent des vannes de fond pour évacuer les crues
et permettre le transit des sédiments. Le barrage de Nathpa est l’un des premiers
barrages des régions himalayennes qui ont adopté ce type de conception. Ses évac-
uateurs doivent faire passer une eau avec uneforte concentration de sédiments en
suspension sur un lit de rivière composée de pavés et de rochers. En raison de
la concentration élevée de sédiments, la résistance à l’abrasion du matériau de
revêtement des déversoirs pour garantir sa durabilité. Cet article met l’accent sur
la performance opérationnelle du revêtement en acier résistant à l’abrasion et des
rails en acier enrobés dans les évacuateurs du barrage de Nathpa.
1. INTRODUCTION
Satluj Jal Vidyut Nigam Limited (SJVNL formerly NJPC) is operating 1500 MW
Nathpa Jhakri Hydro Power Station (NJHPS) in districts Shimla and Kinnaur of
Himachal Pradesh, India on the River Satluj in the foothills of the Himalayas. SJVN
has made a landmark achievement in the history of river valley projects in India
by constructing largest underground hydro power station in the year 2003-2004.
The project has many unique features like the largest underground desilting cham-
bers (4 Nos., each 525 m long), the longest hydropower Hear Race Tunnel (10.15
diameter & 27.40 m long) and deepest Surge Shaft (301 m deep) in the world.
2. INNOVATIVE DESIGN OF UNDER-SLUICE SPILLWAYS (LOW LEVEL

SPILLWAYS) OF NATHPA DAM
The use of large bottom openings as spillways is a relatively modern concept.

These innovative designs of dam spillways have been evolved based on the concept
of flushing of sediments. The following paras discuss about the features of Nathpa
dam and its under-sluice spillways.
2.1. DAM AND ITS UNDER-SLUICE SPILLWAYS
The Nathpa dam located in district Kinnaur of Himachal Pradesh. The dam is a
straight concrete gravity structure with maximum height of 62.50m above the deepest
606
Q.104 - R.44
Fig. 1
Layout Plan of Nathpa dam.
Fig. 2
Upstream Elevation of Nathpa dam.
foundation level on river Satluj. The reservoir provides a live storage capacity of
3.03 MCM & gross storage of 3.43 MCM which can cater 3 hours diurnal peaking
generation during lean season.
The central part of dam body consists of five (05) under sluice spillway blocks
each having 16 m length with total length of 80 m. Central under-sluice block has
two tier spillways system having auxiliary spillway at upper level and under- sluice
spillway at lower level. The auxiliary spillway used for regulating the reservoir level &
passing trash. The total length of dam at top is 185.45 m including seven (07) non-
overflow blocks of 105.45 m. The layout plan & upstream elevation of the dam is as
per “Fig. 1 to 3”.
607
Q.104 - R.44
Fig. 3
Downstream view of the Under sluice spillways.
There are five (05) under- sluice spillways (low-level spillways) blocks. Each
under-sluice spillway block consists of opening with size 7.50 m (wide) × 8.50 m
(high) controlled by radial gates as shown in Figure 2. This low level spillway arrange-
ment is capable of passing the design flood of 5660 m3 /sec by considering one of
the sluices inoperative. The design flood has return period of 1 in 10,000 year. The
design water head for the under-sluice spillway is 33.25 m. The discharge intensity
of the spillway is order of 151 m3 /sec per meter width.
The hydraulics of under-sluice spillways changes with the varying reservoir

level. The flow is free flow for reservoir water levels below the top of the roof of the
sluice. For higher water levels the flow is orifice (pressurised) flow. The crest profile
is required to be designed for orifice flow. The crest profile of under-sluice spillways
is flatter as compared to the overflow spillways crest profile to avoid flow separation
and negative pressures on the crest for small partial gate openings.
The under sluice spillway is divided into three (03) sub components and
detailed in “Table 1”.
The bigger size openings (5 Nos.) of the order 7.50 m (wide) × 8.50m(high)
near the river bed, have been provided for dual purpose of passing flood and draw
down flushing of sediments of the reservoir. This arrangement is found to be effec-
tive in flushing of the sediments which restore the live storage of the reservoir and
encroached during high flow seasons when the sediment concentration is high.
608
Q.104 - R.44
Table 1
Sub components of Under Sluice Spillway
Sr. Name of sub components of Extent Details
No Spillway
1 Under - sluice spillway openings Consist of Bell Mouth entry The bell mouth entry is three
See “Fig 5 to 7” from u/s face of dam connect- side flared with elliptical equation
ing rectangular transition and of x2 /9.502 + y2 /3.1672 = 1 (two
portion till sill beam of the radial side flares) and x2 /8.502 + y2 /
gate 2.8332 = 1(roof flare). Transition
length is 9.50 m.
2 Spillway Profile or Glacis portion. From sill beam of radial gate From u/s of sill beam of the radial
See “Fig 4 ” till tangent point of parabolic gate, the spillway profile/ glacis
spillway profile & start of flip follow equation x2 = 126y.
bucket.
3 Energy dissipation arrangement Tangent Point to lip of the flip Energy Dissipation Arrangement
(EDA) (Flip Bucket or Trajectory bucket bucket has a radius of 24 m.
Bucket type). See “Fig 4”
In India, earlier the spillways in Himalayan regions followed conventional

design viz high ogee spillways in storage dams. However, considering the expe-
rience of silting of dams and damages of underwater parts of turbine, the trend
has changed to design of run-of- the river schemes which utilizes the river flow as
it comes without any large storage. The dams of this type of schemes are using
under-sluices for passing floods and for flushing of the reservoir. Nathpa dam is
among the first dam in Himalayan Regions which adopted these large sized under-
sluice spillways (low level spillways).
3. CHARACTERISTICS OF RIVER SATLUJ
The Satluj river is one of the fastest flowing rivers originating in the mountain
desert with steep slopes and fragile geology, it carries immense silt load during snow
melt season. The river Satluj enters Indian territory from Tibet near Shipki-La. It is
joined by main tributary, river Spiti, at Khab. The total catchment area of river Satluj
till Nathpa dam is about 49,820 km2 out of which 36,900 km2 (71%) is in Tibet and
the rest 12,920 km2 is in India.
The long term data shows that average annual sediment load of river Satluj
is about 38 MCM and average silting rate is 0.682 mm/year or 0.682 thousand
m3 /km2 /year as per compendium on silting of reservoirs in India by Central Water
Commission.
During high flow season (June to September), the average suspended sedi-
ment load of about 973 ppm was expected as per historic data and the same was
used for planning of sediment management of the project. However, after opera-
tion of this project, sediment load has shown increasing trend because of human
interference ecology and other changes in upper catchment area. The sedimentation
609
Q.104 - R.44
Table 2
Suspended Sediment load in ppm
Year No. of days in a year for sediment concentration
2001 to 4000 4001 to 5000 More than 5000
(ppm) (ppm) (ppm)
1971-1996 (Average) 22 4 3
1997-2003 (Average) 38 11 16
2004-2007 (Average) 29 5 26
2008 39 4 10
2009 31 3 1
2010 36 4 27
2011 13 0 5
2012 9 0 7
2013 37 17 8
2014 37 2 3
2015 35 1 10
2016 11 2 8
2017 36 3 9
2018 18 3 14
2019 36 4 4
2020 6 0 3
Table 3
Particle distribution of the sediment load
Coarse (>0.2 mm) 16%
Medium ( 0.75 to 0.2 mm) 22%
Fine (0.075 to 0.75 mm) 62%
Table 4
Typical Mineralogical composition (Petrography profile) of
sediments in river Satluj.
Name of mineral Hardness number on Moh’s Scale % composition of constituent
Quartz 7 45-50
Tourmaline 7 to 7.5 2-4
Lithic Fragments 7 4-6
Feldspar 6-6.50 10-12
Magnetic 6-6.50 8-10
Hornblende 5.50 10-15
Biotite 2.5-4 10-15
Other minerals - Traces
concentration observed in the river Satluj from 2004 to 2020, after commissioning
of the project is shown in “Table 2”.
The average particle size distribution of the suspended sediment load in river
Satluj is shown in “Table 3”.
The silt load carried by the river Satluj has high percentage of quartz which
have hardness of 7 on Moh’s scale as shown “Table 4”.
610
Q.104 - R.44
Apart from the above suspended sediment load, the heavy bed load (10-20%)
comprises of cobbles and boulders with average size of 1.00 m and maximum size
upto about 3.00 m-3.50 m during the monsoon is also characteristic of river Satluj.
4. PROBLEM DUE TO HIGH VOLUME OF SUSPENDED SEDIMENTS AND

BED LOAD
The abrasion protection is required for under-sluice profile/ glacis and flip
bucket portions wherever a combination of high head, high volume and high velocity
flow conditions prevails. Water will be released throughout the monsoon season over
prolonged period of time, both for evacuating the excess water above the turbine
capacity and for flushing out sediments from the reservoir.
This released water will contain high portion of abrasive sediments, cobbles
and boulders which will further roll and bounce over the inverts with high velocities of
about 20m/s or even more. High velocities associated with high heads increase the
susceptibility of cavitation damage to the spillways glacis requiring special attention.
The high abrasive sediment rolling over the under-sluice opening and spillways glacis
necessities use of abrasion resistance steel liner and high performance concrete for
minimising the abrasion damage.
5. SOLUTION ADOPTED FOR UNDER-SLUICE SPILLWAYS DURING

CONSTRUCTION
Various discussions on the topic of abrasion protection took place between

SJVN, Panel of Experts (POE) and retainer consultant for possible solution to
be adopted during constructions stage. The protection measures adopted during
construction is discussed in subsequent para of the paper.
5.1. ABRASION RESISTANCE STEEL FOR UNDER- SLUICE SPILLWAY OPENINGS
Because of extreme silt load in river Satluj, provision has been made for flush-
ing of reservoir periodically during high flow season to preserve live & dead storage
of the reservoir against siltation. Accordingly, sluices have been lined with abrasion
resistant steel having Brinell hardness of 400 BHN.
In 1997, a joint venture of Nippon Koei Company limited Japan, Electro

Watt Engineering Limited, Switzerland and WAPCOS Limited were appointed as a
Retainer consultant for dam detailed engineering and construction. The consultants
suggested abrasion resistance steel liner for the under- sluice spillway openings of
the Nathpa dam instead of mild steel lining. The abrasion resistant steel proposed
because later adaptations / repairs are neither easy nor convenient to be carried out.
611
Q.104 - R.44
Table 5
Specifications of HARDOX 400 Abrasion Resistance steel plate
(20-25 mm thick)
Sr. No Description Permissible values
1. Mechanical Properties
a Ultimate Tensile strength 1250 N/mm2
b Typical Yield strength 1000 N/mm2
c Elongation 10%
2 Chemical Properties
a) Chemical composition (ladle analysis) in %
(i) Alloy Element
Carbon (C) 0.20
Silicon (Si) 0.10-0.70
Manganese (Mn) 1.70
Phosphorus (P) 0.025
Sulphur (S) 0.010
(ii) Micro-alloying Elements used singly or in combination for controlling hardness
Molybdenum (Mo) 0.80
Ni (Nickel) 1.00
Chromium (Cr) 0.80
Boron (B) 0.005
b) Maximum Carbon equivalent corresponding of different liner thickness
Thickness (mm) 15 20-25 40
ii CEQ (Carbon Equivalent) 0.38 0.51 0.57
3 Hardness
(i) (Brinell) BHN 400 BHN (minimum)
4 Impact properties
(i) Charpy V 10x10 mm test specimen 30J at (-) 40◦ C
In all five (05) under-sluice openings, liners involving 1650 T of abrasion resistance
steel were erected and concreted during a single season of construction.
The abrasion steels from HARDOX 400 (Swedish steel) and DILLIDUR 400 V
(German steel) were proposed by the consultant. HARDOX and DILLIDUR possess
more or less equivalent physical and chemical properties. SJVN adopted HARDOX
400 steel plates imported from Sweden. These abrasion resistance steel are low-
carbon ferritic steels having high yield stress and high hardness with good level
of plasticity. To achieve such high strength, these steels are subjected to thermo-
mechanical processing (TMP) during the manufacturing process.
The specifications (physical & chemical) of abrasion resistance steel plate

mentioned in “Table 5”. The detailed drawings showing the abrasion resistance steel
lining system is shown in “Fig. 4 to 7”.
5.2. OPERATIONAL PERFORMANCE OF ABRASION RESISTANCE STEEL LINER
The abrasion resistance steel liner shown long durability life and resistance
to the abrasive sediment load. During the 15-years period (2006 to 2020), data
612
Q.104 - R.44
Fig. 4
Section of Under Sluice Spillway (Low Level Spillway).
Fig. 5
Steel liner arrangement for Under- Sluice opening (bell mouth entry to sill beam).
613
Q.104 - R.44
Fig. 6
Plan showing longitudinal and transverse stiffeners of steel liner.
Fig. 7
Sectional Elevation showing steel liner plate thickness.
614
Q.104 - R.44
has been indicated in “Tables 6 & 7” showing sediment load passed through and
resisted/handled by the under-sluice opening steel liners.
Table 6
Quantum of silt flushed from Nathpa dam
reservoir during drawdown flushing
operations (2006-2020)
Year Nos. of Flushing Total amount of Silt
Cycle performed Flushed (in Tonne)
2006 4 1,140,162
2007 3 1,731,628
2008 2 1,125,250
2009 2 1,685,208
2010 1 1,057,553
2011 2 1,309,458
2012 1 546,093
2013 1 479,671
2014 2 1,793,081
2015 3 2,295,236
2016 4 2,248,771
2017 3 2,693,424
2018 3 2,293,268
2019 2 1,599,618
2020 1 1,078,949
Total 23,077,370
Average Silt flushed in a year 1,538,491
Say, 1.54 Million Tonne
Table 7
Sediment load passed through under sluice openings of steel liners
Sr. No. Description Approximate Quantum Approximate Quantum
of sediments of Silt in Tonnes
(Million Cubic meter) (Million Tonne)
1. Annual Sediments load in river 38 53.2
Satluj at Nathpa dam
2. Annual sediments passed from tur- 5 7
bine measured at Draft Tube.
3. Annual Sediments passed from 5 7
through Silt flushing Tunnels(SFT)
after Desilting Chambers.
4. Annual Sediments load passed 38-(5+5)=28 39.2
through five Under- sluice spill-
ways.
a Annual Sediments load during 1.10 1.54
reservoir flushing
b Annual sediment load carried out 26.90 37.66
by water passed through under-
sluice spillways except flushing
period.
5. Average Sediment load passed (Sr.No.4X15) = 420 (420X 1.4)= 588
through five under- sluice openings
in 15 years from 2006 to 2020.
6. Total average Sediment load 84 117.60 say 118
through each under-sluice open-
ing in 15 years from 2006-2020.
615
Q.104 - R.44
Each sluice liner has resisted sediment load of 118 million Tonne (2006-2020)
with velocity greater than 20 m/s. 69 % of sediments particle have minerals with hard-
ness number on Moh’s scale greater than 6. Liner also resisted impact of boulders
with average size greater than 1 m and even as large as 3.50 m. “Fig. 8” showing
the Steel liner in Block no.5 as intact and without any major impact and damage.
Fig. 8
Intact/ no damage to steel liner after completing 15 high flow seasons.
6. STEEL RAILS BEYOND RADIAL STEEL GATE
For the spillway profile/ glacis portion till lip of energy dissipation arrangement
bucket, 50 lb/yard steel rail having depth 115 mm were placed at 175 mm c/c and
anchored to the base concrete by transverse steel channels have been placed at
3.00 m intervals. The gap between the steel rails was filed with concrete.
The retainer consultant suggested to use one option out of three (03) namely
‘abrasion resistance steel liner’, ‘high performance concrete with silica fume’ and
616
Q.104 - R.44
‘ALAG –concrete’. In addition to this special mortar of SIKA of Switzerland under

the brand name of ABRAROC-M was also suggested.
However, for expediting the construction schedule and consideration of cost,

SJVN adopted steel rails system for glacis portion till the lip of the bucket. The lip of
bucket of energy dissipation arrangement was also protected by abrasion resistance
steel liner of same specifications as adopted in initial reach.
6.1. OPERATIONAL PERFORMANCE OF STEEL RAILS
After 4 years of operation from 2008 onwards, problem of worn out of the steel
rails started in Block No. 6, 7 & 8. For the Block No. 4 & 5, problem was negligible.
The top of rails were worn away by beatings and grinding and the small-
aggregate concrete between the rails was disintegrated/broken and washed away.
Finally the rails were dislodged, bent buckled, in severest cases even the trans-
verse anchor profiles were also destroyed. Abrasion resistance plates were also
pulled out from lip of the energy dissipation arrangement bucket. “Fig. 9 to 12”
depicts damages to steel rails and concrete lining of spillway profile/ glacis & energy
dissipation arrangement bucket.
Fig. 9
Exposed Steel rails after erosion of concrete of Under sluice spillway profile/Glacis
(Block No.8).
617
Q.104 - R.44
Fig. 10
Eroded and damage glacis and EDA steel rails and Concrete (Block No.8).
Fig. 11
Damaged Steel rails and transverse anchor profile of Energy dissipation
arrangement bucket (Closer View).
618
Q.104 - R.44
Fig. 12
Damaged and worn out steel rails (closer view).
6.2. INNOVATION IN REPAIR AND MAINTENANCE OF GLACIS AND EDA BUCKET
From 2008 onwards till 2012, repairs usually was done by using ordinary M-25
grade concrete every year, but concrete used to get eroded after high flow season.
The above problems of erosion and worn out of the steel rails were handled by SJVN
by adopting application of high performance concrete. From 2013 onwards, spillway
profile/ glacis of under-sluice spillways are being repaired with high performance
concrete of M-80 grade. The mixed design adopted has been carried out in house
at Jhakri. The mixed design adopted for M 80 grade concrete is shown in “Table 8”.
619
Q.104 - R.44
Table 8
Mixed Design of High Strength Concrete M80 grade
Sr. No. Material Unit Quantity
1 Cementitious Material
a OPC 53 Grade Cement kg 500
b Silica Fume(kg) kg 50 (10% by weight)
c Total Cement Content (TCC) kg 550
2 Aggregate
a Sand kg 570(33% by weight)
b Coarse Aggregate nominal size 20 mm and 16 mm kg 1200
3 Admixture
a Quantity of Super Plasticizer % 0.9 (By weight of cement)
b Super Plasticizer Type Glenium 806 (SIKA brand)
4 Water
a Quantity of water litre 145
5 Ratios
a Water to Cement (W/C) ratio 0.29
b Water to Binder (W/B) ratio 0.26
c Paste Volume ratio 35
6 Property
a Workability, at 0hrs mm 120 (slump)
b Visual observation Fully Cohesive without bleeding &
segregation.
7 Strength
a Compressive Strength in 7 days MPa 66.70
b Compressive Strength in 28 days MPa 89
d Flexural Strength (28 Days) MPa 7.82
e Split Tensile Strength (28 Days) MPa 7.15
6.3. METHODOLOGY ADOPTED FOR REPAIRS
The repaired concrete in spillway glacis and energy dissipation arrangement

bucket normally resist two (02) high flow seasons and again got repaired after two
high flow season which saved one year cost of repair as mentioned earlier. The
following methodology is being adopted:
a) Dewatering of the spillway glacis and energy dissipation arrangement bucket

portion of Invert lining.
b) Cleaning/removal of sediments, cobbles and boulders in the deepest portion
of the energy dissipation arrangement bucket.
c) Chipping of old damaged concrete so that proper bond can be made with
old concrete.
d) Providing and fixing 25 mm diameter High Yielding Strength Deformed Rein-
forcement bars with yield strength 500 N/mm2 . The bars are anchored at
1000 mm c/c both way after drilling 32 mm dia hole in the old concrete.
620
Q.104 - R.44
e) Steel reinforcement bars or rebars are used to improve the tensile strength
of the concrete. 25 mm diameter bars @250 mm c/c both ways is placed
before concreting.
f) Laying of M25 grade concrete of 250 mm thickness at under- sluice spillway
profile / glacis portion and 750 mm at energy dissipation arrangement bucket
portion over the steel reinforcement.
g) Final layer of 115 mm thickness of M80 grade concrete is laid over the
M-25 grade concrete .“Fig. 13 ” shows the Glacis & energy dissipa-
tion arrangement bucket after repairs with M 80 grade high performance
concrete.
Fig. 13
Spillway Glacis portion and energy dissipation arrangement bucket after repair.
6.4. MODIFICATION IN THE METHODOLOGY ADOPTED FOR THE REPAIRS
The reinforcement / rebars usually get pulled out due to high sediment load
due improper bonding and anchorage. The performance of the above methodology
will be ascertained after passing one high flow season during April 2021 to Sept
2021.
621
Q.104 - R.44
To improve the efficacy of the repairs the following modified methodology is

being proposed for year 2021:
a. The methodology mentioned at 6.3 (a) to (d) shall be the same.

b. Laying of 250 mm thick M80 grade concrete including filling of local pockets
over entire under- sluice spillway profile/ glacis (Sr. No. 2 of “Table 1”).
c. Laying of 750 mm thick M80 grade concrete energy dissipation arrangement
Bucket without reinforcement (Sr. No. 3 of “Table 1”).
6.5. EXPERIENCE GAINED DURING REPAIR AND MAINTENANCE
The experience gained during construction as well as during operation &

maintenance (O&M) stages for repair of spillways glacis and energy dissipation
arrangement has encouraged to adopt better material. These material possess high
resistance against sediment abrasion with exceptional high durability and avoids fre-
quent repair and maintenance during O&M stage. This approach shall provide great
help and benefits for upcoming hydro projects in the Himalayan Region.
7. CONCLUSION
The under sluice spillway / low level spillway with large size opening of the order
6-10 m (width) X 10-14 m (high) as near the river bed as possible, are becoming
popular due to their ability to flush the sediment out of the reservoir. A distinguishing
characteristic of under sluice/ low level type spillways is that water is passed through
spillway as orifice (pressurised) flow conditions instead of atmospheric conditions in
traditional over flow spillways. Thus, these under-sluice spillways having pressurised
flow conditions can pass more flood volume compared to the conventional overflow
spillways.
Typically, this type of spillways would be used in situations where flushing of

reservoir is one of the functions to restore live storage along with flood disposal in nar-
row gorge and a wide range of reservoir levels are anticipated. Decreased water way
in the narrow gorge necessitates special design considerations for increased dis-
charge intensity upto 150-300 m3 /sec per meter width. It is further seen that for under
sluice spillway/ low level spillways the preferably energy dissipation arrangement
may be stilling basins types instead of bucket types. The sediment load, sediment
ratings curves and sediment petrography are required to be scientifically quantified
for planning before construction in case of under sluice spillways.
It is concluded that with high design water head under- sluice spillways open-
ings upto the radial gate sill beam can be provided with abrasion resistance steel
of brands with high yield strength and minimum 400 Brinell Hardness Number for
622
Q.104 - R.44
long durability and resistance to abrasion sediment loads. The steel liner should
be installed with well-designed transverse and longitudinal stiffeners anchored in
concrete for better structural performance and abrasive resistance. It has to be kept
into consideration that fabrication and installation of steel liner is a time consuming
process.
The spillway profile / glacis portion and energy dissipation arrangement (either
bucket type or stilling basing type) shall be laid with high performance concrete with
silica fume or micro silica or ALAG concrete or special mortars with extremely high
durability. To avoid frequent repairs and maintenance during the dam spillways oper-
ations, composite system of steel rail and concrete should be avoided in high velocity
& high head, high sediment load conditions. This composite system is susceptible
to cavitation which starts at contact of the rails and the concrete, thus effecting the
durability of spillway glacis and energy dissipation arrangement.
REFERENCES
[1] Retainer Consultant, a Joint venture of Nippon Koei Company Limited Japan,
Electrowatt Engineering Limited Switzerland and Water and Power Con-
sultancy Services Limited India. Report on Dam Concreting. November
1997.
[2] Electrowatt Engineering Limited Switzerland. Memo No. D13 on Abrasion

Protection of dam-Under-sluices/lip Bucket and Desilting- Flushing Tunnel.
September 1998
[3] Menghini M.,Spriano S.,Electrowatt Engineering Limited Switzerland. Memo

No. Dam-00-02 on Nathpa Dam Sluice Liner. May 2000.
[4] Central Board of Irrigation and Power, New Delhi. Special Issue on 1500 MW
Nathpa Jhakri Hydro Electric Project. June, 2008.
[5] Central Water and Power Research Station, Pune, India. Technical Memo-
randum on research into factors which influence Hydraulic design of Breast
Wall/Sluice Spillways. September, 2008.
[6] Morris, Gregory L. and Fan, Jiahua. Reservoir Sedimentation Handbook.

December 2010.
[7] Central Water Commission, India. Compendium on silting of reservoirs in India.

April, 2015.
[8] Bureau of Indian Standard,Manak Bhawan,New Delhi. IS:11485. Criteria for

Hydraulic Design of Sluice in Concrete and Masonry Dam.
[9] Bureau of Indian Standard,Manak Bhawan, New Delhi. IS:6934. Hydraulic

Design of High Ogee Overflow Spillways –Recommendations.
623
Q.104 - R.45
-------
-------
HYDRAULIC DESIGN OF BARRAGE ON HIMALYAN BOULDERY RIVER- WITH

SPECIFIC REFERENCE TO SJVN’s NAITWAR MORI HEP (60 MW), INDIA (∗)
S.P. BANSAL, Rakesh SEHGAL, Revati RAMAN & Surjeet SINGH

SJVN LIMITED, Shimla
INDIA
SUMMARY
Most of the hydro power development in India is going on by constructing

large dams and barrages on rivers originating from Himalayas in North and North-
Eastern states of the country. The major Himalayan rivers are fed by a large number
of tributaries on their way before entering into the alluvial plains. These tributaries
have steep gradient, heavy boulder bed load and deep pervious strata in their upper
reaches. In Indian context a barrage is a barrier with low crest provided with a series
of gates over its entire length across the river to regulate water surface level and
pattern of flow upstream. It may or may not have a raised sill. At present existing
guidelines provided by Bureau of Indian Standards for planning and design of Bar-
rages do not apply to the structures in boulder dominant river reaches. Naitwar Mori
Hydroelectric Project (60 MW) is one of the such15 Hydro Electric projects of SJVN
under various stages of planning, construction and operation ranging from 44MW
to 1500 MW. This Project (NMHEP) is located on the upper reaches of River Tons,
a tributary of River Yamuna in the Himalayan region of India. Project envisages a
Barrage type of structure for diversion of water for power generation. The project is
in advance stage of construction and this paper covers the hydraulic design aspects
that have been applied while designing the barrage.
∗ Conception hydraulique de barrage sur une rivière himalayenne à fond rocheux – référence
spécifique à SJVN NAITWAR MORI HEP (60 MW), Inde
624 DOI: 10.1201/9781003211808-41

Q.104 - R.45
RÉSUMÉ
La majeure partie du développement hydroélectrique en Inde se fait en con-

struisant de grands barrages et des barrages sur des rivières en provenance de
l’Himalaya dans les États du nord et du nord-est du pays. Les principales rivières
himalayennes sont alimentées par un grand nombre d’affluents avant d’entrer dans
les plaines alluviales. Ces affluents ont une pente abrupte, un lit rocheuxr et des
couches perméables épaisses dans leurs parties supérieures. Dans le contexte
indien, un barrage a une faible hauteur, et comprend plusieurs passes toute sa
longueur de rive à rive pour réguler la cote du plan d’eau et les débits. Il peut ou non
avoir un seuil surélevé. Les lignes directrices existantes fournies par le Bureau des
normes indiennes pour la planification et la conception des barrages ne s’appliquent
pas à ces sites. Le projet hydroélectrique Naitwar Mori (60 MW) est l’un des 15 pro-
jets hydroélectriques de SJVN dans le cadre de diverses étapes de planification, de
construction et d’exploitation allant de 44 MW à 1 500 MW. Ce projet (NMHEP) est
situé dans les sections amont de la rivière Tons, un affluent de la rivière Yamuna
dans la région himalayenne de l’Inde. Le projet prévoit un barrage de dérivation pour
la production d’électricité. Le projet est à l’étape préalable de la construction et ce
rapport couvre les aspects de conception hydraulique qui ont été appliqués.
1. INTRODUCTION
1.1. GENERAL DESCRIPTION OF THE PROJECT
The NMHEP is located about 408 km NE of New Delhi and 173 km N of

Dehradun and on the Tons River in Northern state, Uttarakhand, of India.The Bar-
rage is near to the village Naitwar and powerhouse is nearer to the village Mori.
The project is a medium head, run of river scheme with possible diurnal storage for
peaking power. The geomorphology of the area has largely been determined by its
tectonic and erosional history. The project shall harness the hydropower potential
between Naitwar and Mori villages by utilizing a gross head of 96.76 m. Layout of
project is shown in Fig.1.
1.2. MAIN COMPONENTS OF THE PROJECT
A concrete barrage with riverbed level at 1254 m above MSL and the top level at
1269.50 m elevation. Intake, desilting tank with twin chambers of length 130 m, width
25 m and depth 13.5 m. A 4.3km m long and 5.6 m diameter modified horseshoe
type headrace tunnel. This tunnel terminates into an underground surge shaft of
18 m diameter and 50m depth. Further a 4.0 m diameter 50m deep pressure shaft
625
Q.104 - R.45
Fig. 1
Layout of Project.
bifurcating into two branches leads the water to an underground powerhouse having
an installation of 2 vertical Francis type turbines of 30 MW each. These units will be
housed in an underground power house of size 58m (L)x18.60(W)x33m(H). Seven
number single phase transformers shall be installed in a separate cavern of size
73(L)x12(W)x13.65m(H). The water after generation shall be led to the river through
215m long tail race system. After commissioning the power shall be evacuated
through 220 kV double circuit line.
1.3. BARRAGE
The barrage is 18.5m high above foundation raft, 50m long reinforced concrete
structure having three radial gates of size 9.1 m wide and 8 m height for passing
the design flood of 1600m3 /sec. One auxiliary spillway of size 6.6m wide and 2m
high with a provision of flap gate is also provided for removal of floating debris.
A 40 m long impervious floor of thickness 1.5 m has been provided on upstream
of the barrage. A 40.5 m long horizontal type stilling basin has been provided for
energy dissipation. The total length of impervious floor which consists of upstream
floor, crest, glacis, downstream stilling basin and end sill is 125 m. Concrete blocks
and boulder pitching are proposed to be provided upstream and downstream to
protect impervious floor. Upstream and downstream cut off are provided to cater the
upstream and downstream scour. Longitudinal cut offs have also been provided for
boxing effect. Arrangements for Fish Pass and release of mandatory environmental
flows have also been made in the barrage structure. Location, layout and L-section
of barrage are shown in Fig. 2, 3 and 4.
626
Q.104 - R.45
Fig. 2
Location of Barrage Site (before construction).
Fig. 3
Layout plan of Barrage.
2. DESIGN CONSIDERATIONS
While arriving at final hydraulic design all possible modes of failure have been
considered including foundation failure due to scour, piping of material underneath
due to excessive seepage, exit gradient and suffusion risk. These details have been
discussed in subsequent paragraphs.
627
Q.104 - R.45
Fig. 4
L-section of Barrage.
2.1. SCOUR DEPTH
One of the most popular theory for working out the scour depth has been given
by Lacey. However, Lacey’s formula popularly applied to calculate the scour depth of
rivers in plains, is not applicable in boulder reaches of rivers. Scour depth for boulder
river bed will be different due to armoring effect of large sized boulders. Armoring
is a recognized phenomenon in gravel bed rivers that have been subject to periods
of extended low flows. Coarsening creates a bed surface where greater intergranu-
lar friction angles increase the surface stability and the stress necessary to entrain
the bed. The bed surface of gravel-bed rivers is often coarsened, or armored, rela-
tive to the subsurface. Streambed armoring strongly influences channel hydraulics,
mediates the exchange of water between flow and bed. Flow-sediment interaction
produces drag as a resistance against stream flow, and the head variation over bed-
forms is required to estimate the boundary resistance in gravel-bed streams. The
static armored bed condition exists as a result of an extended period of flows over
a mixed gravel bed. The flows generate shear stresses less than that needed to
entrain the largest particles but large enough to transport the fines and over time
the fines sediment is winnowed from the bed surface. An armor layer increases the
resistance to entrainment of the bed surface. Coarsening creates a rougher surface
with greater intergranular friction angles which increase the stresses necessary to
entrain the bed surface. Flow resistance is increased by alteration of the flow field
and formation of turbulent flow patterns that include vortex shedding. When the
armor layer breaks, the channel bed erodes and incision may occur. Armoring is
a grain sorting process in which larger size particles form an armor coat on the
channel bed reducing further degradation.
In NMHEP barrage the assessment of scour depth therefore has been done
based upon the various available theories as per table below.
628
Q.104 - R.45
Sr. No. Theory Applied Scour Depth Depth of downstream

cut off
1 Lacey’s Formula for alluvial plains 7.88 m 12m
2. U. C. Kothyari & R. J. Garde’s 10.18m 14.5m
formula for gravel bed river
3. U. C. Kothyari & R. J. Garde’s - 9.87 m
formula for gravel bed river with
armoring effect of boulders
The more relevant theory i.e. given by U. C. Kothyari & R. J. Garde’s formula
for gravel bed river was finally adopted with taking into consideration the armoring
effect for boulders to arrive at depth of downstream cut off as shown in the table
above.
2.2. IMPERVIOUS FLOOR
The total length of impervious floor which consists of upstream floor, glacis,
downstream basin and end sill required from consideration of safety against sliding is
fixed in conjunction with the depth of downstream cut-off to satisfy the requirements
of exit gradient, scours and economy.
One of the most popular theory for hydraulic design of impervious floor
on pervious strata was given by Engineer A.N Khosla Ex Vice Chancellor of
the University of Roorkee (Currently IIT Roorkee) a premier institute of country. This
theory is generally applied to determine the length of impervious floor and depth of
downstream cut-off for protection against piping due to sub surface flow for struc-
tures on rivers with fine and medium alluvial sediments. As per this theory, the sub
soil medium is assumed to be homogenous, isotropic and of infinite depth below
the barrage floor. In boulder reach the above stated assumption of Khosla’s theory
may not exist at site. This is mainly because of heterogeneity of soil foundation,
anisotropic permeability characteristics and finite depth of pervious medium due
to presence of any impervious layer or rocky strata. Such kind of departure from
Khosla’s theory will invariably increase the uplift pressure to be exerted under the
barrage floor necessitating more thickness than that to be obtained from Khosla’s
curve.
However, in the absence of any proven theory for such kind of conditions the
length of impervious floor, floor thickness and depth of downstream cut off has been
calculated by Khosla’s theory. The safe exit gradient considered during seepage
analysis by this theory for worst possible condition of static head was 0.25. Further
the design was verified by numerical analysis using Slide software. The bed material
characteristics obtained from bore holes were mapped in Slide and seepage analysis
was done. The boundary condition considered is the static head calculated using
maximum water level. This condition is the most critical with water level upstream
629
Q.104 - R.45
Fig. 5
Flow Net Diagram.
Fig. 6
Exit Gradient Diagram.
of barrage at MWL and no water in the downstream of barrage. From numerical

analysis Exit Gradient was found to be the order of 0.06. This was less than 0.25
considered above during seepage analysis as per Khosla’s theory. Flow net and exit
gradient diagrams are shown in Fig 5 and 6.
2.3. SUFFUSION RISK ANALYSIS
In addition to Khosla’s theory and numerical analysis on Slide software, the

suffusion risk analysis has also been carried out for barrage to check the design.
630
Q.104 - R.45
Fig. 7
Internal instability (Suffusion).
From safety consideration, under-seepage of such structures is of great significance.

Suffusion is the mass movement of fine particles through the pore space of a coarser
matrix driven by seepage forces.
In suffusion phenomena finer particles are removed through the voids between
the larger particles by seepage flow, leaving behind a soil skeleton formed by the
coarser particles, as shown in figure 7.
The suffusion risk analysis is done by applying two theories given by Istomina,
V.S (1957) and Burenkova, V.V. (1993) According to Istomina method the soil matrix
at barrage site was found to be stable against suffusion, where as it was found
suffusive as per Burenkova . Skempton and Brogan (1994) who has investigated
the internal erosion process of an internally unstable soil and observed that when
the hydraulic gradient is less than 0.10 the soil is stable against suffusion. In NMHEP
barrage the exit gradient was 0.06 as per numerical analysis which is less than 0.1.
2.4. LEVEL OF CREST
Barrage crest has been kept at bed level of river. This provision is mainly due
to the reason that the Himalayan Rivers in its upper reaches carry a lot of silt and
boulders as bed load. This arrangement is also better for flushing of sediment and
to keep intake area free from accumulation of bed load.
2.5. UPSTREAM PROTECTION WORKS
To protect the impervious floor form scour, cement concrete blocks of size
1.71x1.71x1.2m have been provided in a reach of 8.5m just beyond the upstream
end of it. Further 1.5m thick boulder pitching in 15m reach upstream of cement
concrete block has been provided as an extra protection against scouring.
631
Q.104 - R.45
2.6. DOWNSTREAM PROTECTION WORKS
Pervious Cement block protection has been provided just beyond the down-
stream end of impervious floor as well. It comprises of cement concrete blocks of
size1.71mx1.71mx1.2m laid over 0.3m thick inverted filter. The length of downstream
block protection is 20.50 m equal to 1.5-time design depth of scour. Further 1.5m
thick boulder pitching in 23 m reach has been provided downstream of cement con-
crete block to prevent them from getting disturbed. A toe wall of concrete extending
up to about 1.5 m below the bottom of filter has been provided at the ends as an
extra protection from scouring.
2.7. GROUTING
For additional safety against seepage, the provision of curtain grouting at

upstream of Glacis has also been kept.
3. PRESENT STATUS
The barrage is in advance stage of construction. The photographs below depict

the barrage during monsoon and present stage of construction. (Fig. 8 & 9). As there
is no deep excavation involved in construction of barrage, no major construction
challenge has been faced since the start of its construction except a minor slope
failure on the left bank of the river. This was surmounted by meticulous design
planning and construction.
Fig. 8
Aerial View of under construction barrage during Monsoon.
632
Q.104 - R.45
Fig. 9
Construction works at barrage.
SUMMARY
India is a country with huge hydro power potential of about 150 thousand
MW mainly in the mountainous regions of the country. Many large dams have been
constructed for diverting the waters of major rivers for power generation. However,
for uniform development of infrastructure, many small hydro power projects are also
being planned and constructed on the tributaries of major Himalayan Rivers. NMHEP
(60 MW) is one of the projects. A very few barrages have been constructed on such
streams carrying heavy bed load with no firm & impervious foundation strata for
large concrete dams and are performing well. The design of Barrage of this project
is also based upon various theories adopted in the design of above barrages in
the absence of any proven guidelines. The project is scheduled for commissioning
during the financial year 2021-22. The experience gained after the performance of
the barrage structure shall be shared with the industry in the future.
REFERENCES
[1] IS 6966: Part1 (1989) ‘Guidelines for Hydraulic Design of Barrages and Weirs:
Part-1 Alluvial Reaches (First revision)’, pub. By Bureau of Indian Standards,
Manak Bhawan, New Delhi
[2] IS 6966: Part-2 ‘Guidelines for Hydraulic Design of Barrages and Weirs: Part-
2 Bouldery Reaches’, IS guidelines under formulation by Bureau of Indian
Standards, Manak Bhawan, New Delhi.
[3] Istomina, V.S., 1957. Filtration Stability of Soils.
633
Q.104 - R.45
[4] Burenkova, V.V. (1993) Assessment of Suffusion in Non-Cohesive and Graded

Soils.
[5] Skempton, W. and Brogan, J.M. (1994) Experiments on Piping in Sandy

Gravels. Géotechnique, No. 3, 449–460.
[6] M.M. patne, S.C. Mittal, s.K. Mazumder and P.B. Deolalikar ‘Pllaning and
design of barrage type diversion structures on bouldary river with steep
gradient’.
[7] IS 7720: (1991) ‘Criteria for Investigation, planning and Layout of Barrages
and Weirs’ Bureau of Indian Standards, Manak Bhawan, New Delhi.
[8] IS: 7349 :( 1989), ‘Guidelines for Operation and Maintenance of Barrages
and Weirs, (First Revision)’ Bureau of Indian Standards, Manak Bhawan, New
Delhi.
[9] Khosla, M.N., Bose, K.K.and Taylor, M.T. (1954), ‘Design of Weirs on
Permeable Foundation’, Publication No.12, CBIP, Malcha Marg, New Delhi.
[10] Lacey, G. (1929) “Stable Channel in Alluvium” J. of Instn. Of Civil Engineers,

U.K., Paper no. 4736,Vol. 229, PP.
634
GENERAL REPORT ON
QUESTION 104
RAPPORT GÉNÉRAL
SUR LA QUESTION 104
GR Q. 104
-------
-------
QUESTION 104
CONCRETE DAM DESIGN INNOVATION DANS LE

Rafael IBÁÑEZ De ALDECOA

Head of Hydraulic, Underground and Environmental Projects Dragados
Chairman, SPANCOLD’s Technical Committee on Concrete Dams

Member, ICOLD and USSD’s Technical Committees on Concrete Dams
SPAIN/ESPAGNE
General Reporter / Rapporteur Général
DOI: 10.1201/9781003211808-46 637

GR Q. 104
TABLE OF CONTENTS
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
2. OVERVIEW OF RECEIVED REPORTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
3. INNOVATIONS FOR ARCH DAM ANALYSIS, DESIGN AND
CONSTRUCTION, INCLUDING RCC ARCH AND ARCH-GRAVITY
DAMS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .644
3.1. CVC arch dam analysis and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .647
3.2. CVC arch dam construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
3.3. RCC arch dam analysis and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
3.4. RCC arch dam construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
4. INNOVATIONS FOR CONCRETE GRAVITY DAM ANALYSIS, DESIGN,
MATERIALS AND CONSTRUCTION, INCLUDING RCC AND
CMD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
4.1. CVC gravity dam analysis and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
4.2. CVC gravity dam construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
4.3. RCC gravity dam analysis and design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 660
4.4. RCC gravity dam materials and mixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
4.5. RCC gravity dam construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664
4.6. CMD analysis, design, materials and construction . . . . . . . . . . . . . . . . . . . 665
5. INNOVATIONS FOR RAISING EXISTING
CONCRETE DAMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
5.1. Raising existing concrete dams with CVC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 667
5.2. Raising existing concrete dams with RCC . . . . . . . . . . . . . . . . . . . . . . . . . . . 668
6. INNOVATIONS FOR EXTREMELY HIGH CONCRETE DAMS . . . . . . . . . . . . . 669
6.1. Extremely high CVC arch dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669
6.2. Extremely high RCC arch dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671
6.3. Extremely high RCC gravity dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671
7. OPERATIONAL PERFORMANCE OF CONCRETE DAMS DURING
THE LIFE CYCLE, INCLUDING UNDER EXTREME CONDITIONS . . . . . . . 674
7.1. Performance of CVC arch dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674
7.2. Performance of CVC gravity and buttress dams . . . . . . . . . . . . . . . . . . . . . . 676
7.3. Performance of RCC gravity dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
8. CLOSING REMARKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
ACKNOWLEDGEMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .680
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 680
638
GR Q. 104
1. INTRODUCTION
I would like to start this General Report for Question 104, “Concrete Dam
Design Innovation and Performance” (Q. 104), highlighting that the last time there
was a Question specifically on concrete dams (i.e., if we do not take into account
themes such as ageing, rehabilitation or upgrading) at an ICOLD Congress, it was
in 1988 in San Francisco (USA), as we will see a little later.
The above fact could lead to think that there have been no significant advances
in the state-of-the-art in design and construction of concrete dams since the San
Francisco Congress, but nothing could be further from the reality. I started my pro-
fessional career as a dam engineer in the mid-1980s and, since then, we have
witnessed very important engineering breakthroughs related to concrete dams.
In fact, in the late 1980s, the emerging roller-compacted concrete (RCC) tech-
nology (and in parallel the rolled concrete dams (RCD) in Japan) was just emerging
from its early stages. This new form of dam construction had an amazing success
since its inception, with a continuous development/refinement of the technique along
the following decades, and with a remarkable number of dams completed worldwide
as of today. It has not been until very few years ago that the technique has reached
a state of certain maturity, while it is not unlikely that substantial new developments
will take place along the coming years.
At the beginning, the RCC’s incursion into the arch dams’ field was very timid
but, with the passing of the years, the confidence in the new technique fostered that
design and construction of RCC arch dams became more frequent.
Although it must be recognized that in the last three decades there have not
been significant breakthroughs in terms of conventional vibrated concrete (CVC)
gravity dams, most likely motivated by the loss of competitiveness compared to new
RCC gravity dams, they have occurred with CVC arch dams, where this loss of
competitiveness against RCC arch dams may not be so evident in certain cases.
The mere fact that in the last decade 6 double curvature arch dams of 270 m or more
in height have been built, or are in an advanced state of construction, is evidence
that invites us to think about the great advances that have been recently made
in the design and construction of CVC arch dams. Three countries in particular,
China, Turkey and Iran, are leading currently the construction of new extremely
high CVC arch dams (in Section 6 we will clarify the meaning of “extremely high”
dams).
Another field in which important achievements have been undertaken is that

of the raising of existing concrete dams, with either CVC or RCC, and there is a field
still to be developed more in-depth: hybrid RCC-rockfill dams (cross-section wise),
as a potentially competitive solution in certain favorable circumstances of foundation
conditions, availability of materials and spillway needs.
639
GR Q. 104
In parallel with the innovations and development of the analysis, design and
construction of CVC and RCC dams, various related disciplines, such as surveying,
monitoring, concrete pathologies, rehabilitation, upgrading, concrete materials and
mixes, impervious geomembranes, etc., have as well experienced important pro-
gresses. It is noteworthy that for all the previous disciplines, one or more reports
have been received for Q. 104.
Glancing into the past, concrete dams and related subjects have been the
main theme of numerous questions in previous ICOLD congresses, namely:
• Q. 1a (1933, Stockholm): Deterioration by ageing of the concrete of weight

dams.
• Q. 1b (1933): Influence of internal temperature and distortion of weight dams.
• Q. 3 (1936, Washington D.C.): Special cements.
• Q. 5 (1936): Study of facing of masonry and concrete dams.
• Q. 8 (1948, Stockholm): Uplift and resulting stresses in dams.
• Q. 11 (1948): Information obtained from the use of testing methods and of
special cements in large dams.
• Q. 15 (1951, New Delhi): Concrete for large dams.
• Q. 17 (1955, Paris): Economics and safety of different types of concrete dams.
• Q. 19 (1955): The relation of the cements content of the concrete to perfor-
mance in practice of: a) gravity dams (interior and exterior); b) arch dams; c)
buttress dams, and its influence on permeability and frost resistance.
• Q. 20 (1958, New York): Heightening of existing dams including methods of
constructing new dams in successive stages.
• Q. 21 (1958): Observation of stresses and deformations in dams and in their
foundations and abutments; and a comparison of these observations with
computations and tests on small scale models.
• Q. 23 (1958): Use of admixtures and pozzolanic materials in concrete for dams
and the influence of the finer sand particles.
• Q. 24 (1961, Rome): The selection, processing and specification of aggregates
for concrete for large dams.
• Q. 26 (1961): Modern techniques of concrete dams for wide valleys and
ancillary works.
• Q. 30 (1964, Edinburgh): Design of concrete for large dams of all types and
influence of age on concrete properties.
• Q. 34 (1967, Istanbul): The behavior and deterioration of dams.
• Q. 39 (1970, Montreal): Recent developments in the design and construction
of concrete dams.
• Q. 43 (1973, Madrid): New ideas for more rapid and economic construction of
concrete dams.
• Q. 57 (1985, Lausanne): Concrete dams – an old problem always present:
cracking; a new technology: rolled concrete (rollcrete).
• Q. 62 (1988, San Francisco): New developments in the construction of
concrete dams.
640
GR Q. 104
• Q. 65 (1991, Vienna): Ageing of dams and remedial measures.

• Q. 82 (2003, Montreal): Ageing and rehabilitation of concrete and masonry
dams and appurtenant works.
• Q. 90 (2009, Brasilia): Upgrading of existing dams.
• Q. 95 (2012, Kyoto): Ageing and upgrading.
• Q. 99 (2015, Stavanger): Upgrading and reengineering of existing dams.
As can be noticed from the above list, since the first ICOLD Congress in 1933,
questions related to concrete dams were assiduously and continuously addressed
until the 1973 Congress. Since then, only two questions so far in the last fifteen
Congresses, that is, over a span of almost half a century, the last being Q. 62 in the
1988 Congress (without including questions on ageing, rehabilitation or upgrading
in this count).
As for ICOLD Bulletins related to concrete dams, the following exhaustive

list is available, which does not include those specific to foundations, monitoring,
seismicity, rehabilitation, safety, etc.:
• Bulletin 15 (1960): Frost resistance of concrete. Comparison of results

obtained in different laboratories.
• Bulletin 20 (1968): Surface-active admixtures for concrete for large dams.
• Bulletin 22 (1972): Pozzolanas and slags for concrete for large dams.
• Bulletin 24 (1973): Accelerating and retarding admixtures.
• Bulletin 25 (1976): Extensibility of concrete for large dams.
• Bulletin 26 (1976): Methods of determining effects of shrinkage, creep and
temperature on concrete for large dams.
• Bulletin 30 (1978-1987): Finite elements methods in analysis and design of
dams.
• Bulletin 36 (1982): Cements for concrete for large dams.
• Bulletin 47 (1983): Quality control of concrete.
• Bulletin 57 (1986): Materials for joints in concrete dams. Notes on regional
influences.
• Bulletin 63 (1988): New construction methods – State of the art.
• Bulletin 64 (1988): Dam heightenings – Register
• Bulletin 71 (1989): Exposure of dam concrete to special aggressive waters –
Guidelines
• Bulletin 75 (1989): Roller compacted concrete for gravity dams – State of
the art.
• Bulletin 76 (1990): Conventional methods in dam construction – Review.
• Bulletin 79 (1991): Alkali-aggregate reaction in concrete dams – Review and
recommendations.
• Bulletin 93 (1994): Ageing of dams and appurtenant works – Review and
recommendations.
• Bulletin 94 (1994): Computer software for dams. Validation – Comments and
proposals.
641
GR Q. 104
• Bulletin 107 (1997): Concrete dams – Control and treatment of cracks.

• Bulletin 117 (2000): The gravity dam: a dam for the future – Review and
recommendations.
• Bulletin 122 (2001): Computational procedures for dam engineering – Relia-
bility and applicability.
• Bulletin 126 (2003): Roller-compacted concrete dams – State of the art and
case histories.
• Bulletin 136 (2009): The specification and quality control of concrete for dams.
• Bulletin 145 (Pre-print): The physical properties of hardened conventional
concrete in dams.
• Bulletin 155 (2013): Guidelines for use of numerical models in dam engineer-
ing.
• Bulletin 165 (Pre-print): Selection of materials for concrete in dams.
• Bulletin 177 (2020): Roller-compacted concrete dams.
• Bulletin 184 (Pre-print): Management of expansive chemical reactions in
concrete dams & hydroelectric projects.
To conclude this Introduction, I deem necessary to reveal that, within the

ICOLD Committee “D” on Concrete Dams (CCD), we welcomed as fantastic news
that in the 87th General Assembly, held in Ottawa (Canada) in June 2019, an
“update” on the state of art of concrete dams was selected as one of the Questions
for the present Congress in Marseille. And we are very grateful to ICOLD President
Michael Rogers, the former Chairman of the CCD, who raised such proposal as
the “Question from the President”, that was unanimously accepted. Q. 104 com-
prises five sub-questions, which correspond to Sections 3 through 7 of this General
Report.
I invite readers to join me in looking over this General Report, where we will
know about some of the latest developments related to concrete dams, taking as a
baseline the excellent reports received for Question 104 from several leading experts
in this field.
2. OVERVIEW OF RECEIVED REPORTS
As contributions for Q. 104, a total of 41 reports have been received, from 19

countries belonging to four continents. The countries that have submitted a greater
number of reports are China, with six reports, and France and Italy with five reports
each. Three reports have been received from India and Switzerland, two from Great
Britain, Indonesia, Iran, Japan and Spain and, finally, nine different countries have
contributed with one report each.
Table 1 below summarizes the distribution of reports by continent, showing

that contributions from Europe and Asia amount to 88% of the total, with Europe
alone providing slightly more than half.
642
GR Q. 104
Table 1
Geographical distribution of reports
CONTINENT REPORTS % COUNTRIES
Europe 21 51% 9
Asia 15 37% 5
America 3 7% 3
Africa 2 5% 2
Total 41 100% 19
However, it should be noted that two of the reports, despite their good quality,
do not precisely fit into any of the sub-questions ascribed to Q. 104, so they have
therefore not been considered in preparing this General Report.
The small number of reports received from national committees of Africa and
the Americas is striking, although if we make a different distribution by the location
of the dams addressed in the reports, we have a somewhat different picture; see
Table 2 (please note that some reports discuss several dams).
Table 2
Geographical distribution of dams addressed in
the reports
CONTINENT DAMS % COUNTRIES
Europe 27 39% 8
Asia 22 32% 12
Africa 13 19% 10
America 7 10% 5
Total 69 100% 34
The majority of the reports received deal with several topics within one of the
five sub-questions assigned to Q. 104, and in some cases, with topics belonging to
different sub-questions. For that reason, Table 3 has been put together with the aim
to have at a glance a summary of the topics covered by the different reports, and
to use it as a tool to decide for each report what is the sub-question in which they
best fit, and to assemble the general layout of this General Report. Topics with zero
contributions received have been omitted from the table.
The distribution of the reports by type of dam discussed is as follows, keeping

in mind that a few reports refer to the same dam, and others address more than
one dam of the same or different typology: 16 reports discussing RCC gravity dams,
11 on CVC gravity dams, 9 on CVC arch dams, 3 on RCC arch dams, 1 on CVC
buttress dams and 1 on CMD dams and, additionally, 2 on raising CVC arch dams
and 1 on raising CVC gravity dams.
643
GR Q. 104
Table 3
Classification by sub-questions and topics of the reports contributing to
Question 104
SECTION IN .
GENERAL NO REPORT
SUB-QUEST. REPORT TOPIC REPORTS NUMBERS
a) 3.1 CVC arch dam analysis 9 R2, R11, R15, R18, R31, R32,
and design R33, R34, R42
3.2 CVC arch dam 2 R18, R34
construction
3.3 RCC arch dam analysis 3 R16, R17, R23
and design
3.4 RCC arch dam 1 R17
construction
b) 4.1 CVC gravity dam analysis 9 R1, R2, R12, R19, R20, R25,
and design R39, R43, R45
4.2 CVC gravity dam 2 R21, R39
construction
4.3 RCC gravity dam analysis 6 R7, R9, R17, R22, R23, R40
and design
4.4 RCC gravity dam 5 R7, R10, R13, R24, R30
materials and mixes
4.5 RCC gravity dam 7 R14, R17, R24, R27, R28,
construction R29, R30
4.6 CMD analysis, design, 1 R8
materials and
construction
c) 5.1 Innovations for raising 3 R34, R36, R43
existing concrete dams
with CVC
d) 6.1 Extremely high CVC arch 2 R31, R32
dams
6.2 Extremely high RCC arch 2 R16, R17
dams
6.3 Extremely high RCC 4 R4, R7, R14, R22
gravity dams
e) 7.1 Performance of CVC arch 5 R18, R31, R33, R36, R42
dams
7.2 Performance of CVC 5 R3, R20, R25, R36, R44
gravity and buttress
dams
7.3 Performance of RCC 2 R4, R38
gravity dams
3. INNOVATIONS FOR ARCH DAM ANALYSIS, DESIGN AND

CONSTRUCTION, INCLUDING RCC ARCH AND ARCH-GRAVITY DAMS
Even though the construction of new concrete arch dams is somewhat limited,
compared with the construction of new concrete gravity dams, mainly due to the fact
that the former require dam sites with better geologic-geotechnical conditions, and
certain proportions with respect to the dam site geometry, it is a type of dam very
competitive when the above requirements are favorable. Particularly for very high
and extremely high dams, as we will see some examples in Chapter 6.
644
GR Q. 104
In fact, most of the highest dams constructed (or in advanced state of con-
struction) in the world, in the last fifteen years, have been double-arch or arch-gravity
concrete structures, and more specifically, all but one of those exceeding 269 m (6
double-arch dams), and 10 out of the 16 in excess of 229 m. These 10 double-
curvature or arch-gravity dams have been constructed (or are under construction)
in China (7 dams), Turkey (2) and Iran (1).
It should be pointed out that those extremely high arch dams are all CVC dams.
The highest RCC arch dam has a more “modest” height of 168 m. Tables 4 and 5
below show, respectively, the tallest completed or under construction dams in the
world, of all types, and the tallest completed RCC arch dams in the world.
Table 4
The highest dams in the world (250 m and above), all types (ICOLD, April 2020)
NAME OF DAM COUNTRY HEIGHT (M) DAM TYPE YEAR OF COMPLETION
Shuangjiangkou China 312 Embankment U.C.*
Jinping I China 305 Double-arch 2013
Nurek Tajikistan 300 Embankment 1980
Xiaowan China 294 Double-arch 2012
Baihetan China 289* Double-arch U.C.*
Xiluodu China 286 Double-arch 2014
Grande Dixence Switzerland 285 Concrete Gravity 1961
Yusufeli Turkey 275* Double-arch 2020*
Enguri Georgia 272 Double-arch 1987
Wudongde* China* 270* Double-arch* U.C.*
Nuozhadu China 262 Embankment 2014
Chicoasén Mexico 262 Embankment 1980
Tehri India 260 Embankment 2006
Laxiwa China 250 Double-arch 2010
Mauvoisin Switzerland 250 Double-arch 1957
*Updated by the General Reporter
Table 5
The highest RCC arch dams in the world (130 m and above) (Dunstan, 2020)
NAME OF DAM COUNTRY HEIGHT (M) DAM TYPE YEAR OF COMPLETION
Wanjiakouzi China 168 Thin-arch 2011
Janneh Lebanon 157 Arch-gravity 2021
Sanhekou China 142 Arch-gravity 2020
Xiangbiling China 142 Thin-arch 2017
Qinglong China 140 Thin-arch 2012
Yunlonghe N◦ 3 China 135 Thin-arch 2011
Dahuashui China 135 Thin-arch 2007
Sanliping China 133 Thin-arch 2011
Gomal Zam Pakistan 133 Arch-gravity 2011
Lizhou China 132 Thin-arch 2014
Shapai China 132 Thin-arch 2003
Looking at both tables above, it is clear that a vast majority of the most recently
constructed high height arch dams, both with CVC and RCC, have been in China,
645
GR Q. 104
a country where several extremely high arch dams are currently being designed to
be built in the next few years.
As a clarification note to Table 5, it is necessary to explain that among the

highest “thin-arch” RCC dams, all of them in China, during the early stage they didn’t
adopt negative slopes, except for some that had slight negative slopes (e.g. Shapai),
so essentially they could be considered more properly to be single-curvature arch
dams rather than double-curvature. However, some of those built in the last fifteen
years are “true” double-curvature arch dams, although the vertical curvature of the
cantilevers is generally not as pronounced as in CVC double arch dams, as well as is
more limited the degree of horizontal curvature, in order to facilitate the construction
with RCC.
Wanjiakouzi in China is currently the highest RCC arch dam in the world, but
planned 185 m high Batoka Dam, on the Zambezi River across the international
border between Zambia and Zimbabwe, could be the next highest RCC arch dam,
although the final dam typology has not yet been selected at this time.
It took several years of building RCC gravity dams to become confident enough
to design and build extremely high RCC gravity dams, close and above 200 m in
height. Similar process can be observed in construction of RCC arch dams where
experience and confidence gained over time will result in dams built of same heights.
Another factor that could be impacting the selection of the RCC typology for high
height arch dams is the fact that the advantages of very rapid construction with RCC
are not so easy to obtain in an arch dam, due to the smaller areas of placement. The
inconvenience of implementing the systems for the post-grouting of the transverse
contraction joints, that was experienced in the beginning, has been overcome, and it
is something that should not hinder significantly the rates of RCC placement (ICOLD,
2020).
If the topography and/or geological conditions of the dam site are not com-
pletely suitable for a thin CVC arch, but they are for a thick-arch or arch-gravity
RCC dam, this kind of structure probably represents the optimal solution. It will have
significant savings in concrete volume compared with an RCC gravity dam and, in
certain cases, by optimizing the arch geometry and materials for the applicable ther-
mal loads, the need for post-cooling and joint grouting can be eliminated, making this
solution even more attractive and competitive with respect to a gravity dam solution
(Nisar et al., 2008 and ICOLD, 2020).
To date, the total number of RCC arch dams (arch-gravity and thin-arch) con-
structed or under construction, with a height in excess of 15 m (de facto all of them
exceed 30 m), amounts to 85, of which 27 (i.e. 32%) are higher than 100 m. It is
interesting that in China, the country leading by far the construction of RCC dams,
with nearly 29% of the total, 31% of the RCC dams are arch dams, whereas in the
rest of the world the RCC arch dams amount to less than 2%; only 11 RCC arch
dams out of 607 total RCC dams, excluding hardfill dams (Dunstan, 2020).
646
GR Q. 104
As a recommended bibliography for RCC arch dams, Chapter 9 of the recent

ICOLD Bulletin 177 on RCC dams (ICOLD, 2020) is a very valuable and updated
source of information, which in turn cites 23 references. Some additional ones of
interest are (DU, 2021) and (Shaw, 2017).
And with regard to CVC arch dams, among many others, including those
incorporated as “References” in the reports received, some papers and publications
of interest that can be consulted, and are listed in the “References” of this General
Report, are the following: (Akbari et al., 2011), (CFBR, 2018), (FERC, 2018), (Jonker
& Espandar, 2014), (Nuss et al., 2012), (Paixão & Silva Matos, 2014), (Shaw, 2015),
(Varghese et al., 2014), (Vicente et al., 2017) and (Wang, 2016).
To complete this general introduction to Section 3, and what follows also

applies to Section 4, it is important to bring up that the expanded computer technol-
ogy in the 1970s allowed implementation of numerical methods in structural analysis
of dams. Over the last thirty years, the analysis methods have further advanced with
several advanced commercial programs available to the engineers. Complex linear
and nonlinear models in frequency and time domains are now used in analysis
and design of new dams and modifications to existing structures. Advancements
in the computation methods supported by the high computing power of the new
computer generation allow realistic simulations of the dam/foundation/reservoir sys-
tem behavior and better understanding of its performance. However, a complexity
of the analyzed engineering and mathematical problems may introduce high levels
of uncertainty in the results. Therefore, the advanced structural analysis models of
dams require appropriate evaluation of efficiency, reliability, and validity. Engineer-
ing judgments and proper interpretation of the results is an important component in
the structural assessment of concrete dams (ICOLD, 2021, and Chopra, 2020).
3.1. CVC ARCH DAM ANALYSIS AND DESIGN
A total of nine reports have been received, in which the subjects of structural
analysis/modelling and design for CVC arch dams are at least one of those they
deal with, as reflected in Table 3.
Noorzad et al. (R2, Iran) carry out an investigation on the role of wave propa-
gation effects in the correct estimation of stress levels in the body of a concrete dam.
In the study, a half-space with a semi-cylindrical valley is analyzed under inclined
harmonic shear waves (SH), with various angles of incidence, for investigation of
surface amplification, using a domain reduction method (DRM). The results of the
analysis are verified with analytical solutions of the problem, indicating the accuracy
of the implemented method. The investigation, although is applicable to both, gravity
and arched concrete dams, realizes that the wave propagation effects result more
significant for the latter.
In his report, Chopra (R11, USA) presents an in-depth discussion on earth-

quake analyses of arch dams. The report begins highlighting the flaws of the
647
GR Q. 104
substructure method, which was a very good tool some decades ago, but became
obsolete after the appearance of newest computational techniques. The report also
emphasizes that until not long ago, and due to the restricted capabilities of most
commercial finite element analysis (FEA) programs, became popular to ignore the
mass and material damping of rock, as well as compressibility of water, although
it grossly underestimated response. After this introduction, the paper focuses in
outlining the direct finite element method (FEM) and explaining its implementation
in the case study of Morrow Point Dam, as well as outlining methods for selection
and scaling of ground motions (GMs) for the dam site. The report explains how the
direct FEM overcomes the limitations of the standard FEM by introducing wave-
absorbing (or non-reflecting) boundaries at two locations: 1) at the upstream end of
the fluid domain, to model its essentially semi-infinite length, and 2) at the bottom
and side boundaries of the foundation domain, to model its semi-unbounded geom-
etry. Additionally, the model of the fluid domain now includes water compressibility
and wave absorption at the reservoir bottom sediments, and the model of the foun-
dation domain includes mass, stiffness, and material damping appropriate for the
rock. Water-foundation interaction is also considered. Consequently, the untenable
assumptions of massless rock foundation and incompressible water in the popular
FEM are avoided.
Robbe et al. (R15, France) describe a study on the shear phenomena at

the concrete-rock contact of arch dams. The report emphasizes that most of the
existing arch dams in France were built over the two decades following World War
II, when the design methods used at the time were based on simplified calcula-
tion methods founded on an assessment of the upstream and downstream face
stresses and, in particular, did not include an analysis of the state of shear at the
concrete-rock contact. The paper refers to the recent French Guidelines for analyz-
ing the behavior of arch dams (CFBR, 2018), which suggest assessing the shear
state at the base of the cantilevers, and describe some related failure mechanisms,
differentiating under the central, and at the side cantilevers. Then, a comparative
numerical analysis of ten existing arch dams of between 23 and 150 m high, is
performed, which reveals considerable shear at the cantilever-foundation contacts
with mobilized cohesion sometimes exceeding 1 MPa, an order of magnitude that
can easily be found from the simple reasoning that, on average, 1/3 of the hydro-
static load bearing on the central cantilever is transferred by shear at the cantilever
toe. The results were obtained using finite-element or finite-difference method cal-
culations, with nonlinear models capable of releasing the tensile stresses, often
at a previously implanted joint. For most of the arch dam analyzed, no down-
stream displacement at the toe was measured. Finally, the specific mechanical
behavior of four double-arch and one single-arch dams, with heights ranging from
23 to 70 m, analyzed by several major consulting companies, is discussed, with
the description of the different modelling applied to each particular case, together
with the reinforcement solutions, where appropriate, implemented: reinforcement
of the downstream rock abutment, reinforcement by anchoring the arch’s supports,
thickening the side cantilevers and piezometric monitoring at the toe of the central
cantilevers.
648
GR Q. 104
Another report from France, Grimal et al. (R18), deals with concrete patholo-
gies and, in this context, with the modelling of concrete affected by chemical
reactions. In the case of the dams, chemical expansion like alkali silica reaction
(ASR) induces damage to the material that can compromise the concrete mechan-
ical properties and overall stability of the structure. Numerical models capable of
describing the internal state of concrete, as well as its evolution, can be an effective
tool in assessment of structural conditions of dams in order to continue their opera-
tion. The models can assess if the structure is needed or not to be strengthen, and
if the outcome is “yes”, then they make it possible dimensioning the most adequate
restoration works and optimize their cost. The report firstly describes a method-
ology developed for the safe operation of concrete dams with swelling, based on
three pillars: surveillance, laboratory tests and numerical modelling for assessing
the behavior of affected structures, with the modelling taking a predominant place.
The methodology is founded on four successive phases. Secondly, in the report it is
presented the phenomenological modelling used at EDF/CIH, in order to simulate the
internal swelling due to the alkali-aggregate reactions (AAR) or the internal sulphate
attack (ISA). This modelling uses the framework of poromechanics and takes into
account the coupling of the various phenomena evolving in the concrete including:
shrinkage, creep, anisotropic damage and reactions of internal swelling. In addi-
tion to explicate the behavior of the structure in its current state, the model allows
prediction of long-term deformations and stresses of concrete structures affected
by swelling, and has been compared with real cases for verification purposes. The
model has been implemented on several dams in France, such as the Chambon
(AAR), Maury and Castelnau (ISA) dams and also on some Hydro-Quebec dams
(AAR) in Canada. Finally, it is described the case history of Bimont double-curvature
arch dam (h = 86.5 m) affected by swelling due to ISA, and the rehabilitation works
proposed and carried out, which are addressed in Section 3.2, based on the results
of a 3D numerical modelling and two investigation campaigns.
The next two reports dealing with CVC arch dam analysis and design have
been received from China, Zhong et al. (R31), and Lijun et al. (R32), and both
papers are referring to the currently highest dam in the World, Jinping I (h = 305 m).
Both reports are discussed in Section 6.1, which addresses extremely high CVC
arch dams.
The first of two reports received from Switzerland for this topic, Ballisat et al.
(R33), describes an interesting experience of decommissioning Spitallamm arch-
gravity dam (h = 114 m), to be replaced by a double-curvature arch dam. The new
dam, located just downstream of the original one, was designed to eventually be
heightened by 22 m in the future. The performance and problems experienced by the
existing dam, that led to its decommissioning and replacement by a new dam, are
discussed in Section 7.1. In the present section, we focus on the part of the report
regarding the design of the replacement structure. After discarding the raise by
constructing a new double-curvature arch dam directly placed on the sound mass
concrete of the existing dam, the concept of building an entirely new dam down-
stream of the existing one presented two important advantages: 1) No interference
649
GR Q. 104
with the operation of the existing reservoir; 2) No need for demolition of the exist-
ing dam, that could be left in place with a bypass system to equilibrate the water
levels at the reservoir and in the open space between both dams. Considering the
space limitations and the excellent foundation conditions, a double-curvature arch
dam was the best alternative, if not the only one feasible solution. The crest of the
dam was settled at the same elevation as the existing dam, with the commented
possibility of a future raising. The geometry of the new dam was defined by sets of
parabolas both horizontally and vertically. 3D stress analyses were performed for
both, the basic configuration and for the raised dam, to make sure that no strength-
ening would be required at a later stage for the raising. For that reason, relatively
thick arches were adopted for the basic configuration. A dynamic analysis of the dam
was also conducted applying three nondependent synthetic accelerograms to the
structure. Maximum horizontal tensile stress was found to reach 3 MPa, but it cannot
be considered as critical, as the vertical block joints will open during strong shaking
and the stress will be relieved. Foundation excavation was carried out throughout
2019 and 2020, with additional measures taken to guarantee the safety of the exist-
ing dam, under operation, and the works in progress. A comprehensive monitoring
program was implemented, with observation that will continue during the concreting
of the new dam, scheduled from 2021 to 2024. Full impounding of the new dam once
the bypass gallery will be operational (2025) will also require a careful monitoring of
the structure and its foundation.
The second report received from Switzerland related with this topic, Leite
Ribeiro et al. (R34), addresses two interesting cases of raising double-curvature
arch dams. As Section 5.1 is specific for innovations for raising existing concrete
dams with CVC, we have considered including the discussion on this paper in that
section, except for the part related to construction, which is included in Section 3.2.
Bjønness et al. (R42, Norway) present a paper describing the general pro-
cess involved in using data from instrumentation in order to verify the capacity of an
existing concrete arch dam. The data from the instrumentation is used as a basis
for load definition and as a calibration of a FE model. It was found that the dam
was highly influenced by temperature load. When emptying and filling the reservoir,
the change in structural response to the hydrostatic load was relatively small com-
pared to the change caused by the varying temperature. Therefore, temperature
loading is very important when assessing the capacity of arch dams. A transient
thermodynamic analysis, using Abaqus, was performed, and the program ShellDe-
sign was used in order to analyze the nonlinear behavior of the dam. The resulting
temperature obtained from Abaqus was used as loads for the nonlinear structural
analysis in ShellDesign. The nonlinear FE model was calibrated by comparing rel-
ative displacements between the nonlinear FEA and the deformations recorded by
the instrumentation, for various time periods. The capacity of the concrete arch was
checked in accordance with Eurocode 2. But, due to there is no specific standard
for dams in the Eurocode system, the capacity was also verified using the “Modified
Compression Field Theory” (MCFT) developed at the University of Toronto. The two
methods resulted in similar results, indicating that the structural capacity of the vault
did not fulfill the requirements.
650
GR Q. 104
3.2. CVC ARCH DAM CONSTRUCTION
Report R18 from France (Grimal et al.), dealing with concrete pathologies, has
already been discussed in Section 3.1 with respect to dam analysis. Although the
report is not really related to the construction of new CVC dams, we have consid-
ered it to be of interest to this section, since a part of the report addresses major
rehabilitation works undertaken at Bimont Dam from 2017 to 2019. The modifica-
tions were based on the combined contribution of the investigation carried out in two
campaigns (coring and TV camera inspection of borehole walls) and the outcome of
the 3D numerical modelling. Briefly described, the scope of the rehabilitation works
was the following:
• Passive anchors positioned to reinforce the structure prior to the grouting of

cracks.
• Grouting boreholes intersecting the laminating cracks.
• Boreholes intersecting the open joints at an angle in order to key the structure
after the crack grouting work.
Additional works performed related with the problematic of swelling were:
• Installation of a CARPI waterproofing geomembrane on the upstream face on

the right bank, in order to limit the inflow of water and avoid a resumption of
swelling.
• Installation of vertical prestressed tie rods in order to stabilize the right bank
abutment, subjected to the forces caused by hydrostatic, summer thermal, and
swelling loads.
After the rehabilitation works, the impounding of the reservoir, emptied to

make possible the job, began in the spring of 2019, carried out in steps, with careful
monitoring of the dam’s behavior throughout the impoundment process.
Report R34, from Switzerland (Leite Ribeiro et al.) is discussed in Section 5.1,
except for the construction part, which fits better in this section. The report describes
the background, design and construction of two cases of arch dam raising: Cam-
bambe Dam in Angola, raised by 20 m between 2012 and 2018, to a new height of
92 m, and Vieux Emosson Dam in the Swiss Alps, raised by 21.5 m between 2012
and 2015, to a new height of 76.5 m. For both cases, the report explains in detail the
preparatory works carried out, including demolition of significant parts of the upper
part of the existing structures. In the case of Vieux Emosson Dam, drilling and
blasting was abandoned (actually, smooth and controlled blasting) and substituted
by using two pavement milling machines. Another critical activity in the construction
was the preparation of the contact between the old and new concrete. At Cambambe
Dam, several measures were taken in order to guarantee adequate contact, and to
monitor the behavior of the contact, including an inspection and drainage gallery. At
Vieux Emosson Dam, all contact surfaces were treated with hydro-blasting, and the
651
GR Q. 104
new concrete in direct contact with the old concrete was of a special concrete mix
with small MSA (32 mm instead 63 mm). Additionally, steel bars were anchored into
the old concrete to strength the bonding between both concretes. Also, an inspection
gallery was arranged above the interface between both concretes. Other important
aspects that greatly affected the construction were the river diversion, in the case of
Cambambe, and the weather conditions in the case of Vieux Emosson. Finally, the
particularities of the joint grouting in both dams are narrated in detail.
3.3. RCC ARCH DAM ANALYSIS AND DESIGN
Report R16 received from France (Si-Chaib et al.), discusses in its first part
possibilities and benefits offered by numerical analyses to assist in the design of
thick RCC arch dams and to understand the long-term thermo-mechanical behavior
of dams in general. The 157 m high Janneh Dam in Lebanon, the second highest
RCC arch dam in the world, was used as a study case. The second part of the report
focuses on a crossed feedback between numerical models and construction, based
on the experience of two thick RCC arched dams, the 121 m high Tabellout Dam in
Algeria, already built, and a 74 m high unnamed dam which is currently under con-
struction in South-East Asia, both of them in contexts of intense seismic activity and
relatively difficult geological conditions. The benefits and limitations of 3D nonlinear
numerical models are presented and discussed, as well as the retrospective valida-
tion of the adopted mechanical properties, and the chosen construction provisions
such as joint grouting system and seismic belt are also discussed.
A second report received from France for this topic (R17, Andrian et al.), also
deals with construction of RCC arch dams (Section 3.4), with RCC gravity dams
design and construction (Sections 4.3 and 4.5) and with very high RCC arch dams
(Section 6.2). In order not to split the presentation of the report into several sec-
tions, keeping in mind that it includes a comparison of several case studies, we will
discuss R17 only in this section. The first part of the report addresses general con-
cepts about the wide possible range of cementitious content in the mixes. The mix
can be designed to provide both, structural strength and watertightness, or only the
required structural strength, with the watertightness ensured by an external facing.
In the report three case studies are presented, the Janneh arch-gravity RCC dam in
Lebanon, and two RCC gravity dams, Nam Ngum 4 in Laos (h = 74 m), and Hatta
in Dubai (h = 70 m). Some of the most important features to highlight are as follows.
The concrete mix for all three dams involves high-cementitious, high-workability
RCC, with a VeBe consistency of 15–18 seconds. For Janneh Dam, the 10 m strip
next to the upstream face uses a richer RCC mix. Upstream zone has (in kg/m3 )
77 cement + 143 ground granulated blast furnace slag (GGBFS), for a total of 220,
whereas the downstream zone has 56 cement + 104 GGBFS, for a total of 160. The
respective mixes for Nam Ngum 4 and Hatta incorporate 82 cement + 100 flyash for
a total of 182, and 50 cement + 110 flyash for a total of 160. The above two mixes
with 160 kg/m3 total cementitious content are close to the lower limit to be considered
high-cementitious RCC mixes. At Janneh and Hatta a retarding admixture was incor-
porated to the RCC mix. As an important feature, at Janneh the transverse joints will
652
GR Q. 104
only be grouted in case of high leakage, as a mitigation measure, taking advantage

of the long-term low shrinkage and creep expected for the RCC, even with the use
of GGBFS instead low lime flyash, which was tested at the Full-Scale Trial (FST).
Cores taken from the FST at Nam Ngum 4 showed that the aggregates in RCC
are rather more in contact between each other when compared to that of the fac-
ing grout-enriched RCC (GERCC). The aggregate skeletal arrangement is believed
to be a factor contributing to the low long-term shrinkage/creep of high workability
RCC. The thermo-mechanical analysis suggested a 24 m transverse joint spacing,
which turned out to be 30 m for the Hatta Dam, where the original concept of an
upstream zone with a systematic bedding mortar and a medium-cementitious RCC
downstream zone, was replaced by a unique high-cementitious RCC zone. The limit
for the hot joint was established at Janneh and Hatta as the initial setting time of the
RCC, and the limit for the warm joint based on a Modified Maturity Factor (MMF) of
1000 ◦ C x h. The limit for the hot joint at Nam Ngum 4 was the initial setting time of the
RCC minus 1 hour, and the limit for the warm joint was stipulated in a table with differ-
ent time limits (hours) for the different seasons. The paper also explains the different
RCC placement methods applied to the three dams and the formworks used. In the
last part of the report, valuable information is given regarding the construction of sev-
eral large RCC dams: Souapiti (Guinea, h = 115 m), Tha Dan (Thailand, h = 92 m),
Gibe III (Ethiopia, h = 250 m), GERDP (Ethiopia, h = 170 m), Bui (Ghana h = 108 m),
and Theun Hinboun (Laos, h = 70 m), with regards to placement methods, RCC pro-
duction and placement rates (up to an average monthly rate of 180,000 m3 and max.
monthly rate of 330,000 m3 at GERDP), RCC transportation, use of rainfall radars
to help maximize RCC placement, and management of floods during construction
allowing overtopping over the portion of the structure already built.
Finally, the report authored by Shaw (R23, South Africa) fits in this section
and in Section 4.3. In the report the author describes the differences between the
early behaviour of RCC and CVC in large dams and how these differences involve
a slightly different design approach, taking cognisance of an additional property of
concrete in the form of stress-relaxation creep (SRC). It is highlighted the differ-
ence between the vertical and horizontal approaches for the construction of CVC
and RCC dams, respectively, which together with the use of high levels of sup-
plementary cementitious materials (SCM) typical in RCC mixes, introduce two key
factors that can together create significantly different early behaviour and related
design requirements. Firstly, the levels of concrete shrinkage experienced during
the hydration process can be substantially less than traditionally assumed for CVC
when high levels of flyash are used. Secondly, horizontal concrete placement implies
that the mass of the dam will not necessarily be carried directly, or vertically down
into the foundation, particularly when concrete shrinkage levels are low. Therefore,
the typical historical design based on the 2D analysis of CVC gravity dams may not
be completely valid for RCC gravity dams. The above fact is more important when
the RCC dam has lower length/height ratio (< 6), and in this scenario, knowledge of
the applicable level of SRC is essential in developing a meaningful understanding of
the 3D structural function. On the contrary, when the gravity dam is long in relation
to its height and adjacent blocks are of similar height, the level of SRC will be of little
importance and the dam can be designed, or analyzed, as 2D structure. In the case
of RCC arch dams, an upstream movement during construction was measured, that
653
GR Q. 104
is effectively equivalent to a lengthening of the dam crest of 30 to 50 microstrain

on a structure of 70 to 100 m in height, implying the presence of horizontal com-
pression along the structure, which substantially impacts the distribution of gravity
loading onto the dam foundation and within the structure itself. The author states
that a full understanding of the structural function of an RCC arch, arch-gravity, or
gravity dam in a narrower valley constructed in low SRC RCC can only be devel-
oped through an appropriate thermo-mechanical analysis that takes cognizance of
the actual variation of SRC with compression stress level.
3.4. RCC ARCH DAM CONSTRUCTION
Report R17, from France (Andrian et al.), is the only one that has a part of it
that deals with the topic of RCC arch dam construction, specifically referring to the
Janneh Dam in Lebanon, but has been presented in Section 3.3.
4. INNOVATIONS FOR CONCRETE GRAVITY DAM ANALYSIS, DESIGN

MATERIALS AND CONSTRUCTION, INCLUDING RCC AND CMD
It is a fact that soon after its inception in the early 1980s, RCC has significantly
replaced CVC as the standard method for the construction of concrete gravity dams.
To date, the total number of RCC gravity dams built or under construction
(excluding hardfill dams), with a height greater than 15 m, amounts to 761, of which
121 (that is, 16%) exceed 100 m, and 4 surpass 200 m (Dunstan, 2020).
Table 6 shows the ten highest completed or under construction RCC dams in
the world, with nine of those ten being gravity dams.
Table 6
The ten highest RCC dams in the world (Dunstan, 2020)
YEAR OF
NAME OF DAM COUNTRY HEIGHT (M) DAM TYPE COMPLETION
Gibe III Ethiopia 246 Gravity 2016
Longtan China 217 Gravity 2009
Huangdeng China 203 Gravity 2019
Guangzhao China 201 Gravity 2009
Miel 1 Colombia 188 Gravity 2002
Ayvali Turkey 177 Gravity 2015
Wanjiakouzi China 168 Thin arch 2011
Guandi China 168 Gravity 2013
Nam Ngiep 1 Laos 167 Gravity 2019
Lushi China 165 Gravity 2018
654
GR Q. 104
From the table above we can see that six of the ten highest RCC dams have
been built in China, and that seven of the ten have been completed within the last
decade, and the latter suggests that there is increasing confidence to build higher
and higher RCC dams.
Despite the proliferation of the construction of RCC gravity dams in recent

decades, to the detriment of CVC gravity dams, for this Section 4 up to 14 reports
have been received dedicated to RCC dams and, surprisingly, up to 10 to CVC dams
(one is actually a buttress dam), and only one report to a CMD (a hardfill dam).
4.1. CVC GRAVITY DAM ANALYSIS AND DESIGN
The first of the two reports received from Iran (Noorzad et al., R1), discusses
that earthquake analysis of concrete gravity dams is mostly performed under the
assumption of elastic foundation, while considering a nonlinear response of the foun-
dation rock has been addressed infrequently. However, it has been observed that
the dam-foundation interface is highly susceptible to cracking due to its geometrical
properties, whereas assuming elastic response for foundation prevents crack prop-
agation within this medium and forces the crack to propagate in a straight direction
at the interface. This is why the paper analyzes a dam-foundation-reservoir system
considering nonlinear responses of both, the dam and the foundation, subjected
to seismic loading. A comparison analysis is also performed with linear response
of the foundation rock. As a prototype for the investigation, Koyna concrete gravity
dam was chosen. The results reveal a completely different crack pattern at the heel
of the dam, depending on whether the foundation is considered elastic, or with a
nonlinear properties. In the first case the crack is forced to propagate along the
dam-foundation interface, whereas in the second case crack propagates within the
foundation rock, and a different overall response for the whole system is obtained.
The second report received from the same authors, Noorzad et al. (R2, Iran),
in addition to CVC gravity dams it is also related to CVC arch dams, and has therefore
been previously presented in Section 3.1.
Report R12, from the Czech Republic (Hrazdira & Švancara), presents an
interesting case of increasing the evacuation capacity of an existing concrete grav-
ity dam. The Orlík Dam (h = 91 m), creates the largest volume reservoir in the Czech
Republic, 716.5 Mm3 , and 68 km long. The dam was built between 1957 and 1960
as a main component of a hydropower scheme, but it also serves other purposes,
including flood control, recreation, and ensuring a minimal water flow in the Vltava
River. Some years ago, much more stringent regulations governing safety assess-
ment of dam structures during floods were adopted by the Czech authorities, and
a safety improvement program was then initiated on the majority of Czech dams.
Belonging to Category I, a 10,000-year flood turned out to be applicable for Orlík
Dam, with a peak discharge of 5,300 m3 /s. As a valuable precedent, in August 2002
the dam experienced the largest flood event along the dam history. The outflow
655
GR Q. 104
reached 3100 m3 /s, and the peak inflow appreciably exceeded the 100-year flood
value. The water level in the reservoir exceeded the design level, and although
the dam withstood the flood reliably, the scheme suffered considerable damage,
mainly concerning the substructures adjacent to the dam. A program of analyses
and preparatory work for the future refurbishment of the dam was initiated based
on the assessment of the 2002 flood event. A physical model of the dam body
and spillways was set up in 2009 to assess the parameters of the discharge and
extrapolating them to water levels matching extreme floods, specifically a 10,000-
year flood. Comprehensive calculations using FEM were made during 2013–2015
based on preceding engineering geological survey of the bedrock and, in partic-
ular, of the contact between the dam body and the bedrock in the dam base. A
study analyzing several possible options to improve the safety of the dam during
extreme floods was carried out in 2010, and later evaluated in the feasibility study
developed in 2013, with respect to technical feasibility, costs, operational aspects,
and safety aspects, as well as to the impacts on the scheme during the construction
phase. The finally selected alternative consisted of 3 new gated spillways outside the
dam, next to the right abutment. Immediately after this decision was taken, detailed
engineering-geological surveys, geodesic surveys and inspections of the concrete
structures were performed. The new spillways design was refined by 3D modelling
(CFM), used to assess and optimize the flow, and subsequently verified and fur-
ther optimized by using the physical model. Their combined capacity amounts to
3 x 588 m3 /s. The construction works are planned to start in mid-2021 and will be
managed based on the BIM methodology.
Report R19, authored by Rashek (Canada), discusses on seismic analysis for

dam appurtenant structures, such as spillways, intakes, outlets, penstocks, gates,
control buildings, and their associated equipment. Failure in any of these structures
could result in loss of reservoir control after an earthquake. But the input earthquake
demand for an appurtenant structure is different from the free-field earthquake haz-
ard used in the analysis of dams. The characteristics and intensity of this demand
depends on the appurtenant location and support conditions; in fact, peak acceler-
ations at the top of some dams can be amplified even ten times the values observed
at its base. The report explains the steps to perform the seismic evaluation of a
dam and its appurtenants, based on the analysis results. Also, it defines the seismic
analysis of the dam and the appurtenant at three levels of sophistication. Then, it
discusses the Response Spectrum Analysis (RSA) of the appurtenant, and how to
develop the Floor Design Spectra (FDS), that can be used to design and evaluate
appurtenant structures, select equipment for a prescribed level of shaking, or deter-
mine if a set of components need to be isolated in order to improve their ability to
withstand the imposed shaking. Since this advanced level of analysis might be pro-
hibitive for a quick evaluation of the appurtenant, the author provides an approximate
solution that can be used for quick seismic evaluations. The proposed approximate
solution is based on modifying the classic RSA procedure to account for the pos-
sible resonance of the appurtenant to the primary modes of vibration of the dam.
In order to achieve this goal, the modal analysis results for the dam-foundation-
reservoir system are obtained using FEA of a complete or simplified model of the
dam. The simplified solution is intended to be used for preliminary design, sensitivity
656
GR Q. 104
and parametric studies, and as a verification tool for more rigorous time-history anal-
ysis. In the last part of the report, the above procedure is established for the case of
a gated spillway structure. A FE model of the dam (including the foundation, spillway,
piers, and the trunnion block) was created using ANSYS software, and the analysis
was performed to estimate the approximate in-structure response spectrum at the
trunnion of a gated spillway.
Malm et al. (R20, Sweden) present a report that conforms to this section, but
also to Section 7.2, and we have considered including the discussion about it in that
section.
Likewise, report R25 received from Bulgaria (Tzenkov et al.) conforms to this
section and also to Section 7.2, but in this case, we include the discussion on it in the
present section. The first part of the report narrates the deterioration along 75 years
of operation of the 50.7 m high Beli Iskar concrete gravity dam, located at an eleva-
tion greater than 1800 m a.s.l. in the Rila Mountain, an area with very harsh winter
conditions. Main problems experienced by the dam include: undesirable construc-
tion joints and cavities (due to poor compaction, only by hand ramming) allowing
water to seep into the dam body deeper and deeper over the years, damage to the
block joints, leaks everywhere in the inspection gallery with carbonized deposits,
damage (corrosion) to the reinforced concrete structure of the valve chamber and,
above all, the ASR and the aging of the concrete. This finding was also confirmed by
concrete samples taken in four different campaigns between 1961 and 2013, and
therefore specimens of different ages were examined. Drainage boreholes in the
dam foundation were inoperative and new ones were drilled three times. The verti-
cal drainage curtain in the dam body does not work since 2002, and this leads to the
development of an inadmissible seepage field inside the dam structure and addi-
tionally affects the stresses field in the dam body and its stability. In addition to the
purely physical reasons for the urgently needed dam rehabilitation, there is another
important reason to carry out a structural refurbishment of the dam. According to
the current regulations in Bulgaria regarding the consideration of seismic excitation
(non-existent at the time of construction) the stability of the dam is not guaranteed in
the event of strong earthquake. Consequently, rehabilitation works were necessary.
In a first phase, the reservoir was emptied in 2002, a PVC sealing geomembrane
was installed on the upstream face, and the contact area with the rock founda-
tion was grouted. In the second phase, the development of a conceptual design
for structural rehabilitation was tendered and awarded. The solution finally chosen
consisted of a massive concrete buttress placed against the downstream surface
of the existing dam. From a seismological study it was determined a peak ground
acceleration (PGA) equal to 0.305 g for a 1000 year return period, and 0.564 g for
a 10,000 year return period. The seismic verification of the current dam structure
was carried out using a 2D FEA for the dam-foundation-reservoir system, using the
DIANA computer program, with the initial assumption of linear-elastic behavior of
the dam and its foundation. The model geometry was defined for the tallest dam
monolith. In case of safety evaluation earthquake (SEE), the results show dynamic
tensile strengths that exceed the allowable limit values for the construction joints
and for the mass concrete. Then, the similar analysis of the strengthened dam was
657
GR Q. 104
carried out following the same approach and, in case of SEE, the results show both
the compressive and the tensile stresses at lower levels when compared with the
original dam model. An exception is at the heel of the dam, where some local crack-
ing may occur, but for earthquake excitation of such intensity level (SEE) this type
of damage would be acceptable since the overall stability of the dam is not endan-
gered. Thus, the efficacy of the particular strengthening solution investigated was
proven.
Report R39 (Sidik et al., Indonesia), is dealing with the design and construc-
tion of a modular precast concrete system for sidewalls, and it is addressed in
Section 4.2.
A report received from India (Chaubey et al., R43), describes the 5 m raise
carried out on a 32 m high CVC gravity dam, Dhanikhari Dam in Andaman & Nicobar
Island. The report provides details of the design of the heightening, consisting on
a concrete buttressing at the downstream side, which included performing different
FEM, namely a thermal analysis, a stress analysis and a shrinkage stress-strain
analysis. One factor that greatly influenced the design and construction was the fact
that the water supply system could not be seriously disrupted. Accordingly, as the
ideal situation of a completely empty reservoir was not possible, the concrete was
placed during the months when the reservoir level was lower, in order to reduce
the reservoir loading on the dam, during the construction period, to the minimum
possible so that the new composite structure reacts to the largest range of water
loading.
The last report received under this topic, also from India (Bansal et al., R45),
deals with the hydraulic design of a 18.5 m high reinforced concrete spillway struc-
ture, equipped with three radial gates. The project is located on the upper reaches
of a Himalayan river with a heavy boulder bed load and deep pervious strata, con-
ditions not appropriate for large concrete dams. This barrage (barrier or dam with
low crest) is the most upstream component of the 60 MW Naitwar Mori Hydroelec-
tric Project (NMHEP), scheduled for commissioning in 2022. The design faced the
hurdle that the existing guidelines provided by the Bureau of Indian Standards for
planning and design of Barrages do not apply to structures in boulder dominant
river reaches. For the final hydraulic design several potential modes of failure were
considered, including foundation failure due to scour, piping of material underneath
due to excessive seepage, exit gradient and suffusion risk, based upon different rec-
ognized theories, in the absence of any proven guidelines. The design was verified
by seepage numerical analysis using Slide software to obtain the flow net and exit
gradient diagrams. The report provides details of the level of barrage’s crest, which
has been kept at the level of the river bed, as well as upstream and downstream
protection works and grout curtain. Several small hydropower projects are being
planned on the tributaries of major Himalayan rivers, involving difficult foundation
conditions similar to those explained above and, for this reason, the experience
gained after the performance of the barrage structure of the NMHEP is of utmost
importance.
658
GR Q. 104
4.2. CVC GRAVITY DAM CONSTRUCTION
The first of the two reports received on the CVC gravity dam construction
topic, R21 (Yamashita et al., Japan), presents the concept of “dam concrete auto-
matic placing system”, whose main objectives are to improve productivity and reduce
needs of qualified personnel by applying information technology. The report states
that concrete placement in concrete gravity dams accounts for about 60% of the
construction cost and about 50% of the construction time. And it is a fact that
the operations for concrete production, transportation, and placement have been
repeated for several years and require many experienced skilled workers for han-
dling of large heavy machines and equipment. Particularly, when rail rope cable
complex 3D operation, so greater experience and skill than usual are required. The
new system integrates the automation technology that has previously been applied
to the batching plant and transfer car with automation of the cable crane. The sys-
tem was tested during placement of about 40,000 m3 out of a total of 213,500 m3 of
concrete volume using the cable crane at the Yanagawa Dam (h = 77 m). The results
obtained were: firstly, the cycle time per placement was reduced by about 10% (from
3 min 40 sec to 3 min 20 sec); secondly, the number of workers required for place-
ment in the dam was reduced by 33% (from 15 to 10); and thirdly, by automation of
the operation of cable cranes with this technology, the operation can be performed
with the same working efficiency as a highly qualified operator. For the future, the
intent is to increase the number of applications of the technology (other types of
crane, etc.) and increase the reliability and versatility of the system.
In the second report, received from Indonesia (Sidik et al., R39), the planned
implementation of a modular precast concrete system on the sidewalls for the
downstream channel of the spillway of Margatiga Dam is explained. The dam is
a composite type structure, with a combination of a concrete gravity dam and a
rockfill dam with vertical core. One of the main goals of using this system was
to speed up the construction schedule time. Additional advantages, when com-
pared to conventional methods, are: the constructability (ease of implementation,
no formwork nor scaffolding required), less need for labor, simplified quality control
(repetitive work carried out in a factory), and improvement of product quality. The
system requires precise planning because the level of flexibility of this method is low
compared with the conventional system. The transportation and installation of the
precast concrete modules must be carefully considered and could be a limiting fac-
tor for the dimensions/weight of these. For the wingwalls of this project 11 different
modules were designed, including 5 for the 3 m-high foundation slab (12 pieces in
total), and 6 for the 9 m-high trapezoidal wall (16 pieces). The prefabricated mod-
ules were designed in such a way that, once placed, they left a significant volume
of the cross section of the wall to be filled with in-situ concrete, but not requiring
formwork, and having reinforcing rebar protruding from the precast element to be
wrapped with the in-situ concrete. The weight of the heaviest module was 5.45 t, and
the work was planned to use a 50 t crawler crane to install the modules, which are
659
GR Q. 104
connected to each other with 30 mm diameter anchor bolts subsequently grouted.

In this particular project, when compared to the use of conventional methods, a
reduction of the schedule from 5 to 4 months was considered during planning, and
a reduction in labor from 100 to 10 workers, for the construction of the walls at the
site
4.3. RCC GRAVITY DAM ANALYSIS AND DESIGN
A first report received for this topic, R7 from Italy (Pietrangeli et al.),
addresses various aspects related to the Grand Ethiopian Renaissance Dam
(GERD) Project, including thermal and seismic behavioral analysis, but due to the
large size and height of this dam we have included the discussion of this report in
Section 6.3.
The next report received for this section is from Spain (G. de Membrillera
et al., R9), and includes a general overview of several numerical models used in
the design of Casupá RCC dam in Uruguay, as well as some details on the design
approach, and of the dam foundation and potential quarries. Casupá is a 31.5 m high
RCC gravity dam which will be constructed with the purpose of ensuring the water
supply to the city of Montevideo and its Metropolitan Area. The dam will have a crest
length of 750 m, with transverse joints spaced at 20 m, a vertical upstream face,
and a stepped downstream face with an equivalent slope of 0.8H:1V. The numerical
models listed in the report are:
• 2D transient thermal model that assesses the maximum temperatures reached

during the construction stage, plus the minimum and maximum expected
thermal gradients in the concrete, both for short and long-term conditions.
• 2D static mechanical-structural model of the cross-section of the dam, from
which the expected strains and stresses in the various load scenarios are
obtained, as well as safety coefficients against sliding of the dam, and safety
factors for maximum stresses in the concrete. This model includes several
sensitivity analyses considering the most relevant resistance parameters for
the stability of the dam at representative cross-sections.
• A static 3D mechanical-structural model, to verify the most critical hypotheses
that were found with the 2D model, has been deemed necessary for the cen-
tral area of the dam, where singular cross-sections are located, such as the
spillway, outlet works and the water intake.
• 2D geotechnical model developed for the dam foundation, considering the typ-
ical cross-section. With this model, the flow net and seepage in the foundation
has been estimated for several load scenarios, in conjunction with stability
calculations in the foundation.
Report R17, from France (Andrian et al.), has a part of it that deals with the
topic of this section, using several case studies, but has been entirely addressed in
Section 3.3.
660
GR Q. 104
Similarly, Report R22, from Japan (Aosaka et al.) has part of it that fits in this
section, but also can be included in Section 6.3, and we have preferred to discuss
the whole report in that section.
Likewise, report R23, authored by Shaw (South Africa), has already been
discussed in Section 3.3, due to it elaborates on a specific characteristic of the
RCC to consider in the design not only for RCC gravity dams, but also for arch and
arch-gravity dams, the stress-relaxation creep.
Finally, R40 from France (Deroo et al.), presents the concept of hybrid RCC-
Rockfill dams. The report assesses the advisability of combining both dam typologies
to build economical and safe composite dams, based on the return of the expe-
rience of similar structures, on the calculation, and on a multi-criteria evaluation
of the advantages and disadvantages. Although it was written considering struc-
tures a few tens of meters high, there is nothing to prevent the concepts used
from being applied to dams over 100 m, but it seems necessary first to have feed-
back from experiences on earlier, more modest structures. Two RCC and rockfill
hybridization options are presented, but many other options are possible, and may
be better suited to the circumstances of each site and each project. The first option
is called RCC-R, with the two parts of the dam body being erected simultaneously,
with the RCC at the upstream side constituting the impervious element, and the
downstream face of the RCC not formed. In the report, the calculations are pre-
sented with a downstream slope of the RCC at 0.4H:1V, but steeper slopes can
be considered. Zoning of the downstream backfill can be considered, to optimize
costs and mechanical behavior, particularly at the interface. The whole range of
RCCs is possible, from “high cementitious” to “hardfill”, and the geometry adapted
accordingly. A geomembrane can be considered upstream to address the issues
of cracking and seepage. The second option is called R-RCC, also with the two
parts of the dam body being erected simultaneously, but with the RCC at the
downstream side. Watertightness is provided by an upstream facing of the rock-
fill (typically in reinforced concrete, or by a geomembrane). Zoning of the backfill
is necessary, as with concrete face rockfill dams (CFRD), to provide support for
the facing, and to ensure increasing permeability from upstream to downstream.
The RCC is designed to be highly permeable, for example by providing drainage
galleries. Under certain conditions, a spillway can be fitted. The report notes that
building mixed dams is not a new idea, although in the past the building materi-
als were neither RCC nor riprap, but rather masonry and backfill. In some cases,
masonry dams, whose stability was considered precarious, have been transformed
into hybrid masonry-embankment dams, and some case histories are mentioned.
Furthermore, it is recalled that the idea of combining concrete and backfill to form
new economic and safe dams is not new: ICOLD Bulletins 83 and 117 feature sev-
eral illustrations of this concept. The last sections of the report are dedicated to
showing the advantages of the RCC-R and R-RCC hybrid dams over the CFRD
and RCC dam types in terms of economy, safety and environmental footprint and
lastly, various numerical simulations of RCC-R and R-RCC are presented and
discussed.
661
GR Q. 104
4.4. RCC GRAVITY DAM MATERIALS AND MIXES
As commented in Section 4.3, report R7 (Pietrangeli et al., Italy), deals with

various aspects of the GERD, including used RCC mixes, but due to the extremely
large size and height of this dam we have included the discussion of this report in
Section 6.3.
Report R10, received from Spain and authored by Ortega, introduces the lat-
est innovation regarding the methods for forming the faces of RCC dams, which is
termed immersion-vibrated RCC, or IV-RCC, and uses as a study case Enciso Dam
in Spain (h = 103 m), one of the largest RCC dams in Western Europe (second high-
est and third by RCC volume, 695,000 m3 ). The author debates that a number of
different techniques have been developed along the past 35 years to form the faces
of RCC dams. But it must be acknowledged that actually none of them fully met to
date the requirements for quality and/or simplicity. The IV-RCC mix, halfway between
the high cementitious RCC (HCRCC) and the traditional very stiff (very low to zero
slump) CVC for dams, has successfully solved most of the critical issues encoun-
tered in early RCC dams. The performance after first impoundment of the recently
completed Enciso IV-RCC dam in Spain has confirmed an extraordinary improve-
ment in quality compared with previously constructed RCC dams. The IV-RCC mix
is a highly workable RCC mix, with consistency zero as measured by the Slump
test, and between 6 and 12 seconds, with a target typical range of 8 to 10 seconds,
as measured by the Loaded VeBe test. For RCC mixes with higher VeBe times, the
consistency needs to be modified by adding water, grout, paste or mortar in order
to achieve the ability for immersion vibration. With IV-RCC mixes this is no longer
required and the same concrete can be consolidated by either method, immersion
vibration or external roller compaction. The report provides the key features of the IV-
RCC mixes and highlights the advantages of their use. Firstly, technical advantages,
due to the fact that for the dam faces (also contact against rock at the abutments,
contact against previously built CVC structures, around instrumentation apparatus
to leave embedded, etc.) the same concrete mix can be used, as for the adjacent
core of the dam, “both” developing similar strength and elastic and thermal proper-
ties, being the consolidation method the only difference. Additionally, the bonding
between “both” concretes is perfect, with no interface, provided that the respec-
tive consolidation of “both” concrete mixes (face and core) are correctly executed.
Secondly, the simplicity and economy of construction gained through avoiding the
additional resources and personnel for production, transportation, placement and
quality control of the CVC or grout otherwise required for consolidation by immersion
vibration. The first experiences of using IV-RCC were gained at Pirris Dam (Costa
Rica, 2009–2010), and De Hoop (2009–2013) and Spring Grove (2012–2013) dams
in South Africa. Most recently, Enciso Dam, completed in 2018, is another exam-
ple of a successful implementation of the IV-RCC technique. In the last part of the
report the characteristics of the mixes tested and used in this dam, as well as some
key details of the construction, are described. Finally, a brief description of the per-
formance of the dam during the reservoir filling is discussed, where some leakage
coming from the foundation has appeared, but the overall behavior of the IV-RCC in
terms of seepage is being flawless.
662
GR Q. 104
In their report, Dunstan et al. (R13, Great Britain), address the Trial Mix Pro-
gram conducted for an RCC dam due to start construction in 2021. The purpose
of the report, along with a full explanation of the Trial Mix Program, is to highlight
the benefits of implementing it early and with an extensive scope. The dam will
be the planned 84 m high Mwache Dam, the first RCC dam in Kenya. It is men-
tioned that an earlier study of the dam (by a firm of consulting engineers with no
RCC experience), concluded that there were no locally-available pozzolans in Kenya
and thus Mwache should be a traditional concrete gravity dam. The main objective
of the Trial Mix Program for the RCC dam was to provide sets of mixture propor-
tions for the various combinations of locally-available cementitious materials for the
Specification of the Bidding Documents for the construction of the Mwache Multi-
Purpose Project. The report describes the potential sources that were identified for
the various materials: aggregates (from upstream and downstream of the dam site),
Portland cement (two types, both from Kenya), low-lime flyash (two types, both from
India), natural pozzolans (one type from Kenya and one from Tanzania), GGBFS
(from India), and retarders (an extensive list of them from BASF, Fosroc, Sika and
Chryso). The next stage was to conduct a Preliminary Trial Mix Program, which
included the optimization of the gradation of the aggregates, the optimization of the
workability and water content of the RCC mix, and the optimization of the retarder
dosage (after defining the requirements for the Setting Times). Then, the Main Trial
Mix Program was undertaken, following the steps below: choice of the mixture pro-
portions (different content ranges of cementitious materials depending on the type
of SCM used; analysis of the development of strength with age of the RCCs con-
taining flyash, containing natural pozzolan, and of those containing GGBFS (finally
discarded) and, lastly, definition of the optimum mixture proportions and retarders
for the different combinations of the cementitious materials. One of the main findings
emphasized in the report is that specific retarders had to be designed in order to
obtain the specified setting times for the RCCs containing the available particular
cementitious materials. This research took time that would not have been available
if the Trial Mix Program had not been undertaken so early. Also, it was found that
some combinations of the potential cementitious materials were extremely efficient.
An RCC mix containing 32.5 kg/m3 of Portland cement and a low-lime flyash con-
tent of 187 5 kg/m3 and a further RCC mix containing a Portland cement content
of 50 kg/m3 and a natural pozzolan content of 170 kg/m3 , both conformed to all
the requirements of the RCC for Mwache Dam. The early and extensive Trial Mix
Program allowed ranges of mixture proportions and specific retarders to be included
in the bidding documents enabling contractors to bid on a reasonably-certain basis.
Another report for this topic has been received from Morocco (Boukaidi
Laghzaoui et al., R24), and narrates the details of the different combinations of
sands that were subsequently used in the RCC mix for Toudgha Dam (h = 67.5 m),
as well as other characteristics of the employed RCC mixes. Up to 5 different sand
types were utilized, which were combined in different proportions throughout the
construction of the dam: 0/4 mm sand obtained after processing at the on-site
crushing plant the dolomitic limestone rock mined from a quarry located about 1 km
upstream of the dam site (unwashed); the same sand but after washing; a high fines
content (65%, non-plastic) 0/4 mm sand obtained by crushing limestone from a
quarry located 20 km from the dam site; a 0/4 mm sand obtained after reprocessing
663
GR Q. 104
in the on-site crushing plant the significant surplus obtained in the production of
the 2/4 mm sandy fraction; and 0/2 mm sand from a second external source. The
last part of the report is dedicated to some specifics of the construction, shortly
commented in Section 4.5.
Report R30 from China (Junjie et al.) is mostly focused on the RCC interlayer
bonding, and it is briefly described in the next Section 4.5. Also, a part of the report
provides in depth information regarding the individual materials used in the RCC mix,
the mixture proportions, and different properties, including mechanic, deformation
and thermal properties of the fresh and hardened RCC used for Kafue Gorge Lower
Hydropower Project in Zambia.
4.5. RCC GRAVITY DAM CONSTRUCTION
The first report received that fits this topic, R14 from Great Britain (Dunstan),
begins with an overview of the nearly one thousand RCC dams either completed
or under construction at the end of 2019, and accentuates the remarkable growth
over the last 35 years of this new form of dam construction. A basic, but at the
same time fundamental idea, is that the original concept for RCC dams was to keep
the design and method of construction simple, because with simplicity would come
rapid construction, which implies quality and economy. It is mentioned that unfor-
tunately, over the past few decades, some modern RCC dams have become too
complicated and the advantages of the speed of construction has in some cases
been lost. Thus, it is strongly recommended that RCC dams revert to the original
concept of simplicity and rapid construction. The report reminds that on the web-
site “www.rccdams.com” there is available a powerful database with details of the
950+ RCC dams completed or under construction worldwide. Statistics are provided
on the number of RCC dams completed per year from the early 1980s until 2019
with indicated locations of RCC dams around the world. The report also discusses
placement rates of RCC in dams, and concludes that a significant number of all
RCC dams (probably more than 95%) are not being designed and/or constructed
as efficiently as they could (or should) be. On the positive side, the report presents
the data for the thirteen RCC and hardfill dams with the highest average rates of
placement. The next part of the report is dedicated to provide interesting details of
the five dams in the previous group that were finished well ahead of schedule, in
order to try to understand why the dams were successful; the dams are: Longtan
(Stage 1) (China), Upper Stillwater (USA), Olivenhain (USA), Beni Haroun (Algeria)
and Nam Ngiep 1 (Laos). And lastly, an overview of Lower Ghatghar (India), Son
La and Lai Chau (both in Vietnam) is provided as three very good examples of the
great benefits achieved by an early completion of hydropower projects, thanks to
the speed of construction of the associated RCC dams.
Report R17, from France (Andrian et al.), has a part of it that deals with the
topic of this section, using several case studies, but has been entirely addressed in
Section 3.3.
664
GR Q. 104
Boukaidi Laghzaoui et al., in their report R24 (Morocco), besides discussing

particulars of the sand types for the RCC mixes used at Toudgha Dam, mentioned
in Section 4.4, describe also a key aspect of the construction at the right bank. As
the slope of the right abutment was very gentle (at 7.9%), this condition made it very
difficult the placement of the RCC in horizontal layers. Thus, tests of RCC placement
of inclined layers, with the same slope of the excavation bottom, were conducted
with successful results. The procedure was adopted throughout the right bank until
the elevation of the upstream gallery roof is reached; from this elevation upwards,
the RCC layers are placed horizontally.
Four separate reports have been received from China, all related to the
same RCC dam which is one of the main components of the Kafue Gorge Lower
Hydropower Project in Zambia. The dam is 130 m-high and the volume of RCC is
1.3 Mm3 . The first report, R27 (author/s not indicated), addresses the problems the
Chinese contractor experienced in terms of the low skills level of the local workforce
needed to build the project, where only for the RCC dam construction 50 Chinese
and 932 Zambian employees were required at peak. The report explains some of
the measures taken to alleviate the problem. The second report, R28 (author/s not
indicated), discusses the Rotec high-speed belt-conveyor equipment used for the
transportation of the RCC from the concrete batching plant to the dam. Total length
of belt-conveyors was 1559 m, with a maximum slope of 27◦ , belt speed 3.6 m/s,
and maximum conveying capacity of 500 m3 /h. The report summarizes some prob-
lems that had to be overcome related with the implementation and operation of the
conveyor system. The third report, R29 (Yuanguang et al.), highlights some aspects
of the construction, as the concrete pre-cooling systems, the Rotec high-speed belt-
conveyor transportation equipment, the only RCC mix used with MSA 63 mm (after
discarding the upstream RCC facing area with MSA 37.5 mm), the use of bedding
grout in between the RCC layers, the use of GEVR at dam faces, and the use of fly-
ash, rock powder and high efficient Naphthalene water reducing agent in the RCC
mix. Finally, the fourth report, R30 (Junjie et al.), focuses on the RCC interlayer
bonding and describes some research conducted, including: optimizing the RCC
mix; study of characteristics of joints under different exposure time and treatment;
establishing the Modified Maturity Factor; and definition of the quality control system.
4.6. CMD ANALYSIS, DESIGN, MATERIALS AND CONSTRUCTION
The only report that has been received on the subject related to Cemented
Material Dams, is R8 from Peru, authored by Mendívil et al. The report discusses the
purpose, analysis, design, materials, and construction for the study case of Asana
River Diversion Dam in southern Peru, completed in 2018. The dam is one of the
components of the infrastructure needed for a large copper mine, and it was built
immediately upstream of the open pit to divert the Asana River and its small tribu-
taries through a tunnel that allows the river bypass the mine site, protecting mining
operations against flooding, and maintaining the natural quality of the water for the
communities located downstream. The site involves high seismicity (0.36 PGA) and
665
GR Q. 104
the geological and geotechnical conditions are very complex. Consequently, since
the dam location was determined primarily by the layout of the mine, and not based
on optimal foundation conditions, the adopted solution was a symmetrical hardfill
dam with a maximum height of 50 m. The design was driven by the highly variable
foundation conditions, the use of plastic colluvial materials available near to the site,
and the ability to construct the spillway over the dam body. To assist design, 2D
and 3D FEM analysis were performed to evaluate the level of stresses and stability
safety factors of the dam at critical foundation conditions. For all foundation and
loading conditions, all performed stability analyses showed adequate sliding stabil-
ity, a dam and a dam-foundation contact in compression, and transfer of low and
uniform stresses on the foundation. Additionally, a 3D FEM analysis was performed
in order to estimate total dam deformations and differential settlements between
monoliths. The report gives a detailed description of the foundation conditions and
various singularities encountered (including an important subvertical fault 8 to 12 m
wide adjacent to the river channel), and mitigation measures taken. Also, the con-
struction of the dam body with hardfill is discussed (50 mm MSA, 120 kg/m3 cement,
VeBe = 35 sec, 5 MPa at 90 days compressive strength), by placing and compacting
horizontal 30 cm thick layers, and the stepped upstream face protection with 40 cm
of conventional unreinforced concrete for durability and water tightness purposes.
A bedding mix (19 mm MSA) was placed at the lift joints over a 2.0 m wide strip
next to the upstream face to improve water tightness between hardfill layers. One
problem to solve was the high plasticity of the fines of the colluvium material used for
the hardfill. Another important characteristic of the selected hardfill mix was the low
modulus of elasticity in the order of 7 GPa, ideal for the variable conditions encoun-
tered in the foundation, as it allows for a higher deformation of the structure without
cracking. Since the completion of the dam, and after a partial filling of the reser-
voir during intense storm events in February 2019, periodic visual inspections and
instrumentation readings have not revealed any major issues. Asana Dam entails
a good example of the selection of the symmetrical hardfill technology for specific
conditions, such as those on this dam site and its surroundings.
5. INNOVATIONS FOR RAISING EXISTING CONCRETE DAMS
There are several possible distinct reasons, or a combination of them, to

decide the heightening of an existing concrete dam. Some of these reasons may be:
increase the reservoir volume (due to increase in demand and/or loss of capacity
because of sedimentation); increase flood control and/or allow a higher design flood;
correction of design flaws with respect to stability (under-design, that can be as a
consequence of more stringent regulations with respect to loading hypotheses) by
adding weight which facilitates the raise of the dam as a consequence and, similarly,
when precarious condition of the dam by poor construction (old masonry dams in
particular) or concrete ageing (due to AAR, ISA, free-thaw, etc.), among other.
ICOLD Bulletin 64 (ICOLD, 1988) on dam heightenings is the latest ICOLD

publication dedicated to raising of dams. An update to this publication, to introduce
666
GR Q. 104
concrete dam projects that have been raised in recent decades all over the world,
has been already assumed by the ICOLD Committee “D” on Concrete Dams. A white
paper on the subject, that it is currently under preparation by the USSD Committee
on Concrete Dams, will be an initial contribution to the new ICOLD bulletin.
As for Q. 104, only three reports have been received describing experiences
of raising existing concrete dams, and two of them have one case in common.
In all cases, the described heightening has been with CVC, and are presented in
Section 5.1. In Section 5.2, however, we will offer a general discussion on the raising
of existing concrete dams with RCC.
In some cases, after analyzing different options, it has been decided to build
a new dam, generally downstream, but in some cases upstream of the existing
dam. The La Breña II RCC gravity dam, and the Puentes IV CVC gravity dam,
both in southern Spain, are respectively good examples of both alternatives for the
“heightening” of an existing dam. Another assimilable example is the one described
in report R33 concerning Spitallamm dam in Switzerland, which is addressed in
Sections 3.1 and 7.1. The primary reason of building a new double-curvature arch
dam was the precarious condition of the existing dam. The new dam was built
immediately downstream of the existing one, with a design that entails the possibility
of a future heightening of 22 m.
5.1. RAISING EXISTING CONCRETE DAMS WITH CVC
Report R34, from Switzerland (Leite Ribeiro et al.), explains the details of
the design and construction of the heightening of two arch dams, Cambambe in
Angola and Vieux Emosson in the Swiss Alps. The aspects related with the con-
struction of both heightening have been explained in Section 3.2. The Cambambe
arch dam, 72 m high, was built between 1959 and 1963. From its conception, the
dam was designed to be raised. The raising of the dam took place from 2012 to
2018, reaching a new height of 92 m (20 m raise). The layout of the heightened
Cambambe dam follows the same shape as per the existing dam, of a three-center
arch dam geometry. On the other side, the original Vieux Emosson dam was a 55 m
high, single-curvature arch-gravity dam, built between 1952 and 1956, and raised
between 2012 and 2015 to a new height of 76.5 m (21.5 m raise). The original dam
was not conceived to be raised. With the dam raising, the original single-curvature
arch-gravity dam is converted into a double-curvature arch dam (parabolic shape).
For geometric constraints at the contact between the original dam and the raise, it
was found compulsory to demolish the upper part of the original dam by 10 m and
additionally 10 more meters only at the downstream side, thus forming a 10 m high
step in the heightened dam section. This solution was believed the most economical,
when considering the reduced contact surface and the reasonable concrete volume,
compared to other raising alternatives that maintained the arch-gravity configuration.
The report makes an interesting comparison of both raises, which presented some
similarities in certain aspects, and important differences in other. Particularly the very
different site conditions (original dam to be raised, its geometry, shape and existing
667
GR Q. 104
concrete and foundation conditions), weather conditions (impacting the concrete mix
design, the pre- and post-cooling and construction schedule) and hydrological con-
ditions (affecting the water management in the reservoir, the raised arch dam joint
grouting operation and the first impounding of the reservoir) which made essential
adapting the dam raising design and works to the specific, local conditions. As a
conclusion, the report highlights that on one hand arch dams are structures which
can readily and economically be raised, but on the other hand are also the most
challenging dam type in terms of structural behavior, design and construction tech-
nologies to be implemented. Consequently, raising such slender structures should
be carefully conceived, designed, planned and managed.
A second report from Switzerland (Droz & Wohnlich, R36), also includes the
heightening of Vieux Emosson as one of several case histories referenced in that
report, but to a lesser degree of detail. The report is discussed in Sections 7.1
and 7.2.
The third and last report on this topic, R43 from India (Chaubey et al.) describ-
ing the raise carried out on the Dhanikhari CVC gravity dam, has been addressed
in Section 4.1.
5.2. RAISING EXISTING CONCRETE DAMS WITH RCC
As previously commented, no reports have been received on the raise of

existing concrete dams with RCC. Although there have not been many realiza-
tions of this typology, some of them are noteworthy, and at this moment a handful
of them remarkable as well are under design, to our knowledge, particularly in
the US.
Taking advantage of the information already collected by the USSD Subcom-

mittee, which is preparing the white paper on raising concrete dams mentioned
above, Table 7 lists selected projects in the USA of heightenings with both, RCC
and CVC, post ICOLD Bulletin 64.
Table 7
Selected concrete dams raised in the USA (USSD white paper under drafting)
DAM HEIGHT RAISE RAISE YEAR
DAM STATE TYPE (M) (M) TYPE COMP.
Gross Colorado Curved gravity 104 39.9 RCC U.D.
San Vicente California Gravity 67 35.7 RCC 2013
Blue Lake Alaska Single arch 60 26.5 CVC 2014
Theodore Roosevelt Arizona Masonry arch 85 23.5 CVC 1996
Halligan Colorado Arch-gravity 27 8.3 CVC or RCC U.D.
Buffalo Bill Wyoming Arch 107 7.6 Reinf. CVC 1994
Bartlett Arizona Multiple arch and 88 6.6 Reinf. CVC 1996
buttress
Shasta California Curved gravity 184 5.6 CVC or RCC U.D.
Note: U.D. = Under Design
668
GR Q. 104
Confidently, this information will be completed in the near future with other
projects around the world as the planned ICOLD bulletin is prepared.
6. INNOVATIONS FOR EXTREMELY HIGH CONCRETE DAMS
It was indicated at the beginning of Section 3 that most of the extremely-high

dams constructed in the world in recent years have been CVC arch dams. But only
two reports on one of them, in fact the highest dam in the world, Jinping I in China,
have been received.
While there is no precise definition for the “extremely high dam” term for the pur-
poses of this report we are considering extremely high CVC dams those exceeding
200 m in height.
Nevertheless, in order to allow more reports to be included in this section, we

are adopting a subjective criterion of considering as extremely high RCC dams those
with a height over 150 m. To this date, nineteen gravity RCC dams have been built
exceeding that height (Dunstan, 2020), from which seven of them are discussed in
the received reports. As for RCC arch dams, only two fit into this category, and two
reports have been received referring to one of the two, Janneh Dam in Lebanon.
Table 6 in the introduction to Section 4 lists the ten highest RCC dams in the world.
6.1. EXTREMELY HIGH CVC ARCH DAMS
Only two reports have been received for the topic of innovations for extremely
high CVC arch dams, and both refer to the same dam, currently the highest dam in
the World, not for much longer, Jinping I (h = 305 m).
The first report, received from China, Zhong et al. (R31), is dedicated, on the
basis of monitoring data and analysis using 3D nonlinear elasto-plastic and visco-
elasto-plastic rheological models, to the prediction and evaluation of the long-term
deformation of the left abutment slope, and its effect on the behavior of the double-
arch dam structure. The overall stability analysis methodology of arch dams and their
foundation is also presented, providing reference for the professionals in the future.
The left bank, with a 50◦ –65◦ steep slope and an extremely complicated founda-
tion, mainly composed of marble and partially of sandstone, has several faults and
lamprophyre dikes, resulting in potential slope failure modes of shear slip, block slip
and wedge shape damage, etc. The stability of the 540 m high excavation slope on
the left bank was ensured by extensive reinforcements, including rock anchors, rock
cables and anti-slide concrete replacement tunnels. The left bank slope was divided
into 6 different deformation zones, and the following was monitored: slope shallow
deformation, slope deep deformation, and valley deformation. According to the data
669
GR Q. 104
collected, four of the six zones are in steady state, while in one zone the deformation
has not yet converged, and in other zone is still in the adjustment period. Since the
abutment slope and downstream abutment slope have been in stable state, it can be
concluded that the long-term safety of the dam will be maintained, and the left bank
slope, as a whole, is in a stable state. In order to study the effect of left abutment
slope deformations on the safety of the arch dam, the following procedures have
been considered: 1) Analysis of geological structures and their rheological parame-
ters; 2) Long-term deformation prediction of the slope; and 3) Analysis of the effects
of the long-term deformation on the arch dam, with the utilization, as fundamental
tools of 3D elasto-plastic and visco-elasto-plastic FE models, as mentioned above.
The second report, also from China, Lijun et al. (R32), is referring to Jin-
ping I Dam as well, in respect to the foundation treatments. It is highlighted that
Jinping I has the most complicated geological conditions among the extremely-high
arch dams, including 10 to 13 times in difference between the deformation modulus
of the two banks, which made the dam design very challenging. The report discusses
that according to the extremely complex geological conditions of the dam site, the
effect of different geological defects on the deformation and stability of the arch dam
have been determined by numerical analyses, as well as systematic treatments
been evaluated and determined to make sure that the arch dam foundation meet
the design requirements for bearing capacity, deformation stability and impervious-
ness. Unfortunately, apart from mentioning that it was performed a 3D nonlinear
numerical analysis and geomechanical model tests, no further specific information
is provided on the numerical analysis results of the dam foundation treatments. Only,
that the results show that, after the foundation treatments, the integrity and unifor-
mity of the rock mass at the left abutment are improved, as well as an increase in
the deformation resistance of the dam foundation and the resistance at the con-
tact dam bodyrock mass. The authors conclude that the dam’s foundation system
can be considered overall safe. The foundation treatment measures included in the
numerical analysis are: concrete cushion, gravity piers, concrete replacement grid,
consolidation grouting, and curtain grouting. In particular, the analysis shows that:
• The concrete cushion can greatly improve the stiffness of the foundation
between certain elevations, with a reduction of up to 30% in displacements
along the river.
• The abutment consolidation grouting can reduce the cross-river displacement
by 13.3% to 21.3% between certain elevations.
According to the results of the 3D numerical analysis and geomechanical

model tests, a comprehensive foundation treatments project was put forward and
carried out, including a 155 m high concrete cushion with a volume of 560,000 m3 ,
5 levels of grouting galleries, 720,000 m of drilling for consolidation grouting, 3 lev-
els of anti-shear galleries, and systematic concrete replacements of fault zone and
lamprophyre dikes, with volume of 146,000 m3 . Regarding the dam performance,
the reservoir level has exceeded the normal water level (NWL) 7 times since Octo-
ber 2013. The seepage is very small, and the displacement of the double-curvature
dam, linked with the water level, is basically elastic.
670
GR Q. 104
6.2. EXTREMELY HIGH RCC ARCH DAMS
The only two reports that partially fit in the topic of high RCC arch dams are
R16 (Si-Chaib et al.), and R17 (Andrian et al.), both received from France, and
both having one part that specifically refers to the Janneh arch-gravity RCC dam in
Lebanon, which is why both have already been covered in Section 3.3. Please refer
to that section.
6.3. EXTREMELY HIGH RCC GRAVITY DAMS
Report R4 received from Italy, authored by Miranda et al., deals with geomem-
brane systems in general, to provide or restore watertightness in dams, and in
particular its application for RCC dams, and more specifically for high RCC dams,
using as a case history the Miel I RCC dam in Colombia (h = 188 m). It is explained
that geomembranes have been used in dams for more than 60 years (since 1959).
Initially they were only applied to embankment dams less than 30 m in height, later
to the rehabilitation of all types of dams and to new RCC dams, without limitation of
the height of the dam and, subsequently, to very high dams of all types, either for
new construction or for rehabilitation, in the dry or underwater. The foregoing coin-
cides with the recommendations of the three successive bulletins that ICOLD has
dedicated to the use of geomembranes in dams: Bulletins 38 (1981), 78 (1991) and
135 (2010). Focusing on La Miel I RCC dam, this project set a milestone in water-
proofing of new RCC dams with exposed geomembranes. For very high dams,
i.e. for conditions of very high hydrostatic pressures at the dam face, key essen-
tials of the design to grant full watertightness are the selection of a proper type of
geomembrane liner, a fastening system capable to resist the applied loads at full
and empty reservoir, a watertight peripheral sealing system connected to the grout
curtain at the foundation, and an efficient drainage system to avoid water infiltration
in the dam body and to allow monitoring of the leakage by separate sectors of the
dam. Additionally, to foresee repair methods to remediate accidental damages even
underwater. The report provides the details of the installed waterproofing system,
using the 3.00 mm-thick SIBELON® CNT 4400 for the lower 62 m of the dam, while
for the rest of the dam the 2.50 mm-thick SIBELON® CNT 3750 was used, both of
them laminated to a 500 g/m2 nonwoven polypropylene geotextile, as well as of the
fastening system, and the description of the installation work by separate horizontal
sections. The good surface finish of the upstream face of the dam, thanks to the
use of GERCC, greatly facilitated the installation of the geocomposite. A need for
stringent Construction Quality Assurance procedures, implemented by a specialty
waterproofing contractor, it is highlighted in the report. An implemented continuous
monitoring program of the dam and geocomposite system performance indicates
stable readings meeting the design criteria. Regular inspections carried out over 18
years of the hydropower plant operation evidenced little damage to the membrane,
and practically no repairs nor maintenance has been required. The experience at
the Miel I Dam confirms the long lasting excellent performance of geomembranes
at high dams.
671
GR Q. 104
A second report received from Italy for this topic, R7 (Pietrangeli et al.),
addresses various specifics of the Grand Ethiopian Renaissance Dam (GERD)
Project, for which the filling of the reservoir began in July 2020. The Main Dam
is a 175 m-high “mega” RCC gravity dam with an RCC volume of 10.2 Mm3 , which
makes it the largest RCC dam by volume in the world [N.B. In the register of RCC
dams (Dunstan, 2020) GERD is listed as 160 m in height, and for this reason it does
not appear in Table 6]. The crest length is nearly 2 km and the dam is composed of 85
monoliths separated by contraction joints at 18 to 27 m spacing. The report explains
the extensive campaign of mix designs and testing that were carried out, showing the
nonlinear stress-strain behavior of RCC mixes and anisotropy of parent (unjointed)
concrete tensile strength value (horizontal vs vertical). These results were used by
the designer to properly characterize the thermal tensile strain capacity of the mixes
and to optimize the RCC zoning in the dam body by the verification of mixes with
lower cementitious content and less heat of hydration for areas of high stress. The
dam body was zoned in five areas in which five different RCC mixes were applied.
The cement used is a pozzolanic cement. Flyash was not used. Enough content of
fine sand in the mix is obtained by using a good mixture of sound and weathered
granite. Cement content in the five RCC mixes vary from 80 to 142 kg/m3 , with guar-
anteed compressive strength values ranging between 8 and 15 MPa. A retarding
admixture from Mapei was incorporated to the mixes. The mixes with higher cement
content are used in the upstream part of the dam, to meet the tensile strength under
extreme seismic load and permeability requirements, and at the downstream toe,
for compressive strength requirements. Mixes with low cement content are used in
the central zone of the dam in order to control the temperature rise and the conse-
quent risk of cracking. A systematic bedding mix is prescribed only in the upstream
portion of the dam, in order to meet the impermeability requirements at lift joints,
and for the first 3 m above the foundation. GERCC is used at the upstream face
and at the downstream face of the non-overflow section. At the downstream over-
flow section, the steps are protected by conventional concrete with silica fume. The
testing campaign included about 500 cores from different zones of the dam body,
which were tested on site, and some cores were also sent to the Mapei labora-
tory in Milan, Italy, for further verification and comparison. The results obtained
from the extensive testing were used to perform a transient thermal analysis by
a finite differences software developed by Studio Pietrangeli, in order to evaluate
the temperature distribution histories in the dam and to establish the necessities
of pre-cooling (i.e. maximum allowable placing temperature of RCC). During the
progress of dam construction, continuous monitoring of the RCC temperatures was
undertaken, and the measurements are periodically compared with the maximum
allowable temperatures deduced from thermal studies. Additionally, an advanced
early warning temperature monitoring procedure was implemented, based on the
tensile strain capacity, to prevent the risk of cracks which could develop in the dam
body. Regarding dynamic stability analyses, at Level One Design FEM analyses
were performed to define a more reliable and complete stress distribution within the
dam body, to optimize the shape of the dam and to outline the RCC and bedding
mix zoning. The behavior of the dam under seismic loads was evaluated by means
of Response Spectrum Analysis for the calculation of maximum expected tensile
and compressive stresses under earthquake, and by means of Equivalent Lateral
672
GR Q. 104
Force method for the purposes of sliding stability analysis. The dam behavior was
assumed to be essentially two-dimensional, given that the height of the dam is much
smaller than the length of the crest. At the Level Two Design stage, further and more
accurate analyses were performed to confirm the final dimensions of the dam and
to study local effects. In order to ponder the possibility of cracks propagating from
the upstream face of the dam, a nonlinear static analysis with gap friction elements
was performed using Straus 7 finite element code.
Report R14 from Great Britain (Dunstan) provides details, mainly related to the
speed of construction, of some of the highest RCC dams built worldwide, including
Gibe III (Ethiopia, h = 246 m), Guangzhao (China, h = 201 m), Longtan (Stage 1)
(China, h = 192 m), Nam Ngiep 1 (Laos, h = 167 m), and Guanyinyan (China, h =
159 m). This report has already been discussed in Section 4.5.
A report received from Japan (R22, Aosaka et al.) deals with a cracking prob-
lem experienced in the 167 m-high Nam Ngiep 1 RCC gravity dam in Laos. The main
components of the RCC mixes used were sandstone aggregate, Type I cement
(ASTM C150) and Class F flyash (ASTM C618). Despite the fact that a thermal
stress analysis was incorporated into the dam design and construction plan in order
to control the thermal stress of the concrete, while the RCC placing on the left bank
side was suspended from October 2017 to January 2018 in order to install the hori-
zontal part of the penstocks, cracks parallel to the dam axis appeared on the surface
of the RCC on the left bank side, and vertical cracks appeared on the exposed sur-
face of the contraction joint CJ12 which was temporarily separating both halves
of the dam at different elevations. Although water curing was applied to the RCC
and CJ12 surfaces, the cracks appeared due to a sudden larger-than-usual drop
in ambient temperature in mid-December 2017 which caused temperatures to fall
below 15◦ C. In order to assess the cause of the cracks, a thermal stress analysis was
conducted using the available real data from monitored ambient and concrete tem-
peratures and from concrete testing. The thermal stress analysis was performed
sequentially according to the RCC placing and initial impounding timeline, using
ASTEA MACS software. The results of the analysis showed, for the horizontal sur-
face where the cracks appeared, maximum tensile stresses of up to 2.5 MPa in the
middle of December 2017, which is beyond the actual 1.5 MPa tensile strength of the
concrete. It was investigated that some other RCC dams in Southeast Asia region
experienced similar cracks during the suspension of the RCC placement for several
months, and probably for the same reason, a significant difference in temperature
between the interior and the surface of the concrete due to the difference in their
cooling rates, and the consequential thermal stresses caused by internal restraints.
The main remediation measure taken was, in most cases, to install a reinforcing
steel mesh over the cracks. A dam stability study was performed by hearing from
the Project’s Dam Safety Review Panel and other dam experts, and observing data
from monitoring for displacement, leakage from the dam body, uplift and other. As
the vertical cracks are parallel to the dam axis, they will not directly lead to water
leakage through the dam, and after impounding, the body will integrate under com-
pressive stresses even where there are vertical cracks, and stress in the dam body
will be transmitted accordingly, which means the dam is stable. The last part of the
673
GR Q. 104
report includes some recommended remediation works for cracks on the horizontal
surface of the RCC and cracks on the vertical surface of contraction joints. Addi-
tionally, it might be preferable to cover the RCC surface with an insulating sheet, in
addition to wet curing by water spraying and ponding, when exposing the concrete
surface for a long time. An important lesson learned is that attention should be paid
to the ambient temperature input into the thermal stress analysis, when a certain
concrete surface is going to be exposed to the ambient temperature for a long time.
For example, applying the ambient temperature at every 12 hours rather than the
mean monthly temperature that is generally used, which will allow to obtain more
realistic concrete temperature and strain, and to improve the accuracy of analyses.
7. OPERATIONAL PERFORMANCE OF CONCRETE DAMS DURING THE

LIFE CYCLE, INCLUDING UNDER EXTREME CONDITIONS
A total of eleven reports, entirely or partially dedicated to the discussion on

the operational performance of a single dam or several dams, have been received.
They describe a varied casuistry of issues experienced by the different concrete
dams, including problems related with the foundation, different concrete pathologies
(AAR, ISA, free-thaw, etc.), scouring downstream of the dam, ineffectiveness of the
impervious and/or drainage curtains and ageing of the monitoring system, among
other, and also the remediation and prevention measures implemented in each case.
One fundamental conclusion extracted from several of the reports is the impor-
tance of the design, implementation and follow-up of the monitoring and surveillance,
for the case of new dams, for existing dams that are not properly monitored and
surveilled, and when major modifications are performed to the dam structure, such
as buttressing or heightening.
7.1. PERFORMANCE OF CVC ARCH DAMS
The first report dealing with this topic of CVC arch dams performance is report
R18 from France (Grimal et al.). The report is also dealing with numerical modelling
of expansive concrete, and rehabilitation works carried out at Bimont Dam, therefore
it has been fully addressed in Sections 3.1 and 3.2.
The second report on this topic is R31 from China (Zhong et al.), and similarly
to the preceding one, it has been discussed in some detail in a previous section, in
this case in Section 6.1.
In their report, partially presented in Section 3.1, Ballisat et al. (R33, Switzer-
land) describe the replacement of an older dam structure by building a new one
immediately downstream, while continuing unrestricted operation of the reservoir.
674
GR Q. 104
In Section 3.1 the design of the new double-curvature arch dam was discussed,
while in this section we will focus on the problematic performance of the existing
dam along its service life. The existing Spitallamm Dam (h = 114 m) was erected
from 1926 to 1932 and it was one of the highest concrete dams in Europe at that
time. Two layouts were originally considered, a straight gravity dam and an arch-
gravity dam with downstream slope of 0.5H:1V and upstream slope 0.1H:1V. Finally,
the latter was selected, taking advantage of the excellent foundation conditions at
the narrow gorge on high strength and low deformability granitic and granodioritic
rock. During its service life, the dam experienced serious ageing problems, including
damages at both faces, due to freeze-thaw, and leaking spots. A grouting campaign
unfortunately provoked the formation, about 12 m below the dam crest, of a nearly
horizontal crack that matched a former cold joint performed during the construc-
tion and led to a need of additional waterproofing measures at the upstream dam
face. Additional monitoring instruments were installed, including inclinometers which
readings were complemented by measures with sliding micrometers, to document
vertical displacements, and showing that the upper dam arch was moving indepen-
dently from the rest of the structure. Although surveying of the dam crest confirmed
that a swelling process was taking place, testing of dam concrete samples did not
reveal significant signs of concrete expansion due to AAR. The possibility of tec-
tonic displacements at the dam abutments, affecting the slender upper part of the
dam, was also discarded by means of cross-valley geodetic measurements. The
report doesn’t reveal clearly what was the cause of the continuous displacement
towards upstream of the dam crest. As the dam was slowly deteriorating with time,
the dam owner took the decision to build an entirely new structure downstream of
the existing dam, and have it designed to allow a subsequent raise of 22 m in the
future.
Another report from Switzerland, Droz & Wohnlich (R36), which is focused on
the safety of dams along the operation lifetime, includes several case histories for
different dam types. With respect to CVC arch dams, the issues experienced at Les
Toules Dam (h = 86 m) in the Swiss Alps, a double-curvature arch dam with a very
slender shape, high vertical curvature towards downstream, and no shear keys, is
discussed. The origin of the problems was the foundation, consisting in alternate
strips of sub-vertical rock formations parallel to the valley, with variable properties
and stiffness, especially on the left bank, where small irreversible displacements
towards downstream were observed since the first reservoir impounding. A stability
analysis was performed and as a result, remedial works were designed and under-
taken between 2008 and 2011. The rehabilitation works consisted in buttressing
from the downstream side, with additional concrete mass, the blocks located at
both abutments. A second case history is the Kariba double-curvature arch dam
(h = 128 m), located on the border of Zambia and Zimbabwe, in which the operation
of the surface spillway provoked a major scouring at the foot of the dam, digging a
90 m deep hole some tens of meters downstream from the toe of the dam. Rehabil-
itation works, consisting in reshaping the plunge pool, were implemented after the
results of laboratory scale model tests, with the aim of optimizing the shape of the
pool in order to avoid further degradation. Additionally, Kariba Dam is affected by
AAR, and rehabilitation works to overcome this problem were arranged.
675
GR Q. 104
The final report received on performance of CVC arch dams is R42 from Nor-
way (Bjønness et al.), in which the data from the instrumentation of an unnamed
CVC arch dam is used as a basis for load definition and for calibration of a FE model.
This subject-matter has been discussed in Section 3.1. With regards to performance,
there is limited information in the paper, explaining only that leakages were observed
towards the foundation of the arch during the first year of operation. The leakage
was sealed with grouting the following year, but some similar leakage was observed
about seventy years later. It was sealed again a few years ago with new grouting
underneath the foundation. Simultaneously, the instrumentation activities and struc-
tural analyses of the dam were initiated in order to improve the documentation of the
dam. The report presents an interesting case of how the documented performance
of a dam, gathered from instrumentation, can be used to obtain a better knowledge
of the structural behavior.
7.2. PERFORMANCE OF CVC GRAVITY AND BUTTRESS DAMS
Report R3, from Italy (Fornari et al.), addresses the overall rehabilitation works
and performance of a waterproofing system that was applied to a CVC gravity dam
at the upstream face, to stop water infiltration through the dam body. The case his-
tory explained is Lago Nero (h = 38 m), a dam built in the years 1924 to 1929 in
the Italian Alps, at high altitude (at elev. 2025 m), in a harsh climate with frequent
free/thaw cycles, temperatures from −24 to +26 ◦ C, thick ice and snow, and the
action of soft alpine waters, conditions that are detrimental for concrete durability.
Repeated traditional repairs were carried out over the years from 1953 to 1962,
to improve the watertightness of both, the foundation and the dam body, including
grouting in the bedrock and complete refacing of the dam with reinforced gunite,
which appeared not to be effective in the long term. A limitation of the water level in
the reservoir was imposed due to non-fulfilment of the stability checks for the dam.
Then, the owner decided to undertake overall rehabilitation works, which lasted from
1979 to 1981 and were executed in three summer campaigns. Modifications to the
dam included the adoption of a watertight CARPI liner system as new upstream
water barrier, which was quite innovative technology in the late 1970s, connected
to a grout curtain injected from the excavation of a new plinth. The report provides
the details of the installed system, using the 1.93 mm-thick SIBELON® CNT 2800
geomembrane heat-bonded during fabrication to a 200 g/m2 anti-puncture polyester
geotextile, to form what is called a geocomposite, the fastening and drainage sys-
tems, as well as the description of the placement work. Since its installation, the
exposed geomembrane system has being monitored for seepage performance and
for geomembrane’s ageing. In 2020, after 40 years in service, the system is per-
forming well, and a long remaining service life (28 years from now) is predicted
for an almost 100 years old dam at this high altitude. Based on this precedent the
owner of the dam, ENEL, waterproofed using CARPI geomembrane systems 5 more
large gravity, multiple arch and masonry dams, 7 reservoirs, and several canals and
hydraulic tunnels. The report highlights also the advantages of this waterproofing
system which led to its selection as the preferred solution over other options, that
676
GR Q. 104
include: simplicity and speed of installation, the watertightness at joints, a relatively

lower cost, the lack of routine maintenance, and the possibility of easy and quick
repair, if needed (nowadays, even underwater repair). Additional rehabilitation works
described in the report are: the placement of additional weight in the form of a new
reinforced concrete block, raising the central part of the dam by approx. 6 m with
an inspection gallery inside it, the restoration of the downstream face, replacing the
most degraded parts of the existing reinforced gunite layer, the increase of the spill-
way capacity up to the thousand-year flood event, and the installation of a de-icing
system to create a water movement impeding the formation of ice in the proximity
of the upstream face which driven by wind could damage the geomembrane.
Report R20 received from Sweden (Malm et al.) it is the only one among
all those received that is related with concrete buttress dams. The purpose of the
report is to show how numerical analyses can be used to plan monitoring of dams,
such as selection of locations and type of sensors, and how they can be used to
predict the future dam behavior, which is important to determine the required sen-
sitivity and accuracy of the sensors. FE models can be used with high accuracy for
defining warning levels early in the project and be used for definition of alert values,
i.e. indications of potential failures. This concept is applicable to many dams that
are approaching the end of their expected service life, where the monitoring of the
dam is crucial to ensure that a high safety of the dam is maintained while the dam
remains in service. The authors recommend that the sensors should be installed in
strategic positions to give early indications if the dam behavior changes and pro-
vide early warnings of potential dam failures. For monitoring of new projects, the
commonly used databased regression methods are not applicable to use to define
warning levels since they require sufficiently long recordings of the dam behavior to
train the algorithms. The report includes two study cases where this concept has
been applied. These dams are two thin concrete flat slab buttress dams in Sweden,
Ramsele and Storfinnforsen, both located 10 km apart on the Faxälven river, both
have a maximum height slightly above 40 m, and both were built in the 1950s. In
the 1960s several cracks were detected, following the same pattern, in both dams.
Structural rehabilitation was performed during the 1990s. But about 10 years after
the performed rehabilitation program, further cracking was observed. The risk of new
potential internal failure modes resulted in undertaking a new extensive strength-
ening and service life-extension program, including the installation of pre-stressed
tendons in the most extensively damaged monoliths to tie the buttresses to the
foundation. Prior to implementation of an extensive monitoring program, numeri-
cal analyses were conducted, to determine potential failure modes and selection of
suitable types of sensors, and their locations at the dam. In addition, the numeri-
cal models were used to define warning and alert values for some of the sensors.
The six monoliths that were chosen for extensive monitoring were also modelled
using Abaqus and Brigade Plus software. The results from the analyses showed
the expected variation in the dam behavior and determined where and how differ-
ent potential failure modes may initiate and develop. Based on these analyses, it
was possible to identify suitable type and locations of sensors for monitoring various
potential failure modes. After 1.5 years of data collection, the recorded measure-
ments have so far shown behavior that is quite close to the expected, with small
677
GR Q. 104
variations in crest displacements and crack widths. Undoubtedly, the report intro-
duces a very valuable tool that can provide accurate predictions that are necessary
for supporting the owner in decision making and planning future maintenance.
Report R25 (Tzenkov et al., Bulgaria), besides dam performance also deals
with a dynamic analysis of the Beli Iskar CVC gravity dam, and for that reason has
already been discussed in Section 4.1.
In their report, Droz & Wohnlich (R36, Switzerland), discussed partially in

Section 7.1, also include case histories of three CVC gravity dams affected by AAR,
Salanfe Dam (h = 52 m), located in the Swiss Alps and built in 1952, and Inga I
(built in 1972) and Inga II (1982) dams in Democratic Republic of Congo. Salanfe
Dam experienced an irremediable deformation towards upstream due to swelling of
the concrete mass induced by AAR. Extensive inspections and specific extension
of the monitoring system were implemented, as well as a restriction in reservoir
operation before the rehabilitation works took place in 2012–2013. Using a numerical
modelling, integrating the AAR effects in the concrete, the rehabilitation works were
defined, consisting in the execution of 22 vertical 11 mm slot cuts, maximum 22 m
high, in order to release the stresses in the structure. Similarly to Salanfe Dam,
Inga I and Inga II dams presented continuous and regular displacements, but in this
cases towards downstream. As a complement of the rehabilitation of the monitoring
system of the dams, a satellite data analysis has been performed in order to have a
better understanding of the dam movements and identify possible deformation of the
foundation. The SAR-interferometry technology was used, which contributed to the
identification of the causes of the large deformations. In-depth analysis of concrete
samples was also performed, confirming the presence of AAR in the concrete. The
improved monitoring system was able to identify the development of this concrete
expansion pathology, which must be followed regularly, although it does not predict
any safety issues in the near future.
A last report received on this topic of CVC gravity and buttress dams perfor-
mance comes from India (R44, Gupta et al.). It addresses the sediment handling
and operational performance of five large size under-sluice spillways (low-level spill-
ways) of the 62.5 m high Nathpa concrete gravity dam in the Himalayan region. The
spillways are designed to allow the passage of water with a high concentration of
suspended sediments, together with bed load comprising cobbles and boulders. For
this reason, abrasion resistance steel liner and steel rails were implemented, along
with high performance concrete, and the report explains the details of the durability
of these protections over time. During the 15-years period from 2006 to 2020, each
sluice liner has resisted sediment load of 118 million Ton with velocity greater than
20 m/s. It is noteworthy that 69% of sediments particle have minerals with hard-
ness number on Mohs scale greater than 6. Liner also resisted impact of boulders
with average size greater than 1 m and even as large as 3.50 m, without any major
damage. But the top of rails, that were installed at the chute of the spillways, was
worn away by beatings and grinding and the small-aggregate concrete between the
rails was disintegrated and washed away and the rails were dislodged and bent. In
order to extend the periods between repairs, the methodology adopted for these was
678
GR Q. 104
modified in 2013, and a new one is being proposed for 2021, including thicker layers
of high performance concrete with silica fume, with a minimum compressive strength
of 80 MPa and high durability. As a lesson learned, composite system of steel rail
and concrete should be avoided in high velocity and high sediment load conditions.
7.3. PERFORMANCE OF RCC GRAVITY DAMS
The first report received for this topic is from Italy (R4, Miranda et al.), which
deals with geomembrane systems and describes as a case history the Miel I RCC
dam in Colombia, which was discussed in Section 6.3.
In their report, Prasetio & Muis (R38, Indonesia) provide the details of the
intensive monitoring program implemented in two dams located in the center of
Sulawesi Island. The dams are in an earthquake-prone area due to the presence of
two major faults, and movements of the earth’s surface that originate from volcanic
activity and earthquakes are quite common occurrences. The dams are Balam-
bano, a 100 m-high RCC gravity dam, and Karebbe, a combined RCC–CVC dam
74 m-high. The report provides the list of instrumentation installed in both dams
and discusses the importance of real-time monitoring, so that the prevention of
dam failure due to deformations induced by horizontal and vertical movements due
to earthquakes, landslides, leaks, etc. can be anticipated, as a way to mitigate
damage or evacuate the community downstream. The monitoring involves visual
monitoring, manual instrumentation data retrieval, and real-time data retrieval using
3D and 1D sensors for both static and dynamic movements for dam deformations
(robotic monitoring system). A pilot step was to implement dam deformation mon-
itoring technology based on active stations GPS/GNSS-CORS (Global Positioning
System/Global Navigation Satellite System-Continuously Operating Reference Sta-
tion). The use of Geodetic Monitoring Systems (Geomos) can be considered as
a step forward in a dam monitoring system based on the latest technology, which
is based on speed, accuracy, reliability and effectiveness, with very little room for
human error in dam monitoring activities. Having real-time data, rapid response after
an earthquake (or other dangerous event) can be triggered immediately if abnormal
conditions are found.
8. CLOSING REMARKS
Although there are fewer and fewer suitable sites available for construction of
new dams, particularly for concrete dams, there is still great potential for the design
and construction of new concrete dams of all possible typologies in many countries,
as well as for the raising of existing concrete dams, while the need for rehabilitation
and upgrading of existing dam structures is increasing. In some cases, the raising
could be a more practical and effective alternative to the construction of a new dam.
679
GR Q. 104
In many developing countries, there is a vast untapped hydro potential, and at

the same time, an enormous need for irrigation and water supply, as well as for flood
control. Climate changes are exacerbating the problems of flooding and drought in
many regions all over the world, and this also affects developed countries. These
problems can be alleviated by the justified and reasonable construction of new dams
or raising and modifications to existing dams, which will surely have a greater chance
of success in a context of international and multidisciplinary collaboration.
As we have witnessed the important advances in design and construction of

concrete dams over the last three decades, the technology has been continuously
developing, thanks to the dedication day by day of hundreds of engineers and other
professionals. Young engineers and scientists have taken the advantage of working
side by side with their mentors, and in this way, the expertise can be maintained and
further evolved.
The nearly forty contributions received to Q. 104, prepared by engineers and

other technicians at the cutting edge of analysis, design, construction and operation
& maintenance of concrete dams, attest the vitality existing in this field.
I believe in a promising future for successful operations of existing and for

construction of new concrete dams in the years to come, to be led by the advance-
ments in the state-of-the-art technology and, why not to witness the emergence of
a new form of dam construction, in the same way that happened with the arise of
the RCC/RCD in the early 1980s?
ACKNOWLEDGEMENTS
The General Reporter would like to express his gratefulness to the following
colleagues for providing very helpful advice and references for the writing of this
report: Dr. Malcolm R.H. Dunstan (Great Britain), Carlos Granell and Francisco
Ortega (Spain), Aled Hughes (USA), and Dr. Quentin H.W. Shaw (South Africa).
My thanks also to Dr. José Marques Filho (Brazil) and Dr. Jerzy Salamon
(USA), Chairman and Vice-Chairman, respectively, for Question 104, for reviewing
and contributing with valuable suggestions.
And finally, grateful thanks again to ICOLD President Michael Rogers (USA)
for proposing the subject of concrete dams for this Question 104.
REFERENCES
AKBARI, J., AHMADI, M.T. & MOHARRAMI, H. “Advances in concrete arch dams
shape optimization”. Applied Mathematical Modelling 35, 2011, pp. 3316–
3333.
680
GR Q. 104
CFBR. “Recommandations provisoires pour la justification du comportement des

barrages-voûtes”. Paris, France, 2018.
CHOPRA, A.K. “Earthquake Engineering for Concrete Dams – Analysis, Design,

and Evaluation”. Willey Blackwell 2020.
DU, C. “Recent advances in RCC Arch Dam design”. International Water Power
& Dam Construction, January 2021.
DUNSTAN, M.R.H. “www.rccdams.co.uk” and “RCC dams database. World Atlas

of International Journal of Hydropower & Dams”. October 2020.
FERC. “Engineering Guidelines for the Evaluation of Hydropower Projects,

Chapter 11: Arch Dams”. Washington DC, USA, March 2018.
ICOLD / CIGB. “Bulletin 64: Dam heightenings – Register”. Paris, France, 1988.
ICOLD / CIGB. “Bulletin 177: Roller-Compacted Concrete Dams”. Paris, France,

2020.
ICOLD / CIGB. “World Register of Dams”. Paris, France, updated April 2020.
ICOLD / CIGB. “Numerical Analysis of Dams – Proceedings of the 15 th ICOLD

International Benchmark Workshop”, Milan, Italy. February 2021.
JONKER, M. & ESPANDAR, R. “Evaluation of Existing Arch Dam Design Criteria

in lieu of ANCOLD Guidelines”. ANCOLD Conference, 2014.
NISAR, A., DOLLAR, D., JACOB, P., CHU, D., LOGIE, C. & LI, G. “Nonlinear
incremental stress-strain analysis for Portugues Dam; an RCC gravity arch
dam”. 28th USSD Conference. Portland, Oregon, USA. April 2008.
NUSS, L.K., MATSUMOTO, N. & HANSEN, K.D. “Shaken but not stirred.
Earthquake performance of concrete dams”. 32nd USSD Conference. New
Orleans, Louisiana, USA. April 2012.
PAIXÃO, J. & SILVA MATOS, D. “Shape optimization in the design of con-

crete arch dams – Two case studies”. Blucher Mechanical Engineering
Proceedings. Vol. 1, num. 1, May 2014.
SHAW, Q.H.W. “The structural function of different arch dam types”. Proceedings.
SANCOLD. Dam safety, maintenance and rehabilitation of dams in Southern
Africa. Cape Town, South Africa, September 2015.
SHAW, Q.H.W. “The influence of stress-relaxation creep on the design of

large RCC dams”. Symposium – Knowledge-based Dam Engineering. 85th
ICOLD Meeting. Prague, Czech Republic. July 2017.
VARGHESE, B., SAJU, A. & JOHN, S. “Finite Element Analysis of Arch Dam”.
International Journal of Research in Engineering and Technology. Volume
03, Issue 07, July 2014.
681
GR Q. 104
VICENTE, D.J., SAN MAURO, J., SALAZAR, F. & BAENA, C.M. “An Interactive
Tool for Automatic Predimensioning and Numerical Modeling of Arch Dams”.
Mathematical Problems in Engineering, Volume 2017.
WANG, R. “Key Technologies in the Design and Construction of 300 m Ultra-High

Arch Dams”. Engineering 2, 2016, pp. 350–359.
682
GR Q. 104
-------
-------
QUESTION 104
CONCRETE DAM DESIGN INNOVATION DANS LE

Rafael IBÁÑEZ De ALDECOA

Head of Hydraulic, Underground and Environmental Projects Dragados
Chairman, SPANCOLD’s Technical Committee on Concrete Dams

Member, ICOLD and USSD’s Technical Committees on Concrete Dams
SPAIN/ESPAGNE
General Reporter / Rapporteur Général
DOI: 10.1201/9781003211808-46 683

GR Q. 104
TABLE DES MATIÈRES
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
2. APERÇU DES RAPPORTS REÇUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688
3. INNOVATIONS POUR L’ANALYSE, LA CONCEPTION ET LA
CONSTRUCTION DE BARRAGES-VOÛTES, Y COMPRIS LES
BARRAGES-VOÛTES EN BCR ET LES BARRAGES
POIDS-VOÛTES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 691
3.1. Analyse et conception d’un barrage-voûte en BCV . . . . . . . . . . . . . . . . . . . 694
3.2. Construction d’un barrage-voûte en BCV . . . . . . . . . . . . . . . . . . . . . . . . . . . . 698
3.3. Analyse et conception du barrage-voûte en BCR . . . . . . . . . . . . . . . . . . . . 699
3.4. Construction d’un barrage-voûte en BCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 701
4. INNOVATIONS POUR L’ANALYSE DE BARRAGES-POIDS EN
BÉTON, LA CONCEPTION DES MATÉRIAUX ET LA
CONSTRUCTION, Y COMPRIS EN BCR ET DE TYPE CMD . . . . . . . . . . . . . 701
4.1. Analyse et conception d’un barrage-poids en BCV . . . . . . . . . . . . . . . . . . . 702
4.2. Construction d’un barrage-poids en BCV . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706
4.3. Analyse et conception d’un barrage-poids en BCR . . . . . . . . . . . . . . . . . . . 708
4.4. Matériaux et compositions es béton des barrages-poids en BCR . . . . . 710
4.5. Construction d’un barrage-poids en BCR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712
4.6. Analyse d’un barrage en matériaux cimentés:
conception, matériaux et construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714
5. INNOVATIONS POUR LA SURÉLÉVATION DE BARRAGES EN
BÉTON EXISTANTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715
5.1. Surélévation des barrages en béton existants en BCV . . . . . . . . . . . . . . . 716
5.2. surélévation des barrages en béton existants en BCR . . . . . . . . . . . . . . . 717
6. INNOVATIONS POUR LES BARRAGES EN BÉTON EXTRÊMEMENT
HAUTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 718
6.1. Barrages-voûtes en BCV extrêmement élevés . . . . . . . . . . . . . . . . . . . . . . . 718
6.2. Barrages-voûtes en BCR extrêmement élevés . . . . . . . . . . . . . . . . . . . . . . . 720
6.3. Barrages-poids en BCR extrêmement élevés . . . . . . . . . . . . . . . . . . . . . . . . 720
7. COMPORTEMENT OPÉRATIONNEL DES BARRAGES EN BÉTON
PENDANT LE CYCLE DE VIE, Y COMPRIS DANS DES CONDITIONS
EXTRÊMES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .724
7.1. Comportement des barrages-voûtes en BCV . . . . . . . . . . . . . . . . . . . . . . . . 724
7.2. Comportement des barrages-poids et contreforts en BCV. . . . . . . . . . . .726
7.3. Comportement des barrages-poids en BCR . . . . . . . . . . . . . . . . . . . . . . . . . 729
8. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 730
REMERCIEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
RÉFÉRENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
684
GR Q. 104
1. INTRODUCTION
Je voudrais commencer ce rapport général pour la question 104, “ Innovation

et performance de la conception de barrages en béton ” en soulignant que la dernière
fois qu’il y a eu une question spécifiquement sur les barrages en béton (c’est-à-
dire si nous ne prenons pas en compte des thèmes tels que le vieillissement, la
réhabilitation ou la modernisation) lors d’un congrès de la CIGB, c’était en 1988 à
San Francisco (USA).
Cela pourrait conduire à penser qu’il n’y a pas eu de progrès significatifs dans
l’état de l’art dans la conception et la construction de barrages en béton depuis le
Congrès de San Francisco, mais rien ne pourrait être plus éloigné de la réalité. J’ai
commencé ma carrière professionnelle en tant qu’ingénieur de barrages au milieu
des années1980 et depuis lors, nous avons assisté à des percées techniques très
importantes pour les barrages en béton.
En fait, à la fin des années 1980, la technologie émergente du béton com-

pacté au rouleau (BCR) (et, en parallèle, les barrages de type RCD au Japon) n’en
était qu’à ses débuts. Cette nouvelle technique de construction a connu un succès
étonnant depuis sa création, avec un développement / raffinement continu de la
technique au cours des décennies suivantes et avec un nombre remarquable de
barrages achevés dans le monde entier à ce jour. Il y a quelques années, la tech-
nique a atteint un certain état de maturité, alors qu’il n’est pas improbable que de
nouveaux développements substantiels aient lieu au cours des prochaines années.
À l’origine, l’incursion du BCR dans le champ des barrages-voûtes a été très

timide mais, au fil des ans, la confiance dans la nouvelle technique s’est accrue lors
de la conception et de la construction des voûtes en BCR.
Bien qu’il faille reconnaître qu’au cours des trois dernières décennies, il n’y
a pas eu de percées significatives en termes de barrages-poids conventionnels
en béton vibré (BCV), très probablement motivés par la perte de compétitivité
par rapport aux nouveaux barrages-poids en BCR, ils se sont produits avec des
barrages-voûtes en béton conventionnel où cette perte de compétitivité par rapport
aux barrages-voûtes en BCR peut ne pas être aussi évidente dans certains cas. Le
simple fait qu’au cours de la dernière décennie, 6 barrages-voûtes à double courbure
de 270 m ou plus de hauteur aient été construits, ou soient dans un état de construc-
tion avancé, est une preuve qui nous invite à réfléchir aux grands progrès qui ont été
récemment réalisés dans la conception et la construction des barrages-voûtes BCV.
Trois pays en particulier, la Chine, la Turquie et l’Iran, sont actuellement à la tête de
la construction de nouveaux barrages-voûtes en BCV extrêmement élevés (dans la
section 6, nous clarifierons la signification des barrages “ extrêmement élevés ”).
Un autre champ dans lequel des réalisations importantes ont été entreprises
est celui de la surélévation de barrages existants, en BCV ou en BCR, et il
685
GR Q. 104
reste un domaine à développer plus en profondeur: les barrages hybrides BCR-

enrochement comme solution potentiellement compétitive dans certaines circon-
stances favorables des conditions de fondation, de disponibilité des matériaux et de
dimensionnement des évacuateurs de crues.
Parallèlement aux innovations et au développement de l’analyse, de la con-

ception et de la construction de barrages en BCV et en BCR, diverses disciplines
connexes, telles que l’arpentage, la surveillance, les pathologies du béton, la réha-
bilitation, la valorisation, les matériaux et mélanges en béton, les géomembranes
imperméables…ont également connu des progrès importants. Il est à noter que,
pour toutes les disciplines précédentes, un ou plusieurs rapports ont été reçus pour
Q. 104.
En jetant un coup d’œil dans le passé, les barrages en béton et les sujets
connexes ont été le thème principal de nombreuses questions lors des congrès
précédents de la CIGB, à savoir:
• Q. 1a (1933, Stockholm): Détérioration par vieillissement du béton des

barrages-poids.
• Q. 1b (1933): Influence de la température interne et distorsion des barrages-
poids.
• Q. 3 (1936, Washington): Ciments spéciaux.
• Q. 5 (1936): Étude du parement des barrages en maçonnerie et en béton.
• Q. 8 (1948, Stockholm): Soulèvement et contraintes qui en résultent dans les
barrages.
• Q. 11 (1948): Informations obtenues grâce à l’utilisation de méthodes d’essai
et de ciments spéciaux dans les grands barrages.
• Q. 15 (1951, New Delhi): Béton pour grands barrages.
• Q. 17 (1955, Paris): Economie et sécurité des différents types de barrages en
béton.
• Q. 19 (1955): La relation entre la teneur en ciments du béton et la perfor-
mance dans la pratique: a) des barrages-poids (intérieurs et extérieurs), b)
des barrages-voûtes, c) des barrages à contreforts et leur influence sur la
perméabilité et la résistance au gel.
• Q. 20 (1958, New York): Renforcement des barrages existants, y compris les
méthodes de construction de nouveaux barrages par étapes successives.
• Q. 21 (1958): Observation des contraintes et des déformations dans les bar-
rages et dans leurs fondations et culées; comparaison de ces observations
avec des calculs et des tests sur des modèles à petite échelle.
• Q. 23 (1958): Utilisation d’adjuvants et de matériaux pouzzolaniques dans le
béton pour les barrages et influence des particules de sable plus fines.
• Q. 24 (1961, Rome): Sélection, traitement et spécification des agrégats pour
le béton pour les grands barrages.
• Q. 26 (1961): Techniques modernes de barrages en béton pour de larges
vallées et travaux auxiliaires.
• Q. 30 (1964, Édimbourg): Conception de béton pour de grands barrages de
tous types et influence de l’âge sur les propriétés du béton.
• Q. 34 (1967, Istanbul): Le comportement et la détérioration des barrages.
686
GR Q. 104
• Q. 39 (1970, Montréal): Développements récents dans la conception et la

construction de barrages en béton.
• Q. 43 (1973, Madrid): Nouvelles idées pour une construction plus rapide et
économique de barrages en béton.
• Q. 57 (1985, Lausanne): Les barrages en béton – un vieux problème tou-
jours présent: la fissuration; une nouvelle technologie: le béton compacté au
rouleau.
• Q. 62 (1988, San Francisco): Nouveaux développements dans la construction
de barrages en béton.
• Q. 65 (1991, Vienne): Vieillissement des barrages et mesures correctives.
• Q. 82 (2003, Montréal): Vieillissement et réhabilitation de barrages en béton
et en maçonnerie et ouvrages associés.
• Q. 90 (2009, Brasilia): Modernisation des barrages existants.
• Q. 95 (2012, Kyoto): Vieillissement et mise à niveau.
• Q. 99 (2015, Stavanger): Modernisation et réingénierie des barrages existants.
Comme on peut le voir sur la liste ci-dessus, depuis le premier Congrès de

la CIGB en 1933, les questions liées aux barrages en béton ont été assidûment
et régulièrement traitées jusqu’au Congrès de 1973. Depuis 1973, on trouve seule-
ment deux questions dans les quinze derniers Congrès, c’est-à-dire sur une période
de près d’un demi-siècle, la dernière étant Q. 62 au Congrès de 1988 (sans inclure
dans ce décompte les questions relatives au vieillissement, à la réhabilitation ou au
renforcement.
La liste ci-après fournit la liste des bulletins de la CIGB relatifs aux barrages en
béton (elle n’inclut pas ceux qui sont spécifiques aux fondations, à la surveillance,
à la sismicité, à la réhabilitation, à la sécurité...):
• Bulletin 15 (1960): Résistance au gel du béton. Comparaison des résultats

obtenus dans différents laboratoires.
• Bulletin 20 (1968): Adjuvants tensio-actifs pour béton pour grands barrages.
• Bulletin 22 (1972): Pouzzolanes et laitiers pour béton pour grands barrages.
• Bulletin 24 (1973): Accélération et retard des mélanges.
• Bulletin 25 (1976): Extensibilité du béton pour les grands barrages.
• Bulletin 26 (1976): Méthodes de détermination des effets du retrait, du fluage
et de la température sur le béton pour les grands barrages.
• Bulletin 30 (19 78-1987): Méthodes par éléments finis dans l’analyse et la
conception des barrages.
• Bulletin 36 (1982): Ciments pour béton pour grands barrages.
• Bulletin 47 (1983): Contrôle de la qualité du béton.
• Bulletin 57 (19 86): Matériaux pour joints de barrages en béton.
• Bulletin 63 (1988): Nouvelles méthodes de construction – État de l’art.
• Bulletin 64 (1988): Surélévations de barrages – Registre
• Bulletin 71 (1989): Exposition du béton de barrage à des eaux agressives
spéciales – Lignes directrices
• Bulletin 75 (1989): Béton compacté au rouleau pour barrages-poids – État
de l’art.
687
GR Q. 104
• Bulletin 76 (1990): Méthodes conventionnelles dans la construction de

barrages – Revue.
• Bulletin 79 (1991): Réaction alcali-agrégat dans les barrages en béton –
Examen et recommandations.
• Bulletin 93 (1994): Vieillissement des barrages et des ouvrages annexes –
Examen et recommandations.
• Bulletin 94 (1994): Logiciel informatique pour les barrages. Validation –
Commentaires et propositions.
• Bulletin 107 (1997): Barrages en béton – Contrôle et traitement des fissures.
• Bulletin 117 (2000): Le barrage-poids: un barrage pour l’avenir – Examen et
recommandations.
• Bulletin 122 (2001): Procédures informatiques pour l’ingénierie des barrages
– Fiabilité et applicabilité.
• Bulletin 126 (2003): Barrages en béton compacté au rouleau – État de l’art et
études de cas.
• Bulletin 136 (2009): Spécification et le contrôle de qualité des barrages en
béton.
• Bulletin 145 (Pré-impression): Les propriétés physiques du béton convention-
nel durci dans les barrages.
• Bulletin 155 (2013): Du bon usage des modèles numériques dans l’ingénierie
des barrages.
• Bulletin 165 (Pré-impression): Sélection des matériaux pour le béton des
barrages.
• Bulletin 177 (2020): Barrages en béton compacté au rouleau.
• Bulletin 184 (Pré-impression): Analyse des réactions chimiques expansives
dans les barrages en béton et les projets hydroélectriques.
Pour conclure cette introduction, je juge nécessaire de révéler que, au sein

du Comité “D” de la CIGB sur les barrages en béton, nous avons accueilli comme
une nouvelle fantastique que, lors de la 87e Assemblée générale, tenue à Ottawa
(Canada) en juin 2019, une “mise à jour” sur l’état de l’art des barrages en béton a
été sélectionnée comme l’une des questions pour le présent Congrès à Marseille.
Nous sommes très reconnaissants au Président de la CIGB, Michael Rogers, ancien
Président du Comité D, qui a fait cette proposition comme “Question du Président”,
proposition qui a été acceptée à l’unanimité. La question 104 comprend cinq sous-
questions, qui correspondent aux sections 3 à 7 du présent rapport général.
J’invite les lecteurs à se joindre à moi pour examiner ce rapport général, où

nous connaîtrons certains des derniers développements liés aux barrages en béton,
en prenant comme base de référence les excellents rapports reçus pour la question
104 de plusieurs experts de premier plan dans ce domaine.
2. APERÇU DES RAPPORTS REÇUS
En tant que contributions pour la question 104, 41 rapports ont été reçus, de
19 pays appartenant à quatre continents. Les pays qui ont soumis un plus grand
688
GR Q. 104
nombre de rapports sont la Chine, avec six rapports, et la France et l’Italie avec cinq
rapports chacun. Trois rapports ont été reçus de l’Inde et de la Suisse, deux de la
Grande-Bretagne, de l’Indonésie, de l’Iran, du Japon et de l’Espagne et, enfin, neuf
pays différents ont contribué avec un rapport chacun.
Le tableau 1 ci-dessous résume la répartition des rapports par continent,

montrant que les contributions de l’Europe et de l’Asie représentent 88% du total,
l’Europe à elle seule fournissant environ de la moitié.
Tableau 1
Répartition géographique des rapports
CONTINENT RAPPORTS % PAYS
Europe 21 51% 9
Asie 15 37% 5
Amérique 3 7% 3
Afrique 2 5% 2
Total 41 100% 19
Cependant, il convient de noter que deux des rapports, malgré leur bonne
qualité, ne correspondent précisément à aucune des sous-questions de la Q. 104 ;
elles n’ont donc pas été prises en compte dans la préparation du présent rapport
général.
Le petit nombre de rapports reçus des comités nationaux d’Afrique et

des Amériques est frappant, bien que, si nous faisons une répartition selon
l’emplacement des barrages abordés dans les rapports, nous avons une image
quelque peu différente comme le montre le tableau 2 (certains rapports traitent de
plusieurs barrages).
Tableau 2
Répartition géographique des barrages cités
dans les rapports
CONTINENT BARRAGES % PAYS
Europe 27 39% 8
Asie 22 32% 12
Afrique 13 19% 10
Amérique 7 10% 5
Total 69 100% 34
La majorité des rapports reçus portent sur plusieurs sujets dans l’une des
cinq sous-questions de la Question 104, et dans certains cas, avec des sujets
appartenant à différentes sous-questions. Le tableau 3 permet d’avoir un résumé
des sujets couverts par les différents rapports, et il fournit un outil pour décider,
pour chaque rapport, quelle est la sous-question dans laquelle ils correspondent le
mieux, et pour assembler la présentation générale de ce rapport général. Les sujets
sans contributions reçues ont été omis du tableau.
689
GR Q. 104
La répartition des rapports par type de barrage discuté est la suivante, en

gardant à l’esprit que quelques rapports se réfèrent au même barrage, et d’autres
traitent de plus d’un barrage de la même typologie ou d’une typologie différente:
16 rapports traitent des barrages-poids en BCR, 11 des barrages-poids en BCV, 9
des barrages-voûtes en BCV, 3 des barrages-voûtes en BCR, 1 des barrages de
contrefort BCV et 1 sur les barrages en matériaux cimentés (CMD) et, en outre, 2
de l’élévation des barrages-voûtes en BCV et 1 sur l’élévation des barrages-poids
en BCV.
Tableau 3
Classification par sous-questions et thèmes des rapports contribuant à la
Question 104
SECTION DU NOMBRE NUMÉROS
SOUS- RAPPORT DE DES
QUESTION GÉNÉRAL SUJET RAPPORTS RAPPORTS
a) 3.1 Analyse et conception 9 R2, R11, R15, R18, R31,
d’un barrage-voûte en R32, R33, R34, R42
BCV
3.2 Construction d’un 2 R18, R34
barrage-voûte en BCV
3.3 Analyse et conception du 3 R16, R17, R23
barrage-voûte en BCR
3.4 Construction d’un 1 R17
barrage-voûte en BCR
b) 4.1 Analyse et conception 9 R1, R2, R12, R19, R20, R25,
d’un barrage-poids en R39, R43, R45
BCV
4.2 Construction d’un 2 R21, R39
barrage-poids en BCV
4.3 Analyse et conception 6 R7, R9, R17, R22, R23, R40
d’un barrage-poids en
BCR
4.4 Matériaux et mélanges 5 R7, R10, R13, R24, R30
des barrages-poids en
BCR
4.5 Construction d’un 7 R14, R17, R24, R27, R28,
barrage-poids en BCR R29, R30
4.6 Analyse d’un barrage 1 R8
en matériaux cimentés:
conception, matériaux et
construction
c) 5.1 Innovations pour 3 R34, R36, R43
l’élévation de barrages
en béton existants avec
BCV
d) 6.1 Barrages-voûtes en BCV 2 R31, R32
extrêmement élevés
6.2 Barrages-voûtes en BCR 2 R16, R17
6.3 Barrages-poids en BCR 4 R4, R7, R14, R22
e) 7.1 Comportement des 5 R18, R31, R33, R36, R42
barrages-voûtes en BCV
7.2 Comportement des 5 R3, R20, R25, R36, R44
barrages-poids et
contreforts en BCV
7.3 Comportement des 2 R4, R38
barrages-poids en BCR
690
GR Q. 104
3. INNOVATIONS POUR L’ANALYSE, LA CONCEPTION ET LA

CONSTRUCTION DE BARRAGES-VOÛTES, Y COMPRIS LES
BARRAGES-VOÛTES EN BCR ET LES BARRAGES POIDS-VOÛTES
Même si la construction de nouveaux barrages-voûtes en béton est limitée, par

rapport à la construction de nouveaux barrages-poids en béton, principalement en
raison du fait que les premiers nécessitent des sites de barrage avec de meilleures
conditions géologiques et géotechniques, et une morphologie adéquate de la val-
lée, c’est un type de barrage très compétitif lorsque les conditions sont favorables,
notamment pour les barrages très hauts et extrêmement hauts, comme nous le
verrons au chapitre 6.
En fait, la plupart des plus hauts barrages construits (ou en cours de construc-
tion) dans le monde, au cours des quinze dernières années, ont été des structures
en béton à double courbure ou poids-voûte: tous sauf un de ceux dépassant 269 m
(6 barrages à double courbure), et 10 sur les 16 au-delà de 229 m. Dix barrages
à double courbure ou poids-voûtes ont été construits (ou sont en construction) en
Chine (7 barrages), en Turquie (2) et en Iran (1).
Il convient de souligner que ces barrages-voûtes extrêmement élevés sont

tous des barrages en BCV. Le plus haut barrage-voûte en BCR a une hauteur plus
“ modeste ” de 168 m. Les tableaux 4 et 5 ci-dessous montrent, respectivement, les
plus hauts barrages achevés ou en construction, de tous types, et les plus hauts
barrages-voûtes en BCR achevés dans le monde.
Tableau 4
Plus hauts barrages du monde (250 m et plus), tous types
(CIGB, avril 2020)
NOM DU TYPE DE ANNÉE D’
BARRAGE PAYS HAUTEUR (m) BARRAGE ACHÈVEMENT
Shuangjiangkou Chine 312 Remblai U.C.*
Jinping Ier Chine 305 Double courbure 2013
Nurek Tadjikistan 300 Remblai 1980
Xiaowan Chine 294 Double courbure 2012
Baïhetan Chine 289* Double courbure U.C.*
Xiluodu Chine 286 Double courbure 2014
Grande Dixence Suisse 285 Gravité du béton 1961
Yusufeli Turquie 275* Double courbure 2020*
Enguri Géorgie 272 Double courbure 1987
Wudongde* Chine* 270* Double courbure* U.C.*
Nuozhadu Chine 262 Remblai 2014
Chicoassur Mexique 262 Remblai 1980
Tehri Inde 260 Remblai 2006
Laxiwa Chine 250 Double courbure 2010
Mauvoisin Suisse 250 Double courbure 1957
*mis à jour par le rapporteur général
691
GR Q. 104
Tableau 5
Plus hauts barrages-voûtes en BCR (Dunstan, 2020)
NOM DU BARRAGE ANNÉE
BARRAGE PAYS HAUTEUR (M) BARRAGE D’ACHÈVEMENT
Wanjiakouzi Chine 168 Voûte mince 2011
Janneh Liban 157 Poids-voûte 2021
Sanhekou Chine 142 Poids-voûte 2020
Xiangbiling Chine 142 Voûte mince 2017
Qinglong Chine 140 Voûte mince 2012
Yunlonghe N◦ 3 Chine 135 Voûte mince 2011
Dahuashui Chine 135 Voûte mince 2007
Sanliping Chine 133 Voûte mince 2011
Gomal Zam Pakistan 133 Poids-voûte 2011
Lizhou Chine 132 Voûte mince 2014
Shapai Chine 132 Voûte mince 2003
En regardant les deux tableaux ci-dessus, il est clair qu’une grande majorité
des barrages-voûtes de haute hauteur les plus récemment construits, à la fois
en BCV et en BCR, se trouvent en Chine, pays où plusieurs barrages-voûtes
extrêmement élevés devraient construits dans les prochaines années.
Pour clarifier le tableau 5, il est nécessaire d’expliquer que les barrages en

BCR “ à voûte mince ” les plus élevés, tous en Chine, n’avaient, dans un premier
temps, pas de pentes négatives, à l’exception de certains qui avaient de légères
pentes négatives (par exemple Shapai), de sorte qu’ils pourraient essentiellement
être considérés comme des barrages-voûtes à courbure simple plutôt qu’à double
courbure. Cependant, certains de ceux qui ont été construits au cours des quinze
dernières années sont de “ vrais ” barrages-voûtes à double courbure, bien que
la courbure verticale des porte-à-faux ne soit généralement pas aussi prononcée
que dans les barrages à double courbure en BCV, ainsi que le degré de courbure
horizontale, afin de faciliter la construction du BCR.
Wanjiakouzi en Chine est actuellement le plus haut barrage-voûte en BCR au

monde, mais le barrage de Batoka prévu de 185 m de haut, sur le fleuve Zambèze de
l’autre côté de la frontière entre la Zambie et le Zimbabwe pourrait être le deuxième
plus haut barrage-voûte en BCR, bien que la typologie finale du barrage n’ait pas
encore été sélectionnée pour le moment.
Il a fallu plusieurs années de construction de barrages-poids en BCR pour

devenir suffisamment confiants, concevoir et construire des barrages-poids en BCR
extrêmement élevés, proches et supérieurs à 200 m de hauteur. Un processus
similaire peut être observé dans la construction de barrages-voûtes en BCR où
l’expérience et la confiance acquises au fil du temps se traduiront par des barrages
construits de même hauteur. Un autre facteur qui pourrait avoir une incidence sur le
choix de la typologie du BCR pour les barrages-voûtes de grande hauteur est le fait
que les avantages d’une construction très rapide avec le BCR ne sont pas si faciles
à obtenir dans un barrage-voûte, en raison des zones de placement plus petites. Le
problème de mise en œuvre des systèmes pour l’injection des joints transversaux,
692
GR Q. 104
qui a été expérimenté au début, a été surmonté, et c’est quelque chose qui ne
devrait pas entraver de manière significative les taux de placement du BCR (CIGB,
2020).
Si la topographie et/ou les conditions géologiques du site du barrage ne sont

pas tout à fait adaptées à une voûte mince en BCV, mais qu’elles le sont pour une
voûte épaisse ou un barrage poids-voûte en BCR, ce type de structure représente
probablement la solution optimale. Il permettra de réaliser d’importantes économies
de volume de béton par rapport à un barrage-poids en BCR et, dans certains cas, en
optimisant la géométrie de la voûte et les matériaux pour les charges thermiques
applicables, le besoin de post-refroidissement et d’injection des joints peut être
éliminé, ce qui rend cette solution encore plus attrayante et compétitive par rapport
à un solution de barrage-poids (Nisar et al., 2008 et CIGB, 2020).
À ce jour, le nombre total de barrages-voûtes en BCR (poids-voûte et voûte

mince) construits ou en construction, d’une hauteur supérieure à 15 m (de facto
tous dépassent 30 m), est de 85, dont 27 (soit 32%) sont supérieurs à 100 m. Il est
intéressant de noter qu’en Chine, de très loin en tête la construction de barrages
en BCR, avec près de 29% du total, 31% des barrages en BCR sont des barrages-
voûtes, alors que dans le reste du monde, les barrages-voûtes en BCR représentent
moins de 2%; seulement 11 barrages-voûtes en BCR sur un total de 607 barrages
en BCR, à l’exclusion des barrages en remblai dur (Dunstan, 2020).
En tant que bibliographie recommandée pour les barrages-voûtes en BCR, le

chapitre 9 du récent Bulletin 177 de la CIGB sur les barrages en BCR (CIGB, 2020)
est une source d’information très précieuse et mise à jour, qui à son tour cite 23
références. D’autres sont d’intérêt (DU, 2021) et (Shaw, 2017).
Et en ce qui concerne les barrages-voûtes en BCV, de nombreux autres, y

compris ceux incorporés en tant que “ références ” dans les rapports reçus, certains
articles et publications d’intérêt qui peuvent être consultés et qui sont énumérés
dans les “ références ” du présent rapport général, sont les suivants: (Akbari et
al., 2011), (CFBR, 2018), (FERC, 2018), (Jonker & Espandar, 2014), (Nuss et al.,
2012), (Paixão & Silva Matos, 2014), (Shaw, 2015), (Varghese et al., 2014), (Vicente
et al., 2017) et (Wang, 2016).
Pour compléter cette introduction générale à la section 3, et ce qui suit

s’applique également à la section 4, il est important de souligner le fait que la
technologie informatique développée dans les années 1970 a permis la mise en
œuvre de méthodes numériques dans l’analyse structurelle des barrages. Au cours
des trente dernières années, les méthodes d’analyse ont encore progressé avec
plusieurs programmes commerciaux avancés à la disposition des ingénieurs. Des
modèles linéaires et non linéaires complexes dans les domaines fréquentiel et tem-
porel sont maintenant utilisés dans l’analyse et la conception de nouveaux barrages
et les modifications apportées aux structures existantes. Les progrès des méth-
odes de calcul soutenus par la puissance de calcul élevée de la nouvelle génération
d’ordinateurs permettent des simulations réalistes du comportement du système
693
GR Q. 104
barrage/fondation/réservoir et une meilleure compréhension de son comportement.

Cependant, une complexité des problèmes d’ingénierie et mathématiques analysés
peut introduire des niveaux élevés d’incertitude dans les résultats. Les modèles
d’analyse structurelle avancée des barrages nécessitent une évaluation appro-
priée de l’efficacité, de la fiabilité et de la validité. Les jugements d’ingénierie et
l’interprétation correcte des résultats sont des éléments importants de l’évaluation
structurelle des barrages en béton (CIGB, 2021 et Chopra, 2020).
3.1. ANALYSE ET CONCEPTION D’UN BARRAGE-VOÛTE EN BCV
Au total, neuf rapports ont été reçus, dans lesquels les sujets de l’analyse,
la modélisation structurelle ou de la conception des barrages-voûtes en BCV sont
traités comme le reflète le tableau 3.
Noorzad et al. (R2, Iran) mènent une enquête sur le rôle des effets de prop-
agation des ondes dans l’estimation correcte des niveaux de contrainte dans le
corps d’un barrage en béton. Dans l’étude, un demi-espace avec une vallée semi-
cylindrique est analysé sous des ondes de cisaillement harmoniques inclinées (SH),
avec différents angles d’incidence, pour l’étude de l’amplification de surface, en util-
isant une méthode de réduction de domaine (DRM). Les résultats de l’analyse sont
vérifiés avec des solutions analytiques du problème, indiquant la précision de la
méthode mise en œuvre. L’étude, bien qu’elle s’applique à la fois aux barrages-
poids et aux barrages-voûtes, met en évidence le fait que les effets de propagation
des vagues sont plus importants pour ces derniers.
Dans son rapport, Chopra (R11, USA) présente une discussion approfondie
sur les analyses sismiques des barrages-voûtes. Le rapport met en évidence
les défauts de la méthode de sous-structure, qui était un très bon outil il y a
quelques décennies, mais qui est devenue obsolète après l’apparition des tech-
niques de calcul les plus récentes. Le rapport souligne également que, jusqu’à il
n’y a pas si longtemps et en raison des capacités limitées de la plupart des pro-
grammes commerciaux d’analyse par éléments finis, il est devenu courant d’ignorer
l’amortissement de la masse et des matériaux de la roche, ainsi que la compressibil-
ité de l’eau, bien que cette hypothèse sous-estime sensiblement la réponse. Après
cette introduction, l’article se concentre sur la présentation de la philosophie des
éléments finis et sur l’explication de sa mise en œuvre dans l’étude de cas du bar-
rage de Morrow Point, ainsi que sur les méthodes de sélection et de mise à l’échelle
des mouvements du sol pour le site du barrage. Le rapport explique comment la
méthode des éléments finis direct surmonte les limites de la méthode standard en
introduisant des limites absorbant les ondes (ou ne les réfléchissant pas) à deux
endroits: 1) à l’extrémité amont du domaine fluide, pour modéliser sa longueur
essentiellement semi-infinie, et 2) aux limites inférieure et latérale du domaine
de fondation, pour modéliser sa géométrie semi-illimitée. De plus, le modèle du
domaine des fluides inclut maintenant la compressibilité de l’eau et l’absorption des
vagues dans les sédiments du fond du réservoir ; la modélisation de la fondation
694
GR Q. 104
prend en compte la masse, la rigidité et l’amortissement des matériaux rocheux.

L’interaction eau-fondation est également prise en compte. Par conséquent, les
hypothèses simplifiées de fondation rocheuse sans masse et d’eau incompressible
dans la méthode classique des éléments finis sont évitées.
Robbe et al. (R15, France) décrivent une étude sur les phénomènes de cisaille-
ment au contact béton-rocher des barrages-voûtes. Le rapport souligne que la
plupart des barrages-voûtes existants en France ont été construits au cours des
deux décennies qui ont suivi la Seconde Guerre mondiale, lorsque les méthodes
de conception utilisées à l’époque étaient basées sur des méthodes de calcul sim-
plifiées fondées sur une évaluation des contraintes sur les parements amont et aval
et, en particulier, n’incluaient pas une analyse de l’état de cisaillement au contact
béton-rocher. L’article fait référence au récent guide pour l’analyse du comportement
des barrages-voûtes (CFBR, 2018), qui recommande d’évaluer l’état de cisaillement
à la base des consoles et décrivent certains mécanismes de défaillance connexes,
différenciant les consoles centrales et latérales. Une analyse numérique compara-
tive de dix barrages-voûtes existants d’une hauteur comprise entre 23 et 150 m est
effectuée, ce qui révèle un cisaillement considérable aux contacts console-fondation
avec une cohésion mobilisée dépassant parfois 1 MPa, un ordre de grandeur que l’on
peut facilement trouver à partir du raisonnement simple que, en moyenne, 1/3 de la
charge hydrostatique sur le plot central est transférée par cisaillement au niveau du
pied de la console. Les résultats ont été obtenus en utilisant des calculs par éléments
finis ou différences finies, avec des modèles non linéaires capables de libérer les
contraintes de traction, souvent au niveau d’une articulation préalablement implan-
tée. Pour la majeure partie des barrages analysés, aucun déplacement du pied aval
n’a été mesuré. Enfin, le comportement mécanique spécifique de quatre barrages
à double courbure et d’un barrage à simple courbure, avec des hauteurs allant de
23 à 70 m, analysés par plusieurs grandes sociétés de conseil, est discuté, avec la
description des différentes modélisations appliquées à chaque cas particulier, ainsi
que les solutions de renforcement, le cas échéant, mises en œuvre: renforcement
de la culée rocheuse en aval, renforcement par ancrage des culées de la voûte,
épaississement des consoles latérales et surveillance piézométrique au pied des
plots centraux.
Un autre rapport français par Grimal et al. (R18), traite des pathologies
du béton et de la modélisation du béton affecté par les réactions chimiques.
Le gonflement interne comme celui dû à la réaction alcali-granulats induit des
dommages au matériau qui peuvent compromettre les propriétés mécaniques du
béton et la stabilité globale de la structure. Des modèles numériques capables
de décrire l’état interne du béton, ainsi que son évolution, peuvent être un outil
efficace dans l’évaluation des conditions structurelles des barrages afin de pour-
suivre l’exploitation. Les modèles peuvent permettre d’évaluer la nécessité d’un
renforcement et, le cas échéant, ils permettent de dimensionner les travaux de ren-
forcements les plus adéquats et d’optimiser le coût. Le rapport décrit tout d’abord une
méthodologie développée pour l’exploitation en toute sécurité des barrages en béton
695
GR Q. 104
avec gonflement, basée sur trois piliers: la surveillance, les tests de laboratoire et
la modélisation numérique pour évaluer le comportement des structures affectées,
la modélisation de la structure prenant une place prédominante. La méthodologie
est fondée sur quatre phases successives. Le rapport présente également le mod-
èle phénoménologique utilisé à EDF/CIH, afin de simuler le gonflement interne dû
aux réactions alcali-agrégat (AAR) ou à la réaction sulfatique interne (ISA). Ce
modèle utilise le cadre de la poromécanique et prend en compte le couplage des
différents phénomènes évoluant dans le béton dont le retrait, le fluage, les dom-
mages anisotropes et les réactions de gonflement interne. En plus d’expliquer le
comportement de la structure dans son état actuel, le modèle permet de prédire
les déformations et les contraintes à long terme des structures en béton affectées
par le gonflement ; il a été comparé à des cas réels à des fins de vérification.
Le modèle a été mis en œuvre sur plusieurs barrages en France, tels que les bar-
rages de Chambon (AAR), Maury et Castelnau (ISA) ainsi que sur certains barrages
d’Hydro-Québec (AAR) au Canada. Enfin, il est décrit l’histoire du cas du barrage
à double courbure de Bimont (h = 86, 5 m) affecté par un gonflement dû à l’ISA, et
les travaux de réhabilitation proposés et réalisés, qui sont abordés à la section 3.2,
sur la base des résultats d’une modélisation numérique 3D et de deux campagnes
d’investigation.
Les deux rapports suivants traitant de l’analyse et de la conception d’un

barrage-voûte en BCV ont été reçus de la Chine par Zhong et al. (R31), et Lijun et
al. (R32). Les deux articles font référence au barrage actuellement le plus haut du
monde, Jinping I (h = 305 m). Les deux rapports sont examinés à la section 6.1, qui
traite des barrages-voûtes en BCV extrêmement élevés.
Le premier des deux rapports reçus de la Suisse, Ballisat et al. (R33), décrit
une expérience intéressante de démantèlement du barrage poids-voûte de Spital-
lamm (h = 114 m), qui sera remplacé par un barrage-voûte à double courbure. Le
nouveau barrage, situé juste en aval de l’original, a été conçu pour être éventuelle-
ment rehaussé de 22 m à l’avenir. Le rendement et les problèmes rencontrés par
le barrage existant, qui ont mené à son déclassement et à son remplacement par
un nouveau barrage, sont abordés à la section 7.1. Dans la présente section, nous
nous concentrons sur la partie du rapport concernant la conception de la structure
du barrage projeté. Après avoir écarté la solution d’un nouveau barrage-voûte à dou-
ble courbure directement placé sur le béton sain du barrage existant, le concept de
construction d’un barrage entièrement nouveau en aval du barrage existant présen-
tait deux avantages importants: 1) aucune interférence avec le fonctionnement du
réservoir existant; 2) pas besoin de démolition du barrage existant, qui pourrait être
laissé en place avec un système de dérivation pour équilibrer les niveaux d’eau
entre le réservoir et dans l’espace entre les deux barrages. Compte tenu des lim-
ites d’espace et des excellentes conditions de fondation, un barrage-voûte à double
courbure était la meilleure alternative, sinon la seule solution possible. La crête
du barrage a été fixée à la même altitude que le barrage existant, avec la pos-
sibilité commentée d’un soulèvement futur. La géométrie du nouveau barrage a
696
GR Q. 104
été définie par des ensembles de paraboles à la fois horizontalement et verticale-

ment. Des analyses 3D des contraintes ont été effectuées pour la configuration
de base et pour le barrage surélevé, afin de s’assurer qu’aucun renforcement ne
serait nécessaire à un stade ultérieur pour la deuxième phase. Pour cette raison,
des arches relativement épaisses ont été adoptées pour la configuration de base.
Une analyse dynamique du barrage a également été réalisée en appliquant trois
accélérogrammes synthétiques non indépendants à la structure. La contrainte de
traction horizontale maximale a atteint 3 MPa, mais elle ne peut pas être considérée
comme critique, car les joints entre consoles s’ouvriront lors de fortes secousses
et la contrainte sera soulagée. L’excavation des fondations a été réalisée tout au
long de 2019 et 2020, avec des mesures supplémentaires prises pour garantir la
sécurité du barrage existant, en exploitation pendant les travaux. Un programme de
surveillance complet a été mis en œuvre, avec des observations qui se poursuiv-
ront pendant le bétonnage du nouveau barrage, prévu de 2021 à 2024. La mise en
eau complète du nouveau barrage une fois que la galerie de contournement sera
opérationnelle (2025) nécessitera également un suivi attentif de la structure et de
ses fondations.
Le deuxième rapport reçu de la Suisse sur ce sujet, Leite Ribeiro et al. (R34),
aborde deux cas intéressants de surélévation de barrages-voûtes à double courbure.
Étant indiqué que la section 5.1 est spécifique aux innovations pour l’élévation de
barrages existants en BCV, nous avons inclus la discussion sur ce document dans
cette section, à l’exception de la partie liée à la construction, qui est incluse dans la
section 3.2.
Bjønness et al. (R42, Norvège) présentent un article décrivant le proces-

sus général impliqué dans l’utilisation des données de l’instrumentation afin de
vérifier le comportement d’un barrage-voûte en béton existant. Les données de
l’instrumentation sont utilisées comme base pour la définition de la charge et comme
étalonnage d’un modèle aux éléments finis. Il a été constaté que le barrage était
fortement influencé par la charge de température. Lors de la vidange et du remplis-
sage du réservoir, le changement de réponse structurelle à la charge hydrostatique
était relativement faible par rapport au changement causé par une variation de tem-
pérature. Par conséquent, la charge de température est très importante lors de
l’évaluation de la capacité des barrages-voûtes. Une analyse thermodynamique
transitoire, à l’aide d’Abaqus, a été effectuée, et le programme ShellDesign a été
utilisé afin d’analyser le comportement non linéaire du barrage. La température
résultante obtenue à partir d’Abaqus a été utilisée comme charge pour l’analyse
structurelle non linéaire dans ShellDesign. Le modèle non linéaire aux éléments
finis a été étalonné à partir des déformations enregistrées par l’instrumentation,
pour différentes périodes. La capacité de la voûte a été vérifiée conformément à
l’Eurocode 2. Mais, en raison de l’absence de norme spécifique pour les barrages
dans le système Eurocode, la capacité a également été vérifiée à l’aide de la “ théorie
du champ de compression modifié ” (MCFT) développée à l’Université de Toronto.
Les deux méthodes ont donné des résultats similaires, indiquant que la résistance
structurelle de la voûte ne répondait pas aux exigences.
697
GR Q. 104
3.2. CONSTRUCTION D’UN BARRAGE-VOÛTE EN BCV
Le rapport français R18 (Grimal et al.), traitant de pathologies du béton, a

déjà été discuté à la section 3.1 en ce qui concerne l’analyse des barrages. Bien
que le rapport ne soit pas vraiment lié à la construction de nouveaux barrages en
béton conventionnel, nous l’avons considéré comme intéressant pour cette section,
puisqu’une partie du rapport traite des grands travaux de réhabilitation entrepris
au barrage de Bimont de 2017 à 2019. Les modifications ont été basées sur la
contribution combinée de deux campagnes d’investigation (carottage et inspection
par caméra TV des forages) et sur le résultat de la modélisation numérique 3D.
Brièvement décrits, la portée des travaux de réhabilitation était la suivante:
• Ancrages passifs positionnés pour renforcer la structure avant l’injection des

fissures.
• Forages d’injection croisant les fissures de “ feuilletage ”.
• Forages coupant les joints ouverts à un angle permettant de claveter la
structure après les travaux de scellement des fissures.
D’autres travaux liés à la problématique de gonflement étaient:
• Installation d’une géomembrane d’étanchéité CARPI sur le parement amont

en rive droite, afin de limiter l’introduction d’eau et d’éviter une reprise du
gonflement.
• Installation de tirants verticaux précontraints afin de stabiliser la culée de la
rive droite soumise aux effets des charges hydrostatiques, thermiques d’été
et de gonflement.
Après les travaux de réhabilitation, la remise en eau du réservoir qui avait été
vidé pour les travaux a commencé au printemps 2019 et a été réalisée par étapes,
avec un suivi attentif du comportement du barrage tout au long du processus de
remplissage de la retenue.
Le rapport R34, de Suisse (Leite Ribeiro et al.) est examiné à la section 5.1,
à l’exception de la partie construction, qui relève de cette section. Le rapport décrit
le contexte, la conception et la construction de deux cas d’élévation de barrage-
voûte: le barrage de Cambambe en Angola, surélevé de 20 m entre 2012 et 2018,
à une nouvelle hauteur de 92 m, et le barrage de Vieux-Émosson dans les Alpes
suisses, surélevé de 21,5 m entre 2012 et 2015, à une nouvelle hauteur de 76,5 m.
Dans les deux cas, le rapport explique en détail les travaux préparatoires effec-
tués, y compris la démolition de parties importantes de la partie supérieure des
structures existantes. Dans le cas du barrage du Vieux-Émosson, le forage et le
dynamitage ont été abandonnés (en fait, un dynamitage léger et contrôlé) et rem-
placés par l’utilisation de deux fraiseuses de chaussée. Un autre point délicat a été
la préparation du contact entre l’ancien et le nouveau béton. Au barrage de Cam-
bambe, plusieurs mesures ont été prises afin de garantir un contact adéquat, et de
698
GR Q. 104
surveiller le comportement du contact, notamment grâce à une galerie d’inspection

et de drainage. Au barrage de Vieux-Émosson, toutes les surfaces de contact ont
été traitées par hydro-décapage, et le nouveau béton en contact direct avec l’ancien
béton était un mélange de béton spécial avec de petits granulats (0-32 mm au lieu
de 0-63 mm). De plus, des barres d’acier ont été ancrées dans l’ancien béton pour
renforcer la liaison entre les deux bétons. Une galerie d’inspection a été aménagée
au-dessus de l’interface entre les deux bétons. D’autres aspects importants qui ont
grandement affecté la construction ont été le détournement de la rivière, dans le cas
de Cambambe, et les conditions météorologiques dans le cas de Vieux-Émosson.
Enfin, les particularités du jointoiement dans les deux barrages sont exposées en
détail.
3.3. ANALYSE ET CONCEPTION D’UN BARRAGE-VOÛTE EN BCR
Le rapport R16 reçu de France (Si-Chaib et al.) discute dans sa première

partie des possibilités et des avantages offerts par les analyses numériques pour
aider à la conception de barrages-voûtes épais en BCR et pour comprendre le
comportement thermomécanique à long terme des barrages en général. Le bar-
rage libanais de Janneh de 157 m de haut, le deuxième plus haut barrage-voûte
en BCR au monde, a été utilisé comme cas d’étude. La deuxième partie du rap-
port se concentre sur une comparaison croisée entre les modèles numériques et
la construction, basé sur l’expérience de deux barrages-voûtes épais en BCR, le
barrage de Tabellout de 121 m de haut en Algérie, déjà construit, et un barrage de
74 m de haut qui est actuellement en construction en Asie du Sud-Est, tous deux
dans des contextes d’activité sismique intense et de conditions géographiques rel-
ativement difficiles. Les avantages et les limites des modèles numériques 3D non
linéaires sont présentés et discutés, ainsi que la validation rétrospective des pro-
priétés mécaniques adoptées, et les dispositions de construction choisies telles que
le système de clavage et la ceinture parasismique sont également discutées.
Un deuxième rapport reçu de la France à ce sujet (R17, Andrian et al.) traite

également de la construction de barrages-voûtes en BCR (section 3.4), de la con-
ception et de la construction des barrages-poids en BCR (sections 4.3 et 4.5) et des
barrages voûtés en BCR très élevés (section 6.2). Afin de ne pas diviser la présen-
tation du rapport en plusieurs sections, en gardant à l’esprit qu’il comprend une
comparaison de plusieurs études de cas, nous discuterons de R17 uniquement dans
cette section. La première partie du rapport aborde les concepts généraux concer-
nant le large éventail possible de contenu cimenté dans les mélanges. Le mélange
peut être conçu pour fournir à la fois la résistance structurelle et l’étanchéité, ou
seulement la résistance structurelle requise, l’étanchéité pouvant être assurée par
un parement extérieur. Dans le rapport, trois études de cas sont présentées, le
barrage-voûte en BCR de Janneh au Liban et deux barrages-poids en BCR, Nam
Ngum 4 au Laos (h = 74 m) et Hatta à Dubaï (h = 70 m). Certaines des caractéris-
tiques les plus importantes à souligner sont les suivantes. Le mélange de béton pour
les trois barrages implique un BCR avec une forte teneur en ciment et hautement
699
GR Q. 104
maniable, avec une consistance Vébé de 15 à 18 secondes. Pour le barrage de

Janneh, la bande de 10 m près du parement amont utilise un mélange de BCR plus
riche. La zone amont a (en kg/m3) 77 de ciment + 143 de laitier de haut fourneau
granulé broyé (GGBFS), pour un total de 220, tandis que la zone en aval a 56 de
ciment + 104 de GGBFS, pour un total de 160. Les mélanges respectifs pour Nam
Ngum 4 et Hatta incorporent 82 de ciment + 100 de cendre volante pour un total
de 182, et 50 de ciment + 110 de cendre volante pour un total de 160. Les deux
mélanges ci-dessus avec une teneur totale en ciment de 160 kg/m3 sont proches
de la limite inférieure pour être considérés comme des mélanges de BCR à haute
teneur en ciment. Pour Janneh et Hatta, un mélange retardateur a été incorporé
au BCR. Comme caractéristique importante, à Janneh, les joints transversaux ne
seront injectés qu’en cas de fuite élevée afin de les réduire, en tirant parti du faible
retrait et du fluage à long terme attendus pour le BCR même avec l’utilisation de
laitier de haut fourneau au lieu de cendres volantes à faible teneur en chaux, qui a
été testé lors de l’essai à grande échelle (FST). Les carottes prélevées dans le FST
à Nam Ngum 4 ont montré que les agrégats du BCR sont plutôt en contact les uns
avec les autres par rapport à ceux du BCR enrichi en coulis de parement (GEBCR).
On pense que la disposition du squelette agrégé est un facteur contribuant au faible
retrait / fluage à long terme du BCR à haute maniabilité. L’analyse thermomécanique
a suggéré un espacement des joints transversaux de 24 m, qui s’est avéré être de
30 m pour le barrage de Hatta, où le concept original d’une zone en amont avec
un mortier de litière systématique et une zone de BCR moyennement cimentaire en
aval, a été remplacé par une zone BCR à haute cimentation unique. La limite pour le
joint chaud a été établie à Janneh et Hatta comme temps de réglage initial du BCR,
et la limite pour le joint chaud basée sur un facteur de maturité modifié (MMF) de
1000◦ C × h. La limite pour le joint chaud à Nam Ngum 4 était le temps de réglage
initial du BCR moins 1 heure, et la limite pour le joint chaud était stipulée dans un
tableau avec différentes limites de temps exprimés en heures pour les différentes
saisons. L’article explique également les différentes méthodes de placement BCR
appliquées aux trois barrages et aux coffrages utilisés. Dans la dernière partie du
rapport, des informations précieuses sont données concernant la construction de
plusieurs grands barrages BCR: Souapiti (Guinée, h = 115 m), Tha Dan (Thaïlande,
h = 92 m), Gibe III (Éthiopie, h = 250 m), GERDP (Éthiopie, h = 170 m), Bui (Ghana
h = 108 m) et Theun Hinboun (Laos, h = 70 m), en ce qui concerne les méthodes de
placement, Taux de production et de placement du BCR (jusqu’à un taux mensuel
moyen de 180 000 m3 et un taux mensuel maximal de 330 000 m3 au GERDP),
transport du BCR, utilisation de radars pluviométriques pour aider à maximiser le
placement du BCR et gestion des inondations pendant la construction permettant
de déborder la partie de la structure déjà construite.
Enfin, le rapport rédigé par Shaw (R23, Afrique du Sud) s’inscrit dans cette
section et dans la section 4.3. L’auteur y décrit les différences entre le comporte-
ment du BCR et du BCV aux jeunes âges dans les grands barrages et comment
ces différences impliquent une approche de conception légèrement différente, en
tenant compte d’une propriété supplémentaire du béton sous la forme d’un fluage
contrainte-relaxation (SRC). Il est souligné la différence entre les approches ver-
ticale et horizontale pour la construction de barrages respectivement en BCV et
700
GR Q. 104
BCR, qui, avec l’utilisation de niveaux élevés de matériaux cimentaires supplémen-

taires typiques dans les mélanges BCR, introduisent deux facteurs clés qui peuvent
ensemble créer un comportement initial significativement différent et donc des exi-
gences de conception. Tout d’abord, les niveaux de retrait du béton subis pendant
le processus d’hydratation peuvent être considérablement inférieurs à ce que l’on
suppose traditionnellement pour le BCV lorsque des niveaux élevés de cendres
volantes sont utilisés. De plus, le placement horizontal du béton implique que la
masse du barrage ne sera pas nécessairement transmise directement ou verticale-
ment dans la fondation, en particulier lorsque les niveaux de retrait du béton sont
faibles. Par conséquent, la conception historique typique basée sur l’analyse 2D des
barrages-poids en BCV peut ne pas être complètement valide pour les barrages-
poids en BCR. Cela est plus important lorsque le barrage en BCR a un rapport
longueur/hauteur inférieur (< 6), et dans ce scénario, la connaissance du niveau de
fluage est essentielle pour développer une compréhension significative de la fonc-
tion structurelle 3D. Au contraire, lorsque le barrage-poids est long par rapport à sa
hauteur et que les blocs adjacents sont de hauteur similaire, le niveau de fluage
aura peu d’importance et le barrage pourra être conçu, ou analysé, comme une
structure 2D. Dans le cas des barrages-voûtes en BCR, un mouvement en amont
pendant la construction a été mesuré, ce qui équivaut effectivement à un allonge-
ment de la crête du barrage de 30 à 50 microdéformations sur une structure de 70 à
100 m de hauteur, générant une compression horizontale le long de la structure, ce
qui a un impact important sur la répartition de la charge gravitationnelle sur la fon-
dation du barrage et à l’intérieur de la structure elle-même. L’auteur affirme qu’une
compréhension complète de la fonction structurelle d’une voûte en BCR, d’un bar-
rage poids-voûte ou d’un barrage-poids dans une vallée plus étroite construite en
utilisant un BCR à faible fluage ne peut être développée que par une analyse ther-
momécanique appropriée qui prend connaissance de la variation réelle du fluage
en fonction de la contrainte de compression.
3.4. CONSTRUCTION D’UN BARRAGE-VOÛTE EN BCR
Le rapport français R17 (Andrian et al.) est le seul qui traite du sujet de la
construction du barrage-voûte mais BCR, en se référant spécifiquement au barrage
de Janneh au Liban, mais a déjà été présenté dans la section 3. 3.
4. INNOVATIONS POUR L’ANALYSE DE BARRAGES-POIDS EN BÉTON, LA

CONCEPTION DES MATÉRIAUX ET LA CONSTRUCTION, Y COMPRIS
EN BCR ET DE TYPE CMD
Peu de temps après sa première utilisation au début des années 1980, le BCR
a remplacé le BCV comme méthode standard pour la construction de barrages-poids
en béton.
701
GR Q. 104
À ce jour, le nombre total de barrages-poids en BCR construits ou en con-

struction (à l’exclusion des barrages en remblai dur), d’une hauteur supérieure à
15 m, s’élève à 761, dont 121 (soit 16 %) dépassent 100 m et 4 dépassent 200 m
(Dunstan, 2020).
Le tableau 6 montre les dix barrages en BCR les plus élevés achevés ou en
construction dans le monde, neuf de ces dix étant des barrages-poids.
Tableau 6
Plus hauts barrages en BCR (Dunstan, 2020)
NOM DU TYPE DE ANNÉE
BARRAGE PAYS HAUTEUR (m) BARRAGE D’ACHÈVEMENT
Gibe III Éthiopie 246 Poids 2016
Longtan Chine 217 Poids 2009
Huangdeng Chine 203 Poids 2019
Guangzhao Chine 201 Poids 2009
Miel 1 Colombie 188 Poids 2002
Ayvali Turquie 177 Poids 2015
Wanjiakouzi Chine 168 Voûte mince 2011
Guandi Chine 168 Poids 2013
Nam Ngiep 1 Laos 167 Poids 2019
Lushi Chine 165 Poids 2018
On constate que six des dix plus hauts barrages en BCR ont été construits
en Chine, et que sept des dix ont été achevés au cours de la dernière décennie,
et cela suggère qu’il y a une confiance croissante pour construire des barrages en
BCR de plus en plus élevés.
Malgré l’augmentation de la construction de barrages-poids en BCR au cours

des dernières décennies, au détriment des barrages-poids en BCV, les rapports
relevant de cette section 4 concernent le BCR pour 14 d’entre eux et 10 pour le
BCV (l’un est en fait un barrage de contreforts), et un seul rapport pour un ouvrage
en matériau cimenté (CMD).
4.1. ANALYSE ET CONCEPTION D’UN BARRAGE-POIDS EN BCV
Le premier des deux rapports reçus de l’Iran (Noorzad et al., R1) indique
que l’analyse sismique des barrages-poids en béton est principalement effectuée
sous l’hypothèse d’une fondation élastique, tout en considérant qu’une réponse non
linéaire de la roche de fondation a été rarement abordée. Cependant, il est observé
que l’interface barrage-fondation est très sensible à la fissuration en raison de ses
propriétés géométriques, alors que l’hypothèse d’une réponse élastique pour la fon-
dation empêche la propagation des fissures dans ce milieu et force la fissure à se
propager parallèlement à l’interface. C’est pourquoi l’article analyse un système de
barrage-fondation-réservoir en tenant compte des réponses non linéaires du bar-
rage et de la fondation, soumis à une charge sismique. Une analyse comparative
702
GR Q. 104
est également effectuée avec la réponse linéaire de la roche de fondation. Comme

exemple d’application, le barrage-poids en béton de Koyna a été choisi. Les résultats
révèlent un motif de fissure complètement différent au pied du barrage, selon que
la fondation est considérée comme élastique ou avec des propriétés non linéaires.
Dans le premier cas, la fissure est forcée de se propager le long de l’interface
barrage-fondation, tandis que dans le second cas, la fissure se propage dans la
roche de fondation et une réponse globale différente pour l’ensemble du système
est obtenue.
Le deuxième rapport reçu des mêmes auteurs, Noorzad et al. (R2, Iran), en
plus des barrages-poids en BCV, s’applique également aux barrages-voûtes en BCV
a donc été précédemment présenté à la section 3.1.
Le rapport R12, de la République tchèque (Hrazdira & Švancara), présente un

cas intéressant d’augmentation de la capacité d’évacuation d’un barrage-poids en
béton existant. Le barrage d’Orlík (h = 91 m), crée le plus grand réservoir de volume
en République tchèque, 716,5 Mm3 , et 68 km de long. Le barrage a été construit
entre 1957 et 1960 en tant que composant principal d’un projet hydroélectrique,
mais il sert également à d’autres usages, notamment le contrôle des inondations,
les loisirs et le soutien d’étiage de la rivière Vltava. Il y a quelques années, des régle-
mentations beaucoup plus strictes pour l’évaluation de la sécurité des structures de
barrages pendant les inondations ont été adoptées par les autorités tchèques, et
un programme d’amélioration de la sécurité a été lancé sur la majorité des bar-
rages tchèques. Appartenant à la catégorie I, une inondation de 10 000 ans s’est
avérée applicable au barrage d’Orlík, avec un débit maximal de 5 300 m3 / s. En
août 2002, le barrage a connu la plus grande crue de l’histoire du barrage. Le débit
sortant a atteint 3100 m3 / s et le débit maximal a sensiblement dépassé la valeur
de crue de 100 ans. Le niveau d’eau dans le réservoir dépassait le niveau de con-
ception et, bien que le barrage lui-même ait parfaitement résisté, l’aménagement a
subi des dommages considérables, principalement en ce qui concerne les ouvrages
annexes. Un programme d’analyses et de travaux préparatoires pour la future réno-
vation du barrage a été lancé sur la base de l’évaluation de la crue de 2002. Un
modèle physique du barrage et des évacuateurs a été réalisé en place en 2009 pour
évaluer les paramètres du débit et les extrapoler aux niveaux d’eau correspondant
aux crues extrêmes, en particulier celle de 10 000 ans de période de retour. Des cal-
culs complets aux éléments finis ont été effectués en 2013-2015 sur la base d’une
étude géologique technique précédente, en particulier pour le contact entre la base
du barrage et le substrat rocheux. Une étude analysant plusieurs options possibles
pour améliorer la sécurité du barrage lors de crues extrêmes a été réalisée en 2010,
puis évaluée dans l’étude de faisabilité développée en 2013, en ce qui concerne la
faisabilité technique, les coûts, les aspects opérationnels les aspects de sécurité,
ainsi que les impacts sur le projet pendant la phase de construction. L’alternative
finalement retenue consistait en 3 nouveaux évacuateurs vannés à l’extérieur du
barrage, à côté de la culée droite. Des études techniques et géologiques détaillées,
des levés géodésiques et des inspections des structures en béton ont alors été
effectués. La nouvelle conception des évacuateurs a été affinée par modélisation
3D pour évaluer et optimiser les débits, puis vérifiée et optimisée en utilisant un
modèle physique. La capacité combinée s’élève à 3 × 588 m3 /s. Les travaux
703
GR Q. 104
de construction devraient débuter à la mi-2021 et seront gérés sur la base de la

méthodologie BIM.
Le rapport R19, rédigé par Rashek (Canada), traite de l’analyse sismique des
structures annexes aux barrages, comme les évacuateurs, les prises d’eau, les per-
tuis, les conduites forcées, les vannes, les bâtiments de contrôle et leur équipement
associé. Une défaillance de l’une de ces structures pourrait entraîner une perte de
contrôle du réservoir après un tremblement de terre. Mais la charge sismique pour
un ouvrage annexe est différente du risque sismique en champ libre utilisé dans
l’analyse des barrages. Les caractéristiques et l’intensité de cette charge dépen-
dent de l’emplacement et des conditions de transfert. En effet, les accélérations
de pointe au sommet de certains barrages peuvent être amplifiées jusqu’à dix fois
les valeurs observées à sa base. Le rapport explique les étapes pour effectuer
l’évaluation sismique d’un barrage et de ses ouvrages annexes en se basant sur
les résultats de l’analyse. Il définit l’analyse sismique du barrage et des ouvrages
annexes à trois niveaux de sophistication. Ensuite, il discute de l’analyse du spectre
de réponse (RSA) de l’ouvrage annexe et de la façon de développer les spectres
de projet, qui peuvent être utilisés pour concevoir et évaluer les structures annexes,
sélectionner l’équipement pour un niveau prescrit de séisme ou déterminer si un
ensemble de composants doit être isolé afin d’améliorer leur capacité à résister
aux vibrations imposées. Étant donné que ce niveau avancé d’analyse peut être
prohibitif pour une évaluation rapide de l’ouvrage annexe, l’auteur fournit une solu-
tion approximative qui peut être utilisée pour des évaluations sismiques rapides.
La solution approximative proposée est basée sur la modification de la procédure
RSA classique pour tenir compte de la résonance possible de l’ouvrage annexe
aux modes primaires de vibration du barrage. Pour cela, les résultats de l’analyse
modale pour le système barrage-fondation-réservoir sont obtenus en utilisant un
modèle aux éléments finis, complet ou simplifié du barrage. La solution simplifiée
est destinée à être utilisée pour la conception préliminaire, la sensibilité et les études
paramétriques, et, comme outil de vérification, pour une analyse plus rigoureuse
de l’historique temporel. Dans la dernière partie du rapport, la procédure ci-dessus
est établie pour le cas d’une structure de déversoir fermé. Un modèle aux éléments
finis du barrage (y compris la fondation, le déversoir, les piliers et le bloc de touril-
lons) a été créé à l’aide du logiciel ANSYS, et l’analyse a été effectuée pour estimer
le spectre d’intervention approximatif dans la structure au niveau du tourillon d’un
déversoir vanné.
Malm et al. (R20, Suède) présentent un rapport qui aurait pu être décrit dans
cette section, mais il se rattache aussi à la section 7.2 et il sera donc évoqué dans
cette dernière section.
De même, le rapport R25 reçu de la Bulgarie (Tzenkov et al.) est relève des
deux mêmes sections, mais dans ce cas, il est pris en compte ici. La première partie
du rapport expose la détérioration au cours des années d’exploitation du barrage-
poids en béton de Beli Iskar, haut de 50,7 m, situé à une altitude supérieure à
1800 m d’altitude dans la montagne de Rila, une région aux conditions hivernales
très difficiles. Les principaux problèmes rencontrés comprennent: des fissures de
construction et des trous (en raison d’un mauvais compactage, à la main) permettant
704
GR Q. 104
à l’eau de s’infiltrer dans le corps du barrage de plus en plus profondément au fil des
ans, des dommages aux joints entre plots, des fuites nombreuses dans la galerie
d’inspection avec des dépôts de calcite, des dommages (corrosion) à la structure
en béton armé de la chambre de vanne et, surtout, le gonflement par alcali-réaction
et le vieillissement du béton. Ce dernier point a été confirmé par des carottages du
béton réalisés au cours de quatre campagnes différentes entre 1961 et 2013, et des
échantillons d’âges différents ont donc été examinés. Les forages de drainage dans
la fondation du barrage étaient inopérants et de nouveaux forages ont été réalisés
trois fois. Le drainage vertical dans le corps du barrage ne fonctionne plus depuis
2002, et cela a conduit au développement d’un champ d’infiltration inadmissible à
l’intérieur de la structure du barrage et affecte en outre le champ de contraintes dans
le corps du barrage et sa stabilité. En plus des raisons purement physiques de la
réhabilitation urgente du barrage, il existe une autre raison importante de procéder
à une rénovation structurelle du barrage. Conformément à la réglementation en
vigueur en Bulgarie concernant la prise en compte d’un cas de charge sismique
(prescription inexistante au moment de la construction), la stabilité du barrage n’est
pas garantie en cas de fort séisme. Par conséquent, des travaux de réhabilitation
ont été nécessaires. Dans une première phase, le réservoir a été vidé en 2002,
une géomembrane d’étanchéité en PVC a été installée sur le parement amont et
la zone de contact avec la fondation rocheuse a été injectée. Dans une deuxième
phase, le projet de réhabilitation structurelle a été mis en appel d’offres et attribué.
La solution finalement choisie consistait en un contrefort massif en béton placé
contre le parement aval du barrage existant. À partir d’une étude sismologique, il a
été déterminé une accélération maximale du sol (PGA) égale à 0,305 g pour une
période de retour de 1000 ans et à 0,564 g pour une période de retour de 10 000 ans.
La vérification sismique de la structure actuelle du barrage a été réalisée à l’aide
d’un modèle aux éléments finis 2D pour le système barrage-fondation-réservoir à
l’aide du programme informatique DIANA, avec l’hypothèse initiale du comportement
linéaire-élastique du barrage et de sa fondation. La géométrie du modèle a été
définie pour le plus haut plot du barrage. Pour le séisme d’évaluation (SEE), les
résultats montrent des contraintes dynamiques de traction qui dépassent la limite
admissible pour les joints de construction et pour la masse de béton. Dans un
deuxième temps, la même analyse pour le barrage renforcé a été effectuée selon la
même approche et, dans le cas du SEE, les résultats montrent que les contraintes
de compression et de traction sont à des niveaux inférieurs par rapport au modèle de
barrage original. Une exception est au pied du barrage, où une certaine fissuration
locale peut se produire, mais pour une excitation sismique d’un tel niveau d’intensité
(SEE), ce type de dommage serait acceptable puisque la stabilité globale du barrage
n’est pas menacée. Ainsi, l’efficacité de la solution de renforcement particulière
étudiée a été prouvée.
Le rapport R39 (Sidik et al., Indonésie) traite de la conception et de la con-

struction d’un système modulaire en béton préfabriqué pour les parois latérales, et
il est traité à la section 4.2.
Un rapport reçu de l’Inde (Chaubey et al., R43), décrit la surélévation de 5

m effectuée sur un barrage-poids en BCV de 32 m de haut, le barrage Dhanikhari
705
GR Q. 104
sur les Îles Andaman et Nicobar. Le rapport fournit des détails sur la conception
de la surélévation consistant en un contrefort en béton du côté aval, à partir de la
réalisation de différents calculs aux éléments finis, à savoir une analyse thermique,
une analyse de contrainte et une analyse de contrainte de retrait. Un facteur qui
a grandement influencé la conception et la construction était le fait que le système
d’approvisionnement en eau assuré par le barrage ne pouvait pas être sérieusement
perturbé. En conséquence, comme la situation idéale d’un réservoir complètement
vide n’était pas possible, le béton a été mis en œuvre pendant les mois où le niveau
du réservoir était plus bas afin de réduire autant que possible la charge du réservoir
sur le barrage, afin que la nouvelle structure composite réagisse à une plus grande
plage de niveau du réservoir.
Le dernier rapport reçu au sujet de ce sujet, également de l’Inde (Bansal et al.,

R45), traite de la conception hydraulique d’une structure de déversoir en béton armé
de 18,5 m de haut, équipée de trois vannes segment. Le projet est situé sur le cours
supérieur d’une rivière himalayenne avec un lit comprenant de gros blocs rocheux
et des couches perméables profondes, des conditions qui ne conviennent pas aux
grands barrages en béton. Ce barrage (barrage mobile ou barrage à faible chute) est
la composante la plus en amont du projet hydroélectrique Naitwar Mori (NMHEP)
de 60 MW, dont la mise en service est prévue en 2022. La conception s’est heurtée
à l’obstacle que les guides existantes fournis par le Bureau des normes indiennes
pour la planification et la conception des barrages ne s’appliquent pas aux structures
dans les tronçons de rivière à dominante rocheuse. Pour la conception hydraulique
finale, plusieurs modes de défaillance potentiels ont été pris en compte, y compris
la défaillance de la fondation due à l’affouillement, l’érosion interne en fondation en
raison de gradients hydrauliques élevés, le gradient élevé au pied et le risque de
suffusion, sur la base de différentes théories reconnues, en l’absence de directives
éprouvées. La conception a été vérifiée par analyse numérique des écoulements
à l’aide du logiciel Slide pour obtenir les diagrammes de flux net et de gradient de
sortie. Le rapport fournit des détails sur le niveau de la crête du barrage, qui a été
maintenu au niveau du lit de la rivière, ainsi que sur les travaux de protection en
amont et en aval et le rideau de coulis. Plusieurs petits projets hydroélectriques
sont prévus sur les affluents des principaux fleuves himalayens, impliquant des
conditions de fondation difficiles similaires à celles expliquées ci-dessus et, pour
cette raison, l’expérience acquise après l’exécution de la structure de barrage du
NMHEP est de la plus haute importance.
4.2. CONSTRUCTION D’UN BARRAGE-POIDS EN BCV
Le premier des deux rapports reçus sur le thème de la construction de

barrages-poids en BCV, R21 (Yamashita et al., Japon) présente le concept de
“ système de coulage automatique de béton de barrage ” dont les principaux objec-
tifs sont d’améliorer la productivité et de réduire les besoins en personnel qualifié
en appliquant les technologies de l’information. Le rapport indique que la mise en
place de béton dans les barrages-poids en béton représente environ 60% du coût
706
GR Q. 104
de construction et environ 50% du temps de construction. Les opérations de pro-

duction, de transport et de placement du béton sont répétées pendant plusieurs
années et nécessitent de nombreux travailleurs qualifiés expérimentés pour la
manutention de grosses machines et équipements lourds. En particulier, lorsque
des grues à câble sont adoptées pour le transport du béton, la manipulation de
la grue devient une opération 3D complexe, de sorte que plus d’expérience et de
compétences que d’habitude sont nécessaires. Le nouveau système intègre la tech-
nologie d’automatisation qui a déjà été appliquée à la centrale à béton et transposé
à l’automatisation de la grue à câble. Le système a été testé lors de la mise en place
d’environ 40 000 m3 sur un total de 213 500 m3 de volume de béton à l’aide de la
grue du barrage de Yanagawa (h = 77 m). Les résultats obtenus ont été les suivants:
premièrement, le temps de cycle par mise en place a été réduit d’environ 10% (de
3 min 40 sec à 3 min 20 sec); deuxièmement, le nombre de travailleurs requis pour
le coulage dans le barrage a été réduit de 33% (de 15 à 10); et troisièmement, par
l’automatisation du fonctionnement des grues avec cette technologie, l’opération
peut être effectuée avec la même efficacité de travail qu’un opérateur hautement
qualifié. Pour l’avenir, l’intention est d’augmenter le nombre d’applications de la
technologie (autres types de grues, etc.) et augmenter la fiabilité et la polyvalence
du système.
Dans le deuxième rapport, reçu de l’Indonésie (Sidik et al., R39), la mise en

œuvre prévue d’un système modulaire en béton préfabriqué sur les parois latérales
du canal en aval du déversoir du barrage de Margatiga est expliquée. Le barrage
est une structure de type composite, avec une combinaison d’un barrage-poids en
béton et d’un barrage zoné en enrochement à noyau vertical. L’un des principaux
objectifs de l’utilisation de ce système était d’accélérer le calendrier de construction.
Les avantages supplémentaires, par rapport aux méthodes conventionnelles, sont:
la constructibilité (facilité de mise en œuvre, pas de coffrage ni d’échafaudage req-
uis), moins de besoin de main-d’œuvre, un contrôle qualité simplifié (travail répétitif
effectué dans une usine) et une amélioration de la qualité du produit. Le système
nécessite une planification précise car le niveau de flexibilité de cette méthode
est faible par rapport au système conventionnel. Le transport et l’installation des
modules en béton préfabriqué doivent être soigneusement examinés et pourraient
être un facteur limitant pour les dimensions ou le poids de ceux-ci. Pour les parois
latérales de ce projet, 11 modules différents ont été conçus, dont 5 pour la dalle
de fondation de 3 m de haut (12 pièces au total), et 6 pour la paroi trapézoïdale de
9 m de haut (16 pièces). Les modules préfabriqués ont été conçus de telle sorte
que, une fois placés, ils laissaient un volume important de la section transversale
du mur à remplir de béton in situ, mais ne nécessitant pas de coffrage, et ayant des
barres d’armature dépassant de l’élément préfabriqué pour être enveloppé avec
le béton coulé en place. Le poids du module le plus lourd était de 5,45 t et les
travaux devaient utiliser une grue sur chenilles de 50 t pour installer les modules qui
étaient reliés les uns aux autres avec des boulons d’ancrage de 30 mm de diamètre,
puis injectés. Dans ce projet particulier, par rapport à l’utilisation de méthodes con-
ventionnelles, une réduction du calendrier de 5 à 4 mois a été envisagée lors de
la planification, et une réduction de la main-d’œuvre de 100 à 10 travailleurs par
rapport à la construction des murs sur le site.
707
GR Q. 104
4.3. ANALYSE ET CONCEPTION D’UN BARRAGE-POIDS EN BCR
Un premier rapport reçu pour ce sujet, R7 d’Italie (Pietrangeli et al.), aborde

divers aspects liés au projet grand barrage éthiopien de la Renaissance (GERD),
y compris l’analyse comportementale thermique et sismique, mais en raison de la
grande taille et de la hauteur de ce barrage, nous avons inclus la discussion de ce
rapport dans la section 6.3.
Le rapport suivant relevant de cette section vient d’Espagne (G. de Membrillera

et al., R9) ; il comprend un aperçu général de plusieurs modèles numériques utilisés
dans la conception du barrage en BCR de Casupá en Uruguay, ainsi que quelques
détails sur l’approche de conception, ainsi que sur la fondation du barrage et les
carrières potentielles. Casupá est un barrage-poids en BCR de 31,5 m de haut
qui sera construit dans le but d’assurer l’approvisionnement en eau de la ville de
Montevideo et de sa région métropolitaine. Le barrage aura une longueur de crête de
750 m, avec des joints transversaux espacés à 20 m, un parement amont vertical et
un parement aval en escalier avec une pente équivalente de 0,8H:1V. Les modèles
numériques énumérés dans le rapport sont les suivants:
• Modèle thermique transitoire 2D qui évalue les températures maximales

atteintes pendant la phase de construction, ainsi que les gradients thermiques
minimaux et maximaux attendus dans le béton, à la fois pour des conditions
à court et à long terme.
• Modèle mécanique-structurel statique 2D de la section transversale du bar-
rage, à partir duquel les déformations et contraintes attendues dans les
différents scénarios de charge sont obtenues, ainsi que les coefficients de
sécurité contre le glissement du barrage et les facteurs de sécurité pour
des contraintes maximales dans le béton. Ce modèle comprend plusieurs
analyses de sensibilité tenant compte des paramètres de résistance les
plus pertinents pour la stabilité du barrage à des sections transversales
représentatives.
• Un modèle mécanique et structurel 3D statique, pour vérifier les hypothèses
les plus critiques qui ont été trouvées avec le modèle 2D, a été jugé nécessaire
pour la zone centrale du barrage, où se trouvent des sections transversales
singulières, telles que le déversoir, les pertuis et la prise d’eau.
• Modèle géotechnique 2D développé pour la fondation du barrage, en ten-
ant compte de la section transversale typique. Avec ce modèle, le débit
et les écoulements dans la fondation ont été estimés pour plusieurs scé-
narios de chargement, en conjonction avec les calculs de stabilité dans la
fondation.
Le rapport R17, de France (Andrian et al.), a une partie qui traite du sujet de
cette section en exposant plusieurs études de cas, mais a été entièrement abordé
dans la section 3.3.
708
GR Q. 104
De même, le rapport R22, du Japon (Aosaka et al.) a une partie qui correspond
à la présente section, mais il sera intégralement traité dans la section 6.3.
Le rapport R23, rédigé par Shaw (Afrique du Sud), a déjà été discuté à la
section 3.3, car il développe une caractéristique particulière du BCR, à savoir le
fluage sous contraigne, à prendre en compte dans la conception non seulement pour
les barrages-poids en BCR, mais aussi pour les barrages-voûtes et poids-voûtes.
Enfin, le rapport R40 de France (Deroo et al.), présente le concept de barrages

hybrides BCR-enrochement. Le rapport évalue l’opportunité de combiner les deux
typologies de barrages pour construire des barrages composites économiques et
sûrs en se basant sur le retour de l’expérience de structures similaires, sur le calcul
et sur une évaluation multicritère des avantages et des inconvénients. Bien qu’il ait
été écrit en considérant des structures de quelques dizaines de mètres de haut, rien
n’empêche les concepts utilisés d’être appliqués aux barrages de plus de 100 m,
mais il semble nécessaire d’abord d’avoir un retour d’expérience sur des structures
antérieures et plus modestes. Deux options d’hybridation BCR et enrochements
sont présentées, mais toutes les autres options sont possibles et peuvent être mieux
adaptées aux circonstances de chaque site et de chaque projet. La première option
est appelée BCR-R, les deux parties du corps du barrage étant érigées simultané-
ment, le BCR du côté amont constituant l’élément imperméable et le parement aval
du BCR non coffrée. Le rapport, les calculs sont présentés avec une pente en
aval du BCR de 0,4H:1V, mais des pentes plus raides peuvent être envisagées.
Le zonage du remblai en aval peut être envisagé, afin d’optimiser les coûts et le
comportement mécanique, notamment à l’interface. Toute la gamme de BCR est
possible, du “BCR à haut dosage en ciment” au “remblai dur”, et la géométrie doit,
bien sûr, être adaptée en conséquence. Une géomembrane peut être envisagée
pour le parement amont pour résoudre le problème de fissuration et d’infiltration.
La deuxième option est appelée R-BCR, également avec les deux parties du corps
du barrage érigées simultanément, mais avec le BCR du côté aval. L’étanchéité est
assurée par un parement amont de l’enrochement (généralement en béton armé,
ou par une géomembrane). Le zonage du remblai est nécessaire, comme pour les
barrages en enrochements à masque amont (CFRD), afin de fournir un soutien au
masque et d’assurer une perméabilité croissante de l’amont vers l’aval. Le BCR est
conçu pour être très perméable, par exemple en prévoyant des galeries de drainage.
Dans certaines conditions, un évacuateur peut être intégré. Le rapport note que la
construction de barrages mixtes n’est pas une idée nouvelle, bien que dans le passé,
les matériaux de construction n’étaient ni du BCR ni des enrochements, mais plutôt
de la maçonnerie et du remblai. Dans certains cas, des barrages en maçonnerie,
dont la stabilité était considérée comme précaire, ont été transformés en barrages
hybrides maçonnerie-digue, et certains cas d’histoire sont mentionnés. En outre, il
est dit que l’idée de combiner béton et remblai pour former de nouveaux barrages
économiques et sûrs n’est pas nouvelle: les bulletins 83 et 117 de la CIGB présen-
tent plusieurs illustrations de ce concept. Les dernières sections du rapport sont
consacrées à montrer les avantages des barrages hybrides BCR-R et R-BCR par
rapport aux types de barrages CFRD et BCR en termes d’économie, de sécurité et
709
GR Q. 104
d’empreinte environnementale et enfin, diverses simulations numériques de BCR-R

et R-BCR sont présentées et discutées.
4.4. MATÉRIAUX ET COMPOSITIONS DES BÉTONS DE BARRAGES-POIDS EN BCR
Comme il est indiqué à la section 4.3, le rapport R7 (Pietrangeli et al., Italie)

traite de divers aspects du barrage de la Renaissance, y compris les compositions
de BCR utilisés, mais en raison de la taille et de la hauteur extrêmement grandes
de ce barrage, nous avons inclus la discussion de ce rapport dans la section 6.3.
Le rapport R10, reçu d’Espagne et rédigé par Ortega, présente la dernière

innovation concernant les méthodes de réalisation des parements des barrages
en BCR, appelée “ BCR vibré ”, ou IV-BCR, et utilise comme cas d’étude le bar-
rage Enciso en Espagne (h = 103 m), l’un des plus grands barrages en BCR
d’Europe occidentale (deuxième plus haut et troisième par volume de BCR,
695 000 m3 ). L’auteur débat du fait qu’un certain nombre de techniques différentes
ont été développées au cours des 35 dernières années pour réaliser les parements
des barrages en BCR. En réalité, aucune d’entre elles ne répondait pleinement à
ce jour aux exigences de qualité et/ou de simplicité. Le mélange IV-BCR, à mi-
chemin entre le BCR à forte teneur en ciment et le BCV traditionnel très rigide
(affaissement très faible à nul) pour les barrages, a résolu avec succès la plupart
des problèmes critiques rencontrés dans les premiers barrages en BCR. Le com-
portement après la première mise en eau du barrage Enciso récemment achevé en
Espagne a confirmé une amélioration extraordinaire de la qualité par rapport aux
barrages en BCR construits précédemment. Le mélange IV-BCR est un mélange
de BCR facile à mettre en œuvre, avec une consistance nulle telle que mesurée
au cône d’affaissement, et entre 6 et 12 secondes, avec une plage cible typique
de 8 à 10 secondes, mesurée par le test Vébé. Pour les mélanges de BCR avec
des temps Vébé plus élevés, la consistance doit être modifiée en ajoutant de l’eau,
du coulis, de la pâte ou du mortier afin d’obtenir la capacité de vibration. Avec les
mélanges IV-BCR, cela n’est plus nécessaire et le même béton peut être consolidé
par l’une ou l’autre méthode, vibration traditionnelle ou compactage au rouleau. Le
rapport présente les principales caractéristiques des mélanges IV-BCR et souligne
les avantages de leur utilisation. Tout d’abord, les avantages techniques, dus au fait
que pour les parements du barrage (également au contact contre la roche au niveau
des culées, contact contre des structures en BCV précédemment construites, autour
des appareils d’auscultation qui doivent être noyés dans le béton, etc.), le même
mélange de béton peut être utilisé, comme pour le noyau adjacent du barrage, les
“ deux ” développant une résistance et des propriétés élastiques et thermiques sim-
ilaires, la méthode de mise en œuvre étant la seule différence. De plus, la liaison
entre les deux bétons est parfaite, sans interface, à condition que la mise en œuvre
respective des “deux” mélanges de béton (parement et corps du barrage) soit cor-
rectement exécutée. Deuxièmement, la simplicité et l’économie de la construction
ont été obtenues en évitant les ressources et le personnel supplémentaires pour la
production, le transport, la mise en place et le contrôle de la qualité du BCV ou du
710
GR Q. 104
coulis autrement nécessaire à la consolidation par vibration. Les premières expéri-

ences d’utilisation du IV-BCR ont été acquises aux barrages de Pirris (Costa Rica,
2009-2010) et de De Hoop (2009-2013) et Spring Grove (2012-2013) en Afrique du
Sud. Plus récemment, le barrage d’Enciso, achevé en 2018, est un autre exemple
de mise en œuvre réussie de la technique IV-BCR. Dans la dernière partie du rap-
port, les caractéristiques des mélanges testés et utilisés dans ce barrage, ainsi que
quelques détails clés de la construction sont décrits. Enfin, une brève description
du comportement du barrage pendant le remplissage du réservoir est discutée: cer-
taines fuites provenant de la fondation sont apparues, mais le comportement global
de l’IV-BCR en termes d’infiltration est parfait.
Dans leur rapport, Dunstan et al. (R13, Grande-Bretagne), abordent le pro-

gramme d’essais de composition mené pour un barrage BCR dont la construction
devrait commencer en 2021. L’objectif du rapport, ainsi qu’une explication complète
du programme de combinaison d’essais, est de mettre en évidence les avantages
de sa mise en œuvre précoce et avec une portée étendue. Le barrage sera celui
de Mwache dont la hauteur envisagée est de 84 m ; ce sera le premier barrage
en BCR au Kenya. Il est mentionné qu’une étude antérieure du barrage (par une
société d’ingénieurs-conseils sans expérience du BCR) a conclu qu’il n’y avait pas
de pouzzolanes disponibles localement au Kenya et que Mwache devrait donc être
un barrage-poids en béton traditionnel. L’objectif principal du programme d’essais
de composition pour le BCR du barrage était de fournir des ensembles de pro-
portions de mélange pour les différentes combinaisons de matériaux cimentaires
disponibles localement pour la spécification des documents d’appel d’offres pour la
construction du projet polyvalent Mwache. Le rapport décrit les sources potentielles
qui ont été identifiées pour les différents matériaux: agrégats (en amont et en aval
du site du barrage), ciment Portland (deux types, tous deux du Kenya), cendres
volantes à faible chaux (deux types, tous deux de l’Inde), pouzzolanes naturels (un
type du Kenya et un de Tanzanie), laitier de haut fourneau (d’Inde) et des ralen-
tisseurs (une longue liste d’entre eux de BASF, Fosroc, Sika et Chryso). L’étape
suivante consistait à mener un programme d’essais préliminaires de composition,
qui comprenait l’optimisation de la granulométrie des agrégats, l’optimisation de la
maniabilité et de la teneur en eau du mélange BCR et l’optimisation de la posolo-
gie du ralentisseur (après avoir défini les exigences pour les temps de réglage).
Ensuite, le programme d’essais principaux de composition a été entrepris, en suiv-
ant les étapes ci-dessous: choix des proportions du mélange (différentes gammes
de teneurs en matériaux cimentaires en fonction du type de matériau cimentaire
utilisé; analyse de l’évolution de la résistance avec l’âge des BCR contenant des
cendres volantes, contenant de la pouzzolane naturel, et de celles contenant du
laitier (finalement écarté) et, enfin, définition des proportions optimales du mélange
et des ralentisseurs pour les différentes combinaisons des matériaux cimentaires.
L’une des principales conclusions soulignées dans le rapport est que des ralen-
tisseurs spécifiques ont dû être conçus afin d’obtenir les temps spécifiés pour les
BCR contenant les matériaux cimentaires particuliers disponibles. Cette recherche
a pris du temps qui n’aurait pas été disponible si le programme d’essais de com-
position n’avait pas été entrepris si tôt. En outre, il a été constaté que certaines
combinaisons de matériaux cimentaires potentiels étaient extrêmement efficaces.
711
GR Q. 104
Un mélange BCR contenant 32,5 kg/m3 de ciment Portland et une teneur en cen-
dres volantes à faible chaux de 187 5 kg/m3 et un autre mélange BCR contenant
une teneur en ciment Portland de 50 kg/m3 et une teneur en pouzzolane naturel de
170 kg/m3 , tous deux conformes à toutes les exigences du BCR pour le barrage de
Mwache. Le programme de mélange d’essais précoce et étendu a permis d’inclure
des gammes de proportions de mélange et des ralentisseurs spécifiques dans les
documents d’appel d’offres, ce qui a permis aux entrepreneurs de soumissionner
sur une base raisonnablement certaine.
Un autre rapport sur ce sujet a été reçu du Maroc (Boukaidi Laghzaoui et

al., R24). Il présente les détails des différentes combinaisons de sables qui ont été
utilisées dans la composition du BCR pour le barrage de Toudgha (h = 67, 5 m),
ainsi que d’autres caractéristiques des compositions de BCR utilisés. Jusqu’à 5
types de sable différents ont été utilisés, qui ont été combinés dans des proportions
différentes tout au long de la construction du barrage: sable de 0/4 mm obtenu
après traitement à l’usine de concassage sur place de la roche calcaire dolomitique
extraite d’une carrière située à environ 1 km en amont du site du barrage (non lavé);
le même sable mais après lavage; une teneur élevée en fines (65%, non plastique)
0/4 mm de sable obtenu par concassage de calcaire provenant d’une carrière située
à 20 km du site du barrage; un sable de 0/4 mm obtenu après retraitement dans
l’usine de concassage sur place le surplus significatif obtenu dans la production de
la fraction sablonneuse de 2/4 mm; et 0/2 mm de sable provenant d’une deuxième
source externe. La dernière partie du rapport est consacrée à certains détails de la
construction, brièvement commentés à la section 4.5.
Le rapport R30 de la Chine (Junjie et al.) est principalement axé sur la liaison
entre couches de BCR, et il est brièvement décrit dans la section 4.5 suivante. Une
partie du rapport fournit des informations détaillées sur les matériaux individuels util-
isés dans la composition du BCR, les proportions du mélange et les différentes pro-
priétés, y compris les propriétés mécaniques, de déformation et thermiques du BCR
frais et durci utilisé pour le projet hydroélectrique de Lower Kafue Gorge en Zambie.
4.5. CONSTRUCTION D’UN BARRAGE-POIDS EN BCR
Le premier rapport reçu qui correspond à cette section, R14 de Grande-

Bretagne (Dunstan), commence par un aperçu de près d’un millier de barrages
en BCR achevés ou en construction à la fin de 2019, et met en évidence la crois-
sance remarquable au cours des 35 dernières années de cette nouvelle forme de
construction de barrage. Une idée de base, mais en même temps fondamentale,
est que le concept original des barrages en BCR était la simplicité de la conception
et de la méthode de construction, car avec la simplicité viendrait la construction
rapide, ce qui implique qualité et économie. Il est mentionné que, malheureuse-
ment, au cours des dernières décennies, certains barrages en BCR modernes sont
devenus trop compliqués et que les avantages de la vitesse de construction ont,
dans certains cas, été perdus. Ainsi, il est fortement recommandé que les barrages
712
GR Q. 104
en BCR reviennent au concept original de simplicité et de construction rapide. Le

rapport rappelle que le site Web “http://www.rccdams.co.uk/” fournit une importante
base de données avec des détails sur plus de 950 barrages en BCR achevés ou en
construction dans le monde entier. Des statistiques sont fournies sur le nombre de
barrages en BCR achevés par an du début des années 1980 jusqu’en 2019 avec
l’emplacement indiqué des barrages BCR dans le monde. Le rapport traite égale-
ment des taux de placement du BCR dans les barrages et conclut qu’un nombre
important de barrages en BCR (probablement plus de 95%) ne sont pas conçus
et/ou construits aussi efficacement qu’ils pourraient (ou devraient) l’être. Le rapport
présente les données pour les treize barrages en BCR et remblai dur avec les taux
moyens de placement les plus élevés. La partie suivante du rapport fournit des
détails intéressants sur les cinq barrages du groupe précédent qui ont été achevés
bien avant la date prévue, afin d’essayer de comprendre pourquoi les barrages ont
réussi; les barrages sont: Longtan – Étape 1 (Chine), Upper Stillwater (États-Unis),
Olivenhain (États-Unis), Beni Haroun (Algérie) et Nam Ngiep 1 (Laos). Et enfin,
un aperçu des barrages de Bas Ghatghar (Inde), de Son La et de Lai Chau (tous
deux au Vietnam) est fourni comme trois très bons exemples des grands avantages
obtenus par un achèvement précoce des projets hydroélectriques, grâce à la rapidité
de construction des barrages en BCR associés.
Le rapport R17, de la France (Andrian et al.), a une partie qui traite du sujet de
cette section à partir de plusieurs études de cas, mais cette partie a été entièrement
abordée dans la section 3.3.
Boukaidi Laghzaoui et al., dans leur rapport R24 (Maroc), en plus de discuter
des détails des types de sable pour le BCR utilisé au barrage de Toudgha, men-
tionnés à la section 4.4, décrivent également un aspect clé de la construction sur
la rive droite. Comme la pente de la rive droite était très douce (à 7,9%), la mise
en œuvre du BCR en couches horizontales s’avérait délicate. Ainsi, des tests de
mise en place du BCR de couches inclinées, avec la même pente que celle de la
fondation, ont été menés avec des résultats positifs. La procédure a été adoptée
sur toute la rive droite jusqu’à ce que l’élévation du toit de la galerie en amont soit
atteinte; à partir de cette élévation, les couches BCR sont placées horizontalement.
Quatre rapports distincts ont été reçus de la Chine, tous liés au même bar-
rage en BCR qui est l’une des principales composantes du projet hydroélectrique
de Lower Kafue Gorge en Zambie. Le barrage mesure 130 m de haut et le volume
de BCR est de 1,3 Mm3. Le premier rapport, R27 (auteur non indiqué), aborde les
problèmes rencontrés par l’entrepreneur chinois en termes de faible niveau de com-
pétence de la main-d’œuvre locale nécessaire à la construction du projet, alors que
seulement pour la construction du barrage en BCR, un maximum de 50 employés
chinois et 932 zambiens étaient nécessaires. Le rapport explique certaines des
mesures prises pour atténuer le problème. Le deuxième rapport, R28 (auteur non
indiqué), traite de l’équipement de convoyeur à bande à grande vitesse Rotec utilisé
pour le transport du BCR de la centrale à béton au barrage. La longueur totale des
convoyeurs à bande était de 1559 m, avec une pente maximale de 27◦ , une vitesse
de bande de 3,6 m/s et une capacité de transport maximale de 500 m3 /h. Le rap-
port résume certains problèmes qui ont dû être surmontés liés à la mise en œuvre
713
GR Q. 104
et à l’exploitation du système de convoyeur. Le troisième rapport, R29 (Yuanguang

et al.) , met en évidence certains aspects de la construction, comme les systèmes de
pré-refroidissement en béton, l’équipement de transport à bande à grande vitesse
Rotec, le mélange BCR utilisé avec une taille maximale d’agrégat de 63 mm (sauf
pour la zone de parement amont en BCR avec une taille 37,5 mm), l’utilisation de
coulis entre les couches de BCR , l’utilisation de BCR enrichi sur les parements du
barrage et l’utilisation de cendres volantes, de poudre de roche et d’agent réduc-
teur d’eau naphtalénique à haut rendement dans le mélange de BCR. Enfin, le
quatrième rapport, R30 (Junjie et al), se concentre sur la liaison intercouches et
décrit certaines recherches menées, notamment: l’optimisation du mélange de BCR;
l’étude des caractéristiques des joints sous différents temps d’exposition et traite-
ment; l’établissement du facteur de maturité modifié; et la définition du système de
contrôle de la qualité.
4.6. ANALYSE D’UN BARRAGE EN MATÉRIAUX CIMENTÉS: CONCEPTION, MATÉRIAUX

ET CONSTRUCTION
Le seul rapport qui a été reçu sur le sujet lié aux barrages de matériaux cimen-
tés, est le rapport R8 du Pérou, écrit par Mendívil et al. Le rapport traite de l’objectif,
de l’analyse, de la conception, des matériaux et de la construction du cas d’étude
du barrage de dérivation de la rivière Asana dans le sud du Pérou, achevé en 2018.
Le barrage est l’une des composantes des infrastructures nécessaires à une grande
mine de cuivre, et il a été construit immédiatement en amont de la mine à ciel ouvert
pour détourner la rivière Asana et ses petits affluents à travers un tunnel qui permet
à la rivière de contourner le site minier, protégeant les opérations minières contre
les inondations et maintenant la qualité naturelle de l’eau pour les communautés
situées en aval. Le site connait une sismicité élevée (0,36 PGA) et les conditions
géologiques et géotechniques sont très complexes. Par conséquent, étant donné
que l’emplacement du barrage était déterminé principalement par la disposition de
la mine et non par des conditions de fondation optimales, la solution adoptée était
un barrage symétrique à remblai dur d’une hauteur maximale de 50 m. La con-
ception a été motivée par les conditions de fondation très variables, l’utilisation de
matériaux colluviaux en plastique disponibles à proximité du site et la possibilité de
construire le déversoir au-dessus du corps du barrage. Pour faciliter la conception,
des analyses aux éléments finis 2D et 3D ont été effectuées pour évaluer le niveau
de contraintes et les facteurs de sécurité de stabilité du barrage dans des conditions
de fondation critiques. Pour toutes les conditions de fondation et de charge, toutes
les analyses de stabilité effectuées ont montré une stabilité au glissement adéquate,
un barrage et un contact barrage-fondation en compression, et le transfert de con-
traintes faibles et uniformes sur la fondation. De plus, un calcul par éléments finis 3D
a été réalisé afin d’estimer les déformations totales du barrage et les tassements dif-
férentiels entre les plots. Le rapport donne une description détaillée des conditions
de fondation et des différentes singularités rencontrées (y compris une importante
faille subverticale de 8 à 12 m de large adjacente au chenal de la rivière), et des
mesures d’atténuation prises. En outre, la construction du corps du barrage en
remblai dur est discutée (taille de granulats inférieur à 50 mm, 120 kg/m3 de ciment,
714
GR Q. 104
Vébé = 35 sec, 5 MPa de résistance à la compression de 90 jours), en plaçant et

en compactant des couches horizontales de 30 cm d’épaisseur, et la protection du
parement amont réalisé avec 40 cm de béton conventionnel non armé à des fins de
durabilité et d’étanchéité à l’eau. Un mortier (taille de granulats inférieure à 19 mm)
a été placé au niveau des joints inter-couches sur une bande de 2,0 m de large à
côté du parement amont afin d’améliorer l’étanchéité entre les couches de remblai
dur. Un problème à résoudre était la grande plasticité des fines du matériau collu-
vium utilisé pour le remblai dur. Une autre caractéristique importante du mélange
de remblai dur sélectionné était le faible module d’élasticité de l’ordre de 7 GPa,
idéal pour les conditions variables rencontrées dans la fondation, car il permet une
déformation plus élevée de la structure sans fissuration. Depuis l’achèvement du
barrage, et après un remplissage partiel du réservoir lors d’événements de tem-
pête intense en février 2019, les inspections visuelles périodiques et les lectures
d’instrumentation n’ont révélé aucun problème majeur. Le barrage d’Asana est un
bon exemple de la sélection de la technologie de remblai dur symétrique pour des
conditions spécifiques, telles que celles de ce site de barrage et de ses environs.
5. INNOVATIONS POUR LA SURÉLÉVATION DE BARRAGES EN BÉTON

EXISTANTS
Il existe plusieurs raisons distinctes possibles, ou une combinaison d’entre

elles, pour décider de la surélévation d’un barrage existant. Certaines des raisons
peuvent être: augmenter le volume du réservoir (en raison de l’augmentation de
la demande et/ou de la perte de capacité due à la sédimentation), augmenter le
contrôle des inondations et/ou permettre une prise en crue d’une crue de dimension-
nement plus importante, corriger des défauts de conception en matière de stabilité
(sous dimensionnement, qui peut être le fait de réglementations plus strictes en
matière d’hypothèses de charge) en ajoutant du poids qui facilite la surélévation du
barrage en conséquence, réhabiliter le barrage lorsque l’état du barrage est précaire
par une mauvaise construction (anciens barrages en maçonnerie notamment) ou le
vieillissement du béton (en raison de phénomène de gonflement, gel/dégel, etc.).
Le Bulletin 64 de la CIGB (CIGB, 1988) sur la surélévation des barrages est

la dernière publication de la CIGB consacrée à ce thème. Une mise à jour de cette
publication, pour présenter des projets de barrages en béton qui ont été surélevés
au cours des dernières décennies dans le monde entier, a déjà été entrepris par le
Comité D de la CIGB. Un premier document sur le sujet, actuellement en prépara-
tion par le Comité de l’USSD sur les barrages en béton, constituera une première
contribution au nouveau bulletin de la CIGB.
Pour la question 104, trois rapports ont été reçus décrivant des expériences
de construction de barrages en béton existants, et deux d’entre eux ont un cas
en commun. Dans tous les cas, l’augmentation de hauteur décrite a été réalisée
en BCV et est présentée à la section 5.1. Dans la section 5.2, cependant, nous
proposerons une discussion générale sur la surélévation des barrages en béton
existants avec BCR.
715
GR Q. 104
Dans certains cas, après avoir analysé différentes options, il a été décidé de
construire un nouveau barrage, généralement en aval, voire en amont du barrage
existant. Le barrage-poids La Breña II en BCR et le barrage-poids Puentes IV BCV,
tous deux situés dans le sud de l’Espagne, sont respectivement de bons exemples
des deux alternatives pour la “ surélévation ” d’un barrage existant. Un autre exemple
assimilable est celui décrit dans le rapport R33 concernant le barrage de Spitallamm
en Suisse, qui est traité dans les sections 3.1 et 7.1. La principale raison de la con-
struction d’un nouveau barrage-voûte à double courbure était l’état précaire du bar-
rage existant. Le nouveau barrage a été construit immédiatement en aval de l’actuel,
avec une conception qui implique la possibilité d’une future surélévation de 22 m.
5.1. SURÉLÉVATION DES BARRAGES EN BÉTON EXISTANTS EN BCV
Le rapport R34, de Suisse (Leite Ribeiro et al.), explique les détails de la

conception et de la construction de la surélévation de deux barrages-voûtes, Cam-
bambe en Angola et Vieux-Émosson dans les Alpes suisses. Les aspects liés à
la construction des deux rehaussements ont été expliqués à la section 3.2. Le
barrage-voûte de Cambambe, haut de 72 m, a été construit entre 1959 et 1963.
Dès sa conception, le barrage a été conçu pour être surélevé. L’augmentation de
hauteur du barrage a eu lieu de 2012 à 2018 pour atteindre une nouvelle hauteur
de 92 m (majoration de 20 m). La disposition du barrage de Cambambe rehaussé
suit la même forme que celle du barrage existant avec une géométrie à trois voûtes
centrales. Le barrage original du Vieux-Émosson était un barrage-voûte gravitaire
à courbure unique de 55 m de haut, construit entre 1952 et 1956 et porté entre
2012 et 2015 à une nouvelle hauteur de 76,5 m (soit un gain de 21,5 m). Le barrage
d’origine n’a pas été conçu pour être surélevé. Avec la surélévation du barrage,
le barrage d’origine à arche à courbure unique est converti en un barrage-voûte à
double courbure (forme parabolique). Pour les contraintes géométriques au con-
tact entre le barrage d’origine et la partie nouvelle, il a été jugé nécessaire de
démolir la partie supérieure du barrage d’origine de 10 m et de 10 mètres supplé-
mentaires uniquement du côté aval, formant ainsi une marche de 10 m de haut.
Cette solution a été jugée la plus économique, compte tenu de la surface de con-
tact réduite et du volume de béton raisonnable, par rapport à d’autres alternatives
de surélévation qui maintenaient la configuration poids-voûte. Le rapport fait une
comparaison intéressante des deux surélévations, qui présentaient certaines simil-
itudes dans certains aspects, et des différences importantes dans d’autres. Ces
différences concernent en particulier les conditions très différentes du site (bar-
rage d’origine à surélever, sa géométrie, sa forme et les conditions existantes
du béton et des fondations), les conditions météorologiques (impact sur la con-
ception du mélange de béton, le calendrier de pré- et post-refroidissement et de
construction) et les conditions hydrologiques (affectant la gestion de l’eau dans le
réservoir, l’opération de clavage du barrage-voûte surélevé et la première mise en
eau du réservoir). Elles ont rendu indispensable l’adaptation de la conception et des
travaux de surélévation du barrage aux besoins spécifiques, les conditions locales.
En conclusion, le rapport souligne que, d’une part, les barrages-voûtes sont des
structures qui peuvent être facilement et économiquement surélevées, mais d’autre
716
GR Q. 104
part, ils constituent également le type de barrage le plus difficile en termes de com-
portement structurel, de conception et de technologies de construction à mettre en
œuvre. Par conséquent, l’augmentation de hauteur de ces structures minces doit
être soigneusement conçue, conçue, planifiée et gérée.
Un deuxième rapport de la Suisse (Droz & Wohnlich, R36) évoque égale-

ment la surélévation du barrage du Vieux-Émosson comme l’un des nombreux cas
référencés dans ce rapport, mais avec un degré moindre de détail. Le rapport est
examiné aux sections 7.1 et 7.2.
Le troisième et dernier rapport sur ce sujet, R43 de l’Inde (Chaubey et al.)

décrivant l’augmentation effectuée sur le barrage-poids en BCV de Dhanikhari a
été abordé à la section 4.1.
5.2. SURÉLÉVATION DES BARRAGES EN BÉTON EXISTANTS EN BCR
Comme nous l’avons déjà dit, il n’y a eu aucun rapport sur la surélévation
de barrages en béton existants avec du BCR. Bien qu’il n’y ait pas eu beaucoup
de réalisations de ce type, certaines d’entre elles sont remarquables, et plusieurs
projets remarquables sont en phase de conception, en particulier aux États-Unis.
Tirant parti des informations déjà recueillies par le sous-comité de l’USSD

qui prépare le rapport déjà mentionné sur la surélévation de barrages en béton,
le tableau 7 énumère les projets sélectionnés aux États-Unis de surélévation au
moyen de BCR ou de BCV et réalisé après le Bulletin 64 de la CIGB.
Tableau 7
Barrages en béton aux États-Unis (livre blanc de l’USSD en cours de
rédaction)
TYPE DE HAUTEUR SURÉLÉVATION ANNÉE
BARRAGE ÉTAT BARRAGE (M) (M) TYPE COMP.
Brut Colorado Poids arqué 104 39.9 BCR U.D.
San Californie Poids 67 35,7 BCR 2013
Vicente
Lac Bleu Alaska Voûte 60 26,5 BCV 2014
Théodore Arizona Voûte en 85 23,5 BCV 1996
Roosevelt maçonnerie
Halligan Colorado Poids-voûte 27 8,3 BCV ou BCR U.D.
Buffalo Bill Wyoming Voûte 107 7.6 BCV (armé) 1994
Bartlett Arizona Voûtes multiples 88 6.6 BCV (armé) 1996
et contreforts
Shasta Californie Poids arqué 184 5.6 BCV ou BCR U.D.
Note: U.D. = En phase de conception
717
GR Q. 104
Ces informations seront complétées dans un proche avenir avec d’autres

projets à travers le monde avec la préparation d’un bulletin de la CIGB.
6. INNOVATIONS POUR LES BARRAGES EN BÉTON EXTRÊMEMENT

HAUTS
Il a été indiqué au début de la section 3 que la plupart des barrages

extrêmement élevés construits dans le monde ces dernières années étaient des
barrages-voûtes en BCV. Mais seuls deux rapports sur l’un d’entre eux, le plus haut
barrage du monde, Jinping I en Chine, ont été reçus.
Bien qu’il n’y ait pas de définition précise du terme “ barrage extrême-
ment élevé ” aux fins du présent rapport, nous considérons les barrages en BCV
extrêmement élevés et ceux qui dépassent 200 m de hauteur.
Néanmoins et afin de permettre d’inclure plus de rapports dans cette section,

seront aussi considérés comme extrêmement élevés les barrages en BCR dont la
hauteur est supérieure à 150 m. À ce jour, dix-neuf barrages-poids en BCR ont été
construits dépassant cette hauteur (Dunstan, 2020), dont sept sont discutés dans
les rapports reçus. En ce qui concerne les barrages-voûtes en BCR, seuls deux
entrent dans la catégorie, et deux rapports ont été reçus faisant référence à l’un des
deux, le barrage de Janneh au Liban. Le tableau 6 de l’introduction de la section 4
fait la liste des dix barrages en BCR les plus élevés au monde.
6.1. BARRAGES-VOÛTES EN BCV EXTRÊMEMENT ÉLEVÉS
Seuls deux rapports ont été reçus pour le sujet des innovations pour les
barrages-voûtes en BCV extrêmement élevés, et les deux se réfèrent au même
barrage, actuellement le plus haut barrage du monde, plus pour très longtemps,
Jinping I (h = 305 m).
Le premier de ces rapports, reçu de Chine, Zhong et al. (R31), est dédié, sur
la base de données de surveillance et d’analyse à l’aide de modèles rhéologiques
élastoplastiques et viscoélasto-plastiques non linéaires 3D, à la prédiction et à
l’évaluation de la déformation à long terme de l’appui rive gauche et de son effet sur
le comportement de la structure du barrage à double courbure. La méthodologie
d’analyse globale de la stabilité des barrages-voûtes et de leur fondation est égale-
ment présentée, fournissant une référence pour les professionnels à l’avenir. La
rive gauche, avec une pente raide de 50o -65o et une fondation extrêmement com-
pliquée, principalement composée de marbre et partiellement de grès, présente
plusieurs failles et filons de lamprophyre ce qui entraîne des modes de rupture de
718
GR Q. 104
pente potentiels de glissement de cisaillement, de glissement de bloc et de dom-

mages en forme de coin, etc. La stabilité de l’excavation de 540 m de haut sur la rive
gauche a été assurée par d’importants renforts, y compris des ancrages rocheux,
des tirants et des clavages par galeries remplies de béton. La pente de la rive
gauche a été divisée en 6 zones de déformation différentes, et les éléments suiv-
ants ont été surveillés: déformation peu profonde de la pente, déformation profonde
de la pente et déformation de la vallée. Selon les données recueillies, quatre des six
zones sont à l’équilibre, tandis que dans une zone, la déformation n’a pas encore
convergé et dans une autre zone est encore dans la période d’ajustement. Depuis
que la pente de culée et la pente de culée en aval sont dans un état stable, on peut
conclure que la sécurité à long terme du barrage est assurée et que la pente de la
rive gauche, dans son ensemble, est dans un état stable. Afin d’étudier l’effet des
déformations de la pente de la culée gauche sur la sécurité du barrage-voûte, les
procédures suivantes ont été envisagées: 1) Analyse des structures géologiques
et de leurs paramètres rhéologiques; 2) Prévision de la déformation à long terme
de l’appui; 3) Analyse des effets de la déformation à long terme sur le barrage-
voûte, avec l’utilisation, comme outils fondamentaux de modèles aux éléments finis
élastoplastiques et visco-élasto-plastiques 3D, comme mentionné ci-dessus.
Le deuxième rapport, également de Chine, Lijun et al. (R32), fait également

référence à Jinping I en ce qui concerne les traitements de la fondation. Il est
souligné que Jinping I a les conditions géologiques les plus compliquées parmi
les barrages-voûtes extrêmement hauts avec notamment un rapport de 10 à 13
fois entre les modules de déformation des deux rives, ce qui a rendu la conception
du barrage très difficile. Le rapport explique que, du fait les conditions géologiques
extrêmement complexes du site du barrage, l’effet de différents défauts géologiques
sur la déformation et la stabilité du barrage-voûte a été déterminé par des analyses
numériques ; des traitements systématiques ont été évalués et déterminés pour
s’assurer que la fondation du barrage-voûte répond aux exigences de conception
pour capacité portante, stabilité à la déformation et imperméabilité. Malheureuse-
ment, à part mentionner qu’il a été effectué une analyse numérique 3D non linéaire et
des tests de modèles géomécaniques, aucune information spécifique n’est fournie
sur les résultats d’analyse numérique des traitements de fondation du barrage. Les
résultats montrent seulement que, après les traitements de fondation, l’intégrité et
l’uniformité de la masse rocheuse sous la culée gauche sont améliorées, et qu’on
obtient une augmentation de la résistance à la déformation de la fondation du bar-
rage et de la résistance du barrage au contact avec le rocher. Les auteurs concluent
que le système de fondation du barrage peut être considéré comme globalement
sûr. Les mesures de traitement des fondations incluses dans l’analyse numérique
sont les suivantes: tapis en béton, piliers et grille en béton, injection de consolidation
et injection du rideau d’étanchéité. En particulier, l’analyse montre que:
• Le tapis en béton peut grandement améliorer la rigidité de la fondation entre

certaines élévations, avec une réduction allant jusqu’à 30% des déplacements
le long de la rivière.
• L’injection de consolidation de la culée peut réduire le déplacement transversal
de 13,3 % à 21,3 % entre certaines élévations.
719
GR Q. 104
Selon les résultats de l’analyse numérique 3D et des essais de modèles

géomécaniques, un projet complet de traitement des fondations a été mis en avant et
réalisé, comprenant un tapis en béton de 155 m de haut d’un volume de 560 000 m3 ,
5 niveaux de galeries d’injection,720 000 m de forage pour l’injection de consoli-
dation, 3 niveaux de galeries anti-cisaillement, et remplacement systématique en
béton dans les zones de faille et de lamprophyre, d’un volume de 146 000 m3 . En
ce qui concerne le comportement du barrage, le niveau du réservoir a dépassé le
niveau de retenue normal 7 fois depuis octobre 2013. Le niveau des fuites est très
faible, et le déplacement du barrage sous l’effet de la pression hydrostatique est
fondamentalement élastique.
6.2. BARRAGES-VOÛTES EN BCR EXTRÊMEMENT ÉLEVÉS
Les deux seuls rapports qui correspondent partiellement au sujet des grands
barrages-voûtes en BCR sont R16 (Si-Chaib et al.), et R17(Andrian et al.), tous deux
reçus de France, et tous deux ayant une partie qui fait spécifiquement référence au
barrage-voûte en BCR de Janneh au Liban, c’est pourquoi les deux ont déjà été
couverts dans la section 3.3 à laquelle on peut se référer.
6.3. BARRAGES-POIDS EN BCR EXTRÊMEMENT ÉLEVÉS
Le rapport R4 reçu d’Italie, rédigé par Miranda et al., traite des systèmes de
géomembranes en général, pour fournir ou rétablir l’étanchéité dans les barrages
et, en particulier, leur application pour les barrages BCR, et plus spécifiquement
pour les barrages en BCR élevés, en utilisant comme cas d’histoire le barrage Miel I
BCR en Colombie (h = 188 m). Il est expliqué que les géomembranes sont utilisées
dans les barrages depuis plus de 60 ans (depuis 1959). En fait, ils n’ont d’abord été
appliqués qu’aux barrages de remblai de moins de 30 m de hauteur, plus tard à la
réhabilitation de tous les types de barrages et aux nouveaux barrages en BCR, sans
limitation de la hauteur du barrage et, par la suite, aux barrages très élevés de tous
types, que ce soit pour de nouvelles constructions ou pour la réhabilitation, avec des
travaux à l’air libre ou en subaquatique. Ce qui précède coïncide avec les recom-
mandations des trois bulletins successifs que la CIGB a consacré à l’utilisation des
géomembranes dans les barrages: bulletins 38 (1981), 78 (1991) et 135 (2010). Le
barrage en BCR de La Miel I a posé un jalon dans l’imperméabilisation des nouveaux
barrages en BCR par géomembranes. Pour les barrages très élevés, donc avec des
pressions hydrostatiques très importantes, les éléments essentiels de la conception
pour assurer une étanchéité totale sont le choix d’un type approprié de la géomem-
brane, un système de fixation capable de résister aux charges appliquées lorsque
la retenue est pleine ou vide, un système d’étanchéité périphérique connecté au
rideau d’injection de la fondation, et un système de drainage efficace pour éviter
les infiltrations d’eau dans le corps du barrage et pour permettre la surveillance des
720
GR Q. 104
fuites par des secteurs distincts du barrage. De plus, il convient de prévoir des méth-
odes de réparation pour remédier aux dommages accidentels, même sous l’eau. Le
rapport fournit les détails du système d’étanchéité installé, utilisant du SIBELON®
CNT 4400 de 3,00 mm d’épaisseur pour les 62 m inférieurs du barrage, tandis que
pour le reste du barrage du SIBELON ®CNT 3750 de 2,50 mm d’épaisseur a été
utilisé, tous deux soudés à un géotextile en polypropylène non tissé de500 g/m2 ,
ainsi que du système de fixation, et la description des travaux d’installation par sec-
tions horizontales distinctes. La bonne finition de surface du parement amont du
barrage, grâce à l’utilisation du BCR enrichi, a grandement facilité l’installation du
géocomposite. Un besoin de procédures strictes d’assurance de la qualité de la
construction, mises en œuvre par un entrepreneur spécialisé en imperméabilisa-
tion, est mis en évidence dans le rapport. Un programme de surveillance continue
mis en œuvre du comportement du barrage et du système géocomposite indique
des lectures stables répondant aux critères de conception. Les inspections effec-
tuées sur 18 ans d’exploitation du barrage hydroélectrique ont mis en évidence peu
de dommages à la membrane et pratiquement aucune réparation ni entretien n’a
été nécessaire. L’expérience au barrage Miel I confirme l’excellent comportement
durable des géomembranes sur les hauts barrages.
Un deuxième rapport reçu de l’Italie pour ce sujet, R7 (Pietrangeli et al.),

aborde diverses spécificités du projet grand barrage de la Renaissance éthiopienne
(GERD), pour lequel le remplissage du réservoir a commencé en juillet 2020. Le
barrage principal est un “méga” barrage-poids en BCR de 175 m de haut avec un
volume de BCR de 10,2 Mm3, ce qui en fait le plus grand barrage en BCR en volume
au monde [N.B. dans le registre des barrages en BCR (Dunstan, 2020), le GERD
est répertorié comme étant de 160 m de hauteur, et pour cette raison, il n’apparaît
pas dans le tableau 6]. La longueur de la crête est de près de 2 km et le barrage
est composé de 85 plots séparés par des joints de contraction à un espacement
de 18 à 27 m. Le rapport explique la vaste campagne de conceptions et d’essais
de compositions qui ont été effectués, montrant le comportement non linéaire en
contrainte-déformation du BCR et l’anisotropie (horizontale vs verticale) de la valeur
de résistance à la traction du béton. Ces résultats ont été utilisés par le concepteur
pour caractériser correctement la capacité de déformation de traction thermique des
mélanges et pour optimiser le zonage du BCR dans le corps du barrage par la véri-
fication des mélanges à faible teneur en ciment et moins de chaleur d’hydratation
pour les zones de contraintes élevées. Le corps du barrage a été zoné en cinq zones
dans lesquelles cinq mélanges différents de BCR ont été appliqués. Le ciment util-
isé est un ciment pouzzolanique. On n’a pas utilisé de cendre volante. Une teneur
suffisante en sable fin dans le mélange est obtenue en utilisant un bon mélange de
granit sain et de granit altéré. La teneur en ciment dans les cinq mélanges de BCR
varie de 80 à 142 kg/m3 , avec des valeurs de résistance à la compression garanties
comprises entre 8 et 15 MPa. Un mélange retardateur de marque Mapei a été
incorporé aux mélanges. Les mélanges à teneur en ciment plus élevée sont utilisés
dans la partie amont du barrage afin de répondre de résistance à la résistance à
la traction sous des exigences extrêmes de charge sismique et de perméabilité, et
au pied aval, pour les exigences de résistance à la compression. Des mélanges
à faible teneur en ciment sont utilisés dans la zone centrale du barrage afin de
721
GR Q. 104
contrôler l’élévation de température et le risque de fissuration qui en résulte. Un

mortier n’est prescrit que dans la partie amont du barrage, afin de répondre aux
exigences d’imperméabilité au niveau des reprises, et pour les 3 premiers m au-
dessus de la fondation. Le BCR enrichi au coulis est utilisé pour le parement amont
et le parement aval des sections non-déversantes. À l’aval de la section déversante,
les marches sont protégées par du béton conventionnel avec des fumées de silice.
La campagne d’essais comprenait environ 500 carottes provenant de différentes
zones du corps du barrage ; elles ont été testées sur place, et certains échantil-
lons ont également été envoyées au laboratoire Mapei de Milan, en Italie, pour une
vérification et une comparaison plus approfondie. Les résultats obtenus à partir des
tests approfondis ont été utilisés pour effectuer une analyse thermique transitoire
par un logiciel de différences finies développé par Studio Pietrangeli afin d’évaluer
les historiques de distribution de température dans le barrage et d’établir les néces-
sités du pré-refroidissement (c’est-à-dire la température maximale admissible de
mise en œuvre du BCR). Au cours de l’avancement de la construction du barrage,
une surveillance continue des températures du BCR a été réalisée et les mesures
sont périodiquement comparées aux températures maximales admissibles déduites
des études thermiques. De plus, une procédure de surveillance détaillée de la tem-
pérature a été mise en œuvre, basée sur la capacité de déformation de traction,
afin de prévenir le risque de fissures qui pourraient se développer dans le corps
du barrage. En ce qui concerne les analyses de stabilité dynamique, des analyses
aux éléments finis de niveau 1 ont été effectuées pour définir une répartition des
contraintes plus fiable et plus complète dans le corps du barrage, pour optimiser
la forme du barrage et pour définir le zonage du BCR et du mortier intercouches.
Le comportement du barrage sous des charges sismiques a été évalué au moyen
de l’analyse du spectre de réponse pour le calcul des contraintes maximales de
traction et de compression attendues en cas de séisme, et au moyen de la méthode
de la force latérale équivalente aux fins de l’analyse de la stabilité du glissement.
Le comportement du barrage a été supposé être essentiellement bidimensionnel,
étant donné que la hauteur du barrage est beaucoup plus petite que la longueur
de la crête. À l’étape de la conception de niveau 2, d’autres analyses plus pré-
cises ont été effectuées pour confirmer les dimensions finales du barrage et étudier
les effets locaux. Afin de réfléchir à la possibilité que des fissures se propagent à
partir du parement amont du barrage, une analyse statique non linéaire avec des
éléments de frottement d’espace a été réalisée à l’aide du code d’éléments finis
Straus 7.
Le rapport R14 de Grande-Bretagne (Dunstan) fournit des détails, principale-

ment liés à la vitesse de construction de certains des plus hauts barrages en BCR
construits dans le monde, notamment Gibe III (Ethiopie, H=246 m), Guangzhao
(Chine, H=201 m), Longtan (étape 1) (Chine, H=192 m), Nam Ngiep 1 (Laos,
H=167 m) et Guanyinyan (Chine, H=159 m). Ce rapport a déjà été examiné à la
section 4.5.
Un rapport reçu du Japon (R22, Aosaka et al.) traite d’un problème de fis-
suration rencontré dans le barrage-poids de Nam Ngiep 1 en BCR de 167 m de
haut au Laos. Les principaux composants des mélanges de BCR utilisés étaient
722
GR Q. 104
des agrégats de grès, un ciment de type I (ASTM C150) et des cendres volantes
de classe F (ASTM C618). Bien qu’une analyse des contraintes thermiques ait
été intégrée dans la conception et le plan de construction du barrage afin de con-
trôler la contrainte thermique du béton, alors que la mise en place du BCR sur
la rive gauche avait été suspendue d’octobre 2017 à janvier 2018 afin d’installer
la partie horizontale des conduites forcées, des fissures parallèles à l’axe du bar-
rage sont apparues à la surface du BCR du côté de la rive gauche, et des fissures
verticales sont apparues sur la surface exposée du joint de contraction CJ12 qui
séparait temporairement les deux moitiés du barrage à des altitudes différentes.
Bien que les surfaces en BCR et le joint CJ12 aient été curées, les fissures sont
apparues en raison d’une chute soudaine plus importante-que d’habitude-de la tem-
pérature ambiante à la mi-décembre 2017, ce qui a fait baisser les températures en
dessous de 15 o C. Afin d’évaluer la cause des fissures, une analyse des contraintes
thermiques a été réalisée à l’aide des données réelles disponibles provenant des
températures ambiantes et du béton surveillées et des essais de béton. L’analyse
des contraintes thermiques a été effectuée séquentiellement selon le calendrier
de mise en œuvre du BCR et du premier remplissage à l’aide du logiciel ASTEA
MACS. Les résultats de l’analyse ont montré, pour la surface horizontale où les fis-
sures sont apparues, des contraintes de traction maximales allant jusqu’à 2,5 MPa
à la mi-décembre 2017, ce qui est au-delà de la résistance à la traction du béton
de 1,5 MPa. Il a été montré que d’autres barrages en BCR dans la région de l’Asie
du Sud-Est ont connu des fissures similaires pendant la suspension de la mise en
place du BCR pendant plusieurs mois, et probablement pour la même raison, une
différence significative de température entre l’intérieur et la surface du béton en
raison de la différence de leur vitesse de refroidissement, et donc les contraintes
thermiques consécutives causées par les contraintes internes. La principale mesure
de réparation a été, dans la plupart des cas, d’installer un treillis d’armatures au-
dessus des fissures. Une étude de la stabilité du barrage a été réalisée en suivant
les avis du comité d’examen de la sécurité du barrage du projet et d’autres experts
du barrage, et en observant les données de la surveillance du déplacement, des
fuites du corps du barrage, du soulèvement... Comme les fissures verticales sont
parallèles à l’axe du barrage, elles n’entraîneront pas directement de fuite à travers
le barrage, et, après mise en eau, le béton sera soumis à des contraintes de com-
pression même là où il y a des fissures verticales, ce qui signifie que le barrage
est stable. La dernière partie du rapport comprend la description de travaux de
réparation recommandés pour les fissures sur la surface horizontale du BCR et les
fissures sur la surface verticale des joints de contraction. En outre, il pourrait être
intéressant de recouvrir la surface du BCR d’une feuille de protection, en plus de
la cure par pulvérisation d’eau et recouvrement lors de l’exposition de la surface
en béton pendant une longue période. Une leçon importante est qu’une attention
doit être accordée à la prise en compte de température ambiante dans l’analyse
des contraintes thermiques, lorsqu’une certaine surface de béton va être exposée
à la température ambiante pendant une longue période. Par exemple, appliquer la
température ambiante toutes les 12 heures plutôt que la température mensuelle
moyenne qui est généralement utilisée, ce qui permettra d’obtenir une tempéra-
ture et une déformation du béton plus réalistes, et d’améliorer la précision des
analyses.
723
GR Q. 104
7. COMPORTEMENT OPÉRATIONNEL DES BARRAGES EN BÉTON

PENDANT LE CYCLE DE VIE, Y COMPRIS DANS DES CONDITIONS
EXTRÊMES
Au total, onze rapports, entièrement ou partiellement consacrés au comporte-

ment réel d’un ou de plusieurs barrages, ont été reçus. Ils décrivent une grande
variété des problèmes rencontrés par les différents barrages en béton, y compris les
problèmes liés à la fondation, les différentes pathologies du béton (AAR, ISA, dégel
libre, etc.), l’affouillement en aval du barrage, l’inefficacité des rideaux d’étanchéité
et/ou de drainage et le vieillissement du dispositif de surveillance ainsi que les
mesures d’assainissement et de prévention mises en œuvre dans chaque cas.
Une conclusion fondamentale tirée de plusieurs des rapports est l’importance

de la conception, de la mise en œuvre et du suivi du contrôle et de la surveillance
que ce soit dans le cas de nouveaux barrages, pour les barrages existants qui ne
sont pas correctement auscultés et surveillés, et lorsque des modifications majeures
sont apportées à la structure du barrage, telles que la mise en place de contreforts
ou la surélévation.
7.1. COMPORTEMENT DES BARRAGES-VOÛTES EN BCV
Le premier rapport traitant de ce sujet du comportement des barrages-voûtes

en BCV est le rapport R18 de France (Grimal et al.). Il traite également de la modéli-
sation numérique du gonflement du béton et des travaux de réhabilitation effectués
au barrage de Bimont, c’est pourquoi il a été entièrement abordé dans les sections
3.1 et 3.2.
Le deuxième rapport est le rapport R31 de Chine (Zhong et al.), et comme le

précédent, il a été discuté en détail dans une section précédente, dans ce cas dans
la section 6.1.
Dans leur rapport, partiellement présenté à la section 3.1, Ballisat et al. (R33,
Suisse) décrivent le remplacement d’une structure de barrage plus ancienne par
la construction d’une nouvelle immédiatement en aval, tout en continuant à utiliser
sans restriction le réservoir. Dans la section 3.1, la conception du nouveau barrage-
voûte à double courbure a été discutée, tandis que dans cette section, nous nous
concentrerons sur le comportement problématique du barrage existant tout au long
de sa vie. Le barrage existant de Spitallamm (h = 114 m) a été érigé de 1926 à
1932 et c’était, à cette époque, l’un des plus hauts barrages en béton d’Europe. À
l’origine, deux conceptions étaient envisagées: un barrage-poids droit et un barrage
poids-voûte avec une pente en aval de 0,5H:1V et une pente en amont de 0,1H:1V.
Ce dernier type a été sélectionné, profitant des excellentes conditions de fondation
de la gorge étroite sur des roches granitiques et granodioritiques à haute résistance
et à faible déformabilité. Au cours de sa vie, le barrage a connu de graves problèmes
724
GR Q. 104
de vieillissement, y compris des dommages sur les deux parements, dus au gel-
dégel et aux fuites. Une campagne d’injection a malheureusement provoqué la
formation, à environ 12 m sous la crête du barrage, d’une fissure presque horizontale
qui correspondait à un ancien joint froid réalisé pendant la construction et rendait
nécessaire des mesures supplémentaires d’étanchement pour le parement amont
du barrage. Des instruments complémentaires de surveillance ont été installés, y
compris des inclinomètres dont les lectures ont été complétées par des mesures
avec des micromètres coulissants, pour mesurer les déplacements verticaux et
montrer que la partie supérieure du barrage se déplaçait indépendamment du reste
de la structure. Bien que l’instrumentation de la crête du barrage ait confirmé qu’un
processus de gonflement était en cours, les analyses d’échantillons de béton du
barrage n’ont pas révélé de signes significatifs d’expansion du béton en raison de
réaction alcali-granulat. La possibilité de déplacements tectoniques au niveau des
culées du barrage, affectant la partie supérieure élancée du barrage, a également
été écartée au moyen de mesures géodésiques transversales. Le rapport ne précise
clairement quelle a été la cause du déplacement continu vers l’amont de la crête du
barrage. Comme le barrage se détériorait lentement avec le temps, le propriétaire a
pris la décision de construire une structure entièrement nouvelle en aval du barrage
existant et de la faire construire pour permettre une augmentation ultérieure de 22 m.
Un autre rapport de la Suisse, Droz & Wohnlich (R36), qui se concentre sur la
sécurité des barrages tout au long de leur durée d’exploitation, comprend plusieurs
études de cas pour différents types de barrages. En ce qui concerne les barrages-
voûtes en BCV, les problèmes rencontrés au barrage des Toules (h = 86 m) dans
les Alpes suisses, un barrage-voûte à double courbure avec une forme très élancée,
une courbure verticale élevée vers l’aval, et aucune clé de cisaillement, sont dis-
cutés. L’origine des problèmes était la fondation, constituée de bandes alternées
de formations rocheuses subverticales parallèles à la vallée, avec des propriétés
et une rigidité variable, en particulier sur la rive gauche, où de petits déplacements
irréversibles vers l’aval ont été observés depuis la première mise en eau de la
retenue. Une analyse de stabilité a été effectuée et, par conséquent, des travaux
de réhabilitation ont été conçus et réalisés entre 2008 et 2011. Ces travaux con-
sistaient à ajouter un contrefort aval en béton au niveau des plots des rives. Un
deuxième cas est le barrage-voûte à double courbure de Kariba (h = 128 m), situé
à la frontière entre la Zambie et le Zimbabwe, pour lequel le fonctionnement du
déversoir de surface a provoqué une importante fosse d’érosion, de 90 m de pro-
fondeur, à quelques dizaines de mètres en aval du pied du barrage. Des travaux
de réhabilitation, consistant à remodeler la fosse, ont été mis en œuvre après les
résultats de tests en laboratoire sur modèles réduits, dans le but d’optimiser la forme
du bassin d’amortissement afin d’éviter une dégradation supplémentaire. En outre,
le barrage de Kariba est affecté par un phénomène de gonflement, et des travaux
de réhabilitation pour surmonter ce problème ont été organisés.
Le dernier rapport sur le comportement des barrages-voûtes en BCV est le

R42 de Norvège (Bjønness et al.), dans lequel les données de l’instrumentation d’un
barrage-voûte en BCV sont utilisées comme base pour la définition de la charge et
725
GR Q. 104
pour l’étalonnage d’un modèle aux éléments finis. Ce sujet a été examiné à la sec-
tion 3.1. En ce qui concerne le comportement, il y a peu d’informations dans le
document, expliquant seulement que des fuites ont été observées vers la fondation
de la voûte au cours de la première année d’exploitation. La fuite a été arrêtée par
injection l’année suivante, mais une fuite similaire a été observée des années plus
tard. Il a été maitrisé à nouveau il y a quelques années avec une nouvelle cam-
pagne d’injection dans la fondation. Simultanément, les activités d’instrumentation
et d’analyse structurelle du barrage ont été lancées afin d’améliorer la connaissance
du comportement du barrage. Le rapport présente un cas intéressant de la manière
dont le comportement documenté d’un barrage grâce à l’auscultation, peut être
utilisé pour obtenir une meilleure connaissance de la structure.
7.2. COMPORTEMENT DES BARRAGES-POIDS ET CONTREFORTS EN BCV
Le rapport R3, d’Italie (Fornari et al.), traite de l’ensemble des travaux de

réhabilitation et du comportement d’un système d’étanchéité qui a été appliqué sur
le parement amont d’un barrage-poids en BCV, pour arrêter les infiltrations à travers
le corps du barrage. Le cas d’étude est celui du barrage de Lago Nero (h = 38 m),
barrage construit dans les années 1924 à 1929 dans les Alpes italiennes, à haute
altitude (2025 m), dans un climat rigoureux avec des cycles fréquents de gel /
dégel, des températures de −24 à +26 o C, de la glace et de la neige épaisses, et
l’action des eaux de montagne douces, conditions préjudiciables à la durabilité du
béton. Plusieurs réparations ont été effectuées au fil des ans de 1953 à 1962, afin
d’améliorer l’étanchéité de la fondation et du corps du barrage, y compris des travaux
d’injection dans le substratum rocheux et le traitement complet du barrage avec de
la gunite armée, ce qui ne semblait pas être efficace à long terme. Une limitation
du niveau de la retenue a été imposée en raison du non-respect des conditions
de stabilité du barrage. Le propriétaire a alors décidé d’entreprendre des travaux
globaux de réhabilitation ; ils ont duré de 1979 à 1981 et ont été exécutés en trois
campagnes d’été. Les modifications apportées au barrage comprennent l’adoption
d’un système de revêtement CARPI étanche comme nouvelle barrière amont, ce
qui était une technologie assez innovante à la fin des années 1970 ; la membrane
est reliée à un rideau injecté à partir d’une nouvelle longrine. Le rapport fournit
les détails du système installé, utilisant la géomembrane SIBELON® CNT 2800 de
1,93 mm d’épaisseur soudée, lors de la fabrication, à un géotextile en polyester anti-
perforation de 200 g/m2 , pour former ce qu’on appelle un géocomposite. Il détaille
les systèmes de fixation et de drainage, ainsi que la description de la méthode de
mise en œuvre. Depuis son installation, ce système de géomembrane fait l’objet
d’une surveillance concernant l’infiltration et le vieillissement de la géomembrane.
En 2020, après 40 ans de service, le système fonctionne bien et, malgré l’altitude
élevée, une longue durée de vie restante (28 ans) est prévue pour un barrage vieux
de près de 100 ans. Sur la base de ce précédent, le propriétaire du barrage, ENEL,
a traité, à l’aide de systèmes à géomembranes CARPI, 5 autres grands barrages,
plusieurs barrages-voûtes et en maçonnerie, 7 réservoirs et plusieurs canaux et
tunnels hydrauliques. Le rapport souligne également les avantages de ce système
726
GR Q. 104
d’étanchéité qui ont conduit à son choix comme solution préférée par rapport à
d’autres options: la simplicité et la rapidité d’installation, l’étanchéité aux joints, un
coût relativement bas, l’absence d’entretien courant, et la possibilité d’une réparation
facile et rapide, si nécessaire (y compris en subaquatique). Les autres travaux de
réhabilitation décrits dans le rapport sont: la réalisation d’un poids additionnel sous
la forme d’un nouveau bloc de béton armé, surélevant la partie centrale du barrage
d’environ 6 m avec une galerie d’inspection à l’intérieur, la restauration du parement
aval, le remplacement des parties les plus dégradées de la couche de gunite armée
existante, l’augmentation de la capacité de l’évacuateur pour une crue millénale et
l’installation d’un système créant un mouvement de l’eau de la retenue pour éviter
la formation de glace à proximité du parement amont qui pourrait endommager la
géomembrane.
Le rapport R20 reçu de la Suède (Malm et al.) est le seul qui soit lié aux bar-
rages à contreforts en béton. Le but du rapport est de montrer comment les analyses
numériques peuvent être utilisées pour planifier la surveillance des barrages et pour
sélectionner des emplacements et des types de capteurs, et comment ces analy-
ses peuvent être utilisées pour prédire le comportement des barrages, ce qui est
important pour déterminer la sensibilité et la précision requises des capteurs. Des
modèles aux éléments finis peuvent être utilisés avec une grande précision pour
définir les niveaux d’alerte et être utilisés pour la définition des valeurs d’alerte,
c’est-à-dire des indications de défaillances potentielles. Ce concept s’applique à de
nombreux barrages qui approchent de la fin de leur durée de vie prévue, où la surveil-
lance du barrage est cruciale pour s’assurer qu’une sécurité élevée du barrage est
maintenue pendant que le barrage reste en service. Les auteurs recommandent
que les capteurs soient installés dans des positions stratégiques pour donner des
indications précoces si le comportement du barrage change et fournir des alertes
à temps sur les défaillances potentielles du barrage. Pour le suivi des nouveaux
projets, les méthodes de régression basées sur des données couramment utilisées
ne sont pas applicables pour définir les niveaux d’alerte car elles nécessitent des
enregistrements suffisamment longs du comportement du barrage pour caler les
paramètres. Le rapport comprend deux cas d’étude où ce concept a été appliqué.
Ces barrages suédois sont deux barrages à contreforts à dalles plates, Ramsele
et Storfinnforsen, situés à 10 km l’un de l’autre sur la rivière Faxälven ; ils ont une
hauteur maximale légèrement supérieure à 40 m, et ont été construits dans les
années 1950. Dans les années 1960, plusieurs fissures ont été détectées, suiv-
ant le même schéma, dans les deux barrages. La réhabilitation structurelle a été
effectuée au cours des années 1990. Cependant, environ 10 ans après la réhabili-
tation effectuée, d’autres fissures ont été observées. Le risque de nouveaux modes
potentiels de défaillance structurale a conduit à entreprendre un nouveau vaste
programme de renforcement et de prolongation de la durée de vie, comprenant
notamment l’installation de tirants précontraints dans les contreforts les plus grave-
ment endommagés pour renforcer la liaison avec la fondation. Avant la mise en
œuvre d’un programme de surveillance, des analyses numériques ont été menées,
afin de déterminer les modes de défaillance potentiels et la sélection des types
de capteurs appropriés, ainsi que leurs emplacements dans le barrage. En outre,
les modèles numériques ont été utilisés pour définir des valeurs d’avertissement et
727
GR Q. 104
d’alerte pour certains des capteurs. Les six contreforts choisis pour une surveillance
approfondie ont également été modélisés à l’aide des logiciels Abaqus et Brigade
Plus. Les résultats des analyses ont montré la variation attendue du comportement
du barrage et ont déterminé où et comment différents modes de défaillance poten-
tiels peuvent s’initier et se développer. Sur la base de ces analyses, il a été possible
d’identifier le type et l’emplacement appropriés de capteurs pour surveiller divers
modes de défaillance potentiels. Après 1,5 an de collecte de données, les mesures
enregistrées ont, jusqu’à présent, montré un comportement assez proche de celui
qui était attendu, avec de petites variations dans les déplacements de crête et les
largeurs de fissures. Par conséquent, le rapport introduit un outil très précieux qui
peut fournir des prévisions précises nécessaires pour aider le propriétaire dans la
prise de décision et la planification de l’entretien futur.
Le rapport R25 (Tzenkov et al., Bulgarie), outre le comportement du barrage,

traite également d’une analyse dynamique du barrage-poids en BCV de Beli Iskar
et, pour cette raison, il a déjà été discuté à la section 4.1.
Dans leur rapport, Droz & Wohnlich (R36, Suisse), déjà évoqué dans la sec-
tion 7.1, exposent des histoires de cas de trois barrages-poids en BCV affectés
par de l’alcali-réaction, le barrage de Salanfe (h = 52 m), dans les Alpes suisses
et construit en 1952, et les barrages Inga I (construit en 1972) et Inga II (1982)
en République démocratique du Congo. Le barrage de Salanfe a connu une défor-
mation irréversible vers l’amont en raison du gonflement de la masse de béton
induit par l’alcali-réaction. Des inspections approfondies et une extension spéci-
fique du système de surveillance ont été mises en œuvre, ainsi qu’une restriction de
l’exploitation du réservoir avant les travaux de réhabilitation en 2012-2013. À l’aide
d’une modélisation numérique intégrant les effets du gonflement dans le béton, les
travaux de réhabilitation ont été définis: exécution de 22 découpes verticales de
11 mm d’épaisseur, d’une hauteur maximale de 22 m, afin de libérer les contraintes
dans la structure. Comme le barrage de Salanfe, les barrages d’Inga I et d’Inga II
présentaient des déplacements continus et réguliers, mais dans ce cas vers l’aval.
En complément de la réhabilitation du système de surveillance des barrages, une
analyse de données satellitaires a été effectuée afin de mieux comprendre les mou-
vements du barrage et d’identifier une éventuelle déformation de la fondation. La
technologie SAR-interférométrie a été utilisée, ce qui a contribué à l’identification des
causes des grandes déformations. L’analyse approfondie d’échantillons de béton a
confirmé la présence d’alcali-réaction dans le béton. Le système de surveillance
amélioré a pu identifier le développement de cette pathologie d’expansion du béton,
qui doit être suivie régulièrement, bien qu’il ne prédise aucun problème de sécurité
dans un avenir proche.
Un dernier rapport reçu sur ce sujet du comportement des barrages-poids

et contreforts en BCV provient de l’Inde (R44, Gupta et al.). Il traite du transport
des sédiments et du fonctionnement opérationnel de cinq grands pertuis vannés
fonctionnant avec un faible dénivelé du barrage-poids en béton de Nathpa de 62,5 m
de haut dans la région de l’Himalaya. Les évacuateurs sont conçus pour permettre
le passage de l’eau avec une forte concentration de sédiments en suspension, ainsi
728
GR Q. 104
que des galets et des rochers. Pour cette raison, des chemisages en acier résistants
à l’abrasion et des rails en acier ont été mis en œuvre, ainsi que du béton haute
performance ; le rapport explique les détails de la durabilité de ces protections au fil
du temps. Au cours de la période de 15 ans allant de 2006 à 2020, chaque pertuis
a résisté à une charge sédimentaire de 118 millions de tonnes avec une vitesse
supérieure à 20 m/s. 69% des particules de sédiments ont des minéraux dont la
dureté à l’échelle de Mohs est supérieure à 6. Le revêtement a également résisté
à l’impact de rochers de taille moyenne supérieure à 1 m et jusqu’à 3,50 m, sans
dommages majeurs. Mais le haut des rails, qui étaient installés dans les coursiers
des évacuateurs, a été usé par les coups et l’érosion ; le béton formé de petits
agrégats entre les rails a été désintégré et emporté et les rails ont été délogés et
pliés. Afin de prolonger les périodes entre les réparations, la méthodologie adoptée
pour celles-ci a été modifiée en 2013, et un nouveau type de réparation est proposé
pour 2021. Il comprend des couches plus épaisses de béton haute performance
avec fumée de silice offrant une résistance à la compression minimale de 80 MPa
et une durabilité élevée. On peut retenir que le système composite de rails en acier
et de béton devrait être évité dans des conditions de vitesse élevée et de charge
sédimentaire élevée.
7.3. COMPORTEMENT DES BARRAGES-POIDS EN BCR
Le premier rapport reçu pour ce sujet provient de l’Italie (R4, Miranda et al.),
qui traite des systèmes de géomembranes et décrit comme un cas d’histoire le
barrage Miel I BCR en Colombie qui a été discuté à la section 6.3.
Dans leur rapport, Prasetio & Muis (R38, Indonésie) fournissent les détails
du programme de surveillance intensive mis en œuvre dans deux barrages situés
au centre de l’île de Sulawesi. Les barrages se trouvent dans une zone sujette aux
séismes en raison de la présence de deux failles majeures, et les mouvements
du sol qui proviennent de l’activité volcanique et des séismes sont assez courants.
Les barrages sont Balambano, un barrage-poids en BCR de 100 m de haut, et
Karebbe, un barrage combiné BCR-BCV de 74 m de haut. Le rapport fournit la
liste des instruments installés dans les deux barrages et discute de l’importance
de la surveillance en temps réel, afin qu’on puisse anticiper la rupture du barrage
due aux déformations induites par les mouvements horizontaux et verticaux induits
par les tremblements de terre, par les glissements de terrain, etc. afin d’atténuer
les dommages ou d’évacuer la communauté en aval. La surveillance implique une
surveillance visuelle, un relevé manuel des données d’instrumentation et un relevé
automatique de données en temps réel à l’aide de capteurs 3D et 1D pour les
mouvements statiques et dynamiques des barrages. Une étape pilote consistait à
mettre en œuvre une technologie de surveillance de la déformation des barrages
basée sur les stations actives GPS/GNSS-CORS (Global Positioning System/Global
Navigation Satellite System-Continuous Operating Reference Station). L’utilisation
de systèmes de surveillance géodésique (Geomos) peut être considérée comme
un pas en avant dans un système de surveillance de barrage basé sur les dernières
729
GR Q. 104
technologies du fait de la vitesse, la précision, la fiabilité et l’efficacité, avec très peu

de place pour l’erreur humaine dans les activités de surveillance des barrages. Avec
des données en temps réel, une réaction après séisme (ou tout autre événement
dangereux) peut être déclenchée immédiatement si des conditions anormales sont
trouvées.
8. CONCLUSIONS
Bien qu’il y ait de moins en moins de sites appropriés disponibles pour la

construction de nouveaux barrages, en particulier pour les barrages en béton, il
existe encore un grand potentiel pour la conception et la construction de nouveaux
barrages en béton de toutes les typologies possibles dans de nombreux pays,
pour la surélévation de barrages en béton existants et pour la réhabilitation et la
mise à niveau des structures existantes. Dans certains cas, la surélévation pourrait
être une alternative plus pratique et plus efficace à la construction d’un nouveau
barrage.
Dans de nombreux pays en développement, il existe un vaste potentiel

hydroélectrique inexploité et, en même temps, un énorme besoin d’irrigation,
d’approvisionnement en eau et de contrôle des crues. Les changements climatiques
exacerbent les problèmes d’inondations et de sécheresse dans de nombreuses
régions du monde ; ceci qui affecte également les pays développés. Ces prob-
lèmes peuvent être atténués par la construction justifiée et raisonnable de nouveaux
barrages ou par la surélévation et la modification de barrages existants, ce qui
aura sûrement plus de chances de succès dans un contexte de collaboration
internationale et multidisciplinaire.
Comme nous avons été témoins des progrès importants dans la conception
et la construction de barrages en béton au cours des trois dernières décennies,
la technologie a été en constante évolution, grâce au dévouement jour après jour
de centaines d’ingénieurs et d’autres professionnels. Les jeunes ingénieurs et sci-
entifiques ont pris l’avantage de travailler côte à côte avec leurs ainés, et de cette
façon, l’expertise peut être maintenue et évoluée.
Les près de quarante contributions reçues à Q. 104, préparées par des

ingénieurs et techniciens à la pointe de l’analyse, de la conception, de la construc-
tion et de l’exploitation et de l’entretien des barrages en béton, attestent de la vitalité
existant dans ce domaine.
Je crois en un avenir prometteur pour l’exploitation réussie de l’existant et

la construction de nouveaux barrages en béton dans les années à venir, grâce
aux progrès de la technologie et, peut-être, pourrions-nous assister à l’émergence
d’une nouvelle forme de construction de barrages, de la même manière que cela
s’est produit avec l’apparition du BCR / RCD au début des années 1980 ?
730
GR Q. 104
REMERCIEMENTS
Le rapporteur général tient à exprimer sa gratitude aux collègues suivants

pour avoir fourni des conseils et des références très utiles pour la rédaction de ce
rapport: Dr. Malcolm R.H. Dunstan (Grande-Bretagne), Carlos Granell et Francisco
Ortega (Espagne), Aled Hughes (États-Unis) et Dr. Quentin H.W. Shaw (Afrique du
Sud).
Je remercie également M. José Marques Filho (Brésil) et M. Jerzy Salamon

(États-Unis), respectivement Président et Vice-Président, pour la question 104,
d’avoir examiné et contribué avec de précieuses suggestions.
Enfin, merci encore au Président de la CIGB Michael Rogers (USA) d’avoir

proposé le sujet des barrages en béton pour cette Question 104.
RÉFÉRENCES
AKBARI, J., AHMADI, M.T. & MOHARRAMI, H. “Advances in concrete arch dams
shape optimization”. Applied Mathematical Modelling 35, 2011, pp. 3316–
3333.
CFBR. “Recommandations provisoires pour la justification du comportement des

barrages-voûtes”. Paris, France, 2018.
CHOPRA, A.K. “Earthquake Engineering for Concrete Dams – Analysis, Design,

and Evaluation”. Willey Blackwell 2020.
DU, C. “Recent advances in RCC Arch Dam design”. International Water Power
& Dam Construction, January 2021.
DUNSTAN, M.R.H. “www.rccdams.co.uk” and “RCC dams database. World Atlas

of International Journal of Hydropower & Dams’. October 2020.
FERC. “Engineering Guidelines for the Evaluation of Hydropower Projects,

Chapter 11: Arch Dams”. Washington DC, USA, March 2018.
ICOLD / CIGB. “Bulletin 64: Dam heightenings – Register”. Paris, France, 1988.
ICOLD / CIGB. “Bulletin 177: Roller-Compacted Concrete Dams”. Paris, France,

2020.
ICOLD / CIGB. “World Register of Dams”. Paris, France, updated April 2020.
ICOLD / CIGB. “Numerical Analysis of Dams – Proceedings of the 15th ICOLD

International Benchmark Workshop”, Milan, Italy. February 2021.
731
GR Q. 104
JONKER, M. & ESPANDAR, R. “Evaluation of Existing Arch Dam Design Criteria

in lieu of ANCOLD Guidelines”. ANCOLD Conference, 2014.
NISAR, A., DOLLAR, D., JACOB, P., CHU, D., LOGIE, C. & LI, G. “Nonlinear
incremental stress-strain analysis for Portuguese Dam; an RCC gravity arch
dam”. 28th USSD Conference. Portland, Oregon, USA. April 2008.
NUSS, L.K., MATSUMOTO, N. & HANSEN, K.D. “Shaken but not stirred. Earth-
quake performance of concrete dams”. 32nd USSD Conference. New
Orleans, Louisiana, USA. April 2012.
PAIXÃO, J. & SILVA MATOS, D. “Shape optimization in the design of concrete arch
dams – Two case studies”. Blucher Mechanical Engineering Proceedings. Vol.
1, num. 1, May 2014.
SHAW, Q.H.W. “The structural function of different arch dam types”. Proceedings.
SANCOLD. Dam safety, maintenance and rehabilitation of dams in Southern
Africa. Cape Town, South Africa, September 2015.
SHAW, Q.H.W. “The influence of stress-relaxation creep on the design of large

RCC dams”. Symposium – Knowledge-based Dam Engineering. 85th ICOLD
Meeting. Prague, Czech Republic. July 2017.
VARGHESE, B., SAJU, A. & JOHN, S. “Finite Element Analysis of Arch Dam”.
International Journal of Research in Engineering and Technology. Volume 03,
Issue 07, July 2014.
VICENTE, D.J., SAN MAURO, J., SALAZAR, F. & BAENA, C.M. “An Interactive
Tool for Automatic Predimensioning and Numerical Modeling of Arch Dams”.
Mathematical Problems in Engineering, Volume 2017.
WANG, R. “Key Technologies in the Design and Construction of 300 m Ultra-High

Arch Dams”. Engineering 2, 2016, pp. 350–359.
732
QUESTION 105
QUESTION 105
I
QUESTION
105
INCIDENTS ET ACCIDENTS CONCERNANT LES BARRAGES
1. Recent lessons from incidents and accidents concerning dams during the life
cycle, including during construction.
2. Evaluation of the flows and flood, estimation and quantification of the

consequences, including social, economic and environmental aspects, in
case of failure or incidents.
3. Emergency planning: regulation, organisation, information of the population

and examples of implementation.
4. Governance of the safety: definition of the responsibilities, periodic reviews,

implementation tests, organisation of lessons learned implementation.
INCIDENTS ET ACCIDENTS CONCERNANT LES BARRAGES

1. Leçons récentes tirées des incidents et accidents de barrages durant tout le
cycle de vie, y compris en construction.
2. Évaluation des débits et de l’inondation, estimation et quantification des

conséquences en cas de rupture ou d’incidents, y compris du point de vue
social, économique et environnemental.
3. Plans d’urgence : réglementation, organisation, information des populations

et exemples de mise en œuvre.
4. Gouvernance de la sécurité : définition des responsabilités, vérifications

périodiques, essais de mise en œuvre, organisation du retour d’expérience.
II
QUESTION
105
R. 1 GEORGES R. DARBRE, PATRICE DROZ, BOUATHEP

MALAYKHAM, HOUMPHANH VONGPHACHANH, SYCHATH
BOUTSAKITIRATH (Laos)
Institutional organization for dam safety in Lao PDR . . . . . . . . . . . . . . . . . . . . . . 1
R. 2 MATTEO SCOLARI, DANIELE GUALCO, LUCA BURASCHI,
(Italy)
Badana dam (Italy) retrofitting after structural damage . . . . . . . . . . . . . . . . . . . 19
R. 3 MARIA CRISTINA BRAMATI, FEDERICA DEL GIZZI,
FRANCESCO DOLCEAMORE (Italy)
The Italian emergency planning for large dams in case of seismic and
flooding hazard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
R. 4 BART VONK, WOUT DE VRIES, ERIC VAN KUIJK, MARCEL
BOTTEMA, LUDOLPH WENTHOLT, ERIC HUIJSKES (Netherlands)
Towards an international handbook for emergency response to flood risk . . . 48
R. 5 RODOLFO DALMATI, AILÍN PERTIERRA, JUAN FACUNDO
SOUTO, IGNACIO ESCUDER-BUENO, ADRIÁN
MORALES-TORRES, DANIEL CERVERA-MIQUEL,
CARINA R. CABALLERO (Argentina)
Risk assessment to inform of the decision-making process on
rehabilitations in Rio Hondo dam (Argentina) . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
R. 6 PAULO CASTRO, JOSÉ ROCHA AFONSO (Portugal)
Emergency planning in Portuguese dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
R. 7 Tomáš IČ, ROMAN IVANČO, BRANISLAV LIPTÁK, MARIÁN
MIŠČÍK, L’UBOMíR UHORŠŠÁK (Slovakia)
The repair of bottom outlet closing facility at the Palcmanská Maša and
Hriňová dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
R. 8 MANUEL IGNACIO SABAT, LARS ØDEGÅRD (Norway)
Flooding incident in the Tinguiririca valley, Chile in 2017 . . . . . . . . . . . . . . . 103
R. 9 WILLIAM F. FOOS, ENRIQUE E. MATHEU, DEAN DURKEE,
MICHELLE YEZIERSKI (USA)
Dam resiliency is more than just dam safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
R. 10 WESLEY CROSBY, KURT BUCHANAN, ALEXANDRA UBBEN
(UNITED STATES OF AMERICA)
U.S. Army Corps of Engineers modeling, mapping, & consequence
production center processes for dam breach analysis . . . . . . . . . . . . . . . . . . . . 133
III
R. 11 DEAN B. DURKEE, DOUGLAS D. BOYER,
(UNITED STATES OF AMERICA)
A review of recent dam failures in the U.S. and the anticipated impact of
implementation of risk-informed decision making . . . . . . . . . . . . . . . . . . . . . . 152
R. 12 LEE MAUNEY, MARK BAKER, IRFAN A. ALVI, NATHANIEL
GEE, GREGORY RICHARDS, DUSTY MYERS, MARK OGDEN
(USA)
A decade of learning from our past and preparing for our future:
the ASDSO dam failures and incidents committee . . . . . . . . . . . . . . . . . . . . . . 169
R. 13 EDWARD STOWASSER, MICHAEL KOON, WHITNEY SORRELS
(USA)
Rapid inundation mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
R. 14 LJUPCHO PETKOVSKI, STEVCHO MITOVSKI (Macedonia)
Contribution on restoration of tailings dams damaged at initial
period of construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
R. 15 RODNEY BRIDLE (UK)
Liquefaction and dam safety – lessons from events at Empingham,
Feijao and Fundao . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
R. 16 A. L. WARREN, P. J. MASON (UK)
Before and after the Toddbrook disaster – a review of UK reservoir
incident management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
R. 17 RÉMY TOURMENT, THIBAUT MALLET, SÉBASTIEN
PATOUILLARD, AKIM SALMI (France)
Accidentologie des digues fluviales de Loire, du delta du Rhône et de
l’Agly, et leçons tirées . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
R. 18 THOMAS LAURENT, QUENTIN BERCHER, THIERRY VINCENT
(France)
Gestion des crues et instabilités de pentes pendant la construction de
barrages : problématiques d’anticipation et de reconnaissances
préliminaires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
R. 19 THOMAS VIARD, JEAN ROBERT COURIVAUD, FRÉDÉRIC
LAUGIER, BENOIT BLANCHER, JEAN JACQUES FRY, PIERRE
SQUILLARI (France)
Évaluation de l’onde de submersion en cas de rupture des ouvrages
en remblai partie I : pratiques de l’ingénierie française pour les ruptures
de barrages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
R. 20 JEAN ROBERT COURIVAUD, LAURENT DEL GATTO, ANDRÉ
PAQUIER, GUILLAUME VEYLON, PIERRE PHILIPPE, ANTHONY
MOUYEAUX, SYLVIE NICAISE, CATHERINE FOUCHIER,
CLAUDIO CARVAJAL, LAURENT PEYRAS, RÉMI BEGUIN,
LAURENCE DUCHESNE, CHRISTOPHE PICAULT, JEAN-JACQUES
FRY (France)
Évaluation de l’onde de submersion en cas de rupture des ouvrages en
remblai partie II: R&D concernant la rupture par surverse ou érosion
interne des barrages en remblai et levées . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
IV
R. 21 LAURENT PEYRAS ET PATRICK DIVOUX, FRÉDÉRIC LAUGIER,
THIERRY GUILLOTEAU, MARIE CUBAYNES, MÉLANIE TRON,
THOMAS ADELINE, MICHEL POUPART, CATHERINE CASTEIGTS,
BENJAMIN DELARUELLE, JEAN-CHARLES PALACIOS, GLADYS
PAVADAY, GUIREC PREVOT, AGNES VALLEE, THIBAULT BALOUIN,
ERIC VUILLERMET (France)
Évaluation de la sûreté des barrages en France: retour d’expérience
et développement méthodologique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
R. 22 LAURENT BESSADI, FRÉDÉRIC LAUGIER, YANN TARAVEL
(France)
Gestion de la sécurité des barrages – illustration chez deux opérateurs
français . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .385
R. 23 M. ACHARYA, C. RICHARD DONNELLY, PRZEMYSLAW
A. ZIELINSKI (Canada)
Evolution of dam safety and emergency management practice for
transboundary dams – a global perspective with a Nepalese context . . . . . . . 404
R. 24 ANNICK BIGRAS, ERIC PÉLOQUIN, SIMON-NICOLAS ROTH,
ÉRIC MAINVILLE, MATHIEU ROY (Canada)
Chute-Bell dam emergency measures and remedial works . . . . . . . . . . . . . . . 424
R. 25 C. RICHARD DONNELLY, ERIC TIEDJE, ING. JESÚS ARANGO,
DANIEL FLORES (Canada)
Assessing the safety of the Ituango cement-bentonite cutoff wall . . . . . . . . . 444
R. 26 JEAN-PIERRE TOURNIER, AHMED F. CHRAIBI,
ANTON J. SCHLEISS (Canada)
Lessons learnt from the Saddle dam d failure of Xe-Pian Xe-Namnoy
project in Laos PDR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
R. 27 G. SNYDER & REGIS BOUCHARD, A. RATTUE, S. O’BRIEN
(Canada)
Muskrat falls project, repairing the upstream cofferdam . . . . . . . . . . . . . . . . . 481
R. 28 VIOLETA MARTIN, DANIEL ADRIA, HELEMAN WONG (Canada)
Inundation modelling of non-Newtonian tailings dam breach outflows . . . . 501
R. 29 D. J. HAGEN, LOUIS C. HATTINGH (South Africa)
Incidents and failures of small earthfill dams in South Africa:
lessons learned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
R. 30 HENRIETTE ANDERSON, PHILIP NICE, LOUIS
C. HATTINGH (South Africa)
Middle lake dam spillway safety incident and lessons learnt . . . . . . . . . . . . . 529
R. 31 ANIBAL MAITA, DR EDUARDO MARTINS BRETAS (Peru)
Malpaso dam incident and post-event actions . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
R. 32 A.F. CHRAIBI, A .NOMBRE, S.RIHI (Morocco)
Rehabilitation of the operating Comoe dam lateritic foundation . . . . . . . . . . 560
R. 33 XU ZEPING (China)
Lessons learnt from the failure cases on seepage control of CFRD . . . . . . . . 579
V
R. 34 YANG JUN, YAN YI (China)
A study on the communication history of the Three Gorges project . . . . . . . 590
R. 35 ROGER BREMEN (Switzerland)
Consequences on the design practice of worldwide dam accidents . . . . . . . . 599
R. 36 JONATHAN FAURIEL, ALEXANDRA BECKSTEIN, OLIVIER
FOURNIER, NICOLAS ADAM (Switzerland)
Emergency planning: lessons learned regarding reglementation,
organisation and implementation in Switzerland . . . . . . . . . . . . . . . . . . . . . . . . 608
R. 37 ALEXANDRA BECKSTEIN, BETTINA GEISSELER,
BURKHARD RÜDISSER (Switzerland)
Emergency planning and dam failure management: state of practice in
Switzerland, Austria and Germany . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
R. 38 PATRICE DROZ, GEORGES R. DARBRE, BOUATHEP
MALAYKHAM (Switzerland)
Emergency dam safety inspections in Lao PDR . . . . . . . . . . . . . . . . . . . . . . . . . 642
R. 39 ANISSA MAYANGSARI, YAYUK WIJAYA, DUKI MALINDO
(Indonesia)
Learning from incident during construction: cofferdam collapse at
Karalloe dam, Indonesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
R. 40 ANTO HENRIANTO, ESTI WULANDARI (Indonesia)
A quantitative approach to the reliability of the dam early warning
system, against the risk of loss of life in people at risk (PAR) facing
dam collapse disaster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664
R. 41 ANANG MUCHLIS, SONNY B.W, CECEP M.M, NAJLAWATI L
(Indonesia)
Analysis of spillway position on the Cibeet dam construction from soil
geological aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676
R. 42 ALTAN ABDULAMIT, DAN STEMATIU (Romania)
Dams and dikes safety management in Romania; past, present and
perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 687
R. 43 MAGELA, GERALDO PEREIRA (Brazil)
Cases of rupture of dams and lessons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 711
R. 44 RICARDO ABRAHÃO, ALEX, CALCINA (Brazil)
Three practical examples of rigid inclusions in concrete dam foundation . . 723
R. 45 FJÓLA GUÐRÚN SIGTRYGGSDÓTTIR (Norway)
Planning and design of temporary cofferdams – the case of a
cofferdam failure in Bergen, Norway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741
R. 46 MAYARI BERNARD-GARCIA, TEW-FIK MAHDI (Canada)
A worldwide database of dam failure case studies . . . . . . . . . . . . . . . . . . . . . . . 759
R. 47 WANNY K. ADIDARMA, ANISSA MAYANGSARI, OKY
SUBRATA (Indonesia)
Flood design evaluation at Batutegi dam, Indonesia . . . . . . . . . . . . . . . . . . . . . 779
VI
R. 48 XIN WANG, YUANJIAN WANG, ENHUI JIANG, XIANG LI
(China)
Game analysis of water and sediment allocation between cascade
reservoirs and lower channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
R. 49 QIANG WANG, ENHUI JIANG, YUANJIAN WANG, LIKE LI
(China)
Sedimentation problems and management strategies of Liujiaxia
reservoir, Yellow river, China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806
R. 50 A. NOMBRÉ, F. MILLOGO & M. KABORÉ (Burkina Faso)
Conséquences (économiques) des ruptures des petits barrages au
Burkina Faso : études de cas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 821
R. 51 A. K. SINGH, SUNIL J. GANVIR (India)
Change in methodology in construction of plastic concrete cut-off wall
adapted in Kishanganga He project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834

MICHEL POUPART (France),
General Reporter / Rapporteur Général . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 849
VII
PAPERS ON
QUESTION 105
RAPPORTS SUR LA
QUESTION 105
VIII
Q.105 - R.1
-------
-------
INSTITUTIONAL ORGANIZATION FOR DAM SAFETY IN LAO PDR (∗)
Georges R. DARBRE
Dam Safety Consultant on Assignment from the Swiss Agency for
Development and Cooperation SDC, Switzerland
Patrice DROZ
Dam Safety Expert, STUCKY Ltd, On Assignment from the Swiss Agency for
Development and Cooperation SDC, Switzerland
Bouathep MALAYKHAM
Director General, Department of Energy Management, MINISTRY OF ENERGY
AND MINES, Laos
Houmphanh VONGPHACHANH
Deputy Director General, Department of Energy Management, MINISTRY OF
ENERGY AND MINES, Laos
Sychath BOUTSAKITIRATH
Formerly Director General, Department of Energy Management, MINISTRY OF
ENERGY AND MINES, Laos
LAO PEOPLE’S DEMOCRATIC REPUBLIC (LAOS)
SUMMARY
An extensive review of the dam safety institutional framework prevailing in

Laos has been undertaken and the results compared to international practice. This
covers institutional organization and enforcement, legislation and regulations as
well as capacity building and development. Based on the findings, recommenda-
tions have been made on ways to strengthen the dam safety institutional frame-
work. Key element is the separation of Dam Safety Supervision from Hydropower
∗ Organisation institutionnelle pour la sécurité des barrages en RDP du Laos
DOI: 10.1201/9781003211808-47 1
Q.105 - R.1
Generation Supervision, with the associated adaptation of the organizational and

legal frameworks as well as development of education and training.
Recommendations include:
◦ Establishing a Department for Dam Safety within the Ministry of Energy and
Mines;
◦ Establishing a Lao National Commission for Dam Safety;
◦ Establishing a Technical Advisory Panel;
◦ Creating a Lao Association on Dams.
They also include:
◦ Enacting a Law on Dam Safety;

◦ Introducing an educational module in dam safety.
RÉSUMÉ
Une revue du cadre institutionnel en sécurité des barrages prévalant au Laos

a été entreprise et les résultats comparés à la pratique internationale. Cela couvre
l’organisation institutionnelle et la mise en œuvre, la législation et la réglementation
ainsi que le renforcement des capacités. Sur cette base, des manières de renforcer
le cadre institutionnel en sécurité des barrages ont été recommandées. L’élément
clé en est la séparation de la surveillance de la sécurité des barrages de celle
de la production hydroélectrique, accompagné de l’adaptation correspondante des
cadres organisationnel et légal ainsi que de l’éducation et de la formation.
Les recommandations incluent les suivants :
◦ Établir un Département de Sécurité des Barrages au sein du Ministère de

l’Energie et des Mines,
◦ Établir une Commission Laotienne de Sécurité des Barrages,
◦ Établir un groupe consultatif technique,
◦ Créer une Association Laotienne des Barrages.
Elles comprennent également l’adoption d’une Loi sur la Sécurité des Bar-
rages, et l’introduction d’un module éducatif en sécurité.
1. INTRODUCTION
1.1. BACKGROUND
Hydropower development in Lao PDR is progressing rapidly, major objec-

tives being domestic electricity supply and export revenues. Development of dam
construction is linked to this development of hydropower. While the country counted
2
Q.105 - R.1
ca. 10 dams with installed capacity of 15 MW or more 10 years ago, it counts ca. 50
today and is expected to count ca. 100 by 2030, to which an even larger number of
smaller hydropower dams as well as dams for irrigation, mining and other purposes
are to be added.
In the wake of the Xepian-Xenamnoy Hydropower Project saddle dam fail-

ure incident on 23 July 2018 and the nationwide flooding during July-August 2018,
the Government of Laos GoL initiated an emergency dam inspection of all large
hydropower dams in the country. World Bank and development partners assisted in
the process by deploying dam safety technical experts who reviewed owners’ safety
reports and performed inspections, among them the Swiss Agency for Development
and Cooperation SDC that sent 3 experts [Ref. 1]. During this process, it became
apparent that the dam safety institutional framework prevailing in the country was no
longer appropriate to cope with the challenges posed by a rapidly growing portfolio
of dams and needed strengthening.
Upon request from the Ministry of Energy and Mines MEM, SDC deployed an
expert in institutional dam safety mid-2019 to analyze and assist in the strengthening
of the institutional framework with focus on hydropower dams.
1.2. ANALYSIS METHODOLOGY
The following methodology is applied in performing this task:
1. Analysis of the present practice in Laos in terms of Institutional Dam Safety,

covering
◦ Institutional organization and enforcement;
◦ Legislation and regulations;
◦ Capacity building and development;
2. Comparison with international practices and with recommendations of
international institutions;
3. Identification and recommendation of strengthening options;
4. Selection of strengthening options and implementation.
Solely dam safety is addressed in the process, exclusive of other aspects

linked to constraints or potential harmful consequences that can arise from dam
operation (environmental impacts, public safety, security, water usage).
Steps 1 to 3 have been completed by this writing, work on step 4 is progressing.
2. ANALYSIS OF CURRENT SITUATION
2.1. DAM SAFETY ORGANIZATION AND ENFORCEMENT
Management of hydropower production including the related dam safety

aspects is entrusted to the Ministry of Energy and Mines MEM, with the Department
3
Q.105 - R.1
of Energy Management DEM being the leading governmental institution. The

Department of Mines DOM of MEM is the responsible institution for tailing and other
mine dams and the Ministry of Agriculture and Forestry MAF for irrigation dams.
Focusing on hydropower, DEM has the duty to manage water use in the coun-
try with focus on hydropower production and distribution. Managing dam safety
(proposing and improving technical dam safety standards, supervising compliance,
etc) is only one of the many tasks assigned to DEM.
3 of the 5 divisions of DEM assume duties related to the safety of hydropower

dams as part of more general and comprehensive duties:
◦ The Project Construction Management Division oversees dam safety during

design and construction;
◦ The Operation Project Management Division oversees dam safety during
operation;
◦ The Standards and Regulations Management Division of DEM drafts and dis-
seminates laws, regulations, standards, etc; it now also advises the provincial
authorities (Provincial Department of Energy and Mines PDEM) in dam safety
matters.
While the dam safety duties are clearly assigned among the divisions on the
paper, they are embedded in many other duties of the divisions concerned related
to power generation. As a consequence, there is no clear separation between “gen-
eration management” (of hydropower) and “supervision” (of safety of hydropower
dams). This is contrary to the principles of good governance.
It further appears that dam safety, when dealt with, is more present during
design and construction and much less to non-existent during operation. A reactive
dam safety practice rather than a proactive one is the consequence.
Promotion of a proactive dam safety practice respectively culture is also

hindered by the two following aspects:
There is no centralized information system on dams and on dam safety within

DEM, e.g. no centrally stored collection of legislation and regulations, no centrally
stored list of dams in operation (large and small) with their technical dam safety
relevant characteristics, no readily accessible overview of state of compliance with
dam safety requirements for the individual dams, no readily accessible overview of
outstanding dam safety issues at individual dams;
The need to balance the technical qualifications of DEM staff among the vari-
ous fields they have to cover leads to these qualifications lying mostly outside of civil
engineering. It is thus virtually impossible for the staff to identify unusual or critical
dam safety situations from safety reports and site inspections. It also weakens their
position during exchanges with dam owners on dam safety matters.
4
Q.105 - R.1
In terms of separation of management duties between the central government

(MEM/DEM) and the provinces (PDEM), it is set based on the installed capacity of
the associated projects (5 MW for new projects, resp. 15 MW for existing ones). This
might lead to adverse situations where PDEM is in charge of supervising dams of
high or extreme hazard classification.
2.2. LEGISLATION AND REGULATIONS
2.2.1. Hydropower dams
The highest legal or regulatory act on dam safety for hydropower dams is the
Law on Electricity of 9 May 2017 [Ref. 2], that applies to all electricity related activ-
ities. Its enforcement is compulsory for all stakeholders concerned. It differentiates
between large and small hydropower projects, depending upon installed capacity
(threshold set at 5 MW; previous legislation put the limit at 15 MW).
The Law assigns the responsibility for safe construction and operation of the
engineered facilities including dams, run-off river dams, penstocks and surge shafts
to the owners. It also assigns to MEM the role of Principal Authority with direct
responsibility to coordinate with other relevant Sectorial Authorities the manage-
ment of the electric activities throughout the country, including dam safety. The
management organizations comprise mainly MEM and PDEMs (Offices of Energy
and Mines of the districts, municipalities and cities also carry responsibilities for very
small projects).
Although the Law does regulate key issues such as assignment of respon-
sibilities for dam safety assurance and control, it suffers from flaws that prevent a
meaningful enforcement. In particular:
1. Dam safety is not the focus of the Law. The related provisions are partly
embedded in more general provisions, from where they must be extracted.
This leads to uncertainties as to applicability and meaning;
2. While roles and responsibilities of the various stakeholders are assigned,
they are based on installed capacity as opposed to dam dimensions or hazard
categories;
3. Insufficient provisions are made for financing the dam safety supervisory
tasks of DEM and PDEMs during the operation phase of projects.
The 2018 Lao Electric Power Technical Standards LEPTS [Ref. 3] is the second
highest act on dam safety for hydropower dams. Similar to the Law on Electricity,
they cover all electricity related activities whereby 2 chapters are of relevance when it
comes to dams (“General Provisions” and “Hydropower Civil Engineering Facilities”).
Conformance with the fundamental requirements of the LEPTS is compulsory, while
the technical contents may be substituted by alternatives fulfilling the fundamental
requirements.
5
Q.105 - R.1
Conformance with 2018 LEPTS is required for newly installed, rehabilitated

and operated facilities. For the facilities already in operation, under construction and
with completed design at the time the 2018 LEPTS entered into force, the owner is
required to do his utmost so that such power facilities conform to the fundamental
requirements within the limits of possibility (the previous version of LEPTS is dated
2004).
The 2018 LEPTS address qualification of the lead technical personnel by

requiring approval of the nomination of the owner’s chief engineer by the Super-
visory Authority at all stages (design, construction and operation). They also specify
what minimal examination and inspection of the facility the owner has to undertake
up to first impounding as well as what reporting to the Supervisory Authority he
has to make (including Emergency Action Plans, Rules of Flood Management and
Reports for Instrumentation and Monitoring Results). In this sense, the 2018 LEPTS
are rather complete.
It is noted that the 2018 LEPTS introduce a hazard rating used in the Dam
Safety Guidelines (cf. below), e.g. in relation with the frequency of inspections. This
rating (Low, Moderate, High, Extreme) is obtained from the dimensions of the dam
(crude substitute of vulnerability) and the consequences of dam failure.
The 3rd highest set of acts for hydropower dams is the Dam Safety Guide-
lines [Ref. 4]. They set out the main principles governing the management of dam
safety and cover the 8 parts of (1) Training, (2) Documentation, (3) Emergency
Action Planning, (4) Geological and Geotechnical, (5) Instrumentation and Monitor-
ing, (6) Quality Management Program, (7) Reservoir Impoundment, (8) Surveillance
(a 9th guideline on Concession Transfer is in preparation). They list and explain the
underlying reasons of dam safety measures, without going into technical speci-
ficities. Together with the 2018 LEPTS, they represent the technical dam safety
standards to follow.
The guidelines are comprehensive in terms of dam safety relevant areas cov-
ered and provide for check-lists for the owner’s staff and owner’s consultants as
well as for the Dam Safety Supervisory Authority. Still, their content remains at high
level; implementation without proper in-depth technical understanding can only be
incomplete and would lead to a false sense of safety.
Aside from the acts mentioned above, the Concession Agreements between
GoL and the project developers/owners may also contain provisions related to dam
safety. These provisions may deviate from those in the acts above.
2.2.2. Irrigation dams, tailing and other mine dams
The Ministry of Agriculture and Forestry MAF oversees design and construc-
tion of irrigation dams (by itself or provincial entities), while provincial entities provide
6
Q.105 - R.1
for maintenance. There is no specific legislation nor guidelines related to dam engi-
neering and safety of irrigation dams; general engineering principles are being
followed by MAF.
The mining industry is governed by the Law on Minerals [Ref. 5] whose enforce-
ment is placed under the surveillance and responsibility of the Department of Mines
DOM of MEM. A guideline for managing the tailing and mine dams has been issued
[Ref. 6]. It is noted that it introduces a consequence categorization, accounting e.g.
for the population at risk.
2.3. CAPACITY BUILDING AND DEVELOPMENT
2.3.1. University-level education
The National University of Laos NUoL offers a bachelor program in Civil Engi-
neering. Core civil engineering topics are being taught, without any special emphasis
on dam engineering nor dam safety related fields.
A part-time 3-year master program in Hydraulic Engineering has also been

offered at NUoL in the period 2016-2019, supported by the German Academic
Exchange Service (attended by 19 students). Dam Monitoring, Maintenance and
Safety has been offered as a selective subject (2 academic credits out of a pro-
gram total of 75), as well as lectures on Hydraulic Structures (3 credits) and
Concrete Dam Engineering (2 credits). The program reportedly remained at an
introductory level in the fields of more technical substance, e.g. geotechnical engi-
neering for embankment dams, dam foundation, strength of materials or earthquake
engineering.
Clearly, a comprehensive, sustained university level education in dam engi-

neering and dam safety is not available.
2.3.2. Continuing education
No professional organization dedicated to dam engineering and safety exists

in the country. In particular, Laos is not a member of the International Commission
for Large Dams ICOLD nor has a corresponding national association (cf. Chap. 4
for current status). Exchanging knowledge and experiences with peers as well as
promoting professional training, two of the several functions of such a professional
organization, thus does not take place
Workshop and lectures are held at DEM by international visiting experts as

well as by development partners. Such events do enhance dam safety awareness
7
Q.105 - R.1
and do contribute to a better understanding of technical issues, also of practi-

cal ones (through field visits). However, these workshops and lectures are not
part of a coordinated education and training program. For example, they have not
been systematically and directly linked to the 2018 LEPTS nor to the Dam Safety
Guidelines.
2.4. CONCLUSION ON CURRENT SITUATION
The analysis of the current situation has concluded that, overall, assignment
of roles and responsibilities in dam safety lacks clarity and that implementation lacks
efficiency Technical capabilities of ministerial dam safety staff need to be developed
as well as a comprehensive enforcement of dam safety regulation supported by
appropriate organizational and legislative measures. In details:
◦ The regulatory framework for hydropower dams in Laos is sufficient and

aligned with international best practice in terms of content. The Law on Elec-
tricity is, however, deficient in terms of clarity and structure for what dam
safety is concerned;
◦ There is no urgent need for further standards or guidelines, nor for their
revision. Rather, a comprehensive and coordinated educational and training
program on the 2018 LEPTS should be developed;
◦ The core elements of a sound dam safety supervision are present in the cur-
rent mandate of DEM. The present organization, however, suffers from a lack
of separation between generation supervision and dam safety supervision;
◦ Separation of Dam Safety Roles and Responsibilities between the central
government and the provinces is presently based on installed capacity, what
is not in agreement with the risks posed by a dam;
◦ Safety supervision of tailing and mines dams as well as of dams for irrigation
and for other purposes is currently separated from that of hydropower dams,
in spite of the fact that the same engineering principles apply. Setting up a
common Dam Safety Supervision should be contemplated;
◦ Enforcement of dam safety at DEM suffers from the absence of a Dam
Information System;
◦ Staff qualification at DEM is insufficiently aligned with the needs in advanced
knowledge in dam engineering and dam safety. Education and training
opportunities in dam safety are further lacking;
◦ There is insufficient integration in the international dam safety community.
3. STRENGTHENING OPTIONS
The dam safety framework for hydropower dams prevailing in Laos has been
compared with international practices and with recommendations of international
8
Q.105 - R.1
institutions, in particular those of World Bank [Ref. 7] and ICOLD [Refs. 8 and
9]. Specific strengthening needs to achieve a proper dam safety management and
effective implementation have been identified and options developed based on the
analysis. The following fundamental principles were thereby considered:
◦ Regulation, organization and capacity development focused on the ulti-

mate objective of protecting life, property and environment from the pos-
sible harmful consequences of inappropriate dam design, construction or
operation;
◦ Organization in compliance with the principles of good governance (separa-
tion between dam safety supervision and generation supervision);
◦ Enforcement respecting the “4-eyes” principle (first pair of eyes belonging to
the dam owner and the second pair to an independent body);
◦ Organization and enforcement tailored to the features of the country and the
size of the dam portfolio;
◦ Organization tailored to the assigned roles reminding that assuring safety of
a dam is the prime responsibility of its owner and that the role of dam safety
supervisory authority is to verify and not to execute).
3.1. ORGANIZATION AND ENFORCEMENT
It is first recognized that the current distribution of responsibilities between

dam owners and dam safety supervisory authority is in line with modern dam safety
principles specifying that it is the responsibility of the dam owner to ensure safety
of the dam and that of the Dam Safety Supervisory Authority to verify that the dam
owner does comply with the dam safety requirements. There is thus no reason to
modify this current distribution.
It is also recognized that supervision of dam safety is presently weak, with

the core origin being the insufficient separation of responsibilities at MEM between
development of hydropower generation on one side and dam safety supervisory
tasks on the other. The most important organizational measure proposed thus con-
sists in separating the two. This step is not only necessary from a good governance
principle point of view; it would also, in the long run, strengthen the dam safety
enforcement capabilities of the Dam Safety Supervisory Authority and increase its
credibility towards dam owners, thus contributing to a more sustainable dam safety.
An organizational structure has been developed on that basis, consisting of

(cf. Fig. 1):
◦ A Department for Dam Safety DDS whose sole task is to supervise dam
safety for all dams in the country;
◦ A permanent Lao National Commission for Dam Safety LNCDS providing for
strategic guidance in matters of dam safety in Laos and weighting interests
in case of conflicts between dam safety and water use;
9
Q.105 - R.1
Fig. 1
Proposed organization chart of dam safety institutions in Lao PDR.
Organisation proposée des institutions en sécurité des barrages au Laos.
◦ A Technical Advisory Panel supporting DDS in his endeavors by providing

expert opinions and assistance (composed of national and international dam
safety experts);
◦ A Lao Association on Dams LAD which, as a professional organization
attached to the government structure, promotes professional develop-
ment and secures the link with the international dam safety community as
represented by ICOLD.
The key roles and responsibilities of these various entities have been pre-
pared, down to the division level at DDS. This includes a proposition on the way to
redistribute the roles and responsibilities of MEM among the existing departments
so as to make room for the new Department for Dam safety DDS without increasing
the number of departments in the structure of MEM.
3.1.1. Department for dam safety DDS
The Department for Dam Safety DDS is meant to be the dam safety regulator
of the government of Laos. It would be entrusted with supervising the safety of all
dams in the country, from the design and construction stages up to operation and
emergency situations. It would also be responsible for developing the applicable
safety requirements and for promoting capacity development. Its activities would
cover dam safety only, not hydropower management or generation, nor irrigation,
flood protection, mining operation or other types of operation.
10
Q.105 - R.1
It is recognized that, in a first step, DDS should focus on hydropower dams

and limit itself to an advisory role to the other sectors.
A detailed mandate has been drafted for DDS. Some key duties are:
◦ Propose, develop and implement strategy, legislation, regulation and techni-

cal standards and guidelines;
◦ Disseminate strategy, legislation, regulations, technical standards and
guidelines;
◦ Advise and assist PDEM in matters of dam safety;
◦ Promote good practice in dam construction and operation;
◦ Approve the appointment of the owners’ chief engineers for design, construc-
tion and operation;
◦ Inspect reservoir site prior and during construction as well as during
operation;
◦ Control owners’ compliance with dam safety requirements during the whole
operation life of a dam;
◦ Provide training in dam safety and management and capacity development
for its own staff and relevant agencies at central and provincial levels;
◦ Prepare dam safety requirements for concession transfers and examine such
transfers from a dam safety standpoint;
◦ Monitor function of PDEM in implementing dam safety requirements;
◦ Issue approvals/non objections in accordance with 2018 LEPTS procedures.
3.1.2. Lao national commission for dam safety LNCDS
The Lao National Commission for Dam Safety LNCDS is meant toprovide
assistance and advices to the government on dam safety in setting strategies and
policies and in providing co-ordination and management to protect people, property
and the environment from harmful effects or misoperation or failure of dams and
reservoirs and to ensure a sustainable development of water use. For these pur-
poses, it acts both as a think tank to the Government and as a supervisory as well
as strategic body to DDS.
Tentative Duties and Scope of Rights have been prepared for LNCDS, encom-
passed to be composed of Officials from Prime Minister Office, Ministry of Energy
and Mines, Ministry of Natural Resources and Environment, Ministry of Agricul-
ture and Forestry, Ministry of Public Works and Transports, Ministry of Science and
Technology, Ministry of Planning and Investment, Ministry of Labor and Social Wel-
fare, National Disaster Management Committee, National University of Laos and
Governments of Provinces and Vientiane Capital.
3.1.3. Technical advisory panel
The Panel is meant to have both a regular advisory role on general matters
of technical dam safety treated by DDS and a more specific technical advisory role
11
Q.105 - R.1
when it comes to new dam projects as well as modifications, operation, repairs

and rehabilitation of existing dams. The Panel, to be composed of national and
international experts, would have no decision nor enforcement power.
On general matters, the Panel is meant to advise DDS by comparing its way
of working with international practice and identifying ways of closing possible gaps,
by looking at the pertinence of steps and decisions taken on specific projects, by
commenting on new and revised legislation and regulations.
On specific technical matters, the Panel would serve as the first source of
external consultants from which DDS would draw when it requires assistance, being
for resources reasons, complexity of the issues at hand or to ensure a broader
decision basis. One of the first such task of the Panel would be in relation with
the implementation of the correcting measures identified from the emergency dam
safety inspections and review [Ref. 1].
3.1.4. Lao association on dams LAD
The purpose of LAD is to promote best practice and progress in project devel-
opment, construction, maintenance, monitoring and operation of dam structures
and their environment. It represents a platform to discuss experiences, technolo-
gies, methods, etc. prior to deciding to endorse or promote their use. LAD shall
organize conferences, workshops and excursions for its members and be active
within ICOLD.
LAD will assemble dam engineering individuals, organizations, enterprises,

institutions that are related and interested in the field of dam business and that sup-
port the development of dams with works related to research, planning, operation,
survey, design, building, manufacturing equipment and materials, information tech-
nology, environment, management, business etc. All activities shall be pursued in
accordance with laws and regulations of Lao PDR.
3.2. SEPARATION OF TASKS & RESPONSIBILITIES OF DAM SAFETY SUPERVISION

BETWEEN CENTRAL GOVERNMENT AND PROVINCES (HYDROPOWER DAMS)
For the time being, tasks of dam safety supervision are shared by the central
government (installed capacity of more than 5 MW for new dams and more than
15 MW for existing dams) and provinces (other hydropower dams). One option for
the future is to pursue this separation, another one is to concentrate all dam safety
supervision tasks at the central government level. A final proposition cannot be
made before more precise information on number, types, purposes and risk type
of dams presently under safety supervision of the provinces are available (also of
12
Q.105 - R.1
the non-hydropower dams). In any case though, it is advised not to place any dam
under the safety supervision of local authorities (districts, municipalities, cities).
Should the provinces continue to exercise dam safety supervision, it is recom-

mended to abandon the current separation threshold based on installed capacity in
favor of one based on dimensions and/or risk type (hydropower dams).
3.3. DAM INFORMATION SYSTEM
Availability of a Dam Information System is central to proper functioning of a

Dam Safety Supervisory Authority. The present lack of such a system at DEM has
been identified as a major flaw of the present dam safety setting. Such a system
has the two following main functions:
1. Providing for uniformity and centralization of dam data;

2. Ensuring consistency, efficiency and oversight of enforcement.
Detailed terms of references have been prepared for a Lao Dam Information
System LDIS, to be accessed through a web-based interface. The system is intended
to be used by MEM in its function as Dam Safety Supervisory Authority to support its
management of dam safety enforcement during operation. Through the application
LDIS, the officers from MEM would at all times have an updated overview of the
technical characteristics of all reservoirs and dams in Laos, of the status of the
submission and examination of safety reports as well as of safety issues resolved
or still open. It is emphasized that LDIS is no document management system.
The selected database structure is depicted in Fig. 2. The individual reservoirs

are the core of the database; all entities are related to them, either directly or indi-
rectly via the dams (a reservoir entails one to several dams). The attributes forming
the database are conceptually stored in individual spreadsheets, with each spread-
sheet representing a separate entity, whereby xls 01 to xls 07 represent facts and
figures that do not change frequently in the life of a reservoir or dam (“static data”),
xls 08 to xls 10 relate to more frequently changing facts and figures (“dynamic
data”), and xls 11 contains the most important regulatory and similar references.
3.4. NON-HYDROPOWER DAMS
Safety of the non-hydropower dams must also be provided for (tailing and
mine dams, dams for irrigation, flood control, water supply for domestic use, cooling
water for thermal plants, etc). It has been suggested to have DDS be ultimately
also responsible for the supervision of the safety of these other dams. This would
ensure for a uniform approach of dam safety in the country as well as create a
13
Q.105 - R.1
Fig. 2
Proposed database structure of Lao Dam Information System.
Structure proposée de la base de données du Lao Dam Information System.
larger and thus more effective dam safety supervisory entity. Such an objective
should, however, be pursued in a progressive manner with the current setting (only
hydropower dams) being consolidated first before adding tasks and responsibilities
from other fields (respective legislation and regulations would also need to be aligned
to one another). As an alternative, the activities of the DDS discussed here may be
limited, in the long run, to an advisory role to, in particular, DOM and MAF.
3.5. DAM SAFETY CONCEPT
The Dam Safety Concept prevailing in Laos is aligned with international

practice; there is thus no need nor reason to modify it. In particular, design & con-
struction, operation & maintenance, and emergency preparedness are the core of
the prevailing concept.
3.6. LEGISLATION
The analysis has concluded that the Law on Electricity is deficient in terms of
clarity and structure for what dam safety is concerned. Because of the large span of
activities covered by the Law, it is feared that its simple revision would not bring the
clarity sought. It has thus been recommended to establish a well-structured, new
Law on Dam Safety based on 3 sources:
◦ The dam safety related provisions of the existing Law on Electricity;

◦ The dam safety related, compulsory fundamental requirements of the 2018
LEPTS;
14
Q.105 - R.1
◦ Additional dam safety provisions necessary to sustain the selected options

on organization and enforcement.
The dam safety related technical contents of the 2018 LEPTS and the Dam
Safety Guidelines would remain unchanged, cf. also Chap. 3.7.
While developing the new Law on Dam Safety as proposed above, it would
have to be verified that it fully supports the fundamentals of dam safety (separation
of dam safety supervision from power generation supervision, roles and responsi-
bilities of Owners and Dam Safety Supervision Authority, etc) as well as the retained
organization options (Department solely dedicated to dam safety, Lao Association
on Dams, Lao National Commission on Dam Safety, Technical Advisory Panel, etc).
On the enforcement side, it should be ensured that the “4-eyes principle” prevails
already at the owner’s level for dams qualifying as large according to ICOLD as well
as at the smaller ones having a risk type of “high” or “extreme” according to the 2018
LEPTS [Ref. 3].
The important aspect of providing for the necessary financial resources of

the Dam Safety Supervisory Authority will also have to be clarified in the Law. It
is proposed to introduce a supervisory fee to be paid by the dam owners to this
end. The motivation is that the setting up and the maintenance of a Dam Safety
Supervisory Authority is necessary because dams have been constructed and are
being operated to the benefits of their owners. It is then only fair that these dam
owners do finance the functioning of the Dam Safety Supervisory Authority.
3.7. REGULATIONS
Sustainable dam safety requires stability in the requirements. Frequent

changes are, in this respect, detrimental. In as much as the 2018 LEPTS and the
Dam Safety Guidelines already largely cover the relevant areas of dam safety, it
would be detrimental to revise them just after the latest versions have been issued.
Some deficiencies of the present regulations are known, e.g. as related to seismic
loading. They can, however, be addressed and corrected without totally revising the
2018 LEPTS and the Guidelines, e.g. through issuance of an addendum for the
most critical ones. One exception relates to penstocks; depending on how many
there are, preparing a guideline on their safety might prove necessary. The existing
draft Guideline for Transfer of Assets should also be finalized.
The main deficiency for what regulations are concerned lies in the lack of
education and training on their implementation. It has thus been recommended to
initiate such a training in a coordinated fashion, addressing both the theoretical
and the practical aspects in particular of surveillance. In the process, the training
material, tailored to the needs of the attendees, will become true aids to dam safety
implementation.
15
Q.105 - R.1
3.8. CAPACITY BUILDING AND DEVELOPMENT
Based on the analysis of the current situation and needs, on the

IAHR/UNESCO comprehensive study on education in Hydraulic Engineering
[Ref. 10], on the results of the ICOLD survey on Capacity Building [Ref. 11] and
on discussions with the engineering faculty at NUoL, it has been suggested to
strengthen higher-level education in fields related to dam engineering and safety
along the 3 following lines (cf. Fig. 3 for overall structure):
1. Strengthening of bachelor level program at NUoL with the objective of

ensuring completeness and quality of bachelor courses central to Hydraulic
Engineering, most importantly Hydraulics, Hydrology, Structures and Soil
Mechanics;
2. Introduction of an educational module in Dam Safety addressing specifi-
cally the capacity development needs in dam safety, in such a way that it
can become a module of a possibly reinstated Master Program in Hydraulic
Engineering at a later time (cf. below);
3. Reinstatement of a Master Program in Hydraulic Engineering at NUoL,
whereby the opportunity to do so must be further assessed.
Ways of implementing these actions have been detailed. For example, the
educational module in dam safety tentatively encompasses 75 hours resp. 5 aca-
demic credits of lectures and 2 weeks resp. 2 credits for practical thesis as follows
(adding a field training is open):
◦ Fundamentals (15 hrs);

◦ Surveillance framework (3 hrs);
◦ Design and construction principles (6 hrs);
◦ Condition and behavior of dam structures (12 hrs);
◦ Extraordinary and extreme situations (12 hrs);
◦ Surveillance instruments (12 hrs);
◦ Surveillance reports (6 hrs);
◦ Emergency situations (9 hrs);
◦ Practical thesis (2 weeks).
Promoting continuing education has also been proposed to supplement aca-

demic education by targeting active professionals in a very reactive manner (short
implementation time) and in being close to practical applications (thus much involve-
ment of practicing professionals as organizers and lecturers). It is broadly structured
in:
◦ National and international conferences on relevant topics;

◦ Workshops and short courses on specialized topics;
◦ Study tours.
16
Q.105 - R.1
Fig. 3
Structure of recommended academic and continuing education.
Structure de l’éducation académique et continue recommandée.
Aside from these sustained proposed endeavors, short-term mentoring

actions focusing on technical aspects of dam safety (intended primarily to indus-
try, in particular Electricite du Laos EDL and EDL-Generation) and supervision
aspects (intended primarily to GoL administration) rounds up capacity building and
development.
4. CURRENT STATUS OF IMPLEMENTATION
As per writing, MEM/DEM is analyzing the strengthening measures proposed.

A MEM internal Taskforce has been appointed to that effect. An obvious challenge
lies in the balance between resources needed and available. In parallel, the Faculty
of Engineering at NUoL has started assessing how it can be more active in Hydraulic
and Dam Engineering education.
Most advanced at this stage is the establishment of the Lao Association on

Dams, which is being finalized and whose ICOLD membership is expected to be
approved by the 2020 ICOLD General Assembly.
5. FINAL REMARK
A professional management of dam safety by owners and government agen-

cies is needed in Laos, as well as appropriate capacity building and development.
17
Q.105 - R.1
The Swiss Agency for Development and Cooperation is one of the development
partners (including World Bank) that is supporting the necessary strengthening
developments, taking a leading role in the institutional field.
ACKNOWLEDGEMENTS
The support and commitment from the development partners active in dam
safety in Laos and from the World Bank are gratefully acknowledged. Particular
thanks go to Tristan Bellingham and Sinongphen (Tik) Bouphanouvong from the New
Zealand-Laos Renewable Energy Facility in Vientiane for the fruitful exchanges and
their hospitality and to Voradeth Phonekeo as well as to the staff of DEM in Vientiane
and of SDC in Vientiane, Bangkok and Bern for their invaluable continuous support.
REFERENCES
[1] DROZ P., DARBRE G.R., MALAYKHAM B., Emergency Dam Safety Inspec-
tions in Lao PDR, 27ème Congrès des Grands Barrages CIGB, Question 105d,
2021.
[2] GOVERNMENT OF LAOS, Law on Electricity, 9 May 2017.
[3] GOVERNMENT OF LAOS, Lao Electric Power Technical Standards,

December 2018.
[4] GOVERNMENT OF LAOS, Dam Safety Guidelines, December 2018.
[5] GOVERNMENT OF LAOS, Law on Minerals, 13 December 2017.
[6] GOVERNMENT OF LAOS, Guidelines for the Management of Tailing and Mine
Dams in Lao PDR, October 2018.
[7] BRADLOW D.D., PALMIERI A., SALMAN S.M.A., Regulatory Frameworks

for Dam Safety: A Comparative Study, World Bank, 25069 Law, Justice and
Development Series, 2002.
[8] ICOLD, Dam Safety Management: Operational phase of the dam life cycle,
Bulletin 154, 2017.
[9] ICOLD, Regulation of Dam Safety: An Overview of Current Practice Worldwide,

Bulletin 167.
[10] Education of Hydraulic Engineers, IAHR/UNESCO Publication, 1996.
[11] ICOLD, Action Plan on Capacity Building, Ad-Hoc Committee on Capacity

Building, Dec. 2016.
18
Q.105 - R.2
-------
-------
BADANA DAM (ITALY) RETROFITTING AFTER STRUCTURAL DAMAGE (∗)
Matteo SCOLARI, Daniele GUALCO & Luca BURASCHI

Civil Engineer, RINA CONSULTING S.P.A, Italy
ITALY
SUMMARY
Badana Dam is a masonry gravity dam with a maximum height of 56.25 m.

The development at the crest is 216 m, with a curvilinear course with a radius of
curvature of about 200 m. The Dam was built in 1914 and in 2006 was subjected
to heavy structural damage which required the adoption of urgent safety measures,
including the rapid emptying of the reservoir.
This paper presents the design of the retrofitting intervention proposed to fully
restore the operation of the reservoir. Such retrofitting mainly consists in the demo-
lition of the damaged part of the existing dam and in the consequent reshaping with
new cast concrete.
The complexity of the analyzed problem needed also the support of sophisti-
cated finite element model, able to consider both the thermal and the structural point
of view and also to evaluate the interaction between the existing masonry dam and
the new cast concrete during the entire curing phase.
The detailed design of the proposed intervention was finally approved by the
Dams General Direction (Italian Ministry of Infrastructures) in 2017.
∗ Rénovation du barrage de Badana (Italie) après des dommages structurels
DOI: 10.1201/9781003211808-48 19
Q.105 - R.2
RÉSUMÉ
Le barrage de Badana est un barrage-poids en maçonnerie d’une hauteur

maximale de 56,25 m. La longueur de la crête curviligne est de 216 m avec un
rayon de courbure d’environ 200 m. Le barrage a été construit en 1914 et en 2006
a été soumis à de lourds dommages structurels qui ont nécessité l’adoption de
mesures de sécurité urgentes, notamment la vidange rapide du réservoir.
Ce rapport présente la conception de l’intervention de modernisation proposée

pour restaurer complètement le fonctionnement du réservoir. Un tel réaménage-
ment consiste principalement en la démolition de la partie endommagée du barrage
existant et en le remodelage consécutif avec du nouveau béton coulé.
La complexité du problème analysé nécessitait également le support d’un

modèle d’éléments finis sophistiqué, capable de considérer à la fois le point de
vue thermique et structurel et également d’évaluer l’interaction entre le barrage
en maçonnerie existant et le nouveau béton coulé pendant toute la phase de
durcissement.
La conception détaillée de l’intervention proposée a finalement été approuvée

par la Direction générale des barrages (Ministère italien des Infrastructures) en 2017.
1. INTRODUCTION
In Italy there are more than 9000 barrages, of which 532 are classified as
“large dams”. The Italian code defines “large dams” the dams higher than 15 m or
characterized by a reservoir volume greater than 1 × 106 m3 [1].
Among the 532 Italian large dams the average year of construction is 1955,
the average height is 38 m and the average reservoir volume is 27 × 106 m3 .
In Fig. 1 the Italian “large dams” are classified on the basis of the construction
year. As it can be seen most of such dams were built before the 1970 in two principal
periods, the first one was between 1921 and 1930 and the second one was between
1951 and 1970, during the economic growth after World War II.
As it can be seen from Fig. 1, most of the Italian “large dams” are characterized
by an advanced age and surely in the next years the assessment and retrofitting of
such existing dams should be an important task for engineers.
This paper analyzes one of such “large dams”, Badana Dam [2–5], built in 1914
and heavily damaged in 2006. In the following the proposed retrofitting intervention
will be presented.
20
Q.105 - R.2
Fig. 1
Construction years for Italian “large dams” [1].
Années de construction des “grands barrages” italiens [1].
Fig. 2
Badana Dam location.
Position du barrage de Badana.
Badana dam blocks Badana stream, a left tributary of the Gorzente stream,
into which it flows near the reservoir of Lake Lavezze. The Gorzente stream is a sub-
tributary of the Orba stream, located on the Po side of the Ligurian Apennines, on the
Genoa-Ovada axis. The reservoir extends partly in Liguria partly in the Piedmont,
in the north of Italy, Fig. 2.
21
Q.105 - R.2
Fig. 3
Badana Dam – Upstream view after reservoir emptying in 2006.
Barrage de Badana - Vue en amont après la vidange du réservoir en 2006.
The hydrographic basin area is approximately 4.8 km2 . The basin shape is
approximately rectangular with the longer side 2.8 km long and the shorter side
1.7 km long.
Badana Dam was built between 1907 and 1914 and has been in operation
since 1914. The reservoir volume is approximately 4.66 × 106 m3 . The maximum
regulation level is at 716.90 m a.s.l. while the maximum reservoir level is at an
elevation of 717.00 m a.s.l.
The operation of the reservoir historically follows the following cycle: full reser-
voir, water level close to the maximum regulation (716.90 m a.s.l.), in the first half of
the year, reduction of the reservoir level to an elevation of 704 m a.s.l. approximately
in September-October, return to full reservoir between November and December.
The typical management of the reservoir therefore has relatively long periods during
which the level is close to the height of maximum regulation.
In February 2006, serious structural failures occurred in the dam body, which
required the adoption of urgent safety measures, including the rapid emptying of the
reservoir, Fig. 3.
The strategic importance of such reservoir as water resource for Genoa munic-
ipality, made necessary a retrofitting intervention on the dam in order to restore the
22
Q.105 - R.2
reservoir fully operation. To this end, an extensive investigation campaign was car-
ried out analyzing both the geological and geomechanical nature of the area, and
the structural characteristics of the dam body.
Based on the results that emerged in such investigations, as well as on the

studies and numerical analyses carried out, including the suggestions provided by
the Dams General Direction (Italian Ministry of Infrastructures), the solution consid-
ered most suitable to restore the efficiency of the plant , from a static, hydraulic and
environmental point of view, it has been developed and herein presented.
2. DAM CHARACTERISTICS
Badana Dam is a masonry gravity dam with a maximum height of 56.25 m.

The crest is located at 717.25 m a.s.l., has a width of about 4.5 m and is bordered
by two parapet walls 1.2 m high. The development at the crest is 216 m, with a
curvilinear course with a radius of curvature of about 200 m.
The upstream face has a variable slope from i = 0.3H : 1V at the base up to
i = 0 (vertical) above 711.3 m a.s.l. For the downstream face the slope varies from
i = 1H : 1V at the base up to i = 0.25H : 1V, immediately below 711.3 m above sea
level. Also in this case, the upper section is vertical.
Badana Dam is made up of blocks of stones bound with hydraulic lime mortar,
sand and pozzolana. There are no tunnels inside.
Fig. 4
Badana Dam – downstream view.
Barrage de Badana - vue en aval.
23
Q.105 - R.2
The dam is not equipped with an upstream grout curtain, nor with a drainage
system to control the seepage and the uplift pressure.
The main discharge system of the Badana Dam consists of a 6-pipe siphon
surface spillway located on the left bank, whose maximum theoretical flow rate is
estimated at 75 m3 /s. The overflowing sill of the siphons (at the beginning of the
overflow) is at 716.90 m a.s.l.
The bottom outlet consists in 2 pipes with a diameter of 600 mm, arranged
transversely to the axis of the dam, able to drain a maximum theoretical flow rate of
about 17.5 m3 /s. The two pipes, with axis height upstream of 668.50 m a.s.l., cross
the dam for about 15 m from upstream to downstream. Each pipeline was equipped
with a service gate at the outlet section, so there was no possibility of closing the
flow upstream: therefore, during operation, the water pressure inside them always
corresponded to the height of the reservoir.
3. HISTORICAL MONITORING AND 2006’S STRUCTURAL PROBLEMS
3.1. MONITORING
The monitoring data available today includes:
◦ 4 piezometers installed in 2000;

◦ Readings of the displacement on the dam crest, starting from 1960 in the
central point and, starting from 2000 in two additional point, on the left and
on the right;
◦ Seepage flow measure.
3.1.1. Piezometers
In the two piezometers of the highest sections (SBa-1 and SBa-2), there was
a constant and sensitive increase in piezometric levels in the years 2001, 2002 and
2003. In particular, one of them rose by about 18 m in such years. Even in the
piezometers of the intermediate height sections (SBa-3 and SBa-4), the readings
between 2000 and 2005 tend to rise, and the maximums were very close to the
reservoir level.
3.1.2. Displacements on the dam crest
The displacement measurements on the crest have been carried out on aver-
age twice a year since 1967. Fig. 1 shows the displacement of the central pillar:
since October 2005 the displacements of the pillar have increased by 8-10 mm and
during February 2006 event, the key of the arch moved a further 16 mm to return,
24
Q.105 - R.2
Fig. 5
Measure of displacement on the dam crest between 2005 and 2006.
Mesure du déplacement sur la crête du barrage entre 2005 et 2006.
when the reservoir is completely empty (685), to a position about 5 mm further

downstream than that of the previous October (reservoir at 714).
It has to be underlined that the maximum displacement values are generally

recorded at full reservoir and at low temperature (this is presumably linked both to
the fact that the maximum reservoir occurs in winter, and to the fact that with low
temperatures the arch that constitutes the upper part of the dam tends to shorten
and therefore reduce its contribution).
The critical event apparently occurred following an abrupt, but modest,

increase in the reservoir level, which soon passed from 715 to 717 m asl. All the
displacement measures recorded an important increment, which was equal to 16,
7 and 4 mm, for points Central, Left and Right, respectively. The maximum value
of displacement of the Central point was equal to 26 mm, about 2.5 times greater
than the maximum values recorded in the past. Such important and discontinuous
displacements with respect to the mean displacement curve seem to be due to a
structural failure of a part of the dam or of its foundation. The distortion created
between the central and the lateral points was equal to 9 mm to the Left and 12 mm
to the Right. Taking into account that the distance between adjacent measurement
25
Q.105 - R.2
points is in the order of 40 m and that the centers of rotation of the corresponding
sections are located at different elevations, these differential displacements nec-
essarily lead to important redistribution of stresses within the dam body. After such
increasing of displacement the reservoir was rapidly emptied and the displacements
gradually decreased, while remaining higher than the available reference values.
3.2. 2006 STRUCTURAL PROBLEMS
3.2.1. Event description
In the period prior to the failure that occurred in February 2006, the reser-
voir level reached 716 m a.s.l., around December 20th , 2005, and remained
approximately constant for the following 2 months.
In the last days of the week February 13th -19th , 2006, 160 mm of rain were
recorded, of which 120 mm in a single event lasting about 2 hours.
The reservoir experienced an increase in level starting from February 19th ,

which led to the activation of the spillway, at an elevation of 716.9 m a.s.l., a level
only 0.35 m below the height of the crest 717.25 m a.s.l.
On February 21st , 2006, the morning inspection, conducted before 10 A.M.,

did not reveal significant losses or in any case such to suggest the existence of an
imminent or potential danger.
Strong leaks were instead identified in a second inspection, conducted in the

afternoon of the same day, after the personnel in charge noticed a significant lower-
ing of the reservoir level. Between 10 A.M. and 12 A.M., therefore, 3 major leakage
suddenly occurred downstream of the dam, localized as follows:
◦ on the right bank at the level of the riverbed, in the contact area between the
dam and the foundation;
◦ on the left bank at 682 m a.s.l., also in the contact area between the dam
and the foundation;
◦ on the left bank at an elevation of 675 m a.s.l., in correspondence with a
rocky outcrop located about ten meters downstream of the dam body. This
situation is also associated with a secondary loss at an elevation of 677 m
a.s.l. on the dam face.
The global extent of these leakage was estimated at 3 m3 /s.
The management of the exercise immediately decided (at about 8 P.M. of the
same day), as a precaution, to empty the reservoir by opening the bottom outlet.
26
Q.105 - R.2
3.2.2. Dam condition after 2006 event
External Cracks
On the downstream face, in the central portion just above the half of the dam,
some horizontal cracks were visible, approximately parallel to each other. More
specifically, an approximately continuous sub-horizontal crack could be observed at
an elevation of 689 m a.s.l., with sub-parallel branches affecting elevations between
688.5 and 694 m a.s.l.
On the upstream face a continuous sub-horizontal crack was at an elevation

of approximately 689-690 m a.s.l., which went up to the right bank up to 692 m a.s.l.
Still on the upstream face, relatively closed sub-vertical cracks were visible,
approximately at the limits of the highest dam portion.
Internal cracks
After the 2006’s event in order to keep the reservoir empty, a transversal tunnel
was built at the base of the dam, approximately in correspondence with the key
section, with the function of completely discharging the incoming flows, in particularly
those concentrated during significant rainfalls.
Inside such tunnel, between Sta. 7.00 and 10.00, a macroscopic discontinuity
was found in the masonry. This crack corresponds to a horizontal displacement
towards the valley of a considerable portion of the dam body, hereinafter referred
to as the "plate". The investigation carried out on such internal cracks led to the
identification of a rather irregular detachment surface inside the dam body, with
dislocations measured on the tunnel of about 120 mm. The estimated volume of the
detached plate was approximately 8500 m3 .
Failure analysis
Several analyses were performed considering the conditions created after the
critical event of February 2006, to better identify the possible causes.
One of the causes of the failure can surely be related to the uplift increasing at
the base of the dam, over the course of almost a century of operation, considering
the absence of watertightness systems (grout curtain) and control systems (drains
curtain). Starting from the available piezometers data, which cover the years after
2000, there is a certain tendency to increase uplift, as well as a tendency to increase
the leakage measured in the drainage channels. This suggests that there has been
a progressive damage of the joints at the dam base.
27
Q.105 - R.2
The dam is arched, but the opening created by the siphons on the left inter-
rupted the continuity of the upper part of the arch. The arched shape redistributed
the stresses with an effect of unloading the central part and overloading the lateral
parts. The effect was amplified when the dam was subjected to low temperature
(especially when the reservoir is full). All of the above had the effect of increasing
the compressive stresses in the downstream part of the dam, especially in the left
part, and of inducing traction upstream. On the left part, some close sub-vertical
joints formed a more compressible foundation and the load that insisted on it was
transferred laterally. The stresses redistribution probably continued to create an
overload situation and the beginning of a crack, which progressed to displace dam
base by a few centimeters. This dislocation is associated with one of the 3 main
leakage of water observed during the event of February 2006.
It is possible that the initial crack then spread to the right bank, involving the
2 bottom outlet pipes that were under pressure. At this point the entire crack went
under pressure causing the internal crack detected in the tunnel inspection.
4. BEDROCK AND DAM BODY MATERIAL INVESTIGATIONS
A wide and complex investigation on the foundation bedrock and on the dam
body materials was performed in 2007 [2] and in 2013 [3] before the final structural
retrofitting design, to complete the data collected in other recent campaigns (1997
[4], 2000 [5]).
Detailed geomechanical surveys in more than 25 areas, drillings with borehole

tests, permeability tests and a very large number of laboratory tests were performed
in order to obtain an accurate reconstruction of the characteristics of the foundation
of the dam.
Many tests were also executed on the original material of the dam: sonic
tomography, georadar, chemical and mechanical laboratory tests (both on masonry
samples and on the mortar component).
Particularly significant were certainly the large-scale direct shear tests per-
formed in a cavity realized at the dam toe in order to determine the tensio-deformative
behavior of the joints between roughened masonry and new cast concrete elements
(representative of the situation foreseen in retrofitting design), Fig. 6. Two different
tests were carried out, changing the cement type in the concrete mix design: the
first one was made using cement CEM TYPE IV/B 32.5 and the second one using
cement CEM TYPE III/A 32.5 N.
Considering the peak of the force-displacement behavior of the joint, the

tests results were interpreted using the Mohr-Coulomb criterion with the following
parameters:
28
Q.105 - R.2
Fig. 6
Large-scale shear tests on the joint between existing masonry and new concrete.
Essais de cisaillement à grande échelle sur le joint entre la maçonnerie existante
et le béton neuf.
◦ Joint CEM TYPE IV/B 32.5 – existing masonry (Cohesion=0.714 MPa, friction
angle=52.41◦ )
◦ Joint CEM TYPE III/A 32.5 N – existing masonry (Cohesion=0.553 MPa,
friction angle= 65.097◦ )
5. STRUCTURAL RETROFITTING
The deep knowledge of the geomechanical characteristics of the rock and of

the materials constituting the dam body, derived from the investigations carried out,
led to the solution that included interventions both on the upstream and downstream
faces of the dam, as well as on the crest. The design solution brings, among other
things, a significant benefit to the potential and functionality of the plant as the
planned increasing of 1 m in the reservoir height will result in a substantial increase
in the availability of water resources. The works designed in this project, the use
of materials with characteristics compatible with those of the current masonry, the
containment of stresses within limits adequate to existing structures, the insertion
of more functional outlet organs, guarantee that the dam, completely retrofitted, can
be considered as a totally renewed one.
To restore the dam’s equilibrium conditions, the project foresees an increase

of the dam mass. The intervention consists of the following main operations:
◦ demolition of the detached portion at the downstream base of the dam;
29
Q.105 - R.2
◦ execution of excavations at the toe of the dam structure;

◦ consolidation grouting in correspondence of the zones on which additional
structures will be founded;
◦ reshaping of downstream and upstream faces. A variable thickness of con-
crete will be cast on the upstream and downstream faces of the existing dam
(Rck=20 MPa). The casting will be performed for segments with open joints.
The slope of the upstream face will be 0.3H/1V below 715.20 m a.s.l. while
that of the downstream face 0.8H/1V below 712.60 m a.s.l.;
◦ reshaping of the dam crest: the crest will be raised by about 4 m up to 721.20
m a.s.l. and will be widened from 4.70 to 6.50 m. Above 712.60 m a.s.l., the
downstream face will have a 0.2H / 1V slope, while above 715.20 m a.s.l. the
upstream face will be vertical;
◦ addition of grout and drain curtain to reduce the seepage and the uplift
pressure acting on the dam;
◦ demolition of existing siphon spillway and construction of new surface
labyrinth spillway.
In the following the main aspects of the project are described.
Fig. 7
Structural retrofitting design.
Conception de rénovation structurelle.
30
Q.105 - R.2
5.1. NEW SURFACE SPILLWAY
The surface spillway enhancement consists in the replacement of the current

siphon overflow with a new free-threshold overflow. The technical solution was stud-
ied with the aim of integrating the new spillway both with the interventions on the dam
structure, in particular with the elevation of the crest of about 4 m (721.20 m a.s.l.),
and with the raising of the maximum regulation of the reservoir to 718.00 m a.s.l.
The free-sill spillway has the advantage of minimizing the necessity of main-
tenance and, also important, of excluding the danger of a partial obstruction of the
inlet by floating material in the reservoir. Compared to the current solution, more-
over, the free-sill spillway eliminates the impulsive and uncontrollable management
of releases that characterizes the siphons.
Considering the complicated structural and topographical situation of Badana

dam, the study of the optimal hydraulic solution recommended the choice of an
unconventional spillway with a labyrinth threshold. This type of works, now widely
used and tested, has the recognized advantage of presenting, with the same capac-
ity, a limited lateral footprint and therefore of being inserted with a minimum impact.
The labyrinth spillway will be built on the left bank in the space obtained from the
demolition of the current siphon spillways. The overflow sill will be placed at an ele-
vation of 718 m a.s.l. The overall width of the labyrinth will be approximately 14.2 m,
approximately equal to that of the original siphon overflow, thus allowing the existing
outlet channel to be reused without major recovery works.
In 2010, the Owner IREN S.p.A., commissioned the Department of Hydraulic,

Maritime, Environmental and Geotechnical Engineering of the University of Padova
(Italy) to create a physical scale model of the new spillway, with the aim of verifying
and optimizing the discharge capacity and the flow conditions, also complying with
Fig. 8
Surface labyrinth spillway: (a) physical model, (b) design.
Déversoir à labyrinthe de surface: (a) modèle physique, (b) conception.
31
Q.105 - R.2
the request for further information made by the Italian Ministry of Infrastructures.
The tests on the physical model allows to acquire useful information about the real
behavior of the labyrinth threshold inserted in the particular topographical context
of the dam. The constructive flexibility of the model made it possible to test different
configurations of the inlet geometry with the aim of improving the flow conditions,
expanding the liquid surface available and decreasing the hydraulic load necessary
in managing the most important flows. The labyrinth draining capacity curve obtained
from the measurements on the physical model was finally used in the calculation of
the design flood lamination with a return period of 1000 years. The results of the
calculation made it possible to define the new maximum reservoir height at 719.15 m
above sea level.
5.2. NEW BOTTOM OUTLET AND TRANSVERSAL TUNNEL
The bottom outlet rehabilitation consists in the replacement of the actual outlet
pipes with a new inspectable pipe and a new flow control and regulation gate.
The new bottom outlet will consist of a single circular pipe with a diameter
of 1.00 m and a total length of approximately 62 m, placed over metal saddles at
the bottom of the transversal tunnel and largely inspectable. The axis height of the
entrance will be set at 670.8 m a.s.l. while the outlet at the bottom of the valley will
be located at an elevation of 668.70 m a.s.l. The intake will be protected from debris
and floating materials by means of a metal grid placed in the new cast concrete
structure.
The bottom outlet interception devices will consist of a gate valve and a
"Howell-Bunger" valve. The gate valve will be installed in the transversal tunnel
just downstream of the intake and will manage the opening and closing of the outlet.
Such valve will allow to stop the flow at the beginning of the outlet, avoiding the
problem occurred in the past to have the entire pipe in pressure when the valve was
close. The "Howell-Bunger" cone valve will be installed at the outlet of the pipeline,
at the base of the dam, and will regulate the discharged flow. The maximum dis-
charge capacity is approximately 14 m3 / s corresponding to a reservoir level equal
to the maximum regulation elevation (718 m a.s.l.). Operating at maximum capacity
(i.e. with the cone valve fully open and considering an initial reservoir level of 718 m
above sea level) the results of the hydraulic test show that the overall duration of the
rapid drawdown is approximately 5 days.
5.3. INSPECTION AND DRAINAGE UPSTREAM LONGITUDINAL TUNNEL
The inspection and drainage longitudinal tunnel of the upstream face is entirely
obtained in the casting of the new abutment and will extend for about 200 m. With
32
Q.105 - R.2
this updated geometry, the tunnel becomes accessible not only from the transversal
tunnel at the base of the dam, in its central area, but also from two entrances located
at the lateral ends, beyond the elevation of 700 m a.s.l. The spillway block was
therefore also modified compared to the previous version, in order to accommodate
the drainage tunnel.
The dimensions of the tunnel are equal to 2.5 m in width and 3.55 m in height,
with the exception of the section below the spillway and the end segments of the
right shoulder, where they are reduced to 0.90 m in width and 2.0 m in height.
5.4. INSPECTION AND DRAINAGE DOWNSTREAM LONGITUDINAL TUNNEL
At the downstream dam toe, obtained in the new cast concrete, an inspection
and drainage tunnel is provided, directly accessible from the external square and
connected to the transversal tunnel in the center of the dam.
The tunnel will have parallel alignment to the dam axis and will extend for about
50 m over the entire width of the dam. Net dimensions will be approximately 2.5 m
in width and 3.55 m in height.
5.5. SEEPAGE CONTROL IN DAM BODY
According to the current legislation, it is considered necessary to equip the

dam with a drainage curtain able to collect flows from the upstream face, creating
a screen of vertical drains of 200 mm in diameter spaced no more than 2.5 m from
each other (drainage pipes).
This device can be easily realized by inserting sub-vertical ducts in the casting
of the new upstream concrete face.
5.6. SEEPAGE CONTROL IN FOUNDATION
The original dam does not have a grout curtain or drain for capturing and
discharging the seepage of waters in the foundation. The modern practice of con-
struction of dams usually provides the treatment of the rock and the presence of a
drain curtain.
5.6.1. Grout curtain
The project involves the construction of a grout curtain along the upstream toe
of the dam, inside the drainage longitudinal tunnel, right up to the abutment. The
curtain will be made according to the indications that will emerge from the preliminary
tests carried out in the test field and will be obtained with holes extended to a depth
33
Q.105 - R.2
of about 40 m below the level of the tunnels. The holes will be both vertical and
inclined, in a single row.
5.6.2. Drain curtain
The project includes the construction of one or two drain curtains drilled,
respectively, from the upstream tunnel, inclined by 30˚ towards downstream, and
from the downstream tunnel, on a vertical plane. The holes will be drilled with a
diameter of 200 mm at a distance of 2.5 m in parallel planes (vertical or inclined)
such as to form a continuous curtain. The length of the drains will be approximately
15 m.
The drains will collect the seepage flows in the foundation rock and will deliver
the waters to their respective drainage tunnels. Finally, the water will be collected in
the transversal tunnel in the center of the dam and will subsequently flow out of the
dam to be measured.
6. NUMERICAL MODEL
The complexity of the phenomenon to be analyzed required the use of spe-

cific software capable of analyzing in detail all the aspects associated with works
conceived.
A two-dimensional model was developed with the finite element software

DIANA [6], by means of which it was possible to investigate all the following aspects:
◦ construction by casting phases with generation of concrete segments and

evaluation of curing times;
◦ evolution of the mechanical properties of concrete as a function of time during
the hardening phases;
◦ thermal analyses such as to consider both the seasonal variation of the exter-
nal temperature and the hydration heat developed by the concrete during the
hardening phases;
◦ evaluation of shrinkage and viscosity deformations for the control of cracking
in the phases immediately following the casting of the concrete segment;
◦ evaluation of the interaction between the new cast concrete and the exist-
ing masonry structure, by means of the definition of appropriate interface
elements such as to simulate, in terms of strength and stiffness, the behav-
ior of the joint. The interface behavior was modeled based on the results
of the large-scale direct shear tests provided on the dam and presented in
Chapter 4.
34
Q.105 - R.2
Fig. 9
Badana Dam – Finite element model adopted for the structural analyses.
Barrage de Badana - Modèle par éléments finis adopté pour les analyses
structurelles.
In Fig. 9 is shown the finite element model adopted for the structural analyses,
where it can be noted the rocky foundation (grey color), the existing dam (yellow
color) and the new cast concrete in which the longitudinal inspection tunnels are
included.
The study of the problem with such complex finite element model allowed
to conduct parametric studies aimed to evaluate the best possible configuration,
considering the following aspects:
◦ effect of the different possible concrete mix design on its behavior during
the hardening phases and, consequently, on the fresh concrete / existing
masonry interaction;
◦ evaluation of the effects generated by the different curing time that elapses
between casting one segment and the following one;
◦ evaluation of the best construction solution, in term of concrete segment
dimension and construction sequence in order to optimize the curing phase
of the concrete together with the duration of the entire retrofitting intervention,
Fig. 10.
35
Q.105 - R.2
Fig. 10
Badana Dam – Thermal analysis in different construction phases.
Barrage de Badana - Analyse thermique dans différentes phases de construction.
7. CONCLUSIONS
This paper analyzes the Badana Dam case study. Badana Dam was built in
1914 and heavy damaged in 2006. The strategic importance of such reservoir as
water resource for Genoa municipality, made necessary a retrofitting intervention
on the dam in order to restore the reservoir fully operation.
One of the main aspects of the proposed retrofitting intervention consists in

the reshaping of downstream and upstream faces together with the reshaping of the
dam crest that also allow to increase the maximum reservoir level. Such reshaping
is made with new cast concrete and allow to preserve the parts of the existing dam
not affected by damage in the 2006 event.
The proposed intervention has been deeply evaluated by means of finite ele-
ment analyses and in-situ investigation, in particular to understand the interaction
between the existing masonry dam and the new cast concrete.
Finally, compared with the demolition and rebuilding of the entire dam, the
proposed design allows to strongly reduce the construction time and the cost of
the intervention due to the lower amount of material needed to fully restore the
reservoir. Moreover, also from the environmental point of view, several advantages
can be highlighted as a consequence of the lower movement of material inside the
construction site.
ACKNOWLEDGMENTS
The authors would like to thank IREN S.p.A. that allowed the publication of
this contribution. The authors would also like to thank all the people involved in the
36
Q.105 - R.2
project in the last almost 15 years, first Piero Sembenelli, all the colleagues in SC
Sembenelli Consulting and then RINA Consulting (Giuseppe Sembenelli, Luciano
Barbaro, Ugo Polettini, Antonella Ruffini, Marco Scarella, Gianluca Gatto, Fabrizio
Cassone, Francesco Cecchini, Raffaella Pirola, Andrea Dal Fabbro, Marco Albano,
Angelo Farnetano, Giuseppe Silvestri, Alessandro Somma), and the University of
Padova for the development of the physical model of the labyrinth spillway. Last
but not least, the authors would also like to thank DIANA FEA BV for the precious
support given in the definition of the numerical model and, more in general, during
years of fruitful cooperation.
REFERENCES
[1] http://dati.mit.gov.it/catalog/dataset/grandi-dighe-italiane
[2] G.E.T. S.r.l., 2007. Interventi urgenti per la messa in sicurezza della Diga
di Badana - Studi geologici, indagini geognostiche e prove geotecniche.
Relazione Finale per Mediterranea delle Acque S.p.A.
[3] Sial.tec Engineering S.r.l., 2013. “Diga di Badana (AL) – Prove di taglio su
giunti calcestruzzo calce idraulica”. Relazione Finale per Mediterranea delle
Acque S.p.A.
[4] G.E.T. S.r.l., 1997. Studio dell’assetto geologico, strutturale, idrogeologico e

delle caratteristiche geomeccaniche relative all’ammasso roccioso che ospita
la diga del Lago di Badana. Relazione tecnica per Acquedotto De Ferrari
Galliera S.p.A.
[5] Università di Genova, 2000. Studio sulle condizioni statiche e funzionali di

cinque opere di sbarramento con riferimento alle caratteristiche strutturali
delle dighe stesse e alle caratteristiche geomeccaniche dell’ammasso roc-
cioso che costituisce il sedime. Convenzione 280/99 DISEG-Acquedotto De
Ferrari Galliera S.p.A.
[6] DIANA FEA BV, 2019. DIANA User’s Manual - Release 10.3.
37
Q.105 - R.3
-------
-------
THE ITALIAN EMERGENCY PLANNING FOR LARGE DAMS IN CASE OF

SEISMIC AND FLOODING HAZARD (∗)
Maria Cristina BRAMATI

ENEL GREEN POWER ITALIA srl – dams & civil infrastructures safety
Federica DEL GIZZI

MINISTERO DELLE INFRASTRUTTURE E TRASPORTI – direzione generale per
le dighe e le infrastrutture idriche ed elettriche
Francesco DOLCEAMORE
ENEL GREEN POWER ITALIA srl – Global Power Generation – Southern Central
Area – Water Management
ITALY
SUMMARY
This paper describes the evolution of the Italian regulation to inform and pro-
tect population in the areas surrounding the large dams from seismic and hydraulic
hazard.
The Civil Protection Document of a dam represents the framework to define a

specific Dam Emergency Planning for the population living in the areas downstream
large dams.
On July 8, 2014, a new regulation of the Council of Ministers was issued,

and it was implemented into the Civil Protection Document. The organization of the
document was changed. As major change, the document includes two sections: one
for the “risk coming from dam” which deals with the effect of an earthquake or an
∗ Plan d’urgence pour les grands barrages italiens contre les risques sismiques et
hydrauliques
38 DOI: 10.1201/9781003211808-49
Q.105 - R.3
artificial flood, and another for the “hydraulic risk for the downstream areas” which
deals with the flooding hazard.
Inside the Civil Protection Document the authorities that are involved for the
information of the population (i.e. supervisory authority such as General Direction for
Dams, operators/owners of the dam, local authorities and prefectures) are defined.
The Civil Protection Document, also, establishes the functional and procedural links
between the various parties involved in the preparation, activation and implementa-
tion of actions aimed to guarantee the safety of the dams and limit the downstream
hydraulic risk.
This paper also describes, as a case history, the preparation and approval pro-
cedure of the Civil Protection Documents of the dams of the Campotosto reservoir,
which represents the most important water reservoir in Italy, for both hydroelec-
tric power generation and water resource, with 218 million of cubic meters storage
capacity.
RÉSUMÉ
Ce rapport décrit l’évolution de la réglementation italienne pour informer et pro-

téger la population dans les zones de grands barrages, contre les risques sismiques
et hydrauliques.
Le Document de Protection Civile d’un barrage représente le cadre pour définir

un Plan d’urgence de barrage spécifique pour la population vivant dans les zones
en aval de grands barrages.
Le 8 juillet 2014, un nouveau règlement du Conseil des Ministres a été publié,

et il a été mis en œuvre dans le Document de Protection Civile. L’organisation du
document a été modifiée. Le changement le plus important, le document comprend
désormais deux sections : une pour le risque provenant du barrage qui traite de
l’effet d’un séisme ou d’une crue artificielle, et une autre pour le risque hydraulique
pour les zones en aval qui traite du danger d’inondation.
Dans ce Document, les autorités impliquées dans l’information de la popu-

lation (c’est-à-dire les autorités de contrôle telles que la Direction Générale des
Barrages qui est l’autorité de contrôle, les exploitants / propriétaires du barrage, les
autorités locales et les préfectures) sont définies. Il établit également les relations
fonctionnelles et procédurales entre les différents intervenants impliqués dans la
préparation, l’activation et la mise en œuvre des actions visant à garantir la sécurité
des barrages et à limiter le risque hydraulique en aval.
Ce rapport décrit également, à titre d’exemple, l’élaboration et la procé-

dure d’approbation des Documents de Protection Civile des barrages du bassin
39
Q.105 - R.3
de Campotosto, qui est le plus important d’Italie, tant pour la production d’énergie
hydroélectrique que pour les ressources en eau, avec 218 millions de mètres cubes.
1. INTRODUCTION: REGULATORY FRAMEWORK FOR CIVIL

PROTECTION IN THE FIELD OF LARGE DAMS
The historical regulatory context for large dams concerning civil protection
starts from 1986, after the disaster of the Val di Stava in July 1985 which caused
the death of 268 people. It can be summarized by Decrees of the President of the
Republic (the Italian acronym is DPR) and Circulars of the Ministry of Public Works
(the Italian acronym is LL.PP.).
In detail, the reference framework is as follows:
1986 Circ. LL.PP. 1125 acoustic signaling devices, warning signs, artificial
flood propagation studies for unloading maneuvers;
1987 Circ. LL.PP. 352 Dam Operation and Maintenance Document (Italian
acronym is FCEM) with the attached document containing the conditions that must
occur for the civil protection system to be activated and the procedures to be
implemented; artificial flood propagation studies for hypothetical dam collapse;
1995 Circ. P.C.M 22806 Technical recommendations for the mapping of areas
at risk of flooding following maneuvers of the discharge organs or hypothetical col-
lapse of the dams; determination of the maximum flow in the downstream riverbed
area;
1996 Circ. P.C.M. 7019 – Provisions relating to civil protection activities within
the basins where dams are located - Civil Protection Document;
2004 Directive P.C.M.27/2/2004 - Operational guidelines for the organizational

and functional management of the national, state and regional alert system for
hydrogeological and hydraulic risk for civil protection purposes.
Considering the alert criteria, this document establishes the activation condi-
tions of alert phases for the safety of dams and risk management for the valley areas
and defines the actions consequent to the activation of alert phases in the case of
events and scenarios, feared or in progress, relevant for the alert and activation of
the Civil Protection System.
The Civil Protection Document, also, establishes the functional and procedural
links between the various parties involved in the preparation, activation and imple-
mentation of actions aimed to guarantee the safety of the barriers and prevent the
downstream hydraulic risk.
40
Q.105 - R.3
In this document are identified the administrations responsible of the emer-

gency plans to contrast dangerous situations connected with the propagation of
a flood wave originating from maneuvers of the unloading organs or from the
hypothetical collapse of the dam.
2. CIVIL PROTECTION DOCUMENT
The Civil Protection Document (the Italian acronym is DPC) is a specific doc-
ument for each dam, which is developed by the Control Authority (Ministry of the
Infrastructure – General Direction for Dams), together with the hydraulic authority in
charge of the downstream riverbed, the operator and the regional civil protection,
and approved by the prefecture. This document includes the directives to implement
actions for the protection of the population in the areas where the large dams are
located.
The document is organized in such a way as to clearly define which actions

are necessary to guarantee the dam safety for the population, and who are the
subjects involved into the preparation and implementation of these actions. It also
defines the interactions between the parties involved, in terms of communications
and technical-administrative procedures.
After the DPCM was issued on 8 July 2014, some aspects and fields of applica-
tion of the document have been expanded and specified better than in the previous
regulations/prescriptions, in particular with reference to the types of risk consid-
ered, the description of the relationships between the subjects involved and the
assignment of specific tasks to certain technical figures such as the Responsible
Engineer.
The risks associated with large dams are divided within the DPC into two
types: dam hazard and downstream hydraulic hazard. The first concerns the possi-
ble events involving the dam or part of it, which may have an impact on the safety
of the dam itself and of the territories downstream of it. The 2014 Directive included
seismic events, in addition to flood events, among those that start the dam hazard
warning.
The downstream hydraulic hazard concerns the activation of the dam dis-
charges, which causes flows in the downstream riverbed that can lead to artificial
flood waves and flooding hazard.
The Civil Protection Document also represents, together with the hydraulic
studies of propagation of artificial floods concerning the specific river (due to the
collapse of the dam or the opening of the discharges), the reference framework
for the definition of the Emergency Plan for the territories downstream that may be
affected by the presence of the dam.
41
Q.105 - R.3
2.1. ORGANIZATION OF THE DOCUMENT
At the beginning of the document, there are all data relevant to the dam:
the geographical location, the geometric characteristics and the discharges are
shown, together with data relating to the reservoir and the maximum and minimum
authorized water levels.
Then all the authorities and offices (prefectures, regions, provinces and munic-
ipalities) that would be involved in case of dam break and in case of uncontrolled
opening of the discharges are listed.
The two flow rates are then provided which define two limit conditions for the
downstream riverbed:
- Q a max , which represents the maximum flow rate that can be contained into
the downstream riverbed, that means, which does not cause flooding down-
stream. This flow rate value is identified by the Operator through a specific
hydraulic study, and is validated by the hydraulic authority.
- Qmin , which is defined as dam discharge caution flow, which corresponds
to the minimum discharge from the dam above which critical situations may
occur along the downstream riverbed. This flow rate value is established
by the hydraulic authority on the basis of the particular characteristics and
conditions of the river.
Incremental flow thresholds (Q) are also established, so that if they are
reached, the dam Operator is required to give specific communications to the Civil
Protection, the hydraulic authority, the involved prefectures and the dam control
authority.
After having provided all the above information, the DPC deals separately with
the two types of risk mentioned above, dividing them into different phases.
The responsibility for activating each of the alert phases provided for in the doc-
ument is in charge of the dam operator. For this reason, the Operator is required to
adopt an organization capable of fulfilling all communication and event management
duties, even in emergency conditions.
In details:
1. Dam Hazard
For the Dam Hazard, the phases of Early Warning, Reinforced Surveillance,
Danger and Collapse of the dam are defined.
The Early warning phase is activated by the Operator, in case of adverse

hydrometeorological conditions, when these conditions determine a certain water
42
Q.105 - R.3
level in the reservoir, which often coincides with the maximum regulation, and the
weather forecast is worsening. There is a specific department of the Region, the
Functional Center, which has the task of monitoring the situation along the water-
courses of its competence and issuing alerts for hydro-meteorological criticalities.
The activation of the Early Warning phase entails for the Operator the immediate
communication of certain information (e.g. activation of the phase, reservoir level
and discharged flow) to a list of authorities and entities affected by the event, defined
within the DPC. The communication must take place using a predefined template
attached to the DPC.
In case of an earthquake, the Early Warning phase is activated when the

magnitude of the event is such as to activate the execution of immediate controls on
the dam. Also in this case, the Operator is required to promptly notify the involved
subjects of the activation of the phase and of the presence / absence of anomalies
and failures on the dam. Also in this case specific sheets are used for communication.
Following the communication from the Operator, the Supervisory Authority

and the alerted offices notify all the other parties involved in this phase, with precise
methods defined in the DPC.
The Reinforced Surveillance phase is activated when the increasing water in

the upstream basin could cause exceeding of the maximum reservoir level, or when
inspections on the dam reveal anomalous behavior of the structures, or additionally,
in case of an earthquake, when the results of immediate post-event controls have
revealed minor damages to the dam.
In this phase the Operator must implement actions for the management and
containment of the event in progress, and communicate with the other parties
involved, which are defined in the specific paragraph of the DPC. Following the
communications from the Operator, the responsible Civil Protection informs and
coordinates the involved offices and alerts the downstream territories for the possible
implementation of the Emergency Plans.
The Danger phase is activated when irreversible damage is highlighted (also

due to an earthquake), failures, cracks or anomalous movements of the dam, or
landslides on the banks, the evolution of which can compromise the water regulation
function of the dam. The Operator must notify the subjects involved of the activation
of the phase, and must keep them informed with an established frequency depend-
ing on the evolution of the situation. The Civil Protection will coordinate the local
authorities for the implementation of emergency plans, and all the subjects involved
will be alerted and kept informed according to precise provisions. At the end of the
danger phase, the Responsible Engineer for the safety of the dam will have to submit
a technical report describing the event and the actions taken to contain it.
The Collapse phase is activated when the Operator detects the beginning of
a dam break, even if only partial, or damage that can lead to uncontrolled release of
water and risk of loss of lives. All involved subjects coordinate for the implementation
43
Q.105 - R.3
of emergency plans, according to precise provisions. The parties involved include

the operators of any downstream dams that may be affected by the event.
2. Downstream hydraulic hazard
For the Downstream Hydraulic Hazard, only the Early Warning and Warning
phases are defined.
The Early Warning phase for hydraulic hazard is activated by the Operator
upon the arrival of a flood event that determines the opening of the dam discharges.
The Operator notifies the parties involved, specifying the conditions of the reservoir.
The hydraulic risk Warning phase is generally activated by the Operator when
the Qmin flow rate threshold set for the downstream watercourse and specified in the
document is exceeded. Communications between the parties involved are always
regulated in the DPC.
3. CASE HISTORY – CAMPOTOSTO BASIN
3.1. SITE DESCRIPTION, USE AND IMPORTANCE OF THE RESERVOIR
Campotosto basin is located in Central Italy, between the provinces of L’Aquila,

Rieti, Teramo and Ascoli Piceno, within the Gran Sasso – Della Laga Mountains.
Campotosto basin was the site of an ancient lake, which later became a peaty
swamp. The Terni Company for Industry and Electricity identified its potential as a
possible site for an artificial lake, at the head of a system of hydroelectric exploitation
of the water of the Vomano river. The reservoir was therefore built thanks to the
construction of three different dams: Poggio Cancelli dam on the Rio Castellano, a
tributary of the Tronto river, Rio Fucino dam on a tributary of the Vomano river, and
Sella Pedicate dam. It initially had a total capacity of 153 million cubic meters and a
maximum water level of 1312.50 m asl; the capacity was raised, in the years 1964–
1971, to 218 million cubic meters, with a maximum regulation level of 1317.50 m asl,
thanks to the structural raising up of the three dams.
The main water supply for the lake is provided by two gutters, about 20 km long
each, which run in an underground tunnel along the eastern and western slopes of
the Monti della Laga, and which intercept the waters of other watercourses.
The hydroelectric power plant scheme of the Vomano river, where the Campo-
tosto lake represents the head reservoir, is made up of other two dams (in addition to
the three of the lake); two additional daily modulation basins and four hydroelectric
power plants, with a total installed power of over 700 MW and a producibility of
around 600 GWh/year.
44
Q.105 - R.3
The strategic importance of the Campotosto reservoir has progressively

increased over the years, combining its hydroelectric power generation with drinking
and irrigation water uses, especially in periods of poor rainfalls, which have become
increasingly frequent due to climate changes.
The lake is also a resource for local tourism, and in fact there are numerous
sports and recreational activities carried out around the lake and on its shores.
To further highlight the environmental value of the site, it is to underline that the
Campotosto lake hosts a natural protected area (SIC IT7120201) and is included in
the perimeter of the Gran Sasso and Monti della Laga National Park.
3.2. REVISION OF THE DPC ACCORDING TO DPCM 8/7/2014 – PREMISES
Campotosto dams are operated on the basis of a specific Operation and Main-
tenance Condition Document for each dam, which establishes their management
and operation methods. Furthermore, each of them is associated with a specific Civil
Protection Document (DPC), which, in the first draft, was drawn up according to the
provisions of the Circular P.C.M. dated March 19, 1996 and approved by Prefecture
of L’Aquila in 2001, with the task of establishing the activation methods and related
actions in various emergency situations.
With the issue of the PCM Decree dtd 8/7/2014, the Presidency of the Coun-
cil of Ministers provides new operational guidelines on civil protection activities in
the basins where large dams are located, establishing the conditions for activating
the emergency phases to guarantee the safety of dams and downstream territories.
Therefore, the actions resulting from the activation of the different phases for alerting
and activating the Civil Protection system are defined, the functional and procedu-
ral links are established between the various subjects involved in the preparation,
activation and implementation of actions aimed at guaranteeing dam safety and
minimizing the hydraulic hazard downstream. Moreover, the institutions officially in
charge of the preparation of emergency plans are identified to contrast the danger-
ous situations connected with the propagation of a flood wave, caused by opening
of the discharges or by a hypothetical dam break.
The activity was promoted by the Italian Dam Authority (i.e. Ministry of Infras-
tructure – General Direction for Dams) by sending the drafts of the new DPCs to the
local authorities (Prefecture, Public Works Offices, Civil Protection, etc ...) and to
the operators of large dams after the release of such directive.
In August, October and November 2016, and later in January 2017, a part of
Central Italy was affected by a series of severe seismic events; totally the sequence
has counted more than 65,000 events, including 2 earthquakes with magnitude
higher than 6 and at least 7 with Magnitude higher than 5 – both on Richter scale.
45
Q.105 - R.3
Such seismic sequence, contributed to increase an emotional state charac-

terized by strong tensions and concerns in the Authorities in charge of managing
the population in the affected territories. The seismic events of January 2017, more-
over, occurred together with an exceptional meteorological situation, characterized
by snowfall in the area affected by seismic crisis, with snow packs of over 2 meters
in height, which had made it extremely difficult to travel, causing great difficulty in
reaching the dam sites.
In this context, drafting of the new DPC of the Campotosto dams started.
The participants were, in addition to the Dam Authority and the Operator, the
Abruzzo Region, the Prefectures of L’Aquila and Teramo, the Public Works Offices
of L’Aquila and Teramo, the Regional Civil Protection, the Fire Brigade and also
representatives of the Faculty of Engineering of the University of L’Aquila.
The kick-off meeting of the working group was held on February 1, 2017.
3.3. THE WORKING GROUP FOR DRAFTING THE NEW DPC
The purpose of the working group was to complete the updating of the Civil
Protection Document of the Campotosto dams starting from the scheme proposed
by the Dam Authority, that was a preparatory document for the development of the
Dam Emergency Plan in charge on the Region.
First, all the potential emergency scenarios caused by the Campotosto basin
were identified and studied according to the new DPCM 8/7/2014, considering two
specific risks: the dam hazard, that is the risk mainly linked to the structural capacity
of the dam, and the downstream hydraulic hazard, that is the risk caused by the
release of water on the downstream territories.
The existing documents on the Campotosto dams were initially analyzed con-
sidering these new criteria; then, the group proceeded to read again the Dam
Condition Documents, the existing Civil Protection Documents, the flood wave prop-
agation studies for discharge opening and dam break, which dated back to 1992,
and then continued with the dam structural assessments, drawn up after the seismic
sequence of 2009.
The Operator updated all the studies, re-elaborating also those for the
downstream dams, directly influenced by the Campotosto basin.
At the same time, the Operator updated the hydrologic-hydraulic studies to

identify the acceptable flow rates downstream in case of discharge opening con-
tained in the natural riverbeds (Qamax ); for these studies was also necessary a deep
and updated knowledge of the territory that the Operator did not hesitate to promote.
46
Q.105 - R.3
The drafting of the DPCs came together with an important study on the seismic
re-evaluation of the three Campotosto large dams, and a study, performed by the
University of Camerino, aimed at the "Identification and Parameterization of Active
Faults of the Campotosto area” which best defined the position and potential of the
tectonic structures of the Campotosto area.
The activity of the working group lasted for several months. The Civil Protection
Documents of the three Campotosto dams, in their new version, were approved by
the Prefecture of L’Aquila in period May 2018 - March 2019.
ACKNOWLEDGEMENTS
The Authors thank the Director of the General Direction for Dams, Eng.
Angelica Catalano, for the suggestions provided in the final draft of the paper.
REFERENCES
[1] Direttiva del Presidente del Consiglio dei Ministri 8 luglio 2014 “Indirizzi oper-
ativi inerenti l’attività di protezione civile nell’ambito dei bacini in cui siano
presenti grandi dighe”
[2] Prefettura U.T.G. di L’Aquila, Protezione Civile della Regione Abruzzo. Doc-
umento di Protezione Civile (Direttiva P.C.M. 8 luglio 2014) Diga di Poggio
Cancelli (n arch. 392A).
[3] Prefettura U.T.G. di L’Aquila, Protezione Civile della Regione Abruzzo. Docu-
mento di Protezione Civile (Direttiva P.C.M. 8 luglio 2014) Diga di Rio Fucino
(n arch. 392B).
[4] Prefettura U.T.G. di L’Aquila, Protezione Civile della Regione Abruzzo. Doc-
umento di Protezione Civile (Direttiva P.C.M. 8 luglio 2014) Diga di Sella
Pedicate (n arch. 392C).
47
Q.105 - R.4
-------
-------
TOWARDS AN INTERNATIONAL HANDBOOK FOR EMERGENCY RESPONSE

TO FLOOD RISK (∗)
Bart VONK
Coordinating advisor Flood Risk Management, Rijkswaterstaat, MINISTRY OF
INFRASTRUCTURE AND WATER MANAGEMENT
Wout de VRIES
Senior specialist Flood Risk Management, Rijkswaterstaat, MINISTRY OF
Eric van KUIJK

Senior advisor Hydraulic Engineering, Rijkswaterstaat, MINISTRY OF
Marcel BOTTEMA
Senior advisor Flood Risk Management, Rijkswaterstaat, MINISTRY OF
Ludolph WENTHOLT
Program coordinator Flood Risk Management, STOWA FOUNDATION FOR
APPLIED WATER RESEARCH
Eric HUIJSKES
Senior advisor Flood Risk Management, HUIJSKES ADVISORY SERVICES
NETHERLANDS
SUMMARY
Dams, levees and other flood defenses are important elements of flood risk
management. However, there is always a possibility that such defenses will fail, with
∗ Vers un manuel international pour les interventions d’urgence en cas de risque d’inondation
48 DOI: 10.1201/9781003211808-50
Q.105 - R.4
great or even catastrophic impact. Dam and levee managers will do all they can to
prevent failure in the case of an immediate flood threat, and emergency measures
can play a key role in this.
The challenges of such flood events may become so large that they require
transboundary involvement, either because the flood impact area reaches across
borders, or because the flood is so catastrophic that transboundary assistance is
needed. In such cases, effective emergency response depends on implementing
the right measures, and doing that correctly. In order to ensure effective emer-
gency management, it is critical that all involved speak the same “emergency
response language”, use compatible techniques and approaches, and follow similar
procedures.
To achieve this, we launched an initiative to develop an International Handbook

for Emergency Response to Flood Risk. It will provide good practices, will share
lessons learnt from recent floods all over the world and will be an invaluable tool to
help train flood fighting staff. Dam and levee managers both will benefit from such
a tool.
The Handbook will heavily draw on practical experience, both from actual flood-
fighting situations and from calamity exercises. The European project Living Lab
Hedwige and Prosperpolder (Interreg5 Polder 2C call) will create a unique opportu-
nity for such exercises and full-scale experiments, helping to define and refine good
emergency response practices. Suggestions about similar facilities and learning
opportunities are more than welcome.
The present ICOLD Congress Paper outlines the plans and contents for such
an International Handbook for Emergency Response to Flood Risk. The 2021 ICOLD
Congress is also a key opportunity to discuss these plans and contents with the dams
community, and to receive their suggestions and inputs, all the more so since Dam
Incidents and Accidents are a central Congress Theme.
The writing phase of the Handbook is to start soon after the ICOLD Congress,
but there is still a possibility to join the writing team at a later moment for people with
the right expertise.
RÉSUMÉ
Les barrages, les digues et les autres structures de défense contre les inon-
dations sont des éléments importants dans la gestion des crues. Cependant, il
existe toujours la possibilité de rupture des tels structures, avec des conséquences
importantes ou même catastrophiques. En cas de menace imminent, les gestion-
naires des digues et barrages feront tout pour éviter des tels ruptures. Des mesures
d’urgence peuvent jouer un rôle clé dans ce cas.
49
Q.105 - R.4
Les défis posés par des grandes crues peuvent être tels qu’ils nécessitent
un déploiement transfrontalier, soit parce que la zone d’impact des inondations
atteint au-delà des frontières, soit parce que l’inondation est si catastrophique que
l’aide transfrontalière est nécessaire. Dans des tels cas, une intervention d’urgence
efficace dépend de la mise en œuvre des bonnes mesures, en faisant cela de
façon correcte. Afin d’assurer une gestion efficace des situations d’urgence, il est
essentiel que toutes les parties concernées parlent le même “ langage d’intervention
d’urgence ”, en utilisant des techniques et des approches compatibles et en suivant
des procédures similaires.
Compte tenu de ce besoin, nous avons lancé une initiative visant à élaborer un
Manuel international d’intervention d’urgence en cas d’inondation. En fournissant
des bonnes pratiques et en partageant des leçons tirées des récentes inondations
dans le monde entier, ce manuel sera un outil important pour aider à former le per-
sonnel de lutte contre les inondations. Non seulement les gestionnaires de digues
bénéficieront d’un tel outil, mais également les gestionnaires de barrages.
Le Manuel s’appuiera fortement sur l’expérience pratique, à la fois des situa-

tions réelles de lutte contre les inondations, et des exercices de calamité. Le projet
européen Livinglab Hedwige et Prosperpolder (appel Interreg5 Polder 2C) créera
une occasion unique pour de tels exercices et expériences à échelle réelle, aidant à
définir et affiner les bonnes pratiques au sujet des interventions d’urgence. Des sug-
gestions concernant des installations et des possibilités d’apprentissage similaires
sont les bienvenues.
La présente publication décrit nos projets et nos idées préliminaires pour un tel
manuel International pour l’Intervention d’Urgence en cas d’Inondation. Le présent
Congrès de la CIGB est également une occasion importante pour avoir des discus-
sions sur ce projet et le contenu du manuel avec la communauté des barrages, et
de recevoir leurs suggestions et leurs contributions, d’autant plus que les incidents
et accidents concernant des barrages sont un thème central du Congrès.
La phase d’écriture du manuel doit commencer peu de temps après le Congrès

de la CIGB, mail il existe encore la possibilité rejoindre l’équipe d’écriture à un
moment ultérieur pour des personnes ayant la bonne expertise.
1. INTRODUCTION
Flood risk can be managed by using a mix of strategies, such as protection

(flood defenses), spatial planning and flood emergency response. The Associated
Program on Flood Management (APFM), a joint initiative of the World Meteorological
Organization (WMO) and the Global Water Partnership (GWP), provides guidance
and promotes the concept of Integrated Flood Management (IFM) see WMO (2009,
2011). Besides Dam Safety (ICOLD, 2017a), ICOLD promotes the approach of
50
Q.105 - R.4
Integrated Flood Risk Management (ICOLD, 2014) as well, while sharing expe-
riences on Dam Incident Management. In practice, absolute protection against
flooding is rarely feasible. Therefore, additional reliance on emergency response
is often essential.
Worldwide, there are different emergency response approaches to minimize

flood risk. Experiences with actually implementing this response vary widely which
provides an excellent opportunity to learn from each other. This mutual learning
should not be restricted to dam and levee experts only. Firstly, emergency manage-
ment includes a much larger range of actors and stakeholders than these experts
alone. Secondly, the flood risk in river basins is often influenced by both dams and
levees, and their state and operation.
International relationships on flood emergency response have developed

recently. This proved to provide benefits to all organizations involved. In line with
this development, Dutch parties are keen to promote an initiative to join forces by
writing an International Handbook on Emergency Response Measures for Flood
Risk and hope that other parties will join this initiative. This handbook could for
example mirror the International Levee Handbook (CIRIA, 2013), while focusing on
emergency response aspects but including dam and levee experts in its target group.
This paper is a first invitation to join forces to do so and present our initial thoughts,
while inviting the communities on dams and levees to share inputs and ideas about
scope, structure and contents of the Handbook.
2. THE HANDBOOK
The first step in developing this International Handbook is to produce an anno-

tated table of contents. The underlying aim is to identify links to existing and potential
users and user communities, to investigate their needs and wishes, and how these
may be addressed. Such needs and wishes not only refer to the contents, but also
to the way the contents are derived (process) and how these will be shaped and
presented (e.g. web-based, hardcopy).
With those involved, we aim for consensus as to what is required, making

sure we meet the needs of the target audience. We also want to make sure that
those involved understand each other, work along commonly agreed lines and share
inspiration and good practices from various countries around the world.
2.1. WHY DO WE NEED A HANDBOOK?
Flood emergency response is a complex issue. Actual flood events have

triggered the development of new concepts and tools. These cover both crisis
management and technical aspects, such as emergency response measures.
Although numerous initiatives have been taken to exchange experiences and
51
Q.105 - R.4
develop coherent approaches, there appears to be no single international initiative

that provides an integrated overview of flood emergency response.
Furthermore, the scale of especially major flood events may require countries
to help each other, even when such a major flood remains within the borders of a
single country. The diversity, rareness and impact of such major events all increase
the need to learn from each other. Both when helping each other or when learning
from each other, it is extremely important that all involved speak the same “emer-
gency response language”, use compatible techniques and approaches, and follow
similar procedures.
Finally, there are various innovations and good practices across the globe
that can be shared. By showcasing such practices and innovations, others can be
inspired and can benefit from valuable lessons learnt elsewhere.
Hence we feel there is ample basis to jointly write an International Handbook.

Our ambition is to write a Handbook that addresses both the so-called “cold phase”
(preparation) and the “warm phase” (immediate flood threat and response), thereby
making a step beyond presently existing good practice documents like (JRC, 2019).
2.2. FOCUS
It is essential to define the type of flood threats and types of flood defenses to be
addressed in the Handbook. Defining the flood threats helps to identify and describe
the characteristics that are critical for emergency response: lead times, durations
and potential impacts. Linking these to the types of flood defenses will provide a
clear picture regarding the infrastructure that requires emergency response, as well
as feasible and realistic responses that can be made available.
2.3. FLOOD EMERGENCY RESPONSE MANAGEMENT PRINCIPLES
It is also important to provide the wider context of flood emergency response

management. This covers organizational aspects, including the policies that provide
the basis and how this is translated to staffing, planning and operationalization. The
crisis management cycle will be described in detail, making sure that a framework
is presented with commonly agreed concepts and protocols. The aforementioned
wider context can also include the operation of (upstream) dams before and during
flood events, see also (ICOLD, 2014).
2.4. FLOOD DEFENSE RELATED ASPECTS
Flood emergency response activities are commonly triggered by inspections.

These may vary from damage (e.g. significant erosion or deformation, cracks, pip-
ing, animal activities) to concluding that a warning threshold will be exceeded (e.g.
52
Q.105 - R.4
Fig. 1
Erosion protection measure (landside slope).
Mesure de protection contre l’érosion (Talus côté zone protégée).
overtopping) such that the defense is likely not to withstand the expected threat. The
next step is to identify failure mechanism(s) that may occur. Thereafter appropriate
response measure(s) can be formulated. Examples vary from placing the traditional
sandbag to more innovative or specialist measures.
It should be noted that aiming for a uniform approach to the prevention and
preparedness phase will provide a more sound basis to the inspections. We notice
that efforts to actually define the actual strength of flood defenses in advance are now
becoming more common. Such knowledge will help prioritize and plan subsequent
actions and make the response more efficient and effective.
2.5. PRACTICAL CONSIDERATIONS
The selection of the appropriate emergency response measure in a situation

with immediate flood risk is not easy. Such selections need to be made under high
time pressure, commonly on the basis of limited field data, the availability of materials
and in combination with many other parallel emergency response activities and
choices. Local conditions, weather and the overall demand for response measures
make it important to keep the overall perspective so as to be able to prioritize and
to plan for the logistics in advance.
53
Q.105 - R.4
When implementing measures, health and safety aspects require ample atten-
tion. An injury or incident during the execution of emergency response activities
requires attention and energy of staff, and may well delay or jeopardize the exe-
cution of measures, increasing the risk of flooding. This is all the more important
as adverse working conditions often go together with real emergency situations, as
opposed to fair weather exercises.
2.6. ARRIVING AT AN EFFECTIVE AND EFFICIENT INTEGRATED APPROACH
An emergency measure can only be implemented efficiently and effectively

when all aspects are properly managed. Therefore the Handbook will devote ample
attention to the reliability of the response, including the actual measure taken. This
can be at both strategic and operational levels.
Maturity levels of the emergency response organization can be defined

through a strategic approach. This identifies the lines along which the level at which
an organization can deal with a flood event challenge can be defined, covering both
capabilities and capacities for the cold phase and warm phase.
The operational approach zooms in on the emergency measures themselves.

Well before deploying these measures, it should be assessed how measures con-
tribute to minimize the flood risks and e.g. to meet the nationally adopted targets or
standards.
Considering both the strategic and operational approaches described above

will definitely facilitate the process of mutual learning. In addition, existing practices
such as provided in the EU Civil Protection Mechanism (see Section 4) can be used.
2.7. POSITION OF INTERNATIONAL HANDBOOK ON EMERGENCY RESPONSE
An International Handbook on Emergency Response to Flood Risk could be

published as a separate document, but it could also be published as an additional
appendix to existing documents. A promising candidate for the latter is the Interna-
tional Levee Handbook (CIRIA, 2013). However, there could also be other options,
perhaps even the twinning of the proposed Emergency Management Handbook with
another related document in development.
2.8. PROPOSED OBJECTIVES, THEMES AND STRUCTURE OF HANDBOOK
The objectives overlying the International Handbook for Emergency Response

to Flood Risk can be summarized as follows:
54
Q.105 - R.4
• Describing commonly accepted methods, approaches and proven practices

regarding flood emergency response, while building upon existing frameworks;
• Documenting good practices and providing guidance for mutual learning;
• Identifying relevant knowledge, experiences and tools (as well as gaps).
Figure 2 presents the initial table of contents. Please note that both the
contents and sequence of chapters are open for discussion.
Fig. 2
Initial table of contents.
Table de matières initiale.
3. NOTHING IS CAST IN STONE
The approach to arriving at the International Handbook for Emergency

Response to Flood Risk and the contents of the Handbook are both open to discus-
sion and agreement. In the following we raise some first discussions that in our view
would need to be addressed.
3.1. DISCUSSION 1 – TARGET AUDIENCE
The Handbook will aim at practitioners in the field of flood risk management /
flood disaster risk reduction. In particular it will be of use to the:
55
Q.105 - R.4
• Owners and managers of both flood defenses and water regulating structures
such as dams and (retention) reservoirs;
• Specialists in the field of flood emergencies: water system behavior
(hydraulics, hydrology), geotechnics and civil engineering, coordination and
communication in the various response activities (e.g. implementation of
measures to prevent failure);
• Flood fighters;
• Crisis managers.
The audience includes the public sector (e.g. national, regional and local water
authorities, civil defense actors), the private sector (e.g. consultants, contractors,
suppliers), NGO’s and academics (e.g. research institutes).
Targeting such a wide audience implies a fairly high ambition level. We feel
this is justified since an integrated approach requires the involvement of many target
groups. This also implies that readers should be facilitated to quickly assess which
chapters are relevant to him / her.
Alternatively, focus can be put on practitioners involved in emergency response

for flood defense structures only. This would imply that matters related to e.g.
evacuation are not addressed.
3.2. DISCUSSION 2 – SCOPE: TYPE OF THREATS
The Handbook will address threats like storm surges, river flooding and pluvial
flooding. Regarding river dynamics, it is suggested to cover a broad array of flood
threats, ranging from slower processes (typically in downstream or lowland reaches)
to quicker ones (typically in more upstream reaches) and even flashfloods. The
question is whether to address other types of threats as well, like floods caused by
landslides, earthquakes or tsunamis.
3.3. DISCUSSION 3 – SCOPE: TYPE OF DEFENSES
The following types of flood defenses are suggested to be addressed in the

Handbook:
• Levees;
• Dams and barriers;
• Structures with a flood defense function (e.g. locks, sluices and weirs);
• Demountable and Temporary Defenses.
With respect to dams, developments that have been pursued in an ICOLD

context, as reported in ICOLD (2017a, 2017b), can be addressed too.
56
Q.105 - R.4
Fig. 3
Example Factsheet (in Dutch).
Exemple de fiche descriptive (en néerlandais).
Storm surge barriers are not included yet, as the scale of such barriers requires
fundamentally different emergency response measures. There is, however, some
arbitrariness in this choice and suggestions are always welcome.
3.4. DISCUSSION 4 – HOW TO PRESENT
It is suggested to present the Handbook in a digital environment. As an addi-

tional thought, the Handbook could also be presented in a Wiki type environment on
the internet. This not only allows for easy display of the contents of the handbook,
but also provides a platform that allows for active engagement with the Handbook
community, e.g. to jointly develop and implement new approaches.
It is critical to present the information in a well-structured and accessible way,

that is easy to grasp. In the Netherlands we have started documenting lessons
learned in a factsheet format. It is our experience that this is a good means to make
the lessons learned explicit (see Figure 3). Nevertheless, we are open to other
approaches.
3.5. DISCUSSION 5 – HOW CAN THE HANDBOOK HELP IN OUR CO-OPERATION?
Working towards the Handbook implies that a ‘coalition of the willing’ is to be

established, of which members are able to contribute. It can be stated that such a
Coalition / Community is in itself an objective! By joining forces and jointly document-
ing the various chapters, members are more or less forced to arrive at commonly
57
Q.105 - R.4
agreed approaches and thus to highlight at what is the general consensus regarding
good practices. Such discussions will benefit those involved.
The question is how far such discussions should go. Time is scarce and costly,
requiring those involved to make choices as to how to arrive at the Handbook.
4. REQUIREMENTS TO HELP EACH OTHER
When flood disasters strike, response teams, technical equipment and other
resources need to be deployed in the shortest time possible to support the flood
emergency response efforts. Being well prepared to intervene immediately in a
disaster is a critical factor in saving lives and minimizing damage.
A joint approach helps to prepare for the optimal course of actions, which is
informed by pooled expertise and capacities of first responders. In this way it is
aimed that duplication of relief efforts is avoided, and that the assistance offered
meets the needs of those affected.
Pooling together civil protection capacities and capabilities allows for a

stronger and more coherent collective response. We are convinced that the Hand-
book will be an invaluable tool to assist in such processes. The use of the Handbook
can be combined with mechanisms already in place, such as the EU Civil Protection
Mechanism. Please see the textbox below (see Figure 4) and to the factsheet it
refers to (see European Commission, 2020).
Fig. 4
Example - EU Civil Protection Mechanism.
Exemple - Mécanisme de Protection Civile en Union Européenne.
58
Q.105 - R.4
Fig. 5
Practical Exercise in Switzerland for Flash Flood Advisors (October 2019).
Exercice pratique en Suisse pour des conseillers au sujet de submersion rapide
(Octobre 2019).
Jointly organized training and exercises (see Figure 5) are also a means to
strengthen the capabilities of those involved and develop shared practices.
Fig. 6
Processes linked to the initiative.
Des processus liés á l’initiative présente.
59
Q.105 - R.4
5. ARRIVING AT A COALITION
Developing the Handbook will undoubtedly result in inspiring discussions that

are needed to meet the aforementioned ambitions. This in itself is already an
opportunity to learn from each other.
Another opportunity that can potentially be linked to this initiative is the Euro-
pean project Living Lab Hedwige- and Prosperpolder (Interreg5 Polder 2C call). This
project can offer a significant contribution, as it creates a unique opportunity for full-
scale experiments and exercises in order to connect practice and policy (see also
Figure 6). Potentially, other initiatives like test and training sites or flood fighting crisis
exercises (see Figure 7) in the different participating countries can provide useful
contributions as well.
It will be important to arrive at an efficient organization of the writing team.

For this we propose a project management team to safeguard meeting planning
deadlines and to ensure that the right people are involved at the right moments in
time. This team would typically also include the lead editors. For each of the chapters
we foresee a small team of authors (we propose a maximum of 4) and reviewers. As
mentioned, we would hope there is sufficient interaction between the authors. These
authors should be experienced professionals that have an intimate knowledge of the
topic at hand.
At present we already have a first commitment from several organizations,

such as the US Army Corps of Engineers (USA), the Environment Agency (England),
the Flood Disaster Prevention Division, the River Department National Institute for
Land and Infrastructure Management (NILIM) of the Ministry of Land, Infrastructure,
Transport and Tourism (MLIT) (Japan), and the Technisches Hilfswerk (Germany).
We are keen for other countries / organizations to join us.
Fig. 7
Exercise Waterwolf, Marken, The Netherlands, 2-4 Nov. 2016: placing of Big Bags
by the military.
Exercise Waterwolf (Marken, Pays-Bas), 2-4 Nov. 2016: placement des Gros Sacs
par l’armée.
60
Q.105 - R.4
By lack of external funding, participants need to cover their own costs. How-
ever, the process of writing the Handbook, starting by mid-2021, can probably take
place within two years. With modern communication tools, the actual costs would
be quite limited. In any case, we plan to make budget estimates in the initial phases,
assisting parties in their decision about joining our initiative.
Ideally, this coalition will grow into an active community of practice, which will
remain active after finalization of the Handbook. As mentioned earlier, we feel that
joining forces internationally is a very effective manner to get access to the good
practices worldwide. A long term approach will safeguard this, and probably will
even provide for new opportunities.
6. CONCLUDING REMARKS AND NEXT STEPS
We are convinced that the Handbook can provide an internationally recognized

integrated overview of flood emergency response measures and procedures. In this
paper we have presented some initial thoughts on this initiative, hoping that others
feel inspired to join. We are fully aware that our initial thoughts will need to evolve
as the group of participants grows, and the Handbook takes shape. Also numerous
topics as well as practical aspects related to the Handbook have not been touched
in this paper. In other words: this paper is an open invitation to join this coalition of
the willing.
We hope there will be ample opportunity to discuss the initiative for an Inter-
national Handbook on Emergency Response to Flood Risk with the dams and
levees community, during the 2021 ICOLD Congress in Marseilles. The ICOLD
Congress is also a perfect opportunity to explore links with ICOLD-activities on Dam
Safety, Dam Incident Management and Safety Governance. During the event we
want to gauge interest with respect to the proposed Emergency Response Hand-
book, arrive at an initial coalition based on soft commitment and try to make sure
that we are confident that the coalition has the width and depth to deliver the
Handbook.
The next major step will be organizing a first meeting (likely via video con-
ferencing) to firm up on matters, including commitment. We also acknowledge that
prior to such a meeting various practical aspects need to be clear. Examples are
the shape and form of the Handbook, hosting and publishing.
Before this first meeting, there may be another opportunity to gather input,
which is the June 2021 FloodRisk Conference, where we hope to organize an
interactive workshop. We expect that any ideas and response received during
the ICOLD Congress is still in time to be included as input for this interactive
workshop.
61
Q.105 - R.4
REFERENCES
[1] CIRIA, The International Levee Handbook, London, United Kingdom, 2013.
[2] EUROPEAN COMMISSION, Factsheet on EU Civil Protection Mechanism,

available online: https://ec.europa.eu/echo/what/civil-protection/mechanism,
2020.
[3] ICOLD, Integrated Flood Risk Management, ICOLD Bulletin 156, 2014.
[4] ICOLD, Dam safety management: Operational phase of the dam life cycle,
ICOLD Bulletin 154, 2017.
[5] ICOLD, Management of dam incidents. ICOLD European Club, Working Group
on dam incidents, 2017.
[6] JRC, Science for disaster risk management 2017, Joint Research Centre of
the European Commission, 2017.
[7] SBRCUR, Handboek Uitvoering Dijken. Chapter 7, Utrecht, The Netherlands,

2017.
[8] WMO, Integrated Flood Management – Draft Paper. Geneva, Switzerland,

2009.
[9] WMO, Integrated Flood Management Tool Series. Flood Emergency Planning.
Issue 11, Geneva, Switzerland, 2011.
62
Q.105 - R.5
-------
-------
RISK ASSESSMENT TO INFORM OF THE DECISION-MAKING PROCESS ON

REHABILITATIONS IN RIO HONDO DAM (ARGENTINA) (∗)
Rodolfo DALMATI, Ailín PERTIERRA & Juan Facundo SOUTO

Organismo Regulador de Seguridad de Presas (ORSEP) and UNIVERSITY OF
BUENOS AIRES (UBA), Argentina
Ignacio ESCUDER-BUENO, Adrián MORALES-TORRES &

Daniel CERVERA-MIQUEL
iPresas Risk Analysis, POLYTECHNIC UNIVERSITY OF VALENCIA (UPV) and
SPANCOLD, Spain
Carina R. CABALLERO
RED INGENIERÍA and NATIONAL UNIVERSITY OF COMAHUE (UNCOMA),
Argentina
ARGENTINA
SUMMARY
The Río Hondo dam is a strategic asset in the northwestern region of

Argentina, fundamentally since it guarantees the irrigation of 270,000 hectares. With
50 years in operation, its resistant capacity against seismic actions is questioned
due to liquefiable soils in the embankment dam toe. In addition, a flood in 2017 put its
routing capacity and the performance of spillway and stilling basin in question. This
last incident prompted a risk analysis to define potential risk reduction measures.
In this context, this article presents the risk analysis work carried out to
identify, evaluate and guide future investments to improve Río Hondo dam safety,
through a qualitative and quantitative analysis aligned with the Risk Management
Plan conducted by ORSEP in its regulated dams.
∗ Analyse des risques pour la prise de décision sur l’assainissement du barrage de Río
Hondo en Argentine
DOI: 10.1201/9781003211808-51 63
Q.105 - R.5
RÉSUMÉ
Le barrage de Río Hondo est un atout stratégique dans la région nord-ouest

de l’Argentine, essentiellement puisqu’il garantit l’irrigation de 270 000 hectares.
Avec 50 ans de fonctionnement, sa capacité de résistance aux actions sismiques
est remise en question en raison des sols liquéfiables en pied de digue. De plus, une
inondation en 2017 a remis en question sa capacité d’acheminement et les perfor-
mances du déversoir et du bassin de tranquillisation. Ce dernier incident a conduit
à une analyse des risques pour définir les mesures potentielles de réduction des
risques.
Dans ce contexte, ce rapport présente les travaux d’analyse des risques

menés pour identifier, évaluer et orienter les investissements futurs afin d’améliorer
la sécurité du barrage de Río Hondo, à travers une analyse qualitative et quantitative
alignée sur le Plan de Gestion des Risques conduit par l’ORSEP pour ses barrages
réglementés.
1. INTRODUCTION
Rio Hondo dam is located on the Dulce river basin, 2 km upstream from the
city of Termas de Río Hondo in the province of Santiago de Estero (Argentina).
The construction was carried out between 1958 and 1967. The reservoir has an
approximate volume at the maximum operating level of 1,343 hm3 and a surface
of 229 km2 . This dam is key for agricultural water supply in the province, providing
water for 270,000 hectares. In addition, it also generates energy with two turbines
of 9.5 MW and it provides flood protection for downstream populations.
Rio Hondo dam (Figure 1) is a mixed dam, made up of two embankments and
a central section of concrete with Marcelino-type buttresses. The maximum height
of the dam is 39.5 meters with a total length of 4,347 meters. The crest elevation is
278.50 meters above sea level. The right bank embankment is 2,367 meters long
and the left bank embankment 1,773 meters long. The concrete section (206 meters
long) includes a gated spillway with 21 vanes, 4 bottom outlets drains, 4 water intakes
and a hydropower station.
In recent years, two issues have been discussed regarding Rio Hondo dam
safety. First, potential liquefiable sands were found in the toe of the left bank embank-
ment, so significant settlements could be produced during earthquakes of high
magnitude. Second, a significant flood (around 50 year return period) occurred
in 2017, reached high levels in the reservoir and put into question its flood routing
capacity. In addition, high outflow discharges produced displacements in the rock
fill that protects the river bed downstream the stilling basin.
64
Q.105 - R.5
Fig. 1
General view of Rio Hondo dam.
Vue générale du barrage de Rio Hondo.
In order to evaluate the need for potential risk reduction actions, a risk anal-
ysis of the Rio Hondo dam has been carried out based on the risk assessment
methodology for dams, which is described in the Technical Guideline published by
SPANCOLD [1] and the Indian CWC Guidelines [2]. As shown in Figure 2, this
methodology is divided in two phases:
- Qualitative risk analysis: A failure mode identification process was car-

ried out through participative workshops. As a result of this process,
recommendations were made to reinforce dam safety.
- Quantitative risk analysis: A risk model was developed to estimate failure
probability, societal risk and economic risk. Finally, risk results are compared
with international tolerability guidelines to evaluate the need for further risk
reductions actions. Risk reduction actions were also prioritized based on risk
results.
The main steps and results of this process are summarized in the following
sections.
2. QUALITATIVE RISK ANALYSIS
The qualitative risk analysis was made through a participative workshop to

identify potential failure modes that could occur in the dam. More than 40 technicians
65
Q.105 - R.5
Fig. 2
Risk assessment steps.
Étapes de l’évaluation des risques.
participated in this workshop, including international experts in different disciplines

and staff responsible for dam safety.
After reviewing exiting information (more than 70 documents) and a site visit,
failure modes were defined individually by each participant. Secondly, these modes
66
Q.105 - R.5
Fig. 3
Failure Modes identified during working sessions.
Modes de défaillance identifiés lors des sessions de travail.
were put in common and then potential failure modes were identified by the group
based on a common discussion. Next, failure modes were classified and selected
to be included or not in the quantitative risk model (next phase). In total, 28 failure
modes were identified for Rio Hondo dam as shown in Fig. 3.
Finally, based on these failure modes, some recommendations were made to

improve dam safety management, mainly focused on improving the monitoring sys-
tem, implementing a flood warning system and new studies to evaluate the seismic
behavior of the embankment dam and the hydraulic behavior of the stilling basin.
67
Q.105 - R.5
3. QUANTITATIVE RISK ANALYSIS
Within the qualitative risk analysis, failure modes 1, 2, 17 and 18 were selected
to develop a quantitative risk model in order to evaluate potential solutions to improve
dam hydrological and seismic response. Risk model architecture was developed
with iPresas software [3] which computes risk based on even trees and influence
diagrams and provides a user-friendly interface for risk representation and analysis.
Risk model architecture is shown in Figure 4 in an influence diagram. In this figure,
blue nodes are linked with load input data, red and yellow nodes are related with the
system response and green nodes are related with consequences. In this sense,
economic and social consequences in downstream areas were analyzed in detail
for several flood and failure scenarios. The event tree of this risk model has more
than 290,000 branches and it was used to compute both incremental and total flood
risks downstream.
Fig. 4
Quantitative risk model architecture.
Architecture de modèle de risque quantitatif.
As can be observed in this event tree, the Rio Hondo risk model is divided
in two sub-models for hydrological and seismic scenarios. The hydrological sce-
nario sub-model includes: maximum water levels reached in the reservoir for all the
combinations of floods, water pool levels and availability of spillway gates, bottom
outlets and water intakes. These results were used to compute failure probability
due to overtopping or stilling basin erosion.
The seismic scenario sub-model includes two failure modes: overtopping

failure due to crest settlement and internal erosion failure as a consequence of
transverse cracking. Both failure modes are related to crest settlement. For this
reason, a structural numerical model has been carried out in FLAC [4] to study
the seismic behavior of the dam [5]. The numerical model mesh is shown in
Figure 5.
68
Q.105 - R.5
Fig. 5
Structural numerical model mesh.
Maillage de modèle numérique structurel.
Thirty-two ground motion inputs have been used to analyze the behav-
ior of the dam. The ground motion inputs adopt artificial seismic records. The
response spectra based on the site of the dam - prepared by the Argentine National
Institute for Seismic Prevention (INPRES) related to various return periods - are
chosen to be the target response spectra. In the model, the crest settlement
and the loss of thickness of the filter-drain have been determined at different
points.
To estimate the probability of several nodes of the risk model, a deep neural
network with 3 hidden layers and 256 nodes in each of the layers has been used.
This neural network makes it possible to perform a multivariate regression and esti-
mate the behavior of the dam under 5000 seismic signals compatible with the target
spectrum. The input variables to the neural network have been the PGA, PGV and
Arias intensity, and the output variables (input variables in the risk model) have been
crest settlement and the loss of thickness of the filter layer. The probability of trans-
verse cracking is stablished at 60% [5] when relative crest settlement reaches 5%
[6]. The filter failure is considered to occur when 50% loss of filter layer thickness is
achieved.
After entering all the input data for the risk model annual failure probability
and societal and economic risk was estimated for the current situation of Rio Hondo
dam. These results were compared with existing international tolerability guidelines
for dams [8, 9] as shown in Figure 6.
As can be observed in this graph, the highest failure probability is obtained

for the stilling basing erosion failure mode, which is above existing tolerability
recommendations for societal risk and failure probability. However, higher social
consequences would be produced for seismic failure modes, since warning times to
evacuate and protect downstream population are much shorter.
69
Q.105 - R.5
Fig. 6
Results of risks for current situation in tolerability chart proposed by USBR.
Résultats des risques pour la situation actuelle dans le tableau de tolérabilité
proposé par USBR.
4. POTENTIAL RISK REDUCTION ACTIONS
Based on these results, different risk reduction measures were proposed to

improve hydrological and seismic response of the Rio Hondo dam. The proposed
measures were:
1. Gravel columns at the left embankment toe to avoid liquefaction problems.

2. New stabilizing embankment over left embankment slope to increase weight
and avoid liquefaction.
3. Improvement in the gates’ reliability by reinforcing the electrical supply.
4. Reinforcement and improvement of the Emergency Action Plan by enhanc-
ing the warning system and building capacity in the population.
5. Increased restitution water level downstream to prevent erosion of the stilling
basin. Two potential options were analyzed: increases of 1 and 1.5 meters.
6. Preventive emptying (10 days) of the reservoir based on a new flood warning
system. The feasibility of this measure is still being evaluated by authorities
considering the potential loss of water reserves for irrigation.
70
Q.105 - R.5
The effect of these measures on dam risk is shown in Figure 7. As can be

observed, measures 1 and 2 produce an important decrease in seismic risks while
measures 5 and 6 have a significant effect on hydrological risks.
Fig. 7
Results of risk reduction measures prioritization.
Résultats de la priorisation des mesures de réduction des risques.
The proposed measures were prioritized combining the costs of the measures,
the dam and hydropower plant maintenance plan and the risk reduction achieved
by each measure, using the EWACSLS indicator. This indicator combines efficiency
and equity principles [10]. The sequence obtained is shown in Figure 8.
Fig. 8
Results of risk reduction measures prioritization.
Résultats de la priorisation des mesures de réduction des risques.
71
Q.105 - R.5
Prioritization results show that the most efficient measure, both from the point
of view of the risk of the dam and the risk of downstream flooding, would be to
implement a flood warning system that allows emptying the reservoir before a flood.
With this measure, the dam would already be aligned with international tolerability
recommendations. Further analysis should be carried out by authorities to determine
feasibility of this measure.
Next, structural measures allow lowering dam risk by several orders of mag-
nitude through solutions that avoid liquefaction processes at the embankment toe
and solve bad hydraulic performance of stilling basin.
Finally, the measures to reinforce the Emergency Action Plan and improve
power supply the gates appear last as they affect non-predominant failure modes,
especially after implementing the previous measures. In any case, and despite the
current good situation of these aspects, these are two key issues that cannot be
neglected and it is advisable to continuously work on improving them.
5. CONCLUSIONS
Rio Hondo dam is an essential infrastructure for the social and economic devel-
opment of the region where some doubts have arisen in the last years about its
performance in hydrological and seismic scenarios. In order to inform decision mak-
ing on potential risk reduction actions, a qualitative and quantitative risk assessment
process was made, evaluating dam safety in an integrated, comprehensive and
participative way.
As a result of this process, valuable recommendations were obtained to

improve dam monitoring and management. In addition, risk results were used to
prioritize potential actions, being the first step of this sequence the implementa-
tion of a complete warning system in the catchment to improve flood management,
aligning the dam with international risk tolerability recommendations, and protecting
important downstream populations.
REFERENCES
[1] SPANCOLD, Risk analysis as applied to dam safety. Technical Guide on Oper-
ation of Dams and Reservoirs, vol. 1. Spanish National Committee on Large
Dams, 2012.
[2] CWC, “Guidelines for Assessing and Managing Risks Associated with Dams,”
no. February. Central Water Commission. Ministry of Water Resources India,
2019.
[3] IPRESAS, “iPresas Calc. User guide.”, Valencia, 2020. www.ipresas.com
72
Q.105 - R.5
[4] ITASCA, “FLAC Version 8. Manual”, 2016.
[5] International Commision on Large Dams, Selecting Seismic Parameters for

Large Dams- Guidelines. Revision of Bulletin No. 72. Committee on Seismic
Aspects of Dam Design, 2010.
[6] R. Fell, M. Foster, J. Cyganiewicz, G. Sills, N. Vroman, y R. Davidson, “Risk

Analysis for Dam Safety: A Unified Method for Estimating Probabilities of
Failure of Embankment Dams by Internal Erosion and Piping”, 2008.
[7] S. Pells y R. Fell, “Damage and Cracking of Embankment Dams by Earthquake

and the Implications for Internal Erosion and Piping”, 21st Internal Congress
on Large Dams, 2003.
[8] USBR, Dam Safety Public Protection Guidelines - A Risk Framework to Sup-
port Dam Safety Decision-Making. United States Bureau of Reclamation,
2011.
[9] USACE, “Safety of dams - Policy and procedures. Engineering and design”.
United States Army Corps of Engineers, 2014.
[10] SERRANO-LOMBILLO, A ET AL. “A new risk reduction indicator for dam

safety management combining efficiency and equity principles,” Structure and
Infrastructure Engineering, 2016.
73
Q.105 - R.6
-------
-------
EMERGENCY PLANNING IN PORTUGUESE DAMS (∗)
Paulo CASTRO
Civil Engineer, Dam Safety Division
APA – PORTUGUESE ENVIRONMENT AGENCY
José ROCHA AFONSO

Civil Engineer – Head of Dam Safety Division, Secretary General PTCOLD
(CNPGB)
APA – PORTUGUESE ENVIRONMENT AGENCY, Portugal
PORTUGAL
SUMMARY
The implementation of an adequate emergency planning framework in the

dam safety context is of the utmost importance in order to deal with the hazards
inherent to the construction and operation of the dam structures and minimize the
impacts in the downstream valley resulting from a possible incident or dam failure.
After 2007, an increased attention was given by the different dam safety stake-
holders for the need to prepare, update and implement Emergency Action Plans
(EAP’s), within the establishment of a revised version of the Portuguese Regulation
for Safety of Dams.
A number of significant developments relating to emergency planning in the

dam safety context have occurred in Portugal recent years, namely with the publica-
tion of specific guidelines and complementary documentation seeking to organize
and clarify applicable rules and procedures.
∗ Plan d’alerte et de secours pour les barrages au Portugal
74 DOI: 10.1201/9781003211808-52
Q.105 - R.6
This paper aims to summarize a number of aspects related to the emer-

gency planning rules and activities developed in Portugal in recent years, namely by
addressing the evolution in the implementation of the EAP’s. It also refers to a major
European civil protection exercise carried out in 2019 (Cascade 19) and updates on
ongoing pilot projects regarding an emergency planning interface with flood warning.
RÉSUMÉ
La mise en œuvre d’un cadre adéquat de planification d’urgence dans le con-

texte de sécurité des barrages est de la plus haute importance afin de faire face
aux dangers inhérents à la construction et à l’exploitation des structures du barrage
et de minimiser les impacts dans la vallée en aval résultant d’un incident ou d’une
défaillance possible du barrage.
Après 2007, les différents intervenants en matière de sécurité des barrages

ont accordé une attention accrue à la nécessité de préparer, de mettre à jour et de
mettre en œuvre les plans d’action d’urgence (PAE), dans le cadre de la mise en
place d’une version révisée du règlement portugais pour la sécurité des barrages.
Un certain nombre d’évolutions importantes concernant la planification

d’urgence dans le contexte de la sécurité des barrages se sont produites au
Portugal ces dernières années, notamment avec la publication de lignes directri-
ces spécifiques et de documents complémentaires visant à organiser et à clarifier
les règles et procédures applicables.
Ce document vise à résumer un certain nombre d’aspects liés aux règles et

activités de planification d’urgence développées au Portugal ces dernières années,
notamment en abordant l’évolution de la mise en œuvre du PAE. Il fait également
référence à un important exercice européen de protection civile réalisé en 2019
(Cascade 19) et à des mises à jour sur les projets pilotes en cours concernant une
interface de planification d’urgence avec alerte aux inondations.
1. INTRODUCTION
Among all the themes related to the implementation of the Dam Safety policies,
the implementation of an adequate emergency planning framework is admittedly of
the utmost importance, in order to deal with the hazards inherent to the construction
and operation of the dam structures and minimize the impacts in the downstream
valley resulting from a possible incident or dam failure.
In Portugal, the Portuguese Environment Agency (APA – Agência Portuguesa

do Ambiente) holds the responsibility to carry out the Dam Safety Regulator role, by
75
Q.105 - R.6
enforcing the legal dam safety regulations, which is done through its Dam Safety Divi-
sion (GSB). This task is accomplished with the support of other departments within
APA, with the collaboration of the National Laboratory of Civil Engineering (LNEC)
and is carried out in cooperation with the Civil Protection Authority (Autoridade
Nacional de Emergência e Proteção Civil – ANEPC).
Though emergency planning in Portuguese dams has been specifically fore-

seen in Portuguese regulations since 1990, after 2007 an increased attention was
given by the different Dam Safety stakeholders for the need to prepare, update and
implement Emergency Action Plans (EAP’s), within the establishment of a revised
version of the Portuguese Regulation for Safety of Dams (RSB) [1].
In addition, a number of other significant developments relating to emergency

planning in the dam safety context have occurred in recent years, namely the publi-
cation of specific guidelines and complementary documentation seeking to organize
and clarify applicable rules and procedures.
This paper aims to summarize a number of aspects related to the emer-

gency planning rules and activities developed in Portugal in recent years, namely by
addressing the evolution in the implementation of the EAP’s. It also refers to a major
European civil protection exercise carried out in 2019 and updates on ongoing pilot
projects regarding an emergency planning interface with flood warning.
2. LEGISLATION AND GUIDELINES
The main Portuguese dam safety legislation relating to emergency planning

currently in force, is Decree-Law no. 21/2018 of 28 March [2], encompassing Annex I,
which defines the criteria for the classification of dams according to their potential
hazard and damages, Annex II, which publishes the new Small Dams Regulation
(RPB) [2], and Annex III, which publishes the revised version of the previous 2007
RSB [1].
Despite the alterations introduced by the 2018 dam safety legislation, it should
be highlighted that the general structure of the 2018 RSB [2] keeps the overall
structure of the previous 2007 version of this regulation, thus allowing an optimized
transition, particularly in what regards the large dams’ rules and procedures.
In Figure 1 it is schematically presented the framework of the current

Portuguese dam safety legislation, including the related technical guidelines:
The scope of RSB [2] defined under the 2018 referred legislation, includes
large dams only (H > 15 m; 10 m ≤ h < 15 m ∧ V ≥ 1 hm 3 ); three classes of
consequences apply: I (high), II and III)). The scope for the 2018 RPB [2] includes
76
Q.105 - R.6
Fig. 1
Current Portuguese dam safety legislation framework.
Cadre législatif portugais actuel sur la sécurité des barrages.
the following types of dams: h < 10 m, any reservoir capacity; 10 ≤ H < 15 m ∧ V ≤ 1

hm 3 ; for these small dams only classes II and III apply.
The dams are classified under a combination of a potential consequences

factor and a dimensional factor∗ , into one the three consequences classes foreseen
in Annex I, and the classification results from the application of the chart presented
in figure 2, where Class I corresponds to higher consequences.
Fig. 2
Classification chart in Annex I of 2018 dam safety legislation.
Tableau de classification de la législation de 2018 sur la sécurité des barrages,
dans l’annexe I.
∗ Which is taken from the ICOLD Bulletin on Small Dams, after the French Regulations
77
Q.105 - R.6
Fig. 3
Zoning of the PEI and PEE areas in RSB emergency planning.
Zonage des PEI et PEE dans le Plan d’Alerte et de Secours du RSB.
Y is a parameter which considers the population affected by the dam fail-

ure flood wave, to be assessed according to the number of permanent residential
buildings located in the corresponding flood area.
X is a hazard factor which constitutes a dimensional weighting affecting the

√
classification, expressed by X = H2 V (H being the height of the dam in metres and
V the capacity of the reservoir in hm3 ).
Chapter III of RSB [2] specifically addresses civil protection measures and
emergency planning to be considered in the dam safety context, consisting in a
body of eleven articles, spanning from emergency zoning to the definition of the
EAP’s.
For emergency planning purposes, in Class I dams RSB [2] considers the
separation of the dam’s downstream valley into two main zones where the respon-
sibilities for early warning and emergency management are clearly identified. The
dam owner is responsible to issue alert messages to the civil protection entities and
deliver early warnings to the population in the emergency area closer to the dam,
referred to as “Zona de Autosalvamento” or ZAS, this being considered an area
where the Civil Protection may not have the time to issue the emergency warnings.
In general terms, however, the Civil Protection structure is broadly responsible for
the emergency management in the downstream valley.
In addition to the zoning areas, RSB[2]) states that the overall Emergency
Planning is based upon two documents, namely the PEI’s (“Internal” Emergency
Plans) focusing on the ZAS, whose elaboration is a Dam Owner’s responsibility, and
PEE’s ( “External” Emergency Plans) focusing on the valley downstream the ZAS,
a specific Civil Protection Structure responsibility.
78
Q.105 - R.6
The information contained in an EAP† submitted under the Portuguese

legislation addresses the following items:
• Identification of the dam owner and the technical staff responsible for
EAP
• General description of the dam
• Plan for dam monitoring and safety control
• Characterisation of the downstream valley of the dam
• Identification of accident scenarios
• Modelling of the chosen accidents scenarios
• Characterization of the “worst-case” scenario
• Procedures for assessment and classification of accidents
• Identification of procedures in case of an accident
• Human and technical resources – Alert to Civil Protection structure
• Human and technical resources – Early warning of the population
• EAP exercises and public awareness procedures
Although in some particular cases (such as the occurrence of incidents), the

Dam Safety Regulator may enforce the implementation of emergency measures,
the elaboration and implementation of EAP‘s as per RSB [2] is only due to Class I –
high consequence dams.
A main alteration introduced by the Portuguese dam safety legislation pub-

lished in 2018 [2] is that, in emergency planning, for Class II large dams, the concept
of simplified emergency procedures must be considered. Under this concept, the
Class II dam owner must prepare a simple document, basically addressing dam
break flood maps and emergency contacts. Nevertheless, it is specifically fore-
seen in the RSB text that the simplified emergency procedures may be subject to
additional measures in situations involving greater harmfulness.
In small dams included in RPB, classified in Class II, consideration of simplified

warning measures by the owner is also foreseen for emergency situations.
The determination of the dam’s class of consequences is therefore a crucial

setting point, to decide for the need to either elaborate a complete EAP, or just
adopt simplified emergency procedures, or have no need for measures in case of
low consequences dams.
After the publication of a revised RSB [1]in 2007, as the main regulations
for large dams, the Civil Protection Authority and the Dam Safety Regulator jointly
produced a set of guidelines to support the elaboration of the EAP’s.
†In this paper we use the acronym EAP as an equivalent of the “Internal Emergency Plan”
(PEI), as established in the Portuguese Regulations (RSB)
79
Q.105 - R.6
These guidelines are gathered in a document entitled CADERNO PRO-

CIV 5[3]) (Guidelines for Developing Emergency Action Plans in Dams), published
in June 2009, which presents, in a detailed form, the contents and organization to
be considered in the elaboration of an EAP.
Furthermore, in October 2011 a complementary set of guidelines was jointly

published by ANPC, LNEC and the Dam Safety Regulator, specifically designed for
the implementation of exercises in the EAP’s context – CADERNO PROCIV 17[4])
(Guidelines for Developing Exercises on Emergency Action Plans in Dams).
As a result from the experience obtained in the EAP’s evaluation and imple-
mentation process, a document gathering the more Frequently Asked Questions
(FAQ’s) [5] was jointly produced between APA and ANPC with the collaboration
of LNEC and was issued in 2019. These FAQ’s intend to help the dam owners,
consultants and others, particularly on the phase of the EAP’s elaboration.
3. EVOLUTION IN EAP SUBMISSIONS AND EAP IMPLEMENTATION
The first years following the approval of the RSB [1]in 2007 may be considered
a transition period where the dam owners, together with their appointed consultants,
had to perform an adapting stage to a number of new dam safety procedures.
The evolution observed in EAP’s submission was influenced by the different

levels of technical knowledge of the dam owners and respective consultants and
availability of financial support required to carry out the EAP related studies and
implementation of the approved solutions.
As in other countries, in Portugal, hydropower dam owners generally prove to

be more technical and financially prepared to deal with the alterations introduced in
dam safety regulations.
Among the issues raised by the 2007 legislation, many financial difficulties
were detected in a number of dam owners at an early stage, and several public
or state controlled entities were no exception to this problem. This matter has been
partially dealt with by assigning state funding or by allocating EU funds to dam safety.
Figure 4 summarizes the evolution of the EAP’s approved and under evaluation
in Portugal from early 2012 to late 2018. More recently, a trend has been noted
towards the stabilization of the submitted documents to the Dam Safety Regulator,
as a significant percentage of Class I dams EAP’s have been submitted.
Though the implementation of the first EAP’s in Portugal date back to more
than a decade ago, it was only more recently that an illustrative number of these plans
was put in place so that some level of experience may be built upon this phase. One
80
Q.105 - R.6
Fig. 4
Evolution in Emergency Action Plans (2012–2018).
Evolution des plans d’action d’urgence (2012–2018).
general point that arose during first EAP implementations was the need to discuss
a number of matters that have not been considered or were not clear enough from
the elaboration phase. These matters had mainly to do with the following items:
• Procedures to be considered in the public awareness sessions to be promoted

by the Dam Owner (i.e. contacts with national and local authorities, programme
of the information sessions and information to be available to the public)
• Installation of the outdoor siren network (i.e. location practicality, technical
features)
• Communications systems (i.e. available mobile phone network in remote
areas, redundancies)
• Aspects of public evacuation in case of emergency (i.e. shelter locations)
Many of the aspects that have to do with EAP implementation have much
to do with the specific conditions of the occupation in the area surrounding the
dam, cultural level of the local population and adaptability of the foreseen technical
equipment to local conditions.
81
Q.105 - R.6
Fig. 5
Public awareness information leaflet (Page 1) – Ribeiradio-Ermida.
Notice d’information de sensibilisation du public (Page 1) – Ribeiradio-Ermida.
Fig. 6
Public awareness information leaflet (Page 2) – (Ribeiradio-Ermida).
Notice d’information de sensibilisation du public (Page 1) – Ribeiradio-Ermida.
Example of an information leaflet that has been distributed to the public

(Ribeiradio-Ermida development), is shown below. An agreed template was pre-
viously discussed between APA, ANEPC and the Dam Owner (originally a folded
A4 size).
82
Q.105 - R.6
Fig. 7
Implementation of early warning systems at Veiguinhas, Alto Ceira II and Baixo
Sabor dams.
La mise en œuvre des systèmes d’alerte précoce aux barrages de Veiguinhas, de
Alto Ceira II et de Baixo Sabor.
This document contains a very brief description of the dam and dam safety
monitoring, ZAS areas, emergency phone numbers, description of the siren warn-
ings, simplified map highlighting dam’s location and access roads, shelter location
and general procedures to be followed by the public under an emergency occurrence.
4. CASCADE’ 2019 CIVIL PROTECTION EXERCISE
Between 28th May and 1st June 2019 the Portuguese Civil Protection Authority
(ANEPC) coordinated the holding of a LIVEX exercise (Live Exercise) called Cas-
cade’2019, under the European Union Civil Protection Mechanism, having invited
Spain, France, Belgium, Croatia and Germany (THW) as international partners.
The objective of Cascade’2019 was to test the joint and integrated response
capacity of the Portuguese Civil Protection System and of the European Union, in a
scenario composed of multiple and complex emergency situations, namely caused
by earthquakes, extreme weather conditions, flooding in watercourses and dam
accidents.
In terms of dam safety emergency planning, the main and more complex
exercise foreseen consisted in simulating an accident in Ribeiradio dam, where
a scenario of rapid and simultaneous opening of the spillway flood gates associated
with adverse weather conditions caused the existence of a flood with a high return
period.
Ribeiradio dam is a concrete gravity dam 76 m high located in the Vouga

River, with a reservoir of 136 hm3 . The spillway includes three floodgates, and
was classified in Class I regarding potential damages in the downstream valley,
83
Q.105 - R.6
Fig. 8
View of Ribeiradio dam from downstream (bottom outlet in operation).
Vue du barrage de Ribeiradio depuis l’aval (sortie de fond en fonctionnement).
thus requiring the implementation of an emergency action plan in accordance with

RSB [2].
During the Ribeiradio Cascade’2019 exercise, a set of procedures foreseen in

the EAP of the dam was implemented, associated with the evolution of the occur-
rence from blue alert level (lower risk alert level) to the orange level (high risk, before
red level, which is activated under dam failure) and also from the orange level back
to the blue level.
The use of the Observation and Control Unit (POC) was promoted in an emer-
gency context, the population warning system test was performed consisting of the
operation of the warning posts located in the downstream valley, the conditioning
of access to the dam was performed and the communications established between
the participating entities were established, among other aspects.
A previously established “time tape,” gave general guidance to the action

of the “players” who moved to the site, in particular the dam owner and APA as
the dam safety regulator, who developed their action between 09:00 and 18:00 on
29.05.2020.
84
Q.105 - R.6
Fig. 9
View of Ribeiradio Observation and Control Unit (POC) during the Cascade’2019
exercise.
Vue de l’Unité d’observation et de contrôle de la barrage de Ribeiradio (POC)
pendant l’exercice Cascade’2019.
On the eve of the exercise, ANEPC disseminated several simulations on

meteorological reports and on the occurrence of earthquakes that framed the
exercise.
Comments and lessons learnt from the Cascade’2019 Ribeiradio EAP exercise
may be summarized as follows:
• It was considered that the “time tape” that served as the basis for the devel-
opment of the EAP exercise was adequate for testing the scenario foreseen,
though it was necessary to make adaptations in the corresponding hydrogram
considered in EAP, in order to adjust it to the time limits established by ANEPC
for the performance of the exercise.
• The EAP states that, in an orange alert, the responsible for the EAP activation
(dam owner representative) should condition access to the dam area, and it
was not explicit how this procedure should be implemented (the top of the dam
is used as a municipal road, which will be integrated in municipal emergency
planning).
• With regard to the provision of meteorological information, it was not clear
which circuit was adopted to circulate it through the different entities and levels
of responsibility.
• The information transmitted to the population regarding the exercise in the
downstream valley would not have been sufficient to inform the population
and companies in the area.
85
Q.105 - R.6
• A possible revision of the conditions for activating the alert of the orange level
of the EAP may consider this activation at a lower water level.
• The exchange of information between the participants of the exercise high-
lighted the interest in having the EAP documentation in files with formats
enabling them to be moved electronically faster, in particular between entities
with intervention in the emergency response.
5. EMERGENCY PLANNING AND FLOOD WARNING
The installation of warning systems with sirens downstream of large dams,

integrated in the EAP’s for Class I dams and designed for dam failure, also created
an urge for its use in smaller natural flood scenarios, as a tool to warn the populations
and help the civil defense actions.
These warning systems, with sirens, are sized for dam break scenarios and
installed in the “self-rescue zones” immediately downstream of the dams. However,
it is expected that they may be used in situations of “operational discharges”, an
aspect considered in particular in the CADERNO PROCIV 5 [3]. In these situations,
the signal emitted by the sirens is different from the one of dam accidents (also see
FAQs [4]) regarding the EAP).
The “high discharge for operational reasons” concept was thought for situa-
tions or discharge related to the dam itself (need to empty the reservoir for safety
reasons, incident of opening of floodgates, others) or floods of very high return
period, implying close monitoring of the behavior of the dam.
Subsequently, the idea of using the warning system for discharges of more
frequent natural floods was a further step, though it raises the question that these
more frequent floods, giving rise to discharges whose cause is precipitation, are not
expected to pose any safety problems in the dam. These warnings are essentially
part of support for the action of civil protection framed in flood protection.
EDP as the dam owner of a significant number of Class I Portuguese dams

has in due course presented APA some studies of the “Critical Flow Rates for High
Discharge Warnings”, in the form of working documents, concerning 3 sets of dams
in relation to which, through the construction of new dams, warning systems were
implemented on the ground: a) Baixo Sabor and Feiticeiro + Pocinho; b) Foz Tua +
Valeira; c) Ribeiradio and Ermida.
In the case of studies on EAPs and respective warning systems, these are
submitted to APA as the dam safety regulator for assessing the proposed solutions,
to this end by consulting ANEPC. Moreover, it is a question of defining, in the face
of local conditions, the warning flow rates of the different sirens, an aspect in which
the local civil protection structures operating on the ground must be involved.
86
Q.105 - R.6
A joint analysis of the APA-GSB/ANEPC on this issue has been carried out,
which needs to be operationalized, and without prejudice to complementary aspects
that have yet to be worked on, it was possible to reach the following conclusions:
• It was considered that there are conditions to proceed with the use of the
warning systems installed in ZAS, in Class I dams, for the relevant floods in
the sections already studied;
• The system used for the dam break scenario may also serve to signal high
discharges in operational situations, with a different siren sound pattern, and
the methodology envisaged, based on case-by-case analyses, seems to be
generally appropriate;
• The flood rates chosen to trigger the warnings, according to local damages
foreseen during floods, which must be unambiguously established and dis-
closed, may when necessary, be adjusted according to its suitability/efficiency
observed in real situations.
Further to the above mentioned developments, it will be expected that in the

2020/2021 wet season, in case of occurrence of high flow rates, the systems that
have been the subject to study can already be operated.
ACKNOWLEDGEMENTS
The authors would like to thank the colleagues from the Dam Safety Division
of the Portuguese Environment Agency:
- Marta Duque and Arnaldo Nisa, for their support in providing data and
cooperation in the Cascade 19 exercise
- Carla Marques, for the support in the revision and preparation of the final text.
REFERENCES
[1] Regulamento Português de Segurança de Barragens (Portuguese Regula-

tions for Safety of Dams (RSB)), Decree-Law no. 344/2007, October 15th
(revision of Decree-Law no. 11/90, January 6th), Lisbon.
[2] Decreto-Lei n.o 21/2018 de 28 de março. Diário da República n.o 62/2018,

Série I. Altera o Regulamento de Segurança de Barragens e aprova o Reg-
ulamento de Pequenas Barragens (Decree-Law No. 21/2018 of 28 March.
Revision of Portuguese Regulations for Safety of Dams (RSB) and publication
of new Portuguese Regulations for Small Dams, RPB).
87
Q.105 - R.6
[3] CADERNO PROCIV 5 – Guia de Orientação para Elaboração de Planos

de Emergência de Barragens, APA/NPC, 2009, (Guidelines for Developing
Emergency Action Plans in Dams).
[4] CADERNO PROCIV 17 – Guia para Elaboração de Exercícios aos Planos

de Emergência Internos de Barragens APA/ANPC, 2011 (Guidelines for
Developing Exercises on Emergency Action Plans in Dams).
[5] FAQ – Perguntas Frequentes em Planos de Emergência Internos de Barra-

gens, APA/ANEPC, 2019 (Frequently Asked Questions in Emergency Action
Plans for Dams).
88
Q.105 - R.7
-------
-------
THE REPAIR OF BOTTOM OUTLET CLOSING FACILITY AT THE

PALCMANSKÁ MAŠA AND HRIŇOVÁ DAMS (∗)
Tomáš IČ, Roman IVANČO, Branislav LIPTÁK & Marián MIŠČÍK

SLOVAK WATER MANAGEMENT ENTERPRISE, SOE, Banská Štiavnica
L’ubomír UHORŠŠÁK
VODOHOSPODARSKA VYSTAVBA, SOE, Bratislava, Slovakia
SLOVAKIA
SUMMARY
Bottom outlet gates belong among the key technological equipment of dams.
Their functionality and reliability are crucial not only for creating conditions for the
fulfilment of individual purposes of water construction, but also for the safety of dams
themselves.
During the operation of dams, great attention must be paid to the inspection,
monitoring and maintenance of bottom outlet gates. Exceptionally, however, there
are situations in the life cycle of the dam where the deteriorated technical condition
of the bottom outlet gates endangers the functionality and safety of the dam. In such
cases, it is necessary to ensure their early repair or, in worse cases, replacement,
sometimes with the need to apply special support solutions to ensure the successful
implementation of the repair itself.
The report offers two individual cases of ensuring the re-functionality of the bot-
tom outlet gates. In some cases of gate repairs, it seems more difficult preparation
phase for ensuring suitable and safe conditions for the repairs than the implemen-
tation phase itself. Such a situation is illustrated by the first example, which is the
Palcmanská Maša Dam and the production of a special submersible temporary
∗ Réparation des vannes de fond des barrages de Palcmanská Maša et Hriňová
DOI: 10.1201/9781003211808-53 89
Q.105 - R.7
facility to arrange suitable conditions for successful repair of the gate. The second
example is the repair of the inlet gate of the bottom outlet on the Hriňová Dam, the
purpose of which is to supply drinking water to the population. This section illus-
trates the case where ensuring the functionality of the bottom outlet gate required
the replacement of the entire device and points out in particular the issues and
aspects that are specific to such a radical solution, especially in the case of the
drinking water reservoir.
Ensuring the operability of individual bottom outlet gates, their maintenance

and especially their repairs or replacements always require an individual approach,
which takes into account not only the type and dimensional parameters of the gate,
but also the construction period of the hydraulic structure itself in the context of
relevant historical relations, as well as currently valid standards, requirements and
circumstances associated with the operation of the structure in a particular location.
The mentioned technical measures represent in all cases a complex process of
preparation and actual implementation, but their appropriate implementation is a
necessary condition for ensuring safe and reliable operation of dams.
RÉSUMÉ
Les vannes de vidange de fond représentent un élément clé du dispositif d’un

barrage. Leur fonctionnement et leur fiabilité s’avèrent essentiels, tant pour la réali-
sation des différents objectifs de la construction hydraulique que pour la sécurité du
barrage.
Une attention particulière doit être portée à la vérification, au suivi et à la main-

tenance des vidanges de fond durant l’exploitation d’un barrage. Des circonstances
exceptionnelles peuvent néanmoins intervenir lors du cycle de vie – notamment une
altération de l’état des vannes – mettant ainsi en péril le bon fonctionnement et la
sécurité de la construction. Il est alors indispensable de procéder à une réparation
rapide, voire à un remplacement des pièces défectueuses dans le pire des cas.
Le rapport article présente deux cas concrets aboutissant à une remise en état de
fonctionnement des vidanges de fond.
Certaines opérations visant à réparer ce type de vannes peuvent s’avérer

difficiles à mettre en place – tant du point de vue de la sécurité que du point de
vue technique. Cette situation est illustrée par le premier exemple, celui du barrage
de Palcmanská Maša, qui a nécessité la construction d’une installation submersible
temporaire permettant de procéder à la remise en service. Le second cas détaille
la réparation de la vanne amont de l’aménagement de Hriňová qui approvisionne
la population en eau potable. Cette partie présente les circonstances menant au
renouvellement de l’ensemble du dispositif afin de permettre le verrouillage de la
conduite forcée et met en évidence certains aspects et questions spécifiques à ce
type de solution radicale, à plus forte raison pour un réservoir d’eau potable.
90
Q.105 - R.7
Assurer le bon fonctionnement des différentes vannes de vidange de fond

implique principalement la réparation ou le remplacement de celles-ci et requiert
une approche individuelle, tenant non seulement compte de paramètres comme le
type de vanne et ses dimensions, mais aussi de la période à laquelle a été bâtie
l’installation (et les conséquences qui en découlent) ainsi que des exigences et
réalités du moment liées à l’exploitation du barrage. Si ces dispositions techniques
représentent, dans tous les cas, un processus de préparation et de réalisation des
plus complexes, le respect de ces mesures est essentiel afin de garantir la sécurité
et la fiabilité lors de l’exploitation du barrage.
1. INTRODUCTION
Bottom outlet gates belong among the key technological equipment of dams.
Their functionality and reliability is crucial not only for creating conditions for the
fulfilment of individual purposes of the hydraulic structure, but also for the safety of
the dam. Within the operation of the dams in the Slovak Republic, great attention is
paid to their inspection, monitoring and maintenance. Exceptionally, however, there
are situations in the life cycle of the dam where the deteriorated technical condition
of the bottom outlet gate endangers the functionality and safety of the dam. In
such cases, it is necessary to ensure their early repair or, in worse cases, their
replacement, sometimes with the need to apply special support solutions to ensure
the successful implementation of the repair itself. The paper offers two special cases
of ensuring the re-functionality of the bottom outlet gates.
2. APPLICATION OF SUBMERSIBLE TEMPORARY CLOSING FACILITY AT

THE PALCMANSKÁ MAŠA DAM
The first part of the paper is devoted to the topic of creating suitable conditions
necessary for the successful repair of the bottom outlet gates on the Palcmanská
Maša Dam. The dam and reservoir Palmanská Maša are located in the National
Park – Slovenský raj (Slovak Paradise), on the Hnilec River (Fig. 1) in Eastern
part of the Slovak Republic. It is the first part of the pumping hydropower system
Palcmanská Maša – Vlčia Dolina – Dobšiná. The hydroelectric use of water accumu-
lated in Palcmanská Maša Reservoir is provided by an inlet structure that releases
the water in the power plant in Vlčia Dolina [1]. The penstock is 2713 metres long,
what creates a head difference of 285.5 meters. In this case a space distribution of
the discharges can be considered because the water from the Hnilec River basin is
released to the catchment area of the Slaná River, where power plant Vlčia Dolina
is situated. The third part of the hydropower system is balance reservoir Dobšiná.
Palcmanská Maša Dam was built in 1954. It is a concrete gravity dam (height 34 m,
91
Q.105 - R.7
length 209 m, reservoir volume 11 mil. m3 ) consisting of 17 blocks. One of them is

a function block with a spillway and a bottom outlet (Fig. 2).
The temporary gates are nowadays one of the inseparable parts of dam
constructions. They are used for the temporary closure of bottom outlets intakes,
turbines, intake pipelines etc. during the technological equipment repairs. Under
the protection of temporary gate, it is possible to make repair without necessity of
significant water level decrease in reservoir, or even draining it off. The design and
construction of dams are always adequate to available knowledge and experiences.
Perhaps, this is the reason, why the temporary gate concept was not drafted in the
case of Palcmanská Maša Dam, built in the years 1950–1954. In the year 2018,
there was necessity to repair the upstream bottom outlets gates. The main point
of the first part of the paper is to show how under the protection of newly made
temporary gates, it was possible to repair bottom outlets gates without necessity of
significant water level decrease in reservoir.
Fig. 1
Palcmanská Maša Dam situated in the National Park – Slovak Paradise.
Barrage de Palcmanská Maša – parc naturel du Paradis Slovaque.
2.1. SUBMERSIBLE TEMPORARY CLOSING FACILITY PPH-1200
The submersible temporary closing facility PPH-1200 is unique, special pur-

pose facility appointed for closing the intake openings of dams with demands on
high level of tightness and safety. Simple overview of the closing facility construction
is visible in Figure 3. It consists of tripartite independent system of closing elements.
In case of damage or a malfunction of one of the elements, the rest can fully secure
the function of the facility. Basic closure of the opening is assured by the massive
conic element no.1. It represents main safety basis of the facility. It allows steady and
solid closure of intake. Behind the main cone follows pair of self-contained inflatable
92
Q.105 - R.7
Fig. 2
Palcmanská Maša Dam – cross section.
Barrage de Palcmanská Maša – coupe transversale.
rubber bags, which are inflated by a separate pressure air supply (elements no.2
and no. 3). A valve controlled bypass is used to balance the pressures between the
reservoir and the outlet pipe when inserting or releasing the closing facility into or
from intake opening. The bypass passes longitudinally through all the body of the
PPH-1200.
2.2. TECHNICAL SPECIFICATION OF PPH-1200
The technical specification of PPH-1200 is as follows:
Circle profile 1200 mm ± 25 mm

Safety operating depth 30 m
Max. filling overpressure for bag 1 0.1 MPa
(to submersion depth)
Max. filling overpressure for bag 2 0.2 MPa
(to atmospheric pressure)
Max. difference of pressures between 0.2 MPa
chambers of bags
Max. loading on base of intake opening 400 kN
93
Q.105 - R.7
Max. total peripheral normal loading of 678,6 kN

conic surface on Ø120
Max. diameter 1300 mm
Typical length 1450 mm
Max. length 1540 mm
Displacement 1000 kg
Dry weight 890 kg
Fig. 3
Section of submersible closing facility PPH-1200.
Secteur de la vanne souterraine PPH-1200.
2.3. MANIPULATION WITH PPH-1200
The facility is designed to allow as simple and effective manipulation as pos-

sible to ensure its secure positioning to appointed position. The positioning must be
done very careful and precise way so that the facility could fulfil its purpose depend-
ably. It is therefore very important that the manipulator is thoroughly familiarized
with the function of the individual components of the device and the sequence of the
prescribed steps before the actual manipulation. Compliance with the manipulation
procedure is essential also for occupational safety and to prevent possible damage
to the facility as well. The basic manipulation steps are as follows:
Balancing. The facility must be carefully laid on the water with a crane firstly.
Then it is necessary to release residual air from both of the rubber bags. The facility
must stay fixed to the crane by a rope to avoid a possible drop on the reservoir bottom.
It is necessary to fill the chambers with the water partly for easy manipulation under
water so that the facility could lightly float under water level (Fig. 4).
94
Q.105 - R.7
Fig. 4
Submersible closing facility during installation.
Fermeture submersible lors de l’installation.
Submersion. Manipulator – scuba diver is supposed to drag slightly the closing

facility stepwise into required depth. If there is increased difference between gravity
and buoyancy it is possible to balance the body of facility by manipulating the amount
of air in chambers so that the facility is floating in desired level.
Insertion the closing facility into intake opening. Prior to PPH-1200 insertion
the manipulator must check the opening, its geometric shape and cleanliness. If
there is a dirt, it must be removed. No discharge is allowed in the outlet pipe during
the work and manipulation with the facility.
Fixing and centring the position. It is hard for manipulator to exert a greater
force on the facility body in operating depth to get it into the pipe. For this reason
there is the bypass in the facility. When the valve of bypass is open it is easier to
insert and fit the body in the opening. As it is fitted, no other shifts can be done.
Main closing force must be transmitted by the first conic element.
95
Q.105 - R.7
Closing. After fitting the body in the opening, the bypass is closed. Then
the rear rubber bag no. 2 is inflated through connected hosepipe to pressure not
exceeding 0.2 MPa over the atmospheric pressure. In the same way the front rubber
bag no. 1 is inflated to pressure higher than 0.1MPa over hydraulic head in operating
depth. At the same time the difference of the pressures in the chamber between the
rubber bags cannot exceed 0.2MPa. After finishing this stage the hosepipes are
detached and removed from the water.
Testing and monitoring. Testing is performed by opening the outlets on the

level not more than 3–5%, so that uncontrolled discharge of water could not arise in
case of incorrect function of closing facility. Functionality of closing facility must be
monitored at least 12 hours by observing the discharge from outlets.
Control and pumping up the rubber bags. The periodic control of seepage and
functionality is needed all the operation time (every 6 hours). The control of pressures
in the bags must be done at 24-hour intervals. In the case of need the bags must
be pumped up on operating pressure, which can be executed via valves situated on
rear – dry side of the facility. Access to the closing facility is possible through the
service entrance into the bottom outlet pipe.
Releasing the closing facility. When the work is finished the bottom outlet must
be totally closed. Before the closing facility is released the scuba diver shall open the
valves from rubber bags and valve of bypass first. After a short time, the pressures
are balanced out and the diver can release the facility from the intake. Pumping up
the air into the chambers and/or also to rubber bags ensures that the body of facility
can float again. Afterwards the facility can be fixed to rope of crane and pulled out
from the water.
In the initial phase of the intention to repair the bottom outlets gates of Palcman-
ská Maša Dam, the situation seemed complicated due to the absence of provisional
temporary gates. In order to maintain the functionality and operability of the hydraulic
structure, the repair was inevitable. Therefore, the dam administrator faced the task
of finding the most appropriate way of repair. There were two possibilities, to make
the repair with empty reservoir, or to provide a suitable design and manufacture of
a temporary gate, which would not have a significant impact on the reservoir water
level and its operation during repair. To avoid the negative economic impacts, drain-
ing the reservoir was an extreme option. Therefore the dam administrator is highly
satisfied that he managed to secure the design of submersible temporary closing
facility PPH-1200, thanks to which the repair was made very effectively and quickly.
The closing was constructed by a Slovak company, which realised also the restora-
tion of old upstream bottom outlets gates itself. Entire repair process took relatively
short time period. Of course, the particular planning and preparatory works along
with manufacturing the temporary closing facility had to last much longer.
The successful repair of the 66 years old Palcmanská Maša Dam, which is the
part of the first water power pumping station in the territory of the Slovak Repub-
lic, allows the dam to work safely for following many years. However, a significant
96
Q.105 - R.7
positive for the future of this hydraulic structure is also the fact that this important
repair eliminated the problem of provisional closing of bottom outlets, which is a very
important factor in terms of trouble-free structure maintenance.
3. THE REPLACEMENT OF BOTTOM OUTLET GATE AT THE

HRIŇOVÁ DAM
The second part of the paper illustrates the case where ensuring the function-
ality of the bottom outlet gate required the replacement of the entire device. This part
of the paper points out in particular the issues and aspects that are specific to such
a radical solution, after decades of operation of the hydraulic structure designed for
the drinking water supply.
The Hriňová reservoir was put into operation in 1966. It was built on the Slatina
River in central Slovakia (Fig. 5). Dam is an earth-stone embankment dam with a
clay sealing core (Fig. 6). The length of the dam in the crown is 242 m, height 41.5 m.
The dam is equipped with one bottom outlet with diameter 2000 mm and discharge
capacity of more than 60 m3 ·s−1 . The bottom outlet consists of an inlet gate (plus a
temporary gate) and a sparging cone gate situated at the outlet part of the pipe. To
ensure the supply of water to the water purification plant in Hriňová Town (325 l.s−1 ),
three intake horizons with supply pipes and subsequently a common discharge pipe
of diameter 600 mm were built in the intake tower. The total volume of the reservoir
is 7,4 mil. m3 with a flooded area of 55 hectares.
Fig. 5
View of the Hriňová reservoir.
Vue du réservoir de Hriňová.
97
Q.105 - R.7
Fig. 6
Cross-section of the dam body of the Hriňová dam.
Coupe transversale présentant la digue du barrage de Hriňová.
In the initial stage of operation of the Hriňová Dam, in 1966, 1968 and 1971,
there were several serious technical problems (seepage, erosion furrows and land-
slides on downstream slope, leaks under the chute and in the entrance gallery),
which required extensive reconstruction carried out in the years 1989–1992 (con-
struction of a sealing cut-off wall, grouting at the contact of the sealing core and
the grout curtain, construction of new uplift boreholes). The subsequent verification
operation proved that it was possible to put the structure into permanent operation.
After decades of operation, technical problems began to appear on the technological
equipment of the dam. The conical spray gate at the bottom outlet pipe was replaced
in 2007. By replacing this gate, the process of several repairs to the technological
equipment of the hydraulic structure was started.
Another major technological repair was the replacement of the inlet gate
(Fig. 7) on the bottom outlet pipe of diameter 2000 mm with a capacity equal to
the flow rate Q100 , which represents more than 60 m3 ·s−1 at the maximum operat-
ing water level. This gate was located in the engine room, immediately behind the
temporary panel gate of the bottom outlet. It was type of a slide gate valve without
the possibility of flow regulation. This means that it could only be set to the fully open
or fully closed position. Manipulation with the gate was possible only with balanced
pressures from both sides of the valve, i. e. while the reservoir is empty or when the
bottom outlet pipe was filled through the bypass with the outlet gate closed.
During the regular inspection manipulation with the gate in February 2019, the
hydraulic locking device got jammed and subsequently the slide gate was crossed.
During further manipulation, the gate was partially closed, but a leak of approx. 300
l.s−1 was being detected. The conclusion of the commission inspection of the gate
by the employees of the dam administrator and the subsequent inspection within
the technical safety supervision by the state-authorized organization was that the
gate needs to be replaced. With regard to the actual condition of the gate, it was
ordered to keep the water level in the reservoir at a lower level in order to create
sufficient retention volume during flood flows. At the same time, the organization
authorized to perform technical safety supervision stated that the Hriňová Dam is
in risky operation and its safety is endangered. The State Administration Office
98
Q.105 - R.7
Fig. 7
View of the inlet gate.
Vanne à lunette.
ordered the replacement of the damaged gate by the end of 2019. Immediately after
the official decision the process of preparing the repair of the gate followed.
Considering the basic purpose of the hydraulic structure, which is the supply
of drinking water for the population, the basic condition for repairing the gate was to
carry out the work at the highest possible water level in the reservoir, same as was
the case with other technological parts of the structure in the past. The hydraulic
structure administrator formulated further conditions as follows:
– Replace the original inlet gate of the bottom outlet with a type of gate suitable
for the given type of dam (e.g. flap), taking into consideration the parameters
of the hydraulic structure (hydrodynamic pressure), with direct connection to
the bottom outlet pipe before and after the gate,
99
Q.105 - R.7
– The possibility of manipulating the inlet gate to full discharge, with regard to the
capacity of the bottom outlet pipe, i.e. 62.70 m3 ·s−1 at the maximum operating
level in the reservoir,
– Because of the risks arising from non-functional bottom outlet during the repair,
to dismantle the original and install a new gate in the time range of maximum
six weeks,
– Maximum production time of a new gate up to twenty weeks,
– Bringing the electrical control of the new bottom outlet gate into the dam
operating building.
With regard to the endangered safety of the hydraulic structure, the process of
tendering a suitable contractor for the replacement of the gate was relatively quick.
After concluding the contract, the phase of processing the implementation project
documentation and detailed survey works began, including dives to the temporary
panel gate, in order to determine its condition, the condition of the racks, or a need
of its sealing, etc.
Due to the size of the gate, before removing it from the engine room, it was
first necessary to cut it to smaller pieces. This process took a relatively long period
of time, due to the thickness of the material and the numerous welds of the entire
gate system.
After the announcement of the completion of the new flap gate, tests were car-
ried out in December 2019 directly at the gate manufacturer in the Czech Republic.
A tightness test was performed at the prescribed pressure for 10 minutes, a strength
test at the prescribed pressure for 15 minutes, a functional test of the gate and a
final check of dimensions and coatings.
Afterwards the process of transporting the flap gate to the Slovak Republic
followed, on the basis of special permits from transport authorities and with the help
of special equipment.
With respect to the limited spatial parameters of the access gallery, which was
only a few centimetres wider than the gate device itself, it was necessary to remove
from the gallery all equipment situated there (parts of automated control systems,
electrical equipment including fuse boxes, dilatometric clamps). The individual parts
of the flap gate – the flap itself, the electric motor, the counterweight – were stepwise
moved into the access gallery. The flap gate was finally assembled from above
mentioned components and fixed to the bottom outlet pipe in the engine room (Figs.
8 and 9).
After the replacement of individual gates of the bottom outlet – the conical
spray gate (in 2007) and inlet slide gate with a flap gate (in 2019), it can be stated
that the bottom outlet of the Hriňová Dam is ready for the next decades of safe and
reliable operation.
100
Q.105 - R.7
Fig. 8
View of the flap gate in the closed and open position.
Vanne à clapet en position fermée et ouverte.
Fig. 9
A view of a new flap valve mounted on the bottom outlet pipe.
Nouvelle vanne á clapet fixée sur les tuyaux de la conduite forcée.
4. CONCLUSION
The bottom outlet is one of the most important parts of the dam, not only in
terms of the "normal" operation of the hydraulic structure, but also in terms of its
basic technical safety and reliability. From this aspect, the functionality and reliability
of the bottom outlet gates is a fundamental prerequisite. Ensuring the operability of
individual bottom outlet gates, their maintenance and especially their repairs or
101
Q.105 - R.7
replacements always require an individual approach, which takes into account not
only the type and dimensional parameters of the gate, but also the construction
period of the hydraulic structure itself in the context of relevant historical relations,
as well as currently valid standards, requirements and circumstances associated
with the operation of the structure in a particular location.
REFERENCE
BEDNÁROVÁ, E. et al.: Dam construction in Slovakia. Originalities – Milestones –

Noteworthy facts. Bratislava, Kuskus 2010, 207 p. ISBN 978-80-970428-0-6.
102
Q.105 - R.8
-------
-------
FLOODING INCIDENT IN THE TINGUIRIRICA VALLEY, CHILE IN 2017 (∗)
Manuel Ignacio SABAT

TINGUIRIRICA ENERGIA, Chile
Lars ØDEGåRD
STATKRAFT, Norway
NORWAY
SUMMARY
Flash floods represent a significant natural hazard in small mountainous catch-

ments of the Patagonian Andes and have repeatedly caused loss of lives and
severe damages to infrastructure. Such an incident occurred in the Tinguiririca Val-
ley in southern Chile on 25th February 2017 where the two hydropower plants of
La Higuera and La Confluencia are located. The flash flood developed very sud-
denly and the powerplants were affected by overtopping of two intake dams and
damages on roads. The infrastructure is located in a valley with altitudes between
1,100 masl and 4,650 masl. Summer meteorological conditions in Chile are sand-
wiched between the Pacific Ocean to the west, influenced by both Humboldt Current
and ENSO; and the Los Andes mountain range to the east, with a considerable oro-
graphic impact given its mean altitude of 4,000 masl. Rain profiles usually develop
as local convective precipitation, which is aligned with summer high temperature in
the valley. The weather forecast for the day of the flooding said only a fraction of the
precipitation received and no preemptive warnings were issued from the national
Emergency Bureau. A flash flood developed on the Tinguiririca Valley with a high
intensity. Short duration rainfall measured at almost 30 mm in 30 minutes. This
activated fifteen ravines upstream of the Tinguiririca intake. This rainfall poured on
a valley with a zero-degree isotherm near 4,000 masl, with 95% of the catchment
area active. This is the main difference to meteorological events in the valley, which
∗ Inondation majeure dans la vallée de Tinguiririca (Chili) en 2017
DOI: 10.1201/9781003211808-54 103

Q.105 - R.8
are developed mainly due to orographic precipitation and during autumn, winter and
spring. Maximum flood flow was estimated at ∼250 m3 /s. The situation was very
challenging for the operation of the plants with a continued high risk of heavy rainfall
the following days.
RÉSUMÉ
Les crues soudaines représentent un risque naturel important dans les petits
bassins versants montagneux des Andes patagoniennes et ont causé à plusieurs
reprises des pertes en vies humaines et de graves dommages aux infrastructures.
Un tel incident s’est produit dans la vallée de Tinguiririca, dans le sud du Chili, le 25
février 2017, où se trouvent les deux centrales hydroélectriques de La Higuera et de
La Confluencia. L’inondation s’est développée très soudainement et les centrales
électriques ont été affectées par la submersion de deux barrages de prise ainsi que
des dommages sur les routes. L’infrastructure est située dans une vallée avec des
altitudes comprises entre 1100 et 4.650 m. Les conditions météorologiques esti-
vales au Chili sont influencées par l’océan Pacifique à l’ouest, sous l’influence du
courant Humboldt et d’El Niño, et par la chaîne des Andes à l’est avec un impact oro-
graphique considérable compte tenu de son altitude moyenne de 4000 m. Les pluies
se développent habituellement sous forme de précipitations convectives locales, qui
dépendent de la température élevée de l’été dans la vallée. Selon les prévisions
météorologiques pour le jour de l’inondation, seule une fraction des précipitations
réelles et aucun avertissement d’alerte n’ont été émis par le Bureau national des
situations d’urgence. Une crue éclair s’est développé sur la vallée de Tinguiririca
avec une intensité élevée et des précipitations de courte durée mesurées à près
de 30 mm en 30 minutes. Cela a activé quinze ravins en amont de la prise de Tin-
guiririca. Ces précipitations se sont déversées sur une vallée avec un isotherme à
zéro degré près de 4000 m, avec 95% de la zone de captage active. C’est la prin-
cipale différence avec les événements météorologiques habituelles dans la vallée,
qui se produisent principalement en raison des précipitations orographiques et pen-
dant l’automne, l’hiver et le printemps. Le débit maximal d’inondation a été estimé
à ∼250 m3 /s. La situation était très difficile pour l’exploitation des plantes avec un
risque toujours élevé de fortes pluies les jours suivants.
1. INTRODUCTION
The two run-of-river hydropower plants of La Higuera and La Confluencia,

owned by Tinguiririca Energía, were hit by a severe flash flood on 25th February
2017. The flash flood was unusual for this time of the year and developed very
suddenly. The weather forecast gave no warning and resulted in loss of lives in the
upstream village of Termas del Flaco. Much of the infrastructure in the Tinguiririca
Valley was impacted, including two intake dams.
104
Q.105 - R.8
2. PROJECT LOCATION
The Tinguiririca Valley is located in the Colchagua Province in the Chilean

Patagonian Andes, some 160 km south of Santiago.
The geographical coordinates of the site are:
Longitude: 70◦ 33’14”W
Latitude: 34◦ 50’3”S
Fig. 1
Tinguiririca Energía HEPP Complex location.
3. FLOODING EVENT
3.1. BACKGROUND
Flash floods represent a significant natural hazard in small mountainous catch-

ments of the Patagonian Andes, affecting small communities and civil structures
located near rivers and ravines. In Chile, a narrow country with a 4,000 km long
stretch of the Los Andes mountain range, dams and structures need to face the
considerable exposure and risks associated with flash floods.
3.2. METEOROLOGICAL CONDITIONS
The Chilean meteorology is characterized by the interaction between the mul-

tiple meteorological effects over the Pacific Ocean, such as the ENSO Humboldt
105
Q.105 - R.8
Current and the South Pacific High. The orographic effect from the Los Andes
Mountain Range with an average elevation of 4,000 masl provides for high summer
temperatures and a rain profile based on local convective thunderstorms events.
In addition, the narrow Chilean territory between the Pacific Ocean and the
peak of the Los Andes Mountain Range, plays a major role on the forecasting capa-
bilities, as the grid size of the existing meteorological models put a limit on the
ability to replicate complex topographies and steep valleys such as the case of the
Tinguiririca Valley.
3.3. 2017 FLOODING INCIDENT
On February 25th 2017, a flash flood developed on the Tinguiririca branch

of Tinguiririca Energía Complex. The precipitation event was not alerted by neither
the Emergency Bureau (ONEMI) which forecast major events and provides alerts
in order to activate the response nor available meteorological forecast from major
models such as ECMWF. The forecast alerted a minor rainfall, but the convective
front was measured on a meteorological station at almost 30 mm in 30 minutes.
Since this convective rain occurred during the summer high temperatures with a
zero-degree isotherm at 4,000 masl, some 95% of the catchment was active in
producing the flood which resulted in the activation of heavy erosion from fifteen
ravines on the Tinguiririca River.
The weather forecast during the days following the event on February 25th
said continued high precipitation which made it very challenging to access the site
for potential interventions to secure the assets.
The magnitude of the event was understood from a site visit shortly after. The
ravines had expanded and the downstream riverbed was extensively transformed.
Big boulders had moved and huge sediment deposits were observed near the Tin-
guiririca intake dam. Boulders had reached the crest of the dam. This was later
confirmed by video analysis from an expert consultant to confirm rheology changes
during the flood.
The flood peaked at an estimated 250 m3 /s and it overtopped two intakes on

the Tinguiririca branch.
4. DAMAGES TO STRUCTURES
4.1. THE PROJECT
La Higuera and La Confluencia are two run-off-river hydroelectric power plants

located in the valley of the Tinguiririca River, in the central part of Chile. Both power
106
Q.105 - R.8
Fig. 2
Tinguiririca Energía HEPP Complex – Intake scheme.
Fig. 3
Hypsometric curve for HLH and HLC intakes.
plants were commissioned in 2010 and jointly formed the Tinguiririca Energía Com-
plex. The Tinguiririca Energía Complex consists of ten intakes over the Tinguiririca
Catchment, on two main branches: Azufre River and Tinguiririca River. Figure 1
presents a general overview of the Tinguiririca Energía HEPP complex. The La
Confluencia intakes are located at 1,450 masl, while La Higuera intakes are located
at 1,100 masl. The snow line for regular events was estimated from the Direccion
General de Agua [6]. The estimation of snow line for pluvial contribution to rains
dependent on the latitude, for the Tinguiririca Catchment it is 2,000 masl. This data
107
Q.105 - R.8
integrated with the hypsometric curves for both La Confluencia and La Higuera’s
intakes present a contributive catchment of 11% and 17% of the total area for plu-
vial events. In hindsight, the unusually high snow line during the flooding event
contributed to the response from the catchment.
4.2. DAMAGES TO THE TINGUIRIRICA AND TRICAHUE INTAKES
The upstream intake, called Tinguiririca, has a 400 m long and 10 m high
earthfill dam that was overtopped.
The erosion of the downstream slope of the dam happened evenly throughout
the length of the dam with a significant loss of material. Luckily, no damage was done
to the inner clay core. The remnants on the dam crest comprised both boulders and
fine material, a testimony of the complex nature of the flash flood that had passed.
In addition, the I-45 public road leading to the Tinguiririca dam was destroyed
in several locations which made access to the site difficult.
The downstream intake, called Tricahue, has a 30 m long and 10 m high

earthfill dam that was overtopped.
Fig. 4
Tinguiririca intake dam after flooding event.
108
Q.105 - R.8
Fig. 5
Tricahue intake dam after flooding event.
The damages to the Tricahue dam were more severe than the Tinguiririca
intake dam, but the dam crest and inner core remained. The flood was recorded
on camera and in less than two minutes, the flash flood arrived and overtopped the
dam.
5. MITIGATION ACTIONS
5.1. BASELINE STUDIES
In order to prepare an action- and mitigation plan, baseline studies for the
catchment were developed. An on-site review of the ravines activated by the flooding
event was carried out in order to assess the different responses and erosion potential
of the catchment. The hydrology of the catchment was updated as well as detailed
characterization of sediment deposits through soil characterization and bathymetric
surveys.
In addition, the soil loss and sediment transport potential of the Tinguiririca
catchment were evaluated by use of the Universal Soil Loss Equation (USLE) with
use of isohyets and satellite information. The results were normalized in order to
present a traffic-light assessment, as shown in Figure 6.
109
Q.105 - R.8
Fig. 6
Hypsometric curve for HLH and HLC intakes.
5.2. SEDIMENT MANAGEMENT
Based on Annandale & Morris [2] and Morris & Fan [3], various sediment
management alternatives were assessed. Considering the complex layout of the
catchment and intake location, a Sediment Flushing Strategy was implemented.
This is based on four main pillars:
1. Sediment-guiding channel: An excavated channel over the delta has its main
objective to guide sediment toward the gated section. The capacity and shape
were calculated based on methodologies from Altunin [1], Griffiths [3], Maza
& Cruickshank [4] and Simons & Anderson [7].
110
Q.105 - R.8
2. Monitoring System: A gauging station with pluviometry, temperature and tur-

bidity sensors is located 5 km upstream the intake providing data every 10
minutes and integrated into SCADA to alert events.
3. Operational Rules: In case of a flood/rain event, the gates will open in order to
empty the small reservoir at the intake and maximize the combined sediment
handling capabilities of the intake.
4. Adaptive Analysis: Monitoring of progress of sediment delta, review of channel
condition and assessment of gates condition provide feedback to review the
operational rules or adapt the channel in order to adapt the theoretical solution
to the real site conditions.
The strategy was tested on a physical scale model in order to assess

the operational rules and verify the sediment transport capacity of the flushing
activities.
5.3. OPERATION
As part of the improvements of the response plan for the Tinguiririca Dam,
additional communication systems were explored, including the use of the existing
medium voltage line. To ensure a reasonable response time for future flash floods,
backups have been put in place for the energy and control systems.
The intakes of La Higuera and La Confluencia powerplants have radial gates

with logic such as to coordinate the opening of gates in the event of floods. The
operation logic for the radial gates has been updated in order to react preemptively
and include several layers of defence, such as nitrogen accumulators, to provide
failsafe response action.
6. CONCLUSIONS
The flood event occurred on February 25th 2017 resulted in many Lessons
Learned with the main takeaway on implementation of new technology to better
forecast “hard-to-find” convective thunderstorms and a proactive approach on pre-
paredness through live data gathering from upstream meteorological stations and
control logics for the safe operation of the assets. In addition, the monitoring of
the catchment through satellite images is beneficial for an adaptive approach to
sediment management.
111
Q.105 - R.8
REFERENCES
[1] Altunin, S.T. 1962. “The Rivers Regulation”.
[2] Annandale, George W. & Morris, Gregory 2016. “Extending the Life of
Reservoirs: Sustainable Sediment Management for Dams and Run-of-River
Hydropower”.
[3] Griffiths, George A. 1989. “Conversion of Braided Gravel-Bed Rivers to Single-

Thread Channels of Equivalent Transport Capacity.” Jour. of Hydrology (N.Z.)
28(1) 63–75.
[4] Maza-Alvarez, J.A. & Cruickshank, C. 1973. Stable channels in alluvium.

International Symposium on River Mechanics, IAHR, Vol 1.
[5] Morris, Gregory & Fan, Jiahua. 2010. “Reservoir Sedimentation Handbook”.
[6] Dirección General de Aguas. 1995. “Manual de cálculo de crecidas y caudales

mínimos en cuencas sin información fluviométrica”.
[7] Simons, D.B. & Anderson, M.L. 1963. “Uniform water conveyance channels in
uniform material”. Transactions ASCE, Vol D, Part D, Paper No 3399.
112
Q.105 - R.9
-------
-------
DAM RESILIENCY IS MORE THAN JUST DAM SAFETY (∗)
William F. FOOS, CPP, PSP

Vice President, Director of Security and Safety Services,
GANNETT FLEMING, Inc. Denver, Colorado
Enrique E. MATHEU, PhD

Associate Director for Sectors and Councils, Stakeholder Engagement Division,
CYBERSECURITY AND INFRASTRUCTURE SECURITY AGENCY,
Arlington, Virginia
Dean DURKEE, PhD

Vice President, Sr. Consultant,
GANNETT FLEMING, INC. Phoenix, Arizona
Michelle YEZIERSKI, PE
Section Chief, Dams Section, Stakeholder Engagement Division,
CYBERSECURITY AND INFRASTRUCTURE SECURITY AGENCY,
Arlington, Virginia
USA
SUMMARY
This paper proposes a more holistic approach to dam safety by considering

the inclusion of crisis management, public safety, and security program areas into a
broader dam resiliency framework. Awareness of what resiliency entails can lead to a
clear understanding of how dam safety, security, public safety, emergency response,
and crisis management can contribute to the overall resiliency of a project through
prevention, mitigation, preparedness, response, and recovery. A multi-disciplinary
approach focused on enhancing resiliency can provide an effective framework to
∗ La résilience des barrages ne se résume pas à la simple sécurité des barrages
DOI: 10.1201/9781003211808-55 113

Q.105 - R.9
integrate dam safety, public safety around dams, security, and crisis management
programs and leverage their common elements. Risk management has become
part of the culture of these different areas and risk management principles guide
the implementation and the measurement of success for each of these programs.
In addition, they share common fundamental principles regarding their approaches
to planning, implementing, testing, exercising, and improvement.
A coordinated approach to enhancing resiliency by appropriately nesting the

individual programs would provide a unified framework towards the safe operations
and maintenance of dams. This would strengthen individual program areas while
providing an integrated platform to deal with incidents or events once they occur.
This type of approach would enhance the overall ability of the facility and overarching
organization to absorb disruptions, maintain operations, manage crises as they
unfold, get services back as quickly as possible, and incorporate lessons learned
from past incidents.
RÉSUMÉ
Ce rapport propose une approche holistique de la sûreté des barrages en

envisageant l’inclusion des domaines de programmes de gestion de crise, de pro-
tection publique et de sécurité dans un cadre plus large de la résilience des barrages.
La prise de conscience de ce qu’implique la résilience peut permettre de compren-
dre clairement comment la sûreté, la sécurité, la protection publique, la réponse aux
urgences et la gestion de crise des barrages peuvent contribuer à la résilience glob-
ale d’un aménagement par la prévention, l’atténuation, la préparation, la réponse
et le rétablissement. Une approche pluridisciplinaire axée sur l’amélioration de la
résilience peut fournir un cadre efficace pour intégrer la sûreté des barrages, la pro-
tection publique autour des barrages, la sécurité et les programmes de gestion de
crise et tirer parti de leurs éléments communs. La gestion des risques est devenue
partie intégrante de la culture de ces différents domaines et les principes de gestion
des risques guident la mise en œuvre et la mesure du succès de chacun de ces
programmes. En outre, ils partagent des principes fondamentaux communs en ce
qui concerne leurs approches de la planification, de la mise en œuvre, des tests,
des exercices et de l’amélioration.
Une approche coordonnée visant à renforcer la résilience en imbriquant

de manière appropriée les différents programmes offrirait un cadre unifié pour
l’exploitation et l’entretien fiables des barrages. Cela renforcerait les différents
domaines de programme tout en fournissant une plate-forme intégrée pour traiter
les incidents ou les événements une fois qu’ils se produisent. Ce type d’approche
renforcerait la capacité générale de l’installation et de l’organisation principale à
absorber les perturbations, à maintenir les exploitations, à gérer les crises au fur et
à mesure de leur déroulement, à rétablir les services le plus rapidement possible et
à intégrer les enseignements tirés des incidents passés.
114
Q.105 - R.9
1. INTRODUCTION
The most recent trend in critical infrastructure management has been an

increased focus on resiliency. There are multiple definitions of resiliency but most of
them refer to the ability to adapt to changing conditions and prepare for, withstand,
and rapidly recover from disruptions. This concept encompasses a broad domain
of hazards, threats, or incidents including natural disasters, pandemics, industrial
accidents, cyber incidents, acts of terrorism, and criminal activity. The increasing
severity of climate-related natural disasters and the number of dam failures over
recent years demonstrate that resilient dams not only require a sound dam safety
program but also complementary programs addressing crisis management, security
and public safety through an integrated framework. The various aspects and princi-
ples associated with crisis management, public safety, and security along with dam
safety can help achieve a level of prevention, preparedness, response, and recovery
capabilities to achieve the desired resiliency objectives.
This paper takes a look at traditional dam safety programs and expanding it
to a more holistic process that incorporates crisis management, public safety, and
security as part of a dam safety framework. Awareness of what resiliency entails pro-
vides a clear understanding of how each of the four areas (dam safety, security, public
safety, and crisis management), contribute to this holistic approach. Resiliency of a
dam safety program can be achieved through prevention, mitigation, preparedness,
response, and recovery.
2. BACKGROUND
In history, dams have been identified and defined as structures that help com-
munities become more resilient. People focused on the benefits provided by the
dams that supported their local communities in the following ways:
• Flood control which helped communities withstand frequent and often devas-
tating flood damage.
• Water supply which provided essential and reliable water to the communities
often helping those communities become more resilient during extreme dry
spells when water could be scarce.
• Economic viability, a dividend from the presence of a dam which often-allowed
communities to harness renewable power, provide recreational benefits, and
aid in the transportation of economic goods resulting in a positive impact on
communities.
The mere presence of a dam often provided the local community with a resilient
capability that it may not have had otherwise. Today however, people’s perspective
has often changed from seeing dams as critical assets to rather a liability for a
115
Q.105 - R.9
Fig. 1
Dam Safety, Public Safety, and Security Domains
Domaines de la sûreté des barrages, de la protection publique et de la sécurité
community. The increasing numbers of dam failures in the U.S. and around the
world are bringing into question the positive contributions of dams to the overall
resilience of their surrounding communities.
This paper highlights the potential for expanding the various components of
traditional dam safety practices by incorporating additional elements and comple-
mentary disciplines into an integrated dam resiliency framework. This concept is
notionally depicted in Figure 1, which provides an illustration of the relationships
between dam safety, public safety, and security domains.
2.1. THREATS, HAZARDS, AND VULNERABILITIES
Both natural and man-made (deliberate or accidental) threats and hazards

have the potential to harm, damage, incapacitate, or destroy critical infrastructure
facilities such as dams. The literature shows different categorizations depending on
the specific context of the analysis, but a general list includes [1]:
• Climatological Events (extreme temperatures, drought, wildfires)

• Hydrological Events (floods)
116
Q.105 - R.9
• Meteorological Events (hurricanes, typhoons, severe convective storms,

severe winter storms)
• Geophysical Events (earthquakes, tsunamis, volcanic eruptions)
• Pandemics (global disease outbreaks)
• Space Weather Events (geomagnetic storms)
• Technological and Industrial Accidents (structural failures, industrial fires,
hazardous substance releases, chemical spills)
• Unscheduled Disruptions (aging infrastructure, equipment malfunction, large
scale power outages)
• Criminal Incidents and Terrorist Attacks (vandalism, theft, property damage,
active shooter incidents, kinetic attacks)
• Cyber Incidents (denial-of-service attacks, malware, phishing)
• Supply Chain Disruptions (physical or cyber incidents affecting manufacturers,
suppliers, or other third parties)
The facility may exhibit specific vulnerabilities, i.e. physical features or oper-
ational attributes that make it susceptible to disruption, destruction, or exploitation
by these threats or hazards [2]. In general, vulnerabilities are identified as part of
the risk analysis process. For example, engineering analysis can be used for the
identification and development of potential failure modes through a Potential Failure
Mode Analysis (PFMA). Risk analysis can be performed using both semi-quantitative
and quantitative approaches taking into account the corresponding likelihood of fail-
ure and the associated consequences. PFMA provides a structured approach to
identify vulnerabilities while categorizing and prioritizing them within a risk-driven
context.
2.1.1. Climate-related factors
Vulnerability to climate change has been defined by the Intergovernmental

Panel on Climate Change as “the propensity or predisposition to be adversely
affected” [3]. In considering the perspective of scenarios of climatic anomalies, vul-
nerability may be regarded as a possibility of “future damage” [4]. These anomalies
can include extremes in temperatures and in rainfall, which appear to be occurring
on a more frequent basis. It is almost certain that these effects will continue, and
likely accelerate, depending on how fast the concentrations of greenhouse gases in
the atmosphere continue to increase [5]. Weather variability and extreme weather
events should be considered as factors leading to potential vulnerabilities. These
elements, often not accounted for during the original design of the dam, should be
appropriately considered during risk and resiliency management planning.
The resulting impacts may include not only changes in frequency and severity
of extreme events (from extreme rainfall events to severe droughts), but also shift-
ing weather and temperature patterns that could reduce the generating capacity of
hydropower plants and exacerbate competing demands for water resources [6].
117
Q.105 - R.9
2.1.2. Socioeconomic factors
Socioeconomic factors refer to pressures on the facility resulting from compet-

ing interests that are not necessarily focused on life safety. These pressures may
come from sources internal to the organization, such as the evaluation or trade-
off of commercial risks versus societal risks, as well as external sources such as
limitations on funding in the case of publicly owned dams operating through gov-
ernment funding. These pressures often limit the ability of the dam operator to
adequately implement a dam safety program. As an example, in the case of the
Oroville Spillway Incident (California, USA) in 2017 the forensics team determined
that the question of “why the incident happened” involved factors specific to the dam
and spillway as well as additional factors involving organizational, regulatory, and
industry aspects [7].
2.1.2.1. Social and Economic Pressure

Social pressure is most often measured in terms of economic benefits or
impacts when related to dams. Today’s society is looking to maximize the use of
the dam and its potential benefits to the community and the public. Updated design
and engineering analysis, modified improvements, and new technology integration
are required during the life cycle of a dam. This is to help in the extension of capacity
to meet service demands, creating a need for linking the system ‘engineering-
on-design’ with the ever-changing domains of technology, society, the economy,
legislation, and politics, which determine the profiles of service demand and cor-
responding expected performance. The economic pressures put on dams is the
ever-increasing scenario of expected integration among other critical infrastructures,
with many of them operating under very different regulatory conditions. These eco-
nomic pressures are challenging the safety margins designed to prevent system
failure. These increased pressures are leaving structures struggling to cope with
the expected and unexpected stresses put onto these systems [8].
2.1.2.2. Environmental Pressure

The general term “environmental security” has been used to describe a broad
variety of issues reflecting the ability of a society or system to withstand environmen-
tal asset scarcity, environmental risks or adverse changes, or environment-related
tensions or conflicts [9]. Disappearance of biodiversity, competing priorities for water
resources, effects of chemicals, and pollution of water and air are challenging water
management professionals to find innovative solutions to address environmental
pressures. The balancing act of meeting increased population demands along with
meeting the needs of the entire ecosystem that depends on the water is becoming
increasingly complex. These pressures impact the timing of when new structures
are built, or old ones are rehabilitated. Delays exacerbate the pressures on the
increasingly limited water resources.
118
Q.105 - R.9
2.1.3. Complacency factors
Complacency refers to the lack of concern for or awareness of a problem,

defect, or potential danger. The concept of complacency has interpretations at
the individual and organizational levels. As described by Weiner [10]: “Compla-
cency is caused by the very things that should prevent accidents, factors like
experience, training and knowledge contribute to complacency.” We only need to
look at the Independent Forensic Team Report of the Oroville Dam Spillway Inci-
dent (January 5, 2018) section 6.1 Overconfidence and Complacency Regarding
Infrastructure Integrity to see this factor called out. “A contributing factor to DWR’s
overconfidence and complacency was a somewhat widespread belief within DWR
that the SWP was designed by the “best of the best” – a believe passed on through
two generations to the current generation, and possibly increasingly mythologized
by each generation.” [7].
2.1.3.1. Overreliance on Technology

Computer systems may not be able to make human errors, but as the result of
human design, they may potentially automate human errors. Proponents of techno-
logical dependence describe cyborgism as a way of outsourcing certain functions
that we can’t otherwise perform on our own or can be routinely performed more effi-
ciently or quickly through automated processes. The challenge is, we may become
dependent on these systems without eventually understanding how the calculations
are performed or the limits of the underlying assumptions. The increased use of
Operational Technology (hardware and software systems that monitor and control
physical equipment and processes) may lead to complacency if not accompanied
by the appropriate contingency protocols or alternative procedures.
2.1.3.2. Operations and Maintenance

Poor operation and maintenance protocols and procedures can contribute just
as significantly to increased vulnerabilities as design limitations. Manmade struc-
tures have inherently factored into their design an operations and maintenance
component. Many dams built in the U.S. were designed with a life expectancy of
50 years. As a result of proper operation and maintenance, many dams today out-
live this life expectancy. These programs can help to extend the life expectancy
of dam structures beyond their intended design. Inversely, improper operation and
maintenance may reduce the life expectancy of a structure causing it to become a
hazard. For example, allowing vegetation to grow unchecked on an earthen dam
may lead to weakened slope stability and possibly channeling of rainwater caus-
ing the potential for erosion. Additionally, burrowing animals if not taken care of
may create natural seepage areas within the structure that could lead to piping
failure.
119
Q.105 - R.9
2.1.3.3. Human Nature

Humans are not perfect and make mistakes. The goal is to manage and limit
the number of mistakes that occur. The problem of human error can be considered
from two different perspectives: the person approach and the system approach
[11]. Each has its model of error causation, and each model gives rise to different
philosophies for error management. Understanding these differences has important
practical implications for coping with the ever-present risk of mishaps. The person
approach focuses on the errors of individuals: forgetfulness, inattention, or moral
weakness. The system approach concentrates on the conditions under which people
work and focuses on measures to avert errors or mitigate their effects, thus leading
to a more resilient environment.
2.2. RESILIENCY
Converging on a universal definition of resiliency may be difficult; everyone

from engineers, social scientists, to emergency managers and others may define it
differently. Resiliency may mean different things in different contexts and it will have
different interpretations for a specific infrastructure facility, a business, a community,
a city, or a region. In general, the concept of resiliency assumes a disturbance in
the system’s normal operating environment that has the potential, if not the effect,
of disrupting normal system performance. In the United States, the Department of
Homeland Security defines resiliency as “the ability to adapt to changing conditions
and prepare for, withstand, and rapidly recover from disruption” [2]. This broad def-
inition encompasses the ability of systems, infrastructures, government, business,
communities, and individuals to resist, tolerate, absorb, recover from, prepare for,
or adapt to an adverse occurrence that causes harm, destruction, or loss
The literature contains multiple definitions of resiliency, but in general they

capture the same fundamental concept. For example, the American Water and
Wastewater Association defines resiliency as “the ability of an asset or system to
withstand an attack or natural hazard without interruption of performing the asset or
system’s function or, if the function is interrupted, to restore the function rapidly” [12].
Some of the definitions vary depending on the object of analysis and the specific
discipline (infrastructure facility, business, organization, or community). For exam-
ple, in the fields of engineering and construction, resiliency is the ability to absorb
or avoid damage without suffering complete failure and is an objective of design,
maintenance and restoration for buildings and infrastructure.
Many authors have characterized the attributes or features associated with

resilient characteristics [13]. These attributes, which are relevant throughout the
entire lifecycle of the system (design, normal operations, incident, response, and
recovery), include robustness to withstand unforeseen demands (this includes the
ability to maintain critical operations and functions during a disruption), resource-
fulness to identify a problem and to respond effectively (this includes the ability to
prepare for, respond to and manage a crisis or disruption as it unfolds), rapidity
120
Q.105 - R.9
to restore functionality quickly (this includes the ability to return to and/or reconsti-
tute normal operations as quickly and efficiently as possible after a disruption), and
adaptability to absorb new lessons that can be drawn from an incident (this involves
revising plans, modifying procedures, and introducing new tools/technologies before
the next crisis).
2.2.1. Achieving enhanced resiliency
Incorporating enhanced resiliency as an additional objective for guiding deci-

sion making allows for the consideration of courses of action to increase prepared-
ness for an incident, implement measures to mitigate its effects, improve emergency
response and business continuity operations, and expedite recovery. This can be
achieved by appropriately linking risk management and crisis management activities
as part of an integrated approach.
Enhanced resiliency objectives supplement risk management efforts with haz-

ard or threat mitigation measures, response/emergency action planning, and actions
undertaken to enhance continuity of operations. The resiliency focus also allows
for the appropriate consideration of dependencies related to key external resources
required by the facility to maintain operations, susceptibility to disruption of the asso-
ciated supply chains, and the ability to identify alternative resources or backups.
In addition, the resiliency focus provides for a holistic consideration of the public
safety, security, and emergency management domains. This includes the capability
to respond to an accident or an incident without needing an immediate response from
external first responders, as well as the mechanisms to interact with first respon-
ders (law enforcement, fire, or emergency medical personnel). Finally, an integrated
approach to enhanced resiliency provides a useful framework to implement activities
and programs designed to be effective and efficient in returning operating conditions
to a level that is acceptable to the facility. This includes restoration agreements with
entities other than emergency responders, procedures/assets required to support
facility restoration, and needs for specialized materials or equipment.
3. RISK MANAGEMENT
Defining risk management may be different depending on the industry, but

universally risk management can be best described as “the practice of identifying
potential risks in advance, analyzing them and taking precautionary steps to reduce
the risk” [14]. The total process of risk analysis includes risk evaluation, comparison
of the risks, identifying what level of risk is tolerable, determining whether risks
are adequately controlled and, if not, what recommendations might be done to
adequately control or mitigate the risk [15]. These risk management principles can
be found in dam safety, public safety around dams, and dam security programs.
121
Q.105 - R.9
In most cases, there are only three ways to manage risk, either accept it,
transfer it, or mitigate it. Within the context of risk management framework, one
must start with the premise that all risks are known. However, in the environment
we live in today, not all risks are known, and in some cases, they are unknown
until the hazard occurs, or the event takes place. These concealed or unrevealed
risks (sometimes referred to as black swan events) may occur in unexpected ways.
Some might say the events that led to the Oroville Dam Spillway incident could be
characterized as a black swan event. The items identified through the Independent
Forensic Report were unknown to the dam owner. The events that led up to the
erosion of the flood control outlet (FCO) spillway chute was not considered as a
potential issue or Potential Failure Mode (PFM), additionally had the chute not failed,
the flaws in the ogee emergency spillway design (never used before) likely would
have gone undetected. A full failure of the ogee emergency spillway was averted
by redirecting water back through the FCO, sacrificing the spillway chute but not
the emergency spillway. The combination of events like this occur, the best crisis
management plan when executed is going have challenges. This type of black swan
event demonstrates the need to have contingency plans that help impower decision
makers to respond and recover from these unexpected incidents.
3.1. ASSET RISK MANAGEMENT
Using an asset risk management approach allows organizations to prioritize

their investments and identify where capital resources should be allocated. In the
context of risk management, asset risk management is only one part of the equation
for an organization. Organizations must manage personnel risk, information risks,
and operational and maintenance risks. Various dam safety programs utilize Risk
Informed Decision Making (RIDM) to inform the asset risk management process.
3.1.1. Dam safety risk management
The fundamental principles of dam safety are not new. The Code of Ham-
murabi is a Babylonian code of law of ancient Mesopotamia, dated to about 1754
BC. Law #53 from the Code of Hammurabi states: “If anyone be too apathetic to
keep his dam in primly condition, and does not so keep it; if then the dam break and
all the fields be flooded, then shall he in whose dam the break occurred be sold for
money, and the money shall replace the crops which he has caused to be ruined.”
Dam safety rules and regulations today are provided by various governmen-
tal agencies to provide the essential actions required by dam owners to maintain
safe dams. Risk analysis can form the basis of a dam safety program. Hartford and
Baecher (2004) state “Risk analysis for dam safety seeks to address several funda-
mental issues including: what can happen to cause the dam to fail or be damaged?
How probable is it? What are the consequences, or risks?” [16].
122
Q.105 - R.9
The US Bureau of Reclamation uses risk analysis as the primary support for
dam safety decision making. The framework for decision making in dam safety as
three distinct components, risk analysis, risk assessment, and risk management.
“Risk management is the overarching activity when risk is used to inform dam safety
decision making and builds on risk analysis and risk assessment phases” [17].
3.1.2. Security risk management
The main purpose of having an effective security program is to manage the risk
of manmade attacks. The principles associated with security are deter and prevent,
detect and assess, delay and deny, respond, and recover if required. The appli-
cation of these principles is best managed through a risk process. Understanding
consequences, vulnerabilities, threats, and overall risks affords dam owners and
operators the opportunity to select and implement physical security measures that
may best address and mitigate those risks. The basic principles of physical secu-
rity are similar across many types of structures. The main differences are given by
the degree of security required and the sophistication of layered protection needed.
Physical security measures are generally designed and installed to perform spe-
cific functions in relation to an attack or disruption of an asset. Depending on the
results of assessments, regulatory requirements, and available resources, owners
and operators may choose to implement physical security measures to improve one
of these functions or implement a comprehensive approach that integrates benefits
across multiple functions [18].
Security is not limited to the physical environment but encompasses the ever-
evolving cyber environment. Enhancing security and resiliency in the cyber realm
begins with identifying and assessing critical cyber assets and systems, as well
as the facility’s and connecting networks’ cyber risks and vulnerabilities. Deficient
cybersecurity functions can then be selected for further analysis. Once the cyber
landscape has been understood, the owner or operator is able to implement tar-
geted cybersecurity practices to improve the cybersecurity posture and address the
risks and vulnerabilities uncovered. In implementing cybersecurity practices, owners
and operators may choose to implement practices to address select deficiencies.
However, given the evolving cyber landscape and myriad unknowns, owners and
operators may best improve their cybersecurity posture through persistent imple-
mentation of a comprehensive cybersecurity risk management approach organized
around the following functions: identify, protect, detect, respond, and recover [18].
Owners and operators must also consider personnel security practices to help
protect their investments and assets. The errant or illicit actions of one person can
cause significant damage to critical assets, disrupt operations, or harm personnel or
the public. Assessing threats includes conducting background screening for poten-
tial and new hires or contractors, periodically rescreening existing personnel, and
determining the vulnerability to insider threats. Personnel security measures include
background investigations, training to increase security awareness and education,
123
Q.105 - R.9
exercises and drills to hone skills and knowledge related to specific types of security
incidents, and development and management of response and recovery plans [18].
3.1.3. Public safety risk management
Today’s public expects and demands access to water retention structures that
provide various forms of recreational activities. Known as an attractive nuisance, the
water impoundments and reservoirs created by dams are natural attractions to the
public’s various outdoor activities. These activities may lead the public to be exposed
to hazards created by the dam structure. These various hazards such as high flow
water releases, water conveyance structures through the dam, as well as high water
velocity and hydraulic jumps to name a few, can result in injury and even death to
the public if they are exposed to them. As of July 2020, the U.S. had experienced 46
incidents, 33 deaths, and 18 injuries related to public safety incidents at or around
dams [19]. The risk management model for public safety incidents around dams is
measured simply by the presence of the hazard potential and actual incident data.
An effective Public Safety Management Program includes several steps that

provide a construct to manage risk and guides response to and recovery from an
incident. The steps include planning (establish policies, accountabilities, identify
hazards, assess risks, and define countermeasures), implementation (conduct
training, set operational controls, establish warning systems, and place physical
barriers where possible, inform the public through education activities, and conduct
inspections), checking and corrective action (when an incident does occur, per-
form a post incident review, identify compliance issues if necessary, manage change
in the program to mitigate future incidences from occurring, make updates to plans
to ensure that they remain current), and reporting (inform decision makers of the
changes required to mitigate future risks and hazards to the public.
Public safety around dams is intended to ensure hazards created by the

presence and operation of dams are appropriately identified for public awareness.
Additionally, the warning should provide enough time and distance to ensure that
the public can remove themselves from the hazard or that they are prevented from
entering a hazardous area for their own protection.
3.2. ENTERPRISE RISK MANAGEMENT
The underlying premise of enterprise risk management is that every entity or

organization exists to provide value for its stakeholders. All entities face uncertainty,
and the challenge for management is to determine how much uncertainty to accept
as it strives to grow stakeholder value. Uncertainty presents both risk and oppor-
tunity, with the potential to erode or enhance value. Enterprise risk management
124
Q.105 - R.9
enables management to effectively deal with uncertainty and associated risk and
opportunity, enhancing the capacity to build value [20]. The purpose of an enter-
prise risk management program is to combine all the different forms of technical risk
programs i.e. dam safety, security, and public safety, along with personnel risks, and
program risks to help organizations make a more informed decision from a common
risk language platform.
The value of this process is it maximizes the opportunity for management

to set strategies and objectives striking an optimal balance between growth and
returns, goals and related risks, and efficiently and effectively deployed resources
in the pursuit of the entity’s objectives. Enterprise risk management aims at aligning
risk appetite and strategic objectives, enhancing risk response decisions, reducing
operational surprises and losses, identifying, and managing multiple and cross-
enterprise risks, seizing opportunities, and improving deployment of capital [20].
4. CRISIS MANAGEMENT
Critical to being resilient is to have an effective crisis management program.

Dam owners and operators can enhance the overall resilience of their facilities
through the development of plans to guide emergency response, rapid restoration,
and continuity of operations. In general, these plans describe response actions that
the facility will take for multiple scenarios of concern. It is important that these plans
are developed and implemented as part of an integrated approach that serves as
an overarching framework outlining how the facility will respond to an emergency or
incident, regardless of the event or triggering source. Crisis management programs
are intrinsically linked to the risk management strategies adopted by the organiza-
tion as they focus on minimizing safety and economic impacts, limiting operational
disruptions, and achieving prompt recovery.
We use the term “integrated crisis management” to denote a holistic approach

to preparedness, emergency response, continuity, and recovery. Therefore, for the
purposes of this paper, this term includes all elements related to emergency man-
agement, emergency operations, recovery, and business continuity. In a stricter
sense, terms such as disaster management or emergency management focus on
incidents that threatens life, property, operations, information, or the environment
whereas the term crisis management is generally associated with an “organizational”
focus (i.e., impacts on the entity’s operations, brand, image, reputation, market
share, ability to do business, or relationships with key stakeholders). An integrated
crisis management approach enhances the ability to manage issues or incidents with
the potential to impact the facility, disrupt operations, affect external stakeholders,
degrade the ability to do business, or affect the brand and reputation of the organiza-
tion. Crisis management efforts should be intrinsically linked to the risk management
strategies adopted by the organization as they focus on minimizing safety and
economic impacts, limiting operational disruptions, and achieving prompt recovery.
125
Q.105 - R.9
Due to the differences in project configurations, engineering details, project

benefits, and potential consequences from adverse events, each dam is unique.
As such, the scope and extent of crisis management activities aligned with the
corresponding risk management strategies will be unique as well. The corresponding
risk assessments, operational needs, regulatory requirements, and availability of
financial resources will inform this determination.
4.1. PLANNING
Depending on the complexity of the facility and its operations, an integrated

crisis management program may include several different plans to guide emergency
response, rapidly restore operations, and continue operations when key personnel
are not available. Of course, dam owners and operators must ensure that the scope
and structure of these plans conform to any applicable regulatory requirements. It is
critical that the plans also incorporate guidelines and specific actions regarding how
the facility will communicate and share information not only with its own personnel
but also with external entities during a time of crisis.
Coordination of plans with the local agencies involved in emergency response

efforts will enhance the effectiveness of those plans. For example, emergency man-
agement agencies may use the information to facilitate the implementation of their
responsibilities and local plans, such as an emergency operation plan or a warning
and evacuation plan. Maintaining close contact with appropriate government agen-
cies can not only assist in the development of effective evacuation plans but can also
strengthen the information sharing protocols required to share timely and accurate
information that can support decision making.
The following includes a summary of the types of plans that could be consid-
ered as part of an integrated crisis management program. Depending on the type of
facility, purposes, and operational characteristics, as well as risk environment and
organizational priorities, some of these plans may not be applicable.
4.1.1. Emergency action plans
Emergency Action Plans enable an organization to mobilize a pre-planned

response to prevent an uncontrolled release of water from the dam and to initiate
community actions to maintain public safety in case of such a release. Dam safety
programs have long relied on emergency action plans to guide response in critical
situations. While general templates are available, each plan should be site-specific
(because conditions are unique at each dam and at locations downstream of that
dam) while conforming to any applicable jurisdictional or regulatory requirements.
126
Q.105 - R.9
4.1.2. Site security plans
Site Security Plans document security systems, procedures, actions, and per-
sonnel responsibilities in response to suspicious situations, potential threats, or
security incidents. Each site-specific plan should be aligned with the priorities estab-
lished by the corresponding security risk assessment in addressing the relevant
security threats and identified vulnerabilities. In addition to describing security proce-
dures and systems (perimeter control, signage, fences, barriers, intrusion detection
systems, access control, security guards, etc.), the plan should detail actions and
notification processes in response to suspicious activities and security incidents
involving staff, contractors, visitors, or outsiders.
4.1.3. Cyber incident response plans
Cyber Incident Response Plans define specific cyber incidents the organiza-
tion expects to encounter and outline the steps that should be taken to enable a
timely response and mitigate potential impacts while properly coordinating across
the organization. The plan should cover the different stages associated with poten-
tial incidents, from detection and identification to containment, eradication, and
restoration.
4.1.4. Recovery Plans
Recovery Plans are used to help prepare for quick damage repair to at least
partially restore project functions as well as long-term repairs to fully restore the
project. Recovery plans address the hazards likely to impact the facility and its
operations and include options to minimize consequences to each critical component
including procurement of equipment or supplies, vendors for rental equipment or
repairs, and communications with relevant authorities. Recovery plans should be
general enough to be useful for a broad spectrum of scenarios or specific enough
for those events that have a high probability or risk of occurring.
4.1.5. Continuity plans
Continuity Plans help facilitate the performance of an organization’s essential

functions during any situation that may disrupt normal operations. The continuity
plan can encompass a wide range of topics such as leadership delegation, physical
relocation of worksites, data preservation, and large-scale absenteeism.
The process of business continuity management requires advanced planning

and preparation to maintain business functions or quickly resume after a disaster has
127
Q.105 - R.9
occurred. Business continuity planning can only occur within the context of risk iden-
tification and management. The potential risks can include natural, manmade, and
or accidental hazardous events. ISO 22301 provides a framework to plan, establish,
implement, operate, monitor, review, maintain and continually improve a business
continuity management system [21]. It is expected to help organizations protect
against, prepare for, respond to, and recover when disruptive incidents arise. To
minimize disruptions, organizations should prepare business continuity plans that
address dependencies on critical vendors and suppliers.
4.2. EXERCISING THE PLANS
While planning is essential for effective crisis management, more complete

preparation requires periodic exercises that test the implementation and execution
of plans, policies, and procedures. Exercises constitute the natural mechanism to
validate them, ensuring they are up to date and clarifying roles and responsibilities
during the stressful times associated with incidents. Other benefits of conducting
exercises include:
• Testing response actions and the adequacy of information and resources

required to make decisions.
• Improving coordination efforts between the facility and the emergency man-
agement entities responding to a situation at the dam.
• Identifying weaknesses and deficiencies in scripted response actions and
recommending improvements to the corresponding plans to improve its
efficacy.
• Testing staff’s knowledge of their roles and responsibilities.
• Updating contact information to reflect changes in personnel and response
responsibilities, to ensure notifications are received by the right personnel at
the right time.
5. CASE STUDY OROVILLE DAM SPILLWAY INCIDENT
The concepts shared in this paper bring into question whether a failure of a
dam can contain multiple, interwoven components of cause and effect, which make
it an ideal candidate for a holistic risk-management approach. Additionally, it might
also be said that by taking a holistic approach to dam safety could potentially lessen
the impacts of an incident or the likelihood of it occurring.
The events of February 2017, leading to the spillway incident at Oroville Dam
(California, USA) resulted in the evacuation of over 180,000 people. The cost of
damage and repair was estimated in excess of $1.1 billion [22]]. The combination of
events contributed to the emergency none of which can be linked to one contribut-
ing cause but should be looked upon as challenges to the overall resiliency of the
128
Q.105 - R.9
structure and organization that maintains and operates the project. The following
are findings and observations as published by various organizations.
Climate-related Factors: Researchers at Scripps Institution of Oceanography

at the University of California San Diego and the University of Colorado analyzed
the event to understand why there was such a large inflow of water into the reservoir
behind the dam leading up to the breach. What was extraordinary, according to the
researchers, was the unusually deep snow recorded in the northern Sierra Nevada
mountains before the storm event. Subsequently, several records were set for how
much snowmelt occurred during the atmospheric river. The melt took place because
of unusually warm and wet conditions during the atmospheric river, and its increased
water available for runoff by 37 percent over rain alone, straining the capacity of
California’s second-largest reservoir.
The study determined that snowmelt in the Lake Oroville watershed reached
200–400 millimeters (8–16 inches) in some areas that additionally received up
to 500 millimeters of rain (20 inches). Collectively, the large influx of water
(71–91 centimeters [28–36 inches] in places) overwhelmed soil storage and resulted
in tremendous runoff. The researchers concluded that without the unusual warmth
that caused extreme snowmelt from the atmospheric river, the inflows to Lake
Oroville would have been less and the situation around the spillway failures may
have been less critical [23].
Complacency Factors: The Independent Forensic Team found six industry-

level lessons learned from the Oroville Dam spillway incident in 2017 that can be
categorized as organization and industry complacency. Systemic failure involved
practices of the owner, its Federal and state regulators, its consultants, and general
industry practices in the USA. The lessons learned span physical inspections, com-
prehensive facility reviews, regulatory compliance, application of Potential Failure
Mode Analysis, consideration of appurtenant structures, and dam safety culture [7].
Organizational culture and internal working relationships are critical and can
do a lot towards eliminating mistakes if that culture exits [7]. Organizations that don’t
have some level of humility and maintain a level of vigilance regarding the risks
associated with infrastructures are likely to relive the Oroville Spillway incident in
some manner or another. Numerous human factors, including organizational and
industry factors, contributed to the physical factors not being recognized and prop-
erly addressed, and to the decision making during the incident [24]. A number of
factors ranging from lack of experience to overconfidence played a role, including a
reactive risk management approach probably driven by socioeconomic pressures.
In addition, the owner’s dam safety program was not sufficiently mature, lacked clear
executive-level accountability, and was excessively reliant on regulatory processes
to identify risks.
Crisis Management Factor: The factors that contributed to the incident may or
may not have come to light through engineering analysis, but by having an effective
integrated crisis management program in place can help to mitigate the engineering
129
Q.105 - R.9
Fig. 2
Dam Resiliency Framework
Structure du programme de résilience des barrages
risks to the public. The Federal Energy Regulatory Commission requires all licensed
Hydropower operators to maintain and update their Emergency Action Plans (EAP)
annually and provide exercises on a reoccurring basis to ensure the dam owner,
all stakeholders and impacted communities are familiar with what to do in case of
a dam emergency. The information that was available at the time and the decision
made by the Butte County Sheriff to call for immediate evacuation when evidence
showed that the Oroville Dam Emergency Spillway was at risk, demonstrates the
effectiveness that a well-organized and managed crisis management plan can have
in support of a dam resilience program. The evacuation of nearly 180,000 people
occurred without death or injury.
The independent forensic team concluded that the Oroville Dam spillway inci-
dent was the result of interactions of multiple factors without a unique root cause
130
Q.105 - R.9
and reflected a failure to recognize and properly address the deficiencies and warn-
ing signs which preceded the incident [7]. However, the holistic resilience approach
to managing the dam provided for contingencies and unknown risks that led to a
positive outcome of no loss of life.
REFERENCES
[1] U.S. DEPARTMENT OF HOMELAND SECURITY (2019). A Guide to Critical

Infrastructure Security and Resilience. Washington DC, November 2019.
[2] U.S. DEPARTMENT OF HOMELAND SECURITY (2010). Risk Lexicon.

Washington DC, September 2010.
[3] INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE (2014). Climate

Change 2014: Synthesis Report. Contribution of Working Groups I, II and
III to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Change, R. K. Pachauri and L. A. Meyer, Eds., IPCC, Geneva, Switzerland.
[4] QUINTÃO , A.F., BRITO, I. OLIVEIRA, F., MADUREIRA, A.P., and CON-
FALONIERI, U. (2017). Social, Environmental, and Health Vulnerability to
Climate Change: The Case of the Municipalities of Minas Gerais, Brazil.
Journal of Environmental and Public Health, Volume 2017, Article ID 2821343.
[5] McCORMICK, B., LUKAS, J., MAHONEY, K. (2020). 21ST Century Dam
Safety Rules for Extreme Precipitation in a Changing Climate. The Journal
of Dam Safety, Volume 17, Number 3.
[6] U.S. DEPARTMENT OF HOMELAND SECURITY and U.S. DEPARTMENT

OF ENERGY (2017). Dams and Energy Sectors Interdependency Study: An
Update to the 2011 Study. Washington DC, April 2017.
[7] INDEPENDENT FORENSIC TEAM REPORT (2018). Oroville Dam Spillway

Incident. January 2018.
[8] ENRICO, Z. (2016). Critical Infrastructures Vulnerability and Risk Analysis.

European Journal for Security Research, 2016, Volume 1.
[9] BELLUCK, D., HULL, R.N., BENJAMIN, S.L., ALCORN, J., and LINKOV, I.
(2007). Environmental Security, Critical Infrastructure and Risk Assessment:
Definitions and Current Trends. Environmental Security in Harbors and Coastal
Areas. Linkov, I., Kiker, G.A., and Wenning, R.J., Editors, Springer, January
2007.
[10] WIENER, E.L. (1981). Complacency: is the term useful for air safety? Proceed-
ings of the 26th Corporate Aviation Safety Seminar. Flight Safety Foundation.
March 1981, Denver, CO.
131
Q.105 - R.9
[11] REASON, J. (2000). Human error: Models and Management. Western Journal
of Medicine, June 2000, Volume 172, Number 6, pages 393–396.
[12] AWWA, “Risk Analysis and Management for Critical Asset Protection (RAM-
CAP®) Standards for Risk and Resilience Management of Water and Wastew-
ater Systems”, ANSI/AWWA J100-10(R13) First Edition, 2010.
[13] NATIONAL INFRASTRUCTURE ADVISORY COUNCIL (2010). A Framework

for Establishing Critical Infrastructure Resilience Goals: Finale Report and
Recommendations, October 2010, Washington DC.
[14] THE ECONOMIC TIMES. Definition of Risk Management. https://economic

times.indiatimes.com/definition/risk-management
[15] ICOLD (2005). Risk Assessment in Dam Safety Management: A Recon-

naissance of Benefits, Methods and Current Applications, International
Commission on Large Dams, Committee on Dam Safety.
[16] HARTFORD, D.N.D. and BAECHER. G.B. (2004). Risk and Uncertainty in Dam
Safety, Thomas Telford Publishing.
[17] U.S. BUREAU OF RECLAMATION and U.S. ARMY CORPS OF ENGINEERS

(2019). Best Practices in Dam and Levee Safety Risk Analysis, July 2019.
[18] U.S. DEPARTMENT OF HOMELAND SECURITY (2015). Dams Sector

Security Guidelines. Washington DC, November 2019.
[19] ASSOCIATION OF STATE DAM SAFETY OFFICIALS (2020). Reports of

Public Safety Incidents in the News. https://damsafety.org/PSAD-2020.
[20] COMMITTEE OF SPONSORING ORGANIZATIONS OF THE TREADWAY

COMMISSION (2004). Enterprise Risk Management — Integrated Frame-
work. September 2004.
[21] ISO 22301:2019 (2019). Security and Resilience — Business Continu-

ity Management Systems — Requirements. International Organization for
Standardization (ISO), Technical Committee 292 (Security and Resilience).
[22] FIMRITE, P. (2020). Feds reimburse California for Oroville Dam repairs,
but expensive new work possible. San Francisco Chronicle, https://www.
sfchronicle.com/news/article/Feds-reimburse-California-for-Oroville-Dam-150
75134.php.
[23] HENN, B., MUSSELMAN, K., LESTAK, L., RALPH, M., and MOLOTCH, N.
(2020). Extreme Runoff Generation from Atmospheric River Driven Snowmelt
During the 2017 Oroville Dam Spillways Incident, Geophysical Research
Letters, June 2020, Volume 47, Number 4.
[24] ALVI, I. (2019). Lessons Learned Case Study Oroville Dam (California,
2017), Association of State Dam Safety Officials (ASDSO), https://damfailures.
org/case-study/oroville-dam-california-2017.
132
Q.105 - R.10
-------
-------
U.S. ARMY CORPS OF ENGINEERS MODELING,

MAPPING, & CONSEQUENCE PRODUCTION CENTER
PROCESSES FOR DAM BREACH ANALYSIS (∗)
Wesley CROSBY, P.E.

Modeling Technical Coordinator,
U.S. Army Corps of Engineers Modeling,
Mapping, and Consequence Mandatory Center of Expertise
Kurt BUCHANAN
Consequence Technical Lead,
Alexandra UBBEN, PMP

Data Management Project Manager,
SUMMARY
Due to the large scale of the United States Army Corps of Engineers (USACE)
dam portfolio, USACE developed a process for the assessment of the risks associ-
ated with the portfolio of USACE dams throughout the United States in support of
the Dam Safety Program (DSP). In order to determine the risk associated with these
∗ USACE: Modélisation, cartographie et centre de gestion des conséquences – processus
d’analyse des brèches dans les barrages.
DOI: 10.1201/9781003211808-56 133

Q.105 - R.10
dams and appurtenant structures, USACE tasked its Modeling, Mapping, and Con-
sequence Mandatory Center of Expertise (MMC-MCX) to perform hydraulic analysis
of dam breaches, consequence analysis, and map production to aid in risk deter-
mination and communication. The MMC developed guidance and procedures for
performing these tasks. The MMC produces dam breach hydraulic models using the
one-dimensional (1D) and the two-dimensional (2D) version of the Hydrologic Engi-
neering Center-River Analysis System (HEC-RAS) model and applies these breach
models over a range of scenarios at multiple locations. The 1D/2D hydraulic analysis
results are used by the MMC’s consequence team to determine the range of conse-
quences using the Hydrologic Engineering Center-LifeSim (HEC-LifeSim) software
package. After modeling, the consequence team uses this software to determine
impacts such as economic damage, population at risk (PAR), and life loss (LL).
Finally, the MMC mapping team uses the results from the hydraulics and conse-
quence analysis to produce maps and documents that aid in the communication of
risk for areas downstream and upstream of USACE dams. These products are used
by others within USACE to inform emergency planning activities, risk assessments,
and risk communication activities.
RÉSUMÉ
En raison de l’ampleur du parc de barrages du Corps des ingénieurs de l’armée

américaine (USACE), l’USACE a mis au point un processus d’évaluation des risques
associés à son parc dans l’ensemble des États-Unis, en appui du programme de
sécurité des barrages (DSP). Afin de déterminer le risque associé à ces barrages
et aux structures annexes, l’USACE a chargé son centre d’expertise sur la modéli-
sation, la cartographie et les conséquences (MMC-MCX) d’effectuer une analyse
hydraulique des brèches dans les barrages, une analyse des conséquences et la
production de cartes pour faciliter la détermination et la communication des risques.
Le MMC a élaboré des guides et des procédures pour l’exécution de ces tâches. Le
MMC produit des modèles hydrauliques de brèches de barrage en utilisant la version
unidimensionnelle (1D) et la version bidimensionnelle (2D) du modèle HEC-RAS
(Hydrologic Engineering Center-River Analysis System) et applique ces modèles
de brèches sur une série de scénarios à des endroits multiples. Les résultats de
l’analyse hydraulique 1D/2D sont utilisés par l’équipe des conséquences du MMC
pour déterminer la gamme des conséquences à l’aide du progiciel Hydrologic Engi-
neering Center-LifeSim (HEC-LifeSim). Après la modélisation, ce logiciel est utilisé
pour déterminer les impacts tels que les dommages économiques, la population à
risque (PAR) et les pertes de vie (LL). Enfin, l’équipe de cartographie du MMC utilise
les résultats de l’analyse hydraulique et des conséquences pour produire des cartes
et des documents qui aident à la communication des risques pour les zones en aval
et en amont des barrages de l’USACE. Ces produits sont utilisés par d’autres mem-
bres de l’USACE pour informer les activités de planification d’urgence, d’évaluation
et de communication des risques.
134
Q.105 - R.10
1. INTRODUCTION
The United States Army Corps of Engineers (USACE) dam portfolio comprises
738 structures located at 558 projects. The 738 structures include appurtenant struc-
tures such as saddle dikes, in-pool levees, etc. located at a dam project. USACE
owns and operates six of the ten largest reservoirs in the United States, six of the ten
largest embankment dams in the United States, and owns & operates fifty percent
of all federally owned dams. USACE dams prevented approximately $485 billion in
damages between 2004 and 2013. USACE flood damage reduction projects avoid
approximately $8.00 in damages for every $1.00 invested [1]. As a result of the impor-
tance of the projects, USACE developed a process for the assessment of the risks
associated with the portfolio of dams and appurtenant structures that USACE owns
and operates throughout the United States in support of the Dam Safety Program.
This process includes a periodic risk assessment (PA), as well as, more detailed
higher-level risk assessments such as Semi-Quantitative Risk Assessment (SQRA)
and Issue Evaluation Studies (IES).
The Periodic Assessment (PA) for a dam consists of a site visit, typically in
conjunction with a periodic inspection, a potential failure modes analysis, and a
reisk assessment based on existing data and estimated consequences. The Mod-
eling, Mapping, and Consequence (MMC) Production Center will typically produce
information to assist with estimating consequences [2]. The information provided
by the MMC includes hydraulic modeling, consequence analysis, and produc-
tion of emergency action plan (EAP) inundation map products. In addition, to
PA’s, the MMC will occasionally assist with higher level risk assessments when
requested.
This paper describes the USACE Modeling, Mapping and Consequences

(MMC) Production Center approach to hydraulic modeling, consequence modeling,
and mapping in support of periodic assessment analysis for the Dam Safety Pro-
gram. Processes and procedures discussed in this paper are based on procedures
documented in the MMC Technical Manual for Dams [3].
2. HYDRAULIC ANALYSIS
The MMC hydraulic modeling team is tasked with creating geo-referenced

unsteady hydraulic models capable of producing accurate flood inundation infor-
mation to support mapping and consequences analysis across a wide range of
scenarios. The current version of the Hydraulic Engineering Center-River Analysis
System (HEC-RAS) software is applied in all situations. The software is endorsed
by the Hydrology, Hydraulics, and Coastal Community of Practice (HH&C CoP)
within USACE and is maintained by the Hydrologic Engineering Center (HEC).
HEC-RAS has the capability to model areas using a one-dimensional (1D) modeling
135
Q.105 - R.10
approach, a two-dimensional (2D) approach, or it can use a combination of both 1D

and 2D.
The following sections provide general information, guidance, and procedures

for completing an MMC dam breach analysis.
2.1. BASIC MODEL INPUTS
High quality terrain data is critical to building accurate hydraulic models. Flood-
plain features such as roadway embankments, underpasses, or large culverts can
have a significant impact on water surface elevations and extents of inundation. Map-
ping team members make an effort to seek out the best available digital elevation
model (DEM), but the assigned hydraulic engineer has the responsibility to verify the
quality and seek out better data as necessary. In some cases, the DEM may have to
be modified to include bathymetry data (surveyed or estimated), levee data, bridge
openings, and/or other topographic features not sufficiently or accurately defined in
the terrain. Regardless of the source or resolution of the DEM, it is critical that the
DEM represents bare-earth elevations. For an MMC dam breach analysis, the DEM
should be at least 10-meter resolution or better.
Information on significant structural features within the watershed can be found

from multiple sources. Dam dimensions and features are typically based off of infor-
mation from the water control manual (WCM) for the study dam. For watersheds that
include levees along the study reach, levee profiles that are incorporated into the
model are primarily based on the National Levee Database (NLD) data. NLD levee
crest data for USACE levees typically comes from ground survey, but can come from
other sources such as aerial LiDAR or truck mounted LiDAR. The NLD crest ele-
vations can potentially include incorrect points that represent something other than
the top of levee. Particular attention is given to isolated low spots in the levee that
could be incorrect and inadvertently allow flow to overtop the levee. The assigned
hydraulic engineer is responsible for verifying the accuracy and consistency of the
profile compared to any other data that may be available. In some cases, the local
USACE district office may have more current surveys that have not yet been incor-
porated into the NLD. Background information sources such as inspection reports,
levee screening data, or as-built drawings have provided valuable information for the
modeling process.
An inflow hydrograph is needed for the modeling and development of model

scenarios. The inflow hydrograph is based on the inflow design flood (IDF) of the
study dam. Depending on the dam, the IDF can be a probable maximum flood (PMF),
spillway design flood (SDF), standard project flood (SPF), etc. For USACE flood risk
management dams, the IDF is typically a PMF and most navigation dams are SDF
or SPF but some may be designed to a PMF. USACE Engineering Regulation ER-
1110-8-2(FR) [4] provides guidance on determining the IDF for Dams and Reservoirs
within the USACE portfolio.
136
Q.105 - R.10
2.2. LOADING LEVELS/MODEL SCENARIOS
A primary objective of MMC Production Center modeling is to provide consis-

tent data that supports the estimation of consequences for breach and non-breach
scenarios over a range of reservoir pool loadings. This range of conditions supports
requirements of the Critical Infrastructure Protection and Resilience (CIPR) and
Dam Safety Programs. The incremental consequences between breach and non-
breach conditions for a given scenario are of particular importance in a dam safety
consequence project. The MMC program considers alternative scenarios for flood
risk management, dry flood risk management, and navigation dams. To more thor-
oughly assess consequences, scenarios will be run for both breach and non-breach
conditions [3]. The MMC Production Center’s model scenarios are presented below.
2.2.1. Flood risk management dam loading levels
The USACE has established a standard set of five modeling scenarios cov-
ering a range of pool elevations, representing minimum and maximum normal
operating conditions as well as select extreme loading conditions designed to
capture the upper limit of potential downstream consequences for flood risk man-
agement dams (Figure 1). The resulting load conditions included in the HEC-RAS
project are [3]:
• Maximum High (MH) Pool: Corresponds to the maximum inflow design flood
(IDF) elevation for the dam. This scenario is associated with the occurrence of
Fig. 1
Flood Risk Management Dam Pool Loadings
Gestion des risques d’inondation Charge des bassins de retenue des barrages.
137
Q.105 - R.10
an extreme flood event and routed using the emergency operation procedures
from the WCM. It includes full utilization of available surcharge, flood, con-
servation, inactive, and dead storage. The most current inflow design flood
hydrograph, antecedent pool, and routing will be used from the most recent
study. If an antecedent pool is not provided, the pool should be initially set to
the top of active storage (TAS) pool and then adjusted until the provided max
pool is met.
• Intermediate High (IH) Pool: Corresponds to an elevation that would occur in
an unusual event, specifically an elevation that corresponds to a maximum
discharge approximately halfway between the maximum TAS discharge and
the maximum MH pool discharge. It does include some surcharge storage
and will likely result in a significant increase in discharge above downstream
channel capacity. Starting reservoir pool will correspond to the antecedent
pool used for the IDF. Reservoir inflow will be a scaled IDF hydrograph. The
scaled hydrograph will be developed by trial and error. Gate operation proce-
dures from the latest water control manual will be applied to the hydrograph
routing.
• Top of Active Storage (TAS) Pool: Corresponds to the highest elevation which
can be obtained under normal regulated operating conditions for authorized
purposes (i.e., without damaging releases). This scenario is associated with
the occurrence of an unusual flood event. It does not include surcharge stor-
age but should represent the pool elevation that, if exceeded, would result in
significant increase in discharge based on the physical properties of the project
and/or operation procedures from the latest WCM. The term active storage has
similar meaning as the terms full pool and flood control pool. For typical flood
risk management dams with uncontrolled spillways, the TAS corresponds to
the emergency spillway crest elevation. For dams with controlled or gated spill-
ways, the TAS commonly corresponds to an elevation that, if exceeded, the
gates would begin to fully open. In some cases this may be at or near the top
of the spillway gates. Starting reservoir pool will correspond to the 10-percent
duration exceedance pool (DEP) elevation. Reservoir inflow will be a scaled
inflow design flood hydrograph. The scaled hydrograph will be developed by
scaling the available hydrograph using trial and error to produce a peak pool
equal to the TAS elevation.
• Security Scenario (SS) Pool: Corresponds to the one-percent duration
exceedance pool (DEP) elevation (equaled or exceeded about one percent
of the time based on the period of record for the reservoir pool) which can
be obtained under normal regulated operating conditions. Depending on the
dam, this scenario may or may not be associated with a flood event. Starting
reservoir pool will correspond to the 10-percent exceedance duration pool ele-
vation. Reservoir inflow will be a scaled inflow design flood hydrograph. The
scaled hydrograph will be developed by scaling the available hydrograph using
trial and error to produce a peak pool equal to the one-percent exceedance
duration pool elevation.
• Normal High (NH) Pool: Corresponds to the ten percent duration exceedance
pool (DEP) elevation (equaled or exceeded about 10 percent of the time based
on the period of record for the reservoir pool) which can be obtained under
138
Q.105 - R.10
normal regulated operating conditions. This scenario represents a relatively-

high, but normal, pool condition that can be expected to occur every year.
This scenario is not associated with the occurrence of a flood event. Start-
ing reservoir pool will correspond to the 10-percent exceedance duration pool
elevation. Reservoir inflow will be a constant inflow hydrograph as required
to maintain the 10-percent exceedance duration pool under normal gate
operations.
2.2.2. Dry dam loading levels
Scenarios for dry flood risk management dams are to be the same as modeled
for normal flood risk management dams. If the reservoir is dry for a particular event,
this will be noted in USACE documentation. In cases where the reservoir remains
dry in all scenarios except TAS, IH, and MH pools, a special scenario may have to
be developed based on judgment of the project lead to capture an event less than
TAS. This additional scenario will be a pool elevation that is halfway between the
SS and TAS pool elevations and will be named Intermediate Low–Dry–Pool (ILD).
Starting reservoir pool will correspond to the 1-percent exceedance duration pool
elevation. Reservoir inflow will be a scaled inflow design flood hydrograph. The
scaled hydrograph will be developed by scaling the available hydrograph using trial
and error to produce a peak pool equal to the ILD pool [3].
2.2.3. Navigation dam loading levels
The USACE has established a standard set of five modeling scenarios cov-
ering a range of pool elevations, representing minimum and maximum normal
operating conditions as well as select extreme loading conditions designed to cap-
ture the upper limit of potential downstream consequences for navigation dams
(Figure 2). Originally, the MMC used the same scenarios for navigation projects as
those used for flood risk management projects. However, due to the operation of
navigation projects which try to maintain a specified pool, several of the scenarios
used for the flood risk management projects were at or near the same pool ele-
vation. As a result, the MMC worked to produce different scenarios for navigation
projects. Since the pool loadings do not vary significantly as compared to flood
risk management dams, the navigation scenarios focused on pool loadings and
flow discharges from the structure which can impact the incremental consequences
downstream. The resulting load conditions included in the HEC-RAS project
are [3]:
• Maximum High (MH) Pool: Similar to the flood risk management dam, this
scenario corresponds to the maximum inflow design flood (IDF) elevation for
the dam. This scenario is associated with the occurrence of an extreme flood
139
Q.105 - R.10
Fig. 2
Navigation Dam Pool Loadings
Charges des bassins de barrage de navigation.
event and routed using the emergency operation procedures from the WCM. It
includes full utilization of available surcharge, flood, conservation, inactive, and
dead storage. The most current inflow design flood hydrograph, antecedent
pool, and routing will be used from the most recent study. If an antecedent
pool is not provided, the pool should be initially set to the top of active storage
(TAS) pool and then adjusted until the provided max pool is met.
• Intermediate High (IH) Pool: Similar to the flood risk management dam, this
scenario corresponds to an elevation that would occur in an unusual event,
specifically an elevation that corresponds to a maximum discharge approxi-
mately halfway between the maximum TAS discharge and the maximum MH
pool discharge. It does include some surcharge storage and will likely result
in a significant increase in discharge above downstream channel capacity.
Starting reservoir pool will correspond to the antecedent pool used for the IDF.
Reservoir inflow will be a scaled IDF hydrograph. The scaled hydrograph will
be developed by trial and error. Gate operation procedures from the latest
water control manual will be applied to the hydrograph routing.
• Navigation Pool Channel Capacity (NPC) Pool: Corresponds to the highest
elevation which can be obtained under normal regulated operating conditions
that allows releases equal to the downstream channel capacity. This scenario
may or may not be associated with the occurrence of a flood event. Starting
reservoir pool will correspond to the normal navigation (NN) pool elevation.
Reservoir inflow will be a scaled IDF hydrograph. The scaled hydrograph will
be developed to produce a peak pool equal to the NPC loading.
• Intermediate Low (IL) Pool: Corresponds to an elevation that produces down-
stream peak discharges approximately halfway between the maximum NN
peak discharge and the maximum NPC peak discharge. Starting reservoir
pool will correspond to the NN pool elevation. Reservoir inflow will be a scaled
IDF hydrograph. The scaled hydrograph will be developed by trial and error.
140
Q.105 - R.10
• Normal Navigation (NN) Pool: Corresponds to a normal pool condition. This

scenario is not associated with the occurrence of a flood event. Starting reser-
voir pool will correspond to the NN pool elevation. Reservoir inflow will be a
constant inflow hydrograph as required to produce the NN pool under normal
gate operations.
2.2.4. Cascading breaches
Downstream dams should be operated according to the WCM for the dam. If
modeled scenarios on the dam being assessed create conditions that exceed the
storage or discharge capacity of the downstream dam(s) and overtopping occurs,
the downstream dam(s) should be considered for breach. Should conditions war-
rant consideration for cascading breaches for a given loading condition an additional
model simulation and associated consequences will be developed. The MH, and top
of active storage (TAS) scenarios will be the only scenarios that will be considered
for a cascading breach scenario. IH, SS, and NH are not considered for cascading
breach. Breach of the downstream dam will occur at the peak overtopping elevation
unless the overtopping exceeds 1.5 meters [5 feet], in which case the breach will be
triggered at a maximum of 1.5 meters [5 feet] of overtopping. The non-cascading
breach response will be considered as the official scenario for a given loading con-
dition, but modeling assumptions and consequences for the cascading scenario will
be documented as additional information in the model report. Cascading breach is
not required for small water supply or run of river navigation dams that would not
increase consequences if breached [3].
2.3. BREACH DETERMINATION
For periodic assessment (PA) level studies, breach parameters are estimated
using simplified methods for earthen embankments, concrete gravity dams, and
concrete arch dams. For higher level risk assessments, better estimates may be
determined using physically based computer models or expert elicitation. Below
describe the methods used for PA level studies to estimate breach parameters.
For earthen embankments, simple regression equations are used to deter-

mine breach parameters. The MMC calculates four sets of breach parameters
using the following four regression equations: MacDonald and Langridge-Monopolis,
Froehlich (1995a), Froehlich (2008), and Von Thun & Gillette. The results are
compared for each loading and engineering judgment is used to determine which
regression equation produces the most reasonable and realistic breach parameters
for the five loadings. The MMC has developed a spreadsheet that calculates the
parameters using all four regression equations and provides additional information
that may help in the decision process. The additional information includes compar-
ing the computed dimensions against the dimensions of the dam to determine if the
141
Q.105 - R.10
breach for each loading will fit within the dam footprint. In addition, the spreadsheet
plots the computed dimensions for each equation and each loading against the
case study data that was used to produce the regression equations. If two regres-
sion equations produce similar values, the hydraulic engineer may choose to run
a sensitivity analysis in the hydraulic model to determine which equation produces
the most conservative, yet realistic flow.
For concrete dams, there does not appear to be much research on breach
dimensions. USACE has determined that a reasonable estimate for concrete gravity
dams is for the breach of 2 to 6 monoliths. For pool loadings with significant volume,
the number of monoliths assumed to breach will be at or near 6 because it is assumed
that the time to drawdown the storage would be enough to erode the foundations of
other monoliths, causing multiple monoliths to breach. Likewise, for pool loadings
with minimal storage, the number of monoliths assumed to breach would be less. For
concrete arch dams such as those built using roller compacted concrete, the entire
dam is assumed to breach since there is no vertical separation of the monoliths and
if a breach were to occur it would most likely erode the entire layer.
3. CONSEQUENCE ANALYSIS
The MMC uses Hydrologic Engineering Center-LifeSim (HEC-LifeSim) soft-

ware to evaluate the potential life loss and direct economic damages due to dam
breach events. HEC-LifeSim is a spatially-distributed dynamic simulation modeling
system with a computation engine that accounts explicitly for the impact of warning
issuance time, warning diffusion, the population at risk’s (PAR’s) protective action
initiation, the PAR’s evacuation potential via vehicles, detailed flood dynamics, and
loss of shelter. Monte Carlo sampling techniques are used in HEC-LifeSim to cap-
ture the natural variability of temporal uncertainty in the warning and evacuation
process as well as the potential for life loss when exposed to threatening flood
situations. The results from the HEC-LifeSim model include damage estimates to
structures and contents, potential life loss within structures and vehicles on roads,
statistical ranges around the results, animations of selected flood and evacuation
simulations, and the capability to display results based on parameter uncertainty
sampling.
The life loss methodology in HEC-LifeSim is based on methodologies devel-

oped by Utah State University’s Institute for Dam Safety Risk Management. The
software performs the following steps to estimate the loss of life for a selected
event-exposure scenario and given structure inventory with population:
• Estimate the dam-breach flood wave arrival times for each structure (i.e., the
time at which the depth of flooding at the location of the structure is estimated
to be large enough that the inhabitants will choose to remain in the structure
and evacuate vertically instead of risk leaving the structure). HEC-LifeSim
142
Q.105 - R.10
estimates arrival times for each structure by interpolation from the hydrograph
provided at the nearest upstream and downstream location to each structure
or by selecting it from the arrival time grid in the specific cell where the structure
is located.
• Estimate the portion of the PAR for each structure that has received a warning
and mobilized using the warning and mobilization relationships developed by
Sorensen and Mileti [5].
• Route groups of PAR from their structure to their likely destination points using
the road network and traffic routing algorithms. Determine if and when vehicles
are caught on flooded road segments while they would be traveling on them.
• Assign a lethality zone based on the mobilization and evacuation outcome for
each structure: none, safe, compromised, or chance. PAR that did not mobilize
is assigned a lethality zone based on the structure stability criteria or maximum
inundation depth at the structure considering the foundation height. Structure
stability criteria is based on the construction type and the maximum depth
multiplied by velocity at the structure location. Whichever method results in
a higher lethality zone governs the result. PAR caught while evacuating in a
vehicle is assigned a fatality zone based on vehicle stability criteria.
• Estimate the life loss at each structure by sampling from a distribution of
probable fatality rates for each lethality zone on each iteration.
• Estimate the total life loss as the sum of the life loss estimates for all structures
and vehicles. Calculate results statistics for all iterations.
3.1. DATA INPUTS
HEC-LifeSim modeling involves several data inputs from multiple sources.

These include hydraulic modeling outputs, a structure inventory with building and
population information, emergency planning zone polygons, a road network, and
evacuation destination points. The required inputs are described further in the
following sections.
3.1.1. Hydraulic modeling data
HEC-LifeSim has the capability to read hydraulic data directly from HEC-RAS
data files. For each event plan simulated in HEC-RAS, a data file is generated with all
of the hydraulic information for each 1D feature and 2D cell within the model. HEC-
LifeSim uses the same file types and can copy the necessary hydraulic information
from the HEC-RAS file directory into its own file structure. This allows HEC-LifeSim
to have the necessary hydraulic information it needs to calculate consequences
including water arrival times, depths, and velocities at each location. The plan name
and terrain data used for the HEC-RAS analysis are also imported into the HEC-
LifeSim model during the data import.
143
Q.105 - R.10
3.1.2. Structure inventory
The structure inventory used by the MMC in HEC-LifeSim is developed from the
USACE National Structure Inventory (NSI). The NSI is a data service with a base
dataset containing estimated information regarding the locations, building types,
population, values, and other relevant information for all residential, commercial,
industrial, agricultural, public and private structures across the United States. The
base dataset utilizes parcel data, Census block data, Microsoft open source building
footprints, and several other data sources that serve as a basis for estimating flood
hazard consequences. Population and structure values are estimated from a variety
of sources, but are intended to reflect 2017 population levels and 2018 price levels.
Structure point locations and attribute values are manually edited based on
aerial imagery to further improve accuracy.
3.1.3. Emergency planning zones
Life loss parameters and warning times are set according to emergency plan-
ning zone (EPZ) polygons. When estimating life loss for a dam breach at load levels
where downstream releases exceed control flows due to high releases or spill-
way flows, EPZs are delineated for each scenario to allow downstream population
impacted by flooding prior to the breach to receive a warning and have some oppor-
tunity to evacuate in advance of the flooding. To accomplish this, the non-breach
inundation polygon is unioned with a study area polygon and a pool polygon. This
creates 3 distinct EPZs for each scenario: the area upstream of the dam, the area
downstream of the dam that would be flooded in a non-breach scenario, and the
area downstream of the dam that would only be flooded in the case of a breach.
3.1.4. Road network and evacuation destinations
For the majority of dam breach consequence estimates, evacuation via road
networks is considered beyond the scope of the MMC work due to the time required
to calibrate the network and destination points over long distance inundation areas
(as opposed to a levee which is typically a single town or city).
3.2. HEC-LIFESIM PARAMETERS
Once all of the required data has been imported into the HEC-LifeSim model
there are multiple model parameters that must be set up in order to calculate
consequences. To determine whether or not the PAR within a structure is able to
receive warning and mobilize, several parameters are used within HEC-LifeSim
144
Q.105 - R.10
Fig. 3
HEC-LifeSim Warning and Evacuation Timeline
Calendrier d’alerte et d’évacuation de HEC-LifeSim.
to estimate the probability of warning and mobilization at each time step. These
parameters and actions include when warnings will be issued (hazard identification
time, hazard communication delay, and warning issuance delay), how long they will
take to become effective (warning diffusion), and the rate at which PAR will mobi-
lize in response (mobilization or protective action initiation). Fig. 3 represents an
example breach warning and mobilization timeline with the matching HEC-LifeSim
parameters.
3.2.1. Hazard identification relative time
The hazard identification time is the time at which a hazard such as breaching
or flooding is identified (relative to when that hazard was initiated). The MMC uses
two different warning scenarios for breach areas with different ranges of hazard
identification time: minimal warning and ample warning. Minimal warning scenarios
have the hazard identification relative time set as a uniform distribution between 2
hours prior to breach and at breach (-2 to 0 hours). Ample warning scenarios have
the hazard identification relative time set as a uniform distribution between 6 hours
prior to breach and 2 hours prior to breach. For in-pool and non-breach flooding
areas, the warning is set at least 72 hours prior to the breach to allow for higher
mobilization potential.
3.2.2. Hazard communication delay
The hazard communication delay is the time elapsed from when the hazard
is identified to when the EPZ representatives would be notified. For example, if a
breach occurs when no one is observing the project, the emergency managers could
145
Q.105 - R.10
be notified an hour after the hazard is identified. The hazard communication delay
is set as a uniform uncertainty distribution between 0.01 hours and 0.5 hours.
3.2.3. Warning issuance delay
The warning issuance delay is the time elapsed from when the emergency
managers receive the notification of the imminent hazard to when they issue the
first evacuation order to the public. The warning issuance delay is set at the preset
configuration of ‘Preparedness Unknown’, which uses a Lindell uncertainty distribu-
tion. The delay is randomly sampled from 0 to 6 hours, but is positively skewed such
that results from 0 to 1.5 hours are more likely.
3.2.4. Warning diffusion/first alert
The first alert parameter defines the warning diffusion curve for daytime and
nighttime. The diffusion curve represents the percentage of the population which will
receive a first alert warning over time during daytime hours from when the warning
was issued. The first alert curves are set at the preset configuration of ‘Unknown’
which samples from a uniform uncertainty distribution where the upper bound curve
warns 100 percent of the PAR after 1.5 hours and the lower bound curve warns
100 percent of the PAR after 6 hours.
3.2.5. Mobilization/protective action initiation
The Protective Action Initiation (PAI) parameter defines the mobilization curve.
The protective action initiation curve represents the percentage of the population
which will take protective action over time from when the first alert is received. Within
the protected area, the PAI/mobilization curve for is set at the preset configuration
of ‘Preparedness: Unknown/Perception: Unknown’ which samples a uniform uncer-
tainty distribution with maximum mobilization rates between 83 and 100 percent
after 72 hours.
3.2.6. Other parameters
The parameters listed above represent those parameters most often asso-
ciated with estimating life loss, but there are several other parameters within
HEC-LifeSim that could be adjusted. These include the depth-damage curves used
to calculate damages based on flood depth and the structure stability criteria curves
which determine when maximum depths and velocities are likely to compromise the
146
Q.105 - R.10
integrity of a structure. In most MMC studies, these parameters are left at the default
values within HEC-LifeSim.
3.3. HEC-LIFESIM RESULTS
HEC-LifeSim results include estimates of the number of structures flooded, the

population characteristics within the flooded area; damages to structures, contents,
and vehicles; and estimated life loss. Each scenario is run with 1000 iterations using
Monte-Carlo methods to capture the effects of uncertainty. This allows results to be
presented in the form of five number statistical summaries which can be plotted on
a log-scale whisker chart to determine the potential order-of-magnitude of life loss.
In the example in Fig. 4, three scenarios, the Maximum High Pool, the Interim
High, and the Top of Active Storage, have a high likelihood of incremental life loss
between 100 and 1,000. The Security Scenario Pool and the Normal High Pool
scenarios have a high likelihood of life loss between 10 and 100. Results can also
be broken out by shapefiles such as an inundation zone (breach or non-breach flood
areas) as well as specific city or community as seen in Figure 5.
Fig. 4
Life Safety Risk Order of Magnitude
Ordre de grandeur des risques pour la sécurité des personnes.
147
Q.105 - R.10
Fig. 5
Downstream Consequences by Community
Conséquences en aval par Communauté.
4. MAPPING PRODUCTS
4.1. BACKGROUND
The MMC developed a standard for mapping products for dam breach studies.
This standard was developed using feedback from stakeholders and following the
federal guidelines for Emergency Action Planning for Dams (FEMA Publication No.
64). The MMC map standard is reviewed and potentially revised every five years to
account for improvements in modeling capabilities and the needs of stakeholders.
The dam breach mapping standard uses depth grids and timing data generated from
the modeling process.
In developing the mapping standard for dams, the original standard was to
show maximum inundation extents for two breach scenarios within one map. This
was used for many years as the standard for dam breach mapping. The low scenario
was shown as being on top of the maximum scenario, so that users could visualize
flooding occurring between the two breach scenarios shown. These maps were
displaying maximum flood extent vector data, no information about the depth of
water was shown.
148
Q.105 - R.10
As improvements in modeling were made and depth grids were available

as a standard product, the use of depth grids as the inundation map standard
was presented to stakeholders in 2019. Depth grids were determined to be use-
ful information as many emergency managers would value the depth information
for planning for a flood event. Additionally, the change to depth grid made the dam
breach map standard more consistent with levee breach map standards developed
by the MMC. The significant difference from the previous standard is the quantity of
map books created as part of the standard; each map book now contains one depth
grid scenario.
4.2. MAPPING STANDARDS
For the MMC standard for dam breach mapping, a map atlas containing
depth grids is created for two breach scenarios and two non-breach scenarios (see
Table 1).
Table 1
Mapping Naming Convention
Dam Loading Scenario Mapping Naming Convention
Maximum High Breach MH-F
Normal High Breach NH-F
Maximum High Non-Breach MH-NF
Intermediate High Non-Breach IH-NF
Having PDF map products for the scenarios shown in table 2 meet the require-
ments for inundation map conditions for emergency action plan maps. In the case
of an actual event, the stakeholders can use the map atlas that is most like the
modeled scenario.
Fig. 6
Color Scheme for Depth Maps
Système de couleurs pour les cartes de profondeur.
149
Q.105 - R.10
Fig. 7
Example of Depth Map. Draft data for example obtained from USACE HEC-RAS
5.0 example data. This data does not represent an actual event.
Exemple de carte de profondeur. Données provisoires obtenues par exemple à
partir de l’exemple de données USACE HEC-RAS 5.0. Ces données ne
représentent pas un événement reel.
The grids are displayed as a standard set of blue coloring with defined depth
ranges of 0.03–0.6 m [0.1–2 ft], 0.6–1.8 m [2–6 ft], 1.8–4.6 m [6–15 ft], and greater
than 4.6 meters [15 ft]. The depth classification was determined based on how
historical fatality rates differ with different flood depths. (see Fig. 6 and Fig. 7).
For dam breach maps, the standard includes flood wave arrival time points cal-
culated for significant locations within the flood inundation. The breach wave arrival
times are calculated as the time at which the difference between the hydrograph
from the breach scenario and the hydrograph from the non-breach scenario reaches
0.6 meters [2 feet]. This data point captures the time for water from the breach to
increase by 0.6 meters [2 feet] at a given location. This calculation was determined
as the best way to determine increasing flooding from a breach; which in many cases
would be in addition to flooding from non-breach operational releases of the dam.
Figure 8 graphically shows the breach wave arrival time calculation occurring as a
change in 0.6 meters [2 feet] from the non-breach water surface elevation.
150
Q.105 - R.10
Fig. 8
Calculation for Breach Waves
Calcul pour les ondes de rupture.
REFERENCES
[1] U.S. Army Corps of Engineers 2020. “Dam Safety Program” Accessed July
2020. http://www.usace.army.mil/Missions/Civil-Works/Dam-Safety-Program/
[2] U.S. Army Corps of Engineers. Engineer Regulation 1110-2-1156: Safety of

Dams – Policy and Procedures, 2014.
[3] U.S. Army Corps of Engineers. MMC Technical Manual for Dams, 2020.
[4] U.S. Army Corps of Engineers. Engineer Regulation 1110-8-2(FR): Inflow

Design Floods for Dams and Reservoirs, 1991.
[5] Mileti D.S., Sorensen J.H. 2014. Protective Action Initiation Time Estima-
tion for Dam Breaches, Controlled Dam Releases, and Levee Breaches or
Overtopping.
151
Q.105 - R.11
-------
-------
A REVIEW OF RECENT DAM FAILURES IN THE U.S. AND

THE ANTICIPATED IMPACT OF IMPLEMENTATION OF
RISK-INFORMED DECISION MAKING (∗)
Dean B. DURKEE, Ph.D., P.E.

Senior Consultant and Vice President, GANNETT FLEMING, INC.
Douglas D. BOYER, P.E., CEG

Chief, Risk-Informed Decision-Making Branch, Federal Energy Regulatory
Commission
SUMMARY
Hydropower owners in the United States have invested considerable resources

in performing Potential Failure Mode Analyses (PFMA) for their dams, to meet Fed-
eral Energy Regulatory Commission (FERC) requirements as an element of dam
safety assessments. While PFMAs provide a much-improved understanding of the
safety of a dam, PFMAs stop short of providing a portrayal of the dam safety risk.
The PFMA process does not typically provide adequate differentiation of the
likelihood or consequences of failure to allow for distinction or prioritization of indi-
vidual potential failure modes for dam safety actions or decisions. As a result, FERC
has released Draft Risk-Informed Decision-Making Guidelines for use by licensees
to supplement their dam safety programs. Implementation of the Draft Guidelines
has been slowed by the voluntary nature of the guidelines, the uncertainty of ben-
efits and costs on the part of licensees, and lack of experience with risk analysis
among consultants.
∗ Un examen des défaillances de barrages aux États-Unis et de l’impact prévu de la mise
en œuvre d’un processus décisionnel fondé sur la connaissance du risque
152 DOI: 10.1201/9781003211808-57

Q.105 - R.11
This paper presents case histories from two recent failures and discusses how
and whether a risk informed approach earlier in the life of the case history projects
presented would have affected the outcome. The two case history examples pre-
sented herein suggest that if performed properly with appropriate dam professionals
for the given structure a risk analysis approach should reduce the likelihood and
consequences of failure.
RÉSUMÉ
Les propriétaires d’hydroélectricité aux États-Unis ont investi des ressources

considérables dans la réalisation d’analyses des modes de défaillance potentiels
(PFMA) pour leurs barrages, afin de répondre aux exigences de la Commission
fédérale de régulation de l’énergie (FERC) dans le cadre des évaluations de la
sécurité des barrages. Alors que les PFMA permettent une meilleure compréhension
de la sécurité d’un barrage, elles ne fournissent pas une représentation du risque
de sécurité des barrages.
Le processus des PFMA ne permet généralement pas de différencier de

manière adéquate la probabilité ou les conséquences d’une défaillance pour perme-
ttre de distinguer ou de hiérarchiser les différents modes de défaillance potentiels
pour les actions ou décisions relatives à la sécurité des barrages. En conséquence,
la FERC a publié un projet de lignes directrices sur la prise de décision en fonction
du risque à l’usage des titulaires de licence pour compléter leurs programmes de
sécurité des barrages. La mise en œuvre du projet de lignes directrices a été ralen-
tie par leur nature volontaire, l’incertitude des avantages, des coûts de la part des
titulaires de licence et le manque d’expérience des consultants en matière d’analyse
des risques.
Ce rapport présente des cas d’étude de deux défaillances récentes et exam-

ine comment une approche fondée sur les risques, adoptée plus tôt dans la
vie des projets d’étude des cas présentés, aurait affecté le résultat. Les deux
exemples présentés ici suggèrent que, si elle est réalisée correctement avec des
professionnels des barrages compétents pour la structure donnée, une approche
d’analyse des risques devrait réduire la probabilité et les conséquences d’une
défaillance.
1. INTRODUCTION
Hydropower owners in the United States have invested considerable resources

in performing Potential Failure Mode Analyses (PFMA) for their dams, to meet Fed-
eral Energy Regulatory Commission (FERC) requirements [1] as an element of dam
153
Q.105 - R.11
safety assessments. While PFMAs provide a much-improved understanding of the

safety of a dam, PFMAs stop short of providing a portrayal of the dam safety risk.
Risk, by definition, includes both likelihood and consequences of failure. The

PFMA process does not typically provide adequate differentiation of the likelihood or
consequences of failure to allow for distinction or prioritization of individual potential
failure modes for dam safety actions or decisions. As a result, FERC has released
Draft Risk-Informed Decision Making (RIDM) Guidelines [2] for use by licensees to
supplement their dam safety programs. Implementation of the Draft RIDM Guidelines
has been slowed by the voluntary nature of the guidelines, the uncertainty of benefits
and costs on the part of licensees, and lack of experience with risk analysis among
consultants.
This paper presents case histories from two recent failures and how the prac-
tice and industry is changing in response. While the dam industry focuses on moving
from a purely standards-based approach to a risk informed approach, it would seem
appropriate to ask whether a risk informed approach earlier in the life of the case
history projects presented would have affected the outcome. There are several
advantages to the risk informed approach, but the primary assumption is that it
will reduce the likelihood, frequency, and consequences of dam failure. The intent of
this paper is to review the previous practice of PFMA and compare it to the intended
future practice of risk analysis, and to review past failures for evidence that this is
indeed the case.
2. BACKGROUND
Under the current PFMA process [1], all available background material for each
dam is collected and thoroughly reviewed by a core team of dam safety specialists.
Based on this review, potential vulnerabilities and associated potential failure modes
are identified and thoroughly described. Factors making each potential failure mode
more or less likely are also identified and key factors are noted. Potential conse-
quences associated with each potential failure mode are typically described based
on an inundation zone and qualitative information relating to the population residing
in that zone. Based on the descriptions and factors the potential failure modes are
characterized using the four categories briefly described below.
Category I – Highlighted PFM. Those potential failure modes of greatest

significance considering need for awareness, potential for occurrence, magnitude
of consequence, and likelihood of adverse response (physical possibility is evident,
fundamental flaw or weakness is identified and conditions and events leading to
failure seemed reasonable and credible) are highlighted.
Category II – PFM Considered, but not Highlighted. These are judged

to be of lesser significance and likelihood. Note that even though these potential
154
Q.105 - R.11
failure modes are considered less significant than Category I, they are all also
described and included with reasons for and against the potential occurrence of
the failure mode. The reason for the lesser significance is noted and summarized in
the documentation report or notes.
Category III – More Information or Analyses are Needed to Categorize.

These potential failure modes to some degree lacked information to allow a confident
judgment of significance and thus a dam safety investigative action or analyses can
be recommended. Because action is required before resolution the need for this
action may also be highlighted.
Category IV – PFM Ruled Out. Potential failure modes may be ruled out
because the physical possibility does not exist, information came to light which
eliminated the concern that had generated the development of the potential for the
failure mode, or the failure mode is clearly so remote as to be non-credible or not
reasonable to postulate.
Not only are the category descriptions of limited value relating to conse-
quences, they also do not allow for proper assessment of the probability of failure.
Because of these limitations the PFMA process does not allow FERC licensees to
properly identify the risk-driver PFMs, nor does it allow for prioritization of projects
within the inventory.
3. DRAFT RIDM GUIDELINES
FERC has developed draft RIDM Policy Guidance Documents for use in work-
ing on FERC RIDM Projects. In general, the FERC follows risk methodologies
developed jointly between the Bureau of Reclamation (Reclamation) and the U.S.
Army Corps of Engineers (USACE) and documented in Best Practices in Dam and
Levee Safety Risk Analyses. The FERC draft guidelines were issued in March 2016
and consist of the following chapters:
Chapter 1 – Introduction to Risk Informed Decision Making
Chapter 2 – Risk Analysis
Chapter 3 – Risk Assessment
Chapter 4 – Risk Management
Implementation of the Draft RIDM Guidelines has been slowed by uncertainty

of the benefits and costs on the part of licensees and lack of experience with risk
analysis among consultants. Because of this, FERC has encouraged licensees to
conduct pilot project risk analyses in specific cases, utilizing the methodologies
155
Q.105 - R.11
presented in the Draft RIDM Guidelines. Pilot projects are closely monitored by
FERC and allow the licensee exposure to the risk process with input from FERC
starting during the development phase and continuing through the report. Lessons
learned from the Pilot Projects will inform the development of guidelines currently
in development. Several licensees have opted to perform pilot projects utilizing the
Draft RIDM Guidelines.
Chapter 2 of the draft guidelines have been used for pilot risk analysis projects.
Chapter 2 covers the definition of risk as well as discussion on likelihood of failure
and consequences. The levels of dam safety risk analysis are included in Chapter
2, and are as follows:
Level 1 – Screening Level Risk Analyses
Level 2 – Periodic Risk Analyses
Level 3 – Semi-Quantitative Risk Analyses (SQRA)
Level 4 – Quantitative Risk Analyses (QRA)
The guideline also provides information on the composition and qualifications

of the “Risk Team” needed to complete a dam safety risk analysis.
4. CASE STUDY 1 – OROVILLE SPILLWAY INCIDENT
Oroville Dam is a component of the Oroville-Thermalito Complex, which

includes Hyatt Power Plant, Thermalito Diversion Dam and Power Plant, Fish Barrier
Dam, Thermalito Power Canal, Thermalito Forebay, Thermalito Afterbay, and Ther-
malito Power Plant. The facility is a multi-use water resource project, constructed
for the purpose of water conservation, power generation, flood control, recreation,
and fish and wildlife management. The Oroville-Thermalito Complex stores approx-
imately 3.6 million acre-feet of water and has a generating capacity of 841 MWs
combined from the three power plants. The Oroville-Thermalito Complex itself is a
component of the California State Water Project (SWP), which is the largest state-
owned water storage and delivery system in the United States. The SWP provides
water to about 25 million people and 750,000 acres of irrigated farmland in Califor-
nia. Planning of the SWP began mainly in the 1950s, with design and construction
of most of the SWP completed in the 1960s and early 1970s. The Oroville Dam
project was constructed in the 1960s, with the project completed in 1968. [3]
Oroville Dam is the tallest dam in the United States. The dam is a zoned
earthen and rockfill embankment dam with a maximum height of 770 feet above
the Feather River and is classified as a “High Hazard Potential” structure under the
FERC guidelines.
156
Q.105 - R.11
The Oroville Dam facility (see Figure 1) consists of an embankment dam, the
Oroville Flood Control Outlet (FCO) referred to as the service spillway in Figure 1, the
Oroville emergency spillway, the Hyatt Power Plant, the River Valve Outlet System
(RVOS), and the Palermo Tunnel and Outlet.
The Flood Control Outlet (FCO) is located at the right abutment of the main
dam and has a total of eight radial gates. It is comprised of two concrete monoliths
with each containing four radial gates and five piers. The individual gates are 17 feet 8
inches wide and 33 feet 6 inches high. The sill elevation of the FCO is at EL 813.6 feet
and the top of the gates (in the closed position) are at approximately EL 847 feet.
The FCO has an unlined approach channel and a concrete-lined chute downstream,
which extends to about 75 feet above the Feather River. The spillway chute is 178 feet
8 inches wide. Prior to the February 2017 emergency, the chute was 3,050 feet long
from the FCO to the Terminal Structure (dentates). A non-controlled Emergency
Spillway is located to the right of the FCO. The Emergency Spillway consists of two
sections: a 930-foot long gravity ogee weir on the left side and an 800-foot long
broad crested weir on the right side. The concrete gravity ogee monoliths and weirs
on the left side have a structural height above bedrock as high as 90 feet. The right
800 feet of the Emergency Spillway is a small ogee weir approximately 1 foot in
height, referred to as the “speed bump.” The crest of both sections is at EL 901 feet,
which is 1-foot above the maximum normal operating reservoir level at EL 900 feet.
Fig. 1
Overview of Oroville Dam facility prior to the February 2017 incident [3]
Vue d’ensemble des installations du barrage d’Oroville avant l’incident de février
2017 [3]
157
Q.105 - R.11
4.1. PREVIOUS PFMA WORKSHOPS
2004 PFMA Workshop: The first PFMA workshop for Oroville Dam was held in
September 2004 and was performed in accordance with FERC Engineering Guide-
lines Chapter 14 dated April 11, 2003. During the 2004 PFMA, one potential failure
mode (PFM) was identified and developed. Additionally, four “Candidate PFMs” were
described in the report along with several “Other Considerations”.
2009 PFMA Workshop: A PFMA Update was performed for Oroville in Septem-
ber 2009 as a part of the FERC Part 12D inspection. The update consisted of a review
of the 2004 PFMA Workshop results, the Supporting Technical Information Docu-
ment (STID), and other project documents relating to operation and dam safety
at Oroville Dam since the previous Part 12D inspection; followed by an informal
one-day workshop discussing these items. Based on the review and workshop, an
updated PFMA report was generated consisting of the 2004 report with revisions
(insertions/deletions) called out in the report. The most significant change to the
report was the addition of a second PFM.
2014 PFMA Workshop: To further understand the project and for DWR to
prepare for the upcoming RIDM guidelines that had been drafted by FERC, a new
comprehensive PFMA was performed for Oroville Dam in 2014. The 2014 PFMA was
conducted in accordance with FERC’s Chapter 14 Engineering Guidelines dated July
1, 2005 and included a workshop that took place from August 13, 2014 to August
22, 2014. A total of 20 PFMs were developed during the 2014 PFMA workshop. Of
the 20 PFMs, eight were related to the Oroville spillways and three were related to
issues at Hyatt Intake or Power Plant, which could reduce the ability to draw down the
reservoir. In addition, numerous other “candidate” potential failure scenarios were
postulated and discussed but were not carried forward past the initial screening
phase. Of these “candidates,” eight were related to the Oroville spillways and three
were related to issues at Hyatt Power Plant that could reduce the ability to draw
down the reservoir.
Of note are two candidate potential failure modes (PFM) were considered but
ruled out from full development [4]. These two PFMs are summarized below.
1. Cavitation or slabjacking results in loss of the concrete lining in the spillway

chute downstream of the flood control outlet (FCO) structure. The rock in the
spillway chute erodes and the FCO is undermined and lost. The rationale for
not carrying this PFM forward included:
• The spillway channel concrete is in good condition and there is no evi-
dence of significant erosion or stress resulting from flows experienced
to date.
• The rock is fresh and hard and resistant to erosion.
• The duration of large flows through the FCO is not sufficient to develop
significant erosion of the rock.
158
Q.105 - R.11
• DWR has performed minor repairs to the spillway concrete as recently

as 2009.
2. A PMF flood event is occurring and over 10 feet of water is spilling over the
emergency spillway at Oroville Dam. Erosion begins where the flow is enter-
ing the Feather River and progresses by head-cutting into the reservoir. The
rationale for not carrying this PFM forward included:
• The broad area below emergency spillway reduces energy of water.
• 50 years of flow occurring at the downstream end of the FCO spill-
way channel, and very little erosion has been observed. A 150,000 cfs
release through the FCO has also occurred.
• Distance from the Feather River to the emergency spillway is approxi-
mately 3/4 mile.
• The rock between the Feather River and emergency spillway is very
competent and resistant to erosion.
4.2. OROVILLE SPILLWAY INCIDENT
On February 7, 2017 damage was observed in the service spillway at Oroville

Dam. Based on the findings of the Independent Forensic Team (IFT) [3] the service
spillway chute failure most likely initiated near Sta. 33 + 50 due to uplift pressures
beneath the slap, see Figure 2.
Fig. 2
Initiation of Spillway Chute Failure [5]
Déclenchement de la défaillance de la goulotte de l’évacuateur [5]
159
Q.105 - R.11
Once the initial section of the slab failed, the underlying weathered rock and
soil-like foundation were exposed to high velocity flow which rapidly eroded founda-
tion materials in the area and removed additional sections of the chute slab. Flow
continued eroding portions of the chute slab and underlying foundation until the
service spillway gates were closed. Figure 3 shows the extent of damage.
Fig. 3
Extent of Initial Damage to the Spillway Chute [5]
Étendue des dommages initiaux de la goulotte de l’évacuateur [5]
The IFT further concluded that the initiation of failure was caused by uplift pres-
sures beneath a section of the slab which exceeded the uplift resistance provided
by the weight of the slab and weight of water above the slab, foundation anchor-
ing, and bonding between the concrete slab and foundation. Contributing factors
included foundation conditions, cracks in the slab, slab delaminations and spalling
with ineffective repairs, corrosion and failure of reinforcing, and joints without water-
stops [4]. Following initiation of the chute slab failure, CA DWR was faced with
difficult decisions that impacted their ability to control the reservoir. The emergency
spillway had never been activated and due to the continued rainfall, the reservoir
continued to rise. There was obvious uncertainty and concern about opening the
service spillway gates and causing additional damage. In addition, due to debris
from the damage in the service spillway, the Feather River had become partially
blocked resulting in tailwater rise and if that continued, the underground power plant
160
Q.105 - R.11
could be flooded. There was additional concern about the power transmission lines
that crossed the service spillway and the possibility of continued use of and damage
to the service spillway resulting in damage and failure at one of the transmission
towers. After several attempts to test the performance of the service spillway and
the observed continued damage that occurred as a result, inflows greater than had
been projected occurred resulting in activation of the emergency spillway.
Flow over the emergency spillway began on February 11, 2017 and continued
for about 36 hours. The flow peaked at about 3:00 PM on February 12th , at about
12,500 cfs, roughly 3 percent of the emergency spillway capacity [4]. Some erosion
of the downstream hillside was anticipated, and attempts were made to clear the
downstream area of surficial vegetation and debris in preparation of flow. As the
flow over the emergency spillway increased, erosion occurred as expected but with
time progressed beyond what was anticipated. By the afternoon of February 12th ,
concentrated areas of erosion were observed to be rapidly progressing upstream
toward the emergency spillway crest structure [4], see Figure 4.
Fig. 4
Damage at the Emergency Spillway [4]
Dommages à l’évacuateur de secours [4]
4.3. WHY THE INCIDENT HAPPENED
Excerpts below were taken directly from the IFT Report [3].
“The February 2017 spillway incident was a result of interactions of numerous

physical and human factors, beginning with the design of the project and continuing
during the half-century until the incident.”
161
Q.105 - R.11
“….. this report separately discusses the more general organizational, regu-
latory, and industry factors which contributed to the incident. While these general
factors contributed to the incident in a less direct way than the factors specific to
Oroville Dam and its spillways, they did have a significant and important contextual
influence in creating the long trajectory of events which led to the February 2017
incident.”
“On the whole, the incident can be viewed as a textbook case of a major dam
incident, in terms of typifying the extent of the contribution from human factors. The
incident was preceded by decades of somewhat complex interactions and effects of
human and physical factors, through which numerous warning signs of the impend-
ing spillway failure were missed, and many barriers, which were intended to provide
checks and balances, were overcome to eventually produce the spillway failure.”
“Overcoming so many barriers could be thought of as involving a degree of

“bad luck,” but more importantly, it indicates a long-term systemic failure of DWR,
regulatory, and general industry practices to recognize and address the deficiencies
and warning signs which preceded the incident. The incident cannot reasonably be
“blamed” mainly on any one individual, group, or organization” [3].
“The conditions that led to the service spillway chute failure have their roots
in the original design and construction of the chute, which, combined with the geo-
logic conditions at the site, resulted in vulnerabilities in the as-built structure. These
vulnerabilities were not recognized in various inspections and evaluations that were
completed throughout the history of the structure. In fact, warning signs of the vul-
nerabilities came to be accepted as “normal,” and then generally not questioned
further. Because the vulnerabilities were never recognized, chute slab repair efforts
were neither well-conceived nor effective, and likely contributed to deterioration of
conditions over time.”
The following paragraphs summarize the IFTs conclusions specifically regard-

ing the PFMA that was performed in 2014.
As mentioned above, the PFM that initiated at the service spillway was iden-
tified in the 2017 PFMA, but categorized as “Category IV” and ruled out, because it
would not result in uncontrolled release of the reservoir.In addition, it was the opinion
of the PFMA participants that the rock underlying the spillway chute foundation was
generally non-erodible. The IFT concluded that this characterization of the rock was
heavily influenced by DWR opinions provided during the PFMA session. In addition,
geologic records were not diligently reviewed by the participants in the PFMA, to a
level that would have identified the mischaracterization of the foundation conditions
beneath the chute slab.
The PFMA participants also concluded that the chute slab was in good condi-
tion. However, the possibility of a chute failure was not investigated in detail, primarily
because the participants concluded that chute slab failure and subsequent headcut-
ting erosion would not progress to fail the FCO headgate structure and result in an
162
Q.105 - R.11
uncontrolled release of the reservoir. The IFT further concluded that based on the
performance of the spillway after the chute failed, the conclusion that the postulated
PFM would not progress to failure of the dam or FCO structure may have been cor-
rect, and further concluded “…. the February 2017 incident shows how, by limiting
the PFMA process to considering mainly failures resulting in uncontrolled release of
the reservoir, the process can miss component failures that can be very significant
to the dam owner and downstream residents.” In addition, by concluding that ero-
sion in the spillway would not progress to failure of the dam or FCO structure, the
workshop participants may have had only limited discussion of the early stages of
chute slab failure. Had there been more indepth discussion of the structural details
of the spillway they could have identified some of the structural and design issues
with the chute slab.
The IFT could find no evidence that a comprehensive review of the spillway
chute structural integrity had ever been performed and it is believed that such a
review could have identified the following deficiencies [3]:
• The relationship between the protruding drains and the cracks

• The relationship between leakage through the cracks and joints in the chute
slab and the drain flows
• The failure of reinforcing steel bars at cracks
• The foundations that, in places, consisted of strongly weathered rock
“Recognition of this collection of factors could have led to a conclusion that the
chute slab was vulnerable to an uplift and slab jacking failure, followed by erosion of
the foundation.” [3]
5. CASE STUDY 2 – TAUM SAUK DAM FAILURE
The Taum Sauk Project is on the East Fork of the Black River approximately
90 miles southwest of St. Louis, Missouri. The project is a reversible pumped storage
project used to supplement the generation and transmission facilities, and originally
consisted of an upper reservoir (Upper Dam); a shaft and tunnel conduit; a 450-MW,
two-unit pump turbine motor-generator plant; and a lower reservoir (Lower Dam).
The Upper Dam was a continuous hilltop approximately 6,562-foot-long concrete-
faced dumped rockfill dam forming a kidney-shaped reservoir, as shown in Figure 5.
A 10-foot-high, 1-foot-thick reinforced concrete parapet wall at the crest of the
embankment fill extended the crest to elevation 1599 feet at the time of original
construction. It was the first of the large capacity pumped-storage stations to begin
operation in the United States.
The Upper Dam was overtopped during the final minutes of the pumping cycle
on the morning of December 14, 2005. Reservoir data indicate that pumping stopped
at 5:15 AM with the initial breach forming at approximately the same time. Once
163
Q.105 - R.11
Fig. 5
Upper Dam cross section from original design drawings
Coupe transversale de la partie supérieure du barrage
d’après les dessins de conception originaux
overtopping began, erosion started at the downstream toe of the 10-foot-high para-
pet wall. Erosion progressed below the parapet wall, likely causing instability and
resulting in the initial loss of one or two parapet wall sections. Subsequent erosion
of the rockfill embankment formed a breach of about 650 feet wide at the top of
the rockfill dam and approximately 500 feet at the base of the dam. The peak dis-
charge from breach was estimated to be about 273,000 cfs which occurred within
ten minutes of the initial breach. The complete evacuation of the reservoir occurred
within 25 minutes. The breach flows traveled down the west side of Proffit Mountain
into the East Fork of the Black River. Flows destroyed the home of the Johnson’s
Shut-Ins State Park superintendent, flooded motorists on Highway N, significantly
damaged the park, campground, and adjacent properties, and entered the Lower
Taum Sauk Reservoir. The Lower Dam stored most of the releases and had a peak
spillway discharge of approximately 1,600 cfs. This equates to about 1.1 feet over
the spillway crest which is well within the capacity of the lower reservoir spillway.
Upon leaving the Lower Dam area, flows proceeded downstream of the Black River
to the town of Lesterville, MO, located about 3.5 miles downstream from the Lower
Dam. The incremental rise in the river level at Lesterville was about two feet which
remained within the banks of the river. Thankfully, no loss of human life occurred
from this event.
Since construction in the 1960’s, settlement of the rockfill varied between one
and two feet with the greatest settlement occurring at the location of the breach at
panel 72. At this panel, the top of the embankment was at elevation 1586.99 feet
and top of the parapet wall was at elevation 1596.99 feet.
164
Q.105 - R.11
Three forensic evaluation teams investigated and evaluated the cause(s) of the
dam failure [6,7,8].Post-breach inspections and evaluations revealed the following
information [6]:
1. The project had historically operated with a minimum of two feet of freeboard on
the lowest section of the parapet wall. Following installation of a geomembrane
liner in 2004, the dam owner operated the project to fill the upper reservoir
within one foot of the lowest section of the parapet wall. Post-breach evidence
shows the reservoir may have been routinely filled to within 0.25 foot of the
lowest section of the parapet wall.
2. The December 14, 2005 breach was preceded by significant wave overtopping
that occurred on September 25, 2005. Factors involved with this event were
waves due to winds from the remnants of Hurricane Rita combined with a
reservoir level pumped to within 0.4 foot of the top of the parapet wall.
3. On September 27, 2005, the dam owner adjusted the reservoir control pro-
gramming to account for the difference between the actual reservoir levels and
the readings from the reservoir level instrumentation.
4. On October 3–4, 2005, the dam owner personnel discovered that the conduit
which housed the instrumentation for monitoring reservoir levels was not prop-
erly secured to the dam. Deterioration of the instrumentation tie-down allowed
the conduits to move adversely impacting the reservoir level readings. The
instrumentation readings showed reservoir levels that were lower than actual
levels. As a safety measure, the dam owner adjusted the reservoir level con-
trol programming to shut down the pumps when the instruments showed the
reservoir levels were two feet lower than normal settings.
5. Two Warrick conductivity sensors were used as a safety system for shutting
down the units in case of high reservoir levels. The sensors would send a signal
to shut down the units when they became wet. The sensors were physically
relocated to a height that was higher than the lowest point on the parapet wall.
Therefore, if the Warrick Sensors were contacted by water, the Upper Dam
would already be in an “overtopping” condition.
6. Modifications made to the reservoir control programming adversely affected
how the signals from the Warrick Sensors were managed and reported. The
modifications required that both sensors make contact with water to initiate
shutdown. This removed a layer of redundancy to the safety system.
A potential failure mode analysis (PFMA) had not been performed for this
project prior to its failure in 2005.Several potential failure modes were postulated by
a failure review panel [7] and include:
1. Potential Failure Mode No. 1 (PFM 1) - The downstream slope of the embank-
ment is very steep (1.3H:1V or 37.5 degrees) and when overtopping occurs it
is very easy for erosion of the downstream slope to occur and a local increase
in phreatic surface parallel to the slope can result in shallow progressive
sloughing of the slope possibly from the toe upward until sloughing begins
to undermine the parapet wall that leads to sliding and overturning of the wall
which then greatly increases the flow as a single 60-foot-wide panel overturns
165
Q.105 - R.11
or slides resulting in a very high flow which greatly accelerates the failure by
immediately imposing a flow of 7,000 cubic feet per second on the slope.
2. PFM 2 – As overtopping initiates the process in PFM 1 above and the progres-
sive sloughing takes place, the flow of water over the top of the 10-foot-high
wall impinges at the dam crest at a velocity of 25 feet/second and begins locally
undermining the wall footing in addition to the sloughing caused by thin lay-
ers becoming saturated and failing deeper with time. This shortens the time
required to reach overturning or sliding of the wall. In addition to undermining
the wall footing, this jet of water at 25 feet/second impinges on the upper finer
rockfill and can locally transfer to a 10-foot pressure head which can change
the stability of the wall by changing the uplift pressures at the wall toe.
3. PFM 3 – It is possible that the local increase in the phreatic surface between
the parapet wall and the upper part of the slope caused by the impinging jet of
water can cause a local wedge just beneath the wall to deform and/or reach
limiting equilibrium without the entire slope below becoming unstable. This is
one possible mechanism which is enhanced by the high parapet wall loading
in excess of 10 feet of water head. It is obvious that this mechanism can occur
combined with PFM 1 and 2 above.
4. PFM 4 – Deep wedges founded on a base of residual soil inclined downhill
at about 10 degrees. The various wedges could have steep backslopes on a
residual soil base.
Based on the results of post-failure stability analyses, the failure review panel
concluded that PFM 1 was very likely and the progressive sloughing and erosion in
PFM 1 accelerated, leading to sliding or overturning of the wall, when taking into
account the local undermining of the wall by the velocity of the water jet impinging
on the downstream side of the parapet wall footing as described in PFM 2 above.
According to calculations, the parapet wall was likely to fail by overturning if under-
mined by 3 feet. PFM 3 seemed possible and was indicated by a FLAC analysis. The
deep wedges of PFM 4 were analyzed and required the phreatic surface near the toe
to build up to about 30 feet above the base of the toe of the dam. This mechanism
was possible but the time for this deep phreatic surface to build up 30 feet was very
unlikely considering that the “dirty” rockfill would have resulted in a high percentage
of water runoff rather than deep infiltration.
The failure review panel [7] concluded that the exact sequence of the potential
failure mode will not ever be fully known at the breach.The panel judged that the
failure was a combination of PFM 1, 2, and 3 described above. The participation
of a deeper mode such as PFM 4 cannot be excluded however especially after any
wall panel overturning results in a huge flow of water.
6. DISCUSSION
To answer the question of whether the implementation of a risk-informed

approach would have prevented the failures that occurred in the two recent case
166
Q.105 - R.11
histories discussed above, as well as other recent failures is difficult to answer.In

both cases presented in this paper, human factors were as important as physical fac-
tors in the years leading up to the failures.It is noted that organizational, regulatory,
and industry factors contributed to the failures and these influences are not always
flushed out during a PFMA or risk analysis workshop.However, many design and
performance technical lessons learned were also uncovered which should inform
the PFMA/risk analysis process.But one key lesson learned related more to how
PFMA/risk analysis teams are formed.In the case of the Oroville Spillway incident
the IFT identified appropriately the absence of a subject matter expert with spillway
chute expertise during the 2014 and prior PFMA workshops.Had this expertise been
represented among the participants there is a much greater chance that discussions
of the initiating physical conditions would have identified the design and performance
deficiencies and vulnerabilities.It is still uncertain whether the mischaracterization of
the foundation conditions would have been discovered and whether that discussion
could have advanced to discussion of the consequences that played out.
For the Taum Sauk case history the contributing factors that lead to the failure
were all known prior to the failure, including the significant wave overtopping that had
occurred months earlier.A thorough evaluation of these factors by an experienced
team of dam safety professionals in a PFMA or risk analysis session should have
identified this potential failure mode and provided short-term (urgent) risk reduction
measures that could have been implemented to address this potential failure mode.
7. CONCLUSION
A well planned and executed PFMA and risk analysis that includes a com-
prehensive review of existing project records, including design, construction,
operations, and performance records by a qualified team that has the requisite
knowledge and experience applicable to the project components should be able
to identify and appropriately evaluate project potential failure modes and associated
risks to recommend appropriate and prioritized dam safety risk reduction measures
that would reduce the likelihood of future dam incidents and failures.
REFERENCES
[1] “Chapter 14, Dam Safety Performance Monitoring Program”, Federal Energy
Regulatory Commission, April 11, 2003.
[2] “Interim Guidance, Risk-Informed Decision Making (RIDM), Risk Guidelines

for Dam Safety, Version 4.1”, Federal Energy Regulatory Commission, Office
of Energy Projects – Division of Dam Safety and Inspections, March 2016.
167
Q.105 - R.11
[3] “Independent Forensic Team Report, Oroville Dam Spillway Incident,” FERC
Independent Forensic Team, January 5, 2018.
[4] “Potential Failure Mode Analysis Report, Oroville Dam, FERC Project No.
2100-CA” Prepared by Gannett Fleming, Inc., for California Department of
Water Resources, November 4, 2014.
[5] “What Happened at Oroville Dam and Why – Findings of the Spillway Inci-
dent Forensic Investigation”, ICOLD 2018 Presentation, John France, Vienna,
Austria, July 2, 2018.
[6] “Forensic Investigation and Root Cause Analysis, December 14, 2005 Inci-
dent, Upper Reservoir Dike, Taum Sauk Plant”, Paul C. Rizzo Associates,
Inc., April 2006.
[7] “Taum Sauk Upper Dam Breach, Technical Reasons for the Breach of
December 14, 2005”, FERC Independent Panel of Consultants, May 24, 2006.
[8] “Report of Findings on the Overtopping and Embankment Breach of the Upper
Dam – Taum Sauk Pumped Storage Project”, Internal Report, FERC Taum
Sauk Investigation Team, April 28, 2006.
168
Q.105 - R.12
-------
-------
A DECADE OF LEARNING FROM OUR PAST AND

PREPARING FOR OUR FUTURE: THE ASDSO DAM
FAILURES AND INCIDENTS COMMITTEE (∗)
Lee MAUNEY, P.E., CFM

Project Manager, HDR Engineering – Colorado
Mark BAKER, P.E.

Principal Engineer, DAMCREST CONSULTING, LLC – Colorado
Irfan A. ALVI, P.E.

President & Chief Engineer, ALVI ASSOCIATES, INC. – Maryland
Nathaniel GEE, P.E.

Regional Engineer, BUREAU OF RECLAMATION – Nevada
Gregory RICHARDS, P.E., CFM

Senior Project Manager, GANNETT FLEMING – Utah
Dusty MYERS, P.E.

Chief Dam Safety Engineer,
EAGLE CREEK RENEWABLE ENERGY – New Jersey
Mark OGDEN, P.E.

ASDSO, Technical Specialist and States Outreach – Kentucky
USA
∗ Une décennie d’apprentissage de notre passé pour préparer notre avenir : le comité des
défaillances et incidents des barrages de l’ASDSO
DOI: 10.1201/9781003211808-58 169

Q.105 - R.12
SUMMARY
In order to help investigate, learn from, and disseminate lessons learned from
dam failures and incidents, the Dam Failures and Incidents Committee (DFIC) of the
Association of State Dam Safety Officials (United States) was formed in 2010. During
over a decade of research and study, the DFIC has produced a voluminous body
of work. The DFIC intends to continue these efforts and will welcome collaboration
with professional organizations in the United States, as well as international partners,
with the aim of contributing to improved dam safety on a global scale.
RÉSUMÉ
Afin d’aider à enquêter, à tirer des enseignements et à diffuser les leçons des
défaillances et des incidents liés aux barrages, le Comité des défaillances et des
incidents liés aux barrages (DFIC) de l’Association des responsables de la sécurité
des barrages (ASDSO - États-Unis) a été créé en 2010. Au cours de plus d’une
décennie de recherches et d’études, le DFIC a produit un volumineux corpus de
travail. Il entend poursuivre ces efforts et accueillera favorablement la collaboration
avec les organismes professionnels aux États-Unis, ainsi qu’avec des partenaires
internationaux, dans le but de contribuer à l’amélioration de la sécurité des barrages
à l’échelle mondiale.
1. IMPORTANCE OF DAM FAILURES AND INCIDENTS
Whenever a dam fails or a major incident occurs, many questions are asked.
Why did this happen? How did this happen? And perhaps the most important ques-
tion: how can we prevent this from happening again? Trying to answer the last
question has led to changes in policies regarding how we design, build, inspect,
evaluate, monitor, operate, and maintain dams. And in recent history, it has led to a
growing industry dedicated to ensuring that dams can fulfill their functions in a safe
manner. Yet, despite these changes, dams continue to fail every year in the United
States and internationally. We must ask ourselves whether we have really learned
all the lessons we can from past failures and have incorporated the lessons into
current dam design, construction, and operational practices.
The most devastating dam failure in the United States was the South Fork
Dam that failed above the town of Johnstown, Pennsylvania on May 31, 1889, killing
over 2,200 people. Despite the devastation, the idea that the government should
enact laws that oversaw private industry was still resisted at the time, and there
were no laws passed either nationally or locally [1] in reaction to this catastrophe.
170
Q.105 - R.12
Unfortunately, modern history shows that a series of noteworthy dam failures some-
times needs to occur in short succession for action to be taken. For example, in
Pennsylvania, people still likely remembered the Johnstown Flood (Great Flood of
1889) when, in 1911, the Austin (Bayless) Dam in Potter County, Pennsylvania failed,
claiming 78 lives. Shortly following this, in 1913, Pennsylvania became one of the
first states to pass legislation related to oversight of dam safety [1].
Other states followed with their own state laws for oversight of dam safety, but
often not until their own disaster. For example, at midnight on March 12, 1928, the
St. Francis Dam failed and sent a 70-foot wall of water down the San Francisquito
Canyon near Los Angeles, California. The flood took over 400 lives and left thou-
sands homeless [2]. On August 14, 1929, the state of California enacted two key
pieces of legislation. One, the Civil Engineers Registration Bill, required engineers
to be licensed in order to approve certain designs [3]. In addition, a statewide dam
safety program was created, which evolved into the California Department of Water
Resources, Division of Safety of Dams. The program gave the state of California
oversight of design, construction, and maintenance of all jurisdictional dams [2].
While states continued to pass their own laws regulating dams, for several
decades there were no federal laws, policies, or guidelines related to the oversight or
design of dams, despite the fact that tens of thousands of dams were being built dur-
ing the mid-twentieth century by federal agencies including the Natural Resources
Conservation Service (NRCS), United States Army Corps of Engineers (USACE),
Tennessee Valley Authority (TVA), and Bureau of Reclamation (Reclamation).
A series of dam failures in the 1970s finally changed dam safety regulation
at the federal level. There were many notable failures in this time period, but the
one that began the conversation in the halls of Congress occurred in West Virginia.
On February 26, 1972, Pittston Coal Company’s coal slurry impoundment dam #3
failed on Buffalo Creek, causing the deaths of 125 people [4]. This led to hearings in
Congress and ultimately the passing on August 8, 1972 of the National Dam Inspec-
tion Act, Public Law 92-367 [5]. But to say that everyone in the federal government
was on board with federal involvement in dam safety would be incorrect. President
Richard Nixon himself, who signed the law, also heavily criticized it, saying that the
federal government should not get involved and should instead leave dam safety to
the states. The actual inspection portion of the National Dam Inspection Act went
unfunded until further failures reiterated the need for proper funding.
The next major dam failure in the United States, which occurred four years
later, was the failure of Teton Dam on June 5, 1976, causing 11 fatalities [6]. The
failure was especially significant because Teton Dam was built by a federal agency,
the Bureau of Reclamation, which was considered a premier dam builder in the
United States. At 305 feet high, it was also the tallest dam to fail in United States
history. This failure reinforced the need for federal guidelines and oversight of the
safety of federal dams. Accordingly, on April 23, 1977, President Jimmy Carter
released a Presidential Memo on Dam Safety that called for, among other things,
federal guidelines for dam safety [5].
171
Q.105 - R.12
The year 1977 served as a reminder of the need to manage the safety of our
dams, as there were several additional deadly dam failures. The failure of Laurel
Run Dam, along with several other dams that failed during the same flood event in
Pennsylvania, killed a total of 85 people. Later that year, the failure of Kelly Barnes
Dam in President Jimmy Carter’s home state of Georgia killed 39 people [7]. These
failures led to Carter working with Congress to fund the National Dam Inspection Act.
One of the stipulations for states to receive funding was that they have an adequate
state dam safety program [5].
Many of the pillars of dam safety which we have today in the United States
began within the next few years. Reclamation’s dam safety law was passed in 1978 to
give the agency their own dam safety program and funding. The Federal Emergency
Management Agency (FEMA) was founded in 1979 and in it a dam safety office was
formed that oversaw the completion of the Federal Guidelines for Dam Safety [8]. In
1983, the Association of State Dam Safety Officials (ASDSO) was formed so that
officials from all states could meet and share best practices in dam safety.
Dam safety regulation is now in place in 49 of the 50 states. Design practices,

analysis models, and understanding of primary failure modes related to dams have
advanced greatly, resulting in measurably improved safety of new and existing dams.
Significant investments have also been made to bolster emergency planning and
preparation so that people can be quickly evacuated if a failure does occur. Practices
for evaluating safety of dams have been evolving in the United States from a reliance
on dam inspections and standards to assessing the risks of dams.
Knowledge of dam failure case studies is an essential element to the engineer-

ing practice of evaluating the risks of dams based on potential failure mode analysis.
Case studies reveal dam failure and incident progression mechanisms, as well as
frequency data used in probability assessment.
While this work is notable and has led to safer dams in the United States, much
work is still required, and dams still too often fail. Based on an ASDSO database of
failures and incidents, 887 dam failures and incidents occurred in the United States
between 2005 and 2019, and 306 of these were failures, averaging about 22 dam
failures per year [9]. While many of these failures and incidents occurred at small
dams, some – such as the spillway failures at Oroville Dam in 2017 – were highly
publicized and the lessons captured afterward have had a large impact on the dam
safety community. Yet the lessons from other well-known failures are gradually being
lost to time. The large majority of dam failures and incidents remain obscure and
have little to no investigation or documentation of lessons that might be learned.
One example of this is Ka Loko Dam on the island of Kauai, which failed on March
14, 2006, killing seven people. A legal investigation was completed by a Special
Deputy Attorney General, but the report contained limited technical analysis, was
used mostly for court proceedings, and was not widely distributed or discussed in
the dam safety community [10].
Feedback from dam failures is only impactful if the profession makes the effort
to investigate new failures and incidents, and researches failures and incidents of
172
Q.105 - R.12
the past. The investigators and researchers must then educate others regarding the
lessons from these events for them to be truly and widely learned. However, it is
often difficult to share these lessons broadly across organizational and international
boundaries. Compounding this issue, the dam safety industry is experiencing sig-
nificant turnover in the United States, with high rates of retirement among senior
engineers and regulatory staff. Much of the knowledge and experience gained from
decades of building new dams and learning from failures will be lost unless these
lessons are conveyed to the rising generation of dam safety professionals.
In order to help investigate, learn from, and disseminate lessons learned from
dam failures and incidents, the Dam Failures and Incidents Committee (DFIC) of
ASDSO was formed in 2010 (originally called the Dam Failure Investigation Com-
mittee). The mission of the DFIC is to discover and distribute lessons to be learned
from past dam failures and incidents, and to support quality investigations of new
failures and incidents. During over a decade of research and study, the DFIC has pro-
duced a voluminous body of work. This paper outlines the many ways the committee
has done this.
2. DAM FAILURES AND INCIDENTS COMMITTEE (DFIC)

FOUNDING AND MISSION
For countries interested in examining and learning from their own dam inci-
dents and failures, it is valuable to understand how the ASDSO DFIC was formed,
its mission, the projects it completes, and how it functions within the larger ASDSO
organization.
The genesis of the DFIC occurred in 2008 when its founding member co-
authored a technical paper titled Dam Failure Investigation Approaches [11]. Seven
major dam failures were reviewed. This paper showed that there was no standard
methodology for conducting dam failure investigations in the United States. At the
2010 ASDSO national conference, a “soapbox session” was held to discuss recent
and best practices for conducting dam failure investigations. State representatives
shared their recent dam failure and investigation experiences. This soapbox led
to the creation of the Dam Failure Investigation Committee (DFIC). In 2011, the
DFIC authored the initial version of the ASDSO Dam Failure Investigation Guideline
[12]. This guideline shares lessons from previous dam failure investigations and
is designed to assist state dam safety agencies and dam owners in making good
decisions during the chaotic and high stress period during and following a dam failure
or incidents.
In 2013, the Committee completed a Charter that renamed the commit-

tee the “Dam Failures and Incidents Committee” as recognition that both failures
173
Q.105 - R.12
and incidents offer valuable lessons learned. The Charter assigned the following
goals:
1. Maintain and update the Dam Failure Investigation Guideline [12] through
ongoing committee research into past dam failures and investigation practices
of others.
2. Collect and share lessons learned from past dam failures and incidents with
the ASDSO membership.
3. Provide support to ASDSO and states during and following dam failures and
incidents, with the goal of advocating for successful investigations by qualified
and trained personnel.
4. Train the state dam safety officials and other ASDSO members in the best
possible industry methods and practices for investigating and reporting on
dam failures and incidents.
The committee is currently led by two co-chairs and is assigned a representa-

tive from the ASDSO Board of Directors. The committee has about a dozen official
members and about two dozen technical advisors who occasionally work on projects.
While the committee’s financial year follows the ASDSO fiscal year, its project plan-
ning schedule is different. Each year in September, the DFIC holds its primary
annual meeting, where attendance has continued to expand, with 25 to 35 peo-
ple participating each year. During this meeting, the past year’s accomplishments
and challenges are recognized and new projects are discussed. At the end of the
meeting, attendees sign up for projects to be completed in the coming year.
3. DAM FAILURES AND INCIDENTS COMMITTEE (DFIC) PROJECTS
The DFIC has many ongoing projects. Activities that have a clear beginning
and end, such as researching a particular dam failure, are considered projects.
Initiatives that are continued year-to-year, such as the Decade Dam Failure Series,
are organized as subcommittees and have at least one chair. Subcommittees are
often responsible for multiple projects.
3.1. DECADE DAM FAILURES SERIES
The Decade Dam Failure Series (“Decade Series”) subcommittee of the DFIC
is an initiative of the DFIC to provide ongoing research and presentations on past fail-
ures and incidents at each of the annual ASDSO national conferences. Dam failures
and incidents that typically occurred during a decade or multi-decade anniversary
of the conference event are covered. For example, Castlewood Canyon Dam in
Colorado failed in 1933, and in 2013 a member of the DFIC presented the case
174
Q.105 - R.12
study during a concurrent session sponsored by the Decade Series. These decade
dam failures are researched by both DFIC members and technical advisors, and
an overview of the failure or incident, relevant findings, and lessons learned are
presented at the annual conference General Session (typically 90-minute presen-
tations), during Concurrent Sessions (15-30 minute presentations), or as posters,
which include “lightning session” talks (5 minutes).
The ASDSO Conference Planning Committee has been very supportive of

the Decade Series, and conference attendees consistently rate the series favorably.
Each of the annual conferences since 2013 has included at least one Decade Series
concurrent session, with the last two annual conferences (2019 and 2020) including
three concurrent sessions. Since the Decade Series began in 2013, there have been
a total of 70 unique failures and incidents presented, with the years of failures and
incidents ranging from 1874 to 2019. The failures and incidents include events from
across 30 states in the United States as well as international events in 13 different
countries. The full list of Decade Series presentations is available online in Appendix
A (https://drive.google.com/file/d/1kUCY9i-ifjZCT_z9uPyqgDQAbmtZ7Nk0/view).
Every decade will have been covered by 2022. Despite the significant amount
of case studies that have been researched, there remain many more noteworthy dam
failures and incidents which can be studied and presented at future conferences.
For example, during the next ten years, there can be more emphasis on learning
from failures and incidents that are not currently as well known to the dam safety
industry, as well as focusing on more international failures and incidents.
The Decade Series is a unique opportunity for young professionals to get

involved with an ASDSO committee and learn research and presentation skills. It is
often intimidating for young professionals to present to a room of experienced dam
safety professionals. The Decade Series often pairs more seasoned experts with
young professionals in order to provide mentoring opportunities. Poster presenta-
tions and lightning session talks are shorter in length, with a smaller audience, so
they offer a good first chance to complete a case study.
Some lessons learned from conducting the Decade Series include:
• A team of dedicated researchers who stay committed during the course of a

decade-long effort can produce voluminous output.
• It can be difficult for the DFIC to obtain information on failures and incidents that
occur internationally, especially in countries where English is not the primary
language.
• Dam failures have many facets such as mechanisms of failure, design/
construction issues, emergency response, efforts to save the dam, conse-
quences of failure, and the investigations.
• There are many significant failures and incidents for which the DFIC has limited
information, especially international events and those that the news media do
not cover.
175
Q.105 - R.12
• The Decade Series includes case studies for individual failures and inci-
dents, but, aside from human factors work done by DFIC members (described
below), there have been limited efforts to study cross-cutting themes or lessons
learned that emerge when analyzing commonalities across multiple events.
• The DFIC has studied many more case studies at dams than at levees. There
are many levee failures that merit further research and documentation.
• There have been many more studies of failures than incidents (where the dam
was not fully breached). Much can be learned from incidents where a breach
was averted.
• On average, failures and incidents were studied approximately 50 years after
the event. There were only four case studies in the Decade Series from the
1800s, but these were well known at the time. It can be challenging to obtain
quality information years or decades after an event. The case studies that
were completed many years after an incident rely upon documented investi-
gations that occurred shortly after the failures and incidents, highlighting the
importance of forensic investigations that gather perishable information.
• The historical research efforts have been successful. It has proven beneficial
to provide researchers with “tips” for researching and presenting dam failure
and incidents, along with a list of key areas to cover when developing a case
study. An example list is available online in Appendix B (https://drive.google.
com/file/d/1vUfff4h8bDBb1My4knlin4EMWQJ5D-pY/view).
3.2. ASDSO DAMFAILURES.ORG WEBSITE
In September 2015, ASDSO, in cooperation with the FEMA, released the

“Lessons Learned from Dam Incidents and Failures” website (also known as
DamFailures.org). The primary goal of this website is to convey educational infor-
mation relating to and resulting from dam failures and incidents in an innovative,
user-friendly manner. It is not intended to be a comprehensive database, but rather
an educational resource. The website addresses a wide range of failure modes, dam
types, and dam safety practices, including lessons learned relating to a spectrum
of dam safety topics including engineering and design practices, human factors as
they relate to dam failures and incidents, emergency planning and response, opera-
tion and maintenance, and other dam safety regulatory issues. The target audience
is broad and includes dam safety engineers, operators, dam owners, regulators,
emergency managers, academia, and students.
The DamFailures.org website allows users to access a series of web pages

that address specific lessons learned with corresponding case studies (see screen-
shots of the website home page and a typical case study page in Figure 1). Each web
page is peer reviewed and includes a brief narrative as well as pertinent graphics,
photographs, videos, and links to best practice resources and supporting technical
papers. The web format of DamFailures.org allows ASDSO to share information
that has traditionally had limited distribution via conference presentations or pro-
ceedings. As of July 2020, the website included over 20 lessons learned supported
176
Q.105 - R.12
Fig. 1
Home Page and Typical Case Study Page on DamFailures.org
Educational Website.
Page d’accueil et page d’étude de cas typique sur le
site éducatif DamFailures.org.
by 30 case studies of dam incidents or failures that occurred both within the United
States and internationally. The website also provides free access to several ASDSO
on-demand training webinars describing the basics of dam safety for engineers and
dam owners.
Since its release, over 160,000 users worldwide have accessed DamFail-
ures.org to learn more about various dam safety topics. Website usage has steadily
increased over the past several years. In 2020, the average number of sessions
exceeded 7,000 per month. The DFIC, in partnership with ASDSO staff, currently
oversees maintenance and development of the website. Thanks to additional FEMA
funding and volunteer efforts of the DFIC, the website content continues to be
expanded and diversified. The DFIC has plans to add at least 10 more case study
or lesson learned pages in the next year.
3.3. ASDSO DATABASE OF FAILURES AND INCIDENTS
The ASDSO Dam Safety Incident Database provides basic information on dam
safety incidents and failures in the United States to ASDSO members, dam safety
stakeholders, the media, and the public. It is available in a searchable format on
the ASDSO website at damsafety.org/incidents. The database began as a coop-
erative effort between the Department of Homeland Security (DHS) Dams Sector
and ASDSO in 2010. After a few years, DHS was no longer able to support the
database, but ASDSO has continued the project. The initial database contained a
177
Q.105 - R.12
limited number of dam failure records from a previous ASDSO dam failure database
effort. In 2010, ASDSO began to gather dam safety incident information (both failure
and non-failure) from the state dam safety programs. The majority of the incident
records have been collected from state programs following the year in which they
occurred. Some states have been able to provide historic incident information and
ASDSO continues to obtain as much information as possible. The database is not
considered comprehensive of all dam safety incidents, both historic and current, and
reflects only the data that ASDSO has been able to collect. Much of the identify-
ing information on specific dams is obtained from the USACE National Inventory of
Dams (NID). Data on the information fields collected can be found on the ASDSO
website.
The database currently contains 1,074 total incident records – 390 failure and
684 non-failure. As shown in Figure 2, the large majority of the database records
(73%) are for incidents from 2010 to 2019 (data for most of 2019 and 2020 has not
yet been collected). We have learned, unsurprisingly, that many incidents occur over
a period of days as part of a large hydrologic event and many of the failure incidents
occur at smaller, low and significant hazard potential dams (84%). The data also
confirms that failure and non-failure incidents continue to occur every year.
Fig. 2
ASDSO Incident Database Failure and Non-Failure Incidents (2010–2019)
Base de données des incidents avec défaillances et incidents non défaillants de
l’ASDSO (2010–2019).
Information from the database has been very useful to ASDSO in answering
questions from the media and others. Over the years, ASDSO frequently received
inquiries about the number of dams that fail every year, and now the database
178
Q.105 - R.12
provides a source of information with actual numbers for those requests. Dam safety
professionals and others have used the database as a starting point for more in-
depth analysis of dam failure and incident history. Figures 3 and 4 show summaries
of incident mechanisms and dam failure drivers.
Fig. 3
ASDSO Incident Database, Primary Incident Mechanism (2010–2019).
Base de données d’incidents de l”ASDSO, mécanisme d’incident primaire
(2010–2019).
Fig. 4
ASDSO Incident Database, Incident Drivers (2010–2019).
Base de données d’incidents de l’ASDSO, facteurs d’incidents (2010–2019).
179
Q.105 - R.12
Some requests for incident data lead to requests for more detailed information.
It was never the intent of the database to provide comprehensive information about
a particular incident, but rather to be able to answer basic questions about the
number of failure and non-failure incidents, etc. ASDSO currently does not have
plans to expand the detail of the incident data but is actively seeking additional
incident information (both historic and current) to add to the database.
3.4. HUMAN FACTORS RESEARCH AND APPLICATIONS
Dam failures and incidents are typically preceded by interactions of physical

and human factors which go back years and often decades. These human factors
include situations, judgments, decisions, actions, and inactions at individual, group,
organizational, industry, state, national, and overall system levels. The dam industry
has always implicitly recognized the influence of human factors in dam failure and
safety, but prior to the work done by the DFIC, the industry lacked an adequate
framework to systematically understand and manage the full spectrum of human
factors. This is despite the fact that substantial academic research related to human
factors has been done by psychologists and other social scientists over the past
several decades, and in the past few decades this research has been increasingly
and usefully applied by practitioners in fields such as aviation, health care, nuclear
power, and the chemical industry.
In the dam industry, a milestone for application of human factors research

occurred in 2010, when human factors were explicitly discussed during and immedi-
ately after the initial DFIC “soapbox” session on dam failures and incidents conducted
at the ASDSO national conference. This discussion led to revision of the DFIC’s Dam
Failure Investigation Guidelines to add guidelines for investigation of human factors
[12] and initiated research into human factors by one of the DFIC members (Alvi).
During the past decade, members of the DFIC have continued to play a leading
role in the United States and internationally in making human factors an active area of
research and application in the dam industry. The output of DFIC members related
to human factors has included numerous publications, conference presentations,
webinars, and participation in forensic investigations. A brief chronology of this work
is outlined below.
At the 2013 ASDSO national conference, a member of the DFIC presented

a seminal paper on human factors which used the failure of St. Francis Dam in
California as a case study. This presentation generated much interest and discussion
among the conference attendees [13].
At the 2014 ASDSO national conference, three members of the DFIC (Myers,
Alvi, and Baker) participated in a soapbox session on the failure of Big Bay Dam
in Mississippi, discussing both the physical and human factors which contributed
to the failure. This soapbox session was followed by a 2015 paper published in the
180
Q.105 - R.12
ASDSO Journal of Dam Safety [14] and a 2015 presentation at an ASCE meeting
in Baltimore [15].
In 2015, another key milestone was an ASDSO webinar by a DFIC mem-

ber which presented a comprehensive framework for understanding and managing
human factors in dam failure and safety, illustrated with several case studies [16].
At the 2015 ASDSO national conference, a member of the DFIC presented

a paper on the failure of the Sella Zerbino secondary dam in Italy, which explicitly
described both the physical and human factors contributing to the failure [17].
At the 2015 ASDSO Northeast regional conference, a member of the DFIC

presented a keynote address which described the physical and human factors con-
tributing to the failure of Ka Loko Dam in Hawaii [18]. This was followed by a
presentation on this failure at the 2016 ASDSO national conference [19].
Following the 2017 spillway failures at Oroville Dam in California, the federal
regulator required that the owner retain a consultant team to perform an indepen-
dent forensic investigation. Members of ASDSO, the DFIC, and the United States
Society on Dams (USSD) participated in a task force to help define the scope of
the investigation and recommend individuals to serve on the independent forensic
team, and a decision was made by the federal regulator that human factors should
be included in the investigation scope. A member of the DFIC served on the indepen-
dent forensic team to focus on human factors, and the findings of the investigation
included extensive discussion of contributing human factors and associated lessons
to be learned [20]. To the DFIC’s knowledge, the Oroville Dam investigation was the
first investigation of a major dam failure or incident which gave the same level of
attention to human factors as physical factors. The forensic report was followed by
a 2018 ASDSO webinar which overviewed the forensic investigation and its findings
[21], followed by another 2018 ASDSO webinar which focused on the human factors
findings [22].
Following the 2019 failure of Spencer Dam in Nebraska, the owner contracted
with ASDSO to perform an independent forensic investigation of the failure. The
investigation followed the model of the Oroville Dam investigation, giving similar
attention to physical and human factors. A member of the DFIC served as the inves-
tigation team leader and lead for investigation of human factors (Baker), another
member of the DFIC served as human factors technical advisor (Alvi), and another
member of the DFIC served on the Oversight Group for the investigation (Myers).
The forensic report was completed in 2020 [23].
The work related to human factors completed by members of the DFIC during
the past decade has been highly influential in the United States and internationally.
Beyond the examples described above, many members of the DFIC and ASDSO
have made dozens of presentations on dam failures at ASDSO conferences as part
of the Dam Decade Failure Series, most of which have discussed both physical and
human factors.
181
Q.105 - R.12
The Oroville Dam forensic report - particularly its human factors analysis and
findings - have been widely cited in media, as well as papers, conference presenta-
tions, and books related to dam safety (e.g., [24]). In addition, the federal regulator
for Oroville Dam, the Federal Energy Regulatory Commission (FERC), which has
one of the largest dam safety programs in the world, is currently in the process of
substantially revising its guidelines for inspection and evaluation of dams based on
human factors lessons learned which were presented in the Oroville forensic report.
The Spencer Dam forensic report, only recently published in April, has already
had an impact on how the dam industry considers failure modes related to ice
flows. The report also highlighted the lack of consistency across states in how dams
are classified in terms of hazard potential based on downstream consequences.
Discussions have been initiated on how best to address these inconsistencies.
3.5. ASDSO JOURNAL OF DAM SAFETY, QUARTERLY BOOK CLUB
A recent DFIC initiative is the Quarterly Book Club, published in each issue of
the ASDSO Journal of Dam Safety. This Book Club reviews recent or older books
pertinent to ASDSO members with the hope to share a reading list and promote life-
long learning in the dam safety industry. The initial book list was developed through
an online survey of dam safety industry leaders. The results of that survey were pre-
sented at the 2019 USSD Conference [25], and the first two book reviews have been
published in the Summer 2020 [26] and Fall 2020 [27] issues of the Journal of Dam
Safety. A list of useful books specifically on dam failures and incidents is available
online in Appendix C (https://drive.google.com/file/d/1p6qQ8wIjEHd_5GlWiq8yw5
QU47HtkQjI/view).
4. DAM FAILURES AND INCIDENTS COMMITTEE (DFIC): LOOKING

AHEAD
4.1. CONTINUATION OF CURRENT INITIATIVES
The DFIC continues to be a very active committee. Much attention is paid to

continuing with projects it has begun during the past decade. While the DFIC has built
a voluminous body of work by investigating 70 failures as part of its Decade Dam Fail-
ure Series, there are still more failures and incidents to be investigated and new audi-
ences to which to present these case studies. The DFIC will continue to make failure
case study presentations as an important part of ASDSO national conferences.
In addition, lessons learned from past failures will continue to be published

and widely disseminated. DamFailures.org will continue to expand its content as
182
Q.105 - R.12
new case studies and lessons learned are added each year. This website is already
seen as a trusted source for dam safety information, and with time and expanded
content, the audience and usage of the website are expected to increase.
ASDSO, in conjunction with DFIC, will continue to track dam failures and
incidents. As more professionals learn about the ASDSO Dam Safety Incident
Database, it is hoped that there will be increased reporting to the database. The
DFIC knows that many failures and incidents still occur that never get reported. As
ASDSO continues outreach, it is hoped that the number of unreported failures and
incidents will diminish and thereby improve the usefulness of the data.
The Dam Failure Investigation Guideline has been useful in recent forensic
investigations. The DFIC sees its use expanding over time, but to make the guide-
line as useful as possible, it will need to be updated. The DFIC plans to update
the guideline based on the lessons it has learned from recent investigations and
will regularly update it as needed. There is also a need for a uniform dam failure
investigation policy (discussed below) and this guideline will assist in that effort.
The understanding of human factors continues to be an important goal. One

of the DFIC members (Alvi) will be looking at how human factors affected the failures
of two dams in Michigan in 2020 as part of that dam failure forensic investiga-
tion team. The ASDSO monthly webinar program will feature a presentation in
late 2020 on human factors which influence judgment and decision-making in dam
engineering.
4.2. EFFECTIVELY SHARING LESSONS LEARNED
As noted previously, dam failure investigations and research advance the

industry only if the lessons gleaned from these events are shared and learned
by the industry at large. The DFIC has made considerable strides in reaching a
larger audience via modern means such as the DamFailures.org website. However,
disseminating the information is only the first step. We need to gain a better under-
standing of how to help others learn and apply the lessons we are sharing. For
example, website usage statistics from DamFailures.org indicate that users spend
over three hours looking at case studies on the site for every hour spent looking at
lessons learned. With this information, the DFIC is making efforts to weave more
of the technical lessons into the case study content rather than relying on users to
follow the links to the separate lessons learned pages.
It may also be necessary, in this rapidly changing technological world, to adapt

teaching methods in order to effectively train the next generation of dam safety pro-
fessionals. A recent study found that young people in ‘Generation Z’ (those born
between 1997 and 2012) preferred learning new things by watching YouTube® or
similar type videos over in-person group activities or printed books [28]. Future
efforts to share lessons learned from dam failures and incidents will need to adapt
183
Q.105 - R.12
to the rising generation and could possibly include new formats such as short instruc-
tional videos, webinars, or podcasts. The DFIC will collaborate with others both in
and out of the dam safety community to understand these trends and adapt our
efforts to share lessons learned from dam failures and incidents accordingly.
4.3. NEED FOR UNIFORM DAM FORENSIC INVESTIGATION POLICY
The DFIC has supported the profile of ASDSO as an independent, credible

organization for participating in, initiating, and conducting independent dam forensic
investigations. Two recent major dam forensic investigations in which ASDSO and
the DFIC had involvement were the Oroville Dam and Spencer Dam investigations.
These investigations were perceived by stakeholders and the public to be thorough
and unbiased, and garnered valuable lessons learned.
However, challenges remain to the performance of timely and independent

dam forensic investigations, the findings of which are released to the public. As illus-
trated by the recent experience in initiating the independent forensic investigation
of two dam failures in Michigan, which involved significant delays, misunderstand-
ings by stakeholders and politicians regarding the scope of the investigation, and
questions about how the forensic team was formed, the dam industry in the United
States lacks a uniform law, policy, guideline, or process for initiating and conduct-
ing investigations. After a major failure, it is usually unclear what the scope of the
investigation should be (particularly the extent to which human factors should be
investigated), how investigation team members should be selected in a way that
assures independence and multidisciplinary expertise, and who should pay for the
investigation. For example, it took more than four months to initiate the Spencer Dam
Failure investigation, in part because of legal concerns. Precious data can be lost
during such delays as people’s memories fade and perishable evidence becomes
lost.
How do we get over the hurdles that impede initiating and conducting investiga-
tions? Dam failures and incidents often have major financial, legal, and reputational
implications for dam owners, regulators, consultants, construction contractors, and
local communities. Findings from investigations may be adverse to the interests
of one or more of the parties. Due to all of these factors, there may naturally be
reluctance to fund or even cooperate with a forensic investigation.
Nevertheless, in order to advance the dam engineering profession and safe-

guard the public and environment, conducting independent forensic investigations
for significant dam failures and incidents is necessary in order to reach unbiased
and accurate conclusions regarding the physical and human factors that contribute to
these failures and incidents, and to identify and disseminate the associated lessons
to be learned.
184
Q.105 - R.12
The DFIC believes that this situation should be corrected starting with a
national policy or guidelines for initiating and conducting “official” dam forensic
investigations soon after a major failure or incident. Such a policy should address
procedures for establishing an independent investigation team, funding of the inves-
tigation, liability of the investigative team, access to sites, subpoena power and
confidentiality, preservation and gathering of evidence, report review processes,
fact checking, acceptable uses of the forensic reports in legal proceedings, and
communication with the media and stakeholders. With this goal in mind, the DFIC
initiated efforts to help establish a national dam failure investigation program in 2018
in partnership with FEMA [29], however a dam failure investigation program has not
yet been established.
4.4. ADDITIONAL INTERNATIONAL COLLABORATION
One observation made by the DFIC is that language barriers, and to some
extent a natural tendency towards insularity among nations, limit international col-
laboration on work related to dam failures and incidents. Illustrating this, the DFIC’s
work has generally been in isolation from work which may be underway by any sim-
ilar groups in other countries around the world. Members of the DFIC have often
found that their ability to learn about dam failures and incidents outside the United
States has been limited when documents related to those failures and incidents
were not available in the English language.
The DFIC hopes to find ways to overcome these barriers and share information
and lessons to be learned across borders. In particular, the DFIC would benefit from
having international partners who could assist in two-way sharing of information and
lessons learned from dam failures or incidents across international borders.
5. CONCLUSIONS
The DFIC has been highly productive over the course of the past decade due
to strong leadership, effective meetings, interesting subject matter, and a sincere
desire by the committee members to research and share valuable lessons learned
from dam failures and incidents.
Henry Petroski, acclaimed engineering historian, writes that the concept of

failure is central to advancement of engineering, specifically because engineering
design has, as its “first and foremost objective the obviation of failure,” and failures
and incidents are “unplanned experiments that can teach on how to make the next
design better.” [30]
185
Q.105 - R.12
The dam safety industry has made considerable and measurable progress
in large part by learning from past failures and incidents. The large majority of
dams do not fail, but the relatively rare outliers are the ones that get the press: “To
speak of engineering failures is indirectly to celebrate the overwhelming numbers of
successes.” [30]
The DFIC’s original goal of developing a failure investigation guideline was

accomplished and could have been the singular focus for the committee. However,
the committee members realized that there were numerous ways that the DFIC could
use its research into dam failures and incidents to develop educational tools which
help drive the industry forward. The DFIC intends to continue these efforts and will
welcome collaboration with professional organizations in the United States (e.g., the
American Society of Civil Engineers and the United States Society on Dams), as
well as international partners, with the aim of contributing to improved dam safety
on a global scale.
REFERENCES
[1] Rose, A. T. (2013). The Influence of Dam Failures on Dam Safety Laws in
Pennsylvania, Association of State Dam Safety Officials, Proceedings of Dam
Safety 2013.
[2] St. Francis Dam Failure (California, 1928). Lessons Learned Case Study on
DamFailures.org website, Association of State Dam Safety Officials, accessed
June, 2020.
[3] Wilkman, J. Floodpath: The Deadliest Man-Made Disaster of 20th-Century

America and the Making of Modern Los Angeles. New York: Bloomsbury. Page.
237. 2016.
[4] Buffalo Creek Dam Failure (West Virginia, 1972). Lessons Learned Case
Study on DamFailures.org website, Association of State Dam Safety Officials,
accessed June, 2020.
[5] Gee, N. (2017). The Impact of Dam Failures on the Development of Dam
Safety Legislation and Policy in the 1970’s. Journal of Dam Safety, 2017.
[6] Teton Dam Failure (Idaho, 1976). Lessons Learned Case Study on DamFail-
ures.org website, Association of State Dam Safety Officials, accessed June,
2020.
[7] Kelly Barnes Dam Failure (Georgia, 1977). Lessons Learned Case Study on
DamFailures.org website, Association of State Dam Safety Officials, accessed
July, 2020.
186
Q.105 - R.12
[8] Federal Guidelines for Dam Safety. Federal Emergency Management Agency
website, U.S. Department of Homeland Security. Site accessed July 2020.
[9] ASDSO Dam Safety Incident Database Search, Association of State Dam
Safety Officials website, accessed July, 2020.
[10] Godbey, R. (2007). Report of the Independent Civil Investigation of the March
14, 2006, Breach of Ka Loko Dam. January, 2007, Volume 1.
[11] Baker, M.E., Graham, W.J., (2008). Dam Failure Investigation Approaches.
Dam Safety 2008, Association of State Dam Safety Officials, Indian Well, CA.
[12] Dam Failure Investigation Guideline (2017), completed by the ASDSO Dam
Failure & Incidents Committee, latest update in 2017.
[13] Alvi, Irfan A. (2013). Human Factors in Dam Failure. Dam Safety 2013,
Association of State Dam Safety Officials, Providence, RI.
[14] Myers, D., Ferguson, K., Alvi, I. & Baker, M. (2015). Reexamination and
Lessons Learned from the 2004 Failure of Big Bay Dam, Mississippi, Journal
of Dam Safety, 13:2. Lexington, KY: Association of State Dam Safety Officials.
[15] Alvi, I. (2015) 2004 Failure of Big Bay Dam, Lamar County, Mississippi, ASCE
Geo-Institute Presentation. Baltimore, Maryland: American Society of Civil
Engineers.
[16] Alvi, I. (2015). Human Factors in Dam Failure and Safety, ASDSO Webinar.
Lexington, KY: Association of State Dam Safety Officials.
[17] Alvi, I. (2015). Failure of Sella Zerbino Secondary Dam in Molare, Italy,
ASDSO Annual Conference. New Orleans, LA: Association of State Dam
Safety Officials.
[18] Alvi, I. (2015). Human Factors in Dam Failure & Safety, Case Study: Ka
Loko Dam Failure, ASDSO Northeast Regional Conference. Ocean City, MD:
Association of State Dam Safety Officials.
[19] Alvi, I., Richards, G., and Baker, M. (2016). 10th Anniversary of Ka Loko Dam
Failure, Hawaii, Presentation. ASDSO Annual Conference. Philadelphia, PA:
Association of State Dam Safety Officials.
[20] France, J., Alvi, I., Dickson, P., Falvey, H., Rigbey, S. & Trojanowski, J. (2018).
Independent Forensic Team Report: Oroville Dam Spillway Incident.
[21] France, J. (2018). Findings of the Oroville Dam Spillway Forensic Investigation.
ASDSO Webinar. Lexington, KY: Association of State Dam Safety Officials.
[22] Alvi, I. (2018). Human Factors in the Oroville Dam Spillway Incident, ASDSO
Webinar. Lexington, KY: Association of State Dam Safety Officials.
187
Q.105 - R.12
[23] Baker, M., Ettema, R., Teal, M., Trojanowski, J. (2020). Spencer Dam Failure
Investigation Report. Association of State Dam Safety Officials. April, 2020.
[24] Oboni, F., Oboni, C. (2020). Tailings Dam Management for the Twenty-First
Century: What Mining Companies Need to Know and Do to Thrive in Our
Complex World. Springer, 2020.
[25] Mauney, L. (2019). Dam and Levee Safety Leaders Library. U.S. Society on
Dams, 2019 USSD Conference and Exhibition, Chicago, IL.
[26] Mauney, L. (2020). To Engineer is Human: The Role of Failure in Successful

Design, by Henry Petroski. ASDSO Quarterly Book Club, The Journal of Dam
Safety, Summer 2020, Vol. 17, Issue 3, p. 44–45.
[27] Mauney, L. (2020). The Signal and the Noise: Why So Many Predictions Fail –
but Some Don’t, by Nate Silver. ASDSO Quarterly Book Club, The Journal of
Dam Safety, Fall 2020, September 2020.
[28] Pearson (2018). Beyond Millennials: The Next Generation of Learners, Global
Research & Insights, August 2018.
[29] Baker, M., Alvi, I., Durgin, S., Mauney, L., Pittman, J., Wooten, L., and
Roberts, T. (2018). Establishing a National Dam Failure Investigation Pro-
gram: Requirements and Approach. Developed by Dam Failure & Incidents
Committee, Association of State Dam Safety Officials. Sponsored by FEMA
National Dam Safety Program, March 2018.
[30] Petroski, H. (1994). To engineer is human: the role of failure in successful

design. New York: Barnes & Noble Books, 1994.
188
Q.105 - R.13
-------
-------
RAPID INUNDATION MAPPING (∗)
Edward STOWASSER, PE
Hydraulic Engineer, U.S. Army Corps of Engineers
Michael KOON, CMS-RS, GISP

GIS Specialist, U.S. Army Corps of Engineers
Whitney Sorrels, PE
Hydraulic Engineer, U.S. Army Corps of Engineers
USA
SUMMARY
The U.S. Army Corps of Engineers (USACE) Modeling, Mapping, and Con-
sequences Production Center (MMC) has developed a unique tool to improve
inundation mapping response times and enhance risk communication planning for
flood events. Having several projects in the USACE portfolio undergoing dam safety
modifications combined with the availability of real-time forecast models, it has been
recognized that there is a need to ensure an efficient and effective way to produce
inundation mapping rapidly. The MMC has developed a web-based GIS tool to assist
emergency management agencies during flood events and planning of emergency
exercises. Risk communication and response is critical to USACE and when dis-
cussing flood risks it is vital to be able to provide both breach and non-breach risk
inundation mapping for potentially impacted areas.
As inundation mapping evolves within USACE so have the tools used to com-
municate risk. This tool supports the risk communication responsibility of the USACE
Dam and Levee Safety program. The MMC team has created this tool to support
∗ Cartographie rapide des inondations
DOI: 10.1201/9781003211808-59 189

Q.105 - R.13
decision makers. It provides a web-based map that displays National Weather Ser-
vice forecasts along with gage-adjusted mapping for reaches throughout an entire
watershed. The maps identify critical infrastructure, population at risk, economic
effects, and key road locations for the full range of flood events. The tool is flexible
and can be used to add other globally accessible web-based data sources that would
be helpful when analyzing critical points of interest based on each community. The
tool can be used for dam breach communication as well as non-breach risk commu-
nication. The existing mapping data can be shared with emergency management
agencies within minutes. The tool will improve coordination with stakeholders and
can adapt to the many needs of multiple stakeholders, ultimately leading to improved
risk communication with emergency managers and essential stakeholders.
RÉSUMÉ
Le Centre de production de modèles, de cartes et de conséquences (MMC) du

Corps des ingénieurs de l’armée américaine (USACE) a développé un outil unique
pour améliorer les temps de réponse de la cartographie des inondations et la plan-
ification de la communication des risques en cas d’inondation. Plusieurs projets du
portefeuille de l’USACE ayant subi des modifications en matière de sécurité des
barrages, combinées à la disponibilité de modèles de prévision en temps réel, il a
été reconnu qu’il est nécessaire de garantir un moyen efficace et efficient de pro-
duire rapidement une cartographie des inondations. Le MMC a développé un outil
GIS basé sur le web pour aider les organismes de gestion des urgences lors des
inondations et de la planification des exercices d’urgence. La communication et la
réponse aux risques sont essentielles pour l’USACE et, lors de l’examen des risques
d’inondation, il est vital de pouvoir fournir une cartographie des risques d’inondation,
avec ou sans brèche, pour les zones potentiellement touchées.
Les outils utilisés pour communiquer les risques ont évolué au sein de l’USACE
au fur et à mesure que la cartographie des inondations se développait. Cet outil con-
tribue à la responsabilité de communication des risques du programme de sécurité
des barrages et des digues de l’USACE. L’équipe du MMC a créé cet outil pour
aider les décideurs. Il fournit une carte en ligne qui affiche les prévisions du Service
météorologique national ainsi qu’une cartographie ajustée au niveau du sol pour
les étendues d’un bassin versant entier. Les cartes identifient les infrastructures
essentielles, la population à risque, les effets économiques et les principaux axes
routiers pour toute la gamme des inondations. L’outil est flexible et peut être utilisé
pour ajouter d’autres sources de données accessibles sur le web au niveau mondial
qui seraient utiles pour analyser les points d’intérêt critiques en fonction de chaque
communauté. L’outil peut être utilisé pour la communication sur les ruptures de bar-
rage ainsi que pour la communication sur les risques autres que les ruptures. Les
données cartographiques existantes peuvent être partagées avec les organismes
de gestion des urgences en quelques minutes. L’outil améliorera la coordination avec
les parties prenantes et peut s’adapter aux nombreux besoins des multiples parties
190
Q.105 - R.13
prenantes, ce qui permettra en fin de compte d’améliorer la communication sur les

risques avec les gestionnaires des urgences et les principales parties prenantes.
1. INTRODUCTION
The Modeling, Mapping, and Consequences Production Center (MMC) has

developed an innovative mapping tool utilizing a robust inventory of existing hydraulic
models. The Rapid Inundation Mapping (MMC-RIM) tool uses a library of stored
inundation maps for multiple flood profiles to help users build an event-specific inun-
dation picture for an entire watershed. The first iteration of the tool was developed
to enhance dam safety risk communication efforts downstream of U.S. Army Corps
of Engineers (USACE) flood risk management dams. It has since evolved to include
consequence estimation, levee closure elevations, impacted critical infrastructure,
and data from other agencies such as traffic data and hurricane tracks in order to
meet the needs and purposes of various users and stakeholders.
Risk communication is important whether interacting internally with dam safety

and emergency management stakeholders, or externally with communities and spe-
cial interest groups. Striving to balance an effective risk communication plan is
critical to the exchange of ideas, issues, concerns and study information. Dam
safety projects cannot achieve success without effective communication channels
and tools in place to convey risk in a way that’s accessible to stakeholders with
varying backgrounds and levels of understanding. Inundation mapping is one such
visual communication tool. It has seen growing acceptance and wider dissemination
throughout the United States following extreme events such as Hurricane Harvey in
2017 and Hurricane Florence in 2018.
The purpose of the MMC-RIM tool is to supply dam owners with a mecha-
nism to rapidly produce inundation mapping at the watershed level, enhance risk
communication to the public, and provide support to local government emergency
managers. Output from the tool can be used to educate about risks associated with
both a dam failure and the risk of flooding that could occur without a dam failure.
2. BACKGROUND
The U.S. Army Corps of Engineers has been conducting risk assessments of
the Corps dam portfolio since 2008. In recent years, dam safety teams have been
working to improve risk communication plans to ensure that USACE communicates
key messages to stakeholders and the public downstream of the projects. Particu-
larly, it was important to address the misconception that a dam safety modification
eliminates the risk of flooding downstream of a dam. The team quickly identi-
fied this risk communication breakdown and began implementing several different
191
Q.105 - R.13
risk communication measures to balance the existing USACE risk communication

message.
USACE has taken steps to reduce dam safety risks associated in its portfolio.
However, even if a dam performs exactly as it is designed, there is still flood risk
to residents and businesses downstream of the dam. Pamphlets, fact sheets, and
public service announcement videos were used to educate the public on the following
measures for reduction of flood risk downstream:
• Understand that downstream flooding can occur

• Know your risk, your role, and actions to reduce your risk
• Strongly consider purchasing flood insurance
• Following instructions from emergency management officials
• Contact your elected officials to make sound flood risk decisions in your area
The concept of an inundation mapping tool was originally developed to sup-

port risk communication efforts associated with Bluestone Dam located in Southeast
West Virginia in the United States. The need for the tool came from years of working
with local and state emergency managers to provide inundation mapping specific to
several different scenarios that the stakeholders were interested in. The ever chang-
ing requests spurred the concept of the tool’s development. The goal was to be able
to provide several flow profiles for the entire watershed and split those profiles into
key reaches in which a gage could reflect a stage, flow, or water surface elevation.
Having a static library of floods indexed allowed for rapid retrieval to provide the
inundation mapping more quickly and efficiently to stakeholders.
Great thought was put into the workflow of the program, making sure to provide
all the tools a user might need during a flood event, resulting in the incorporation of
several crucial outside datasets. The workflow incorporated key pieces of information
from other agencies at the click of a button such as: NWS Forecast website links,
live traffic feeds, and the ability to add critical infrastructure data to the maps. The
first prototype of the tool was successful at providing rapid inundation mapping at
the click of a button for the Bluestone Dam Safety Assurance Project.
During the Hurricane Florence flood mapping response in 2018, Bluestone

Dam had NWS ensemble traces indicating potential flooding that would require a
200,000 cfs release from the spillway gates. Within minutes, the prototype tool was
used to overlay the Hurricane’s track and the stage slider bar was adjusted to identify
which downstream communities were vulnerable to flooding at that rate of discharge.
The information was provided to emergency managers within USACE to help the
state of West Virginia prepare for potential flooding. Ultimately, the storm stalled out
in the Appalachian Mountains and shifted the storm track to the east, resulting in less
flooding than predicted. However, the usefulness of the tool stood up to the challenge
and proved to be very beneficial in predicting consequences and economic effects
for the potential incident.
192
Q.105 - R.13
The success caught the attention nationally within USACE and quickly spurred
development of a scalable watershed approach. The MMC, tasked by USACE
with developing Corps Water Management System (CWMS) models for 201 water-
sheds nationwide, was a natural partner to scale and expand the reach of the tool.
CWMS is the automated information system supporting the USACE Water Man-
agement mission. CWMS integrates real-time data acquisition, database storage,
flow forecasting of watershed runoff, reservoir operation decision support, river pro-
file modeling, inundated area determination, consequence/damage analysis, and
information dissemination into a comprehensive suite of software supporting water
management decision processes. Utilizing the calibrated hydraulic modeling per-
formed under the CWMS program, the developers were able to build the index of
libraries required to feed into the MMC-RIM tool
In 2019, the MMC began a pilot program that utilized Hydraulic Engineering
Center’s River Analysis System (HEC-RAS) river profile modeling previously devel-
oped under the CWMS program to develop 5 watersheds, each varying in terrain
and model types. The Kanawha, Pearl, Red River of the North, Arkansas, and lower
Connecticut Rivers were selected to guide the first phase of development. Each
watershed was carefully selected to test and refine the MMC-RIM tool in the pilot
project phase. The models differed from mountainous to flat terrain, steady flow and
unsteady flow, tidal and tributary influences, and 1D to 2D model types.
3. RAPID INUNDATION MAPPING TOOL
3.1. TOOL FEATURES
The MMC-RIM tool is created using an ArcGIS online server platform to display
the mapping profiles and is coded such that once a slider bar is adjusted it will query
an index of maps for a certain flow profile along that reach of the river and will display
the inundation maps quickly, as shown in Figure 1.
Within the web viewer, users can select their watershed of interest within a drop
down menu. Upon selection, a list of relevant gauges is queued along with links to the
approved and publicly released real-time National Weather Service (NWS) forecast.
The user selects the forecast gage links to view the hydrographs (shown in Figure 2)
and utilizes any other forecasted release knowledge of the system to predict the
correct values to assign at the gage locations. Using a slider bar displaying stage,
flow, and elevation, the user can select the appropriate flood values that corresponds
to the NWS forecast stage and the maps will automatically update for the selected
gages. The inundation layers can be displayed as depth grids (shown in Figure 3)
or shapefile polygons. Users can change the ESRI-supplied background maps to
aerial or street view base maps.
193
Q.105 - R.13
Fig. 1
MMC-RIM Tool Interface
Interface de l’outil MMC-RIM.
Fig. 2
Live Gage Data and Weather Warnings
Données en temps réel et alertes météorologiques.
Once the user has the watershed verified and all the stages selected, they can
create a link, add metadata about the assumptions made for the flood event, and
share the inundation map selection with another user who can view the map and
metadata but not edit the selected stages without consent from the map creator.
194
Q.105 - R.13
Fig. 3
Depth Grids & Corresponding Legend
Grilles de profondeur et légende correspondante.
In addition to the primary functionality, additional modules allow for further

decision-making capabilities and analysis. A key feature of the tool is dynamically
updated consequence estimates of population at risk for day/night, structures
impacted, and economic damages. Once the stages are selected at each gage,
the user can select a consequence button that will query the selections and retrieve
the consequence data for the specific flood event.
The tool is also capable of displaying georeferenced data from other sources
on the web, allowing users to perform on-the-fly, in-browser analysis. Users have
access to all shapefiles associated with the National Levee Database (NLD) includ-
ing levee centerline data (shown in Figure 4) and protected area polygons. Links to
layers such as live traffic feeds (shown in Figure 5), NWS gages with flood indicators,
radar, and hurricane forecast cones have been prepopulated in the MMC-RIM tool
to enhance visualization of the flood event. For Official Use Only (FOUO) Homeland
Infrastructure Foundation-Level Data (HIFLD) was acquired from the Department of
Homeland Security. The dataset allows users to display critical infrastructure such
has hospitals, nursing homes, and schools.
195
Q.105 - R.13
Fig. 4
National Levee Database Centerline Data
Base de données nationale sur les niveaux d’eau (National Levee Database).
Fig. 5
Live Traffic Data
Données en direct liées au trafic routier.
3.2. TOOL DEVELOPMENT
The MMC-RIM tool application was produced using the ESRI Web AppBuilder
Development Kit. This development environment allows for custom widgets to be
created using the ESRI JavaScript API version 3.x. The final application files are
deployed to a web server for access by the appropriate personnel.
196
Q.105 - R.13
The tool utilizes widgets that can be opened from the applications toolbar. The
primary widget is the Inundation Analysis widget. It also utilizes the Add Data widget
which provides access to the numerous data feeds listed in Section 3.1 including
NWS gage information, live animated radar, traffic data, hurricane data, and much
more.
The application assumes a workflow for a real-time event. The user will open
the inundation widget and analyze the current and forecast conditions at the gage
of interest in the watershed. An inundation level slider bar is then adjusted to the
forecasted or current stage level. This triggers the application to display only the
inundation for that stage. After configuration a button can be selected that generates
a dynamically adjusted set of consequence data. Each gage selection is recorded in
a table for the user to visually see the stage/flow selections made to have an overall
picture of the forecast for the entire watershed.
The data, analysis, and product development methods used to produce inun-
dation delineation and consequence data utilized desktop GIS spatial analysis via
automation scripts. The MMC-RIM tool uses ESRI ArcGIS Spatial Analyst, Python
3.x programming language. Utilizing the spatial analysis with customized scripting
tools allowed for in-depth analysis that permitted logic to be applied beyond the
default software functionality.
Team collaboration meetings took place with other similar tools developed by
the NWS, USGS, FEMA, and the State of North Carolina to establish best practices
and lessons learned from the other web viewers. The MMC-RIM tool is unique from
the other tools because it not only predicts accurately near gages but it encompasses
the entire watershed with inundation map coverage by allowing a water manager to
use the tool and select the appropriate forecasted stages at each of the key gage
locations within the watershed.
4. PRODUCTION PROCESS AND LESSONS LEARNED
4.1. HYDRAULIC MODELING
The inundation data, both polygon inundation extents and depth grids, was
produced from HEC-RAS 5.0.7 software. Model scenarios were set up to capture a
designated range of flows at varying increments, and the resulting vector polygon
data was output in ESRI shapefile format. The inundation polygons are then cut into
reaches representing the individual segments that will be adjusted by the end user
with a slider bar.
The flow range is selected by identifying the smallest flow needed on the small-
est tributary and the largest possible flow from a Probably Maximum Flood (PMF)
197
Q.105 - R.13
dam failure from any reservoir within the watershed. For the purposes of the MMC-
RIM tool, channel capacity is often selected as the lower bound. All tributaries have
the minimum to maximum flow ranges computed for water surface profile creation,
but, when setting each reach’s specific flow range in the tool interface, the range
can be limited to the most credible values needed for that gauge.
Flow increments can vary within the selected flow range in order to capture
appropriate stage increases for the subject basin. Stage increments are largely
determined by the basin topography. For a steep watershed, a one-foot increment
may produce the appropriate density to capture flood impacts. Flat, wide floodplains
however, may require smaller a higher density of stages. Once target stages were
selected, modelers used rating curves to determine the appropriate corresponding
flow increments. In general, steep sections of the rating curve require smaller flow
increments to capture the desired stage density.
Each critical gage identified in the watershed was assigned to represent a

reach of the river. Identifying and drawing the reach polygons is a very crucial task
that requires experienced hydraulic engineers to distinguish the extents that are
applicable to each gage. Once the reach polygons are created, the water surface
profile inundation for the entire watershed is cut into separate polygons that represent
each reach, building a library of flow/stage/elevation index maps for the key gages.
Reach polygons delineated for the Kanawha River basin in West Virginia are shown
in Figure 6.
4.2. LESSONS LEARNED
In some cases, the extents between gages were too expansive for accu-
rate stage prediction therefore interpolation was used to identify key flow change
locations. In those cases, long reaches were split and assigned a drainage area
versus river mile ratio to approximate additional flow entering the basin between
gages. When selecting this reach in the tool, the user will be prompted with slider
bars for the upstream and downstream gages. Once the appropriate selections
are made for the bounding reaches, an algorithm interpolates the change in flow
between the two gages and assigns it to the appropriate reach.
Storm surge was a challenge encountered with the Pearl River, a coastal basin
in central Mississippi. In this case, the modeler developed a separate 2-Dimensional
HEC-RAS model for the area that would be impacted by storm surge and identified
the appropriate stage range using historical gage data from the nearest National
Oceanic and Atmospheric Administration (NOAA) tidal gauge station. Using stage
as a boundary condition, inundation layers were created for one-foot increments and
added to the MMC-RIM tool as an additional slider bar reach at the mouth of the
basin. The user can select between a riverine flooding condition and storm surge
tidal condition to accurately reflect the source of flooding in that reach.
198
Q.105 - R.13
Fig. 6
Reach Polygon Delineation
Délimitation du polygone de tronçon de rivière.
4.3. STANDARD OPERATING PROCEDURES
The five selected pilot projects guided the development of a Standard Operat-
ing Procedures (SOP) document which will be used to develop the static map layers
for all 201 watersheds within the next five years. The MMC is a virtual production
center that utilizes resources from district offices, allowing the team to scale to meet
the demands of a national implementation of the tool. Figure 7 captures the process
identified for mass production of the static inundation layers, consequences, and
processing into the MMC-RIM tool. Table 1 summarizes the production process laid
out in the SOP.
5. TOOL TESTING AND VALIDATION
Since this tool has been developed, the USACE Vicksburg District experienced
a significant flood in the Pearl River watershed in February of 2019. The CWMS
forecast models were used to produce accurate inundation mapping of the event
shown in Figure 8. In order to validate the MMC-RIM tool, an effort was made
to compare the results of the MMC-RIM tool to the event-specific CWMS results
199
Q.105 - R.13
Fig. 7
Static Map Layer Production Workflow
Processus de production des couches de cartes statiques.
shown in Figure 9. The results indicate that if the tool was used during the event the
inundation maps would have been accurate enough to issue warnings to residents
with confidence. The layers indicate a negligibly higher water surface elevation than
the observed condition would have been used due to the nature of the selection
of indexed maps available. The benefit of the tool is that it significantly reduces
the production time to model an event and create a map. Having inundation maps
readily available provides local emergency responders time to react and prepare
for a flood event. This tool, once fully implemented, will allow USACE to effectively
communicate risks to local communities quickly and in a way that the public can
understand. Tool effectiveness will be continually evaluated and refined as planning
studies and real-time flood events occur.
6. BENEFITS AND CONCLUSIONS
The MMC developed the MMC-RIM tool to give dam owners the ability to
rapidly develop and disseminate accurate inundation maps during real-time flood
events and for emergency planning and preparedness. The web-based GIS tool
allows maps to be shared both internally with decision makers and externally with
local stakeholders and partners. It will be used within USACE to more effectively
communicate flood risks to stakeholders. The tool allows for customized flood
scenarios for emergency planning exercises that are specific to elevations of key
infrastructure located within the watershed. This will benefit local communities in
helping them better understanding flood impacts.
200
Q.105 - R.13
Table 1
Static Map Layer Production Process
Data Collection Data collection is critical step at the beginning of the process. The modeler will work
with the district staff to obtain all necessary reference material, gauge data, and model
files. It is important that this work be completed prior to the kickoff meeting, so that a
productive kickoff meeting will occur, and all participants understand the scope of the
work and the required effort it will take to complete the project.
Project Kickoff The kickoff meeting is held to clarify the scope of the study as well as the schedule
Meeting and budget. The MMC H&H technical lead and modeler will discuss key information
like flow range, flow increments, and reach polygons with the district personnel. Tech-
nical decisions will be captured within a Project Work Plan. Primary data sources will
also be identified, such as existing hydraulic models, existing consequences models,
high-resolution terrain, reservoir locations, gauge locations, existing profiles, existing
inundation maps, and project datum conversions.
H&H Modeling The H&H modeler uses the best available models to first generate the high and low
flow bounding model runs. The project team reconvenes for to review the high and low
flow inundation polygons to minimize future rework. The goal of generating sample
flows is to evaluate and update the polygon extents and familiarize the modeler with
the existing hydraulic model. Once this is established, the team will review the data
and verify if the current path forward is acceptable. The SOP documents the process
for incorporating various model types such as steady flow, 1D unsteady flow, and 2D
flow models naming conventions, model validation, rating curve extraction, transfer
of data to the GIS and consequence team members, and addresses unique issues,
such as how to handle tidal flow situations.
GIS Data Several requirements must be met to seamlessly load to the final enterprise database.
Processing Inundation output must be properly processed to a GIS suitable format for database
import and for the web application.
Consequences An economist will first calibrate National Structure Inventory (NSI) data if not provided
with the existing consequences modeling. All flow scenarios will then be run through
the consequences model and the results imported into the Rapid Flood Response
Toolbox (RFRT) to approximate a curve using a linear piecewise approach.
Consequences Consequence data from the library of indexed inundation layers is stored into tables
Data Processing and uploaded into the database.
Web Interface Once all GIS data is processed and loaded, a web interface review takes place to
Review familiarize the district point of contact with the RIM tool and obtain feedback regarding
the data displayed in the web viewer.
Testing and After the data is loaded into the local file geodatabase, the final file geodatabase is
Enterprise reviewed through the test website. This is done through the ArcGIS Portal and/or
Data Loading ArcGIS Server. This allows the contents of the database to be displayed in the test
viewer and any database changes to be made prior to data loading into the enterprise
database. After all comments and changes are implemented, the data is loaded into
the enterprise Flood Inundation Mapping (FIM) database.
Documentation An appendix report documenting the MMC-RIM project development is completed to
accompany the CWMS final report. The report captures flow range, flow increments,
reach polygon delineation, and consequence data development for future reference.
Critical infrastructure and levee data can be visualized to determine quickly

identify which schools to evacuate and which levees may be impacted by flood
waters. Future development of the tool will allow for the query of external data layers
to return lists of key infrastructure intersecting with selected inundation stages within
minutes.
The MMC is working directly with the National Water Center (NWC) led by
the National Weather Service (NWS) to ensure one federal voice provides the best
201
Q.105 - R.13
Fig. 8
Observed Imagery vs Inundation from CWMS Model
Comparaison de l’imagerie observée avec l’inondation du modèle CWMS.
Fig. 9
Observed Imagery vs Inundation from CWMS Model
Comparaison de l’imagerie observée avec l’inondation du modèle CWMS.
available inundation mapping products to emergency managers, stakeholders, and

the public. The NWC will host the inundation layers on their servers where dam
owners can make the most appropriate forecast selection to be displayed on the
NWC viewer. The NWC will utilize the USACE data in conjunction with other data
202
Q.105 - R.13
sources to build a continuous, seamless inundation picture for the entire United
States.
By 2025, the MMC plans to have inundation layers incorporated into the MMC-
RIM tool for 201 basins, or 66,000 kilometers of river (41,000 miles). While the tool is
still in its infancy stage, it has received positive feedback from other agencies and has
been used in a handful of real-time flood events The MMC-RIM tool will continue
to evolve with each flood event, planning study, and interaction with emergency
management agencies.
REFERENCES
[1] State of North Carolina. Flood Inundation Mapping and Alert Network.
https://fiman.nc.gov/fiman/#
[2] U.S. Army Corps of Engineers (USACE) (2015) Bluestone Dam Safety
Modification Study, Hydrologic Engineering Branch, Engineering Division,
Huntington District, June 2015.
[3] http://www.lrh.usace.army.mil/Missions/Civil-Works/Current-Projects/Bluestone-
DSA/
203
Q.105 - R.14
-------
-------
CONTRIBUTION ON RESTORATION OD TAILINGS DAMS

DAMAGED AT INITIAL PERIOD OF CONSTRUCTION (∗)
Ljupcho PETKOVSKI
Full professor,
Chair of Hydraulic Structures, Ss CYRIL AND METHODIUS UNIVERSITY,
Civil Engineering Faculty – Skopje
Stevcho MITOVSKI
Associate Professor,
Chair of Hydraulic Structures, Ss CYRIL AND METHODIUS UNIVERSITY,
Civil Engineering Faculty – Skopje
MACEDONIA
SUMMARY
The similarities between the tailing dams and the embankment dams for water
storage have contributed a great number of procedures and techniques in the design,
construction and maintenance of the conventional dams, to be applied to tailings
dams. However, the numerous reports of collapses of the tailing dams in the last
three decades, all over the World, indicate that the safety in the phases of design
and construction were not controlled with the same rigor and carefulness - as for
the embankment dams. This fact, in part, results from the long-term construction
of the tailings dams, where as a building material is used sand obtained by sep-
arating of the waste material from floatation process during the exploitation of the
mine. In this paper are presented results from the research and insights of the
tailings dam Sasa 4 of mine Sasa, Makedonska Kamenica, Republic of North Mace-
donia, damaged in the initial period of construction in September, 2020, in order
∗ Contribution à la restauration des barrages de stériles endommagés à la construction
204 DOI: 10.1201/9781003211808-60

Q.105 - R.14
to: (1) detect the reasons for damaging, (2) determine urgent measures for provid-
ing of the required stability of the tailings dam at elevation 910.0 m asl, (3) to define
measures for landscaping of the surrounding terrain and upgrade of the structures
and the equipment for restarting of the tailings dam by current elevation of dam crest
at 910.0 m asl and (4) to foreseen measures for further safe service of the tailings
and construction of the tailings dam, from current elevation of 910.0 m asl till final
elevation at 952.0 m asl.
RÉSUMÉ
Les similitudes entre les barrages de stériles et les barrages en remblai pour le
stockage de l’eau ont contribué à un grand nombre de procédures et de techniques
dans la conception, la construction et l’entretien des barrages conventionnels, à
appliquer aux barrages de stériles. Cependant, les nombreuses ruptures de bar-
rages de stériles au cours des trois dernières décennies, partout dans le monde,
indiquent que la sécurité dans les phases de conception et de construction n’a
pas été contrôlée avec la même rigueur et la même prudence – que pour les bar-
rages en remblai. Ce fait résulte en partie de la construction à long terme de ces
barrages, où du sable obtenu en séparant les déchets du processus de flottation
pendant l’exploitation de la mine est souvent utilisé comme matériau de construc-
tion. Dans cet article sont présentés les résultats de recherche et perspectives du
barrage de résidus Sasa 4 de la mine Sasa, Makedonska Kamenica, République
de Macédoine du Nord, endommagé pendant la phase initiale de construction en
septembre 2020, afin de: (1) détecter les raisons des dommages, (2) déterminer
des mesures urgentes pour assurer la stabilité requise du barrage à une altitude de
910 m, (3) définir des mesures d’aménagement paysager du terrain environnant et
de modernisation des structures et de l’équipement pour le redémarrage du barrage
de stériles par l’élévation actuelle de la crête du barrage à 910,0 m d’altitude et (4)
aux mesures prévues pour une exploitation plus sûre et la construction du barrage
depuis l’élévation actuelle de 910,0 m d’altitude jusqu’à l’élévation finale à 952,0 m
d’altitude.
1. INTRODUCTION – SAFETY OF TAILINGS DAMS
The tailings dams are complex engineering structures, composed of starter

dam, sand dam, waste lagoon, drainage system, water conveyors for taking out of
the cleared water and structures for protection in case of inflow (external) water. The
tailings dams, on one hand, due to the numerous structures of which are composed,
should be inspected on great number of safety cases at static and dynamic load-
ings, similar as for conventional fill dams [1], [2], [3], and on other hand, due to the
enormous volume of the waste lagoon, they are fill structures with highest potential
hazard for the surrounding [4]. Due to the great importance of the tailings dams, one
205
Q.105 - R.14
Fig. 1
Location in Europe and map of Republic of North Macedonia
Situation en Europe et carte de la République de Macédoine du Nord.
of the ICOLD’s Technical Committees specifically deals for such hydraulic structures
- ICOLD Committee on Tailings dams and Waste Lagoons, that has published 10
Bulletins, [5], [6]. Due to the long construction period, the approach for conventional
dams (for creation of water reservoirs) for confirmation of proper accomplishment
of the hydraulic structures – with full supervision of the construction and control of
the first reservoir filling, as well and the assessment of the dam’s proper behaviour
with construction parameters throughout comparation with monitoring data, at most
cases is not applied fully in case of tailings dams. Unfortunately, such main differ-
ence between the conventional and tailings dams is amplified in case of technical
solutions with combined construction method [7], [8] and in case of heightening [9],
[10] thus providing increase of the deposit space of the tailings dams. The aims
of the research of the tailings dam Sasa 4 of mine Sasa, Makedonska Kamenica,
Republic of North Macedonia (fig. 1), damaged in the initial period of construction
in September, 2020 are: (1) to detect the reasons for damaging of the tailings dam,
occurred on 14 September, 2020; (2) to determine urgent measures for providing of
the required stability of the tailings dam at elevation 910.0 masl; (3) to define mea-
sures for landscaping of the surrounding terrain and upgrade of the structures and
the equipment for further service of the tailings with the required safety, for actual
crest elevation of the dam at 910.0 masl, (4) to foreseen measures for further safe
service of the tailings and construction of the tailings dam, from current elevation of
910.0 masl till final elevation at 952.0 masl.
2. TAILINGS DAM SASA 4 – BASIC PARAMETERS
In the past period of service of tailings dam of mine Sasa in M. Kamenica,

they were intended for deposition of the flotation tailings obtained with the
206
Q.105 - R.14
Fig. 2
Tailings dam and waste lagoon Sasa 4, February, 2020
Barrage de Tailing et lagune de déchets Sasa 4, février 2020.
technological process of flotation of minerals lead and zinc. It was transported by

hydraulic transport (gravity pulpline) all to the dam crest of active tailings no. 1, 2,
3-1, 3-2 and 4, from where by cycloning was separated in two fractions. The down-
stream dams were constructed by the coarser dry fraction (sand), while the fine
liquid fraction was deposited in the waste lagoons. According to the current state of
the tailings, tailings dam no. 4 is active, that will be created by construction of com-
bined tailings dam no. 4, by final crest elevation of 952.0 masl and waste lagoon
at maximal active level of 950.0 masl. During construction of the starter dam Sasa
no.4 (fig. 2), certain modifications took place (compared to the Basic design from
2014); (1) covering by geosynthetic impermeable lining of: (a) river bed upstream
of the starter dam, (b) valley abutments up to elevation 950.0 masl upstream of the
sand dam and (c) downstream slope of dam no. 3-2 below elevation 950.0 masl; (2)
replacement of the material in the starter dam (constructed up to elevation 905.0
masl), so that low permeable graphite shale was replaced by permeable tailings
mine rock, due to the plan for lining of the upstream slope with geomembrane (that
later was not installed). The appropriate mathematical model for as built state of
dam Sasa no. 4, at final stage is displayed on Fig. 3.
During service period of tailings dam Sasa no. 4, in period February-

September, 2020 was carried out technical monitoring for the tailings dam, the
appurtenant structures and waste lagoon, according appropriate Design for tailings
dam monitoring. The measured values from the monitoring process of tailings dam
Sasa no. 4, such as: surface displacements in the dam obtained by survey methods,
water levels in the piezometers, seepage discharge in the drainage system, quality
of the drained water, visual inspection of the dam and the surrounding terrain, were
systemized in appropriate monthly reports. Afterwards an assessment of the safety
of tailings dam Sasa no. 4 was done, by comparation of the measured values from
monitoring with calculated values by the mathematical models and permanently was
confirmed the global stability of the maximal cross section of the dam.
207
Q.105 - R.14
Fig. 3
Representative cross section of tailings dam no. 4, according to as built state,
model with geomembrane in the riverbed.
Coupe transversale représentative du barrage de résidus no. 4, selon l’état de
construction, modèle avec géomembrane dans le lit de la rivière.
0 Rock foundation, 0 Fondation rocheuse
1 Deposit (variable depth) 1 Gisement (profondeur variable)
2 Starter (initial) dam (crest width b = 5.0 m, 2 Barrage de départ (initial) (largeur
designed at elevation 906.0 m asl, de crête b = 5,0 m, conçu à 906,0 m,
constructed at 905.0 m asl) construit à 905,0 m)
3 Sand dam (final crest elevation 952.0 m 3 Barrage de sable (élévation finale
asl, crest width b = 5.0 m, profile axis de la crête 952,0 m, largeur de la
X = 100 m) crête b = 5,0 m, X = 100 m)
4 Waste lagoon (maximal water level 4 Lagune de déchets (niveau d’eau
950.0 m asl) maximal 950,0 m d’altitude)
5 Downstream embankment of mine rock 5 Remblai en aval de la roche minière
(crest elevation at 917.0 m asl) (élévation de la crête 917)
3. DAMAGING OF THE TAILINGS DAM SASA NO. 4 IN

INITIAL PERIOD OF CONSTRUCTION ON 14.9.2020
According to eye witness of the event, an employee in mine Sasa, morning on

14.9.2020, approximately at 05h45 started process of penetration of water from the
waste lagoon at crest of sand dam Sasa no. 4 (constructed till elevation 910.0 masl)
in nearby of the right valley bank. Rapidly, the water penetration became progressive,
the tailings sand from the downstream slope in nearby of the right bank was taken
away (fig. 4), the material became loose and the material from the upper zone settled,
that enabled creation of opening in the zone of the maximal water level in the lagoon,
till crest of the sand dam. Through the opening was discharged approximately 5,000
m3 of water from the waste lagoon, that was above the tailings silt in the lagoon (fig.
5) and due to the denivelation from the upper to lower water (approximately 23
m), the discharge developed high cinematic energy. Such energy was released by
significant erosion in the downstream part of the tailings dam (fig. 6). The height
of the opening from the crest to the dam toe is H = 910.0 − 885.0 = 25.0 m, bottom
width of approximately 2.0 m, and in the crest 10.0 m, by average depth (transverse
208
Q.105 - R.14
Fig. 4
Damage in the right bank of the tailings dam, state on 14.9.2020
Dommages en rive droite du barrage à résidus, état au 14.9.2020.
Fig. 5
Opening at dam crest where the discharge of the stored waste water took place
Ouverture à la crête du barrage où a eu lieu l’évacuation des eaux usées stockées.
to the original slope with inclination 1:2.7) of approximately 5.0 m. It is estimated

that it was discharged approximately 3,000 m3 tailings sand from the dam body.
The mixture of water from the waste lagoon that leaked uncontrollably through
the breach point in the upper zone in the right bank of the tailings dam and tailings
sand of the downstream part of the dam has propagated downstream of the tailings
209
Q.105 - R.14
Fig. 6
Tailings sand that was taken away from the downstream body of the dam
Sable de résidus qui a été enlevé du corps en aval du barrage.
Fig. 7
Stopped erosion of the slope of tailings dam Sasa no. 4 on 14.9.2020
Arrêt de l’érosion de la pente du barrage de résidus Sasa no. 4 le 14.9.2020.
dam toe. Such mixture of water from the technological process (deposited in the
lake) of approximately 8,000 m3, along with pure water downstream of the terminal
structure of the diversion tunnel, has transported, spreaded and deposited on large
section of Kamenicka river. As soon as the water level has lowered and equaled with
elevation of sufficient length of the beach from tailings mud, directly upstream from
the dam crest, the process of further impermissible seepage through the tailings
dam and surface flow along the downstream zone of the dam has stopped, that
enabled also and the dam erosion to be stopped (fig. 7).
210
Q.105 - R.14
4. REASONS FOR DAMAGING OF TAILINGS DAM SASA NO. 4
According to the event description from the staff of mine Sasa and terrain
inspection conducted on 14.9.2020m it can be stated that there is no case of human
error, equipment malfunction or systematic incorrect operation with tailings dam
Sasa no. 4. It is obvious that tailings dam damaging is not caused by excessive
inflow and overtopping of dam crest (at current elevation 910.0 masl) nor due to
insufficient heightening from maximal water level to the dam crest. Namely, such
heightening was constantly maintained above 2.0 m, as designed. At the same time,
the damaging did not occur due to insufficient casual seepage strength or endan-
germent of the dam stability due to irregularly constructed geometry of the cross
section. Confirmation for such statement is (1) geometry of the constructed part of
the sand dam is according to the design documentation, by which is confirmed the
static stability and (2) measured seepage pore pressures in the installed piezome-
ters are in accordance with the expected values, and by regular visual monitoring of
the drainage system there was no detected occurrence of turbidity of the seepage
water in case of eventual process of mechanical erosion (suffusion).
We are on opinion that initial leakage of water in zone of the dam crest in nearby
of the right bank of valley is case of impermissible seepage. It cannot be described
by established models for stationery and non-stationery seepage, based on princi-
ples of mechanics, but it can be explained by engineering logic. The impermissible
seepage consists of paths of concentrated seepage that in case of progressive
type can endanger stability of some parts or of the full embankment structure.
It is caused by incidental occurrence of loose porous medium exposed at active
action of the flow, due to design or construction errors or due to unfavorable natural
ambient.
In this case, the occurrence of impermissible seepage is caused by super-

position by number if influences, that according to the contribution can be divided
in primary (significant) and secondary (ancillary). Primary for occurrence of the
impermissible seepage are:
(1) Unfavorable hydraulic contact of the tailings sand (from the dam shell) with
coefficient of permeability k = 2 × 10−6 m/s and geomembrane with coefficient
of permeability k = 10−12 m/s. It is a case of hydraulic contact of two materials
that extremely differs by its permeability. This led to fully redirecting of the flow
lines towards the les permeable material – tailings sand and it created initial
contact seepage.
(2) Insufficient length of the beach from tailings mud, upstream of the upstream
edge of the dam crest that conditioned close vicinity of the water from the
waste lagoon to the dam sand. Such vicinity does not result with significant
loss of water head of the seepage flow before the contact with the sand dam
apropos it had maximal head along the exactly along the upstream slope of
the sand dam.
211
Q.105 - R.14
Secondary reasons for occurrence of impermissible seepage are:
(1) Unfavorable natural configuration of the terrain above the dam crest, that in
case of intensive rainfalls creates concentrated surface flow, directed exactly
towards the dam crest or directly downstream of the dam. Such surface flow
additionally weakens the contact of the dam with valley bank.
(2) Unfavorable mechanical contact of the tailings sand with geomembrane apro-
pos smooth layer between the two materials with extremely different stiffness
and reduced friction along the contact that facilitated downstream displace-
ment of the sand grains in direction of the active force of the seepage flow.
(3) Great natural slope in the right bank of the valley, that conditioned “hanging” of
the soft tailings material on the stiff rock foundation, that is manifested by unfa-
vorable transfer of stresses and occurrence of loose (insufficiently compacted)
earth, through which is eased hydraulic breach if exposed to seepage flow.
(3) Unfavorable natural configuration of the terrain by widening of the valley down-
stream of the sand dam and eased displacement in downstream direction at
action of the hydrostatic pressure that conditioned the tailings sand not to be
pressured in the contact with the rock in the bank but to be loosened, that
additionally worsens the seepage through the dam.
It is stated that the dimensions of the created crater in the right bank of the
dam, by variable width of 2m in the bottom up to 10 m in the crest, by average depth
of approximately 5m, cannot endanger global safety of the embankment structure.
Namely, it has crest length of approximately 100m, and the central section, with
maximal height and it is of key importance for the global stability, and is at sufficient
distance from the crater. But the crater muse be filled with ballast material as soon
as possible, so that can be prevented its further expanding and widening.
5. URGENT MEASURES FOR PROVIDING OD REQUIRED STABILITY

OF THE TAILINGS DAM SASA NO. 4, AT ELEVATION 910.0 MASL
The first stage of the damaging restoration of the tailings dam Sasa no. 4
regards the dam stability. Such stage is urgent and should be used current favorable
weather conditions (without rainfalls) and the taken material in the right bank of
the dam (tailings sand) to be replaced by appropriate local material, for which we
emphasize following recommendations:
(1) The material should be non-coherent and by higher or equal strength and
permeability, compared to the tailings sand. Most rational is to use tailings
mine rock, available in mine Sasa and it will be used for the downstream prism
of this combined dam (till elevation 917.0 masl), that is sufficiently investigated
as material and by known geomechanical parameters.
(2) The material will be spreaded in horizontal layers by height of 50 cm, from the
breach bottom (approximately at 885.0 masl) till starter dam crest at elevation
212
Q.105 - R.14
906.0 masl (or till elevation according the as built drawing). The material should
be sufficiently compacted by available machinery of mine Sasa.
(3) During the spreading and compaction in layers of the mine rock, special care
should be taken for employees safety and firstly the steep surfaces of the
tailings sand (that are almost vertical), created by the accident, should be
soften. Such vertical surfaces we believe that are in state of limit equilibrium
and any moment (by small variation of the humidity) is possible occurrence of
local instability and sliding on the steep slope.
(4) During placement of the mine waste rock the continuity from the left to the right
side of the created crater should be preserved, and by scale advancement
in height. The greatest part from the dam opening should be filled with such
material, so that till the final shaping of the geometry to be left approximately 1.0
m of thickness, that will be filled by tailings sand, by placement in sloped layers
from the dam crest (as soon as the use of the tailings dam has commenced).
(5) Above elevation of the starter dam (according to as built drawing) in horizontal
layers should be embedded tailings sand, by moderate compaction in the final
10 m with the contact in the valley right bank. Such embankment should be
by crest elevation 910.0 masl, crest width 5.0. m, dispositioned in layout so
that upstream slope of the sand dam is formed by slope 1:1.5, and on the
downstream face slope less than 1:2 (in order to have satisfying temporary
stability). Required downstream slope of 1:2.7 m in order to permanent safety
to be provided, will be filled with tailings sand, by deposition in inclined layers
from the dam crest (as soon as the use of the tailings dam has commenced).
(6) The tailings sand should be borrowed from the downstream slope of dam Sasa
no. 4m, in section less than 10 m, that will be filled instantly by mine rock or
from the crest of dam Sasa no. 4, that has great width (approximately 30 m),
in zone of the downstream edge, by previously temporary removal of eventual
layers of mine rock.
(7) At contact of the embankment from tailings sand and the right bank the three-
layer geosynthetic lining to be removed (at least 2 m upstream from the contact
of the sand dam with the terrain, and in zone of crest support (width of 5 m) the
local material should be removed, at depth of 0.5 m or at level of rock founda-
tion. Namely, such material was placed in order the unevenness to be leveled
for installation of the geosynthetic lining. Such material is extremely permeable
and it must be removed in trench by approximate dimensions of 5 m (in width
as the dam crest width) and 0.5 m (depth in permeable local material or till
rock foundation in the bank), at the contact of the dam crest with the terrain.
6. NECESSARY MEASUERS FOR PREPARATION OF THE TERRAIN,

STRUCTURES AND EQUIPMENT FOR RESTARTING OF
THE SERVICE OF TAILINGS DAM SASA NO. 4
The second stage of the restoration of the damaging of tailings dam Sasa no.
4 regards preparation of the terrain, civil engineering structures and equipment for
213
Q.105 - R.14
restarting of the technological process and resumption with use of the tailings. The
second stage from the restoration should take place upon completion of the first
stage, and the necessary measures regards two aspects: hydraulic-civil engineer-
ing and mining-technological process. The measures from aspect of hydraulic-civil
engineering are:
(1) If the tailings sand would rely on the diluvium, the natural slope material along
the valley bank, then specified hydraulic and mechanic diversity would be
reduced to a minimum, and due to the natural unevenness the contact would
have higher passive friction force. Therefore, in the further use of the tailings
dam and advancement from the current dam elevation of 910.0 masl till final
elevation of 952.0 masl, the construction to be according to design documen-
tation, prepared by recommendations for lining from the Commission form
Ministry of Environment and Physical Planning. So, it is necessary removal of
the geosynthetic lining on contact between tailings dam and natural ground.
After that the surface material from organic substances (trees, roots, rock and
etc.) in zone of the dam crest should be cleared out.
(2) Common practice in Dam engineering, upon serious damage of the dam body
(as in case of damaging during construction of tailings dam Sasa no. 4) is to
identify and apply additional measures for providing of greater stability. There-
fore, the unfavorable hydraulic contact of the tailings mud (from the beach,
upstream of the sand dam) and geotextile on which he is relied on should be
taken in consideration. The geotextile is drainage material with much higher
seepage coefficient, compared with tailings mud from the beach. It is a thin
water conveyor between two isolators, geomembrane from below and tailings
mud from above. Of course, the geotextile in this zone does not affect on the
central seepage (in maximal cross section), that has dominant influence on the
dam’s global stability. But, the geotextile significantly affects the lateral seep-
age due to the reduction of the head of the lateral seepage flow through the
beach from the tailings mud, in direction of the lagoon with waste water towards
the sand dam and it can cause impermissible contact seepage. Therefore, we
recommend that beside removal of the geo-synthetic lining on contact of tail-
ings dam and natural ground, additional removal (upstream of such contact for
2 m) of the geotextile and the geomembrane, and only the geo-synthetic clay
liner to be kept (that is easily adaptable on unevenness). Also, we recommend
removal of at least 30 m upstream of the crest, as the beach from tailings mud
is planned. Namely, the geotextile in such zone has no any useful function,
because it has nothing to protect from the geomembrane due to filling of the
waste lagoon with mud and waste water and not with rock-fill and concrete
slabs with sharp edges.
(3) Intake pipe Ø200 mm to be installed through the steel shield, upstream at
around 5 m of the joint of the spillway collector D = 2.0 m (horizontal tunnel
according to the as built drawing) with the diversion tunnel of Kamenicka River.
The steel shield to be constructed with height of 1.0 m, relied on steel sloped
element (buttress) from the downstream side. The joints to be filled with per-
manently elastic putty. The sealing with the specified putty to be from the
214
Q.105 - R.14
upstream and the downstream side, and also at all joints apropos joint of steel
with steel and steel with existing concrete lining.
(4) The upper edge of the intake pipe to be installed on the steel shield at height
0.2 m from the upper edge of the shield. Downstream of the intake pipe to be
installed pipeline in the internal part of the tunnel, by leveling that will always
be lower than the leveling of the opening in the steel partition. The pipeline
will convey spillway water all to the existing reservoir (or previously to new
temporary settling tank) by permanent pump station (for recirculation of the
drained water for the need of the technological process). At the downstream
ending of the pipeline two gates are to be installed, that will regulate spillway
outflow whether it should be released in the river or in the settling tank and then
in the reservoir. So, if it is estimated that the existing reservoir is with insufficient
volume, in such case the water should be kept in the temporary settling tank.
(5) The location of the temporary settling tank should met following criteria: (a) not
to obstruct construction of the downstream part of the combined tailings dam
Sasa no. 4, (b) to be located in nearby of the exit structure of the diversion
tunnel (in the right bank of the valley), (c) to be located in nearby of the existing
reservoir, but on higher elevation, so the water can flow by gravity from the
settling tank in to the reservoir, (d) to have eased access for water tank for
occasional cleaning of the settling tank, (e) to be upstream from the terminal
structure of the channel for Petrova Reka (in the left bank of the valley) and (f)
the construction period of the settling tank to be reduced to minimum.
(6) One of the reasons for the impermissible seepage on 14.9.2020 and damaging
in the right bank of tailings dam Sasa no. 4 is unfavorable natural configuration
of the terrain above sand dam crest, that in case of intensive rainfalls creates
concentrated surface flow, directed exactly towards the dam crest or directly
downstream of the dam. Such surface flow additionally weakens contact of the
dam with the valley bank and contributes for erosion of the deposited cycloned
tailings sand. Therefore in the further course of the construction of the tailings
dam it is necessary to be captured external surface water (occurred from inten-
sive rainfalls), that gravitate towards the crest and the downstream slope of
the dam. The tailings dam Sasa no. 4 is earth dam, but it is built in stages,
simultaneously with filling of the waste lagoon apropos construction duration is
equated with service of tailings dam. Therefore in case of tailings dams there
is no possibility to be constructed durable protection channels in service stage
of waste lagoon but, if required, are constructed temporary channels (by exca-
vation in the valley banks), that are appropriate for some stage of construction
of the tailings dam and method of construction advancement.
7. REQUIRED MEASURES FOR FURTHER SAFE SERVICE OF

TAILINGS DAM SASA NO. 4 AND CONSTRUCTION OF
TAILINS DAM FROM ELEVATION 910.0 MASL TO 952.0 MASL
Upon completion of the measures from the second stage of restoration of

tailings dam Sasa no. 4, with which preconditions are acquired for restarting of the
215
Q.105 - R.14
technological process for use of the tailings and further construction of tailings dam
from elevation 910.0 m asl till final elevation of 952.0 m asl, in order to be provided
satisfying safety level, we recommend required measures that regards two aspects
hydraulic civil engineering and mining technological. The measures form hydraulic
civil engineering aspects are the following:
(1) As it is progressed in height with the dam crest simultaneously to be removed

geo-synthetic lining on contact of the tailings dam and the natural ground. After-
wards, to be cleared out surface material of organic substances (trees, roots,
humus, rocks etc.) in dam crest zone and to be removed surface degraded
material in depth of at least 50 cm.
(2) If the future technological process is realized by occasional outflow of non-
cycloned tailings (by spigots) for creation of beach with slope 1:15 from the
upstream edge of the dam in direction of waste lagoon and if there is no
sufficient quantity of tailings sand, in such case in the crest should be placed
layer of mine rock. The horizontal layer of mine rock to be with height of 50 cm
by track transport and by moderate compaction with the available machinery
of mine Sasa. In such case, due to the increased needs for maneuver for the
track transport, the crest width to be increased to 6.0 m
(3) If the future technological process is realized only by cycloning, by increase
of the number of cyclones from the current two to three cyclones, deployed
on dam crest, in such case the water to be distanced from the crest at least
28.5 m, with estimation that slope of the beach from the cycloned mud is 1◦ or
1:57, the discharge of the overflow of the cyclones should be less than 1.5 m
below the crest.
(4) The protective height in the lowermost part of the dam crest till the water level
in the lagoon to be at least 2.0 m.
(5) In case when there is outflow of cleared water through the spillway openings,
then the water in the pipeline downstream of the steel shield will be discharged
directly in to the river (open only gate 1). In case of outflow of turbid water
through the spillway openings (distributed in the collector of the right bank of the
valley), in order to be provided required protective height from the water level till
the dam crest, in such case the water will be discharged in the settling tank and
afterwards in the reservoir (open only gate 2). Upon reaching the maximal level
in the reservoir, the water will be recirculated in to the technological process
(sprinklers) by pump station.
(6) The downstream body from mine rock, designed with final crest elevation of
917.0 masl, is to be constructed in 4 stages. With the stage construction, with
crest width of 3.0 m and upstream slope 1:1.5, will be provided two functions.
First, support of the inclined layers from the deposited tailings sand and sec-
ond, providing of reserve volume. So, if in further period there is eventual taken
out tailings material downstream of the dam site, in that case it will be retained
and it will not propagate in the downstream river bed of Kamenicka River.
(7) The terrain in the zone above the crest should be inspected and improved
permanently and always to be shaped so that eventual concentrated surface
flow along the slope (in case of intensive rainfalls) to be redirected downstream
of the downstream toe of the sand dam. If such is not possible, then to be
216
Q.105 - R.14
redirected towards the waste lagoon, at least 30 m upstream of the dam crest.
However, it is not to be allowed ever again concentrated surface flow from the
valley banks to be directed towards the crest and towards the downstream
slope of the dam.
(8) The tailings sand along the crest surface of the sand dam to be compacted
by the available machinery of mine Sasa, on sections with length of mini-
mum 10 m, in nearby of the right and the left bank of the valley, that will
reduce the eventual looseness of the deposited tailings sand and it will improve
its seepage strength in critical zone for occurrence of impermissible contact
seepage.
(9) The key for stability of the tailings dam is regular operation of the drainage
system that consists of 4 drainage blankets embedded transversally on the
valley. They are constructed in the river bed from filter material and lined with
geotextile, in order to be prevented eventual plugging of the drainage material.
Upon the accident, the upper geotextile of drainage blankets no. 3 and 4
is covered by sediment and it is plugged apropos it is not able to operate.
Therefore, the upper part of the geotextile should be removed and replaced
with new layer with same properties.
(10) On daily basis to be inspected occurrences of: (a) humidity along the surface
of the downstream slope, (b) the surface craters, from eventual settlement of
the material due to mechanical erosion in the dipper layers, (c) turbid water in
the drainage system, as result of unfavorable seepage process and in case
of such registered occurrences the tailings dam service should be stopped
temporary.
(11) The future planning and designing of the protective channels, their construction
apropos excavation in the valley banks, should fit the specific terrain conditions
that will suit for the next state of construction of the tailings dam. Such tem-
porary structures belong in maintenance of the hydraulic structures by the
operator in service period of the tailings dam. We underline that in future are
possible occasional erosion along the surface of the embankment structure, if
in case of more intensive rainfalls are is created surface flow that will exceed
the capacity of the protective channels. It is a case in the service period at per-
manent protective channels of conventional earth dams with water reservoir,
so of course it is possible and for temporary protective channels at tailings
dams with waste lagoons. Therefore there is visual auscultation during the
service period of the dams, because they are civil engineering structures with
active influence of the water. What matters is that the small surface damages
should be timely detected and restored as soon as possible, that is obligation
of the dam operator.
8. CONCLUSIONS
The unfavourable contact seepage in upper zone of the right bank of the dam
conditioned progressive seepage in the right bank of the tailings dam Sasa no. 4,
217
Q.105 - R.14
in initial period of construction, at elevation 910.0 masl in September, 2020. At

action of the progressive seepage was developed process of taking out of sand
grains from the dam body, loosening of the material and settlements under the
loading of the upper layers. Cumulative settlement was maximal at the dam crest.
As soon as the crest was on lower elevation to the water level in the lagoon, there
was uncontrolled out flow of the stored water. The maximal kinetic energy was in
the nearby of the downstream toe of the dam and caused material erosion and
creation of crater. Afterwards the crater was deepened and got to the dam crest but
by lowering of the water level till elevation of the tailings mud the surface flow and
crater widening was stopped. Having in consideration the reasons for occurrence of
unfavourable contact seepage and the scope of damaging of tailings dam Sasa no.
4, were adopted measures for restoration in two stages (urgent for providing dam
stability and preparation for restarting of the tailings dam) and measures for further
safe use of the tailings and construction of the tailings dam. The tailings dam Sasa
no. 4 must not be in service before all measures from both stages of restoration are
not completed in full.
REFERENCES
[1] PETKOVSKI L., TANČEV L., MITOVSKI S. A Contribution to the standardiza-

tion of the modern approach to assessment of structural safety of embankment
dams", 75th ICOLD Annual Meeting, International Symposium “Dam Safety
Management, Role of State, Private Companies and Public in Designing,
Constructing and Operation of Large Dams”, St.Petersbourg, Russia, 2007.
[2] PETKOVSKI L., ILIEVSKA F. Estimation of hydro tailings’s dynamic response

during the influence of strong earthquakes, III Symposium of MAG, Struga,
2010.
[3] PETKOVSKI L., ILIEVSKA F. Comparison of Different Advanced Methods for

Determination of Permanent Displacements of Tailings Dams in Earthquake
Condition“, 14th Europian Conference on Earthquake Engineering, Ohrid, R.
Macedonia, 2010.
[4] PETKOVSKI L., MITOVSKI S. Creating of tailings space by stage construction

of rockfill dam, 25. Congress on Large Dams, ICOLD, Stavanger, Norway,
2015.
[5] ICOLD, Bulletin 45, Manual on tailings dams and dumps, 1982.
[6] ICOLD, Bulletin 139, Improving tailings dam safety – Critical aspects of
management, design, operation and closure, 2011.
[7] PETKOVSKI L. Modified conventional dam in phase construction – an alterna-

tive to tailings dam, UDK: 627.821, Original scientific paper, Watereconomy,
ISSN 0350-0519, Vol. 47, No. 273–275, p. 41–50, Beograd, Serbia, 2015.
218
Q.105 - R.14
[8] PETKOVSKI L., MITOVSKI S. Comparison Of Tailings Dams Dynamic

Response In Case Of Central And Downstream Method Of Construction,
82. Annual Meeting of ICOLD, International Symposium: Dams in Global
Environmental Challenges, Bali, Indonesia, 2014.
[9] PETKOVSKI L., MITOVSKI S. Structural safety of tailings dams at increasing

of the deposit space, Conference, Talinigs dams in Republic of Macedonia,
MACOLD, Shtip, R.Macedonia, 2012.
[10] PETKOVSKI L. Static and dynamic stability of tailings dams at heightening“,

Journal Pressing, Volume 2, no.12, ISSN 1857-7-44X, 2012.
219
Q.105 - R.15
-------
-------
LIQUEFACTION AND DAM SAFETY – LESSONS FROM

EVENTS AT EMPINGHAM, FEIJAO AND FUNDAO (∗)
Rodney BRIDLE
Civil Engineer,
DAM SAFETY LTD
UNITED KINGDOM
SUMMARY
Parallels between events at Empingham Dam and the failure and liquefaction
of Fundao and Feijao tailings dams have provided insights into the behaviour of such
dams, as follows:
The stability of ‘upstream’ construction tailings dams depends on a finely

balanced continuous ‘cycle’ of stress change and pore pressure drawdown as con-
struction proceeds. The drawdown is caused by high shear stresses deep in the
dam which cause dilatancy and low pressure at depth. The low pore pressures gen-
erated by this ‘cycle’ keep the dams stable as load increases during deposition of
tailings. The ‘cycle’ is broken by sudden events, most likely earthquakes, that raise
pore pressures throughout the dam and failure follows.
Liquefaction on failure may be explained because the sudden increase in pore

pressure would have caused the failure, not the disturbing force from the retained
tailings. The disturbing forces would then be applied only to the failing mass of
saturated tailings, raising pore pressures further to cause fracture, bursting and
liquefaction.
∗ Liquéfaction et sécurité des barrages – leçons tirées des événements d’Empingham, Feijao
et Fundao
220 DOI: 10.1201/9781003211808-61

Q.105 - R.15
A simple effective stress analysis using modelled pore pressures from the
Feijao report demonstrated that the dams may be safer than previously thought.
They would be stable while deposition of tailings continued, and failure is most likely
to occur as a consequence of sudden events that break the cycle by suddenly raising
pore pressures throughout the dam. Such events are most likely to be earth tremors,
but by infiltration by very heavy rainfall in the case of Feijao, a ‘retired’ dam, no longer
used for tailings disposal.
Pore pressures respond to stresses and stress changes in all directions in a

dam. Measuring pore pressures directly by piezometers installed in the dams would
make it possible to check the stability of the dam routinely, to establish if stability
was improving or worsening with height – whether the ‘cycle’ would eventually fail
to reduce pore pressures sufficiently to maintain stability and whether a modest or
a substantial earthquake would cause failure – from which decisions on the future
of the dam and the urgency and extent of precautions to protect downstream com-
munities can be made. On-site teams of embankment dam engineers could install
piezometers, carry out the routine analyses, prepare emergency plans and perform
the many other duties on which the safety of the dam depends, as it was raised.
Embankment dam engineering, including both mine dams and conventional

earth dams, is a specialism within dam engineering, applying soil mechanics to the
engineering of water-retaining embankment dams. Closer integration between tail-
ings embankment dam engineers and earth and rock embankment dam engineers,
with opportunities to work on both types, could lead to better practice and better
understanding of the safety of all water and liquid retaining embankment dams. The
ICOLD Technical Committees on Embankment Dams and on Tailings Dams may be
able to take a lead on this.
RÉSUMÉ
Des parallèles entre les événements survenus au barrage d’Empingham et la

rupture et la liquéfaction des barrages de stériles miniers de Fundao et de Feijao
ont permis de mieux comprendre le comportement de ces barrages, comme suit:
La stabilité des barrages de stériles miniers de type «en amont» dépend d’un
équilibre entre les variations de contraintes et de pressions interstitielles durant la
progression de la construction. Les diminutions de pressions interstitielles sont dues
aux contraintes de cisaillement élevées en profondeur dans le barrage qui causent
la dilatation et les diminutions conséquentes des pressions. Les faibles pressions
interstitielles ainsi générées maintiennent les barrages stables à mesure que la
charge augmente pendant la déposition des stériles. Le «cycle» est brisé par des
événements soudains, plus souvent des secousses sismiques, qui augmentent les
pressions interstitielles au droit du barrage qui peuvent mener à sa défaillance.
La liquéfaction peut être expliquée par l’augmentation soudaine des pres-

sions interstitielles qui auraient causé la défaillance, et non par l’action des forces
221
Q.105 - R.15
déstabilisatrices des stériles retenus. Celles-ci ne s’appliqueraient alors qu’à

la masse déstabilisée des stériles saturés, ce qui augmenterait les pressions
interstitielles pour provoquer des fractures, des éclatements et la liquéfaction.
Une simple analyse en contraintes effectives à l’aide des pressions intersti-

tielles modélisées du rapport d’analyse sur le barrage Feijao a démontré que les
barrages auraient pu être plus sûrs qu’on ne l’aurait pensé. Ils auraient été stables
durant la progression de la construction. La rupture aurait la plus susceptible de
survenir à la suite d’événements soudains ayant rompu le cycle en augmentant
soudainement les pressions interstitielles dans tout le barrage. De tels événe-
ments seraient le plus souvent des tremblements de terre mais pourraient être aussi
l’infiltration par des très fortes pluies comme dans le cas de Feijao, un barrage «en
inactivité» qui n’est plus utilisé pour le stockage des stériles.
Les pressions interstitielles réagissent aux contraintes et aux changements

de contraintes dans toutes les directions dans un barrage. La mesure directe des
pressions interstitielles par des piézomètres installés dans les barrages permettrait
de vérifier leur stabilité de manière routinière; d’établir si la stabilité s’améliore ou
diminue avec la hauteur et l’accumulation de stériles – si le «cycle» ne réduisaient
pas suffisamment les pressions interstitielles pour maintenir la stabilité et si une
secousse sismique modeste ou substantielle provoquait des défaillances – à partir
desquelles des décisions sur l’avenir du barrage et l’urgence et l’étendue des pré-
cautions à prendre pour protéger les communautés en aval pourraient être prises.
Des équipes d’ingénieurs en barrages en remblais localisés sur place pourraient
installer des piézomètres, effectuer les analyses de routine, préparer des plans
d’urgence et effectuer les nombreuses autres tâches dont dépend la sécurité du
barrage, au fur et à mesure de la progression de la construction.
L’ingénierie des barrages de remblai, y compris les barrages miniers et les bar-
rages en remblais conventionnels, est une spécialité de l’ingénierie des barrages
qui applique la mécanique des sols à l’ingénierie des barrages de remblai de réten-
tion d’eau. Une intégration plus étroite entre les ingénieurs des barrages de remblai
de stériles et les ingénieurs de barrages de remblai en terre et en enrochements,
avec des occasions de travailler sur les deux types, pourrait mener à une meilleure
pratique et à une meilleure compréhension de la sécurité de tous les barrages de
remblai de retenue d’eau et de liquide. Les comités techniques de la CIGB sur les
barrages en remblais et sur les barrages de stériles pourraient être en mesure d’en
prendre l’initiative.
1. INTRODUCTION
Liquefaction failures of dams and the loss of lives that such failures have
caused undermine the principal objective of ICOLD (the International Commission
on Large Dams) – to make all dams safe. This has recently been re-stated in the
World Declaration on Dam Safety[1] .
222
Q.105 - R.15
The liquefaction failures and fatalities at Fundao[2] and Feijao[3] tailings dams
prompted the thought that the safety of tailings dams might be improved by find-
ing parallels with conventional embankment dams. Question 105 for the ICOLD
Congress in 2021 on ‘Incidents and Accidents concerning dams’ provides a vehicle
for papers on this topic.
At a British Dam Society presentation Cambridge and Shaw[4] noted that the
failure of Feijao (Brumadinho) mine dam occurred three years after completion of
‘construction’ – placement of tailings.
There was a similar delayed deformation at Empingham dam[5] , fortunately

not resulting in liquefaction. On further examination, other unexpected similarities
between the behaviour of the clay dam at Empingham and the ‘upstream construc-
tion’ mine tailings dams at Fundao and Feijao were found. This report describes
these similarities which lead to possible explanations of the two mine dam failures
and to recommendations on engineering the safety of similar dams in future.
2. THE EMPINGHAM EVENT
Empingham[5] dam is a clay fill dam on clay foundations; it retains Rutland

Water. Local extension of the foundation below the crest of the dam continued
for nine years after completion of construction before stopping, as shown in Fig.1.
This deformation was ‘hidden’ in the interior of the dam, detected by purpose-built
monitoring.
Fig.1(a) shows the monitoring plate positions, the deformations after construc-
tion and the piezometer records. Fig.1(b) shows that after construction there was an
extension under the crest (plates 2-1) of about 200 mm. Under the slope (plates 3-2)
very little deformation occurred. Near the toe (plates D-3) there was compression of
about 100 mm. Note that the 48-month readings and 108-month readings should be
transposed[6] , but show that pore pressures were highest at the end of construction
and slowly declined over the years afterwards.
Later, a finite-element strain softening analysis[7] demonstrated that the defor-

mation resulted from local yield (Fig.2). The yield zone was part of an incipient shear
failure plane that did not develop further. It had been forming slowly as pore pres-
sures responded to the strains generated by the high local shear stresses caused
by the non-uniform base load below a ‘triangular’ dam[8] (Fig.3).
3. EVENTS AT FEIJAO AND FUNDAO
Feijao[3] and Fundao[2] dams were iron ore tailings dams constructed by the
‘upstream’ method, tailings transported in water in pipes and discharged behind
223
Q.105 - R.15
Fig. 1
Empingham Dam[5] – Barrage d’Empingham[5]
(a) Monitoring devices and deformations over 9 years after construction
(a) Dispositifs de surveillance et déformations plus de 9 ans après la construction
(b) Deformations: extension under crest (2-1), compression near toe (D-3)
b) Déformations: extension sous la crête (2-1), compression près du pied (D-3).
construction berms added to retain the wet tailings as the dam is raised. The
coarser tailings particles (‘sands’) are deposited near the berms and gradually form
the tailings dam. The tailings are deposited in water and are saturated. Initially the
tailings are wet and almost fluid, resulting in breakouts on to the slopes. As more
224
Q.105 - R.15
Fig. 2
Local yield in foundation of Empingham dam resulting from non-uniform load on
foundation (from presentation[7] )
Plasticité locale dans la fondation du barrage d’Empingham résultant de la charge
non uniforme sur la fondation (d’après la présentation[7] ).
Fig. 3
Contours of maximum shear stress below a dam obtained using finite difference
analysis assuming dam is elastic[8] .
Isovaleurs de contrainte de cisaillement maximale sous un barrage d’après une
analyse en différences finies en supposant le barrage élastique [8] .
tailings are placed the non-uniform loading imposes varying stresses on the early
layers (as in Fig.3). To convert this loading into a shear plane requires expansion
under the high part of the dam and contraction near the toe as at Empingham
(Fig.1). This requires pore volume changes, dilatancy, which in incompressible (and
non-expandable) water and solid particles, causes marked pore pressure changes.
Fig.4 shows the pattern of predicted displacements at Fundao dam. It would

result from a shear stresses distribution similar to that in Fig.3, caused by the
non-uniform loading imposed on the tailings low down in the Fundao dam. This
was the situation shortly before failure (i.e. when Mohr Coulomb Factor of Safety
was approaching unity). About 0.5 m downstream displacement was predicted. To
achieve such substantial deformations, pore expansion would have to occur, not pos-
sible in incompressible water and tailings particles[9,10] . This causes dilation, which
reduces pore pressures locally and sets up gradients causing very small seepage
flows towards the high stress-strain zone and drawing pore pressures down in the
overlying embankment. Although not obvious in Fig.4, the strains near the toe of the
embankment are compressive causing contraction, driving pore pressures up. This
corresponds to the compression measured near the toe at Empingham (Fig.1b).
A similar situation existed before failure at Feijao (Brumadinho)[3] tailings dam,

shown in Fig.5. These predicted deformations have been calculated from creep
225
Q.105 - R.15
Fig. 4
Displacement contours on incipient shear plane at Fundao dam (Figure 6-9[2] )
Isovaleurs de déplacement horizontal sur le plan de cisaillement naissant au
barrage de Fundao (d’après la figure 6-9[2] ).
Fig. 5
Predicted creep deformations at low level in Feijao tailings dam (From Figure 51[3] )
Déplacements horizontaux de fluage prévus à un niveau bas dans le barrage des
stériles de Feijao (À partir de la figure 51[3] ).
assessed from shear stress levels, so the pattern is similar to those at Fundao
and at Empingham. Very thorough searches of all data were made but revealed
no evidence of any deformations of the Feijao dam during the three years of its
‘retirement’ after deposition of tailings ceased. Detectable deformation of saturated
fill can only occur if water escapes from it by drainage. The creep approach, which
was also tried at Empingham[5] , was used at Feijao to break the bonding that had
been observed in tailings found after the failure. The bonding may have been thought
to explain the lack of deformation, however, the principle of effective stress applies in
concrete[11] and in other bonded materials, consequently the large predicted strains
(assumed to have been achieved by creep) would have played out as pore pressure
226
Q.105 - R.15
changes, as described above, developing more dilatancy at low level and drawing
pore pressures down further.
4. PORE PRESSURE RESPONSE AND THE ‘CYCLE’
The effect of the pore pressure changes is shown in Fig.6. On careful exami-
nation it can be seen that the 25 January 2019 pore pressure line on the right above
910 m elevation, shows an increased pore pressure (as a consequence of loss of
suction) following infiltration by ‘gentle’ rainfall, but it is on the left, under the other
pore pressure lines, below the 910 m elevation, showing a decrease in pore pres-
sure. The loss of suction at high level increases shear stress at depth, the shear
strain causes dilatancy and pore pressures are drawn down
The computed low pore pressures are shown in Fig.7(b). Although computed,
many do correspond to pore pressures recorded in the piezometers shown on the fig-
ure. Some records taken soon before failure show increases in pore pressures near
the toe, and others nearer the crest show decreases. Note that the pore pressures
are low, in suction (negative) far below the crest, where the shear strain and dilatancy
Fig. 6
Showing that pore pressure (i.e. load) increase at high level decreases pore
pressure in the tailings fill below (from Figure 48[3] ) Montrant que l’augmentation
de la pression interstitielle (i.e. la charge) en surface diminue la pression
interstitielle dans les stériles en dessous (d’après la figure 48[3] )
227
Q.105 - R.15
at depth are great, but high near the toe, where contraction was expected to occur,
driving pore pressures up.
This ‘cycle’ drawing down pore pressure, thereby increasing effective stress
and shear strength and resistance to failure along any incipient shear failure planes,
and thus maintaining stability, would have continued as long as ‘gentle’ rainfall con-
tinued. However, very heavy rainfall occurred over the period before the failure on 25
January 2019. The rainwater may have infiltrated into the crest, probably by 20 m,
as the gentle rain had already done (as shown on Fig.6, 940 m to 920 m eleva-
tion) and perhaps deeper, suddenly raising pore pressure and reducing suction and
increasing the disturbing force. The report estimated a pore pressure increase of
15 kPa, only 1.5 m water pressure, but it was probably greater because the portion
of the pressure thought to have been applied to cause creep would have remained
and been applied in raising pore pressures further. The pore pressures increased
too suddenly for the ‘cycle’ to operate to maintain stability by a further decrease in
pore pressure in the tailings. Consequently, the embankment failed, the tailings in
the dam liquefied and the retained liquid tailings were released.
5. EFFECTIVE STRESS STABILITY ANALYSIS – FAILURE AND

LIQUEFACTION
The Feijao pore pressure assessments made it possible to do a very simple

conventional[12] effective stress wedge failure stability analysis (as used for design at
Empingham[5] ). Fig.7 shows the failure plane at depth in the dam selected from the
results of the various analyses given in the reports and from observations from the
videos. Note that the failure takes place within the tailings (Fig.7(a)). At this depth
pore pressures on the failure plane were positive (Fig.7(b)). Preliminary analysis
assumed that the dam was close to failure (as it had failed and liquefied shortly
after), therefore the disturbing force imposed by the retained tailings was equal to
the resistance.
However, on closer examination, the factor of safety that the low pore pres-
sures, high effective stress and high shear resistance generated by the ‘cycle’ as
construction proceeds was found to be about 2.0. As the dam had failed, it was clear
that the dam, although adequately stable in normal operation as the dam height
increased (or drying out in the case of a ‘retired’ dam) was also near to failure from
substantial sudden loadings – seismic shaking or from sudden increases in pore
pressure as at Feijao. The sudden loadings would raise pore pressures through-
out the dam, too rapidly for the ‘cycle’ to operate, thereby weakening the dam and
reducing its resistance to the increased disturbing force.
This probably explains why a worldwide survey of tailings dam incidents found
that the majority were preceded by tremors 2 or 3 days before the incident[13] . The
majority includes Fundao, where an earthquake occurred a few days before the
failure, although the report[2] concluded that it had not influenced it.
228
Q.105 - R.15
Fig. 7
(a) Location of failure wedge within tailings dam (Figure 50 [3] )
a) Position du coin de rupture dans le barrage de stériles (figure 50 [3] ).
(b) Failure plane for simple effective stress wedge failure analysis
(adapted from Figure 47 [3] )
b) Plan de rupture du coin pour une simple analyse de stabilité en contraintes
effectives (adapté de la figure 47 [3] ).
The minority of failures may have included ‘retired’ dams such as Feijao, dams
where the ‘cycle’ had failed to deliver pore pressures low enough to maintain stability,
and other causes. Note that ‘retired’ dams such as Feijao, no longer used for disposal
of tailings are vulnerable to failure by both earthquakes and sudden infiltration of
rainfall.
6. LIQUEFACTION ON FAILURE
As the failure in the effective stress analysis was caused in part, possibly
entirely, by raised pore pressures, not all of the disturbing force would have been
deployed to overcome shear resistance along the failure plane. The balance would
229
Q.105 - R.15
have been applied horizontally to the failing mass. In saturated tailings, this would
have raised pore pressures sufficiently to cause vertical fractures, bursting and
disruption leading to liquefaction.
This explanation of the cause of liquefaction seems simple but should comply
with rigorous examinations of the situation near and after failure in conventional and
critical state failure modes [14,15] .
7. STABILITY AND SAFETY OF ‘UPSTREAM CONSTRUCTION’

TAILINGS DAMS
The simple effective stress stability analysis has shown that a knowledge of
pore pressures makes it possible to check the stability of this type of tailings dams
readily. The pore pressures are stress controlled, there is no drainage. Stability
depends on the low pore pressures, which are generated by the high shear stresses
deep in the dam. These shear stresses can be estimated by various numerical
analyses and methods, as indicated in Figs.3, 4 and 5. The Feijao pore pressures
were modelled, and compared well to measured values, but determining the pore
pressures to be expected from shear stresses is not usually possible, although it
can be assessed from pore pressure parameters [9] .
The analysis is ‘conventional’[12] but the process is not a form of staged con-
struction; here the pore pressures are stress-generated during the on-going process
of shear to form a shear plane (as indicated by the change from Fig.3 to Fig.2). There
is no drainage, the pore pressures respond solely to stresses.
As deposition continues, shear stresses increase, and the pore pressures

change in response to the shearing. The shear stress changes cause dilatancy at
depth, and contraction locally elsewhere, without any changes from drainage, the
stress-induced changes rippling through the dam and changing the pore pressures
accordingly.
Pore pressures respond to stresses and stress changes in all directions in a

dam. Measuring pore pressures directly by piezometers would make it possible to
check the stability of the dam routinely, to establish if stability is improving or wors-
ening with height as tailings were deposited. The trend would assist in determining
how to address future actions. If the trend suggests on-going stability, the magnitude
and probability of the earthquake that would cause failure could be estimated and
used to estimate risk and its acceptability[16] . If the trend suggests future instability in
minor earthquakes or non-earthquake situations, construction of a conventional dam
downstream could be initiated to collect and contain the debris and liquid tailings
from the failed dam.
A new dam would become necessary because no obvious remediation to over-

come the vulnerability of ‘upstream construction’ tailings dams seems to be available.
230
Q.105 - R.15
It would be best not to build dams of this type. As Feijao[3] shows, ‘retiring’ them
is not a safe solution. Long berms were provided at Empingham[5] to dissipate the
disturbing force over a long length of the weak foundation, not entirely successfully
as events in Fig.1 show. Berms ‘attached’ to a liquefiable existing dam do not offer
a secure remedy. Geotechnical solutions (grouting or cut-offs) may be considered,
but how to use them effectively in an ever-growing embankment is problematic, and
widespread disturbance of the embankments during installation may cause failure.
Research building on earlier work[17] to understand why the tailings settle in such a
way may provide insights into overcoming the long-term vulnerability of these dams.
A network of piezometers installed extensively within the dam, including at

depth below the crest as ‘construction’ advances would provide the data for routine
stability analyses. The continuous record of pore pressures will show the effects
of the ‘cycle’ and confirm that it is drawing down pore pressures confirming that it
causes dilatancy at depth. The piezometers will also confirm the extent and position
of the water table, saturation zone and negative pore pressures above the phreatic
line and the positive pore pressures below it (as on Fig.7(b)).
Installing piezometers and other instrumentation, taking readings and using

the data to check stability would be a primary duty of experienced engineers
and technicians engaged on site, following normal practice during construction of
conventional water-retaining earth embankment dams. They would supervise con-
struction, in co-ordination with mine operations. Incidents such as that at Fundao
where slimes were allowed to enter the sand tailings zone should be avoided. The
on-site engineers would be responsible for the civil engineering safety of the dams,
including preparing and testing emergency plans, and the many other engineering
duties on which efficient tailings disposal and the safety of the dam depend.
The emergency plan should be prepared to limit deaths and damage from fail-
ures. The consequences of failure on downstream communities can be examined
by dambreak analyses, determining the extent of the dambreak flood and people
and properties art risk. The plan includes details of how to warn downstream com-
munities if a dambreak is expected and identifies places of safety. The plan should
be tested from time to time. As many failures have occurred two or three days after
earthquakes, seismometers could be installed locally to provide early warnings for
people and workers to stay clear until it is plain that the tremors have not caused
failure.
8. CONCLUSION – INTEGRATED EMBANKMENT DAM ENGINEERING
The paper points out that upstream construction tailings dams similar to Fun-
dao and Feijao may be safer during operation than has been previously thought.
They are kept stable by substantial resistance to shearing resulting from the low
pore pressures generated by dilation caused by high shear stresses at depth. Failure
occurs if the low pore pressures are raised suddenly – most likely as a consequence
231
Q.105 - R.15
of earthquake shaking or similar sudden loading. The failure of Feijao dam after three
years of ‘retirement’ – no longer being used for tailings disposal – indicates that such
dams are also safe in normal conditions while drying proceeds gradually, but vul-
nerable to failure when loaded by sudden infiltration from very heavy rainfall as well
as by sudden loading from earthquakes.
Measured pore pressures could be used in routine effective stress analyses

to check stability in normal operation (or ‘retirement’) and examine the probability of
failure by earthquake shaking to apply in quantitative risk assessment[16] , from which
decisions on future actions and precautions against failure can be made. This follows
safety practice for conventional earth embankment dams and is in compliance with
the ICOLD declaration on dam safety[1] .
The extent to which the findings on stability can be applied to other tailings
dams is not clear (to the author), but the direct measurement of pore pressures and
the application of the principles of effective stress will usually improve understanding
of behaviour and simplify analysis of any embankment dam. Undrained (total stress)
analyses, although necessary in very low permeability embankment fills and foun-
dations, are inevitably making estimates of pore pressure based on tests which may
or may not match reality. The elaborate trials used to determine undrained strength
at Empingham[5] were not entirely successful, as Fig.1 and Fig.2 show.
Tailings dams such as Fundao and Feijao are unusual embankment dams.
They form ‘naturally’ in water. They are watertight. They retain settlement ponds
containing near-liquid and liquid iron ore tailings, the liquid is denser than water
as particles drop during sedimentation. The iron tailings make an unusual fill. The
specific gravity is around 4.0 (compared to 2.6 in many soils). The void ratio, e,
is high, occasionally as high as one, meaning that in parts the fill contains equal
volumes of tailings particles and water. Although the water volume is large, the
water content by weight of this very wet fill is about 25%. The density of this ‘heavy’
fill is about 25 kN/m3 .
These properties provide conditions to generate and sustain (until an earth-

quake occurs) the very low pore pressures shown in Fig.7(b) – the 200 kPa pressure
at the upstream end of the failure plane is at a depth of 70 m, where hydrostatic pres-
sure would be 700 kPa! The zero pore pressure (phreatic) line is about 40 m below
crest. Fig.6 shows large negative pore pressures (suctions) in the saturated tailings
above the phreatic line up to the ‘gradually dropping’ water table, where capillary
action commences, at about 30 m below crest.
By contrast, in low permeability clay at Empingham, Fig.1(a) shows high pore

pressures at end of construction, which dissipate gradually over the following nine
years.
Although there are differences, mine embankment dams and earth embank-
ment dams have much in common, notably that applying soil mechanics, with pore
pressure measurement as far as possible, aided by numerical methods when nec-
essary, can be used to examine the stability of upstream construction mine dams
232
Q.105 - R.15
similar to Feijao and Fundao, and probably other types of mine dams also, as well
as conventional earth embankment dams.
Embankment dam engineering, embracing mine dams and conventional earth

dams, is a specialism within dam engineering, applying soil mechanics to the engi-
neering of water-retaining embankment dams. Closer integration between tailings
embankment dam engineers and earth and rock embankment dam engineers, with
opportunities to work on both types, could lead to better practice and better under-
standing of the safety of all water and liquid retaining dams. The ICOLD Technical
Committees on Embankment Dams and on Tailings Dams may be able to take a
lead on this.
REFERENCES
[1] ICOLD. World Declaration on Dam Safety. Approved Porto 18 October 2019.
International Commission on Large Dams, Paris.
[2] FUNDAO Report. MORGENSTERN, N.R., VICK, S.G., VIOTTI, C.B., WATTS,
B.D. Report on the Immediate Causes of the Failure of the Fundão Dam, 2016.
From: http://fundaoinvestigation.com/.
[3] FEIJAO Report. ROBERTSON P.K., DE MELO L., WILLIAMS D.J., WARD
WILSON G. Report of the Expert Panel on the Technical Causes of the
Failure of Feijão Dam, 2019. From: http://www.b1technicalinvestigation.com/
report.html.
[4] CAMBRIDGE, M., SHAW, D. Preliminary reflections on the failure of the Bru-
madinho tailings dam in January 2019. Dams and Reservoirs, Proceedings of
the British Dam Society, Volume 29, No 3 Sept, 113–123, 2019.
[5] BRIDLE, R.C., VAUGHAN, P.R., JONES, H. N. Empingham Dam – design,

construction and performance. Proc. Institution of Civil Engineers, Part 1, ,78,
Apr., 247–289, Paper 8828, 1985.
[6] WINDER, A.J.H. et al. The Empingham Reservoir Project Discussion Proc.
Institution of Civil Engineers, Part 1, 1986, 80, Aug., 247–289, Discussion
Papers 8827-9, 1986.
[7] KOVACEVIC, N., HIGGINS, K.G., POTTS, D.M., VAUGHAN, P.R. Undrained
behaviour of brecciated Upper Lias Clay at Empingham Dam. Géotechnique
57 (2), 181–195, 2007.
[8] WESLEY, L. The Bishop Method, the life and achievements of Professor Alan
W Bishop, soil mechanics pioneer. Whittle Publishing, London, 2019a.
233
Q.105 - R.15
[9] BISHOP, A.W. The use of pore pressure coefficients in practice. Géotechnique
4, 148–152, 1954.
[10] WESLEY, L. Professor AW Bishop’s finest papers, a commemorative volume.

Whittle Publishing, London, 2019b.
[11] SKEMPTON, A.W. Effective stress in soils, concrete and rocks. In Pore
pressure and suction in soils. Conf sposm.02050.0014. 1960.
[12] LADD, C.C. Safety evaluation during staged construction. Journal of Geotech-
nical Engineering 117(4) 540-615 April 1991.
[13] LEGGE, Kelvin. Personal communication, e-mail 13 March 2020: ‘if we look
at the worldwide survey of tailings dam incidents, we note the majority are
preceded by tremors within 2 or 3 days of the incident’.
[14] GENS, A. Hydraulic fills with special focus on liquefaction. Proceedings of the
XVII ECSMGE-2019 Geotechnical Engineering foundation of the future ISBN
978-9935-9436-1-3, 2019.
[15] CARRERA, A., COOP, M., LANCELLOTTA, R. Influence of grading on

mechanical behaviour of Stava tailings. Géotechnique 61(11), 935–946
[doi.org/10.1680/geot.9.P.009], 2011.
[16] BRIDLE, R. Estimating the risk to dams from earthquakes, floods and internal
erosion. Trans.26th ICOLD Congress, Vienna, Q101, R3, 33–43, International
Commission on Large Dams, 2018.
[17] ECKERSLEY, J.D. Instrumented laboratory flowslides. Géotechnique 40(3),

489–502, 1990.
234
Q.105 - R.16
-------
-------
BEFORE AND AFTER THE TODDBROOK DISASTER – A REVIEW OF

UK RESERVOIR INCIDENT MANAGEMENT (∗)
A.L. WARREN
MOTT MACDONALD
P.J. MASON
DAMSOLVE LTD
UNITED KINGDOM
SUMMARY
The UK has an excellent reservoir safety record with no loss of life from dam
breach since 1925. Until the turn of the century only major or unusual incidents were
reported and captured in databases. Centralised regulation of UK reservoirs led to
opportunities for improving the capture and analysis of incidents and in identifying
those incidents worthy of more detailed, independent investigation. In 2013, incident
reporting became mandatory for regulated reservoirs located in England and Wales.
There have only been a small number of serious incidents in the UK over the last 20
years. However, a very significant incident occurred in 2019 at Toddbrook Reservoir
where damage to an auxiliary spillway had to be repaired through a multi-agency
and military response and the reservoir drained to avert the destruction of a town.
The paper summarises developments in UK reservoir incident management

over the last twenty years and reviews the impact of mandatory reporting. The
∗ Avant et après la catastrophe de Toddbrook – un examen de la gestion des incidents dans
les réservoirs au Royaume-Uni
DOI: 10.1201/9781003211808-62 235

Q.105 - R.16
technical findings of the UK government’s investigation of the Toddbrook incident

are summarized, together with the efforts being made to further review reservoir
safety management processes and legislation in light of this major incident.
RÉSUMÉ
Le Royaume-Uni n’a connu aucune perte de vie à la suite d’une rupture de

barrage depuis 1925. Jusqu’au tournant du siècle, seuls des incidents majeurs ou
inhabituels étaient signalés et enregistrés dans des bases de données. La réglemen-
tation nationale des réservoirs britanniques a permis d’améliorer la connaissance
et l’analyse des incidents et d’identifier les incidents nécessitant une enquête plus
détaillée et indépendante. En 2013, la déclaration des incidents est devenue obli-
gatoire pour les réservoirs réglementés situés en Angleterre et au Pays de Galles.
Il n’y a eu qu’un petit nombre d’incidents graves au Royaume-Uni au cours des 20
dernières années. Toutefois, un incident très important s’est produit en 2019 au
réservoir Toddbrook, où les dommages causés à un déversoir auxiliaire ont dû être
réparés par une intervention militaire et multi-organismes et le réservoir vidangé
pour éviter la destruction d’une ville.
Le rapport résume l’évolution de la gestion des incidents de barrages au

Royaume-Uni au cours des vingt dernières années et passe en revue l’impact des
rapports obligatoires. Les conclusions techniques de l’enquête du gouvernement bri-
tannique sur l’incident de Toddbrook sont résumées, ainsi que les efforts déployés
pour examiner plus avant les processus de gestion de la sécurité des réservoirs et
la législation à la lumière de cet incident majeur.
1. INTRODUCTION
In the UK there were many serious incidents at dams in history requiring

emergency response activities to avert potentially dangerous escapes of water
from reservoirs (CIRIA, 2014). Legislation was first introduced in 1930 to introduce
regulatory controls for the design, construction, periodic inspection and removal
of reservoirs. The current legislation for England and Wales is the Reservoirs
Act 1975 (the Act). Scotland is covered by similar but separate legislation, the
Reservoirs (Scotland) Act 2011. The Act sets out the responsibilities of reservoir
owners/operators in maintaining records, carrying out monitoring and surveillance
activities, appointing a Supervising Engineer to oversee the safety management
of the reservoir and Inspecting Engineers to carry out periodic independent safety
reviews. For England, the government regulatory body is the Environment Agency
(EA), responsible for over 2000 regulated reservoirs.
236
Q.105 - R.16
Since the introduction of legislation there has been no loss of life due to dam
failure. The introduction of mandatory requirements for reservoir supervision under
the Act in the 1980’s is also credited with advancing reservoir safety in the UK.
In the 1990’s work was started to collate information on the characteristics

of regulated reservoirs and their associated incident history (Tedd et al, 1992).
A system of post-incident management has been administered by the EA since 2007
(Hamilton-King et al, 2008). Through an amendment to the Act in 2013, it became
a mandatory requirement to report any serious incident. The EA publishes annual
reports on all reported incidents. Typically, there are a few serious incidents each
year requiring urgent intervention activities. The dam failures that have occurred in
recent times have occurred at small non-statutory reservoirs of less than 25,000m3
capacity (the current lower bound for regulated reservoirs in England). No large reg-
ulated reservoir in the UK has failed since 1970 when the Warmwithens dam failed
due to internal erosion (CIRIA, 2014).
A number of incidents in the present century indicate that there may be more
that can be done to improve reservoir safety management in the UK. The most
notable of these are the incidents at Ulley in 2007 (Hinks et al, 2008) and at Tod-
dbrook in 2019 (Warren and Bennett, 2019). Both of these incidents involved flood
events and an emergency response to preserve dam safety. The Toddbrook incident
was particularly serious, instigating the evacuation of 1500 residents of the town of
Whaley Bridge.
The aim of this paper is to summarise the work done over the last 15 years in
the UK to identify/register incidents, investigate selected incidents, learn from them
and inform the regulatory framework and tools to promote reservoir safety. The tech-
nical findings of the government review of the Toddbrook incident are summarised
together with the current process of government review.
2. VOLUNTARY REPORTING OF INCIDENTS, 2007 – 2013
The reporting of reservoir incidents in the UK up to the turn of the century was
generally limited to those incidents requiring some type of an emergency response
or those which were of sufficient interest or magnitude to warrant the engineers
involved to write about their experiences in identifying or managing incidents.
The EA developed a specification for recording incidents at reservoirs and

undertaking investigations in 2006 (Warren and Hope, 2006). Building on the earlier
database of UK incidents developed by the Building Research Establishment, the
new system provided the means for any serious incident to be reported on a volun-
tary basis and recorded in a central database. The database allowed for incidents
to be recorded against any regulated reservoir in England, Wales or Scotland. Inci-
dents at non-regulated reservoirs could also be added. The system also set out the
237
Q.105 - R.16
principles for the investigation of serious or unusual incidents and for dissemination
of the reported incidents and investigations to the reservoir industry. The system
was referred to as the post-incident reporting (PIR) system in recognition of the
need to discriminate between immediate flood-risk reporting to the authorities and
the reporting of information sometime after the event. Typically, more information
on the causes of the incident and the effectiveness of the response become avail-
able over the year following the incident, typically following investigations by the
owner.
The system adopted recognises that:
• incidents that by chance do not lead to a dam breach should attract equal
attention to those that do;
• investigations into the root causes of incidents can be particularly valuable
where the circumstances are unclear or unusual;
• reservoir safety should be improved nationally by disseminating the nature of
incidents that do arise and using the information to inform related research
into the vulnerability and the effectiveness of incident management through
guidance and legislation.
3. MANDATORY REPORTING IN ENGLAND AND

WALES FROM 2013
Mandatory reporting of incidents at regulated reservoirs in England and Wales

was introduced in 2013. Reporting of incidents at reservoirs in Scotland remains
voluntary at the present time. Incidents at non-regulated reservoirs (i.e. less than
25,000 m3 capacity in England or 10,000 m3 in Wales and Scotland).
Under the regulations, the reservoir operator must provide a preliminary report
to the regulator, as soon as an incident is under control. This report must contain:
• the date and time of the incident

• the location of the reservoir
• immediately observed facts
Within a year from the day after the incident, the operator must send a final
report containing:
• information about what happened

• analysis of its circumstances
• the conclusions that can be drawn from it
• any lessons learned from the incident
238
Q.105 - R.16
Fig. 1
Temporal distribution of reportable reservoir incidents by severity for all reservoirs
(England 2004-19; Scotland and Wales 2004–16).
Figure 1 shows the total number of reportable incidents since 2004, excluding
incidents in Wales and Scotland for 2017–2019. Level 1 incidents are dam failures
(uncontrolled sudden large release of water); Level 2 incidents involve an emergency
drawdown, emergency works or a serious operational failure in an emergency; Level
3 incidents are those involving a precautionary drawdown of the reservoir, unplanned
physical works to preserve reservoir safety or a human error leading to a major
(adverse) change in operating procedure. Most of the Level 1 failures have been at
small non-regulated reservoirs. The large number of incidents in 2007 and 2012 are
mostly associated with significant flood events that occurred in England.
It is of note that there has been no marked increase in the number of reported
incidents since the introduction of mandatory reporting for regulated reservoirs in
2013. This could be viewed as a failure of the legislation to expose all reportable
incidents or an indication of the good level of reporting that existed under the volun-
tary system. It is the view of the authors that the completeness of reporting prior to
2013 was not good and that mandatory reporting was necessary. Since 2013, either
the legislation has not been effective in its aims or there an unusually low number
of incidents have occurred. For regulated reservoirs incidents, reporting is often
instigated by the Supervising Engineer familiar with the reporting system. Reporting
239
Q.105 - R.16
of incidents at non-regulated reservoirs (which are not supervised) has often been
through the involvement of the EA in managing the associated flood risk.
Devolution in the UK has led to differences in legislation for reservoirs located

in England, Wales and Scotland. This has the potential to negatively impact the
effectiveness of incident management through the introduction of different national
systems, legislation and guidance. Ideally, the arrangements and format for reporting
incidents and the required content of reports shall continue to be consistent across
the UK. This will position the UK to best contribute to international databases and
studies on the causes of dam incidents and incident management.
4. DISSEMINATION
The EA has published an annual report to the UK reservoir industry every year
since 2007. Unless the incident is already nationally recognized, the name, location
and reservoir owner are not disclosed. This has been to encourage reporting and has
continued since mandatory reporting, recognizing that in most cases this information
is superfluous to the aims of the system. Each summary report includes:
• Dam type
• Legal status (regulated or not regulated)
• Dam height
• Incident type (e.g. internal erosion)
• Incident severity (1/2/3)
• A description of the events leading to an incident being declared
• The key lessons to be learned from the factors leading to the incident and the
incident management.
5. LEARNING from PREVIOUS INCIDENT INVESTIGATIONS
Since the start of the PIR system in 2007 there have been six independent
investigations on behalf of the EA/UK government. The aim of these investigations
has been to ensure that where unusual or serious incidents occur, sufficient informa-
tion is gathered to fully inform the reservoir industry on the causes of the incident or
to inform further research to better understand what might have led to the incident.
In most cases, bulletins have been produced to provide the industry with a summary
of key points of learning and to inform research needs. The Ulley incident featured
in a review of the 2007 summer floods by Sir Michael Pitt (2008).
Ulley reservoir near the town of Rotherham in the UK, was constructed 1873.
It is retained by an earth embankment dam with a central clay core and features two
240
Q.105 - R.16
Fig. 2
Ulley Reservoir, 2007: masonry spillway wall collapse and
erosion of embankment fill.
original stepped masonry spillway chutes, one running down each embankment
mitre. Although a new, larger spillway was added in 1943, at times of flood, water
still discharged down these original masonry structures. On the 25th of June 2007
and in response to heavy rainfall, all three spillway chutes discharged and the small
masonry chute on the left flank started to break-up with the released flow eroding
the adjacent downstream fill of the main embankment (Figure 2).
Efforts to draw down the reservoir by pumping were hampered by the heavy
rain and catchment run-off which continued to fill the reservoir. An emergency was
declared, the nearby M1 motorway was closed for 24 hours and approximately 1,000
people downstream were evacuated. The incident, and that of a similar spillway
collapse in 2005 at Boltby reservoir near Thirsk in the UK, were later described in a
paper by Hinks, Mason and Claydon (2009).
There were a number of consequences. One was a government funded

research programme in the technical reasons for the collapses. The resulting report
(Winter et al, 2010) identified trapped rotating vortices in the horizontal steps which
generated high hydro-dynamic pressure differentials on the adjacent walls. These
effectively “plucked out” the masonry units, causing the walls to collapse internally.
It also emerged that because no downstream inundation mapping was avail-

able, many people downstream were put on alert of evacuation unnecessarily. This
241
Q.105 - R.16
led to a programme of downstream inundation mapping, initially focused on just

England and Wales’ highest consequence reservoirs but later expanded to include
all regulated reservoirs.
Incidents such as Ulley, together with certain international incidents, have

played an important part in shaping UK reservoir safety management research,
guidelines and legislation (CIRIA, 2014).
6. INTRODUCTION THE TODDBROOK INCIDENT

INVESTIGATION
Toddbrook Reservoir was constructed in 1840, providing a feed to the Peak

Forest Canal. It is owned and managed by the Canal & River Trust (the ‘Trust’). It is a
high-risk dam posing a risk to lives in a community downstream of the reservoir and
holds 1.3 million m3 of water behind a 24m high earthfill dam that features a central
puddle clay core. The reservoir has a bywash on the northern bank of the reservoir
which can be used to divert a proportion of the reservoir inflow around the reservoir
and into the primary spillway channel down the left mitre of the dam embankment. A
gated inlet to the bywash channel is provided at the inlet weir located at the head of
the reservoir. The dam also features an auxiliary spillway which was added in 1970
to increase the flood discharge capability of the works. It operated in 1973, 1983,
1998, 2007 and then in 2019 when it partially failed.
The prelude to the failure was two periods of heavy rain starting on 27th July
2019. It is believed that about 160mm of rain fell on the catchment between then
and 31st July 2019. The rise in reservoir level caused the auxiliary chute to oper-
ate throughout 31st July. Plumes of spray were thrown up by flows impacting onto
embedded rock plums on the chute invert and regular burst of silt laden water could
be seen discharging from the base of the chute. The reservoir reached its maximum
water level at around 9pm on 31st July.
By the following morning, the 1st August, flows down the chute were much
less. By then it became apparent that some lower chute slabs had lifted and were
releasing mud and silt from their joints. Later in the morning some upper slabs
collapsed into a void which had developed in the supporting embankment beneath
them and this situation progressed until late afternoon. It also became apparent
that if the erosion of the embankment continued, it would start to threaten the main
water-retaining core of the dam. An emergency was declared, pumps were brought
in to lower the reservoir water level, a military helicopter was deployed to place
large bags of aggregate in the void (Figure 2) and approximately 1500 people were
evacuated from the downstream town of Whaley Bridge as a precaution, informed by
the reservoir inundation maps. Further details on the incident response are provided
by Warren and Bennett (2020).
242
Q.105 - R.16
Fig. 3
Scour hole infilling.
The investigations into the cause of the collapse formed Phase A of a review
headed by Professor David Balmforth. They concluded that the chute was in gen-
erally poor condition, that its design did not comply with modern practice and that
maintenance had at times been quite poor. All this had facilitated the injection of
flows into the underlying fill during the flood due to water impacting on the embed-
ded rock plums. This in turn had led to “stagnation” pressures developing on the
chute where locally high impact pressures had facilitated flow injection into local
cracks in the chute invert. Photographic evidence demonstrated the historic pres-
ence of extensive cracking in and just above the area that failed and research reports
from US practice (USBR, 2007) enabled the likely amounts of injection water to be
quantified.
The report by Professor Balmforth (2020) also drew other conclusions and
made recommendations about inspections and reporting practice based specifi-
cally around what he had seen at Toddbrook. These recommendations have been
accepted by the UK government and are in the process of implementation.
7. NEXT STEPS
Phase B of the UK government review commenced in 2020 and is scheduled

to complete in 2021. The purpose is to review the application of current legislation for
reservoir safety and report on whether the regulation of reservoirs remains effective
and robust in securing the ongoing safety of dams. It will include:
• how far the interpretation and application of current legislation is consistent

with all those who have responsibility for reservoir safety, and secures the
ongoing safety of reservoirs;
243
Q.105 - R.16
• consideration of the competency needs, assessment, appointment and roles

of panel engineers to ensure appropriate, timely and independent advice to
undertakers and the regulator;
• whether current legislation and/or associated regulations require updating.
One reason that the emergency at Toddbrook Reservoir was so successfully

and promptly dealt with in 2019 was because the owner, the Canal & River Trust,
had such plans in place for all their major reservoirs and routinely practiced their
implementation by way of simulation exercises. The government review does not
include a review of incident emergency response by the UK emergency responders.
However, in 2020 the UK government consulted the industry on proposals to intro-
duce mandatory emergency on-site planning for all regulated reservoirs in England.
A questionnaire in 2019 found that 78% of 883 high-risk reservoirs in England did
have an emergency on-site plan.
8. CONCLUSIONS
Following its introduction in 2007, the UK’s incident management system has
enabled the recording of in excess of one hundred reservoir incidents and has been
used to promote learning and inform research requirements. The most serious UK
incidents in this century, notably those at Ulley and Toddbrook have been the subject
of government investigation and industry review. The immediate recommendations
arising from the Toddbrook incident are now being implemented. The outcome of
the wider review on the application of current legislation will be published in 2021.
REFERENCES
Balmforth, D. 2020. Toddbrook Reservoir independent review report. Available on

www.gov.uk [accessed 21 September 2020].
Rigbey, S.J. and Hartford, D.N.D. 2019. Oroville in retrospect – what needs to
change? ICOLD annual meeting, 2184-2195, Ottawa, 2019.
CIRIA. 2014. Lessons from incidents at dams and reservoirs – an engineering guide.
SP167. CIRIA Publishing.
Hamilton-King, L., Hope, I. and Warren A. 2008. Post-incident reporting: learning

from experience to promote reservoir safety. British Dam Society conference
proceedings, University of Warwick, Thomas Telford.
Hinks, J., Mason, P. and Claydon J. 2008. Ulley Reservoir and high velocity spill-
way flows. Proc of the British Dam Society conference, University of Warwick.
Thomas Telford.
244
Q.105 - R.16
Hinks JL, Mason PJ and Claydon, 2009, The 2007 Event at Ulley Reservoir near
Sheffield, Q91-R21, ICOLD Congress, Brasilia.
Pitt M, Lesson learned for the 2007 floods, an independent review by Sir Michael
Pitt, UK Gov, London, June 2008.
Institution of Civil Engineers. 2015. Floods and Reservoir Safety. Fourth edition. ICE
Publishing.
Tedd, P., Holton, I. and Charles, J. 1992. The BRE dams database. Proc of the
British Dam Society conference, University of Stirling. Thomas Telford.
US Bureau of Reclamation, 2007. Uplift and crack flow resulting from high velocity
discharges over open offset joints. Report DSO-07-07.
Warren, A. and Bennett, N. 2020. The Toddbrook reservoir spillway incident and
emergency response. ICOLD annual meeting, New Delhi, 2020 (publication
pending).
Warren, A. and Hope, I. 2006. A New Incident Reporting System for UK Dams. Proc
of the British Dam Society conference, University of Durham. Thomas Telford.
Warren, A. and Patten, B. 2018. Learning from reservoir incidents – a summary of

the causes and management of incidents in the UK. Proc of the British Dam
Society conference, University of Swansea. Thomas Telford.
Winter C, Mason PJ and Baker R, Guidance for the Design and Maintenance of
Stepped Masonry Spillways, Project SC080015, Environment Agency, UK, Oct
2010.
245
Q.105 - R.17
-------
-------
ACCIDENTOLOGIE DES DIGUES FLUVIALES DE LOIRE,

DU DELTA DU RHÔNE ET DE L’AGLY, ET LEÇONS TIRÉES (∗)
Rémy TOURMENT
Ingénieur expert, INRAE
Thibaut MALLET
Directeur général, SYMADREM
Sébastien PATOUILLARD
Ingénieur expert, DREAL Centre – Val de Loire
Akim SALMI
Directeur général délégué, ISL
FRANCE
RÉSUMÉ
Ce rapport présente le retour d’expérience liés aux crues et aux inondations

sur trois sections de fleuves français, la Loire moyenne, Le Rhône aval et l’Agly mar-
itime, au cours de la période récente (3 dernières décennies) et en tenant compte
aussi des événements des siècles passés. Les principales leçons qui en ont été
tirées, qui ont été partagées au niveau national et pour certaines ont d’ores et
déjà été suivi d’effet sont présentées. Les plus importantes en sont: l’importance
d’une bonne gouvernance, la nécessité de gérer les surverses à l’aide de systèmes
résilients, l’utilité du retour d’expérience et de la pérennisation des informations sur
l’accidentologie. Aussi bien l’ingénierie de conception que les diagnostic et l’analyse
∗ Accidents and incidents on fluvial levees on the Loire, Rhône and Agly rivers, and lessons
learned
246 DOI: 10.1201/9781003211808-63

Q.105 - R.17
de risque bénéficient de cette connaissance. Par ailleurs la prise de conscience col-

lective de l’importance des ouvrages de protection contre les inondations conduit
à une structuration de la communauté professionnelle et à une amélioration des
connaissances et des pratiques.
SUMMARY
This report presents the experience feedback on floods and inundations on

three sections of French rivers, the Loire middle course, the downstream Rhône
and the Agly River, during the recent period (last 3 decades) and also taking into
account events of past centuries. The main lessons learned, which have been shared
at national level and some of which have already been followed up, are presented.
The most important are a good governance, the need to manage overflows using
resilient systems, the usefulness of feedback and the sustainability of information
on accidentology. Both design engineering, diagnosis, and risk analysis benefit
from this knowledge. In addition, collective awareness of the importance of flood
protection works leads to a structuring of the professional community and to an
improvement in knowledge and practices.
1. INTRODUCTION
Voici près de trente ans, la communauté française a redécouvert l’existence

et l’importance de son patrimoine de digues de protection contre les inondations,
fluviales, maritimes et torrentielles, à l’occasion de crues sur le Rhône aval ayant
endommagé les digues de Camargue et inondé son delta. Depuis, les épisodes
se sont multipliés dans des contextes géographiques divers, et parfois de manière
répétée sur certains secteurs: Rhône aval en octobre 1993, janvier 1994, décembre
2003, Gard et Rhône septembre 2002, Aude 1999, Agly 1999, 2013, 2014 et 2020,
tempête Xynthia 2010, crues torrentielles de juin 2013 dans les Pyrénées, pour ne
citer que les plus importants en termes de conséquences humaines ou matérielles.
Ces événements récents sont survenus après plus d’un siècle de “tranquillité”, les
derniers événements majeurs de crue mettant en jeu des digues datant du XIXème
siècle, même si d’autres événements ont pu amener à des prises de conscience du
patrimoine de digues et à des actions de reprise en main plus locales, comme par
exemple sur la Loire dans les années 1960 ou dans le Centre–ouest du pays suite
aux crues de 1982–83.
Nous présentons dans cet article les retours d’expérience (REX) qui ont pu
être effectués sur trois secteurs de fleuves français, la Loire moyenne, le delta du
Rhône et l’Agly maritime et l’ensemble des évolutions et perspectives concernant
les digues.
247
Q.105 - R.17
Il a pu être constaté en faisant l’analyse des comptes rendus d’événements

un certain nombre de difficulté d’interprétation suite à des usages variables du
vocabulaire, entre autres sur le terme de brèche. Depuis une période récente, la
terminologie s’est “normalisée”. La brèche est une rupture au sens de la Commis-
sion Internationale des Grands Barrages (CIGB) dont le lâcher d’eau ne peut plus
être maîtrisé. Elle est obtenue à la fin d’un processus impliquant des mécanismes
de dégradation et de ruptures, qui se développe en trois étapes [1]: initiation, for-
mation et élargissement. Elle est donc caractérisée par une phase d’élargissement
et d’approfondissement dans la fondation, et des entrées d’eau dans la zone pro-
tégée qui ne peuvent plus être interrompues. Le départ de brèche se distingue de
la brèche par le fait que la phase d’élargissement n’est pas encore en cours et/ou
que le lâcher d’eau a été maîtrisé ou sans impact sur la zone protégée ; ces entrées
d’eau peuvent être interrompues par un effondrement de la digue, une intervention
d’urgence ou la diminution de la charge hydraulique liée à la décrue. Le désordre
est défini quant à lui comme le signe observable ou quantifiable d’une dégradation
de l’état initial de la digue qui peut être l’amorce d’une rupture et d’un processus de
brèche. Une brèche peut être partielle (seule une partie de la hauteur de la digue a
disparu) ou totale (toute la hauteur de la digue a disparu).
2. LA LOIRE MOYENNE: REX DES CRUES DU XIXEME SIÈCLE
L’histoire des digues de Loire, dont l’origine remonte au Moyen-âge, est mar-
quée d’évènements catastrophiques dus à des crues exceptionnelles. Les dernières
crues exceptionnelles remontent à la fin du XIX◦ siècle; au nombre de trois, elles se
sont succédé en 1846, 1856 et 1866 avec des conséquences humaines et sociétales
difficiles à imaginer aujourd’hui. La période de retour de ces crues a été évaluée à
170 ans, c’est à peu de chose près le temps qui nous sépare de ces évènements
dramatiques. Les hommes qui ont vécu ces évènements, ont aussi assisté à la
destruction des grandes digues de Loire dont la fiabilité semblait garantie par des
siècles d’amélioration apportée par un savoir-faire empirique. Cette période a aussi
été celle des grandes remises en cause et des innovations en matière d’ouvrage de
protection contre les inondations dans le bassin de la Loire.
2.1. DESCRIPTIF (HYDRAULIQUE) DES CRUES ET INONDATIONS

HISTORIQUES DU XIXÈME SIÈCLE
Les crues de la fin du XIXème siècle qui ont fortement impacté la Loire
Moyenne sont qualifiées de crues mixtes. Comme illustré dans la figure 1, il s’agit
d’une combinaison de crues océanique et méditerranéenne (dont l’origine est liée à
des phénomènes météo venant respectivement de l’Océan Atlantique ou de la Mer
Méditerranée).
248
Q.105 - R.17
Fig. 1
Représentation de la Loire Moyenne et schéma d’une crue mixte
Loire Moyenne location and mixed flood drawing
(source DREAL CVL/DETL).
La crue de 1846 a été la plus catastrophique en termes de pertes humaines

et de bétail. Le niveau d’eau est monté de 5 mètres en 24 h au niveau des ponts
ligériens et la crue a submergé les digues, occasionnant plus de cent brèches et
inondant le lit majeur. Comme pour 1856 et 1866, les débits en Loire moyenne
ont été estimés localement entre 7000 et 8000 m3 /s pour un débit moyen de
l’ordre de 500 m3 /s. Sur ces trois évènements, 376 brèches affectant les levées
ont été recensées. Elles constituent le fondement de la base de données servant
à l’analyse sur les levées de Loire [2], les huit brèches ou départs de brèches
observés lors des crues suivantes (1895, 1902, 1968, 1982) n’ayant pas eu d’impact
comparable.
2.2. LA CAPITALISATION DU RETOUR D’EXPÉRIENCE:

UNE RICHESSE EXEMPLAIRE
À la suite de la crue de 1846, les ingénieurs ont entrepris des renforcements

conséquents et mis en œuvre plusieurs outils innovants pour suivre les crues du
fleuve et capitaliser les données liées à l’accidentologie: le réseau de bornes repères
et d’échelles hydrométriques, une cartographie détaillée représentant le lit majeur
sur près de 600 km avec la structure des digues, les brèches connues et les fosses
d’érosion d’anciennes brèches, ainsi qu’un nivellement général des deux rives. Ces
outils seront utilisés pour capitaliser les données relevées sur les brèches des deux
crues à suivre comme le montre la figure 2 ci-après. Aujourd’hui ces outils sont
toujours utilisés pour le suivi de l’évolution du lit de la Loire et la prévision de crue,
complétés de nouveaux développements: notamment la modélisation hydraulique,
le traitement de donnée et la géomatique.
249
Q.105 - R.17
Fig. 2
Extrait de la carte de 1850 au niveau d’une brèche de 1846 (en noir)
surchargée par 2 brèches de 1856 (en jaune)
Extract from the 1850 map at the location of a breach of 1846
(written in black) and two breaches of 1856 (added in yellow)
2.3. COMPLÉTER L’ACTION DES DIGUES PAR L’ACTION DE BARRAGES AMONT
Après la crue de 1856, la stratégie de rendre les digues insubmersibles, est

remise en question. Pour éviter les brèches et écrêter les grandes crues, deux
systèmes sont proposés: un ensemble de réservoirs modérateurs des crues sur la
Loire amont et sur l’Allier et/ou des déversoirs régulateurs des crues dans les levées
de tous les vals de la Loire [3]. La crue de 1866 conduira à précipiter l’orientation
vers un programme ambitieux de restauration des systèmes de protection assortis
de déversoirs.
Par la suite, le système de protection sera effectivement complété par la mise

en service du barrage écrêteur de crues de Villerest sur la Loire amont en 1985.
3. LE DELTA DU RHÔNE
3.1. CONTEXTE
Le Delta du Rhône couvre une superficie de 1500 km2 . Il est protégé par trois
systèmes d’endiguement fluviaux et un système d’endiguement maritime constitués
250
Q.105 - R.17
respectivement de 225 km de digues fluviales et 50 km de digues maritimes, gérés

par le Syndicat Mixte Interrégional d’Aménagement des Digues du Delta du Rhône
et de la Mer (Symadrem), un établissement public créé en 1999 suite à la défaillance
des associations de propriétaires riverains qui géraient, depuis le milieu du XIXème ,
les ouvrages.
Les systèmes d’endiguement ont été créés après les inondations de 1840 et
1856 (périodes de retour estimées respectivement à 400 et 250 ans) en lieu et place
d’autres ouvrages encore plus anciens, dont certains remontent au XIIème siècle. Du
fait de leur mode de réalisation (compactage manuel avec des dames de 15 kg, non
prise en compte de la teneur en eau à l’optimum découverte en 1933 par Ralph Proc-
tor) et de l’effet «mille-feuilles» dû aux phases successives de rehaussement, les
digues du Rhône sont fortement exposées au risque de brèche par érosion interne
[4]. Dans le cadre des études de dangers exigées par la réglementation française,
une étude accidentologique a été menée sur l’ensemble du parc de digues fluviales.
Les principaux résultats sont présentés ci-après, on les retrouve de manière plus
détaillée dans [5].
3.2. STATISTIQUES DES ÉVÉNEMENTS RÉCENTS ET ANCIENS
La figure 3 indique les débits de pointe, en tête de delta, des crues maxi-
males annuelles de 1830 à 2016. En rouge figurent les crues ayant donné lieu à
inondation par brèche, de la zone protégée. Depuis 1840, neuf crues ont donné
lieu à inondation: 1840, 1841, 1843, 1856, 1993, 1994, 2002, 2003, 2016. La
figure indique également les périodes analysées: la période ancienne de 1840 à
1886 et la période récente de 1993 à 2016. La période intermédiaire de 1886 à
1993, correspondant à la «veille» hydrologique du Rhône, n’a pas été analysée
faute de documents d’archive suffisamment détaillés pour procéder à une analyse
de qualité. En sus de ces crues avec inondation, six crues sans inondation ont
également été analysées. L’analyse de la figure 3 permet de constater que la prob-
abilité d’inondation par brèche a été inférieure à 5% pour les crues de périodes
de retour comprises entre 5 et 10 ans. Elle a été comprise entre 5 et 50% pour
les crues de période de retour comprise entre 10 et 20 ans et supérieure à 50%
pour les événements de 20 à 50 ans. Elle est de 100% pour les crues de péri-
odes de retour supérieures à 50 ans. Ces probabilités, issues de l’observation des
crues passées, sont considérées comme un bon indicateur de la gestion de crise.
Sur les 15 crues analysées, 57 brèches, 57 départs de brèche et 133 désordres
ont été recensés. 3 brèches par retour vers le fleuve ont également été recensées
en 1840, ce qui donne un nombre total de 117 brèches et de départs de brèche
en 176 ans, ce qui est conforme à l’évaluation faite dans le cadre des études de
dangers [6].
Les brèches sont globalement localisées sur le Rhône et le Petit Rhône amont,
là où les hauteurs de digues sont les plus importantes (4,5 à 7 m) et les hauteurs
251
Q.105 - R.17
Fig. 3
Débits des crues et inondation de la zone protégée (1830–2016)
Flood discharges and inundation of the protected area (1830–2016).
de charge (3 à 6 m) les plus élevées par le passé. Les départs de brèche sont,
quant à eux, plutôt localisés sur le Grand Rhône et le Petit Rhône aval, là où les
hauteurs de digues (2 à 4,5 m) sont plus faibles et les hauteurs de charge par le
passé également moins élevées (1,5 à 2,5 m).
Le tableau 1 ci-dessous dresse un bilan par initiateur de rupture selon les

périodes. On constate que l’accidentologie des digues du delta du Rhône suit
une tendance similaire à celle des barrages en remblai [1] avec une proportion
d’accidents liés à l’érosion interne plus élevée.
Les brèches par surverse sont au nombre de 19 sur toute la période. 16 se

sont produites en 1840 avant le rehaussement généralisé des digues et 3 lors de la
crue centennale de décembre 2003. Elles ont toutes été initiées sous une très faible
lame d’eau (moins de 10 cm) mais avec une hauteur de chute supérieure à 4 m.
Pour ces trois brèches de 2003, certaines hétérogénéités ont été observées, après
crues, dans les corps de digue. Des mécanismes d’érosion interne ne sont donc
pas à exclure, ce qui pourrait donner un total de 100% des brèches par érosion
interne pour la période récente. On note également deux départs de brèche par
surverse en 2002 et 2003 sur un linéaire de 2 km en partie aval du delta (analysées
plus bas).
252
Q.105 - R.17
Tableau 1
Statistique des brèches et des départs de brèche par initiateur de rupture
Breaches and breach initiations by initiating failure mode
PÉRIODE STATISTIQUE DES BRÈCHES (57) STATISTIQUES DES BRÈCHES (57)
PAR INITIATEUR DE RUPTURE ET DÉPARTS DE BRÈCHE (57)
PAR INITIATEUR DE RUPTURE
SURVERSE EROSION AFFOUILLEMENT SURVERSE EROSION AFFOUILLEMENT
INTERNE INTERNE
1840–1886 36 % 64 % 0% 19 % 73 % 8%
1993–2016 25 % 75 % 0% 17 % 83 % 0%
(0 % ?) (100 % ?)
1840–2016 33 % 67 % 0% 18 % 76 % 6%
Barrages 48 % 46 % Glissement 6 %
p.m. [1]
3.3. ANALYSE SUCCINCTE ET ENSEIGNEMENT POUR L’ANALYSE DE RISQUE
Pour la période 1993–2016, toutes les brèches par érosion interne sont
dues à l’érosion de conduit [4] à l’exception d’une, pour laquelle l’initiateur reste
à confirmer. Pour la période 1840–1886, la dénomination par les ingénieurs de
l’époque de “brèche par infiltration” laisse un doute sur la possibilité de cas d’érosion
régressive, notamment au droit de paléo-chenaux. Néanmoins, compte tenu du
retour d’expérience récent et l’évaluation de la probabilité de brèche par érosion
régressive [7], l’érosion de conduit semble être le mode de rupture privilégié des
brèches par érosion interne de la période ancienne. L’étude a également montré
que l’affouillement pouvait être une source de désordre, voire de départ de brèche
quand la crue était longue (crue de 1856), mais sans conduire à la formation de
brèche.
L’analyse rétrospective des brèches récentes par érosion de conduit a montré

que, dans tous les cas, la contrainte critique (caractérisant le seuil d’érosion) des
matériaux était inférieure à 10 Pa, voire moins. Les Index d’érosion (caractérisant
la cinétique d’érosion) étaient, quant à eux, de l’ordre de 3 (entre 2,7 et 3,2). A
contrario, un cas de grosses fuites dans des terriers de lapins a été constaté et n’a
pas donné lieu à brèche quand la contrainte critique était de l’ordre de 50 Pa. Les
gradients initiateurs, lors du démarrage du processus de brèche dans les terriers de
blaireaux ou de lapins, ont tous été compris entre 0,01 et 0,05. Ils ont été légèrement
supérieurs à 0,1 pour les brèches par érosion de conduit le long des ouvrages
traversants. On se reportera à [4] pour l’évaluation quantitative de la probabilité de
brèche par érosion de conduit. L’analyse a montré également que les paramètres
«charges d’eau mesurées depuis le côté protégé» et «revanche/hauteur (R/H)»,
étaient fortement corrélés entre eux et jouaient un rôle primordial dans le processus
de brèche [5].
Concernant la surverse, l’étude a montré, qu’avec une hauteur de chute

inférieure à 1,5 m et une végétation dense en pied d’ouvrage permettant de ralentir
253
Q.105 - R.17
les vitesses, le processus de brèche par surverse n’était pas allé à son terme avec
des lames d’eau inférieures à 20 cm et des durées de déversement de plusieurs
dizaines d’heures. A contrario avec des hauteurs de chute de 4 m, le processus
de brèche par surverse est allé à son terme avec moins de 10 cm de déversement
pendant quelques heures.
3.4. ANALYSE DU PROCESSUS DE FORMATION DE BRÈCHE ET LEÇONS TIRÉES

POUR L’ANALYSE DE RISQUES ET LA GESTION DE CRISE
L’étude accidentologique a montré que l’initiation d’un mode de rupture

n’engendrait pas automatiquement une brèche et que cette dernière était le
processus de plusieurs étapes décrites dans [1] et quantifiées dans [4], [7] et [6].
Tableau 2
Désordres, départs de brèche et brèche par initiateur de rupture
Damages, breaches initiation and breaches by failure initiation mode
INITIATEUR DE RUPTURE CRUE NOVEMBRE 2002 CRUE DÉCEMBRE 2003
DES. URG. DP.BR. BR. DES. URG. DP.BR. BR.
surverse 6 5 1 0 13 4 4 3
érosion de conduit par terrier 11 9 1 1 12 5 0 0
érosion de conduit le long d’un OHT 1 1 0 0 6 4 1 0
érosion contact (ancienne brèche) 3 1 0 0 8 0 0 0
érosion interne – autre 0 0 0 0 1 0 1 1
affouillement 4 0 0 0 11 1 0 0
glissement côté terre 1 0 0 0 0 0 0 0
glissement côté fleuve 1 1 0 0 1 0 0 0
défaillance fonctionnelle (OHT) 4 4 0 0 9 2 0 0
Total 31 21 2 1 61 16 6 4
En novembre 2002, sur les 33 désordres relevés, un seul a donné lieu à un

départ de brèche et un autre à une brèche. En décembre 2003, sur les 67 désordres
relevés, 2 ont donné lieu à des départs de brèche et 4 à des brèches. Le tableau
(2) donne respectivement pour les crues de novembre 2002 (période de retour 20
ans, durée de crue 15 jours) et de décembre 2003 (période de retour de 100 ans,
durée de crue 3 jours), le nombre respectivement de désordres (DES.), le nombre
d’interventions d’urgence (URG.), le nombre de départs de brèche (DP.BR.) et le
nombre de brèches (BR.) par initiateur de rupture.
Ces statistiques croisées avec celles de 1993 et 1994 [6] permettent

d’appréhender, selon l’initiateur de rupture, l’évolution possible d’un désordre en
départ de brèche, dans un premier temps et en brèche, dans un second temps.
Elles permettent également de mesurer l’impact des interventions d’urgence. Pour
ces deux crues, on constate globalement le succès des interventions d’urgence sur:
254
Q.105 - R.17
Fig. 4
Exemple d’arbre d’événements [6]
Example of event tree [6].
les surverses localisées, même en cas de gradient de montée de crue rapide; les
terriers d’animaux fouisseurs visibles; les défaillances fonctionnelles des ouvrages
hydrauliques traversants; les érosions de conduit le long des conduites traversantes.
A contrario, les interventions d’urgence n’ont pas été possibles quand les terriers
étaient masqués. Dans ce scénario décrit en détail et évalué dans [4], le terrier par-
tiellement colmaté apparait soudainement après claquage hydraulique du bouchon
de colmatage. Le gradient hydraulique est alors sensiblement supérieur au gradi-
ent critique et la cinétique d’érosion trop rapide pour envisager une intervention
d’urgence. L’affouillement et l’érosion de contact très présentes n’ont, quasiment
pas nécessité d’interventions d’urgence compte tenu de leur cinétique d’érosion
très lente. Il est en est de même pour le glissement.
L’étude accidentologique des ouvrages et des conséquences de leur ruine

sur la zone protégée a permis de déterminer pour l’analyse de risques liés aux
ouvrages en remblai, 11 scénarios de brèche impliquant un ou plusieurs modes
de rupture et 1 scénario de défaillance fonctionnelle [6]. Elle a également permis
d’établir des arbres d’événements (Fig. 4) pour chacun des scénarios tenant compte
non seulement des phases d’initiation caractérisées par l’apparition des désordres
mais également des autres étapes, que sont la continuation et la progression [1].
Ces arbres d’événements ont également intégré les étapes de détection des désor-
dres et d’intervention d’urgence, qui ont permis d’éviter, lors des crues de 2002 et
2003 et plus récemment en 2014 et 2016, la formation ou l’extension de brèches.
L’évaluation des scénarios de brèches a été conduite suivant une méthode proba-
biliste calée sur le retour d’expérience accidentologique. On se reportera sur [4], [6]
et [7] pour de plus amples détails.
255
Q.105 - R.17
4. L’AGLY
4.1. HISTORIQUE DES DIGUES ET DESCRIPTIF DES

ÉVÉNEMENTS RÉCENTS
Les 13 derniers kilomètres aval du fleuve côtier méditerranéen Agly sont

endigués. Les crues de l’Agly sont généralement dues à des “épisodes méditer-
ranéens”, de violents systèmes pluvio-orageux qui apportent des précipitations
intenses, jusqu’à plusieurs mois de précipitations en quelques heures. Le bassin
à l’amont du couloir endigué a une superficie de 1044 km2 . Entre 1969 et 1974, le
lit a été recalibré en portant sa largeur de 30 m à 65 m, et un endiguement continu
d’environ 3 m de hauteur a été érigé au sommet de chaque berge reprofilée. Ces
ouvrages protègent aujourd’hui environ 30 000 personnes. L’ouvrage a subi de nom-
breuses crues (7 fois en 18 ans) ayant causé des désordres [8] [9] dont 4 ont conduit
à des surverses par-dessus les digues (1999, 2013, 2014 et 2020). Ces crues ont
conduit le gestionnaire à engager à partir de 2009 un diagnostic approfondi. Il a
ainsi été mis en évidence une capacité relativement faible du couloir endigué de
l’ordre de 700 à 900 m3 /s (près de deux fois moins que la valeur prévue par les
études de conception réalisées dans les années 1960) qui correspond à une crue
de période de retour comprise entre 5 et 10 ans (contre 25 ans selon les études de
conception). Il a aussi été démontré une fragilité structurelle des ouvrages et de leur
fondation, dont les compositions sont hétérogènes avec la présence de matériaux
graveleux perméables (k ∼ 10−3 m/s) et de sables fins propres uniformes partic-
ulièrement sensibles à l’érosion interne. Les conclusions du diagnostic ont conduit
à estimer que la rupture était quasi-certaine pour une crue décennale. Ce point a
été vérifié par la suite avec les crues de mars 2013, mars 2014 et janvier 2020. La
première a généré des surverses sur plusieurs kilomètres, deux brèches dont une
de plus de 80 m de longueur et de nombreux désordres par érosion interne de la
digue et de sa fondation (dont des centaines de sandboils, certains ayant généré
des instabilités de talus). La seconde a conduit à de légères surverses et de nou-
veaux désordres par érosion interne et la troisième a encore une fois conduit à des
surverses par-dessus les digues et de nouveaux désordres (érosions et sandboils)
sans pour autant causer de brèche.
4.2. PROBLÉMATIQUE DES INCERTITUDES DANS LES

DIAGNOSTICS
À notre connaissance, la fréquence des événements ayant conduit à des

écoulements à plein bord avec surverses (4 crues en 20 ans) fait de l’Agly un cas
unique en France. Elle permet de confronter les observations aux prédictions de
comportement que l’on pouvait en faire en appliquant l’état de l’art pour les dif-
férents modes de rupture. Le cas de l’Agly est en particulier intéressant pour mettre
256
Q.105 - R.17
en évidence les incertitudes sur la capacité à évaluer les niveaux et vitesses corre-
spondant à un débit donné et celles sur l’évaluation des probabilités des modes de
rupture, en particulier les deux modes prédominants: surverse et érosion interne.
Concernant le premier point, des jaugeages ont été réalisés par le service de prévi-
sion des crues pour 3 crues ayant conduit à des surverses, ces derniers ayant été
réalisés à des hauteurs voisines du pic des crues (2013, 2014 et 2020). Les incer-
titudes sur ces jaugeages ont été estimées comprises entre 5 et 15 %. Par ailleurs,
de nombreux repères de crue ont été relevés dans le couloir endigué. L’exploitation
de ces données dans le cadre de plusieurs modélisations hydrauliques détaillées a
conduit aux conclusions suivantes:
• bien qu’elles se soient produites dans un laps de temps assez court, les
conditions d’écoulement se sont révélées instationnaires avec une capacité
maximale du couloir endigué (avant surverse) qui a varié dans une fourchette
encore incertaine (entre 700 et 900 m3 /s); les analyses, incluant la photo-
interprétation de la végétation dans le couloir, ont montré que l’impact de la
végétation était déterminant (réduction de la vitesse et sans doute fixation et
exhaussement des bancs dans le lit);
• il a été démontré que les coefficients de rugosité (déduits par modélisation
à partir des repères de crue et des jaugeage) pouvaient être drastiquement
réduits (jusqu’à des valeurs de 15 pour les coefficients de Strickler); à noter,
les études de conception avait retenu une rugosité correspondant à un canal
bien entretenu alors que le couloir endigué constitue aujourd’hui un corri-
dor écologique remarquable, en particulier du fait de la végétation; l’impact
des évolutions morphologiques demeurent encore incertain faute de données
suffisantes pour les caractériser.
Pour quatre crues les digues ont surversé sur plusieurs kilomètres et sur
chaque rive, avec des lames d’eau pouvant atteindre 0,3 m pendant plusieurs heures
sur des talus majoritairement enherbés avec des remblais érodables (sables peu
limoneux) et une hauteur de chute de l’ordre de 2 à 3 m. Les érosions sur les talus
déversant ont été importantes avec localement des talus raidis jusqu’à atteindre
des pentes subverticales sur des portions de plusieurs dizaines de mètres. Parmi
ces quatre crues, une seule brèche (2013) s’est produite dans une zone ayant
potentiellement surversé, sans qu’il soit possible de l’associer à l’érosion externe
au vu des observations de terrain (laisses de crue sous le niveau de la crête aux
extrémités de la brèche, nombreux désordres liés à l’érosion interne en aval immé-
diat de la brèche). Dans ces conditions d’écoulement et une cinétique rapide qui a
conduit le gestionnaire à interdire de manière précoce la surveillance linéaire des
ouvrages et toute autre intervention en crue, les approches classiquement utilisées
pour prédire le comportement auraient toutes conduit à évaluer la rupture comme
certaine pour les 4 crues vis-à-vis des modes de rupture prépondérants: surverse
(seuils critiques d’érosion externe par surverse des abaques du CIRIA [10, section
8.4.2.1] dépassés) et érosion interne (nombreux et importants sandboils, terriers, et
ouvrages traversants sans disposition à l’interface, gradients 5 à 10 fois supérieurs
aux gradients critiques d’érosion déduits des essais de type HET réalisés sur les
matériaux majoritaires dans la digue et sa fondation).
257
Q.105 - R.17
4.3. RÉPERCUSSION DE CES INCERTITUDES SUR LES CONCEPTIONS DE

CONFORTEMENTS OU RÉPARATIONS
En ce qui concerne la conception et les pratiques des réparations ou conforte-

ments de ces ouvrages, les incertitudes relatives aux conditions d’écoulements
et à leurs instationnarités doivent être mieux prises en compte. D’autant qu’elles
s’ajoutent aux incertitudes fortes relatives à l’attribution d’une fréquence à un débit
donné. Les axes d’amélioration les plus évidents sont les suivants:
• encourager et assister les maîtres d’ouvrage pour l’amélioration des suivis

hydrométrique (jaugeage) et morphologique des couloirs endigués (suivi
pluriannuel de la végétation, des atterrissements, de la topographie et
bathymétrie);
• associer à tout projet de confortement un plan de gestion morphologique du
cours d’eau (traitement de la végétation, gestion des bancs …);
• mobiliser des compétences en morphologie des cours qui dépassent le cadre
habituel de celles mobilisées pour étudier ce type de projet;
• mieux prendre en compte ces incertitudes dans la conception: sur un couloir
endigué de plusieurs km à plusieurs dizaines de km, adopter partout des
marges de sécurité identiques (comme par exemple une revanche de 0,5 m
par rapport à un débit donné) n’est pas cohérent avec la variabilité spatiale
des évolutions morphologiques et ce qu’il est raisonnable de faire (du point de
vue écologique) en matière de gestion morphologique des cours d’eau;
• réduire ces incertitudes par la mise en place de déversoirs.
En ce qui concerne l’évaluation des performances des ouvrages existants,

on relève souvent un certain décalage entre les approches théoriques utilisées par
l’ingénierie et le retour d’expérience. Pour un ouvrage existant, les gestionnaires
doivent tout mettre en œuvre pour capitaliser ce retour d’expérience (notamment
lors des relevés post-crue) et l’ingénierie doit en prendre compte dans les approches
développées. Une approche théorique conduisant à un résultat qui n’est pas con-
firmé par le REX est une approche inadaptée. Selon notre expérience, ces écarts
peuvent être évités par une meilleure appréciation des paramètres régissant les
phénomènes, de l’enchainement des phénomènes conduisant à la rupture (à titre
d’exemple, une inondation par une brèche induite par une instabilité ne se résume
par un simple calcul de stabilité) et des spécificités de l’ouvrage étudié.
5. PRINCIPALES LEÇONS TIRÉES DES ÉVÉNEMENTS PASSES
Les conséquences qui en ont été tirées par la communauté furent multiples et
le paysage a bien changé depuis la fin du XXème siècle en ce qui concerne la prise
en compte des digues de protection, qui jouent un rôle essentiel dans la gestion du
risque inondation (en association bien sûr avec la panoplie des mesures non struc-
turelles et d’autres mesures structurelles comme les barrages ou bassins écrêteurs
258
Q.105 - R.17
de crues). Les trois principales des leçons tirées de ces REX sont présentées dans
cette section: l’importance de la gouvernance, structures gestionnaires et réglemen-
tation, la nécessaire gestion des surverses et l’utilisation de l’accidentologie pour
l’analyse de risque.
5.1. LA GOUVERNANCE
Nombre d’événements passés ont mis en évidence le rôle important voire

essentiel du gestionnaire des systèmes de digues pour leur efficacité et leur sûreté.
Sur le Rhône aval entre autres, les structures de gestion mises en place au XIXème
siècle ont montré leurs limites lors des crues de 1993 et 1994, ce qui a conduit à
la création et à l’organisation du Symadrem (voir 3.1). Le gestionnaire a en charge
la surveillance et l’entretien des ouvrages, y compris en période de crue, afin de
garantir leur efficacité et d’éviter la formation de brèches.
En parallèle aux évolutions locales des structures gestionnaires et de leurs

pratiques, souvent initiées par des événements de crue majeures et d’inondations
résultantes, l’Etat a pris conscience de la nécessité de cette bonne gestion à l’échelle
nationale. Après avoir engagé depuis 1999 un recensement des ouvrages, de leurs
zones protégées et de leurs gestionnaires (bases de données nationales gérées
par l’administration), une réglementation concernant les ouvrages de protection
contre les inondations a progressivement été développée [11]. Cette réglementation
a d’abord été purement technique, puis a précisé les rôles respectifs du gestionnaire
et de l’Etat, et enfin a clarifié l’attribution de la responsabilité de la gestion des
systèmes d’endiguement et autres ouvrages de protection contre les inondations
(nouvelle compétence GEMAPI).
La réglementation, comme la bonne pratique des gestionnaires conduit à

réaliser de manière régulière diagnostics et analyses de risque, ce qui fournit
des outils d’aide à la décision pour la définition des programmes de travaux et
l’organisation de la surveillance des ouvrages.
5.2. GESTION DES SURVERSES
Parmi les phénomènes à l’origine des ruptures de digues, la surverse est un

des plus importants, à la fois en termes de statistique qu’en termes de mécanismes
qui y sont liés: l’érosion par surverse d’un ouvrage en remblai, phénomène auto-
accélérant est bien entendu le plus important, mais même en l’absence d’érosion
une surverse peut amener à déstabiliser l’ouvrage en saturant sa partie aval. Il est
donc nécessaire de concevoir des ouvrages, ou plutôt des SYSTEMES, résilients
à la surverse, ce qui amène à distinguer différents niveaux de sollicitation [12] [13,
chapitre 3]: un niveau de protection, en dessous duquel l’eau ne pénètre pas dans la
259
Q.105 - R.17
zone protégée, un niveau de sûreté, en dessous duquel l’eau peut pénétrer dans la
zone protégée avec une très faible probabilité de brèche, donc de manière contrôlée,
et un niveau de danger au-delà duquel il y a un risque élevé de rupture.
Sur les levées de la Loire, après la crue de 1866, l’ingénieur Comoy a conduit
la mise en œuvre du programme de déversoirs régulateurs de crue sur les bases
suivantes:
• les ouvrages en remblai ne résistent pas à la surverse et la rehausse des

levées ne permet pas de résoudre cette fragilité;
• l’approche alternative de faire «la part de l’eau» avec l’invention de «déversoirs
modernes» de grande dimension (Fig.5) permettant au fleuve de s’étendre
dans son lit majeur (zone d’expansion de crue) tout en protégeant les ouvrages
(déchargeoir);
• l’association du public pour les ouvrages de gestion de surverse vers les vals
habités.
Faisant suite aux réflexions engagées dès le XVIe siècle sur les déchargeoirs,
ces nouveaux ouvrages devaient avoir trois fonctions: la maîtrise des brèches,
l’écrêtement des eaux et la réduction de la pression sur les digues. Le programme
a permis la réalisation 8 déversoirs entre 1870 et 1892 sur les 19 prévus initiale-
ment. À noter que les projets ont été soumis à l’avis des riverains entre 1868 et
1871 sous la forme d’enquête d’utilité publique (démarche également innovante à
cette époque). Après quarante ans, ce programme a été arrêté, mais les grands
principes de réflexion demeurent et sont toujours d’actualité pour l’étude de gestion
de surverse d’un système d’endiguement [14].
Fig. 5
Déversoir de type Comoy à Jargeau sur la levée d’Orléans
Comoy spillway at Jargeau on the Orleans levee
(Source DREAL CVL/DETL).
260
Q.105 - R.17
Les digues du delta du Rhône ont été sensiblement rehaussées après les
grandes inondations de 1840 et 1856. Qualifiées d’insubmersibles à l’époque, ces
ouvrages, qui protègent plus de 100 000 personnes et le triple en saison estivale,
ont montré leur limite lors des crues de 1993, 1994, 2002 et 2003. Durant ces
événements, 17 départs de brèche et 12 brèches ont été recensés [5]. Au lendemain
de la crue centennale de 2003, l’État et les régions ont mis en place un vaste de plan
de prévention des inondations à l’échelle du bassin versant, appelé Plan Rhône, qui
prévoit notamment la rénovation complète des digues du delta du Rhône sur 20 ans.
Plutôt que de rehausser les digues, il a été opté pour une solution d’acceptation
de l’inondation pour des crues rares (période de retour 100 ans en amont d’Arles
et 50 ans en aval) tout en considérant la formation de brèches comme inacceptable
jusqu’à des événements exceptionnels (période de retour 1000 ans). Ce choix passe
notamment par la réalisation de digue résistante à la surverse, qui consiste à ren-
forcer avec des enrochements bétonnés le talus de digue côté val de manière à résis-
ter aux vitesses élevées en cas de surverse. En amont et aval, les digues sont calées
50 cm au-dessus de la crue millénale pour éviter tout risque de contournement en
phase de déversement. On se reportera à [16] [17] pour plus de détails.
Sur l’Agly, selon les études de conception (1967), l’aménagement était censé
pouvoir contenir une crue de période de retour 25 ans, dont le débit était estimé à
1250 m3 /s, en disposant d’une revanche de 0,5 m par rapport à la crête de digue. Le
projet initial prévoyait, à l’extrémité amont du couloir endigué, deux déversoirs de
600 m (un sur chaque rive) permettant de limiter le débit. Ces déversoirs n’ont jamais
été réalisés. Un programme de confortement a été proposé dès 2014 avec un ou
plusieurs déversoirs et sans modification du niveau de protection après l’étude de
multiples scénarios et la réalisation d’analyse multicritère. Malgré la criticité du sys-
tème, la probabilité de rupture et la gravité de la rupture étant toutes deux élevées,
ce projet n’est pas aujourd’hui réalisé du fait de:
• la difficile acception sociale de déversoirs de sécurité,

• la difficulté à récupérer le foncier nécessaire aux travaux pour limiter leurs
impacts environnementaux,
• l’importance du montant de l’investissement (estimé de l’ordre de 45 MC = ) au
regard du niveau de protection obtenu (période de retour comprise entre 5 et
10 ans).
20 ans après les premières surverses (1999) et malgré les enjeux et la récur-
rence des crues ayant nécessité la mise en sécurité des populations (1999, 2013,
2014, 2020), aucune solution n’a fait l’objet d’un consensus et de nouvelles études
vont être engagées pour proposer de nouvelles alternatives.
Beaucoup de systèmes d’endiguement, voire une grande majorité, en France,

mais aussi à l’étranger, ne sont pas capables de gérer correctement les surverses.
Même si les digues sont en bon état et résistent pour un niveau d’eau atteignant leur
crête, au-delà de ce niveau de protection il y a un risque élevé voire certain de rup-
ture et un risque d’inondation résultant catastrophique, vu les conséquences d’une
261
Q.105 - R.17
brèche à pleine charge. Il est souhaitable que l’approche présentée de concep-

tion de systèmes de protection contre les crues fluviales résilients à la surverse se
généralise. En corollaire il faut faire prendre conscience à la société d’une limite à la
protection, assortie d’une gestion contrôlée des surverses, avec ces conséquences
bien moindres que des ruptures en des lieux aléatoires. Cette approche est détaillée
dans un guide [18] dont la traduction en anglais est en cours.
5.3. L’APPORT DE L’ACCIDENTOLOGIE DANS LES ANALYSES DE RISQUE
L’analyse de risque des systèmes d’endiguement [13], qui se traduit au niveau

règlementaire à travers les Etudes de Dangers, est un outil très riche d’aide à la
décision. Néanmoins, s’agissant d’une activité récente, de nombreuses lacunes per-
sistent à la fois en termes de méthodes générales et de données locales. Le retour
d’expérience sur les accidents et incidents permet de pallier à un certain nombre de
ces lacunes. Le risque est une combinaison de l’aléa et des conséquences. On a
vu plus haut (3.4) comment le retour d’expérience permettait d’identifier et de prob-
abiliser les aléas de rupture, il permet aussi de caractériser les brèches potentielles
et les inondations qui en résultent.
La modélisation de l’inondation dans la zone protégée en cas de brèche, utile

aussi bien pour la connaissance du risque résiduel que pour la préparation à la ges-
tion de crise, nécessite de connaître les caractéristiques de la brèche (géométrie et
dynamique) ainsi que des données de calage relatives à la zone protégée. Dans le
cadre des études de dangers des digues d’Orléans, de Tours et d’Authion [15],
l’analyse s’est notamment appuyée sur les données des crues du XIXe siècle
pour définir à partir des dimensions des fosses historiques la zone de dissipation
d’énergie derrière une brèche, estimée à 100 fois la hauteur de la digue (longueur
dans le sens de l’écoulement) et pour reconstituer l’hydrogramme de la brèche
(Fig.6) survenue à Jargeau en 1856 (débit de 3700 m3 /s et volume total de 270 Mm3 ).
Fig. 6
Hydrogrammes historiques de la Loire en amont et en aval de la brèche de
Jargeau et reconstitution de l’hydrogramme de la brèche
Historical hydrographs of the Loire upstream and downstream of the Jargeau
breach and reconstitution of the breach hydrograph
(Source: Archives départementales du Loiret, Cerema).
262
Q.105 - R.17
Sur le Petit Rhône, l’analyse des largeurs de brèches et des hauteurs de

charge [5] ont permis de déduire la loi suivante:
Largeur de brèche = 72 × Hauteur de charge −123 m.
Sur l’ensemble du Rhône aval, la largeur cumulée de brèches lors d’une même
crue et le volume de déversement en fonction du débit en tête de delta (Fig. 7) a
permis de constater l’évolution exponentielle de ces paramètres primordiaux pour
l’analyse de risques, notamment dans la définition des scénarios d’inondation par
brèche.
Fig. 7
Largeur cumulée de brèches et volumes de déversement
Cumulative width of breaches and spilled volumes.
À partir des REX sur les brèches sur les différents cours d’eau, ceux présen-
tés ici ou d’autres, il serait intéressant de réaliser une analyse comparative des
différents paramètres: géométrie et constitution de la digue, largeurs du lit mineur
et du lit majeur, débit dans le lit, géométries de la brèche et de la fosse d’érosion,
débit estimé dans la brèche…Cela permettrait entre autres que de mieux “caler” les
modélisations nécessaires lors des analyses de risque.
263
Q.105 - R.17
6. CONCLUSION ET PERSPECTIVES
Les nombreux événements de crue et d’inondation dans les territoires

endigués qui se sont succédé en France depuis la fin du XXème siècle, ainsi que
la connaissance relative aux événements plus anciens, constituent une base très
riche d’analyse. Les leçons qui ont d’ores et déjà été tirées de ces événements sont
multiples:
• La gouvernance a un impact essentiel sur la sécurité et l’efficacité des

ouvrages: importance d’un gestionnaire compétent et d’un cadre réglemen-
taire national,
• La gestion des surverses au-delà du niveau de protection des systèmes est
fondamentale pour minimiser les dommages lors des crues majeures,
• Les diagnostics et l’analyse de risque bénéficient largement des informations
tirées du retour d’expérience (REX), aussi bien pour la mise au point des
méthodes que pour l’obtention de données sur des cas particuliers. Entre
autres, le REX permet de prendre conscience d’incertitudes, et dans des cas
favorables d’évaluer celles-ci.
• L’ingénierie de conception des ouvrages s’adapte aux mécanismes de rupture
et à la connaissance qu’on peut en avoir et s’enrichit au fur et à mesure des
REX. Les méthodes de confortements et de réparations utilisées par la pro-
fession sont partagées en son sein, en faisant l’objet d’un groupe de travail
(GT) du CFBR. Le livrable de ce GT, qui sera disponible en 2021, consistera
en un document classiquement rédigé en chapitres et décrivant les principes
des méthodes de confortement et de réparation. Il sera lisible de manière
autonome mais s’accompagne de 32 fiches descriptives présentant dans le
détail des techniques de confortement/réparation, organisées par fonction
(étanchéité, drainage, filtration, stabilité, rehausse, protection contre l’érosion
externe) que les techniques remplissent de manière principale ou secondaire.
Ces fiches seront consultées individuellement au besoin. Ce document sera
évolutif et pourra être enrichi au fur et à mesure du retour d’expérience sur de
nouvelles techniques.
Au-delà de ces éléments acquis on peut aussi noter un certain nombre

d’évolutions en cours et de perspectives d’améliorations. La communauté des
digues s’organise pour progresser. La création de France Digues, l’association des
gestionnaires de digues, permet la professionnalisation de ce “nouveau” métier.
L’intégration des digues comme objet d’étude au sein du CFBR en France, et
de la CIGB au niveau international, permet les échanges entre ingénieurs de
tout type d’intervenant sur les ouvrages hydrauliques (gestionnaires, consultants,
entrepreneurs, administrations). On peut aussi noter la mise en place ces dernières
années de sessions de formations dédiées aux digues, comme de nombreux pro-
grammes de recherche liés à ces ouvrages. Parmi les pistes de progrès identifiées
sur la base du REX, notons l’intérêt qu’il y aurait à lier les projets de confortements
avec des plans de gestion de la morphologie des couloirs endigués.
264
Q.105 - R.17
Les connaissances sur les ouvrages et les événements historiques ou récents

qu’ils ont rencontrés sont donc une richesse qu’il faut continuer à capitaliser, à
l’échelle locale comme à l’échelle nationale, pour assurer le suivi de l’accidentologie.
Une certaine “normalisation” dans la manière de présenter et d’analyser les événe-
ments est souhaitable pour leur utilisation et leur pérennisation sur des bases
partagées. Mais, au final, il faut garder à l’esprit de rester humble et attentif pour
que la collectivité garde en mémoire sur le long terme tous ces événements passés
et les éléments présentés ici: on a en effet malheureusement souvent constaté par
le passé un relâchement de l’attention après une longue période sans crues…
REFERENCES
[1] ICOLD, 2015. Internal erosion of existing dams, levees and dikes, and their
foundations – bulletin n◦ 164 – Volume 1: internal erosion processes and
engineering assessment.
[2] Blondiaux P. & Patouillard S. 2019. Base de données sur les brèches
historiques de la Loire, stage Master I. DREAL CVL
[3] Fournier M. 2008. Quelle place pour les riverains dans la gestion des
inondations de Loire – les leçons du passé – Freude am Fluss
[4] Mallet T., Outalmit K. & Fry J.J. (2014) Probability of failure of an embankment
by internal erosion using the Hole Erosion Test. ICOLD BALI International
Symposium
[5] Mallet, T., Fry, J.J., Tourment, R., Mériaux, P. Accidentologie des digues du
delta du Rhône de 1840 à nos jours. Digues maritimes et fluviales de protection
contre les inondations – 3ème colloque national – Digues2019 – 20-21 mars
2019 – Aix-en-Provence
[6] Mallet T., Dast C. Requi M. Chardès C. Castagnet A. et Fry J.J. (2018). Etude
de dangers du système d’endiguement rive gauche du delta du Rhône –
ICOLD Vienna Congress
[7] Mallet T., Fry J.J. (2016) – Probabibility of failure of an embankment by

backward erosion using the formulas of Sellmeijer and Hoffmans – ICOLD
Johannesburg International Symposium
[8] Tourment, R., Benahmed, N., Nicaise, S., Mériaux, P., Salmi, A., Rougé, M. –
2018. Retour d’expérience sur les désordres des digues de l’Agly maritime.
Analyse des phénomènes de “sand-boils”. CIGB, 26ème congrès.
[9] CFBR, 2021, Barrages, le savoir-faire français, Chapitre 9 “Digues de

protection contre les inondations” (20 pages). ISBN 979-10-96371-11-2
265
Q.105 - R.17
[10] CIRIA, MEDDE, USACE – 2013. The International Levee Handbook. Van
Hemert, H., Igigabel, M., Pohl, P., Sharp, M., Simm, J., Tourment, R., Wallis,
M. (ed.), CIRIA, Londres, 1349 p.
[11] Tourment et al. – 2021 – The new French régulation on flood protection
structures; conséquences on risk management. FLOODrisk2020 conference,
Budapest, 11 p.
[12] Mallet et al., 2013. Niveaux de protection, de sûreté et de danger pour

les digues fluviales. Digues maritimes et fluviales de protection contre les
submersions 2ème Colloque national, Aix-en-Provence, 12–14 juin 2013
[13] Tourment, R., Beullac, B. – 2019. Inondations – Analyse de risque des sys-
tèmes de protection – Application aux études de dangers. Tourment, R.,
Beullac, B. (ed.), Lavoisier, Paris, 356 p.
[14] Maurin, J., Patouillard, S., Tourment, R. – 2013. Les déchargeoirs au XVIIe
siècle, les déversoirs au XIXe et quelles perspectives pour la gestion des
surverses au XXIe siècle. Congrès SHF Événements extrêmes d’inondation,
Lyon, FRA. 9 p.
[15] Maurin, J., Boulay, A., Piney, S., Le Barbu E., Tourment, R. – 2012. Les
brèches des levées de la Loire: brèche de Jargeau 1856. Congrès SHF
Événements extrêmes fluviaux et maritimes, Paris, FRA. 8 p.
[16] Cheetham, M., Mallet, T., Chastel, E., Tourment R., Robustelli, P., Pelt, P.,
2015. Building a resilient system of defence against flooding from the Rhône.
Proceedings of the Institution of Civil Engineers-Water Management, vol. 168,
n◦ 2, p. 74–84
[17] Mallet et al., 2013, Sécurisation des digues du Rhône entre Beaucaire et
Arles: conception hydraulique du système et des digues résistantes à la sur-
verse. Digues maritimes et fluviales de protection contre les submersions
2ème Colloque national, Aix-en-Provence, 12–14 juin 2013
[18] Degoutte, G. (coord.) – 2012. Les déversoirs sur digues fluviales. Editions
Quae, Versailles, 184 p.
266
Q.105 - R.17
1. INTRODUCTION
Almost thirty years ago, the French community rediscovered the existence
and the importance of its heritage of flood protection levees, during floods on the
downstream Rhône that damaged the Camargue levees and flooded its delta. Since
then, episodes have multiplied in various geographic contexts, fluvial, coastal and
torrential, and sometimes repeatedly in certain sectors: downstream Rhône in Octo-
ber 1993, January 1994, December 2003, Gard and Rhône September 2002, Aude
1999, Agly 1999, 2013, 2014 and 2020, Xynthia storm 2010, torrential floods in
June 2013 in the Pyrenees, to name only the most important in terms of human or
material consequences. These recent events occurred after more than a century of
“tranquility”, the last major flood events involving levees dating from the 19th cen-
tury, even if other events may have led to awareness of the importance of levees
and more local recovery actions, for example on the Loire in the 1960s or in the
center-west of the country following the floods of 1982–83.
We present in this paper the feedback that has been carried out on three
sectors of French rivers, the middle course of the Loire, the Rhône delta and the
Agly maritime, and the related developments and perspectives about levees.
When analyzing reports about such events, it was noted that certain difficulties
of interpretation resulted from varying uses of the vocabulary, including the term
“breach”. Recently, terminology has become “standardized”. In the meaning of the
International Commission on Large Dams (ICOLD), a breach is a rupture which
water release can no longer be controlled. It is reached at the end of a process
involving mechanisms of deteriorations and failures, which develops in three stages
[1]: initiation, formation and expansion (enlargement and deepening). It is therefore
characterized by a phase of widening and deepening in the foundation, and water
inflows into the protected area which can no longer be interrupted. The breach in
progress is distinguished from the breach by the fact that the expansion phase is not
yet underway and/or that the release of water has been controlled or has no impact
on the protected area; these water inflows can be interrupted by a collapse of the
levee, an emergency response or the reduction in the hydraulic load associated with
the flood recession. A damage is defined as the observable or quantifiable sign of
a degradation of the initial state of the levee which can be the start of a rupture and
a breach process. A breach can be partial (only part of the height of the levee has
disappeared) or total (the entire height of the levee has disappeared).
1.1. LOIRE MIDDLE COURSE: FEEDBACK ON THE 19TH CENTURY FLOODS
The history of the Loire levees, whose origin dates back to the Middle Ages,
is marked by catastrophic events due to exceptional floods. The last exceptional
floods date back to the end of the 19th century; three such floods followed one
another in 1846, 1856 and 1866 with human and societal consequences that are
267
Q.105 - R.17
difficult to imagine today. The return period of these floods has been estimated at 170
years, which is roughly the time that separates us from these dramatic events. The
men who lived through these events also witnessed the destruction of the great Loire
levees, the reliability of which then seemed guaranteed by centuries of improvement
brought about by empirical know-how. This period was also one of major challenges
and innovations in the field of flood protection in the Loire basin.
1.2. DESCRIPTION (IN TERMS OF HYDRAULICS) OF THE HISTORICAL FLOODS AND

FLOODINGS OF THE 19TH CENTURY
The floods at the end of the 19th century which had a strong impact on the
Loire Moyen are qualified as mixed floods. As illustrated in Figure 1, this is a combi-
nation of Oceanic and Mediterranean floods (the origin of which is linked to weather
phenomena coming respectively from the Atlantic Ocean or the Mediterranean Sea).
Fig. 1
Représentation de la Loire Moyenne et schéma d’une crue mixte
Loire Moyenne location and mixed flood drawing
The 1846 flood was the most catastrophic in terms of loss of life and livestock.
The water level rose by 5 meters in 24 hours at the level of the Loire bridges and the
flood submerged the levees, causing more than a hundred breaches and flooding
the whole floodplain. In 1856 and 1866, the average Loire flow rates were locally
estimated to be between 7000 and 8000 m3 /s, compared to an average flow rate
of around 500 m3 /s. During these three events, 376 breaches on the levees were
identified. They form the basis of the database used for the analysis of the levees of
the Loire [2], the eight breaches or start of breaches observed during the following
floods (1895, 1902, 1968, 1982) not having had any comparable damage impact.
268
Q.105 - R.17
1.3. CAPITALIZATION OF EXPERIENCE: AN EXEMPLARY WEALTH
Following the 1846 flood, the engineers undertook substantial reinforcements

and implemented several innovative tools to monitor the river’s floods and capital-
ize accidentology related data: a network of mileposts and hydrometric gauges,
a detailed cartography representing the floodplain over nearly 600 km with levee
structure, known breaches and erosion pits from old breaches, as well as a general
elevation model of both sides of the floodplain. These tools have been used to record
data on the breaches of the two subsequent floods, as shown in Figure 2. Today,
these tools are still used for monitoring the evolution of the bed of the Loire and
for flood forecasting, supplemented by new developments: in particular hydraulic
modeling, data processing and geomatics.
Fig. 2
Extrait de la carte de 1850 au niveau d’une brèche de 1846 (en noir) surchargée
par 2 brèches de 1856 (en jaune)
Extract from the 1850 map at the location of a breach of 1846 (written in black) and
two breaches of 1856 (added in yellow)
1.4. COMPLEMENT THE ACTION OF LEVEES WITH UPSTREAM DAMS
After the 1856 flood, the strategy of making levees “non-submersible” was
called into question. To avoid breaches and stop large floods, two systems are
proposed: a set of flood storage reservoirs on the upstream Loire and on the
Allier (a Loire tributary) and/or flood regulating spillways in the levees of all the
269
Q.105 - R.17
Loire “Vals1 ” [3]. The flood of 1866 will lead to accelerate an ambitious program of
rehabilitation of the protection systems together with spillways.
Subsequently, the protection scheme will be effectively supplemented by the

commissioning of the Villerest flood control dam on the upstream Loire in 1985.
2. THE RHÔNE DELTA
2.1. CONTEXT
The Rhône Delta covers an area of 1500 km2 . It is protected by three river
levee systems (225 km) and a coastal levee system (50 km), managed by the Syn-
dicat Mixte Interrégional d’Aménagement des Digues du Delta du Rhône and Mer
(Symadrem), a public institution created in 1999 following the failure of the associa-
tions of riparian owners who had been managing the structures since the middle of
the 19th century.
The levee systems were created after the floods of 1840 and 1856 (return
periods estimated at 400 and 250 years respectively) in place of other even older
structures, some of which date back to the 12th century. Due to their construc-
tion technique (compaction with 15 kg manual tamping devices of, not taking into
account the optimum water content concept discovered in 1933 by Ralph Proctor)
and the “onion” effect due to successive rising phases, the Rhône levees are highly
exposed to failures by internal erosion [4]. As part of the hazard studies required
by French regulations, an accidentology analysis was carried out on all the entire
levee systems. The main results are presented below, and can be found in more
detail in [5].
2.2. STATISTICS ON RECENT AND OLD EVENTS
Figure 3 presents peak flows, at the head of the delta, of maximum annual
floods from 1830 to 2016. In red are the floods that initiated flooding of the protected
zone by breach. Since 1840, nine floods have led to flooding: 1840, 1841, 1843,
1856, 1993, 1994, 2002, 2003 and 2016. The figure also indicates the analyzed
periods: the early period from 1840 to 1886 and the recent period from 1993 to
2016. The intermediate period from 1886 to 1993, corresponding to the hydrological
“standby” of the Rhône, was not analyzed for lack of sufficiently detailed archive
documents to carry out a quality analysis. In addition to these floods leading to
1 in local terminology, a Loire “Val” is a part of the floodplain protected by a levee system.
270
Q.105 - R.17
flooding, six floods without flooding were also analyzed. The analysis of Figure 3
shows that the probability of flooding per breach was less than 5% for floods with
return periods of between 5 and 10 years. It was between 5 and 50% for floods
with a return period of between 10 and 20 years and greater than 50% for events
of 20 to 50 years. It is 100% for floods with return periods greater than 50 years.
These probabilities, derived from the observation of past floods, are considered a
good indicator for crisis management. On the 15 analyzed floods, 57 breaches, 57
breaches in progress and 133 damages were identified. 3 breaches caused by water
going back to the river were also recorded in 1840, which gives a total number of
117 breaches and breaches in progress in 176 years, which is consistent with the
assessment made within the framework of the hazard studies [6].
Fig. 3
Débits des crues et inondation de la zone protégée (1830–2016)
Flood discharges and inundation of the protected area (1830-2016).
Breaches are generally located on the Rhône and the upstream Petit Rhône,
where levees are the highest (4.5 to 7 m) and hydraulic loads the highest in the
past (3 to 6 m). Starts of breaches are, for their part, rather localized on the Grand
Rhône and the downstream Petit Rhône, where levees are lower (2 to 4.5 m) and
the hydraulic loads also lower in the past (1.5 to 2.5 m).
Table 1 below provides an overview by breach initiator according to the period.

It can be seen that the accidentology of levees in the Rhône delta follows a similar
trend to that of embankment dams [1] with a higher proportion of accidents linked
to internal erosion.
271
Q.105 - R.17
Table 1
Statistique des brèches et des départs de brèche par initiateur de rupture
Breaches and breach initiations by initiating failure mode
RETURN BREACHES (57) BREACHES (57) AND
PERIOD BY INITIATION CAUSE STARTS OF BREACHES (57)
BY INITIATION CAUSE
OVERFLOWING INTERNAL EXTERNAL OVERFLOWING INTERNAL EXTERNAL
EROSION EROSION, EROSION EROSION,
SCOURING SCOURING
1840–1886 36 % 64 % 0% 19 % 73 % 8%
1993–2016 25 % (0 % ?) 75 % (100 % ?) 0% 17 % 83 % 0%
1840–2016 33 % 67 % 0% 18 % 76 % 6%
Dams p.m.[1] 48 % 46 % Sliding 6 %
There were 19 breaches caused by overflowing throughout the period. 16

occurred in 1840 before the general rising of the levees and 3 during the centennial
flood of December 2003. They all were initiated under a very low water level (less
than 10 cm) but with a fall height greater than 4 m. For these three breaches in 2003,
certain heterogeneities were observed, after flooding, in the levee bodies. Internal
erosion mechanisms cannot therefore be ruled out, which could give a total of 100%
of breaches by internal erosion for the recent period. We also note two breaches by
overflow in 2002 and 2003 over a line of 2 km downstream of the delta (analyzed
below).
2.3. SUCCINCT ANALYSIS AND LESSONS FOR RISK ANALYSIS
Between 1993 and 2016, all breaches by internal erosion are due to concen-
trated leak erosion [4] with the exception of one, for which the initiator remains to be
confirmed. Between 1840 and 1886, the designation by the engineers of the time
of “breach by seepage” leaves some doubt on the possibility of backward erosion,
in particular at the place of former river beds. Nevertheless, taking into account
recent experience and the assessment of breach probability by backward erosion
[7], concentrated leak erosion seems to be the preferred mode of breaching by inter-
nal erosion in the past too. The study also showed that scour could be a source of
damage, even the start of a breach when the flood was long (1856 flood), but without
leading to a full breach.
The forensic analysis of recent breaches by concentrated leak erosion showed

that, in all cases, the critical stress (characterizing the erosion threshold) of the mate-
rials was less than 10 Pa, or even less. The Erosion Rate indexes (characterizing
the erosion kinetics) were on the order of 3 (between 2.7 and 3.2). On the other
hand, a case of large leaks in rabbit burrows was observed and did not lead to a
breach when the critical stress was of the order of 50 Pa. The initiating gradients,
when the breach process started in badger or rabbit burrows, all ranged from 0.01 to
0.05. They were slightly higher than 0.1 for breaches caused by pipe erosion along
272
Q.105 - R.17
the crossing structures. Refer to [4] for a quantitative assessment of the probability
of breach by concentrated leak erosion. The analysis also showed that the param-
eters “water loads measured from the protected side” and “freeboard/height (R/H)”
were strongly correlated with each other and played a key role in the breaching
process [5].
Regarding overflow, the study showed that with a fall height of less than 1.5 m
and dense vegetation at the foot of the structure making it possible to slow down the
speeds, the breaching process by overflow had not gone to its end with water layers
less than 20 cm and discharge times of several tens of hours. In contrast, with fall
heights of 4 m, the overflow breaching process was completed with less than 10 cm
of discharge during a few hours.
2.4. ANALYSIS OF THE BREACHING PROCESS AND LESSONS LEARNED FOR RISK
ANALYSIS AND CRISIS MANAGEMENT
The accidentological study showed that the initiation of a failure mode did
not systematically lead to a breach and that breaches were a process with several
steps described in [1] and quantified in [4], [7] and [6]. In November 2002, out
of the 33 identified damages, only one led to a start of breach and another to a
breach. In December 2003, out of the 67 damages identified, 2 gave rise to starts
of breaches and 4 to breaches. Table (b) presents respectively for the floods of
November 2002 (return period 20 years, duration of flood 15 days) and of December
2003 (return period of 100 years, duration of flood 3 days), the number respectively of
damages (DES.), emergency interventions (URG.), breaches in progress (DP.BR.)
and breaches (BR.) per breach initiator.
These statistics, crossed with those of 1993 and 1994 [6] make it possible to
assess, according to the breach initiator, the possible evolution of a damage first
Table 2
Désordres, départs de brèche et brèche par initiateur de rupture
Damages, breaches in progress and breaches by failure initiation mode
INITIATION CAUSE NOVEMBER 2002 FLOOD DECEMBER 2003 FLOOD
DES. URG. DP.BR. BR. DES. URG. DP.BR. BR.
overflowing 6 5 1 0 13 4 4 3
concentrated leak erosion in burrow 11 9 1 1 12 5 0 0
concentrated leak erosion 1 1 0 0 6 4 1 0
along a crossing pipe
contact erosion (ancient breach) 3 1 0 0 8 0 0 0
internal erosion- other 0 0 0 0 1 0 1 1
scour 4 0 0 0 11 1 0 0
land side sliding 1 0 0 0 0 0 0 0
water side sliding 1 1 0 0 1 0 0 0
functional failure (pipe) 4 4 0 0 9 2 0 0
Total 31 21 2 1 61 16 6 4
273
Q.105 - R.17
toward a breach in progress, then as a breach. They also make it possible to mea-
sure the impact of emergency interventions. For these two floods, the success of
the emergency interventions on: localized overflows, even in the event of a rapid rise
of water level; visible animals burrows; functional failures of through hydraulic struc-
tures; concentrated leak erosions along pipes. Conversely, emergency interventions
were not possible when the burrows were hidden. In this scenario described in detail
and evaluated in [4], partially plugged burrows appear suddenly after hydraulic frac-
turing of the plug. The hydraulic gradient is then significantly greater than the critical
gradient and the erosion kinetics too fast to consider an emergency intervention. The
very present scour and contact erosion hardly required any emergency interventions,
given their very slow erosion kinetics. It is the same for sliding.
The accidentological study of the structures and of the consequences of

their ruin on the protected area made it possible to determine for the risk analysis
associated with the embankment structures, 11 breach scenarios involving one or
more failure modes and 1 functional failure scenario [6]. It also made it possible to
establish event trees (Fig. 4) for each of the scenarios, taking into account not only
the initiation phases characterized by the appearance of damage but also the other
stages, which are continuation and progression [1]. These event trees also inte-
grated the phases of damage detection and emergency intervention, which made it
possible to avoid, during the floods of 2002 and 2003 and more recently in 2014 and
2016, the initiation or extension of breaches. The assessment of breach scenarios
was carried out using a probabilistic method based on feedback from experience.
Refer to [4], [6] and [7] for more details.
Fig. 4
Exemple d’arbre d’événements [6]
Example of an event tree [6]
.
274
Q.105 - R.17
3. THE AGLY RIVER
3.1. LEVEES HISTORY AND DESCRIPTION OF RECENT EVENTS
There are levees all along both banks of the last 13 kilometers downstream of
the Mediterranean coastal river Agly. The Agly floods are generally due to “Mediter-
ranean episodes”, violent rainstorm systems that bring intense rainfall, up to several
months of rainfall in a few hours. The basin upstream of the leveed corridor cov-
ers an area of 1044 km2 . Between 1969 and 1974, the river bed was recalibrated
increasing its width from 30 m to 65 m, and a continuous embankment of about 3
m in height was erected at the top of each re-profiled bank. These structures pro-
tect around 30,000 people today. They have suffered numerous floods (7 times in 18
years) causing disturbances [8] [9], 4 of which led to overflows over the levees (1999,
2013, 2014 and 2020). These floods led the manager to initiate an in-depth diagno-
sis starting in 2009. A relatively low capacity of the leveed corridor, of the order of
700 to 900 m3 /s, was highlighted (almost twice less than the value predicted by the
design studies carried out in the 1960s) which corresponds to a flood return period
between 5 and 10 years (versus 25 years based on the design studies). Structural
fragility of the structures and their foundations has also been demonstrated, the
composition of which is heterogeneous with the presence of permeable gravelly
materials (k ∼ 10−3 m/s) and clean, uniform fine sands particularly sensitive to inter-
nal erosion. The conclusions of the diagnosis led to estimate that failure was almost
certain for a ten-year flood. This point was subsequently verified with the floods of
March 2013, March 2014 and January 2020. The first one generated overflows over
several kilometers, two breaches, one of which is over 80 m in length and numerous
damages by internal erosion of the levee and its foundation (including hundreds
of sand boils, some of which have generated bank instabilities). The second led to
slight overflows and new damages by internal erosion and the third led once again to
overflows over levees and new damages (erosions and sand boils) without causing
any breach.
3.2. THE ISSUE OF UNCERTAINTIES IN ASSESSMENTS
To our knowledge, the frequency of events leading to full-edge runoff with

overflows (4 floods in 20 years) makes Agly a unique case in France. It makes it
possible to compare the observations with the predictions of behavior that could be
made by applying up-to-date assessment methods for the different failure modes.
The case of Agly is particularly interesting for highlighting the uncertainties on
the ability to evaluate the levels and speeds corresponding to a given flow and
those on the evaluation of the probabilities of failure modes, in particular the
two predominant modes: overflow and internal erosion. Regarding the first point,
gauging was carried out by the flood forecasting service for 3 floods that led to
overflows, carried out at heights close to the flood peak (2013, 2014 and 2020).
275
Q.105 - R.17
The uncertainties on these gauges have been estimated to be between 5 and

15%. In addition, many flood marks were noted in the leveed corridor. The use
of this data in the context of several detailed hydraulic models led to the following
conclusions:
• although they occurred in a fairly short time, the flow conditions turned out to
be unsteady with a maximum capacity of the leveed corridor (before overflow)
which varied within a range that is still uncertain (between 700 and 900 m3 /s);
the analyzes, including photo-interpretation of the vegetation in the corridor,
showed that the impact of the vegetation was decisive (reduction in speed and
probably fixing and raising shoals in the bed);
• it has been shown that the roughness coefficients (deduced by modeling from
flood benchmarks and gauging) could be drastically reduced (down to values
of 15 for Strickler’s coefficients); It should be noted that the design studies
retained a roughness corresponding to a well-maintained canal, whereas the
leveed corridor today constitutes a remarkable ecological corridor, in particular
due to the vegetation; the impact of morphological changes is still uncertain
due to the lack of sufficient data to characterize them.
For 4 floods, levees overflowed for several kilometers and on each bank, with
loads of up to 0.3 m for several hours on erodible embankments (sands with little
silt) slopes, only protected by grass, with a drop height of about 2 to 3 m. Erosion
on the land side slopes was significant with locally resulting steep slopes until they
reached subvertical slopes over portions of several tens of meters. Among these
four floods, only one breach (2013) occurred in an area that potentially overflowed,
without it being possible to associate it with external erosion in view of field obser-
vations (flood marks under the crest level at both ends of the breach, numerous
damages linked to internal erosion immediately downstream of the breach). Under
these flow conditions and rapid kinetics, which led the manager to prohibit at an
early stage inspection of the levees and any other flood intervention, conventionally
approaches used to assess failure would have all led predict certain failure for the
4 floods with respect to the predominant failure modes: overflow (critical thresholds
of external erosion by overflow of the CIRIA charts [10, section 8.4.2.1] exceeded)
and internal erosion (numerous and important sand-boils, burrows, and crossing
structures without provision at the interface, gradients 5 to 10 times greater than the
critical erosion gradients deduced from HET type tests carried out on the dominant
materials in the levee and its foundation).
3.3. CONSEQUENCES OF THESE UNCERTAINTIES ON THE DESIGN OF

REHABILITATION OR REPAIR
With regard to the design of repairs or reinforcements of these structures, the

uncertainties relating to flow conditions and their instability must be better taken into
account, all the more so as they are added to the strong uncertainties relating to the
276
Q.105 - R.17
allocation of a frequency to a given flow. The most obvious areas for improvement
are as follows:
• encourage and assist levee managers to improve hydrometric (gauging)

and morphological monitoring of leveed corridors (multi-year monitoring of
vegetation, landings, topography and bathymetry);
• include a watercourse morphological management plan (vegetation and banks
management, …) in any reinforcement project;
• involve specialists skilled in watercourses morphology, beyond the usual level
of those mobilized in this type of project;
• take better account of these uncertainties in design: on a levee corridor of
several km to several tens of km, adopt identical safety margins everywhere
(such as for example a freeboard of 0.5 m in relation to a given flow) is not
consistent with the spatial variability of morphological evolutions and what it is
reasonable to do (from an ecological point of view) in terms of morphological
management of watercourses;
• reduce these uncertainties by installing spillways.
When it comes to assessing the performance of existing structures, there

is often a certain discrepancy between the theoretical approaches used by engi-
neering and feedback. For an existing structure, managers must do everything
possible to capitalize on this feedback (especially during post-flood surveys) and
engineering must take this into account in the development of methods. A theoret-
ical approach leading to a result that is not confirmed by feedback is an unsuitable
approach. According to our experience, these deviations can be avoided by a better
appreciation of the parameters governing the phenomena, of the sequence of phe-
nomena leading to failure (for example, a flood by a breach induced by an instability
is not summarized by a simple stability calculation) and the specifics of the structure
studied.
4. MAIN LESSONS LEARNED FROM PAST EVENTS
Many consequences have been drawn by the community and the landscape
has changed a lot since the end of the 20th century with regard to the inclusion of
protection levees, which play an essential role in the management of the flood risk (in
association of course with the portfolio of non-structural measures and other struc-
tural measures such as dams or flood storage reservoirs). The three main lessons
learned from these feedbacks are presented in this section: the importance of gov-
ernance, management structures and regulations, the necessary management of
overflowing and the use of accidentology for risk analysis.
4.1. GOVERNANCE
Many past events have highlighted the important and even essential role of
the managers for the efficiency and safety of levee systems. On the downstream
277
Q.105 - R.17
Rhône among others, the management structures put in place in the 19th century
showed their limits during the floods of 1993 and 1994, which led to the creation and
organization of the Symadrem (see 3.1). The manager is in charge of monitoring
and maintaining the structures, including during flood periods, in order to guarantee
their efficiency and avoid the formation of breaches.
In parallel with local changes in management structures and their practices,

often initiated by major flood events and resulting floodings, the French State has
become aware of the need for such good management at the national level. After
having undertaken since 1999 a census of the levees, their protected zones and
their managers (national databases managed by the administration), a regulation
on flood protection levees was gradually developed [11]. This regulation was initially
purely technical, then clarified the respective roles of the manager and the State, and
finally clarified the allocation of responsibility for the management of levee systems
and other flood protection structures (the new GEMAPI2 competence).
Regulations, like managers good practice, leads to regular performance

assessments and risk analyzes, which provide decision-support tools for the
definition of work programs and the organization of the surveillance of the structures.
4.2. OVERFLOWING MANAGEMENT
Among the phenomena at the origin of levee failure, overflow is one of the most
important, both in terms of statistics and in terms of related mechanisms: erosion by
overflow of an embankment structure, self-accelerating phenomenon is of course
the most important, but even its absence, an overflow can lead to destabilize the
structure by saturating its land side part. It is therefore necessary to design struc-
tures, or rather SYSTEMS, resilient to overflow, which leads to distinguish different
levels of loading [12] [13, chapter 3]: a protection level, below which water does not
enter the protected area, a safety level, below which water can enter the protected
area with a neglectible probability of breach, therefore in a controlled manner, and
a danger level beyond which there is a high hazard of failure.
On the Loire levees, after the 1866 flood, the engineer Comoy led the
implementation of the flood control spillways program on the following bases:
• embankment structures do not resist overflow and the heightening of the

levees does not resolve this weakness;
• the alternative approach of “giving water its share” with the invention of large
“modern spillways” (Fig. 5) allowing the river to expand into its whole floodplain
(flood expansion zone) while protecting the works;
• involvement of stakeholders in matters relating to spillways overflowing water
toward inhabited areas.
2management of aquatic environments and flood prevention (Gestion des milieux aquatiques
et prévention des inondations)
278
Q.105 - R.17
Fig. 5
Déversoir de type Comoy à Jargeau sur la levée d’Orléans
Comoy spillway at Jargeau on the Orleans levee
(Source DREAL CVL/DETL).
Following on from the reflections initiated in the 16th century on diversion spill-
ways3 , these new structures were to have three functions: avoid breaches, lower
peak discharges and reduce pressure on levees. The program allowed the construc-
tion of 8 spillways between 1870 and 1892 on the 19 initially planned. It should be
noted that projects were submitted for the opinion of local residents between 1868
and 1871 in the form of a public utility survey (also innovative at that time). After
forty years, this program was stopped, but the main principles of reflection remain
and are still relevant for the study of overflow management in levee systems [14].
The Rhône delta levees were significantly enhanced after the great floods
of 1840 and 1856. Qualified as “non-submersible” at the time, these structures,
which now protect more than 100,000 people and three times more in the summer
season, showed their limit during the floods of 1993, 1994, 2002 and 2003, when
17 starts of breaches and 12 breaches were recorded [5]. In the aftermath of the
2003 centennial flood, the State and the Regions put in place a vast flood prevention
plan at the watershed scale, called the Rhône Plan, which provides for the complete
renovation of the levees of the Rhône over 20 years.
Rather than raising the levees, it was chosen to accept flooding for rare floods
(10−2 annual probability upstream from Arles and 2.10−2 downstream) while con-
sidering the formation of breaches as unacceptable until exceptional events (10−3
3 locally called “déchargeoir”, this particular type of spillway derives a part of the flood dis-
charge from the river into a channel that goes back to the river downstream, in order to limit
the discharge in a segment of the river too narrow to allow safely high
279
Q.105 - R.17
annual probability). This choice involves the construction of overflow resistant lev-
ees, reinforcing the embankment on the valley side with concrete rip-rap so as to
withstand high speeds in the event of overflow. Upstream and downstream, levees
crest are 50 cm above the thousand-year flood to avoid any risk of bypassing during
overflow. We refer to [16] [17] for more details.
On the Agly, according to the 1967 design studies, levees were supposed to
be able to contain a flood with a return period of 25 years, the flow rate of which was
estimated at 1250 m3 /s, with a freeboard of 0,5 m. The initial project included, at the
upstream end of the leveed corridor, two 600 m spillways (one on each bank) to limit
the flow. These spillways were never built. A reinforcement program was proposed
in 2014 with one or more spillways and without changing the protection level, after
analyzing multiple scenarios and carrying out a multicriteria analysis. Despite the
criticality of the system, the probability of failure and the severity of its consequences
being both high, this project has not yet been carried out due to:
• difficult social acceptance of safety spillways,

• difficulty in recovering the land required for the works in order to limit their
environmental impact,
• the important amount of the investment (estimated at around = C 45 million) with
regard to the level of protection obtained (return period of between 5 and 10
years).
20 years after the first overflows (1999) and despite the stakes and the recur-
rence of floods that have necessitated the evacuation of the populations (1999,
2013, 2014, 2020), no solution has reached a consensus and new studies will be
undertaken to propose new alternatives.
Many levee systems, if not a large majority, in France, but also abroad, are
not able to properly manage overflows. Even if levees are in good condition and
withstand a water level reaching their crest, beyond this level of protection there is
a high or even certain risk of rupture and a risk of catastrophic flooding, given the
consequences of a breach at full load. It is desirable that the presented approach
for designing overflow-resilient river flood protection systems becomes generalized.
As a corollary, society must be made aware of a limit to protection, together with a
controlled management of overflows, with consequences much lower than randomly
placed breaches. This approach is detailed in a handbook [18] whose translation
into English is in progress.
4.3. CONTRIBUTION OF ACCIDENTOLOGY TO RISK ANALYSIS
The risk analysis of levee systems [13], which in France is reflected at the
regulatory level through Hazard Studies, is a very rich tool for decision-making.
However, as this is a recent activity, many gaps persist both in terms of general
methods and local data. Experience feedback on accidents and incidents makes it
280
Q.105 - R.17
Fig. 6
Hydrogrammes historiques de la Loire en amont et en aval de la brèche de
Jargeau et reconstitution de l’hydrogramme de la brèche
Historical hydrographs of the Loire upstream and downstream of the Jargeau
breach and reconstitution of the breach hydrograph
(Source: Archives départementales du Loiret, Cerema).
possible to remedy a number of these shortcomings. Risk is a combination of hazard

and consequences. We saw above (3.4) how experience feedback made it possible
to identify and probabilize the contingencies of failure, it also makes it possible to
characterize the potential breaches and the resulting floods.
Modeling the inundation in the protected area in the event of a breach, useful
both for understanding the residual risk and for the preparation of crisis manage-
ment, requires to know the characteristics of the breach (geometry and dynamics)
as well as calibration data relating to the protected area. Within the framework of
the hazard studies of the levees of Orleans, Tours and Authion [15], the analysis
relied in particular on data from the 19th century floods to define the area from the
dimensions of the historic pits formed by energy dissipation behind a breach, esti-
mated at 100 times the height of the levee (length in the direction of flow) and to
reconstruct the hydrograph of the breach (Fig. 6) that occurred at Jargeau in 1856
(flow of 3700 m3 /s and total volume of 270 Mm3 ).
On the Petit Rhône, the analysis of breaches widths and loading heights [5]
made it possible to deduce the following law:
Breach width = 72 × Load height – 123 m.
Over the whole of the downstream Rhône, the cumulative width of breaches
during the same flood and the volume of discharge as a function of the flow
at the head of the delta (Fig. 7) made it possible to observe the exponential
281
Q.105 - R.17
Fig. 7
Largeur cumulée de brèches et volumes de déversement
Cumulative width of breaches and spilled volumes.
evolution of these essential parameters for risk analysis, particularly in the definition
of inundation-by-breach scenarios.
From the lessons learned on the breaches on the various watercourses, those
presented here or others, it would be interesting to carry out a comparative analysis
of the various parameters: geometry and constitution of the levee, widths of the
minor bed and of the floodplain, flood discharge, geometries of the breach and of
the erosion pit, estimated flow in the breach, etc. This would allow, among other
things, to better calibrate the models required during risk analyzes.
5. CONCLUSION ET PERSPECTIVES
The numerous events of floods and flooding in leveed areas which have fol-
lowed one another in France since the end of the 20th century, as well as the
knowledge relating to older events, constitute a very rich base of analysis. There
are many lessons that have already been learned from these events:
• Governance has an essential impact on the safety and efficiency of structures:

importance of a competent manager and a national regulatory framework,
282
Q.105 - R.17
• The management of overflows beyond the level of protection of the systems

is fundamental to minimize damage during major floods,
• Assessments and risk analysis benefit greatly from information drawn from
experience feedback, both for the development of methods and for obtaining
data on specific cases. Among other things, experience feedback makes it
possible to become aware of uncertainties, and in favorable cases to assess
them,
Design engineering of structures is adapted to failure mechanisms and to the

knowledge that we may have of them and is enriched as and when feedback is
provided. The reinforcement and repair methods used by the profession are shared
within it, being the subject of a CFBR4 working group (WG). The deliverable of
this WG, which will be available in 2021, will consist of a document conventionally
written in chapters and describing the principles of the reinforcement and repair
methods. It will be readable autonomously but is accompanied by 32 descriptive
sheets presenting in detail the reinforcement/repair techniques, organized by func-
tion (waterproofing, drainage, filtration, stability, enhancement, protection against
external erosion) that the techniques fulfill mainly or secondary. These sheets will
be consulted individually as needed. This document will be evolving and may be
enriched as the experience feedback on new techniques.
Beyond these acquired elements, we can also note a certain number of current
developments and prospects for improvement. The levees community is organizing
to progress. The creation of France Digues, the French association of levees man-
agers, allows the professionalization of this “new” profession. The integration of
levees as an object of study within the CFBR in France, and ICOLD at the interna-
tional level, allows exchanges between engineers from all types of stakeholders in
hydraulic structures (managers, consultants, contractors, administrations). We can
also note the setting up in recent years of professional training sessions dedicated to
levees, as well as numerous research programs linked to these structures. Among
the leads for progress identified on the basis of feedback, let us note the benefit of
linking the reinforcement projects with plans for managing the morphology of the
leveed corridors.
Knowledge of the structures and the historical or recent events that they have
encountered is therefore a wealth that must continue to be capitalized on, both locally
and nationally, in order to monitor accidents. A certain “standardization” in the way
of presenting and analyzing events is desirable for their use and sustainability on
shared bases. But, in the end, we must keep in mind to remain humble and attentive
so that the community keeps in the long-term memory all these past events and the
elements presented here: indeed we have unfortunately often observed in the past
a relaxing attention after a long period without floods ...
4 Comité Français des Barrages et Réservoirs, the French COLD
283
Q.105 - R.17
REFERENCES
[1] ICOLD, 2015. Internal erosion of existing dams, levees and dikes, and their
foundations – bulletin n◦ 164 – Volume 1: internal erosion processes and
engineering assessment.
[2] Blondiaux P. & Patouillard S. 2019. Base de données sur les brèches
historiques de la Loire, stage Master I. DREAL CVL.
[3] Fournier M. 2008. Quelle place pour les riverains dans la gestion des
inondations de Loire – les leçons du passé – Freude am Fluss.
[4] Mallet T., Outalmit K. & Fry J.J. (2014) Probability of failure of an embankment
by internal erosion using the Hole Erosion Test. ICOLD BALI International
Symposium.
[5] Mallet, T., Fry, J.J., Tourment, R., Mériaux, P. Accidentologie des digues du
delta du Rhône de 1840 à nos jours. Digues maritimes et fluviales de protection
contre les inondations – 3ème colloque national – Digues2019 – 20-21 mars
2019 – Aix-en-Provence.
[6] Mallet T., Dast C. Requi M. Chardès C. Castagnet A. et Fry J.J. (2018). Etude
de dangers du système d’endiguement rive gauche du delta du Rhône –
ICOLD Vienna Congress.
[7] Mallet T., Fry J.J. (2016) – Probabibility of failure of an embankment by

backward erosion using the formulas of Sellmeijer and Hoffmans – ICOLD
Johannesburg International Symposium.
[8] Tourment, R., Benahmed, N., Nicaise, S., Mériaux, P., Salmi, A., Rougé, M.
– 2018. Retour d’expérience sur les désordres des digues de l’Agly maritime.
Analyse des phénomènes de “sand-boils”. Commission Internationale des
Grands Barrages. Vingt-sixième congrès des grands barrages 01/07/2018-
07/07/2018, Vienna, AUT. 31 p.
[9] CFBR, 2021, Barrages, le savoir-faire français, Chapitre 9 “Digues de

protection contre les inondations” (20 pages). ISBN 979-10-96371-11-2.
[10] CIRIA, MEDDE, USACE – 2013. The International Levee Handbook. Van
Hemert, H., Igigabel, M., Pohl, P., Sharp, M., Simm, J., Tourment, R., Wallis,
M. (ed.), CIRIA, Londres, 1349 p.
[11] Tourment et al. – 2021 – The new French régulation on flood protection
structures; conséquences on risk management. FLOODrisk2020 conference,
Budapest, 11 p.
[12] Mallet et al., 2013. Niveaux de protection, de sûreté et de danger pour

les digues fluviales. Digues maritimes et fluviales de protection contre les
submersions 2ème Colloque national, Aix-en-Provence, 12–14 juin 2013.
284
Q.105 - R.17
[13] Tourment, R., Beullac, B. – 2019. Inondations – Analyse de risque des sys-
tèmes de protection – Application aux études de dangers. Tourment, R.,
Beullac, B. (ed.), Lavoisier, Paris, 356 p.
[14] Maurin, J., Patouillard, S., Tourment, R. – 2013. Les déchargeoirs au XVIIe
siècle, les déversoirs au XIXe et quelles perspectives pour la gestion des
surverses au XXIe siècle. Congrès SHF Événements extrêmes d’inondation
2013 13/11/2013-14/11/2013, Lyon, FRA. 9 p.
[15] Maurin, J., Boulay, A., Piney, S., Le Barbu E., Tourment, R. – 2012. Les
brèches des levées de la Loire: brèche de Jargeau 1856. Congrès SHF
Événements extrêmes fluviaux et maritimes 01/02/2012-02/02/2012, Paris,
FRA. 8 p.
[16] Cheetham, M., Mallet, T., Chastel, E., Tourment R., Robustelli, P., Pelt, P.,
2015. Building a resilient system of defence against flooding from the Rhône.
Proceedings of the Institution of Civil Engineers-Water Management, vol. 168,
n◦ 2, p. 74–84.
[17] Mallet et al., 2013, Sécurisation des digues du Rhône entre Beaucaire et
Arles: conception hydraulique du système et des digues résistantes à la sur-
verse. Digues maritimes et fluviales de protection contre les submersions
2ème Colloque national, Aix-en-Provence, 12–14 juin 2013.
[18] Degoutte, G. (coord.) – 2012. Les déversoirs sur digues fluviales. Editions
Quae, Versailles, 184 p.
285
Q.105 - R.18
-------
-------
GESTION DE CRUES ET INSTABILITÉS DE PENTES

PENDANT LA CONSTRUCTION DE BARRAGES: PROBLÉMATIQUES
D’ANTICIPATION ET DE RECONNAISSANCES PRÉLIMINAIRES (∗)
Thomas LAURENT, Quentin BERCHER & Thierry VINCENT

ARTELIA
FRANCE
RÉSUMÉ
Les projets de construction de barrages et aménagements hydroélectriques

sont des ouvrages très techniques nécessitant au préalable un temps d’études
spécifiques menées par des spécialistes du domaine. Néanmoins, face à la multipli-
cation des projets et une volonté de réduire les coûts et temps de développement,
des projets sont démarrés en phase de construction avec des conceptions non
abouties, des vérifications et optimisations souvent non effectuées et des méth-
odes travaux quasiment non existantes. C’est dans ce contexte que cet article
recense trois expériences de terrain récemment vécues en phase de construction
de l’aménagement.
Au Malawi, le barrage de Bwanje a démarré sa construction sans phasage

de construction adapté à l’hydrologie saisonnière de la zone, et par une entreprise
désignée par le client mais sans réelle expérience dans ce type de construction.
ARTELIA a été mandaté au démarrage des travaux et est intervenue pour modifier et
sécuriser l’évacuateur de crue de chantier deux semaines juste avant les premières
crues qui auraient pu détériorer le barrage lui-même.
∗ Flood management and slope instabilities during dam construction: anticipation and
preliminary site investigation issues
286 DOI: 10.1201/9781003211808-64

Q.105 - R.18
Sur l’aménagement de Nam Ngum 3 au Laos, l’entreprise travaux a en charge

et dans ses responsabilités l’étude et la réalisation des travaux de soutènement des
talus d’excavation du bâtiment de la centrale électrique. Quelques mois après sa
mobilisation sur site, les équipes ARTELIA ont mis en évidence l’apparition de désor-
dres sur le talus supérieur de la centrale conséquents d’un sous-dimensionnement
du soutènement et des méthodes de travaux non adaptées. Les mouvements
alors mis en évidence et analysés se sont conclus par un effondrement du talus
occasionnant des retards dans la construction.
Pendant la construction du barrage de Delsitanisagua en Equateur, l’ouverture

des fouilles a déclenché une instabilité de versant profonde impactant l’appui de rive
gauche du barrage. La vitesse de déplacement, forte au début a remis en ques-
tion la faisabilité du barrage. Cette instabilité a été déclenchée par la mobilisation
d’une faille profonde située dans cette rive et qui n’a pas été caractérisée en phase
de conception. La conception du barrage a été adaptée pour pouvoir anticiper un
déplacement irréversible de ses rives. Cependant, des solutions robustes et fiables
ont dû être trouvées en cours de construction pour stabiliser ce versant.
SUMMARY
Dam and hydropower development projects are highly technical projects

requiring specific studies, which must be completed by specialists in the field and
finalised prior to the start of works. However, with an increasing number of projects
and pressure to reduce the costs and time spent on the project preparation stages,
the works phase of some projects begins before the design has been finalised. In
addition, checks and optimisations are often not carried out and a works method-
ology is almost non-existent. Within this context, this article describes three recent
experiences on site during the works phase of development projects.
In Malawi, construction began on the Bwanje dam with a scheduling of works

that did not take the local seasonal hydrology into consideration, and with a con-
tractor that had been appointed by the client despite having no real experience with
this type of construction. ARTELIA was commissioned at the beginning of the works
and intervened to modify and secure the works site flood spillway just two weeks
before the first flooding event that could have caused damage to the dam itself.
On the Nam Ngum 3 development in Laos, the works contractor was in

charge of and responsible for the design and construction of retaining works on
the excavated slopes around the powerhouse. A few months after arriving on site,
the ARTELIA teams pointed out the appearance of damage on the powerhouse
backslope, a consequence of under-sizing of the dimensional design of the retain-
ing structure and an unsuitable works methodology. The movements that were
then observed and analysed resulted in a collapse of the slope, causing delays
to construction.
287
Q.105 - R.18
During construction of the Delsitanisagua dam in Ecuador, the beginning of

excavations triggered a deep slope instability that affected the left bank abutment
of the dam. The speed of progression of the displacements, which was initially very
high, cast doubt on the feasibility of the dam project. This instability was triggered by
the activation of a deep fault located on this bank and which had not been identified
during the design phase. Design of the dam was modified to anticipate an irreversible
displacement of its banks. However, robust and reliable solutions were required
during the works to stabilise the slope.
INTRODUCTION
Dans un contexte international où le développement de projets de barrages

et aménagements hydroélectriques est fortement dépendant de la volonté des dif-
férents acteurs et des capacités de financement, il a été constaté à plusieurs reprises
l’engagement de travaux sur la base d’études de faisabilité et sans l’implication d’un
bureau d’études spécialisé en conception de barrages.
Au cours de ces dernières années, ARTELIA a recensé sur plusieurs amé-

nagements, dans le cadre de missions d’Assistance à Maitrise d’Ouvrage ou
de Maitrise d’Oeuvre pour la construction et après adjudication des entreprises,
des problématiques majeures non anticipées en amont des travaux par manque
d’investigations ou de préparation. Ces aléas de chantier peuvent entrainer des
risques sécuritaires, des dégâts sur les ouvrages réalisés et occasionnent des
surcoûts et retards sur le planning travaux.
Le présent article décrit et analyse le retour d’expérience d’ARTELIA sur trois

aménagements où des problématiques de gestion de crues et d’instabilités de talus
ont été gérées en phase travaux.
1. LE BARRAGE DE BWANJE: ANTICIPATION ET PRÉVISION DU

RISQUE DE CRUE
1.1. CONTEXTE ET OUVRAGE
1.1.1. Description du projet
Le barrage de Bwanje est un ouvrage financé par l’Union Européenne et qui

va permettre un accès à l’eau et de réguler l’irrigation. C’est un en remblai de 40 m
de haut situé dans la Région Centrale du Malawi. ARTELIA a pour mission sur ce
projet une revue d’étude de conception et la supervision des travaux. Les travaux ont
288
Q.105 - R.18
Fig. 1
Vue en plan du barrage et de l’évacuateur
Top view of the dam and the spillway.
été attribués à PLEM Construction, une entreprise locale sans référence «barrage»
en août 2016. ARTELIA est intervenue sur le projet après l’attribution du marché de
travaux et lorsque toute la conception été terminée.
Le barrage, d’une longueur de 150 m en crête, est réalisé en enrochements

avec un noyau d’étanchéité argileux. Une tour de prise pour l’irrigation est également
réalisée, avec une conduite enterrée passant sous le barrage. Sur la rive droite,
l’évacuateur de crues est un canal courbe en escalier et excavé directement dans
le rocher.
1.1.2. Phasage prévisionnel
La construction est prévue en deux phases. Lors de la première, une hauteur

voisine de 40 m sera atteinte, avant d’être rehaussée à 60 m lors de la seconde
phase (autre marché). La capacité du réservoir, qui est prévue à 5,6 M m3 dans un
premier temps, passera alors à 31,5 M m3 . La phase 1 dont il est question ici a été
livrée en janvier 2019 et la seconde phase sera réalisée ultérieurement.
Artelia a commencé sa mission en revoyant et modifiant tous les concepts des

ouvrages barrages et évacuateur de crues car ceux-ci n’étaient pas satisfaisants.
289
Q.105 - R.18
Fig. 2
Coupe révisée du barrage
Revised Dam Cross Section.
Compte tenu du climat tropical du Malawi, le phasage est fortement impacté

par l’arrivée de la saison des pluies. Dans une première conception qui se basait
sur des hypothèses favorables, la construction devait prendre 12 mois. Débutant
lors de la saison sèche de 2017, il était prévu que le barrage atteigne l’élévation
639 avant la fin de l’année, protégeant l’ouvrage contre les crues et permettant une
poursuite des travaux jusqu’à la cote finale (645 m) dans les mois suivants.
1.2. EVALUATION ET ANTICIPATION DES RISQUES
1.2.1. Anticipation de retard dans les travaux
Le planning de construction avait été établi par l’Entreprise en même temps

que couraient les détails du contrat. L’Entreprise avait à sa charge le phasage et les
moyens techniques pour y arriver. En particulier, les travaux de dérivation étaient
de la responsabilité de l’entreprise, néanmoins Artelia, en tant que concepteur et
surveillant des travaux et de leur avancement, a rapidement alerté sur l’existence
d’un risque lié à ce planning en cas de retard. En effet la saison des pluies pouvait
alors interrompre les travaux et menacer l’ouvrage.
Dès le démarrage des travaux, il a été constaté des variations sur les
conditions géologiques de la fondation nécessitant une surexcavation et une aug-
mentation des quantités du voile d’étanchéité. Rapidement, une dérive du planning
travaux a été constatée et il est apparu comme clair que l’avancée du barrage ne per-
mettrait pas d’atteindre un niveau suffisant avant la saison des pluies de fin d’année
pour éviter les risques de surverse.
290
Q.105 - R.18
1.2.2. Etude des crues et analyse de risques
Le débit de la rivière augmente normalement en décembre, avec des épisodes

de crue survenant à la mi-décembre. Le réservoir étant relativement petit (7% du
volume annuel écoulé), il peut se remplir très rapidement en cas de crue. Il est donc
crucial de savoir si le système de dérivation sera suffisant, et d’évaluer le risque
encouru dans le cas contraire.
La hauteur requise pour assurer la déviation des eaux par le déversoir est
supérieure à 30 m. A cette hauteur, l’ouvrage passe dans la catégorie des grands
barrages et le risque hydrologique à prendre en compte lors de la construction
évolue avec cette classification. Pour la durée de construction de 12 mois prévue,
soit avec une saison des pluies, la sécurité des populations devrait être assurée face
à la même probabilité d’événement que pour l’ouvrage complet. Dans le cas d’un
barrage en enrochements, une crue de récurrence de 10 000 ans est une sécurité
acceptable. Pour le cas de Bwanje, dont la durée de vie est de 50 ans, cela amène
pour une période de 12 mois de construction théorique, à considérer une crue de
récurrence de 200 ans imposée au marché (très sécuritaire). Cette fréquence a donc
été utilisée dans l’évaluation des risques hydrologiques lors de la construction.
L’analyse des débits de crues, basée sur les mesures historiques, montrent
qu’il existe un risque fort que le niveau d’eau atteigne la cote 637 m avant le 21
décembre. Ce niveau étant 32 m au-dessus des fondations, il aurait fallu assurer une
élévation de 24 m supplémentaires entre octobre et décembre 2017. De cette façon,
le barrage aurait été en mesure de dévier les eaux vers le déversoir. Néanmoins,
au moment de ces études le constructeur avait décidé de retarder l’excavation du
déversoir, rendant impossible son utilisation face à la crue.
Pourtant la protection de l’ouvrage face à des crues est indispensable: la

pénétration d’eau dans le noyau aurait compromis la suite de la construction ainsi
que l’intégrité structurelle de l’ouvrage. S’agissant d’un grand barrage, le risque
pour les populations était élevé et la ruine devait à tout prix être prévenue. Il a donc
été nécessaire de trouver une solution permettant le maintien du barrage pendant
l’épisode de crues.
1.3. SOLUTION DE PROTECTION
Artelia a proposé différentes solutions et cela dès le début du chantier au cas

où la construction ne serait pas assez avancée pour faire face à l’arrivée de la saison
des pluies. Une solution envisagée consistait notamment en une sur-excavation au
droit du seuil déversoir pour le passage de la crue de chantier avant la réalisation
du seuil déversant. Une seconde solution prévoyait l’accélération du remblaiement
du barrage en utilisant une technique alternative. Ces deux solutions ont toutefois
été écartées en raison de l’état d’avancement des travaux et du délai restant avant
la période de crues.
291
Q.105 - R.18
La solution retenue consiste à protéger le noyau par une couverture imper-

méable, protégée par un géotextile et stabilisée par des enrochements pour
autoriser le déversement du réservoir par-dessus la crête si le débit dépasse la
capacité de la conduite de prise. L’écoulement sur le parement aval demanderait
alors d’adoucir la pente à 5/1 voire 10/1, contre 1,7/1 pour l’ouvrage final. Cette
solution a pour avantage d’être rapide à mettre en place et à retirer, autorisant une
reprise rapide de la construction.
Fig. 3
Disposition prévue des couches protectrices
Proposed organization of the protective layers.
La membrane imperméable PEHD ainsi que le géotextile BIDIM A10 sont

arrivés sur le site le 15 décembre 2017. Un aplanissement et un compactage de la
partie supérieure du noyau ont été réalisés, puis deux tranchées de 500 mm ont été
creusées pour ancrer les membranes en amont et en aval de la crête.
Suite au raidissement du talus aval par rapport aux préconisations d’Artelia,

la protection géotextile a finalement été prolongée sur tout le parement et le pied
aval comme mesure complémentaire. De même, une couche de 200 mm de sable a
été placée par sécurité sur les membranes au niveau de la crête avant d’ajouter une
couche d’enrochement de 2 m de haut. Bien que les membranes soient résistantes,
ces dispositions permettent de limiter les risques de déchirure.
Fig. 4
Mise en place de la membrane HDPE sur le noyau argileux
HDPE membrane disposal on the clay core.
292
Q.105 - R.18
Fig. 5
Parement aval recouvert de géotextile
Downstream slope covered with geotextile.
Les mesures préventives ont été mises à l’épreuve dès le 20 décembre,

où le niveau d’eau a atteint la crête et a engendré un léger débordement. Dès
lors, la protection a montré son efficacité. Une surveillance de l’ouvrage a été
menée par le constructeur durant tout l’épisode de pluies. Lors de certaines crues,
de légères réparations des protections ont été nécessaires sur le parement aval,
mais la solution mise en place a tenu bon. Lorsque le niveau du réservoir était
assez bas, certains travaux étaient possibles à l’amont. Les protections ont été
retirées du 4 au 7 mai 2018. Après une vérification favorable de l’humidité de
l’argile et un re-compactage, les travaux d’élévation du noyau ont pu reprendre dès
le 9 mai.
Fig. 6
Crue du 12/03/2018, atteignant la cote de 624.30 m
Flood event on 12/03/2018, reaching the elevation of 624.30 m.
293
Q.105 - R.18
1.4. CONCLUSION
La saison des pluies 2017–2018 n’a pas été particulièrement sévère, de sorte
que les débordements ont été limités en temps et en intensité. Les protections tem-
poraires ont résisté malgré quelques réparations occasionnelles, mais des crues
plus importantes auraient pu occasionner des dégradations plus importantes sur le
talus aval raidi par l’entreprise par rapport aux préconisations. Vu le faible élance-
ment du barrage au moment de la mise en place des protections, le risque de
ruine était relativement faible. En revanche, avec l’augmentation de la pente aval
par l’entreprise, la stabilité des blocs s’est trouvée compromise. Le respect des
préconisations d’une pente maximale de 5/1 aurait été souhaitable bien qu’aucun
dégât majeur n’ait été constaté.
La responsabilité de la gestion des crues de chantier a été déléguée à l’entre-

prise sans que celui-ci n’ait l’expérience requise. Il aurait été préférable d’imposer
à l’entreprise une méthode de gestion de crues dès le cadre de l’appel d’offre.
2. AMÉNAGEMENT HYDROÉLECTRIQUE DE NAM NGUM 3 – LAOS:

GLISSEMENT DU TALUS AMONT DE LA CENTRALE
Le projet hydroélectrique Nam Ngum 3 (NN3) est situé au Laos sur la rivière
Nam Ngum qui est un affluent majeur du Mékong. Il prend sa source dans la province
de Xiangkhoang et coule vers le sud à travers la province de Vientiane pour rejoindre
le Mékong près de la capitale du Laos, Vientiane. Le bassin versant s’étend sur
16 640 km2 et abrite environ un million de personnes. Cet aménagement d’une
puissance de 480 MW est notamment constitué d’un barrage CFRD de 212 m de
haut, de 11 km de tunnel et d’une centrale hydroélectrique équipée de 3 turbines
Francis.
La construction du projet a débuté en 2015 pour une mise en service en

2022 Le schéma contractuel pour la construction du projet est de type conception-
réalisation. Ainsi, l’entreprise générale est en charge de l’ensemble de la conception
et de la réalisation des travaux. ARTELIA intervient sur le projet en tant qu’Assistant
à Maitre d’Ouvrage depuis fin 2017 une mission de revue de conception et la
supervision des travaux.
2.2. CONDITIONS GÉOLOGIQUES
La construction de la centrale hydroélectrique de Nam Ngum 3 nécessite la

création d’une fosse d’excavation dans un nez de terrain dont la coupe est présentée
sur la figure suivante:
294
Q.105 - R.18
Fig. 7
Profil d’excavation et vue de face du talus excavé de la centrale de Nam Ngum 3
(fond de fouille à un niveau de 358 m)
Excavation profile and view of the excavated slope of the powerhouse of Nam
Ngum 3 HPP.
La partie supérieure du talus est composée de sols résiduels et de roches

fortement altérées tandis qu’en son sein se développe une matrice rocheuse
marquée par une foliation très resserrée dont le pendage, parallèle à la pente,
se trouve être très raide par rapport à la ligne d’excavation (70◦ ∼ 80◦ ). Cet état du
massif constitue une situation défavorable pour les travaux d’excavation. En effet, il
conduit de manière inévitable à la décompression de la masse rocheuse le long des
plans de foliation à cause de la diminution de la contrainte de confinement induite
par les travaux. Une instabilité du massif rocheux par basculement est ainsi initiée.
Fig. 8
Instabilité par basculement du massif rocheux
Flexural toppling of the rock mass.
295
Q.105 - R.18
De plus, dans le cas du talus de la centrale, ce mouvement de basculement

est amplifié par le poids de l’épaisse couche de sol résiduel constituant la partie
haute du talus.
2.3. CONCEPTION
Le système de soutènement, dans sa conception initiale et prévue par

l’Entreprise, prévoit ainsi la mise en place d’un maillage de clous (3 m × 3 m) ancrés
dans le rocher à des profondeurs allant de 8 à 12 mètres. Ceux-ci sont reliés à
un treillis soudé en acier (200 mm × 200 mm), la surface du talus étant recouverte
d’une couche de béton projeté d’épaisseur minimale 10 cm.
Fig. 9
Coupe axial du dispositif de soutènement initialement prévu pour le talus amont de
la centrale
Section of excavation supports initially designed for powerhouse backslope.
2.4. EVOLUTION DE L’ÉTAT DU TALUS AU COURS DES TRAVAUX
Dès les premières phases des travaux, des défauts de réalisation et de con-
ception sont relevés lors de l’excavation de la fouille de la centrale et de la mise
en place du soutènement sur la partie haute du talus. Il est ainsi constaté d’une
296
Q.105 - R.18
part que la liaison mécanique entre les ancrages et le ferraillage réparti sur la paroi
est uniquement assurée par des soudures qui limitent la transmission des efforts
en comparaison avec une liaison par platine en tête des ancrages. D’autre part,
le dispositif de drainage prévue dans la conception initiale est limité avec unique-
ment une galerie de drainage située à mi-hauteur du talus et de quelques drains.
La combinaison de ces éléments a conduit aux premières instabilités observées sur
le soutènement du talus, à savoir l’initiation du flambement de la partie basse du
béton projeté avec une dissociation du contact sol/béton projeté.
Par ailleurs, un système d’auscultation composé de mires X ,Y ,Z a été installé

sur le talus amont pour en assurer le suivi. Suite aux premières observations de site,
les déplacements mesurés sur ces points d’observations ont été mis en parallèle
avec les activités d’excavation et de soutènement du talus pour identifier les causes
des mouvements observés.
Fig. 10
√
Déplacements cumulés (X2 + Y2 + Z2 ) du talus amont de la centrale de Nam
Ngum 3 et planning des travaux exécutés en dessous du niveau 413 m.
Displacement rates observed on powerhouse backslope and schedule of
excavated and support works below el. 413.
297
Q.105 - R.18
Fig. 11
Coupe des tirants précontraints sur la hauteur de l’excavation
Cross section of pre-stressed anchors installed on excavation height.
Etablir un lien entre les travaux d’excavation et la magnitude des déplacements

de la surface du talus n’est pas évident malgré l’évolution certaine de la situation.
Toutefois, cette analyse permet de mettre en évidence un mouvement global du
massif vers l’aval et ce malgré la mise en place de mesures correctives tel que le
déchargement du talus avec la surexcavation de la partie haute et l’installation de
tirants précontraints en partie supérieure (tension de blocage variant de 100t à 150t
pour un total de 80 unités). Ces mesures, appliquées en urgence, ont été mises en
place avec un certain retard au regard de l’apparition des déplacements en raison
d’une mobilisation tardive de l’assistant Maitre d’Ouvrage vis-à-vis du démarrage
des travaux et des divergences de point de vue entre l’entreprise et ARTELIA sur
l’analyse du phénomène et les actions à mettre en œuvre.
L’analyse des données mesurées sur les dynamomètres installés sur certains
tirants précontraints a montré que les mesures correctives en cours de mise en
œuvre ne permettaient pas de ralentir les déplacements du talus.
Fig. 12
Evolution de la tension dans le câble et localisation du tirant d’ancrage n◦ PR4−10
Tension evolution and location of pre-stressed anchorage n◦ PR4−10 .
298
Q.105 - R.18
Malgré les signes d’alerte et motivée par le respect du planning des travaux,
l’entreprise a poursuivie l’excavation du fond de fouille à l’aide d’explosif, sans plan
de tir de minage particulier ni d’installation d’instruments de mesure, pour vérifier
les impacts sur les structures existantes et le massif rocheux, comme requis par
l’Assistant Maitre d’Ouvrage.
L’effondrement partiel du talus amont de la centrale advient le 21 mai 2018.

La partie concernée se situe entre les élévations 380 et 385, sous la dalle en
béton servant de support aux têtes des rangées 5 et 6 des tirants précontraints en
cours de finalisation. Quelques heures après l’évènement, l’entreprise entreprend le
remblaiement de la base de la partie effondrée afin de conforter l’instabilité comme
en atteste les photos suivantes.
Fig. 13
Effondrement du talus de la centrale entre les niveaux 380 et 385
Collapse of the powerhouse backslope between
elevations 380 and 385.
299
Q.105 - R.18
2.5. ANALYSES DU PHÉNOMÈNE ET MESURES CORRECTIVES
Suite à l’analyse des études de conception, des conditions de réalisation et

de la chronologie des phénomènes, il est évalué que l’effondrement partiel du talus
de la centrale est principalement dû aux défauts de conception et de réalisation
suivants pour de telles conditions géologiques.
– Une pente de talus non adaptée en partie supérieure et des excavations réal-
isées sans prise en considération des caractéristiques et du comportement
du terrain en place
– Un dispositif de soutènement insuffisant avec des ancrages sans platines ni
centreurs et des défauts dans le béton projeté (faible épaisseur et enrobage
avec le treillis soudé)
– Un dispositif de drainage insuffisant avec drains uniquement en pied de talus,
peu profond et avec un espacement élevé entre drains
– Un phasage d’excavation non adapté avec des hauteurs excavées trop
importantes sans soutènement ni drainage
– Une possible augmentation des pressions interstitielles consécutive au début
de la saison des pluies et à un drainage inefficace du talus.
– L’exécution de tire d’excavation non contrôlés dans le fond de fouille, à prox-
imité du talus, conduisant à des vibrations excessives dans un massif déjà
affaibli par une déformation de type «basculement».
Après sécurisation de la zone et l’arrêt des travaux d’excavation pendant toute

la durée de la saison des pluies, les mesures correctives suivantes ont été mise en
œuvre:
– Actualisation des études de conception en tenant compte des conditions

réelles du site de manière à définir au mieux les mesures correctives;
– Nivellement complet de la partie haute du talus pour décharger l’ensemble du
talus;
– Amélioration du système de drainage avec la réalisation de deux galeries sup-
plémentaires creusées perpendiculairement à la galerie de drainage existante
et parallèlement à la surface excavée;
– Augmentation significative du dispositif de soutènement en partie basse du
talus avec l’ajout d’ancrage précontraints auscultés;
– Amélioration du dispositif d’auscultation avec la mise en place de piézomètres,
extensomètres et de de mires supplémentaires en partie basse du talus;
– Revue de l’organisation sur site pour assurer le suivi et l’interprétation des
données d’auscultation dont la prise de mesures a été revue et les seuils
d’alerte redéfinis.
Seuls des dégâts matériels ont été déplorés suite à l’effondrement du talus
qui aurait certainement pu être évité avec des investigations géotechniques plus
300
Q.105 - R.18
approfondies au stade de la conception, moyennant des coûts et délais supplé-

mentaires, et une adaptation des travaux en fonction des mesures d’auscultation.
Le schéma contractuel du projet est également en cause dans cet évènement
avec peu de marge de manœuvre pour l’Assistant Maitre d’Ouvrage pour imposer
des solutions techniques à une entreprise de travaux maitresse de la concep-
tion et de la réalisation. Cet évènement a engendré un retard des travaux estimé
à 5 mois.
3. BARRAGE DE DELSITANISAGUA: INSTABILITÉ DE LA

RIVE GAUCHE
Le barrage de Delsitanisagua est un barrage-poids en béton de 35 m de haut

et de 200 m de long en Equateur dont les travaux se sont déroulé de 2013 à 2018. Le
schéma contractuel pour la construction du projet est de type conception-réalisation
où l’entreprise de travaux est responsable des études de conception et de l’exécution
des travaux.
L’ouverture des fouilles du barrage a déclenché une instabilité de versant pro-

fonde impactant l’appui de rive gauche du barrage. La vitesse de déplacement,
forte au début (de l’ordre du centimètre par jour) a remis en question la faisabilité
du barrage. Cette instabilité a été déclenchée par la mobilisation d’une faille pro-
fonde située dans cette rive et qui n’a pas été caractérisée en phase de conception.
La conception du barrage a été adaptée pour pouvoir anticiper un déplacement
irréversible de ses rives. Cependant, des solutions robustes et fiables ont dû être
trouvées pour stabiliser ce versant.
3.2. PROBLÉMATIQUES RENCONTRÉES EN RIVE GAUCHE DU BARRAGE
La hauteur maximum du talus rive gauche est de 120 mètres et son emprise
horizontale est de 300 mètres. Les pentes des excavations varient de 0.3 H/1 V à
0.6 H/1 V. Les hauteurs de talus entre chaque risberme varient entre 10 et 15 m. Le
massif rocheux de la rive gauche est composé principalement de gneiss pouvant
être très altéré ou fracturé. L’épaisseur des dépôts alluviaux en couverture sur les
talus varie de 6.00 à 18.00 m.
Les premières excavations du talus ont débuté en février 2013. A son arrivée
sur site et après revue des documents d’avant-projet détaillé entre mars et juillet
2013, ARTELIA, en tant qu’Assistant du Maitre de l’Ouvrage, a mis en évidence des
301
Q.105 - R.18
manquements concernant les hypothèses géotechniques retenues par l’entreprise

pour dimensionner le soutènement du talus rive gauche principalement composé
d’ancrages passifs de longueur d’environ 6 mètres et de béton projeté.
En août 2013 un plan de glissement subhorizontal du talus a été identifié. Dès

Janvier 2014 des mouvements significatifs du talus ont été détectés. Les excavations
ont dû être stoppées entre Février et Avril 2014. Des investigations géotechniques
complémentaires ont été réalisées, principalement des carottages horizontaux, et
un nouveau calcul de stabilité a été réalisé. Ces études complémentaires ont conclu
à une demande d’ajout de nouveaux ancrages passifs de longueur variant entre 12
et 15 mètres. Un total de 673 nouveaux ancrages a été mis en œuvre dans le talus.
Fig. 14
Excavation rive gauche, Août 2014
Left bank excavation, August 2014.
A partir de Janvier 2015, à la demande d’ARTELIA, l’entreprise a complété son

dispositif d’auscultation du talus en installant des extensomètres supplémentaires.
Ces extensomètres ont très vite mis en évidence des mouvements en profondeur
du talus de l’ordre de 20 mm par mois. Dans les semaines qui ont suivi l’installation
de ces appareils complémentaires, une dégradation du béton projeté en surface a
été constatée avec l’apparition de plusieurs fissures et épaufrures.
302
Q.105 - R.18
Fig. 15
Apparition de fissures, Mars 2015
Cracks initiation, March 2015.
En avril 2015, de nouvelles fissures sont apparues sur le talus et l’entreprise

émet l’hypothèse d’une instabilité par un phénomène de renversement du massif
rocheux (toppling). Dans les semaines suivantes, l’entreprise met en place 50 pre-
miers ancrages actifs de capacité de 1000 kN, chacun ancrés à une profondeur de
60 à 70 mètres. Le dispositif d’auscultation du talus est également renforcé avec
l’ajout d’extensomètres de piézomètres et d’inclinomètres.
En septembre 2015, lors de la finalisation des excavations, il apparait en pied

de talus l’inclusion d’un matériau inconsistant dans le massif rocheux sur 3 zones
ce qui étaye l’hypothèse émises en 2013 d’un plan de glissement subhorizontal
du talus. En octobre 2015, une analyse des données d’auscultation montre une
décélération des déplacements, toutefois l’équilibre de stabilité n’est pas atteint.
La réalisation d’investigations complémentaires a permis l’actualisation du

modèle géologique du massif de la rive gauche parcouru par une faille subhori-
zontale ainsi qu’une famille de failles subverticales.
Fig. 16
Modèle géologique de la rive gauche actualisé suite aux excavations
Geological model of left bank updated post excavation.
303
Q.105 - R.18
3.3. SOLUTION DE REMÉDIATION
Les méthodes de stabilisation des massifs s’axent généralement sur

l’utilisation d’ancrages passifs et actifs, sur la mise en place de drains et sur une
adaptation de la pente d’excavation. Cette dernière solution n’est pas envisage-
able dans le cas de Delsitanisagua puisque l’instabilité d’ensemble du massif s’est
déclenchée suite à l’excavation de la pente.
Ainsi, une solution lourde a été préconisée. Celle-ci, en plus de l’utilisation

d’ancrages actifs et passifs, consiste en la mise en œuvre d’une galerie bétonnée
pour servir de clé de cisaillement dans le plan du glissement. Cette solution a été
complétée par la réalisation une galerie de drainage et d’injection et la substitu-
tion du matériau inconsistant identifié en pied de talus par du béton. Par ailleurs,
la conception du barrage a été revue pour intégrer la mise en œuvre d’un joint
compressible entre plots.
Très rarement employé et uniquement dans les cas les plus critiques, ce type
de soutènement ne dispose d’aucune méthode de dimensionnement et de peu de
retour d’expérience. L’entreprise en charge de la conception et de la réalisation des
travaux a ainsi mis en œuvre ce soutènement de manière empirique. Cette solution
a toutefois fait l’objet de simulations numériques par ARTELIA, dans l’optique d’une
analyse contradictoire, sur la base de la géométrie du talus avant excavation et des
données d’auscultation enregistrées au cours des travaux.
3.4. CONCLUSION
Les modélisations menées par ARTELIA, après installation du soutènement et

dans le cadre d’un calcul contradictoire, pour étudier les phénomènes d’instabilité
ont mis en évidence une influence relativement négligeable du tunnel servant de
clef de cisaillement (amélioration de 2% de la marge de stabilité). D’un point de vue
physique, il s’agit d’un décalage de la zone de rupture d’avant à après la position
du tunnel. Ce dispositif semble donc peu adapté à la problématique de stabilisation
du massif.
Néanmoins ces travaux de modélisation ont permis de mettre en avant

l’influence significative du système d’ancrages, dont certains ancrages dépassent
les 50 m et sont précontraints à plus de 200 tonnes. Après installation, le facteur
de sécurité global est alors de 2.80 (contre 1.03 avant, 1 signifiant l’équilibre exact
sans marge de sécurité).
Toutefois, l’étude des cas particuliers tels que la vidange rapide du réservoir
entraîne des risques pour la stabilité d’ensemble du massif, laissant craindre des
ruptures d’ancrage et des effondrements. Des dispositions particulières, notamment
304
Q.105 - R.18
Fig. 17
Coupe et vue en plan de galerie utilisée en clé de cisaillement
Section and plan view of the gallery used as shear key.
dans les protocoles de vidange ont été préconisées par ARTELIA. A la fin des
travaux, le barrage a été soumis à trois cycles de remplissage et de vidange de son
réservoir pour analyser l’évolution de la piézométrie et le comportement global de
la rive gauche. Ces opérations se sont déroulées avec succès.
305
Q.105 - R.18
Fig. 18
Modèle de calcul après installation du soutènement
Calculation model after support installation.
4. CONCLUSION
Les aléas de chantier sont courants en phase réalisation et dépendent

généralement de l’évaluation des conditions géologiques et hydrologiques des sites.
La caractérisation des sites de projet est l’un des objectifs majeurs des études de
conception et la précision de sa définition est fonction du niveau d’étude, et des
investigations de terrain associées, engagé avant construction. Toute anticipation
des aléas permettra de définir une conception la plus adaptée des ouvrages.
Ces aléas ont toujours un impact économique sur les projets et le retour
d’expérience sur les trois aménagements présentés dans cet article rappelle
l’importance des études de conception détaillées avant la réalisation d’ouvrages
majeurs tels que les barrages. Le risque d’aléa en construction pourra certes dif-
ficilement être supprimé vu la complexité et la grandeur de ce type d’ouvrages.
Toutefois, il convient toujours d’évaluer au démarrage du projet les risques sécuri-
taires et économiques acceptables pour définir le niveau d’étude nécessaire avant
toute réalisation.
306
Q.105 - R.19
-------
-------
ÉVALUATION DE L’ONDE DE SUBMERSION EN CAS DE RUPTURE DES

OUVRAGES EN REMBLAI
PARTIE I: PRATIQUES DE L’INGÉNIERIE FRANÇAISE POUR LES RUPTURES
DE BARRAGES (∗)
Thomas VIARD, Jean Robert COURIVAUD, Frédéric LAUGIER &

Benoit BLANCHER
Ingénieurs, EDF
Jean Jacques FRY

Expert, indépendant
Pierre SQUILLARI
Ingénieur, GEOPHYCONSULT
FRANCE
RÉSUMÉ
En France, le cadre règlementaire impose aux maîtres d’ouvrage de grands

barrages – de classe A et B – de réaliser des analyses d’ondes de submersion
à travers trois cadres règlementaires : les plans d’évacuation en cas de rupture
(PPI) s’ils font plus de 20 m de hauteur avec un volume de plus de 15 millions de
mètres cube, les analyses de risques des barrages (EDD) et vis-à-vis des sites
nucléaires (guide ASN). Il existe aussi des cas où des ondes de submersion peu-
vent s’avérer utiles hors cadres règlementaires (barrages naturels et dommages
différentiels).
∗ Evaluation of embankment hydraulic structures’ failure flood wave. Part1: French engineer-
ing practices for dam failures.
DOI: 10.1201/9781003211808-65 307

Q.105 - R.19
Pour répondre à ces besoins, deux étapes sont nécessaires : la détermination

de l’hydrogramme de brèche (dont le débit maximum de rupture), et la propagation
de cet hydrogramme à l’aval. Pour les barrages en remblai, les digues de canaux ou
fluviales en permanence en eau, la physique du phénomène étant complexe et mal
connue, l’expertise de l’ingénieur s’appuie sur le retour d’expérience et des formules
empiriques issues de bases de données des ruptures historiques de barrages.
De nombreuses méthodes existent ensuite pour propager les débits à l’aval, la

plus courante faisant référence est la modélisation hydraulique 1D ou 2D résolvant
les équations de Saint Venant, éventuellement comparative (dommages différen-
tiels), mais certaines méthodes simplifiées conservent tout leur intérêt en cas de
crise (barrages naturels) ou lorsqu’un grand nombre de scénarios est à traiter (EDD).
SUMMARY
French regulation imposes dam break flood wave analyses for large dams on
owners through 3 laws: Emergency plan law, Dam risk analysis law, guidance for
nuclear power plant safety. Sometimes, dam break flood waves are also necessary
out of regulations (natural dams or dam break during flood analysis).
Dam break flood wave analyses are done in two steps: hydrograph building
(with maximal flow discharge estimation) and wave propagation. For embank-
ment dams, physics is complex and experience feedback and empirical formula
established with databases analyses have to be used by engineer with caution.
For wave propagation, many techniques exist, but numerical modelling (1D, 2D
with shallow water equations, eventually with comparative approach) is the reference
method. However, few simplified methods are still used in case of emergency (natural
dam) or in case of numerous scenarios to study (dam risk analysis).
1. INTRODUCTION
Depuis plusieurs décennies, le nombre de ruptures de grands barrages dans le

monde entrainant une onde de submersion est de l’ordre de 4 à 5 par an. Le nombre
de ruptures de petits barrages est plus difficile à estimer, car les données sont plus
difficiles à collecter, mais il est certainement bien supérieur. Ainsi, sur la décennie
2010 – 2020, on dénombre au moins 42 ruptures de grands barrages, plus de 300
ruptures de petits barrages et au moins 46 ruptures de barrages de stériles miniers
[1]. La plupart de ces ruptures a causé la disparition de vies humaines et/ou des
dégâts économiques très importants. Toujours au cours de cette dernière décennie,
des incidents graves n’ayant pas entrainé la rupture de l’ouvrage ont conduit à des
évacuations préventives de populations, comme pour les barrages d’Oroville, USA
308
Q.105 - R.19
(2017, 188 000 personnes évacuées), ou de Toddbrook, Royaume-Uni, (2019, 6

500 personnes évacuées).
L’efficacité de ces évacuations de population et de la sauvegarde des biens

économiques en cas de rupture de barrage repose notamment sur la prédiction de la
propagation de l’onde de submersion. La caractérisation de l’inondation consécutive
à cette propagation de l’onde de submersion est donc un élément essentiel parmi les
dispositions de sûreté associées à tout barrage qui présente des enjeux aval signi-
ficatifs. La prédiction de cette inondation est déterminée par un calcul hydraulique,
dont la condition limite amont est l’hydrogramme de brèche (débit s’écoulant en
fonction du temps à travers la brèche formée dans le barrage). Cette brèche con-
stituant une section de contrôle du débit de l’onde de rupture, la cinétique de son
évolution et ses dimensions, qui dépendent directement des processus physiques
qui la génèrent, sont des paramètres clés dans cette prédiction d’inondation.
Les statistiques les plus récentes concernant les ruptures de barrages [2], qui
sont issues de 1443 cas de rupture parmi 50 pays, montrent que la très grande
majorité de ces ruptures concerne des barrages en remblai (90%), dont les modes
de défaillance principaux sont l’érosion externe par surverse (58%), l’érosion interne
(37%) et moins fréquemment le glissement (5%), conduisant ensuite à la formation
d’une brèche. Les outils d’ingénierie permettant de représenter les processus de
rupture d’un barrage en remblai par érosion par surverse ou par érosion interne sont
encore actuellement très limités, certains mécanismes physiques restant encore
peu ou pas décrits et non modélisés.
La difficulté principale pour l’ingénieur en charge d’un calcul d’onde de sub-

mersion d’un barrage en remblai est donc, pour l’évaluation de l’hydrogramme
de brèche comme pour le calcul de l’onde de submersion, de bien connaître les
approches d’ingénierie qui sont à sa disposition, avec leurs hypothèses et leurs lim-
ites, de manière à bien évaluer les incertitudes associées aux résultats de l’étude,
en fonction de l’approche adoptée.
Ce rapport est divisé en deux parties : la partie I présente les pra-

tiques d’ingénierie utilisées en France, et notamment par EDF, pour déterminer
l’hydrogramme de brèche et calculer l’onde de rupture à l’aval des barrages en rem-
blai. La partie II présente plusieurs actions récentes de recherche et développement
qui ont visé à mieux caractériser la rupture par érosion par surverse ou par érosion
interne des barrages en remblai et des levées.
Dans cette partie I, les ouvrages considérés sont les ouvrages en remblai à
charge permanente : barrages réservoirs, remblais de canaux ou digues fluviales
à charge permanente. Après avoir présenté le cadre dans lequel sont réalisées
en France les études d’ondes de submersion, ce rapport s’attache à présenter les
méthodes simplifiées disponibles pour déterminer les débits maximum de brèche
et les hydrogrammes simplifiés de rupture, puis les méthodes de propagation des
ondes de submersion et, enfin, il fournit des indications sur le choix des méthodes
ainsi qu’une réflexion sur les incertitudes existantes.
309
Q.105 - R.19
2. CONTEXTE DE RÉALISATION DES ONDES DE SUBMERSION
2.1. CADRE RÈGLEMENTAIRE EN FRANCE
Aujourd’hui en France, trois principaux textes règlementaires encadrent la

réalisation des études d’ondes de submersion en cas de rupture de barrage.
Historiquement, la rupture du barrage de Malpasset en 1959 est un évène-

ment fondateur de la législation pour la sécurité des barrages en France. Ainsi,
suite à cet évènement, le comité technique permanent des barrages et ouvrages
hydrauliques (CTPBOH, anciennement CTPB) est créé pour rendre un avis sur les
projets de barrages de plus de 20 m de hauteur. A partir de 1968, les barrages
de plus de 20 m de hauteur et d’un volume supérieur à 15 hm3 sont soumis à
obligation de plan d’alerte aux populations en cas de rupture [3] puis aux plans
particuliers d’intervention (PPI) [4]. Ces PPI comprennent la réalisation d’une onde
de submersion définissant trois zones : la zone inondée en moins de 15 minutes
dans laquelle des sirènes doivent se faire entendre en cas de rupture, le linéaire
de rivière sur lequel la rupture du barrage engendre une inondation supérieure aux
crues historiques (traditionnellement une crue de période de retour de 100 ans) et la
zone aval jusqu’à un point correspondant à “ une inondation à risque limité pour les
personnes ” (traditionnellement le CTPBOH demande un débit inférieur à une crue
de temps de retour de 10 ans ou une onde de submersion restant sans déborde-
ment du lit mineur ou du lit endigué de la rivière). La logique des PPI est basée sur
l’anticipation et l’évacuation préalable des populations exposées avec une gestion
de crise sous la responsabilité des préfets (l’exploitant exécute les instructions du
préfet coordonnateur, pilote de la crise).
67 grands barrages exploités par EDF sont soumis à PPI. Certaines vallées
comportent plusieurs de ces ouvrages, dont la rupture hypothétique de l’un d’entre
eux entraînerait des conséquences sur l’ensemble des ouvrages situés à l’aval.
C’est d’ailleurs une hypothèse de l’établissement des PPI. Conformément aux plans
d’organisation interne (POI), des essais réglementaires sont réalisés, tous les 3
mois pour les dispositifs d’alerte aux populations, et tous les mois pour les disposi-
tifs d’alerte aux autorités. Un compte rendu trimestriel des essais est adressé à
l’administration. Chaque équipe concernée par un POI effectue à minima tous les 5
ans un exercice de mise en œuvre de POI.
En parallèle, depuis 2007, les barrages sont classés en différentes catégories

[5], des plus grands (classes A et B) aux plus petits (C et “ non-classés ” ou D
avant 2015). Les barrages de classe A et B sont soumis à “ Etude de Dangers ”
(EDD), c’est-à-dire une analyse de risque fournissant les différents évènements red-
outés centraux (ERC) qui pourraient survenir (rupture de barrage, rupture de vanne,
ouverture intempestive de vanne…). Pour chacun de ces ERC, les conséquences
aval sont évaluées, notamment à travers une onde de submersion pour la rupture de
barrage et, si nécessaire, pour les autres accidents. Ce cadre règlementaire intègre
310
Q.105 - R.19
ainsi la notion “ d’acceptabilité ” du risque à travers une analyse de la probabilité

d’occurrence et de la gravité.
En 2018, un nouvel arrêté [6] met à jour le contenu des EDD et stipule que
les EDD servent de base à l’établissement des PPI, mettant ainsi en cohérence ces
deux premiers cadres règlementaires. L’onde de submersion qui, par convention
dans le cadre des PPI, était réalisée avec un niveau initial aux PHE et une rivière
sur fond sec à l’aval, est à présent le scénario défini dans l’EDD.
Le troisième cadre règlementaire dans lequel des ondes de submersion

doivent être réalisées est la règlementation concernant la sécurité nucléaire [7].
En effet, avec la construction des premières centrales nucléaires dans les années
80, la règle fondamentale de sûreté de 1984 détaille la prise en compte du risque
d’inondation d’origine externe pour les sites nucléaires, le risque de rupture de bar-
rage en faisant partie. Cette règle a fait l’objet d’une mise à jour en 2013 avec le
guide n˚ 13 de l’Autorité de Sûreté Nucléaire (ASN). De légères différences existent
entre les exigences de ce guide et la règlementation PPI. Ce guide demande une
modélisation globale puis une modélisation raffinée locale autour du site nucléaire.
Le modèle local prend en donnée d’entrée les résultats du modèle global. Dans la
majorité des cas, la modélisation issue du PPI (avant l’arrêté EDD de 2018) peut
servir de modèle global pour le site nucléaire.
De légers écarts ont pu exister sur le contenu des ondes de submersion

(rupture sur fond sec ou en crue, niveau initial dans la retenue, comporte-
ment des barrages aval…) en fonction des cadres règlementaires. Une tendance
de fond tend toutefois à une convergence des exigences dans ces différents
cadres.
2.2. HORS OBLIGATION RÈGLEMENTAIRE
Il peut arriver que la simulation d’onde de submersion s’avère utile hors du

cadre règlementaire. Deux cas couramment rencontrés sont : l’analyse des con-
séquences de la formation de barrages naturels (avalanche ou glissement de terrain
barrant le cours d’eau) et la comparaison des conséquences, en crue extrême,
d’une rupture de tout ou partie de barrage en l’état à celles de ce barrage avec son
évacuateur de crues recalibré pour assurer la conformité réglementaire (notion de
dommages incrémentaux, ou dommages différentiels évités).
Le premier cas est lié à un aléa naturel. En effet, régulièrement on observe

des glissements de terrain ou des avalanches dans les vallées montagneuses
comme les Alpes. Ces effondrements, s’ils atteignent le fond d’une vallée ou une
311
Q.105 - R.19
retenue génèrent un barrage naturel (obstruction ou partition) pouvant avoir une

incidence sur la sécurité (remplissage incontrôlé suivi d’une rupture). Ils sont donc
surveillés avec attention et des études d’onde de submersion peuvent s’avérer
nécessaires, parfois en urgence. On peut citer en exemple le glissement de terrain
du pas de l’Ours en 2017 dans la vallée du Guil détaillé dans la seconde partie du
rapport.
Le second cas où la réalisation d’onde de submersion présente un intérêt

concerne le juste redimensionnement des évacuateurs de crues de barrages sur la
base d’une approche coût-bénéfice, le coût reposant principalement sur les travaux
de mise aux normes à réaliser et le bénéfice reposant principalement sur les dom-
mages incrémentaux ou différentiels évités en cas de rupture en crue. Cette piste a
fait l’objet d’un groupe de travail du CFBR entre 2013 et 2017 et de recommanda-
tions provisoires [8]. Cette démarche peut potentiellement faire sens pour relativiser
l’impact des barrages mobiles en rivière qui sont conçus pour être effacés en péri-
ode de crue et qui engendrent une faible perte de charge (par exemple cas d’une
onde de submersion causée par la rupture d’une vanne levante, remise en charge
par l’écoulement). Cela peut être intéressant également pour relativiser l’urgence
(dispositifs et moyens particuliers transitoires) des chantiers de mise en confor-
mité réglementaire des évacuateurs de crues des barrages. Il est ainsi apparu
clairement que certains barrages ne semblent présenter que peu, voire pas, de
conséquences en cas de rupture en crue, principalement en raison de leur capac-
ité de rétention d’eau très limitée comparativement au volume des crues (aléa de
surdébit de rupture de barrage très limité en intensité et dans l’espace) et compte-
tenu du fort dépeuplement des vallées encaissées situées en contrebas (enjeux
quasi-inexistants).
Il existe d’autres cas de figure de réalisation d’onde de submersion (en cas de

rupture d’une poche sous-glaciaire, en cas de rupture de conduite forcée, …) mais
qui restent relativement anecdotiques pour le moment.
3. LA DÉTERMINATION DU DÉBIT ET DE L’HYDROGRAMME DE RUPTURE
La première étape du calcul de l’onde de submersion est la détermination de

sa condition aux limites amont, à savoir le débit s’écoulant à travers la brèche en
fonction du temps, appelé hydrogramme de brèche. Cet hydrogramme peut être
déterminé par une approche simplifiée ou sophistiquée :
– le calcul du débit maximum de brèche suivi de la détermination d’un

hydrogramme simplifié, de forme triangulaire ;
– le calcul d’un hydrogramme à l’aide d’une modélisation couplant les proces-
sus d’écoulement et d’érosion.
312
Q.105 - R.19
3.1. APPROCHE DE L’HYDROGRAMME DE BRÈCHE SIMPLIFIÉ
L’approche de l’hydrogramme de brèche simplifié consiste à calculer dans

un premier temps un débit maximum de brèche puis, dans un second temps, à
déterminer un hydrogramme simplifié. Les deux approches les plus utilisées pour
estimer le débit maximum de brèche sont les suivantes:
– estimation du débit maximum de brèche à l’aide de formules empiriques à

partir de paramètres caractérisant le barrage et sa retenue.
– estimation du débit maximum de brèche à l’aide d’une loi de seuil, d’une
hypothèse sur la géométrie de la brèche (généralement considérée dans
ces approches comme rectangulaire) et d’une estimation de la largeur
maximale de la brèche, cette dernière étant estimée à l’aide de formules
empiriques.
3.1.1. Formules empiriques de débit maximum de brèche
Les formules empiriques disponibles ont été conçues pour estimer de manière
rapide le débit maximum de brèche dans les barrages en remblai. Elles sont en
revanche mal adaptées pour être utilisées sur des remblais de canaux ou des digues
fluviales à charge permanente, en raison principalement des différences entre les
volumes d’eau présents en amont et entre les caractéristiques hydrodynamiques
des écoulements.
Les premières formules empiriques de brèche utilisées par la profession ne

prennent en compte qu’au plus deux paramètres pour déterminer le débit maximum
de brèche : le volume du réservoir et la hauteur du barrage (ou plutôt la hauteur
d’eau en amont du barrage). Elles ont été établies par des régressions linéaires
basées sur ces deux paramètres et en utilisant des bases de données de cas réels
de rupture plus ou moins étendues. En ne tenant pas compte de la résistance à
l’érosion du matériau constitutif du remblai, la plupart de ces formules négligent
ainsi un paramètre essentiel dans le processus d’érosion et dans l’estimation du
débit qui passe à travers la brèche.
Les formules ayant montré les meilleures capacités prédictives sont présen-
tées dans le Tableau 1. Les deux plus récentes prennent en compte la résistance
à l’érosion du sol constitutif du remblai de manière empirique. QP est le débit maxi-
mum de brèche, exprimé en m3 /s. Hw est la hauteur d’eau au-dessus du fond de la
brèche, exprimée en mètres. Hd est la hauteur du barrage, en m. Vr est le volume
du réservoir avant la rupture, exprimé en m3 . V est le volume d’eau au-dessus du
fond de la brèche, en m3 .
313
Q.105 - R.19
Tableau 1
Formules empiriques recommandées de débit maximum de
brèche
Empirical formulas for maximal flow discharge estimation
Auteurs - Authors Formule - Formulas
Froehlich (1995) QP = 0, 607(Vr )0,295 (HW )1,24
CLF (2006) QP = a(Hd )2 + cHd (Vr )0,5
Xu et Zhang simple (2009) QP = 0, 133(g(V )5/3 )0,5 (V 1/3 /Hw )−1,276 eFM+ER
La formule de Froehlich (1995) [9], à deux paramètres (volume de la retenue

et hauteur d’eau au-dessus de la brèche), a été établie sur la base d’une vingtaine
de cas de rupture. Parmi les formules classiques à deux paramètres, c’est celle qui
offre les meilleurs résultats de prédiction.
La formule CLF (Courivaud, Lempérière et Fry, 2006) [10], à quatre

paramètres, présentée au 22ème Congrès CIGB, était une première tentative de
calage réalisé sur 15 cas de rupture. Dans le rapport [10], il est écrit que des efforts
ultérieurs sont nécessaires pour continuer à évaluer les valeurs de ces paramètres,
en partant d’une base de données nettement plus élargie. Le paramètre a, pouvant
faire l’objet d’ajustements, est pris égal à 1. Le paramètre c, représentant de manière
empirique la cinétique d’érosion du remblai, dépend du type de barrage: c = 0,15
pour les barrages en terre peu compactée, sans masque amont ; c = 0,1 pour ceux
dotés d’un masque amont et c = 0,01 pour les barrages zonés, bien compactés et
disposant d’un épais noyau argileux.
La formule de Xu et Zhang (2009) [11], à quatre paramètres, a été établie à

partir d’une base de données de 41 cas de rupture, dont 23 par surverse, le reste
étant par érosion interne. FM est un paramètre dépendant du mode de rupture.
FM = -0,788 si la rupture est par surverse et FM = -1,232 si la rupture est par
érosion interne. ER est un paramètre représentant l’érodabilité. ER = - 0,089 pour
une érodabilité élevée ; ER = -0,498 pour une érodabilité moyenne et ER = - 0,433
pour une érodabilité faible.
L’acquisition de nouvelles données sur les ruptures de barrage, présentées

dans le Tableau 2, est l’opportunité d’améliorer le calage des paramètres de la
formule CLF, à défaut de valider la première estimation de 2006. Ces données
constituent une base de 56 brèches totales par érosion interne ou par surverse.
L’analyse de cette base de données montre d’abord le poids prépondérant du

volume de la retenue. Il existe une relation simple et un coefficient de corrélation
significatif entre le débit de pointe QP (m3 /s) et le volume de la retenue Vr (m3 ).
QP = 0, 0561(Vr )0,6716 avecR 2 = 0, 73]
314
Q.105 - R.19
Tableau 2
Base de données de 56 cas de rupture par érosion interne ou surverse
Data base of 56 dam failed by internal erosion or overtopping
Nom du barrage QP Nom du barrage QP Nom du barrage QP
Dam name (m3 /s) Dam name (m3 /s) Dam name (m3 /s)
Apishapa 6 900 Glashuette 117 Noppikoski 1500
Baldwin Hills 1 130 Goose Creek 565 Norwegian test 1B-03 240
Banqiao 78 000 Gouhou 2050 Norwegian Test 2C-02 74
Bayi 5000 Hatch Town 3 100 Opuha 1900
Belci 3 200 Hatfield 3 400 Oros 40 000
Big Bay 4 200 Hell Hole 7 000 Qielinggou 2000
Break Neck 2 37 Horse Creek 3 890 Quail Creek dike 3 100
Briseis 220 Huqitang 50 Rito Manzanares 180
Castlewood 2 3 600 Johnstown 10 000 Salles de Oliveira 7 200
Chenying 1200 Ka Loko 770 Shimantan 30 000
Danghe 2500 Kodaganar 1280 Staw Starzycki 118
Davis reservoir 510 Laurel Run 1160 Swift 25 000
Dehli 1 950 Lawn Lake 510 Taum Sauk 7 700
Dongchuankou 21000 Lijaju 2950 Teton 65 000
El Guapo 64 000 Liujiaitai 28000 Union Bay 1 420
Euclides da Cunha 1 020 Machhu 2 87 000 Whitewater Brook upper 70
Frankfurt 79 Mahe 4950 Zuguou 11200
French Landing 929 Mammoth 2 520 Zuocun 23600
Frias 400 Martin Cooling pond 3 120
Fig. 1
Courbes de fréquence cumulée du coefficient c (CLF 2020).
Cumulated frequency curves of coefficient c (CLF 2020).
Erosion/Erosion Surverse /Overtopping interne / Internal
Coefficient c A C
Fréquence cumulée/Cumulated frequency B D
L Low Erodability Erodabilité faible
M Medium erodability Erodabilité moyenne
H High erodability Erodabilité forte
315
Q.105 - R.19
Le rapport Qp /(V0,5
r Hd ) permet de visualiser une première évaluation de la
plage de variation du paramètre c. Cette plage de variation est réduite en répartis-
sant les données par type de rupture et par classe d’érodabilité du remblai (Fig. 1)
et en prenant la hauteur d’eau et le volume d’eau juste avant rupture.
La figure 1 souligne l’incertitude résultant du choix de la valeur médiane pour

le coefficient c. Cette incertitude est forte car elle cumule l’erreur sur les valeurs
observées, en particulier le débit (qui peuvent varier de 1 à 3 selon les observateurs),
au manque d’information sur les matériaux constitutifs (méconnaissance de l’indice
de plasticité, absence d’essais d’érosion) et à la rusticité des classes de barrage.
Le choix du coefficient a ne diminue pas clairement cette incertitude. Connaissant
l’importance de cette incertitude, une expression plus simple de la formule CLF est
proposée, avec un coefficient a=0 et un coefficient c dont les valeurs sont regroupées
dans le tableau 3:
QP = cHd (Vr )0,5 CLF(2020) (1)
En reprenant les valeurs des coefficients a et c de la relation CLF(2020), le

carré du coefficient de corrélation entre débits prédit et observé atteint R2 = 0, 858
avec les 56 données. La corrélation est meilleure pour la rupture par érosion interne :
R2 = 0, 885. Ce résultat est légèrement meilleur que celui obtenu par la relation de
Froehlich et équivalent à celui de Xu et Zhang.
Tableau 3
Valeurs du coefficient c de la formule CLF(2020) avec a=0
Values of the coefficient c used in CLF(2020) formula (a=0)
Classe d’érodabilité Type de barrage Surverse Erosion interne
Class of erodability Type of dam Overtopping Internal erosion
L Zoné, noyau en argile bien compacté c=0,035 c=0,015
M CFRD, zoné noyau en silt c = 0,06 C= 0,04
H Homogène peu ou pas compacté c=0,1 c=0,08
Enfin, la formule CLF(2020) fournit la meilleure prédiction pour les deux plus
grands débits de rupture connus : Machhu 2 (Inde), QP = 87 000 m3 /s et Banqiao
(Chine), QP = 78 000 m3 /s (Tableau 4).
Tableau 4
Comparaison des deux plus grands débits observés avec ceux
prédits
Comparison between the two largest observed and estimated
peak flows
Barrage QP mesuré (m3 /s) QP Estimé QP Estimé QP Estimé
Dam Observed QP Estimated QP Estimated QP Estimated QP
(m3 /s) Froehlich 1995 Xu et Zhang CLF(2020)
Machhu 2 87 000 7 600 14 100 18 600
Banqiao 78 000 16 700 53 000 76 100
316
Q.105 - R.19
Le fort écart entre la prédiction et la mesure du débit à Machhu 2 incite à

intégrer la longueur du barrage dans la formule, pour prendre en compte l’occurrence
de plusieurs brèches. En connaissant l’incertitude de ces formules, il est prudent,
lors de chaque étude, d’utiliser l’ensemble des trois formules précédentes. Ensuite,
l’ingénieur jugera de la plage de valeurs du débit maximum de brèche à retenir en
fonction du site et des conséquences de la rupture.
3.1.2. Estimation du débit maximum de brèche par une loi de seuil
Cette approche est utilisée à EDF plus spécifiquement pour les remblais de
canaux et les digues fluviales à charge permanente, pour lesquelles les formules
empiriques de débit maximum de brèche présentées ci-dessus pour application à
des barrages ne sont pas bien adaptées.
Le débit maximum de brèche peut être estimé à partir d’une largeur maxi-
male de brèche et de l’application d’une loi de débit de seuil. Dans cette approche,
l’incertitude provient principalement de l’estimation de la largeur maximale de brèche
et reste importante.
Deux approches peuvent être considérées, selon la configuration de l’ouvrage

étudié :
1. La partie en remblai qui est soumise à un écoulement de surverse est

encadrée par des frontières non-érodables (structures en béton) relative-
ment proches (quelques dizaines à centaines de mètres). Dans ce cas,
la largeur maximale de brèche est bornée, en première approche, par la
distance entre les deux structures rigides (estimation généralement très
conservative dont la pertinence doit être évaluée en fonction de chaque
site).
2. La largeur maximale de brèche peut être estimée à l’aide de formules
empiriques.
Lorsque la première approche n’est pas applicable (cas le plus fréquent),

seule la seconde méthode est utilisée, en considérant une incertitude estimée par
l’ingénieur en fonction du contexte de l’ouvrage.
Les formules empiriques proposées par la littérature pour estimer la largeur

maximum de brèche sont présentées dans le tableau 5. B désigne la largeur max-
imale de brèche. H représente la hauteur entre le niveau d’eau en amont de la
brèche et le fond de la brèche. V désigne le volume du réservoir, en m3 . La for-
mule de Froehlich, qui recourt au volume du réservoir en amont de l’ouvrage est à
réserver aux remblais de canaux ou de rivières compartimentées en biefs séparés.
Les formules de la FERC et de l’USBR, qui ne prennent en compte que le paramètre
H, sont à utiliser avec beaucoup de réserve dans le cas des digues. En complément
à l’utilisation de ces formules, l’ingénieur peut également considérer qu’en milieu
317
Q.105 - R.19
fluvial, la largeur maximale de brèche dans une digue sera de l’ordre de grandeur
de la largeur du lit mineur.
Tableau 5
Formules empiriques de largeur de brèche
Empirical formulas for breach width
Formule de la FERC Valeurs les plus probables : 2 × H ≤ B ≤ 4 × H
Valeurs extrêmes : H ≤ B ≤ 5 × H
Formule de l’USBR B=3×H
Formule de Froehlich B = 0, 1803 × K × V 0,32 × H 0,19 avecK = 1, 4
pour les ruptures par surverse
La formule de débit de seuil qu’il est recommandé d’utiliser est la formule

utilisée dans le code WinDAM [12], à savoir :
Qmax = g 1/2 L(2H /3)3/2
Qmax est le débit maximum de brèche, en m3 /s ; L est la largeur de brèche en

mètres, H est la hauteur entre le niveau d’eau à l’amont de la brèche et le niveau
d’eau aval, en mètres.
3.1.3. Hydrogramme de brèche simplifié
Pour les barrages en remblai, deux approches simplifiées sont utilisables:
1. estimation du débit maximum de brèche à l’aide d’une formule empirique,

puis construction d’un hydrogramme de brèche simplifié, de forme triangu-
laire.
2. estimation du débit maximum de brèche à l’aide d’une loi de seuil et
d’une estimation de la largeur maximale de brèche, puis construction d’un
hydrogramme de brèche simplifié se basant sur une vitesse moyenne
d’élargissement de brèche.
A partir du majorant et du minorant choisis pour encadrer le débit maximum

de brèche, un hydrogramme de brèche triangulaire (Fig. 2) peut être défini par les
trois points suivants:
– (T0 , QP = 0) : point origine de l’hydrogramme ;

– (T1 , Q = Qmax ) : point du pic de débit. T1 est estimé égal à T0 + d, d étant
la durée de montée du débit jusqu’au pic. Cette valeur de d est estimée à
dire d’expert, dans une plage de valeurs variant généralement d’environ 30
minutes à quelques heures, en fonction des informations disponibles sur la
hauteur de l’ouvrage, sa section transversale et la connaissance que l’on a
de la résistance à l’érosion des matériaux qui le constituent.
318
Q.105 - R.19
Fig. 2
Construction d’un hydrogramme de brèche simplifié.
Simplified hydrograph for breach.
– (T2 , Q = 0) : point de fin de l’hydrogramme (retour à un débit nul). T2 est

déterminé de manière à ce que l’intégrale de l’hydrogramme soit égale au
volume du réservoir.
Lorsque le débit maximum de brèche a été évalué à partir d’une estima-

tion de la largeur maximale de brèche, l’hydrogramme simplifié est déterminé en
considérant des ordres de grandeur de la vitesse moyenne d’élargissement de la
brèche. Ces ordres de grandeur de vitesse d’élargissement sont issus de l’analyse
de données disponibles sur des cas de rupture de barrage ou de digues fluviales
à charge permanente (cette analyse n’a pas encore fait l’objet de publication) et
sont estimés en fonction de la résistance à l’érosion de l’ouvrage, selon les quatre
catégories suivantes :
– Barrages anciens constitués de matériaux très érodables (compactés sans

moyen mécanique) : 10 à 15 m/min.
– Barrages modernes (compactés avec des moyens mécaniques) mais con-
stitués de matériaux érodables à très érodables : 0,5 à 3 m/min.
– Barrages modernes constitués d’un corps homogène ou d’un noyau moyen-
nement résistant à l’érosion : 0,1 à 0,5 m/min.
– Barrages modernes constitués de matériaux très résistants à l’érosion : 0,01
à 0,1 m/min.
3.2. APPROCHE DU CALCUL D’HYDROGRAMME DE BRÈCHE PAR UNE MODÉLISATION

COUPLANT ÉCOULEMENT ET ÉROSION
A l’heure actuelle, pour les barrages en remblais, la modélisation détermin-

iste basée sur la physique des mécanismes d’initiation et de développement d’une
brèche dans le remblai reste généralement non validée et donc indisponible pour
des études opérationnelles d’ingénierie, excepté pour l’érosion par surverse d’un
remblai homogène, de géométrie simple, constitué de matériaux cohésifs. Pour
ce dernier mécanisme, un important effort de recherche a été mené depuis la fin
des années 1990 sur la description physique des phénomènes et leur modélisation
319
Q.105 - R.19
expérimentale à grande échelle et numérique, notamment au laboratoire HERU de

l’USDA-ARS à Stillwater, OK, USA. Cet effort de recherche a permis d’aboutir à la
validation de deux codes de calcul adaptés aux besoins de l’ingénierie, WinDAM
[12] et EMBREA∗ [13], dans le cadre d’un projet international d’évaluation mené par
CEATI [14] et auquel EDF a contribué.
Pour les autres mécanismes d’érosion (érosion par surverse dans des matéri-
aux peu ou non cohésifs et érosion interne), aucun code de calcul adapté à
l’ingénierie n’a encore fait l’objet d’une évaluation et d’une validation similaire.
Un calcul déterministe à l’aide d’une modélisation basée sur la physique et

validée au niveau international est préconisé, avec la mise en œuvre du code
WinDAM ou du code EMBREA, dans les hypothèses suivantes :
– l’ouvrage est un barrage réservoir ou le barrage latéral d’un canal ou d’un

bief délimité de fleuve endigué ;
– l’ouvrage peut être considéré comme étant constitué d’un matériau
homogène fin et cohésif ;
– les parements amont et aval peuvent être considérés par un fruit constant
sur chaque face ;
– la condition de chargement hydraulique amont peut être représentée par
une courbe de capacité de la retenue, une condition initiale de cote du plan
d’eau et un hydrogramme de crue entrant défini à partir de l’instant initial de
la modélisation ;
– le sol constitutif du remblai est caractérisé par sa densité humide, sa cohé-
sion non drainée (Cu) et les deux paramètres de résistance à l’érosion
mesurés par des essais de Jet Erosion Test : KD (qui représente la cinétique
de l’érosion) et TauC (qui représente la contrainte de cisaillement critique à
partir de laquelle débute l’érosion).
Ces deux codes mettent en œuvre une modélisation physique conceptuelle

3D du processus d’érosion de type “ Head Cut Migration ”. Ils se basent sur
une représentation 2D verticale de la géométrie du remblai, des caractéristiques
géotechniques du matériau constitutif du remblai et des conditions hydrauliques
amont et aval. Ils prennent également en compte la résistance à l’érosion
par surverse due au type de revêtement sur la face aval et sur la crête de
l’ouvrage : enherbement, rip-rap. Ils fournissent comme résultats l’hydrogramme
de l’écoulement qui franchit l’ouvrage (qu’il y ait formation de brèche ou pas) ainsi
que la géométrie de la zone érodée (largeur, profondeur).
∗Le projet de validation mené par CEATI a évalué et qualifié le code HR Breach développé
par HR Wallingford qui a ensuite évolué vers le code EMBREA qui est actuellement
commercialisé par HR Wallingford.
320
Q.105 - R.19
4. PROPAGATION DE L’HYDROGRAMME EN AVAL DE LA BRÈCHE
4.1. CHOIX DES MÉTHODES
Différents outils existent pour modéliser une onde de submersion. Le plus

simple est l’utilisation de formules empiriques ou analytiques et d’abaques. Le plus
répandu est la modélisation numérique 1D ou 2D résolvant les équations de Saint-
Venant (plus rarement les équations de Navier Stokes en 3D). Les plus complexes
pouvant être la modélisation numérique de type Smoothed Particule Hydrodynamics
(SPH) ou la modélisation physique à échelle réduite. Deux facteurs principaux
entrent dans le choix de la méthode. Premièrement le temps disponible ; en effet,
en période de “ crise ” avec risque de rupture imminente sans information existante,
comme cela peut être le cas pour un barrage naturel, les méthodes rapides de type
analytiques vont naturellement être privilégiées. Si le temps disponible pour réaliser
une étude plus approfondie est suffisant, la précision et la qualité des données
d’entrée vont orienter le choix. La modélisation numérique qui a fait ses preuves
pour modéliser ce type de phénomène est en général privilégiée. La rupture du
barrage de Malpasset, très bien documentée, fait ainsi partie des cas test de logi-
ciels de modélisation 1D tel que MASCARET ou 2D tel que TELEMAC [15]. Les
ondes de submersion concernent régulièrement plusieurs dizaines voire centaines
de km de cours d’eau, écartant de fait les solutions de type 3D, SPH ou modèle
physique.
4.2. LA MODÉLISATION NUMÉRIQUE
La modélisation numérique est aujourd’hui la méthode de référence pour

étudier une onde de submersion. Dans la majorité des cas, une modélisation 1D suf-
fit. En effet, beaucoup de barrages sont situés dans des vallées plutôt encaissées où
les écoulements suivent bien le fond de la vallée sans recirculation ou autre écoule-
ment complexe. Toutefois, des modélisations 2D peuvent s’avérer nécessaires dans
les grandes plaines d’inondation, en cas de fort débordement des cours d’eau. His-
toriquement, les premières ondes de submersions ont été réalisées pour des très
grands barrages (PPI) engendrant des débits très supérieurs aux débits du cours
d’eau en crues extrêmes. Ainsi pour ces barrages, aucune donnée ne permettait de
caler le modèle numérique ; et le débit initial présent dans la rivière n’influençait pas
les résultats. Les modélisations étaient réalisées sur fond sec avec des coefficients
de Strickler choisis à dire d’expert. Les incertitudes liées à cette absence de calage
incitent à prendre des précautions : le CTPBOH demande pour les PPI des hau-
teurs d’eau majorées de 15 % et des temps d’arrivée de l’onde de crue diminués
de 13 % pour prendre des marges de sécurité, utiles pour cadrer l’évacuation des
populations en cas de survenue d’accident. Pour les barrages de taille plus mod-
este, notamment les barrages mobiles en rivière, leur rupture engendre des débits
comparables à de fortes crues du cours d’eau. Ainsi des données de calage sont
321
Q.105 - R.19
souvent disponibles et il est préférable de caler le modèle plutôt que d’appliquer

la majoration des hauteurs d’eau. En effet, une majoration de 15% de la hauteur
d’eau par analogie avec les PPI pour une onde de submersion en limite de déborde-
ment du lit mineur ou en léger débordement n’est pas adaptée et peut amener à très
fortement surévaluer la zone inondée, notamment dans les plaines d’inondation peu
vallonnées. Dans le cas de ces barrages de taille modeste, les conditions initiales
peuvent aussi fortement influencer les conséquences de la rupture. Dans ces cas,
les études de danger peuvent fournir des informations sur les scénarios de rupture
les plus probables et donc à étudier.
4.3. MODÉLISATION SPÉCIFIQUE AUX DOMMAGES INCRÉMENTAUX OU

DIFFÉRENTIELS
Un Groupe de Travail du CFBR a produit des recommandations provisoires

en date d’octobre 2017 [8] pour encadrer l’application de cette méthode.
La démarche consiste à réaliser des simulations des crues et des ruptures de

barrages en l’état actuel et en supposant un recalibrage des évacuateurs de crues, à
établir les dommages aux personnes et aux biens dans les différentes configurations
et à évaluer l’optimisation possible au regard d’analyse de proportionnalité des coûts
et d’analyse coût-bénéfice, le coût se résumant bien souvent aux travaux de mise
en conformité et le bénéfice (ou le gain de sûreté) aux dommages évités par le
recalibrage (différentiel de dommages intégré sur l’échelle de probabilité). Il ne s’agit
donc pas d’un différentiel avec et sans présence du barrage ou d’un différentiel avec
et sans rupture du barrage mais bien d’un différentiel avec et sans recalibrage des
évacuateurs de crues.
Les ruptures de barrages s’envisagent en crue et peuvent être partielles (en

fonction des fragilités admises des différentes parties, comme la remise en charge
des vannes levantes par exemple) ou totales (en l’absence de justification), instanta-
nées ou progressives. Plusieurs hypothèses sont à établir : cote de danger au-delà
de laquelle la stabilité ne peut plus être garantie, influence de l’ennoiement aval
sur la limitation des débits de rupture, influence des débits des affluents en aval
en crue, calage des coefficients de rugosité sur la base des laisses de crues sig-
nificatives et du boisement des berges, influence des ruptures des barrages et des
ponts en aval, fonctions de dommages aux biens en fonction des hauteurs d’eau et
des vitesses d’écoulement, formule empirique d’évaluation des personnes poten-
tiellement impactées, fréquentation des abords des cours d’eau en crue (attirance
du public pour ces écoulements spectaculaires), conditions d’évacuation des étab-
lissements recevant du public, influence des incertitudes sur la topographie, …Par
ailleurs, il est nécessaire de procéder à une mise à jour régulière de l’analyse pour
tenir compte des évolutions non maitrisables des enjeux (évolution des plans locaux
d’urbanisme, …).
322
Q.105 - R.19
Fig. 3
Abaque de décroissance du débit en fonction de la distance au barrage.
Flow discharge decrease depending on dam distance.
4.4. LA MÉTHODE SIMPLIFIÉE DU CTGREF
Cette méthode historique date de 1978 [16]. Basée sur des abaques, elle per-
met en quelques heures d’évaluer le risque en cas de rupture de barrage. Une fois
le débit de rupture (Qb ) déterminé à partir d’une formule empirique ou d’une modéli-
sation, pour une rivière de pente J et de frottement K réguliers, la décroissance de
ce débit (Qmax /Qb ) est une fonction de la distance X au barrage adimensionné par
la racine cubique du volume V de la retenue (X/V1/3 ) et d’un facteur défini comme
la pente multipliée par le frottement au carré (JK2 - Fig. 3).
Une fois le débit connu en un point d’enjeu (habitations, ville, village, barrage
aval…), la hauteur d’eau locale peut être approximée par une hauteur d’eau normale
pour estimer la zone inondée. Cette méthode présente de fortes incertitudes, des
marges de sécurité importantes doivent donc être prises, mais elle permet une
grande réactivité en cas de crise.
323
Q.105 - R.19
Fig. 4
Exemple de simulation Hayami.
Example of Hayami simulation.
4.5. LA MÉTHODE SIMPLIFIÉE HAYAMI
Un module de propagation 1D des débits en rivière a été développé par le Lab-

oratoire National d’Hydraulique et Environnement (LNHE) d’EDF à Chatou. L’objectif
de ce module est de disposer d’un calcul robuste, facilement paramétrable (sans
bathymétrie) et très rapide en temps de calcul, permettant de modéliser le transfert
de débit entre deux points éloignés d’une rivière (sans s’intéresser particulièrement
aux valeurs de débit ou de niveaux intermédiaires).
Le modèle Hayami classique à coefficients variables est un modèle de

propagation-diffusion des débits avec des coefficients de célérité et de diffusion
dépendant du débit. Il donne des résultats très satisfaisants pour les crues. En
revanche ce modèle n’est pas conservatif sur les volumes d’eau pour les très forts
gradients de débit (fronts raides de type échelon), ce qui est le cas des ruptures de
barrage instantanées.
Pour la propagation des fronts raides (ruptures instantanées), un nouveau

modèle, propageant une grandeur homogène à une section mouillée (débit divisé
par une célérité) a été mis au point. Ce modèle conservatif permet de reproduire
(pour les géométries simples où ils sont comparables) les résultats obtenus par une
résolution complète des équations de Saint-Venant.
324
Q.105 - R.19
Cette solution, intermédiaire entre la méthode du CTGREF et la modélisa-

tion numérique peut s’avérer une bonne alternative lorsque le nombre de scénarios
étudiés dans une Etude de Dangers est important, entrainant de nombreuses ondes
de submersion différentes, ou pour déterminer quel barrage présente le risque
principal pour un site nucléaire et doit être étudié en détails.
5. ANALYSE DES INCERTITUDES ET PRECONISATIONS
L’ingénieur doit garder à l’esprit que les formules empiriques de débit maximum
de brèche pour les barrages en remblais ne sont capables au mieux que de fournir
des ordres de grandeur. Leur utilisation n’assure pas l’obtention de débits maximum
de brèche conservatifs. Les écarts entre prédiction et valeur issue des observations
peuvent atteindre un rapport de 2 à 10. Il semble préférable d’utiliser l’ensemble des
formules présentées ci-dessus, qui fournissent actuellement les meilleurs résultats
de prédiction et de ne jamais fournir une estimation qui serait le résultat d’une seule
formule. Après avoir utilisé plusieurs formules, l’ingénieur doit faire son propre juge-
ment de la plage de valeurs de débit maximum de brèche à retenir. Le choix de
l’hydrogramme de brèche simplifié doit prendre en compte cette incertitude en four-
nissant au moins deux hydrogrammes, basés sur les valeurs supérieure et inférieure
du débit maximum de brèche.
Pour la propagation aval des débits, il est important de prendre des marges
de sécurité dans deux cas : en cas de méthode simplifiée (CTGREF ou Hayami)
et en cas d’absence de données de calage (très grands barrages). Ainsi, seule
la modélisation numérique calée d’un barrage de moyenne importance engendrant
des débits comparables aux fortes crues du cours d’eau et la méthode incrémentale
ne présentent pas de fortes incertitudes.
6. CONCLUSION
Le sujet des ondes de submersion est un sujet qui a souvent été traité [9],
[17], mais qui reste toujours d’actualité avec quelques ruptures de grands barrages
par an dans le monde. La France a su tirer l’expérience des ruptures sur son sol
(Malpasset) et ailleurs dans le monde pour se munir d’un arsenal règlementaire
permettant de gérer ce risque. Toutefois, même avec une littérature foisonnante,
ce phénomène extrême présente toujours de nombreuses incertitudes, notamment
pour les débits de rupture des barrages en remblai qui représentent la majorité
des ruptures actuellement dans le monde. Ainsi, l’expérience et l’expertise des
ingénieurs dans ce domaine restent un besoin important permettant d’apporter une
réponse adaptée à chaque situation règlementaire ou autre.
325
Q.105 - R.19
REFERENCES
[1] Vogel A., Courivaud J.R., Jarecka K., Notable Dam Failures in the Last Decade,
Dam World 2020, Lisbon, September 21-25, 2020.
[2] Zhang L., Pend M., Chang D., Xu Y., Dam Failure Mechanisms and Risk
Assessment, Wiley, 499p., 2016.
[3] Décret n˚ 68-450 du 16/05/1968
[4] Loi 87-565 du 22/07/87, Décret n˚ 92-997 du 15/09/1992, Arrêté du

01/12/1994, Arrêté du 02/02/2002 et Décret n˚ 2005-1158 du 13/09/2005
[5] Décret n˚ 2007-1735 du 11/12/2007, Arrêté du 12/06/2008, Décret n˚ 2015-

526 du 12/05/2015 et Arrêté du 03/09/2018
[6] Arrêté du 06/08/2018
[7] Règle Fondamentale de Sûreté 1.2.e – 1/04/1984 et Guide ASN n˚ 13, 2013
[8] Dimensionnement des évacuateurs de crue de barrage par les dommages

incrémentaux ou différentiels : Recommandations pour la mise en œuvre d’une
méthode applicable aux barrages en France, CFBR, 2017
[9] CIGB, Etude de l’onde de rupture d’un barrage, Synthèse et recommandations,

Bulletin n˚ 111, 1998.
[10] Courivaud J.-R., Lempérierre F., Fry J.-J., A New Analysis of Embank-
ment Dam Failures by Overtopping. 22ème Congrès International de la CIGB,
Barcelone, 16-23 juin 2006, Q. 86 – R. 66.
[11] Xu Y., Zhang L., Breaching parameters for earth and rockfill dams, J. of
Geotechnical and Geoenvironmental Engineering, Vol. 135, N◦ . 12, 2009.
[12] WinDAM C Software | NRCS, https://www.nrcs.usda.gov
[13] EMBREA, http://www.dambreach.org
[14] CEATI Dam Safety Interest Group, Evaluation of Numerical Models for Simulat-
ing Embankment Dam Erosion and Breach Processes. DSO-2017-02, USBR,
2017.
[15] TELEMAC – MASCARET opensource software, http://www.opentelemac.org/
[16] Calcul d’une onde de submersion due à la rupture d’un barrage, CTGREF,
ministère de l’Agriculture, 1978
[17] Bernchmark Workshop on numerical analysis of dams – ICOLD – 02-

04/10/2013 – Theme C – Estimation for Risk Assessment: Numerical
Modelling, Uncertainty Quantification, and Communication of Results
326
Q.105 - R.19
1. INTRODUCTION
For several decades, there have been approximately 4 to 5 flood waves each
year, consecutive to the failure of large dams. The number of failures of small dams
is harder to estimate since data is scarcer but is certainly much higher. During the
2010-2020 decade, there have been at least 42 large dams’ failures, more than
300 failures of small dams and at least 46 failures of tailings dams [1]. Most of the
failures led to casualties and/or major economic damage. During this same decade,
major incidents, not leading to failure of the dam, conducted to preventive population
evacuations, such as at Oroville dam, USA (2017, 188 000 people evacuated), or
Toddbrook dam, UK (2019, 6 500 people evacuated).
The efficiency of population evacuations and of the safeguarding of economic

goods in case of a dam failure, strongly rests upon the prediction of the flood
wave’s propagation. The characterisation of the resulting flood is thus an essential
component among safety measures associated with high hazard dams. This flood
prediction is determined by hydraulics calculation, the breach hydrograph being
the upstream boundary condition (evolution with time of the flow through the dam
breach). The breach standing as a control section of the failure wave’s flow, the kinet-
ics of its evolution and its dimensions, directly dependant on the physical processes
that generate it, are key parameters in the flood wave prediction.
Most recent statistics about dam failures [2], taken from 1443 failures in 50
countries, show that the vast majority of failures occur on earthen dams (90 %), with
the main failure mechanisms being overtopping (58 %), internal erosion (37 %), and
less frequently sliding (5 %), ending with the formation of a breach. Engineering
tools allowing to represent the failure processes of an earthen dam by overtopping
or internal erosion are still very limited, some physical mechanisms remaining little
or not described nor modelled.
The main difficulty for the engineer in charge of flood wave calculations of an
earthen dam is then, for the breach hydrograph as well as for flood wave calculation,
to have sufficient knowledge about available engineering approaches, their under-
lying hypothesis and their limits, in order to evaluate the uncertainties associated
with the study results, depending on the approach used.
This article is divided into two parts: part I will present engineering practices
used in France, especially by EDF, to determine the breach hydrograph and calculate
the flood wave downstream of earthen dams. Part II will present several recent R&D
actions undertaken to better characterise overtopping or piping failures of earthen
embankments.
In this first part, only earthen structures with a permanent hydraulic loading
are considered: reservoir dams, earthen canals, or fluvial dikes with permanent
hydraulic loading. After presenting the French regulatory framework for flood wave
327
Q.105 - R.19
studies, this article will present the simplified methods available to estimate the
breach peak outflow and the breach hydrographs, then methods of flood wave prop-
agation and finally, it provides some guidelines about the choice of methods as well
as a discussion on uncertainties.
2. CONTEXT OF FLOOD WAVE STUDIES
2.1. FRENCH REGULATORY FRAMEWORKS
Nowadays in France, three main regulatory texts supervise the dam failure
flood analysis.
Historically, Malpasset (France) dam failure in 1959 is a founding event for

the French dam safety legislation. Following this event, the Permanent Technical
Committee of Dams and Hydraulic Structures (CTPBOH in French) was created to
provide notice on projects of dams higher than 20 m. Since 1968, for dams higher
than 20 m with a reservoir capacity larger than 15 hm3 , the owner must provide a
population alert plan in case of failure [3] as well as a Particular Intervention Plan
(PIP) [4]. PIPs include a dam failure flood analysis defining 3 zones: a less than
15 minutes flooded zone where sirens must be heard in case of failure; the linear
of the downstream river where the failure of the dam would cause a flood greater
than historical ones (typically 100-year flood); and the downstream zone until a point
corresponding to “a flood with limited risk for the population” (traditionally, CTPBOH
asks for a flow lower than a 10-year flood or a flood wave without overflowing the
minor bed or the diked bed of the river). PIPs’ logic is based on anticipating and
prior evacuation of the exposed population with a risk management under the local
state authority’s responsibility (the owner executes the instructions of the local state
authority, who manages the crisis).
67 dams operated by EDF must provide PIPs. Some valleys have several
structures where the hypothetical failure of one of them would lead to consequences
on all the structures downstream. This is indeed a hypothesis taken into account in
PIPs. In accordance with Internal Organisation Plans (IOPs), regulatory trials are
conducted, every 3 months for the alert of the population devices, and monthly
for the alert of authorities’ devices. A quarterly report of the trials is handed to
the administration. Each operation team having to perform IOPs has to conduct a
training test at least every 5 years.
In parallel, since 2007, dams have been classified in different categories [5],
from the tallest (A and B) to the smallest (C and “unclassified” or D before 2015).
Class A and B dams must provide a Danger Study (DS), which is a risk analy-
sis detailing the different central feared events (CFE) that may arise (failure of the
dam, failure of a gate, undesired opening of a gate…). For each CFE, downstream
328
Q.105 - R.19
consequences are evaluated, especially through a dam failure flood analysis and,
if necessary, for other accidents. This regulatory framework takes into account the
notion of “acceptability” of the risk through an analysis of the probability of occurrence
and of the gravity.
In 2018, a new decree [6] updates the contents of DS and stipulates that DS
are the basis for the creation of PIPs, thus linking together these two regulatory
frameworks. The dam failure flood analysis that was calculated by convention with
the Highest Water Levels and dry downstream river in PIPs, is now the scenario
used in DS.
The third regulatory framework in which dam failure flood analysis must be
performed is included in the nuclear safety regulations [7]. Indeed, with the first
nuclear power plants constructed in France during the 80’s, the fundamental safety
rule of 1984 details how to take into account a flood from external origin for nuclear
plants, dam failure floods being included in it. This rule was updated in 2013 with
the Guide n˚ 13 of the Nuclear Safety Authority. Slight differences exist between
the requirements of this guide and the PIP regulation. This guide requires a global
modelling and then a refined one around the nuclear plant. The local model uses the
results from the global model as input data. In most cases, PIP modelling (before
the DS decree of 2018) can be used as the global model for the nuclear site.
Minor differences may have existed between dam failure flood analysis (failure
on dry or flooded river, initial water level in the reservoir, behaviour of downstream
dams…) depending on the regulatory framework. A global trend now tends to
converging requirements within these different frameworks.
2.2. OUTSIDE REGULATORY OBLIGATIONS
It may occur that the dam failure flood analysis of interest outside of reg-
ulatory frameworks. Two frequently encountered cases are: the analysis of the
consequences of the formation of natural dams (avalanche or landslide blocking
the stream) and the comparison of the consequences, during an extreme flood, of
the failure of part or of the entire dam body, with the consequences of the same
dam considering a recalibrated spillway to ensure regulatory conformity (notion of
incremental damages or avoided differential damages).
The first case is linked to a natural hazard. As a matter of fact, avalanches or

landslides are often observed in mountainous valleys such as the in the Alps. These
collapses, if they reach the bottom of the valley or a reservoir, create a natural dam
(partial or total obstruction) that may have safety consequences (uncontrolled filling
followed by a failure). They are thus carefully monitored, and flood wave analysis
may prove necessary, sometimes in emergency. The Pas de l’Ours landslide in 2017
in the Guil valley can be cited as an example and is detailed in the second part of
this article.
329
Q.105 - R.19
The second case where flood wave studies show an interest is for the correct
re-design of spillways on the basis of a cost-benefit approach, the cost relying mainly
on the compliance of rehabilitation works and the benefit being mostly the incremen-
tal damages or the avoided differential damages in case of failure during a flood.
This issue was addressed by a CFBR (French ICOLD Committee) working group
between 2013 and 2017 and by temporary recommendations [8]. This process may
potentially make sense for gate-structure dams on rivers, that are designed to be
transparent during a flood, creating only a small water head drop (for instance in
case of a flood wave caused by the failure of a rising gate, put under load by the
flow). It may also be interesting to relativise the urgency (temporary devices and
specific means) of regulatory compliance construction works of dams’ spillways. It
thereby appears that some dams seem to present little to no consequences in case
of a failure during a flood, mainly because of their limited water retention in com-
parison with the flood volume (hazard of an increased flow caused by the failure
very limited spatially and in time) and taking into account the depopulation of some
valleys downstream (almost non-existent stakes).
Other scenarios exist of a flood wave study (rupture of an ice pocket, rupture
of a penstock, …) but they remain relatively anecdotal for the moment.
3. DETERMINATION OF THE FLOW AND OF THE FAILURE HYDROGRAPH
The first calculation step of the flood wave is the determination of its upstream
boundary conditions, namely the flow through the breach over time, called the breach
hydrograph. This hydrograph can be determined through a simplified or a physically-
based approach:
– the calculation of the breach peak outflow followed by the determination of a

simplified hydrograph, with a triangular shape;
– The calculation of the hydrograph with a modelling coupling flow and erosion
processes.
3.1. APPROACH OF THE SIMPLIFIED BREACH HYDROHGRAPH
The approach of the simplified breach hydrograph consists in first calculating

the breach peak outflow, then determining a simplified hydrograph. Most often used
methods to calculate the breach peak outflow are:
– Estimation of the breach peak outflow using empirical formulae from the
parameters characterising the dam and the reservoir;
330
Q.105 - R.19
– Estimation of the breach peak outflow using a weir formula, a hypothesis on the
breach geometry (generally considered rectangular with these approaches)
and an estimation of the maximum width of the breach with empirical formulae.
3.1.1. Empirical formulae of the breach peak outflow
Available empirical formulae have been designed to rapidly estimate the

breach peak outflow of earthen dams. However, they poorly adapt to canal embank-
ments or fluvial dikes with a permanent water load, mainly because of the differences
between the upstream water volumes and the hydrodynamic characteristics of the
flow.
The first empirical breach formulae used consider at most two parameters: the
reservoir volume and the height of the dam (or rather the height of water upstream of
the dam above the breach invert). They were developed through linear regressions
based on these two parameters and using data bases of real failures. Not considering
the erosion resistance of the material of the dam, most of the formulae neglect an
essential parameter in the erosion process and finally in the estimation of the breach
flow.
Formulae displaying the best predictive capacities are shown in Table 1.
Table 1
Formules empiriques recommandées de débit maximum de
brèche
Empirical formulas for maximal flow discharge estimation
Auteurs - Authors Formule - Formulas
Froehlich (1995) QP = 0, 607(Vr )0,295 (HW )1,24
CLF (2006) QP = a(Hd )2 + cHd (Vr )0,5
Xu et Zhang simple (2009) QP = 0, 133(g(V )5/3 )0,5 (V 1/3 /Hw )−1,276 eFM+ER
The two most recent formulae take into account the dam material resistance to
erosion empirically. QP is the breach peak outflow, in m3 /s. HW is the height of water
above the bottom of the breach, in m. VR is the reservoir’s volume before failure, in
m3 . V is the water volume above the breach bottom, in m3 .
Froelich’s formula (1995) [9], with two parameters (reservoir’s volume and
height of water above breach) was established on around twenty cases. It gives the
best predictive results among similar two-parameter formulae.
CLF formula (Courivaud, Lempérière et Fry, 2006) [10], with four parameters,
introduced at the 22nd ICOLD congress, was a first calibration attempt based on
331
Q.105 - R.19
15 failures. In the report [10], it is written that further efforts are necessary to bet-
ter assess the values of the parameters, using a more abundant data base. The
parameter a, that may be adjusted, is taken equal to 1. The parameter c, empirically
representing the kinetics of the embankment erosion, depends on the dam type:
c = 0.15 for poorly compacted earthen dams; c = 0.1 for concrete-faced dams and
c = 0.01 for zoned, well-compacted and with a thick clay core dams.
Xu and Zhang’s formula (2009) [11], with four parameters was established with
41 failure cases, including 23 through overtopping and the rest through internal ero-
sion. FM is a parameter depending on the failure mode. FM = −0.788 for overtopping
et FM = −1.232 for internal erosion. ER is a parameter representing the erodibil-
ity. ER = −0.089 for highly erodible dams; ER = −0.498 for medium erodibility and
ER = −1.433 for low erodibility.
New data on dam failures, shown in Table 2, are the opportunity to improve the
calibration of CLF’s formula parameters, in the absence of validation of validation of
the 2006 first formula. These data come from 56 total breach case studies, including
overtopping or internal erosion.
Table 2
Base de données de 56 cas de rupture par érosion interne ou surverse
Data base of 56 dam failed by internal erosion or overtopping
Nom du barrage QP Nom du barrage QP Nom du barrage QP
Dam name (m3 /s) Dam name (m3 /s) Dam name (m3 /s)
Apishapa 6 900 Glashuette 117 Noppikoski 1500
Baldwin Hills 1 130 Goose Creek 565 Norwegian test 1B-03 240
Banqiao 78 000 Gouhou 2050 Norwegian Test 2C-02 74
Bayi 5000 Hatch Town 3 100 Opuha 1900
Belci 3 200 Hatfield 3 400 Oros 40 000
Big Bay 4 200 Hell Hole 7 000 Qielinggou 2000
Break Neck 2 37 Horse Creek 3 890 Quail Creek dike 3 100
Briseis 220 Huqitang 50 Rito Manzanares 180
Castlewood 2 3 600 Johnstown 10 000 Salles de Oliveira 7 200
Chenying 1200 Ka Loko 770 Shimantan 30 000
Danghe 2500 Kodaganar 1280 Staw Starzycki 118
Davis reservoir 510 Laurel Run 1160 Swift 25 000
Dehli 1 950 Lawn Lake 510 Taum Sauk 7 700
Dongchuankou 21000 Lijaju 2950 Teton 65 000
El Guapo 64 000 Liujiaitai 28000 Union Bay 1 420
Euclides da Cunha 1 020 Machhu 2 87 000 Whitewater Brook upper 70
Frankfurt 79 Mahe 4950 Zuguou 11200
French Landing 929 Mammoth 2 520 Zuocun 23600
Frias 400 Martin Cooling pond 3 120
The analysis of this database shows primarily the preponderant weight of the
reservoir volume in the breach peak outflow. A simple relationship and a significant
correlation coefficient exist between the breach peak outflow QP (m3 /s) and the
332
Q.105 - R.19
Fig. 4
Courbes de fréquence cumulée du coefficient c (CLF 2020).
Cumulated frequency curves of coefficient c (CLF 2020).
Erosion/Erosion Surverse /Overtopping interne / Internal
Coefficient c A C
Fréquence cumulée/Cumulated frequency B D
L Low Erodability Erodabilité faible
M Medium erodability Erodabilité moyenne
H High erodability Erodabilité forte
reservoir volume VR (m3 ):
QP = 0, 0561(Vr )0,6716 avecR2 = 0, 73 (2)
The ratio Qp /(V0,5

r Hd ) allows a primary evaluation of the range of variation of
the parameter c. This range is shrinked by classifying data by failure modes and
erodibility classes (Fig. 4) and by considering the height of water just before failure.
Figure 1 highlights the uncertainty caused by the choice of the parameter

c. This uncertainty is strong as it cumulates error on observed values, especially
on the outflow (that may vary from 1 to 3 depending on the observer), the lack
of information on materials (ignorance of the plasticity index, absence of erosion
tests) and the simplicity of the dams’ classes. The choice of the parameter a does
not clearly reduce that uncertainty. Knowing the importance of this uncertainty, a
simpler expression of the CLF formula is proposed, with a = 0 and a parameter c
whose values are detailed in Table 2:
QP = cHd (Vr )0,5 CLF(2020) (3)
With the values of the CLF(2020) relationship, the squared correlation coefficient
reaches R2 = 0.858 with the 56 data case studies. Correlation is better for internal
333
Q.105 - R.19
Table 2
Valeurs du coefficient c de la formule CLF(2020) avec a=0
Values of the coefficient c used in CLF(2020) formula (a=0)
Classe d’érodabilité Type de barrage Surverse Erosion interne
Class of erodability Type of dam Overtopping Internal erosion
L Zoné, noyau en argile bien compacté c=0,035 c=0,015
M CFRD, zoné noyau en silt c = 0,06 C= 0,04
H Homogène peu ou pas compacté c=0,1 c=0,08
Table 3
Comparaison des deux plus grands débits observés avec ceux prédits
Comparison between the two largest observed and estimated peak flows
Barrage QP mesuré (m3 /s) Observed QP Estimé QP Estimé QP Estimé
Dam QP Estimated QP Estimated QP Estimated QP
(m3 /s) Froehlich 1995 Xu et Zhang CLF(2020)
Machhu 2 87 000 7 600 14 100 18 600
Banqiao 78 000 16 700 53 000 76 100
erosion failures: R2 = 0.885. This result is slightly better than that from Froelich’s
formula and equivalent to that of Xu and Zhang formula.
Finally, CLF(2020) formula provides the best prediction of the two largest
known breach outflows: Macchu 2 (India), QP = 87 000 m3 /s and Banqiao (Chine),
QP = 78 000 m3 /s (Table 3).
The important difference between predicted and observed outflows with Mac-
chu 2 motivates to integrate the length of the dam, to take into account the occurrence
of several breaches. Knowing the uncertainties of these formulae, it is prudent, for
each study, to use all three of them. Then, the engineer will have to determine the
range of the breach peak outflow, in regards with the site and the consequences of
the failure.
3.1.2. Estimation of the breach peak outflow with a weir formula
This approach is used by EDF more specifically for canal embankments or

fluvial dikes with a permanent water load, for which the empirical breach peak outflow
formulas shown above are poorly adapted.
The breach peak outflow can be estimated from a maximum breach width and
the use of a weir formula. In this approach, uncertainties mainly come from the
maximum width of the breach’s estimation and remain high.
Two approaches may be considered depending on the studied structure:
334
Q.105 - R.19
3. The earthen part subject to overtopping is between non-erodible borders

(concrete structures) relatively close (from 10 to a few hundreds of meters). In that
case, maximum breach width is limited, primarily, by the distance between the two
rigid structures (this estimation is generally very conservative, and its relevance
should be discussed for each site).
4. The maximum breach width can be estimated by empirical formulae.
When the first approach is not applicable (most frequent case), only the sec-
ond method is used, while considering an uncertainty estimated by the engineer in
regards with the context of the structure.
Empirical formulae proposed in the literature to estimate the maximum breach

width are shown in Table 4. B stands for the maximum breach width. H stands for the
height between the upstream water level and the bottom of the breach. V stands for
the reservoir volume, in m3 . Froelich’s formula, that uses the upstream reservoir’s
volume, should only be applied to canal embankments or rivers split into separated
reaches. FERC and USBR’s formulas, that only consider the parameter H, are to
be used cautiously in case of dikes. In addition to these formulas, an engineer may
consider that the maximum breach width is of the same order of magnitude of the
minor bed.
Table 4
Formules empiriques de largeur de brèche
Empirical formulas for breach width
Formule de la FERC Valeurs les plus probables : 2 × H ≤ B ≤ 4 × H
Valeurs extrêmes : H ≤ B ≤ 5 × H
Formule de l’USBR B = 3×H
Formule de Froehlich B = 0,1803 × K × V0,32 × H0,19 avec K = 1,4
pour les ruptures par surverse
The weir formula that is recommended and used in WinDAM code [12] is:
Qmax = g 1/2 L(2H /3)3/2 (4)
Qmax is breach peak outflow, in m3 /s; L is the breach width, in m; H is the

difference between the upstream and downstream water levels, in m.
3.1.3. Simplified breach hydrograph
For earthen embankments, two approaches can be used:
3. Estimation of the breach peak outflow from an empirical formula, then

construction of a simplified breach hydrograph with a triangular shape.
335
Q.105 - R.19
4. Estimation of the breach peak outflow from a weir formula and an estima-
tion of the maximum breach width, then construction of a simplified breach
hydrograph using a mean breach widening celerity.
From the upper and lower bounds of the maximum breach peak outflow, a
triangular breach hydrograph can be built (Fig. 2) from the following three points:
– (T0 , QP = 0): origin of the hydrograph;

– (T1 , Q = Qmax ): peak outflow. T1 is estimated equal to T0 + d, with d the duration
of the rise to the peak. The value of d is estimated by an expert judgement,
in a range generally going from 30 minutes to a few hours, depending on the
available information about the structure’s height, its cross section and the
knowledge of the resistance to erosion of the materials.
– (T2 , Q = 0): final point of the hydrograph (flow is zero). T2 is estimated so that
the integral of the hydrograph is equal to the reservoir’s volume.
Fig. 5
Construction d’un hydrogramme de brèche simplifié.
Simplified hydrograph for breach.
When the breach peak outflow has been evaluated from an estimation of the
maximum breach width, the simplified hydrograph is determined considering orders
of magnitude of the breach widening celerity. These orders of magnitude of the
breach widening celerity come from the analysis of available data about failures of
dams or fluvial dikes with permanent water loading (such analysis has yet to lead
to an article) and are estimated in regards to the structure’s resistance to erosion,
according to four categories:
– Old dams, made of highly erodible materials (no mechanical compaction): 10

to 15 m/min.
– Modern dams (mechanical compaction) but made of erodible to highly erodible
materials: 0.5 to 3 m/min.
– Modern dams made of a homogenous fill or a medium erosion-resistant core:
0.1 to 0.5 m/min.
– Modern dams made of highly erosion-resistant materials: 0.01 to 0.1 m/min.
336
Q.105 - R.19
3.2. BREACH HYDROGRAPH CALCULATION BY FLOW AND EROSION COUPLED

MODELLING APPROACH
Currently, for earthen dams, deterministic modelling based on the physics

of initiating mechanism and breach development inside the dam remain unvali-
dated and so unavailable for operational engineering studies, except for overtopping
erosion of an homogeneous embankment with a simple shape-geometry made
of cohesive materials. For this mechanism, since the late 90’s, great research
effort have been put into describing the physical phenomena and their large-scale
experimental and numerical modelling, especially at the USDA-ARS’s HERU lab,
in Stillwater, OK, USA. In the context of an international project, led by the CEATI
[14], and to which EDF contributed, this research effort succeeded in validating
two numerical codes, that are adapted to engineering’s needs: WinDAM [12] and
EMBREA [13].
For other erosion mechanisms (overtopping in little or non-cohesive materials

and internal erosion), no numerical code adapted to engineering’s needs, has been
subject to a similar evaluation nor validation.
A deterministic calculation, with a physically-based modelling and internation-

ally validated is recommended, with the use of WinDAM or EMBREA software, within
the following hypothesis:
– The structure is a reservoir dam or a lateral dam of a canal or a river split into
separated reaches;
– The dam body can be considered made of fine, cohesive, homogeneous
materials;
– Upstream and downstream faces have a constant slope;
– The hydraulic loading condition can be represented by a stage-volume curve,
an initial condition of the upstream water level and a inflow hydrograph defined
from the initial time of modelling onwards;
– The embankment soil’s following properties are known: wet density, undrained
cohesion (Cu) and the two resistance to erosion parameters determined
by the Jet Erosion Tests: KD (representing the kinetics of the erosion pro-
cess) and TauC (representing the critical shear stress above which erosion
initiates).
These two calculation codes use a 3D conceptual physical modelling of the “

Head Cut Migration ” erosion process. They are based on a 2D vertical representa-
tion of the embankment geometry, on the geotechnical properties of the materials of
the embankment and on the upstream and downstream hydraulic conditions. They
also take into account the resistance to overtopping due to the presence of grass
or rip-rap on the crest and downstream face of the dam. They provide the outflow
hydrograph (whether there is formation of a breach or not) as well as the geometry
of the eroded zone (width, depth).
337
Q.105 - R.19
4. DOWSTREAM PROPAGATION OF THE BREACH HYDROGRAPH
4.1. CHOOSING THE METHODS
Different tools exist for the modelling of a flood wave. The simplest is to
use empirical or analytical formulae or abacuses. The most common is 1D or
2D numerical modelling, solving the Saint-Venant’s equations (more rarely Navier-
Stokes’ 3D equations). The most complex being Smoothed Particle Hydrodynamics
(SPH) numerical modelling or small-scale experimental modelling. Two main fac-
tors account for the choice of the method. Firstly, time available; indeed, in crisis
times with imminent failure risk without existing information, as it can be the case
for natural dams, quick analytical methods will naturally be favoured. If sufficient
time to conduct a thorough study is available, precision and quality of input data will
guide the choice. Numerical modelling, that proved to be efficient for such phenom-
ena is usually favoured. Malpasset dam failure, very well documented, is among
the test cases of 1D or 2D modelling software such as MASCARET or TELEMAC
[15]. Dam failure flood analysis often involve spatial extension of tens or even
hundreds of km of a rivers, ruling out the use of 3D modelling, SPH or physical
models.
4.2. NUMERICAL MODELLING
Nowadays, numerical modelling is the reference method for dam failure flood
analysis. In most cases, 1D modelling is sufficient. Indeed, many dams are situated
in deep valleys where flows follow the bottom of the valley, without recirculation or
other complex flow. However, 2D modelling may prove necessary for wide flood
plains in case of major overflows of rivers. Historically, the first dam failure flood
analysis were conducted for large dams (PIP) leading to flows much greater than
those of extreme floods. For these dams, no available data allowed for a numerical
calibration; and the initial flow in the river did not influence the results. Modelling was
performed on dry bottoms, with Strickler coefficients expertly chosen. Uncertainties
related to this absence of calibration incite to consider the results very carefully:
CTPBOH demands for PIP that the water levels must be increased by 15 % and
the arrival times of the flood wave reduced by 13 %, so as to take security mar-
gins useful to manage the evacuation of the population in case of an accident.
For smaller dams, especially gate-structure dams, failure leads to outflow similar
to that of a major flood of the river. In that case, calibration data is available, and
it is preferable to calibrate the model rather than artificially increase water levels.
Indeed, increasing water levels by 15 %, as done for the PIP, for a flood wave at
the limit of overflowing the minor bed or with a slight overflow can lead to greatly
overestimate the flooded area, especially in flat areas. In case of small dams, initial
conditions can also have a major influence over the consequences of a failure. In
338
Q.105 - R.19
those cases, danger studies can provide information on most likely failure scenario to
consider.
4.3. MODELLING DEDICATED TO INCREMENTAL OR DIFFERENTIAL DAMAGES
A working group of CFBR has produced temporary recommendation in

October 2017 [8] to supervise the application of this method.
The process consists in realising flood and failure simulations in the current
state of the structure and supposing a recalibration of the spillways, in determining
the damages to people and goods in the different configurations and to evaluate the
possible optimisation in light of the proportionality of costs and costs-benefits analy-
sis, the cost consisting mainly in the compliance rehabilitation work and the benefit
(or the safety gain) in the avoided damages (differential of damages integrated on
the probability scale). So, it does not correspond to a differential with and without
the dam or with and without failure of the dam but indeed to a differential with and
without spillways’ recalibration.
Dam failures are considered during a flood and can be partial (depending on
the fragility of the different parts, such as the reloading of rising gates for instance) or
total (in the absence of justification), instantaneous or progressive. Several hypothe-
sis need to be determined: danger level above which stability cannot be guaranteed,
influence of the downstream flooding on the limitation of the failure flows, influence
of the downstream tributaries’ flows, calibration of rugosity coefficients on the basis
of the past significant flood leashes and of the afforestation of the river banks, influ-
ence of the downstream dam and bridge failures, functions of damages to goods
versus water levels and velocities, empirical formula for the evaluation of potentially
affected people, frequentation near the water course during a flood (public attraction
for such a spectacular flow), evacuation conditions of major buildings, influence of
the uncertainties about topography, …Besides, it is necessary to regularly update
the analysis to take into account uncontrollable evolutions of the stakes (evolution
of local urbanisation plan, …).
4.4. CTGREF’s SIMPLIFIED METHOD
This historical method was developed in 1978 [16]. Based on abacuses, it

allows to evaluate in a few hours the risk in case of a dam failure. Once the failure
flow (Qb ) has been calculated from an empirical formula or a modelling, for a river
with a slope J and constant friction coefficient K, the decrease of the flow (Qmax /Qb )
is a function of the dimensionless distance (X/V1/3 ) and of a factor defined as the
product of the slope by the friction coefficient squared, where X is the distance to
the dam and V the volume of the reservoir (Fig. 6).
339
Q.105 - R.19
Fig. 6
Abaque de décroissance du débit en fonction de la distance au barrage.
Flow discharge decrease depending on dam distance.
Once the flow in a zone of interest (inhabited zone, city, village, presence
of a downstream dam) is known, the local water levels can be approximated by
a normalised water level to estimate the flooded area. This method shows great
uncertainties, and great security margins need to be taken, but allows for a swift
reaction in case of a crisis.
4.5. HAYAMI’S SIMPLIFIED METHOD
A computer model for the 1D propagation of river flows has been developed by
the Laboratoire National d’Hydraulique et Environnement (LNHE) of EDF in Chatou.
The goal of this model is to provide robust calculations, easily configurable (without
bathymetry), allowing to model the flow transfer between two distant points of a river
(without particular interest for intermediate flow or level values).
Hayami’s classical model with variable coefficients is a model of propagation-

diffusion of the flows with celerity and diffusion coefficient function of the flow. It
provides very satisfying results during floods. However, it is not conservative in
340
Q.105 - R.19
Fig. 7
Exemple de simulation Hayami.
Example of Hayami simulation.
terms of water volumes for very high flow gradients (very steep slopes such as
steps), which is the case for instantaneous dam failures.
For the propagation on steep slopes (instantaneous failures), a new model,

propagating a homogeneous value to a wetted section (flow divided by a celerity),
has been developed. This conservative model allows for the reproduction (for simple
geometries where a comparison makes sense) of the results given by a complete
resolution of Saint-Venant’s equations.
This solution, in between the CTGREF’s method and numerical modelling,

turns out to be a good alternative when a great number of scenarios are considered
in a Danger Study, leading to many flood wave calculations, or to determine which
dam constitutes the main risk on a nuclear site and thus needs to be thoroughly
studied.
5. UNCERTAINTY ANALYSIS AND RECOMMENDATIONS
The engineer must bear in mind that empirical formulae for breach peak
outflows provide orders of magnitude at best. Their use does not guarantee conser-
vative breach peak outflows. The difference between observed and predicted values
can reach a ratio of 2 to 10. It seems to be preferable to use all of the above formu-
lae, that currently provide the best prediction results, and to never give an estimation
that would come from a single formula. After using several formulae, the engineer
341
Q.105 - R.19
must on his own, judge the range to consider as values for the breach peak outflow.
The choice of the simplified hydrograph must take this uncertainty into account by
providing at least two hydrographs, using the lower and upper values of the breach
peak outflow.
For downstream propagation of the flow, it is important to consider safety

margins in two cases: in case of a simplified method (CTRGEF or Hayami) and in
the absence of calibration data (very large dams). Indeed, only calibrated numerical
modelling of a medium dam causing flows comparable to major flows of the river
and the incremental method do not display major uncertainties.
6. CONCLUSION
The topic of dam breach flood analysis has often been dealt with [9], [17], but
remains topical regarding the few large dam failures still occurring worldwide every
year. France has learnt from failures on its territory (Malpasset) and elsewhere in the
world and managed to create a regulatory framework allowing for the management
of such a risk. Nevertheless, even with an abundant literature, this extreme phe-
nomenon remains uncertain in some aspects, especially for the flows of an earthen
dam’s failure, which represent the majority of failures worldwide. Finally, experience
and expertise of engineers in this area remain an important requirement to provide
a response adapted to every situation.
REFERENCES
[1] Vogel A., Courivaud J.R., Jarecka K., Notable Dam Failures in the Last
Decade, Dam World 2020, Lisbon, September 21-25, 2020.
[2] Zhang L., Pend M., Chang D., Xu Y., Dam Failure Mechanisms and Risk
Assessment, Wiley, 499p., 2016.
[3] Décret n˚ 68-450 du 16/05/1968.
[4] Loi 87-565 du 22/07/87, Décret n˚ 92-997 du 15/09/1992, Arrêté du

01/12/1994, Arrêté du 02/02/2002 et Décret n˚ 2005-1158 du 13/09/2005.
[5] Décret n◦ 2007-1735 du 11/12/2007, Arrêté du 12/06/2008, Décret n◦ 2015-

526 du 12/05/2015 et Arrêté du 03/09/2018.
[6] Arrêté du 06/08/2018.
[7] Règle Fondamentale de Sûreté 1.2.e – 1/04/1984 et Guide ASN n◦ 13, 2013.
342
Q.105 - R.19
[8] Dimensionnement des évacuateurs de crue de barrage par les dommages

incrémentaux ou différentiels : Recommandations pour la mise en œuvre
d’une méthode applicable aux barrages en France, CFBR, 2017.
[9] CIGB, Etude de l’onde de rupture d’un barrage, Synthèse et recommanda-

tions, Bulletin n◦ 111, 1998.
[10] Courivaud J.-R., Lempérierre F., Fry J.-J., A New Analysis of Embank-
ment Dam Failures by Overtopping. 22ème Congrès International de la CIGB,
Barcelone, 16-23 juin 2006, Q. 86 – R. 66.
[11] Xu Y., Zhang L., Breaching parameters for earth and rockfill dams, J. of
Geotechnical and Geoenvironmental Engineering, Vol. 135, N◦ . 12, 2009.
[12] WinDAM C Software | NRCS, https://www.nrcs.usda.gov.
[13] EMBREA, http://www.dambreach.org.
[14] CEATI Dam Safety Interest Group, Evaluation of Numerical Models for Sim-
ulating Embankment Dam Erosion and Breach Processes. DSO-2017-02,
USBR, 2017.
[15] TELEMAC – MASCARET opensource software, http://www.opentelemac.org/.
[16] Calcul d’une onde de submersion due à la rupture d’un barrage, CTGREF,
ministère de l’Agriculture, 1978.
[17] Bernchmark Workshop on numerical analysis of dams – ICOLD – 02-

04/10/2013 – Theme C – Estimation for Risk Assessment: Numerical
Modelling, Uncertainty Quantification, and Communication of Results.
343
Q.105 - R.20
-------
-------
ÉVALUATION DE L’ONDE DE SUBMERSION EN CAS DE RUPTURE DES

OUVRAGES EN REMBLAI
PARTIE II : R&D CONCERNANT LA RUPTURE PAR SURVERSE OU ÉROSION
INTERNE DES BARRAGES EN REMBLAI ET LEVÉES(∗)
Jean Robert COURIVAUD, Laurent DEL GATTO

Ingénieurs, EDF
André PAQUIER, Guillaume VEYLON, Pierre PHILIPPE, Anthony MOUYEAUX,

Sylvie NICAISE, Catherine FOUCHIER, Claudio CARVAJAL & Laurent PEYRAS
Chercheurs, INRAE
Rémi BEGUIN
Ingénieur, GEOPHYCONSULT
Laurence DUCHESNE
Directrice adjointe Ingénierie et Grands Projets, CNR
Christophe PICAULT
Ingénieur, CNR
Jean-Jacques FRY
Expert, J-J FRY CONSULTING
FRANCE
∗ Evaluation of embankment hydraulic structures’ failure flood wave. Part 2: Research and
Development related to embankment dam and levee failure by overflowing/overtopping or

internal erosion
344 DOI: 10.1201/9781003211808-66

Q.105 - R.20
RÉSUMÉ
L’évaluation de l’hydrogramme de rupture des ouvrages hydrauliques en rem-

blai, barrages ou digues, est un enjeu essentiel pour les maîtres d’ouvrage, bureaux
d’étude et autorités de sûreté. Bien que l’érosion par surverse constitue l’une des
principales causes de rupture des barrages en remblai et des levées, tous types
d’ouvrages confondus, de nombreux processus physiques restent non décrits et
non modélisés. Ce rapport présente brièvement l’état de l’art dans ce domaine, en
pointant les limites d’utilisation des outils validés à la disposition de l’ingénieur et en
présentant les sujets où description physique et modélisation validée sont encore
absentes. L’étude de l’influence de la prise en compte du transport sédimentaire à
l’aval de la brèche sur la modélisation de l’hydrogramme de brèche lui-même et sur
le calcul d’inondation, basée sur trois cas d’ouvrages réels ou d’expérimentation à
grande échelle, montre que cette influence est faible pour des pentes du terrain aval
fortes, mais qu’elle peut être significative lorsque cette pente est faible. L’analyse
de quatre petits barrages du Sud-Ouest de la France qui ont subi des écoulements
de surverse sans rompre, menée en caractérisant les résistances à l’érosion des
matériaux constitutifs de ces ouvrages ainsi que les conditions de surverse, met en
évidence l’influence importante de la végétation recouvrant les parements aval, de
la variabilité des caractéristiques des matériaux, ainsi que de la saturation super-
ficielle du parement aval. Enfin, l’intérêt des essais expérimentaux est illustré par
la représentation du développement d’une érosion de contact jusqu’à la rupture
par une expérimentation à grande échelle menée en laboratoire qui a permis de
reproduire des enchaînements de mécanismes complexes et de quantifier leurs
cinétiques, grâce notamment à l’instrumentation déployée.
SUMMARY
Breach outflow evaluation of embankment dams and levees is an essential task

for dam owners and consulting engineers. Although overtopping erosion is one of the
two main causes of embankment dam and levee failures, including all types of dams
or levees, many physical processes remain not described and not modelled numer-
ically. This report has briefly presented the state-of-the-art in this topic pointing out
the limits of use of the validated tools available to engineers and presenting issues
where a physical description and validated numerical modelling remain absent. The
study of the influence of sediment transport downstream from the breach on its out-
flow and on the wave propagation, based on three case studies of real dams or large
field test shows that this influence is low in case of steep downstream slopes, but it
can be significant when low slopes. The analyses of four small embankment dams
from South-West of France which overtopped without breaching, conducted by char-
acterizing both the erosion resistance of materials constituting these embankments
and the overtopping conditions, shows the great influence of vegetation covering the
downstream face, of the variability of material characteristics and of the saturation
345
Q.105 - R.20
of the superficial part of the downstream fills. Finally, the interest of experimental
tests is illustrated by the large scale modelling of contact erosion up to failure, which
allowed to represent complex chained mechanisms and to quantify their kinetics
thanks especially to the monitoring used.
1. INTRODUCTION
Après avoir présenté les pratiques françaises d’ingénierie pour l’évaluation de

l’hydrogramme de brèche et le calcul de la propagation de l’onde de rupture des
barrages en remblai, la seconde partie de ce rapport présente plusieurs actions
de recherche et développement menées récemment en France pour contribuer à
une meilleure caractérisation de l’érosion par surverse ou par érosion interne des
barrages en remblai et des levées.
Les outils d’ingénierie permettant de représenter les processus de rupture

d’un barrage en remblai ou d’une levée par érosion par surverse ou par éro-
sion interne sont encore actuellement très limités, certains mécanismes physiques
restant encore peu ou pas décrits et non modélisés. En particulier, la rupture d’un
ouvrage en remblai entraîne l’érosion d’un volume de matériaux relativement impor-
tant en provenance du corps de l’ouvrage. Ces matériaux peuvent s’accumuler à
l’aval proche du remblai et, en modifiant la ligne d’eau, avoir une influence sur le
développement de la brèche. Si ces matériaux sont emportés loin de l’ouvrage, leur
dépôt peut néanmoins entraîner une surélévation du niveau d’eau atteint par l’onde
de rupture et donc une augmentation des inondations mais, à l’opposé, l’érosion
du fond générée par les fortes vitesses de l’écoulement peut entraîner un abaisse-
ment général de la ligne d’eau à la pointe de débit et donc réduire globalement
l’étendue de l’inondation [1]. D’autre part, les conditions de hauteur et de durée
de surverse conduisant à la rupture, en fonction des caractéristiques de l’ouvrage,
constituent également des questions encore ouvertes auxquelles les ingénieurs de
bureaux d’étude comme les maîtres d’ouvrage ont besoin de pouvoir apporter des
réponses.
Après avoir présenté l’état de l’art de la modélisation des ruptures d’ouvrages

hydrauliques en remblai par surverse (barrages, remblais de canaux et levées),
incluant l’impact de la prise en compte du transport sédimentaire à l’aval de la
brèche, ce rapport présente un retour d’expérience sur les conditions de hauteur
et de durée de surverse conduisant ou non à la rupture, basé sur des données de
petits barrages en remblai du Sud-Ouest de la France ayant subi de telles condi-
tions récemment. Puis sont abordées plusieurs questions scientifiques relatives à
l’érosion par surverse des ouvrages en remblai qui restent non résolues, nécessi-
tant le lancement de programmes de recherche futurs pour apporter à l’ingénierie
une meilleure compréhension de certains processus et des outils de prédiction plus
performants. Enfin, ce rapport illustre l’apport des essais expérimentaux dans une
telle démarche de recherche en présentant les essais menés à grande échelle pour
évaluer les cinétiques de rupture par érosion interne.
346
Q.105 - R.20
2. ETAT DE L’ART DE LA MODELISATION DES BRÈCHES PAR SURVERSE
2.1. DESCRIPTION GÉNÉRALE DES PROCESSUS PHYSIQUES DE SURVERSE
Quatre mécanismes d’érosion par surverse pouvant conduire à la forma-

tion d’une brèche sont aujourd’hui identifiés pour les ouvrages hydrauliques en
remblai [2] :
– Remblais constitués de sols fins cohésifs. Le mécanisme d’érosion est celui

du “ Head-Cut ”, ou formation progressive, à partir le plus souvent du pied
aval, de marches avec alternance d’érosion par jet impactant et détachement
par blocs.
– Remblais constitués de sol granulaire à granulométrie fine (sableux). Le
mécanisme est appelé “ érosion de surface ”, l’érosion est répartie sur
l’ensemble de l’ouvrage (crête et parement aval).
– Remblais constitués de sols majoritairement granulaires, à granulométrie
souvent étalée, incluant des gros blocs jusqu’aux matériaux fins. L’érosion
par surverse pour ce type d’ouvrage n’est pas encore décrite ni connue à ce
jour et fait l’objet de recherches en cours.
– Barrages en enrochements. L’érosion par surverse de ce type d’ouvrage est
caractérisée par l’importance du couplage entre écoulement de surface et
écoulement interne ainsi que par le couplage entre différents mécanismes :
glissements et érosion externe par surverse.
Les mécanismes de “ Head Cut ” et d’érosion de surface sont illustrés sur la

Figure 1 ci-dessous.
Fig. 1
Illustration des mécanismes d’érosion par surverse par “ Head Cut ” (à gauche) et
par érosion de surface (à droite).
Drawing of Head-Cut-type and surface erosion-type overflowing erosion
mechanisms.
347
Q.105 - R.20
Parmi ces quatre mécanismes d’érosion par surverse, seul le mécanisme de

“ Head Cut ” est représenté par des modèles numériques conceptuels orientés vers
l’ingénierie (modèles utilisant des lois physiques avec des approches simplifiées)
qui ont fait l’objet d’une validation à partir d’essais expérimentaux à grande échelle
et de données de cas réels de rupture de barrage [3]. Ces modèles numériques
permettent de réaliser un calcul déterministe de l’érosion par surverse et du débit
s’écoulant à travers celle-ci dans les conditions suivantes :
– L’ouvrage est un barrage réservoir muni ou non d’un évacuateur de crue ;

– L’ouvrage peut être considéré comme étant constitué d’un matériau
homogène fin et cohésif ;
– La géométrie des parements amont et aval peut être représentée par un fruit
constant sur chaque face ;
– La condition de chargement hydraulique amont peut être représentée par
une courbe de capacité de la retenue, une condition initiale de cote du plan
d’eau, un hydrogramme de crue entrant défini à partir de l’instant initial de la
modélisation et une courbe de débitance de l’évacuateur de crue ;
– Le sol constitutif du remblai est caractérisé par sa densité humide, sa
cohésion non drainée (cu ) et les deux paramètres de résistance à l’érosion
mesurés par des essais de Jet Erosion Test : K D (qui représente la cinétique
de l’érosion) et τC (contrainte de cisaillement critique au-delà de laquelle
débute l’érosion). Le taux de masse de sol érodé au cours du temps est relié
aux paramètres KD et τC par la relation :
– ε = KD (τ − τC ), avec ε, le taux de masse érodée et τ , la contrainte de
cisaillement appliquée par l’écoulement sur le sol.
L’érosion par surverse des remblais constitués de sols granulaires à gran-

ulométrie fines (sols sableux) a fait essentiellement l’objet, comme travaux de
recherche, de modélisations physiques en laboratoire à petite échelle (hauteurs
de remblais de quelques dizaines de centimètres) et de modélisations numériques
utilisant les lois de capacité de transport de sédiments par charriage en rivière.
L’absence de validation de ce type de modélisation sur des essais expérimentaux
à grande échelle ou des cas réels limite l’application de ces modèles pour des
applications d’ingénierie.
2.2. PRISE EN COMPTE DE L’INFLUENCE DU TRANSPORT SÉDIMENTAIRE À L’AVAL

D’UNE BRÈCHE
Pour prendre en compte l’influence des dépôts et érosions lors de la prop-

agation de l’onde de rupture, des modèles hydrosédimentaires 1D ou 2D ont été
développés. Ces modèles couplent la représentation des écoulements par les équa-
tions de Saint Venant à une ou plusieurs équations de transport des sédiments
complétées par des lois d’érosion-dépôt. Pour la rupture d’un barrage, les modèles
hydrosédimentaires 1D ont montré leur efficacité à estimer l’influence des sédi-
ments transportés sur le niveau d’inondation [1] même si l’incertitude des résultats
est très grande compte tenu de la difficulté à paramétrer un tel modèle en l’absence
348
Q.105 - R.20
d’observations d’événements représentatifs. Lorsque les dépôts et érosions ont lieu

à proximité de la brèche, on s’attend à une influence non seulement sur le niveau
d’inondation mais aussi sur l’hydrogramme au droit de la brèche et donc potentielle-
ment sur la totalité de l’inondation en aval de l’ouvrage. Pour simuler ce phénomène,
il est nécessaire de mettre en œuvre un modèle hydrosédimentaire qui permette de
coupler le processus de formation de la brèche avec le calcul des lignes d’eau en
amont et en aval de la brèche.
Pour l’étude des trois cas présentés ci-après, afin d’avoir une représentation
détaillée des dépôts en aval de la brèche, une modélisation 2D a été choisie pour
l’aval de la brèche. A l’amont, la modélisation 2D a aussi été choisie dans le cas
d’une retenue alors que la modélisation 1D a été préférée pour un lit endigué. Dans
ce dernier cas, on utilise donc un modèle couplé 1D/2D avec un couplage par la
brèche mais aussi par les déversements au-dessus des digues du lit endigué [4].
Les trois cas présentés correspondent à trois situations différentes :
1. La rupture d’un petit barrage en travers d’une rivière : l’événement corre-

spond à une expérience effectuée sur le terrain en Norvège pendant le projet de
recherche européen Impact [5]. Le barrage en terre d’une hauteur de 6 m envi-
ron surverse à cause d’un lâcher d’eau contrôlé en provenance d’un grand barrage
situé 600 mètres à l’amont. L’hydrogramme au droit du barrage présenté en figure 2
montre trois phases : un faible déversement sans érosion notable du barrage, une
faible augmentation du niveau d’eau qui conduit à l’ouverture brutale de la brèche
presque jusqu’à sa largeur maximale de l’ordre de 30 mètres en moins de deux
heures puis le maintien d’un débit légèrement supérieur au débit entrant dans la
retenue pendant la phase de vidange et la fin de l’érosion du remblai.
Trois simulations (le modèle décrit dans [6] utilise le logiciel 2D Rubar 20TS)
sont effectuées : l’une ne tient pas compte du transport de sédiments, la seconde
transporte uniquement les sédiments en provenance du barrage, la troisième mod-
élise l’érosion du barrage et du fond de la rivière et produit des dépôts et des érosions
éparses conduisant à des variations verticales du fond inférieures à 2 cm, ce qui
est globalement conforme à la réalité où le fond alluvionnaire de la rivière a très
peu été érodé et les dépôts de sédiment étaient très faibles à proximité du bar-
rage. Les différences en débit entre les trois simulations sont inférieures à 10%
(Figure 2) soit de l’ordre des fluctuations de débit et ne peuvent être considérées
comme significatives.
2. La rupture des digues en basse plaine de l’Agly pendant la crue de 1999

(période de retour de l’ordre de 100 ans) où la surverse au-dessus des digues due
à l’encombrement du lit a finalement conduit à l’ouverture d’une brèche (largeur de
30 mètres sur une hauteur de 4 mètres).
Trois simulations (avec le modèle 1D/2D décrit dans [4] qui utilise les codes
de calcul RubarBE et Rubar20TS) sont effectuées : l’une ne tient pas compte du
349
Q.105 - R.20
Fig. 2
Hydrogrammes de brèche du barrage expérimental norvégien.
Modelled breach outflows of Norwegian field test.
transport de sédiments, la seconde transporte uniquement les sédiments en prove-

nance de la brèche, la troisième permet l’érosion du fond de la plaine d’inondation à
l’aval de la brèche. Pour estimer l’effet du seul transport de sédiments à l’aval, dans
toutes les simulations, l’érosion verticale de la digue est limitée au niveau du terrain
naturel à l’aval (contrairement à la réalité qui a vu une fosse d’érosion se former
au droit de la brèche [6]). La seconde simulation donne un dépôt d’une hauteur
maximale de 36 cm à 100 mètres au Nord-Est de la brèche mais ce dépôt n’a que
peu d’influence sur l’hydrogramme au droit de la brèche (courbe confondue sur la
figure 2). En revanche, la troisième simulation conduit à une diminution des débits
corrélée à un moindre élargissement de la brèche vers l’Est (aval de l’Agly); comme
pour la seconde simulation, un dépôt se forme au Nord-Est de la brèche mais il
n’atteint que 20 cm de hauteur par suite de son érosion en lien avec une érosion
généralisée de l’aval de la brèche (fosse d’érosion); cette érosion est plus marquée
à l’Ouest de la brèche et conduit à des écoulements plus centraux par rapport à
l’axe de la brèche, d’où une forme de brèche moins asymétrique.
3. La troisième étude porte sur la simulation de la rupture d’un barrage qui

pourrait être créé par le glissement de terrain du Pas de l’Ours dans les Hautes Alpes
[7]. Les simulations présentées prennent en compte l’hypothèse d’un barrage d’une
hauteur de 40 mètres d’où un volume de matériaux considérable (2 millions de m3
pour un volume d’eau stocké de 8 millions de m3 au maximum) du fait des faibles
pentes des parements du barrage. Compte tenu de l’incertitude sur les caractéris-
tiques du barrage potentiel, deux hypothèses ont été prises pour le diamètre des
sédiments : 60 mm correspondant au diamètre médian des sédiments du glissement
ou 1 mm pour représenter un matériau plus facilement érodable.
350
Q.105 - R.20
Fig. 3
Simulations de l’hydrogramme de brèche de la digue de l’Agly.
Modeled breach outflows of Agly dike.
Fig. 4
Hydrogrammes pour le glissement du Pas de l’Ours.
Pas de l’Ours sliding outflows.
La figure 4 montre que le débit de pointe est plus que doublé quand le
diamètre est réduit, ce qui correspond à une largeur de brèche passant de 31
à 75 mètres. En revanche, les simulations menées avec ou sans transport de
sédiments donnent des résultats très similaires car la simulation du transport de
351
Q.105 - R.20
sédiments donne des dépôts et des érosions de faible amplitude verticale (quelques
centimètres).
Ces trois exemples montrent un faible impact des dépôts de sédiments en

provenance du remblai érodé sur l’hydrogramme au droit d’une brèche. L’exemple
de la digue le long de l’Agly montre toutefois que dans le cas d’une faible pente à
l’aval de la brèche, dépôts mais aussi érosions peuvent conduire à une modification
des écoulements qui influencera le développement de la brèche d’où la nécessité
de simuler les deux. Pour le glissement de terrain, malgré les volumes importants
mis en jeu, la forte pente de la vallée semble empêcher des dépôts à l’aval immédiat
de la brèche. Cette conclusion ne peut être généralisée : le glissement de terrain
de la Josefina en 1993 a généré des épaisseurs de dépôts dépassant 20 mètres à
l’aval immédiat du glissement [8].
3. RETOUR D’EXPÉRIENCE SUR LES CONDITIONS DE HAUTEUR ET DE

DURÉE DE SURVERSE CONDUISANT OU NON À LA RUPTURE
En octobre 2018, la région Occitanie en France a été frappée par un épisode

hydrométéorologique méditerranéen très intense. Au cours de cet évènement, des
cumuls de pluie dépassant 200 mm en 24 heures ont été observés sur une zone
couvrant plusieurs centaines de kilomètres carrés. Plusieurs cas de surverse de
petits barrages en remblai ont été observés. Les caractéristiques géométriques de
ces barrages, identifiés par trois lettres dans la suite du rapport, sont synthétisées
dans le Tableau 1.
Table 1
Caractéristiques géométriques des barrages en remblais et caractéristiques de
l’hydrologie et de la surverse
Geometrical characteristics of earth dams and characterization of hydrology and
overflow
ID Hauteur Volume de Largeur en Pente du Débit de Lame d’eau Durée
(m) la retenue crête talus aval pointe* maximale de estimée de
(m3 ) (m) (V/H) (m3 /s) surverse (m) la surverse (h)
CAP 6 20 000 2 1/1 30 0,4 2h
HOR 7 30 000 2 1/3 85 1,0 3h
FAJ 10 100 000 3 1/1 4 0,1 1h
PEC 8,25 95 000 3 1/2,5 23 0,2 1h
∗ Il s’agit d’un débit calculé en utilisant le modèle développé par Piotte et al. [9]
Désordres observés. Les ouvrages ont subi des surverses d’intensité très
différentes, générant ainsi des désordres de nature variée. Une représentation
schématique des désordres observés est proposée sur la Figure 5.
352
Q.105 - R.20
Fig. 5
Représentation schématique des désordres observés : vue en plan (à droite) et
sections transversales représentatives (à gauche).
Schematic representation of observed damages from above (right) and
representative cross section (left).
Le barrage CAP est un barrage intégralement recouvert d’une végétation

constituée d’arbres, d’arbustes et de plantes de sous-bois. Il a subi une surverse
généralisée sur l’ensemble du remblai dont la lame d’eau maximale a été estimée
à 40 cm au regard des laisses de crues observées dans les branchages des arbres
situés au sommet du talus amont. Les hautes herbes (ronces de 50 cm à 1 m) recou-
vrant le barrage en crête et sur le parement aval sont couchées dans le sens des
écoulements. Une zone érodée, située juste en aval de la crête, est constatée en
partie centrale du parement aval, avec une largeur d’environ 3 m et une profondeur
de l’ordre de 3 m. En dehors de cette zone érodée (qui reste assez localisée), le
parement n’est pas endommagé malgré la surverse généralisée.
Le barrage HOR est recouvert de végétation herbacée de 10 à 20 cm de

hauteur. Son parement amont est colonisé par une végétation dense d’arbres et
353
Q.105 - R.20
Fig. 6
Zone d’érosion sur le parement aval du barrage HOR.
Erosion zone on the downstream slope of HOR dam.
d’arbustes. Il a subi une surverse généralisée de l’ordre de 1 m. En aval, deux

zones d’érosion (12 m de largeur et 5 m de profondeur environ chacune) se sont
développées depuis le pied aval du barrage. Elles ont régressé jusqu’à une distance
de 3 m de la crête (Figure 6). L’existence de deux zones d’érosion distinctes résulte
de la présence d’un merlon en terre d’environ 40 cm de hauteur sur le tiers central
du remblai et qui a séparé les écoulements en deux parties.
Le barrage de FAJ est également recouvert d’herbes et d’arbrisseaux. Deux

zones de surverse localisées ont été observées. Une zone de glissement/érosion
s’étend sur 3 mètres de largeur en partie centrale sur la moitié haute du remblai. Les
laisses de crue autour des deux points bas en crête du barrage permettent d’estimer
une lame d’eau de l’ordre de quelques centimètres. En raison de l’initiation d’une
brèche dans l’appui rive droite du barrage, la surverse est restée limitée et de courte
durée. Le mécanisme d’érosion s’est donc vraisemblablement développé après celui
du glissement provoqué par la saturation superficielle du remblai.
Le barrage de PEC est également très végétalisé. Il est recouvert d’herbacées

et d’arbrisseaux sur ses parements amont et aval. Plusieurs zones érodées (largeur
et profondeur de l’ordre de 50 cm) sont visibles sur le parement aval une dizaine de
mètres en amont du pied du barrage. Une zone érodée s’est développée en aval
de la crête (2,5 m de hauteur pour 4 m de largeur environ). Elle se prolonge par
une ravine d’environ 1 m de profondeur pour 60 cm de largeur qui s’étend jusqu’au
pied aval du remblai. Les indices recueillis sur le site suggèrent une lame d’eau de
surverse de l’ordre de 20 à 30 cm sur environ la moitié de la longueur du barrage.
Le barrage étant vide au début de la crue, la surverse a été de courte durée.
354
Q.105 - R.20
Chacun des mécanismes de dégradation observés aurait pu conduire à la

rupture du barrage. La possibilité d’observer les conséquences des mécanismes
d’érosion avant rupture tient à la brièveté de l’épisode hydrologique qui a interrompu
le scénario de rupture avant l’initiation d’une brèche.
Essais en laboratoire. Des échantillons intacts et remaniés ont été prélevés

sur chaque barrage dans les zones érodées des remblais. Ils ont fait l’objet
d’essais d’identification qui ont permis leur classification et d’essais d’érosion pour
caractériser leur érodabilité.
Table 2
Résultats des essais géotechniques
Geotechnical test results
ID ρd D50 FC w WL Tc kd
(-) (mm) (%) (%) (%) (cm3 ·N−1 ·s−1 )
(Pa) CAP 1,64 0,47 33 17,2 - 78* 30*
HOR - 0,011 81 - 32 - -
FAJ 1,75-1,79 0,010 80 16,7-15,1 31 300-500** 0,1-0,3**
PEC 1,66 0,050 58 15,1 40 265* 31*
∗ Jet erosion test (JET), ∗∗ Hole Erosion Test (HET)
Les matériaux constitutifs des remblais se répartissent en deux familles de

matériaux correspondant à des formations géologiques bien distinctes. Le remblai
du barrage HOR est constitué de séries d’argiles marrons et de marnes au sein
desquelles s’intercalent des assises résistantes de grès et conglomérats et des lits
calcaires micro-détritiques. Le remblai du barrage FAJ est constitué d’argiles beiges
issues de la même formation que le barrage HOR. Il n’est donc pas surprenant
que ses propriétés physico-chimiques soient très proches. Les barrages CAP et
PEC présentent également des propriétés physico-chimiques proches. Il s’agit de
matériaux à matrice argilo-sableuse présentant des éléments grossiers – calcaires
et grès pour PEC, schistes pour CAP.
Des essais d’érosion au JET ou au HET ont été conduits sur des échantil-
lons intacts prélevés des quatre barrages. L’essai JET réalisé sur le matériau de
CAP a estimé une contrainte critique de l’ordre de 80 Pa et un coefficient d’érosion
de Hanson de l’ordre de 30 cm3 ·N−1 ·s−1 . L’essai a également permis d’évaluer la
vitesse d’impact critique entre 4 et 5 m.s−1 . L’essai JET réalisé sur le matériau de
PEC a abouti à une estimation de la contrainte critique de l’ordre de 265 Pa et un
coefficient d’érosion de l’ordre de 31 cm3 ·N−1 ·s−1 . La vitesse d’impact critique a
été évaluée à 8 m·s−1 . Les valeurs du coefficient d’érosion de ces deux matériaux
correspondent à celles d’un matériau très érodable [9]. Deux essais HET ont été
réalisés sur les échantillons prélevés sur le barrage FAJ. Ils ont mis en évidence
des contraintes critiques comprises entre 300 Pa pour l’échantillon prélevé horizon-
talement et 500 Pa pour l’échantillon prélevé verticalement, avec des coefficients
d’érosion faibles (0,1 à 0,3 cm3 ·N−1 ·s−1 ). La vitesse d’écoulement critique corre-
spondante est comprise entre 4 et 7 m·s−1 . Les paramètres estimés permettent
de caractériser la vitesse d’érosion au-delà de la contrainte critique comme étant
modérément rapide [11].
355
Q.105 - R.20
4. QUESTIONS SCIENTIFIQUES SOULEVÉES
4.1. MODÉLISATION DE L’ÉROSION PAR SURVERSE
Alors que l’érosion par surverse des barrages en remblai représente l’un des
deux modes de rupture majoritaires (l’autre étant l’érosion interne) de barrages dans
le monde, tous types de barrages confondus, et des levées, les modèles numériques
validés représentant l’érosion par surverse, qui sont adaptés pour une utilisation
dans l’ingénierie, sont limités aux barrages homogènes en matériaux cohésifs et de
géométrie simple. De nombreux processus physiques n’ont toujours pas été décrits
et nécessiteront un effort de recherche conséquent pour compléter la boîte à outils
de l’ingénieur. il s’agit notamment :
– de la description et de la modélisation de l’érosion par surverse des remblais

constitués de matériaux grossiers et à granulométrie étalée ;
– de la modélisation dissymétrique de l’élargissement de brèche dans les lev-
ées et remblais de canaux, caractérisés par un écoulement à l’amont qui est
parallèle à l’axe de l’ouvrage;
– de la modélisation de l’érosion de la fondation et du développement d’une
fosse d’érosion en aval, ce qui joue un rôle non négligeable, en particulier,
pour les levées;
– de la prise en compte de l’interaction possible entre différents mécanismes
de rupture au cours de la formation de la brèche.
4.2. CONDITIONS DE HAUTEUR ET DE DURÉE DE SURVERSE CONDUISANT À LA

RUPTURE
Rôle de la végétation. Les essais en laboratoire ont montré des propriétés

physico-chimiques et d’érodabilité des matériaux des barrages CAP et PEC très
proches. De même, les désordres observés sont assez similaires. Pourtant, la
géométrie des ouvrages et les sollicitations hydrauliques auxquelles ils ont été
soumis sont très différentes : le barrage PEC a soutenu une surverse de l’ordre de 20
cm alors que le barrage CAP a été sollicité par une lame d’eau deux fois supérieure
avec une pente de talus aval supérieure. De plus, alors que le matériau a été car-
actérisé comme très érodable, les zones d’érosion initiées en crête ne se sont pas
développées au-delà de quelques mètres de profondeur. Notre retour d’expérience
montre que la végétation présente sur ces ouvrages (herbes et ronces hautes et
denses), a joué le rôle de protection vis-à-vis des écoulements. En particulier, il
semble que les hautes herbes couchées par l’action de l’eau aient pu contribuer à
diminuer les turbulences, à laminer les écoulements, et à rendre l’écoulement moins
agressif pour le remblai sous-jacent (Fig. 7). Ces constatations incitent à se poser
la question de l’effet d’une végétation dense sur i) la minoration de la sollicitation
hydraulique par la création d’un tapis végétal sur la crête et le parement aval et ii)
356
Q.105 - R.20
l’influence des racines sur la résistance à l’érosion des talus et au développement

des mécanismes de brèches. Ces questions doivent selon nous être traitées en
intégrant l’existence d’une variabilité des propriétés des matériaux et la réalité de la
non saturation du remblai avant la crue.
Rôle de la non saturation du remblai. D’après le retour d’expérience du

barrage FAJ, qui n’a connu qu’une surverse très modeste (quelques centimètres),
le glissement superficiel de la couche végétalisée du parement aval serait à l’origine
des désordres observés. La perte du couvert végétal comme protection anti-érosion
peut alors conduire au développement rapide de griffes d’érosion puis de ravines qui
s’accentuent et in fine conduire à la création d’une brèche. Donc dans les approches
par scénario, il convient de ne pas oublier le phénomène de saturation progressive
du remblai et son impact sur la stabilité de peau et la résistance à l’érosion.
Fig. 7
Vue de la crête du parement aval du barrage CAP après la crue.
View of the downstream slope crest of the CAP dam after the flood.
Rôle de la variabilité des propriétés du remblai. Toutes nos visites de terrain

montrent que les désordres débutent en un ou plusieurs points isolés des barrages.
Ce processus de localisation des désordres s’explique par la variabilité initiale de la
géométrie de l’ouvrage (points bas ou irrégularités de la ligne de séparation crête-
talus aval par exemple), mais aussi par la variabilité intrinsèque des propriétés
mécaniques du remblai. Ainsi, pour évaluer le risque de développement d’une
brèche par surverse, il convient selon nous de développer des modèles stochas-
tiques tenant compte de la variabilité spatiale des paramètres hydro-mécaniques
et leurs distributions spatiales pour pouvoir détecter les conditions d’initiation et de
développement du scénario de brèche par surverse. La quantification de la vari-
abilité spatiale des paramètres est aujourd’hui encore un problème ouvert dans le
domaine des ouvrages hydrauliques.
357
Q.105 - R.20
5. APPORT DES REMBLAIS D’ESSAI POUR EVALUER LES CINETIQUES

DE RUPTURE PAR EROSION INTERNE
Les mécanismes conduisant de l’initiation d’une érosion interne à l’ouverture

d’une brèche sont multiples (érosion de contact, suffusion, conduit d’érosion, fil-
tration…) et mettent en jeu des phénomènes à différentes échelles spatiales et
temporelles [12], dont l’enchainement, non accessible aux essais de laboratoire,
conditionne la cinétique associée à l’ouverture de la brèche [13].
Pour analyser cet enchainement de phénomènes dans le cas d’une érosion

de contact à l’interface entre le noyau peu cohésif d’un barrage et sa fondation
alluviale pulvérulente, des remblais d’essais ont été construits de 2010 à 2014 par
la CNR et EDF au laboratoire CACOH de la CNR à Lyon [14], [15]. L’objectif était de
décrire mais aussi quantifier ces phénomènes, pour fournir des données d’entrées
pour des modèles phénoménologiques, analytiques et numériques. Ce chapitre
décrit les cinétiques observées dans la phase initiatrice d’une brèche, ainsi que les
moyens instrumentaux qui permettent de les quantifier.
Le dispositif consiste en une structure en béton armé hébergeant une tranche

de digue de 4 m de largeur, 8 m de longueur et de 2,20 m de hauteur (Fig. 8). Pour
ce projet, la digue est fondée sur une couche de 20 cm de grave propre et con-
stituée d’un remblai de limon sableux (cohésion inférieure à quelques kiloPascals)
de 2,00 m d’épaisseur, taluté à 3H /1V pour une partie des essais. Une grille à
l’extrémité aval empêche l’érosion régressive de la fondation. Ces dimensions corre-
spondent à la borne basse des tailles de digues fluviales du Rhône et du Rhin, mais
sont suffisantes pour représenter les principaux composants de l’ouvrage (noyau,
fondation, recharge) avec des matériaux réels, et sont donc aptes à reproduire les
enchainements de mécanismes.
Le mode opératoire utilisé consistait à 1) construire la digue expérimentale,

en multipliant les mesures de contrôle ; 2) appliquer par paliers le chargement
hydraulique, jusqu’à la l’initiation du conduit ou la charge maximale ; 3) démonter le
modèle avec observations et prélèvements.
En première partie de l’essai, une érosion transitoire est observée après

chaque augmentation du chargement hydraulique. Cette érosion se traduit par un
pic de matières en suspension et de dépôts sédimentaires à l’aval de l’ouvrage,
suivi d’une décroissance exponentielle du transport solide (Fig. 9). Aucune anoma-
lie n’est observée sur les autres paramètres mesurés. Le modèle tend rapidement
vers une stabilisation : le transport solide observé est lié à un lessivage de la fon-
dation et de l’interface et il n’y a pas initiation de rupture à l’échelle de temps testée
(paliers de 24 h à 1 semaine suivant les essais) [14]. Au-delà d’une certaine vitesse
de Darcy dans la fondation, une deuxième phase débute avec la formation d’un
conduit au-dessus de l’interface par érosion des limons vers les graviers, qui pro-
gresse vers l’amont et l’aval. Quand le conduit devient traversant, c’est le début de la
358
Q.105 - R.20
Fig. 8
Schéma du dispositif (en haut) ; vue depuis l’aval du dispositif vide (en bas à
gauche) et lors d’un essai (en bas à droite).
Sketch of the experimental set-up (top); view from downstream of the empty device
(bottom left) and during a test (bottom right).
phase finale : la formation d’une brèche (Fig. 8). Cette phase est observée quelques
minutes à quelques heures après l’atteinte du palier de chargement provoquant la
rupture.
Les débits sont stables ou légèrement fluctuants durant la progression vers

l’aval et l’élargissement du conduit, puis augmentent exponentiellement lorsque le
conduit devient traversant, atteignant en quelques minutes les capacités maximales
du dispositif. Les taux de matière en suspension et les dépôts à l’aval augmentent
brusquement (Fig. 9). Les pressions interstitielles impactées par la progression du
conduit montrent un aplatissement de la ligne piézométrique. Les mesures de défor-
mation par les fibres optiques implantées juste au-dessus de l’interface permettent
la localisation et le suivi de la progression du conduit (zones de déformations fortes
et évolutives). La vitesse de progression des conduits ainsi identifiés est de l’ordre
de 2 cm/min pour l’essai 6 (Fig. 10).
Le conduit devient traversant entre 4 et 23 h après l’augmentation de

charge initiant la rupture dans le cas où il ne s’effondre pas. La formation
359
Q.105 - R.20
Fig. 9
Évolution de la hauteur d’eau à l’amont H (m), du débit Q (l/s) et de la concentration
solide de l’effluent c (g/l) fonction du temps T (h) pour l’ensemble de l’essai 8 (en
haut) et centré sur la phase de progression du conduit des essais 4 et 6 (en bas).
Evolution of the upstream water level H(m), the flowrate Q (l/s) and solid
concentration of the exit flow c (g/l) function of time T (h) for the whole test 8 (Top),
and focused on the progression phase of the pipe, tests 4 and 8 (Bottom).
Fig. 10
Évolution de la longueur du conduit L (m) au-dessus de l’interface en fonction du
temps T (min) d’après les mesures fibre optique (gauche) et photo du conduit
après essai (droite).
Evolution of the pipe length L(m) above the interface function of time T (min) based
on the optical fiber measurements (left) and picture of the pipe after test (right).
360
Q.105 - R.20
d’un conduit est détectée de quelques dizaines de minutes à 1 h avant la rup-

ture. Cette échelle de temps (de l’ordre de l’heure) est cohérente avec celle
du modèle de rupture par érosion de conduit des sols peu résistants (limons
sableux) [16].
Les conditions limites hydrauliques et structurelles et le type et l’état des sols

en jeu impactent fortement la cinétique de l’érosion et l’apparition de la rupture.
Le moteur de l’érosion est la vitesse de l’eau au contact du limon. La vitesse de
Darcy fournit donc le critère d’initiation de l’érosion. Pour le limon sableux de Bourg-
Lès-Valence testé, la valeur critique est vD = 2 cm/s. L’apparition du conduit est
le critère d’initiation de la brèche. En dehors de l’observation visuelle, le débit de
fuite ponctuelle est le paramètre le plus pertinent. Mais les valeurs dépendent des
conditions aux limites et de la résistance du limon. La valeur la plus faible en dessous
de laquelle le conduit n’est pas apparu dans ces essais est Q=8 l/s (sur 4 m de
largeur). Le temps de rupture au-delà de ce débit d’initiation peut être estimé avec
la relation de Bonelli [16].
6. CONCLUSION
Bien qu’étant un sujet crucial pour la sécurité des barrages en remblai et des
levées, l’évaluation de l’hydrogramme de rupture reste un domaine où les outils
de prédiction et de modélisation validés et disponibles pour l’ingénieur sont peu
développés. Un effort de recherche important reste à accomplir pour décrire cer-
tains processus physiques déterminants et développer des outils de modélisation
validés. Lorsqu’un tel effort de recherche est conduit, comme cela a été le cas pour le
mécanisme d’érosion de contact pouvant déboucher sur la formation d’une brèche,
à partir d’essais expérimentaux à grande échelle, une description d’enchaînements
de mécanismes complexes peut être menée, conduisant à une méthode d’analyse
de ce risque de rupture. L’étude de l’influence de la prise en compte du transport
sédimentaire à l’aval de la brèche sur l’hydrogramme de brèche, basée sur la modéli-
sation de trois cas d’ouvrages réels ou d’expérimentation à grande échelle, montre
qu’indépendamment du volume de sédiments mobilisé, cette influence peut s’avérer
faible pour une pente du terrain aval forte, et significative lorsque cette pente est
faible. L’analyse de quatre petits barrages du Sud-Ouest de la France qui ont subi
des écoulements de surverse sans rompre, menée en caractérisant les résistances
à l’érosion des matériaux constitutifs de ces ouvrages ainsi que les conditions de
surverse, met en évidence l’influence importante de la végétation recouvrant les
parements aval, de la variabilité des caractéristiques des matériaux, ainsi que de la
saturation superficielle du parement aval. Ce retour d’expérience sur des ouvrages
ayant surversé sans rompre met en évidence le besoin de mieux cerner le rôle de la
végétation sur l’écoulement et la résistance du remblai et la variabilité spatiale des
propriétés mécaniques et hydriques des matériaux.
361
Q.105 - R.20
7. REMERCIEMENTS
Nous remercions très sincèrement les maîtres d’ouvrages des barrages ayant
servi à notre retour d’expérience ainsi que la DREAL Occitanie et la DDT de l’Aude
pour l’identification des barrages endommagés et leur accompagnement lors des
visites de terrain.
REFERENCES
[1] EL KADI, K., PAQUIER, A., 2009. One-dimensional numerical modeling of

sediment transport and bed deformation in open channels. Water Resources
Research vol. 45, n˚ W05404, 20 p.
[2] COURIVAUD J.-R., DEROO L., BONELLI S., 2019. Erosion externe des
barrages et des digues. Colloque CFBR 27-28/11/2019.
[3] CEATI – DSIG, Evaluation of Numerical Models for Simulating Embankment

Dam Erosion and Breach Processes, DSO-2017-02, USBR, August 2017.
[4] MEZBACHE S., PAQUIER A., HASBAIA M., 2020. A coupled 1D/2D model
for simulating river sediment transport and bed evolution. Journal of Hydroin-
formatics.
[5] MORRIS M. W., HASSAN M.A.A.M., VASKINN, K. A. (2007). Breach forma-

tion: Field test and laboratory experiments, Journal of Hydraulic Research, 45
(sup1), 9-17.
[6] PAQUIER, A., 2015. Comparison of three methods assessing the inunda-
tion caused by the erosion of an embankment. Houille Blanche - Revue
Internationale de l’Eau, n˚ 4, p. 32-38.
[7] QUEFFELEAN Y., FOUQUET M., 2017. Etude des risques hydrauliques
induits par le glissement du Pas de l’Ours (Commune d’Aiguilles - 05).
ONF-RTM, Gap, France. 43 pages.
[8] ABRIL, J. B., KNIGHT, D., 2004. Stabilising the Paute River in Ecuador.
Proceedings of ICE vol. 157, p. 32-38
[9] PIOTTE O., MONTMERLE T., FOUCHIER C., BELLEUDY A., GARANDEAU
L., JANET B., JAUFFRET C., DEMARGNE J., ORGANDE D, (2020). Le ser-
vice d’avertissement sur les pluies intenses et les crues soudaines en France.
Colloque SHF : “ Ruissellement, Lyon 30 nov-2 déc 2020 ”
[10] HANSON G.J., SIMON A. (2001). Erodibility of cohesive streambeds in the

loess area of the midwestern USA, Hydrological Processes, 15, p. 23–38.
Doi : 10.1002/hyp.149
362
Q.105 - R.20
[11] WAN C.F., FELL R. (2004). Investigation of Rate of Erosion of Soils in Embank-
ment Dams, Journal of Geotechnical and Geoenvironmental Engineering,
130(4), pp. 373–380; Doi : 10.1061/(ASCE)1090-0241(2004)130:4(373)
[12] FELL R. and FRY J.-J., 2007. Internal Erosion of Dams and their Foundations.
Taylor & Francis ed.
[13] FloodProBE (2012), D3.1 ‘Guidance on improved performance of urban flood

defences’, 2012.
[14] BEGUIN R., 2011. Etude multi-échelle de l’érosion de contact au sein des
ouvrages hydrauliques en terre.
[15] PICAULT C., COURIVAUD J.-R., BEGUIN R., PHILIPPE P., Justification de la
stabilité d’une recharge granulaire aval sur les digues de canaux : développe-
ment d’un critère de rupture par claquage. - In Colloque Justification des
barrages : Etat de l’art et Perspectives. Chambéry : CFBR - 27,28 novembre
2019. doi : 10.24346/cfbr_colloque2019_c28.
[16] BONELLI, S. (ed.). Erosion in geomechanics applied to dams and levees. John
Wiley & Sons, 2013.
363
Q.105 - R.21
-------
-------
ÉVALUATION DE LA SÛRETÉ DES BARRAGES EN FRANCE?

RETOUR D’EXPÉRIENCE ET DÉVELOPPEMENT MÉTHODOLOGIQUE(∗)
Laurent PEYRAS et Patrick DIVOUX

INRAE et EDF
Frédéric LAUGIER, Thierry GUILLOTEAU, Marie CUBAYNES, Mélanie TRON

EDF
Thomas ADELINE, Michel POUPART

Experts indépendants
Catherine CASTEIGTS
SCP
Benjamin DELARUELLE
ARTELIA
Jean-Charles PALACIOS
SAFEGE
Gladys PAVADAY
CNR
Guirec PREVOT
PoNSOH
Agnes VALLEE et Thibault BALOUIN

INERIS
Eric VUILLERMET
BRLi
FRANCE
∗ Assessment of dam safety in France – feedback and methodological development
364 DOI: 10.1201/9781003211808-67

Q.105 - R.21
RÉSUMÉ
La réglementation française a introduit en 2007 l’obligation de réaliser des

analyses de risques pour les grands barrages à travers les Études de Dangers
(EDD). Au cœur du dispositif de prévention des risques, les EDD exigées par la
réglementation française se réfèrent explicitement à l’analyse de risques. Plus de
600 EDD ont été ainsi réalisées lors de cette dernière décennie. Ce caractère
systématique de la réalisation de ces études est donc inédit au niveau mondial.
L’article s’attache à présenter le retour d’expérience de ces études en s’appuyant
sur le rapport du groupe de travail du CFBR. Il présente également des développe-
ments méthodologiques spécifiques sur la recherche des modes de défaillance et
la construction d’arbres de causes.
SUMMARY
French regulations introduced in 2007 the obligation to carry out risk analyses
for large dams through Safety Review Risk Assessment studies (SaRRA). At the
heart of the risk prevention system, SaRRA required by French regulations explicitly
refer to Risk Analysis. Over 600 SaRRA have been performed in the past decade.
This systematic nature of carrying out these studies is therefore unprecedented
worldwide. The article sets out to present the feedback from these studies based on
the report of the FRCOLD working group. It also presents specific methodological
developments on the search for failure modes (FMEA) and the construction of Fault
Trees.
1. INTRODUCTION – CONTEXTE
La réglementation française a introduit en 2007 [1] l’obligation de réaliser des

analyses de risques pour les grands barrages∗ à travers les études de dangers
(EDD). Au cœur du dispositif de prévention des risques, l’EDD constitue une base
de travail dans l’élaboration des stratégies de prévention. Réalisée sous la respon-
sabilité du propriétaire ou concessionnaire, elle s’attache à identifier les risques,
quantifier leur occurrence et leurs conséquences et démontrer comment sont gérés
les risques d’accidents majeurs associés au barrage. Les EDD sont réalisées par
des chargés d’études – bureaux d’études, services techniques du responsable
de l’ouvrage – agréés par l’État pour leurs compétences démontrées. Elles sont
soumises à l’examen de l’autorité de contrôle de l’État.
Les EDD exigées par la réglementation française se réfèrent explicitement à

l’analyse de risques [2, 3, 4]. Les méthodes d’analyse, incluant l’évaluation de la
∗La réglementation française couvre les barrages et les digues de canaux qui, dans la suite
de l’article, sont dénommés sous le terme générique barrages
365
Q.105 - R.21
Fig. 1
La démarche des EDD en France [5].
Safety Review Risk Assessment (SaRRA) framework in France [5].
sûreté et des enjeux, sont laissées au choix du responsable de l’ouvrage et de son

chargé d’études, en fonction des caractéristiques de l’ouvrage. L’article présente le
retour d’expérience des EDD rassemblé dans la publication du CFBR [5]. Il détaille
les étapes des EDD (Fig. 1) ainsi que des développements méthodologiques spé-
cifiques sur la recherche des modes de défaillance et la construction d’arbres de
causes.
2. LES DONNÉES D’ENTRÉE DES EDD
Une analyse de risques nécessite d’avoir la meilleure connaissance possible

du barrage, de son environnement, de son comportement, ainsi que de l’état de
fonctionnement et du niveau de performance de ses organes et systèmes de sécu-
rité. Ainsi la réalisation des EDD intègre en préliminaire la recherche des données
d’entrée, dont les principales sont synthétisées en Fig. 2 [5]:
Fig. 2
Les données d’entrée des EDD.
Input data for Safety Review Risk Assessment.
366
Q.105 - R.21
La collecte des données d’entrée est complétée par une analyse critique des
études caractérisant les aléas naturels susceptibles d’avoir un impact sur la sûreté
du barrage. Si besoin, des études sont reprises ou engagées en cas d’absence.
Pour compléter ces données collectées, un examen exhaustif des ouvrages et

de leurs équipements est effectué. Il vise à fournir l’image la plus complète et précise
possible de leur état et de leur fonctionnement. Celui-ci est réalisé par des spécial-
istes (génie civil, mécanique, contrôle-commande), avec des moyens adaptés au
contexte d’exploitation (avec vidange, retenue pleine ou avec abaissement partiel
du plan d’eau qui constitue le cas le plus fréquent).
Enfin, les données relatives à l’accidentologie du barrage étudié -s’il en existe-

ou de barrages de même type, sont intégrées [6, 7].
3. L’ANALYSE FONCTIONNELLE
Les EDD en France comportent une analyse fonctionnelle [5]. Celle-ci permet
de recenser et caractériser les fonctions du barrage, afin de fournir les informations
nécessaires à la recherche des modes de défaillance. Elle décrit les usages assurés
par le barrage étudié : stockage d’eau, production hydroélectrique, écrêtement des
crues, soutien d’étiage, et définit les fonctions des différents composants ainsi que
les interactions entre le barrage et son environnement.
L’analyse fonctionnelle définit le périmètre d’étude qui comprend le barrage et

tous les ouvrages et matériels associés contribuant directement aux fonctions de
sécurité (Fig. 3).
Fig. 3
Illustration du périmètre d’étude [5].
Illustration of the study perimeter [5].
367
Q.105 - R.21
Elle identifie les fonctions accomplies par les composants du barrage : les fonc-
tions principales qui représentent la fonction majeure du composant pour laquelle
il a été conçu, et les fonctions technologiques que le composant doit remplir pour
maintenir son intégrité et garantir sa performance. L’analyse fonctionnelle décrit les
interactions entre l’environnement en amont et en aval et le barrage lui-même. Cela
permet d’identifier les facteurs d’agression externe de l’ouvrage ou les aléas, et les
enjeux potentiels en termes de sécurité.
4. ANALYSE DES MODES DE DÉFAILLANCE
Sur la base de l’analyse fonctionnelle, l’EDD met en œuvre une analyse des
modes de défaillance du barrage, qui vise à déterminer les événements redoutés,
initiateurs ou intermédiaires, des scénarios de défaillance. Les deux principales
méthodes utilisées en France sont l’Analyse Préliminaire des Risques (APR) et
l’Analyse des Modes de Défaillance et de leurs Effets (AMDE).
La méthode d’APR, utilisée dans 77% des EDD [8], repose sur une analyse
simple des risques basée sur une appréciation experte : l’identification des dangers
(Fig. 4) est effectuée grâce à l’expérience et aux connaissances du panel d’experts
constitué pour l’étude. La méthode est pertinente pour les barrages, dans la mesure
où le jugement de l’expert reste essentiel pour la détermination d’un certain nombre
de dégradations et de défaillances. Il faut noter qu’en l’absence de normalisation
de la méthode, on constate une grande variabilité dans les APR produites. Ainsi,
certaines APR apparaissent très complètes dans l’analyse des modes de défaillance
et s’apparentent au final plus à des AMDE.
Fig. 4
Exemple d’application de l’APR pour un barrage-voûte.
Example of PRA application for an arch dam.
La méthode AMDE, employée dans 23% des études [8], permet d’aboutir à
une analyse détaillée des modes de défaillance du barrage (Fig. 5). Dans un cadre
normalisé, son avantage est la précision et la qualité de l’analyse. La méthode AMDE
constitue donc une analyse exhaustive des modes de défaillance des composants
368
Q.105 - R.21
Fig. 5
Exemple d’AMDE pour la recharge amont d’un barrage en enrochements.
Example of FMEA for the upstream embankment of a rockfill dam.
du barrage, après une analyse fonctionnelle de qualité. Elle est donc bien adaptée
à une analyse de risques complète, voire quantitative ultérieure.
5. MODÉLISATION DES SCÉNARIOS PAR LES MÉTHODES

ARBORESCENTES
Sur la base de la connaissance des modes de défaillance, l’EDD produit

une représentation des scénarios de défaillance à partir de méthodes arbores-
centes. Ces méthodes reposent sur une visualisation graphique des séquences
accidentelles susceptibles de se produire sur l’ouvrage. Les méthodes mises en
œuvre sont [5, 8]:
• La méthode des arbres d’évènements : elle décrit les scénarios de défail-

lance à partir d’un évènement initiateur, conduisant à la défaillance d’un
composant ou d’une partie du système pour déterminer les évènements
qui en découlent. On détermine l’enchaînement des modes de défaillances
jusqu’à l’occurrence d’un accident potentiel, par une démarche inductive ;
• La méthode des arbres de défaillances : elle décrit les combinaisons pos-
sibles d’évènements qui permettent la réalisation d’un évènement redouté
369
Q.105 - R.21
préalablement défini. Le point de départ de la construction d’un arbre de

défaillances est l’évènement redouté lui-même et le principe est de définir
des niveaux successifs d’évènements tels que chacun est une conséquence
d’un ou plusieurs évènements du niveau précédent, à travers une démarche
déductive;
• La méthode du nœud papillon: elle combine un arbre de défaillances et
un arbre d’évènements autour d’un même Evènement Redouté (ER). Les
ER sont des situations dangereuses conduisant principalement à la perte
d’intégrité partielle ou totale de l’ouvrage ou d’un de ses éléments (vanne
de fond ou de surface, bouchon de galerie …). Le nœud papillon (Fig. 6)
permet d’avoir une vision globale des différents scénarios d’accidents et de
leurs conséquences sur leur environnement. La méthode du Nœud Papillon
est utilisée de façon prépondérante (90% des EDD [8]).
Fig. 6
Exemple de nœud papillon relatif à la défaillance d’une vidange de fond.
Example of bow tie for the failure of a bottom outlet.
6. L’ÉVALUATION DE LA SÛRETÉ
Le principe consiste à évaluer individuellement les probabilités de défaillance

des événements élémentaires d’un scénario, puis de les agréger par des règles
mathématiques adaptées à la méthode choisie, permettant d’obtenir au final la
probabilité globale du scénario.
La pratique la plus utilisée pour l’évaluation des probabilités des défaillances

est une approche semi-quantitative d’évaluation de la sûreté, dans laquelle les éval-
uations sont obtenues au moyen de grilles d’analyse faisant correspondre des dires
d’experts à des intervalles de probabilité (Fig. 7).
Le retour d’expérience de cette méthode semi-quantitative est globalement

positif. Il permet de tenir compte de manière équilibrée des aspects qualitatifs de
370
Q.105 - R.21
l’analyse (avis d’expert) et de ceux qui peuvent être mieux quantifiés (occurrence
de crues).
Une autre pratique moins fréquente consiste en une approche quantitative

probabiliste d’évaluation de la sûreté, dans laquelle les évaluations se font au moyen
de modèles probabilistes, d’analyses statistiques, ou à défaut par jugement expert.
Dans tous les cas, l’évaluation de la sûreté est produite en s’appuyant sur un groupe
de travail regroupant un panel d’experts et d’exploitants. Ce panel assure le contrôle
et la validation de la modélisation des scénarios, l’évaluation des probabilités des
défaillances sous forme d’un intervalle de probabilité ou d’une probabilité unique et
la validation de l’évaluation de la probabilité du scénario dans son ensemble, notam-
ment par l’inter-comparaison avec les évaluations des différents scénarios et leur
hiérarchisation entre eux.
Fig. 7
Tableau de correspondance pour l’évaluation des probabilités.
Correspondence table for the probabilities evaluation.
Les EDD introduisent des barrières de sécurité dans les scénarios. Ces bar-
rières sont des éléments techniques ou organisationnels assurant une fonction de
sécurité. Les EDD considèrent essentiellement des barrières de prévention qui
sont les mesures visant à diminuer la probabilité d’un événement indésirable. Une
barrière doit être indépendante de l’événement redouté, fiable, avoir un temps de
réponse en adéquation avec la cinétique de l’événement redouté et doit pouvoir être
testée. Les exemples suivants de barrières de prévention peuvent être cités (Fig.
6) : (1) Intervention par société agréée et Essai de réception, (2) Contrôle de l’état
lors des visites périodiques, (3) Test annuel.
371
Q.105 - R.21
7. ÉVALUATION DE LA GRAVITÉ DES SCÉNARIOS
L’évaluation générale de la sûreté est traduite, pour chaque scénario de

défaillance, en termes de conséquences sur les enjeux humains, matériels, organ-
isationnels et environnementaux à l’amont et à l’aval du barrage et de sa retenue.
Les conséquences de chaque accident potentiel sont caractérisées par les débits
et volumes relâchés à partir de l’ouvrage, l’intensité, la cinétique et la durée de
ces effets. Cela permet l’évaluation de la gravité des conséquences pour la zone
impactée, notamment au droit des enjeux importants.
Les investigations mises en œuvre sont adaptées à l’importance des enjeux

et reposent sur la modélisation hydraulique de l’onde de submersion pour les acci-
dents identifiés dont les résultats sont la cartographie des zones submergées à une
échelle adaptée (1/25 000), mentionnant les principales caractéristiques de l’onde
de submersion, et l’évaluation de la population impactée selon un ordre de grandeur
global suffisant pour les besoins de protection civile (populations à l’hectare des
zones d’habitation, commerciales, industrielles…).
Au moyen d’une grille de gravité (Fig. 8), une classe de gravité est attribuée à
chaque scénario, selon l’impact le plus pénalisant sur le territoire concerné en aval
du barrage.
Fig. 8
Exemple de grille d’attribution de classes de gravité [5].
Example of a grid for assigning severity classes [5].
8. ANALYSE DE LA CRITICITÉ DES SCÉNARIOS
Une fois déterminée la gravité des conséquences et l’occurrence des

scénarios, les EDD croisent ces deux informations afin d’établir la criticité
des scénarios [4,5]. Pour cela, la pratique française s’est largement appuyée sur
372
Q.105 - R.21
des matrices de criticité, consistant en un affichage et une hiérarchisation des dif-

férents scénarios dans un tableau synthétique. On retrouve ainsi les deux entrées
“ gravité ” et “ occurrence ” des scénarios potentiels (Fig. 9).
En définitive cette synthèse doit permettre au responsable de l’ouvrage

de proposer des mesures permettant de maîtriser ou réduire les risques aussi
raisonnablement que possible (principe ALARP). Les objectifs atteints sont alors
représentés dans une matrice de criticité résiduelle qui tient compte de l’apport des
différentes mesures de réduction des risques [5].
Fig. 9
Exemple de grille de criticité avec affichage des limites d’acceptabilité.
Examples of criticality grids with acceptability limits.
9. DÉVELOPPEMENTS MÉTHODOLOGIQUES AUTOUR DE L’ER1

“RUPTURE BARRAGE”
La démarche développée dans les chapitres précédents a été mise en œuvre

pour la production des EDD de première génération. A l’issue de l’analyse prélim-
inaire des risques, la rupture du barrage ou d’une partie significative du barrage
ressort de manière systématique comme l’évènement le plus redouté. Il est désigné
Evènement Redouté n˚ 1 ou ER1.
Le retour d’expérience des premières d’EDD a mis en évidence le besoin

d’une approche méthodologique plus précise pour plus de cohérence lors de la con-
struction de l’arbre des causes de l’ER1 “ Rupture barrage ”. Ce chapitre présente
l’approche méthodologique développée par EDF [9] qui sera déployée lors des
études à venir.
373
Q.105 - R.21
9.1. LES POINTS COMMUNS À TOUS LES BARRAGES
Tous les barrages sont différents et sont autant d’histoires particulières. Un

barrage et sa fondation constituent un ensemble complexe dont la résistance est
particulièrement difficile à appréhender. L’incertitude sur le niveau de chargement
est forte également. Mais les barrages ont des points communs et ont en général
deux fonctions principales:
– retenir un grand volume d’eau ;
– faire transiter de l’eau (débit d’équipement, crues, vidange, débit réservé, …).
Ces fonctions sont assurées par des composants principaux :
– une structure solidaire de sa fondation retient le grand volume d’eau ;
– des circuits hydrauliques dédiés font passer de l’eau (évacuateur de crue,
prise d’eau, circuit de vidange, dispositif de restitution du débit réservé, …).
Les barrages sont dimensionnés pour un niveau d’eau amont (et parfois aval)
maximal nommé Cote des Plus Hautes Eaux (PHE). Jusqu’à cette cote, ils doivent
fonctionner normalement.
Ils doivent tous résister aux mêmes agressions extérieures :

– celles qui conduisent à dépasser la cote des PHE;
– le séisme et ses conséquences post-sismiques;
– les autres agressions involontaires (chutes de blocs sur le barrage, choc de
bateau, ...);
– les agressions délibérées (celles de l’homme mal intentionné).
L’idée est de s’appuyer sur ces points communs à tous les barrages pour
structurer la recherche des causes principales de rupture.
9.2. LES NOTIONS DE DéFAILLANCE ET D’AGRESSION
La rupture d’un barrage est fonction de deux choses:
– la résistance R du barrage et de sa fondation qui dépend des caractéristiques

internes de l’ouvrage et est très liée à son passé.
– la sollicitation F qui s’applique sur le barrage et sa fondation, fonction des
actions extérieures qui s’exercent ou s’exerceront sur l’ouvrage à l’avenir.
En l’absence d’incertitude, la rupture se produirait très exactement lorsque F > R.

Pour qu’il y ait une rupture, c’est qu’il y a eu à l’origine:
– un défaut de résistance: la résistance R a diminué.

– un excès de sollicitation : la sollicitation F a augmenté.
374
Q.105 - R.21
Fig. 10
Probabilité de rupture – Extraits de Lacasse & Hoeg [10].
Failure probability – From Lacasse & Hoeg [10].
La résistance R et la sollicitation F ont des valeurs précises mais qui ne sont pas
connues précisément. Notre niveau de connaissance est incomplet et donne à la
notion de rupture une nature probabiliste. Il faut donc considérer (Fig. 10) :
– une fonction de probabilité de résistance (R),

– une fonction de probabilité de sollicitation (F),
– un aléa de rupture (M) (surface d’intersection entre (R) et (F)) caractérisé,
de manière quantitative, par une probabilité de rupture Pf .
L’aléa de rupture augmente (Fig. 11) en cas :
Fig. 11
Augmentation de l’aléa de rupture (défaillance et agression).
Increase of failure hazard.
375
Q.105 - R.21
– de défaillance (baisse de la résistance ou augmentation de l’incertitude sur

la résistance);
– d’agression (augmentation de la sollicitation ou augmentation de
l’incertitude sur la sollicitation).
À l’origine de la rupture d’un barrage, il y a donc soit un défaut de résistance

(une défaillance), soit un excès de sollicitation (une agression). Par analogie, une
personne peut être victime d’une maladie ou d’un accident. On ne recherche pas les
causes d’une défaillance (maladie) comme on recherche les causes d’une agression
(accident).
Dans un cas, la défaillance (maladie), on se tourne en priorité à l’intérieur

de l’ouvrage (de la personne) pour rechercher ce qui est fragile. Dans l’autre
cas, l’agression (l’accident), on observe d’abord l’environnement extérieur pour
rechercher ce qui peut arriver et comment.
La base de la méthode proposée est assez simple et consiste à construire

l’arbre des causes de l’ER1 autour des notions de défaillance et d’agression.
9.3. LES BRANCHES PRINCIPALES DE L’ARBRE DES CAUSES DE L’ER1
9.3.1. Les ruptures par défaillance
Qu’est-ce qui est fragile dans l’ouvrage ? Il peut être victime de la perte d’une
de ses fonctions principales, ce qui conduit à distinguer :
• La défaillance structurelle ou la perte de la fonction “ retenir l’eau ”,

• La défaillance (d’un circuit) hydraulique ou la perte de la fonction “ faire tran-
siter l’eau ”. Cela se produit lorsque la défaillance d’un circuit hydraulique
entraîne la rupture du barrage, ce qui n’est pas systématique.
L’ouvrage se rompt dans une situation qualifiée de normale ou exceptionnelle

de crue (cote ≤ PHE). Les sollicitations sont connues et auraient dues être support-
ées sans dommage. L’analyse des risques de rupture par défaillance est menée en
analysant les caractéristiques du barrage et de sa fondation et en recherchant les
scénarios de rupture potentiels. La démarche comprend deux étapes principales:
1. Recherche des scénarios de rupture par défaillance qui comprennent chacun:

• un mode de rupture;
• un contexte de référence (en crue, hors-crue, vannes fermées, …);
• un ou plusieurs évènements initiateurs (surcharges, défauts de résis-
tance).
2. Evaluation de la sureté au travers de l’évaluation de la marge (aléa de rupture)
pour les différents scénarios de rupture.
376
Q.105 - R.21
L’arbre des causes présentera chacun des scénarios de “ défaillance” qui aura
été identifié.
La défaillance structurelle
Une partie de la structure, de sa fondation ou de l’interface structure-fondation

n’assure pas son rôle dans la fonction “ retenir l’eau ”. La défaillance structurelle
regroupe les notions “ Instabilité barrage ” et “ Instabilité fondation ”.
La rupture trouve sa cause dans des défauts structurels ou Evènements Initia-

teurs (EI) tels qu’un défaut de conception (cas de charge oublié ou mal appréhendé,
mauvais dimensionnement, ...), un défaut de réalisation et/ou un défaut d’entretien
et maintenance (dégradations et/ou vieillissement).
Ces défauts génèrent un scénario de défaillance structurelle et chaque scénario

comprend:
• des évènements initiateurs;

• un contexte de référence (ensemble de sollicitations) considéré comme
défavorable (en crue, hors crue, vannes ouvertes, vannes fermées, ...);
• un mécanisme de rupture.
Pour les barrages, une défaillance structurelle est susceptible d’engendrer des
mécanismes de rupture de type :
• instabilité mécanique d’ensemble;

• érosion interne;
• contraintes excessives.
Le groupe des défaillances structurelles comprend, par exemple, les ruptures

historiques du barrage de Gleno (Italie -1923), du barrage de Malpasset (France –
1958) ou du barrage de Teton (Etats-Unis – 1976). Elles se produisent majoritaire-
ment lors de phases de mise en eau (lorsque l’incertitude sur le niveau de résistance
est plus grande) ou pour des structures vieillissantes (lorsque la résistance diminue).
La défaillance (d’un circuit) hydraulique
Ce type de rupture est associé à un circuit hydraulique qui n’assure plus la fonc-
tion “ faire transiter l’eau ”. Un scenario de “ défaillance hydraulique ” est caractérisé
par :
• un circuit hydraulique (évacuateur de crue, vidange de fond, …);

• des conditions de fonctionnement particulières;
• un mécanisme de défaillance du circuit hydraulique;
• un mécanisme de rupture du barrage consécutif à la défaillance du circuit
hydraulique.
377
Q.105 - R.21
Les principaux mécanismes de rupture du barrage suite à une “ défaillance

hydraulique ” sont :
• affouillement ou sous-cavage;
• érosion interne/externe.
La plupart des défaillances hydrauliques sont causés par la rupture d’un com-
posant de l’évacuateur de crues (y compris dissipation aval) mais les autres circuits
hydrauliques (vidange de fond, prise d’eau, ancienne dérivation provisoire, …)
peuvent aussi engendrer ce type de défaillance. Les évènements et dégradations
survenues sur les barrages d‘Oroville (Etats-Unis – 2017), de Toodbrook (Royaume-
Uni – 2019) ou de Guatajaca (Porto Rico -2017) illustrent le risque de défaillance
hydraulique. Ils montrent également que la rupture du barrage n’est pas systéma-
tique en cas de rupture d’un circuit hydraulique. Ce risque est majoritairement lié
à des difficultés d’appréhension de certains phénomènes hydrauliques (dissipation
d’énergie, pression dynamique, ...) dans des conditions de fonctionnement peu ou
pas éprouvées par le passé. Mais les circuits hydrauliques vieillissants (avec des
aciers corrodés ou des bétons altérés) peuvent aussi conduire vers ce type de
rupture.
9.3.2. Les ruptures suite à une agression
Qu’est-ce qui peut venir mettre en danger l’ouvrage et comment ? Les

agressions extérieures peuvent être classées en 4 catégories:
• dépassement de la cote PHE;

• agression sismique;
• autre agression involontaire;
• agression délibérée.
La situation à la rupture est extrême ou accidentelle. Le terme “ extrême ” est

habituellement utilisé dans les études de comportement en France pour désigner
les cas de charge sismiques.
L’analyse des risques de rupture par agression est menée en se tournant vers
l’extérieur du barrage. L’approche méthodologique introduit la notion de “ situation
de danger ”. Il s’agit dans un premier temps de définir le niveau de sollicitation
pour lequel le risque de rupture est jugé trop important pour l’ouvrage dans son
état actuel. Cette situation dite “ situation de danger ” est considérée comme inac-
ceptable. L’analyse de risque recherche, dans un deuxième temps, l’occurrence
des agressions qui mènent à cette situation. C’est une approche qui peut être mise
en cause dans le sens où la définition d’une situation de danger conduit à définir
une résistance donnée pour l’ouvrage et à ne pas pleinement tenir compte des
incertitudes sur cette résistance. En contrepartie, cette approche est facilement
compréhensible, pragmatique et peut être qualifiée de robuste.
378
Q.105 - R.21
Cette approche est inscrite dans la législation française au travers de la déf-

inition d’une cote de danger (CDD) pour chacun des barrages étudiés. L’analyse
d’une agression comprend trois étapes:
1. Analyse et recherche des agressions potentielles comme par exemple un

séisme ou un dépassement de la cote des PHE;
2. Définition de la situation de danger correspondante;
3. Evaluation de la sureté au travers de l’évaluation de l’occurrence d’atteinte de
la situation de danger.
Dépassement de la cote PHE – Atteinte de la cote de danger
Le barrage se rompt suite au dépassement de la cote des Plus Hautes Eaux

(PHE). Les ruptures après dépassement de la cote PHE comprennent, entre autres,
l’ensemble des ruptures par surverse sur les barrages en remblai, l’ensemble des
ruptures consécutives à des embâcles de l’évacuateur de crue et l’ensemble des
ruptures consécutives à des ondes de submersion et niveaux de crue difficiles à
appréhender.
Les cas de rupture de barrage par exhaussement de la cote de retenue au-

delà de la cote des PHE sont analysés à partir d’une étude préalable dont l’objectif
est de définir la cote de danger (CDD) de l’ouvrage. L’analyse se concentre alors
sur l’occurrence d’atteinte de la CDD.
La cote de danger peut être atteinte dans les conditions suivantes:
• hors-crue (exemple du barrage de Taum Sauk – USA – 2005);

• en crue en tenant compte des risques d’embâcles et de disfonctionnement
des vannes (exemple du barrage de Spencer – USA – 2019);
• en cas d’onde de submersion provoquée par des glissements dans la retenue,
des évènements naturels dans la vallée plus en amont (glissement de terrain,
avalanches, débâcles glaciaires, …) ou par la rupture d’un barrage existant
plus en amont.
Sur cette base, un arbre des causes type, identifiant chacun des évènements
initiateurs pouvant conduire à l’atteinte de la cote de danger, est élaboré.
Agression sismique
La rupture du barrage est la conséquence d’un séisme. Lors de l’analyse, il

faut différencier:
• La rupture immédiate lors du séisme;

• La rupture post-sismique.
379
Q.105 - R.21
Les cas historiques de rupture de barrage suite à un séisme sont plutôt rares
comparativement aux autres types de rupture. Cette catégorie comprend par exem-
ple les ruptures du barrage de Shih-Kang (Taïwan – 1999) ou celle du barrage de
Van Norman (Etats-Unis – 1971 – Séisme de San Fernando).
L’analyse du risque de rupture par agression sismique est particulièrement

complexe. Malgré le développement de modèles de plus en plus sophistiqués,
les incertitudes sur l’aléa sismique, sur les caractéristiques des matériaux et sur
les comportements non linéaires des ouvrages soumis à de forts séismes restent
élevées.
L’évaluation des scénarios et de la sureté sera principalement qualitative et

menée à partir d’une appréciation croisée des éléments suivants:
• l’étude de l’aléa sismique;

• la sensibilité du barrage;
• les calculs et études spécifiques de la résistance au séisme de l’ouvrage;
• le retour d’expérience international sur la résistance aux séismes des bar-
rages.
Autre agression involontaire
Une agression involontaire autre qu’un séisme ou un dépassement de la cote

PHE est caractérisée par un choc sur le barrage qui engendre sa destruction comme
par exemple la chute d’une grosse masse rocheuse sur l’ouvrage ou un choc de
bateau.
Le risque de rupture consécutif à une autre agression involontaire est très

faible mais il doit néanmoins être recherché et analysé au cas par cas avant d’être
éliminé.
Agression délibérée
Il s’agit de rupture liée à des actes terroristes, actes de guerre, …L’analyse

des agressions délibérées sort du périmètre des EDD. Ces analyses sont menées
par ailleurs avec le degré de confidentialité adéquat.
9.4. L’ARBRE DES CAUSES DE L’ER1 “ RUPTURE BARRAGE ”
Les notions développées précédemment permettent de structurer la base de

l’arbre des causes de l’ER1 “ Rupture barrage ” (Fig. 12). Il comporte six branches
principales. Deux d’entre elles (autre agression involontaire, agression délibérée)
seront en général exclues au stade de l’analyse préliminaire. Quatre d’entre elles
(défaillances structurelles, défaillances hydrauliques, atteinte de la cote de danger,
380
Q.105 - R.21
agression sismique) sont développées plus en détail jusqu’à l’identification des

principaux évènements initiateurs de la rupture.
La structuration proposée présente l’avantage d’être commune à tous les

barrages et oblige à bien situer le contexte de la rupture. Cela permet aussi une
évaluation plus rationnelle des évènements initiateurs sur l’échelle de A à E présen-
tée au chapitre 6, tout gardant à l’esprit le fait qu’un niveau d’expertise élevé et
spécifique est requis pour évaluer chacun d’entre eux.
Fig. 12
Structure type de l’arbre des causes de l’ER1.
Typical fault tree for dam failure analysis.
9.5. VALIDATION ET DÉCLINAISON AUX AUTRES ÉVÈNEMENTS REDOUTÉS
Une manière de valider les développements méthodologiques proposés pour

construire l’arbre des causes de l’ER1 “ Rupture barrage ” consiste à regarder
si les ruptures historiques trouvent leur place dans l’arbre type proposé. C’est le
cas pour celles mentionnées dans cet article à titre d’exemple. Ces développe-
ments méthodologiques s’inscrivent néanmoins dans une démarche de progrès
pour laquelle le retour d’expérience des études à venir sera très utile.
381
Q.105 - R.21
La méthodologie proposée pour l’ER1 peut se décliner à tous les autres ER en

recherchant et en analysant séparément les modes de rupture par défaillance et les
modes de rupture par agression. Les défaillances se produisent dans la gamme
de fonctionnement prévue pour l’ouvrage. Les agressions sont les sollicitations
extérieures excessives et considérées inacceptables pour l’ouvrage.
10. CONCLUSION: LES APPORTS DE L’ANALYSE DE RISQUES SUR LA

SÛRETÉ DES BARRAGES
10.1. LES APPORTS DES EDD
Les analyses de risques menées à travers les EDD en France, réalisées de

manière systématique sur les quelque 631 grands barrages du parc français, ont
eu, de l’avis de tous les acteurs, des impacts positifs majeurs pour la sûreté des
barrages. Parmi ces apports, on peut citer:
– La sûreté des barrages a longtemps été examinée essentiellement sous

l’aspect “ génie civil ”. Les EDD ont permis une nouvelle approche fonction-
nelle multi-métiers;
– Une amélioration notable de la connaissance sur les ouvrages. L’EDD
constitue une excellente synthèse documentaire et descriptive de la con-
naissance du barrage;
– La mise en lumière des incertitudes par l’introduction des concepts de prob-
abilités. Cela a été une démarche nouvelle pour beaucoup d’acteurs de
l’évaluation de la sécurité.
– Une identification et un interclassement des scénarios de défaillance et des
risques sûreté majeurs. Des défaillances potentielles ont parfois pu être
mises en lumière en examinant des scénarios complexes associant des
défaillances des organes hydromécaniques, du contrôle-commande, des
facteurs organisationnels et humains avec des scénarios hydrologiques pou-
vant conduire à la rupture de l’ouvrage. Les EDD ont permis de disposer
d’une vision globale des risques sur un barrage, avec une proposition de
hiérarchisation, constituant ainsi un outil précieux d’aide à la décision pour
les mesures à mettre en œuvre par le responsable d’ouvrage.
Par ailleurs de nombreux échanges entre l’autorité de contrôle en charge

d’instruire les EDD et les responsables de l’ouvrage et leurs chargés d’études ont
été nécessaires, d’une part pour permettre l’application de la réglementation sur
le contenu des EDD, et d’autre part sur les conclusions de l’étude relatives aux
mesures de maitrise et de réduction du risque. Sur ce dernier point, les mesures de
réductions des risques proposées sont tracées et assorties de délais dans des actes
administratifs pour leur mise en œuvre. Par ailleurs, l’autorité de contrôle s’assure
382
Q.105 - R.21
régulièrement de la bonne réalisation des engagements du responsable d’ouvrage,

en particulier lors des inspections périodiques.
L’EDD est également un document de synthèse utile lors des inspections

de l’ouvrage. En mettant en exergue les matériels et procédures technico-
organisationnelles garantissant la sécurité de l’ouvrage en exploitation, le contrôle
est en particulier axé sur ces éléments et permet de gagner en pertinence et
efficacité.
10.2. UNE DÉMARCHE DE PROGRÈS
Pour les acteurs œuvrant sur les barrages, les EDD donnent du sens à
certaines actions de maintenance ou d’auscultation et peuvent amener à revoir
certaines pratiques pour mieux prendre en compte les risques. Elles aident fréquem-
ment à mieux comprendre et donc accepter les contraintes de disponibilité des
organes sûreté par exemple. Les EDD permettent alors un nouveau cadre de
référence pour tous les acteurs de l’exploitation et de la maintenance.
Les EDD viennent compléter de manière significative la connaissance des

risques majeurs et autres risques liés à l’exploitation des barrages. Avant la mise
en place des EDD, les seuls éléments connus relatifs aux risques des barrages
provenaient des plans d’alerte prévus pour les plus grands barrages. Elles sont
donc, pour le public et pour les collectivités, une approche nouvelle qui s’inscrit dans
une démarche de progrès continu. Le retour d’expérience des premières EDD est
actuellement exploité et conduit vers des développements méthodologiques comme
par exemple une approche plus cohérente et plus structurée pour la construction
des arbres des causes des évènements redoutés.
10.3. VERS UNE DEUXIÈME GÉNÉRATION D’EDD
Les EDD sont désormais en France un document essentiel, par lequel le

responsable de l’ouvrage démontre la sûreté de son barrage en toutes situations
d’exploitation. Il s’agit d’un document utile à tous les acteurs, y compris à l’autorité
de contrôle, qui permet d’identifier les points critiques de la sûreté du barrage.
Avec l’évolution de la réglementation en 2015 [11], l’État a voulu réaffirmer

et renforcer le rôle de l’EDD. Le plan et le contenu des EDD sont consolidés au
travers de l’approfondissement des données d’entrée: un examen exhaustif des
ouvrages et équipements, un bilan de leur état, de leur comportement et de leur
conception sont désormais demandés explicitement en préalable de l’analyse de
risques. Des éléments plus précis sur la gestion de la sûreté du barrage (gestion de
la sous-traitance, définition des responsabilités…) sont à fournir. L’évaluation des
risques est plus complète avec notamment une fourniture systématique des ondes
383
Q.105 - R.21
de submersion au format 1/25 000, y compris pour les scénarios autres que celui
de la rupture du barrage mais pouvant conduire à des conséquences graves.
REFERENCES
[1] Ministère de l’Ecologie, du Développement Durable et de l’Energie (2007).

Décret n˚ 2007-1735 du 11 décembre 2007 relatif à la sécurité des ouvrages
hydrauliques et au comité technique permanent des barrages et des ouvrages
hydrauliques. J. O. 13/12/2007.
[2] CIGB / ICOLD (2005). Bulletin n˚ 130 – Evaluation du risque dans la gestion
de la sécurité du barrage.
[3] Norme NF EN 31010 “ Gestion des risques – Techniques d’évaluation des

risques ”, Juillet 2010
[4] Ministère de l’Ecologie, du Développement Durable et de l’Energie (2012).

Dam Risk Assessment Guidelines. Last Update: August 2012- English transla-
tion: February 2015. https://www.barrages-cfbr.eu/IMG/pdf/risk_assessment_
french_guidelines.pdf
[5] CFBR (2020). Pratiques françaises de l’analyse de risques et de l’évaluation

de la sûreté des barrages. Juin 2020.
[6] Gastaud, C. Gauthier, A. (2015). Enseignements tirés des événements impor-

tants pour la sureté hydraulique (EISH) et éléments de retour d’expérience
sur les organes hydromécaniques et de contrôle-commande des barrages,
Colloque “Vantellerie et contrôle-commande des barrages ”, Chambéry, 2–3
décembre 2015. CFBR: 353–367.
[7] ICOLD (1995). Bulletin 99: Dam Failures – Statistical Analysis. Paris.
[8] Dieudonné S., Peyras L., Beullac B., Gastaud C., Prévot G., (2016) Retour
d’expérience de la première génération des études de dangers en France:
analyse méthodologique et comparaisons à l’accidentologie. Colloque CFBR
“Sûreté des barrages et enjeux.
[9] Divoux P., Méthodologie des EDD – L’ERC1 Rupture barrage – L’arbre des
causes – EDF – Juin 2020.
[10] Lacasse S., Höeg K., Reliability and risk approach for the design and safety
evaluation of dams – Approche fiabilité et risque pour la conception et
l’évaluation de la sécurité des barrages – NGI – Norway – ICOLD 2019.
[11] Ministère de l’Écologie, du Développement Durable et de l’Énergie (2015).

Décret n˚ 2015–526 du 12 mai 2015 relatif aux règles applicables aux ouvrages
construits ou aménagés en vue de prévenir les inondations et aux règles de
sûreté des ouvrages hydrauliques. J. O. 14/05/2015.
384
Q.105 - R.22
-------
-------
GESTION DE LA SÉCURITÉ DES BARRAGES – ILLUSTRATION CHEZ DEUX

OPERATEURS FRANÇAIS (∗)
Laurent BESSADI
Chargé de mission Génie Civil, EDF HYDRO
Frédéric LAUGIER
Expert Sécurité des barrages, EDF HYDRO
Yann TARAVEL
Ingénieur Gestion d’Actifs Génie Civil, CNR
FRANCE
RÉSUMÉ
Ce rapport présente des facteurs de réussite dans la gestion de la sécurité

des barrages chez deux grands opérateurs hydro-électriciens français.
En premier lieu l’accent est mis sur des “ bonnes pratiques ” en matière de
gouvernance à EDF Hydro : mise en œuvre du politique de sécurité axées sur
trois risques majeurs (risques liés à l’exploitation, risque Crue et risque de Rup-
ture d’Ouvrage), implication du mangement, contrôles internes et indépendants,
formation, instances dédiées au pilotage des actions nationales, gestion du référen-
tiel, modalités de traitement des écarts et prise en compte du retour d’expérience
des incidents internes, nationaux et internationaux. Un focus est fait sur la prise
en compte du risque Crues au travers des prévisions hydrométéorologiques, des
consignes de crues et de la capitalisation du retour d’expérience. La gestion du
risque de Ruptures d’Ouvrages est aussi illustrée par les actions de surveillance des
∗ Dam Safety Management – Example of two French dam owners
DOI: 10.1201/9781003211808-68 385

Q.105 - R.22
ouvrages et des matériels et les opérations de requalification après interventions ou

travaux.
Un second chapitre expose le retour d’expérience d’EDF Hydro en matière

d’Etude De Dangers, élément essentiel de la gestion de la sécurité des barrages.
Tout d’abord sont présentés les avantages de l’approche semi-quantitative mise en
œuvre, utilisant des arbres de défaillance de type “ nœud papillon ” : simplicité
de mise en œuvre, facilités de compréhension, rôle des experts. Puis les béné-
fices apportés par cette démarche sont développés : décloisonnement des métiers
grâce à une nouvelle approche fonctionnelle multi-métiers, développement d’une
culture sécurité partagée par tous, amélioration significative de la connaissance
des ouvrages, identification et interclassement des scenarios de défaillance, réduc-
tion des risques par des mesures adaptées, vision globale de la sécurité d’un grand
parc d’ouvrages, et enfin outil de communication interne et externe.
Enfin un troisième chapitre illustre la traduction opérationnelle d’une politique

de sécurité en termes de choix d’investissements dans un système économique
contraint, au travers de la gestion d’actifs à la CNR pour prioriser les opérations
de maintenance lourde des protections à l’érosion en aval des barrages. Après un
rappel sur les fondamentaux de la gestion d’actifs, la déclinaison opérationnelle à
la CNR est présentée : inventaires de l’ensemble des actifs, analyse de l’impact
des projets sur les objectifs de la CNR (respect de la règlementation, disponibil-
ité des ouvrages de production, navigation et évacuation des crues), gestion des
données et partage de la connaissance, management de la qualité, planification et
ordonnancement des projets, évaluation et suivi de la gestion des ouvrages et des
incidents en lien avec la sécurité en vue d’optimiser les budgets, développement
des compétences. L’exemple des protections à l’érosion en aval des barrages vient
illustrer le propos, au travers du cas du barrage du Pouzin.
SUMMARY
This report presents keys of success of dam safety management for two large
French dam operators.
Firstly, dam safety management good practices at EDF Hydro are outlined:
dam safety policy based on three major risks (risk during normal operation, risk
during floods, risk of dam failure), commitment of management, internal and inde-
pendent controls, training, central committees dedicated to broad scale actions
steering, internal dam safety manual, treatment of deviations and integration of inter-
nal, national and international incidents. Risks during flood are emphasized through
weather and flood forecasts, flood routine procedures and proper integration of feed-
back. Management of dam failure risk is highlighted by actions of surveillance and
monitoring of structures and materials and recommissioning procedure after works
or operation.
386
Q.105 - R.22
Secondly, the report presents EDF Hydro lessons learnt from the achievement
of more than 300 dam safety assessment risk analysis in 15 years. These periodic
safety reviews are an essential part of dam safety management. The advantages
of the used semi-quantitative approach based on bow-tie diagrams are explained:
ease to implement, ease to understand by all parties, role of experts. Benefits of the
approach are developed: multi field functional approach which improves communi-
cation between individual specialists, development of common dam safety culture
shared by all, significant improvement of dam knowledge, identification and clas-
sification of failure modes and mechanisms, risk reduction by suitable measures,
global vision of dam safety for a large dam portfolio, and finally tool for internal and
external communication.
Finally, the last part shows an actual example of dam safety policy in terms of
decision making and investment choice in a constrained economical background.
It is exemplified through CNR asset management policy which helps to prioritize
expensive maintenance works concerning protections against erosion at down-
stream of dams. Basics of asset management and operational applications at
CNR are presented: exhaustive inventory of assets, impact assessment on CNR
objectives (compliance to regulation, power plant availability, shipping, and flood
management), data management and sharing knowledge, quality management,
planning and scheduling, assessment and budget monitoring of works and incidents
with regard to safety in order to optimize budgets and skill development. A case study
illustrates these principles: Protection against erosion at Pouzin dam downstream.
1. INTRODUCTION
La sécurité des barrages est un enjeu primordial pour tout exploitant

d’aménagement hydro-électrique. Elle fait partie intégrante, et de façon prioritaire,
de la gestion globale et durable de ses barrages. Face à cet enjeu, tout opérateur
responsable doit mettre en place une organisation robuste et efficace pour assurer
la protection des populations et des biens. Comme présenté dans le bulletin 154
de la CIGB [1], cette sécurité repose sur plusieurs fondamentaux, parmi lesquels
figurent notamment une politique de gestion de la sécurité clairement définie, une
organisation disposant d’une gouvernance efficace, des rôles bien identifiés pour
des prises de décision optimales ainsi que des revues de sécurité périodiques.
Ce rapport illustre plusieurs facteurs de réussite permettant d’atteindre un

niveau d’exigences élevé dans ces composantes essentielles, au travers de
l’exemple de deux grands opérateurs français.
Le premier chapitre présente les principes de gouvernance et d’organisation

de la gestion de la sécurité des barrages à EDF Hydro.
Le second chapitre met l’accent sur l’importance des revues périodiques de

sécurité, au travers du retour d’expérience des Etudes De Dangers (EDD) à EDF
Hydro.
387
Q.105 - R.22
Enfin le troisième chapitre décrit la traduction opérationnelle d’une politique

de sécurité en termes de choix d’investissements dans un système économique
contraint, en présentant la gestion d’actifs à la CNR pour prioriser les opérations de
maintenance lourde des protections à l’érosion en aval des barrages.
2. EXEMPLE DE MODALITES DE GOUVERNANCE ET D’ORGANISATION

DE LA SECURITE DES BARRAGES A EDF HYDRO
Avec plus de 250 grands barrages sous sa responsabilité, EDF Hydro est le
principal opérateur hydroélectrique en France.
EDF Hydro a toujours placé la sécurité des barrages comme priorité absolue
de sa mission d’exploitant, et à ce titre a mis en place depuis de nombreuses années
une gouvernance et une organisation dédiées à cet objectif prioritaire.
2.1. POLITIQUE DE SÉCURITE
La sécurité des personnes et des biens est pleinement intégrée dans la

politique générale d’EDF Hydro, dans un esprit de responsabilisation et d’innovation.
EDF Hydro a défini trois principales familles de risques à maitriser pour assurer
la sécurité de ses aménagements hydroélectriques :
• Les risques liés à l’exploitation : la présence et le fonctionnement des installa-

tions en exploitation hors crue peuvent générer des risques, qui doivent être
limités par une action constante d’identification et de réduction des impacts.
• Le risque Crue : les aménagements hydroélectriques sont soumis aux crues
et doivent pouvoir en assurer le transit sans aggravation des conséquences
naturelles à l’amont et à l’aval.
• Le Risque de Rupture d’Ouvrage : il s’agit du risque majeur, d’occurrence
très faible mais dont les conséquences potentielles sont très graves pour les
biens et les personnes, et qui doit être maîtrisé par des études et expertises,
des contrôles permanents et un haut niveau d’exigences dans la conception,
maintenance et exploitation.
Les grandes orientations de la Direction sont actualisées tous les 3 ans afin
de fixer les priorités pour l’ensemble des entités d’EDF Hydro.
2.2. MANAGEMENT DE LA SÉCURITE
Chaque responsable d’exploitation a pour mission vis-à-vis de la sécurité des

barrages :
388
Q.105 - R.22
• d’assurer le fonctionnement des installations placées sous sa responsabilité,

en maîtrisant en toutes circonstances les risques à caractère hydraulique,
• de surveiller l’état des installations.
La responsabilité du management est clairement affichée : il doit affirmer au

quotidien l’importance et la priorité de la sécurité, organiser sa prise en compte,
assurer le contrôle interne et favoriser le partage au sein du collectif.
L’animation régionale des démarches d’amélioration ainsi que le partage des

bonnes pratiques sont assurés par une filière dédiée en appui du management,
mobilisant une vingtaine de personnes.
Un contrôle interne est réalisé au sein des différentes entités et par le niveau
national.
Un Inspecteur Sureté Hydraulique, directement rattaché à la Présidence du

Groupe EDF à laquelle il rapporte en toute indépendance, a en charge de veiller à
la bonne prise en compte dans les pratiques de terrain des exigences en matière
de sécurité des ouvrages, et établit un rapport annuel (accessible au public) avec
des propositions d’axes de progrès et d’objectifs d’amélioration.
La Direction promeut la diffusion d’une culture de sécurité auprès de

l’ensemble du personnel autour de quatre valeurs principales :
• maintenir une attitude interrogative avant et pendant les interventions,

• déployer une démarche rigoureuse (comprendre et suivre les procédures,
procéder à un autocontrôle, solliciter de l’appui si nécessaire, rendre compte
et tracer les résultats),
• détecter les évènements et écarts, analyser et partager le retour d’expérience,
• coopérer et partager ses expériences, ses informations et savoirs.
Des pratiques de fiabilisation des interventions sont mises en œuvre en vue

d’améliorer la qualité des gestes professionnels (réunions d’enclenchement, min-
utes d’arrêt, autocontrôle, contrôles croisés, communication sécurisée, débriefing).
2.3. FORMATION
La formation du personnel, ainsi que sa qualification pour certaines activités

à fort enjeu en matière de sécurité, sont essentielles pour assurer le respect des
exigences internes et règlementaires.
Les activités liées à la surveillance des ouvrages telles que les inspec-
tions visuelles périodiques, les essais de manœuvre, l’auscultation et certaines
opérations de maintenance, font l’objet d’autorisations délivrées par le manage-
ment, sur la base d’entretiens d’évaluation et à l’issue de cursus de formation /
professionnalisation avec recyclages périodiques. Dans les cas où certaines
389
Q.105 - R.22
prestations sont externalisées, un processus d’évaluation adapté des prestataires

s’applique.
De même le personnel en charge de la gestion des aménagements en crue

doit disposer lui aussi d’une autorisation correspondant à ses prérogatives et son
domaine d’intervention. Il peut bénéficier d’une plate-forme de simulation numérique
offrant un modèle complet des différents systèmes composant son aménagement
et leurs interactions, lui permettant de s’entraîner en réalité virtuelle à faire face à
différents scenarios de crues.
2.4. ORGANISATION DE L’APPUI AU MANAGEMENT DE LA SÉCURITÉ
Un Comité National de Sûreté Hydraulique a en charge d’organiser le domaine

de la sécurité hydraulique en élaborant les politiques et prescriptions associées, de
piloter et contrôler les actions nationales relevant de ce domaine, de proposer des
évolutions stratégiques et d’assurer la cohérence et la prise en compte du retour
d’expérience.
Sur le champ technique, un Comité Technique Hydraulique constitué d’experts

nommés intuitu personae au sein des différentes structures d’EDF Hydro, a pour
mission d’émettre des avis techniques sur les thématiques et dossiers à fort
enjeu, en mettant l’accent prioritairement sur la maîtrise des risques de sécurité
hydraulique.
EDF Hydro dispose d’ingénieries intégrées, agréées par l’Etat, lui permettant
de sécuriser en interne les compétences et principales ressources nécessaires à la
réalisation des études et travaux liés à la sécurité des barrages.
Les Etudes De Dangers et opérations de mise en conformité règlementaire

des barrages sont traitées en mode projet pour en assurer la qualité de réalisation
et simplifier les processus de décision sur un parc important. Le rôle de client est
assuré par les gestionnaires d’actifs en région, le niveau national pilotant les activités
relatives à la méthodologie et au retour d’expérience.
2.5. RÉFÉRENTIEL
Le retour d’expérience du management de la sécurité a conduit EDF Hydro à

s’orienter depuis 2015 dans un processus de modernisation et de simplification de
son référentiel pour une meilleure compréhension et application sur le terrain.
Cette simplification s’est notamment traduite par la création d’un document

unique regroupant l’ensemble des prescriptions de sécurité hydraulique à EDF
390
Q.105 - R.22
Hydro, appelé “ Référentiel de sûreté hydraulique ” (le terme sûreté étant équivalent
en interne EDF Hydro à celui de sécurité au sein de la CIGB).
En complément, un memento accessible à l’ensemble du personnel a été

produit pour expliquer le sens des prescriptions et rappeler de façon pédagogique
les principes essentiels définis dans ce référentiel.
Ces documents sont adossés à la règlementation en vigueur en France.
2.6. GESTION DU RISQUE CRUE
Les crues représentent environ 1/4 des ruptures recensées de barrages dans
le monde.
Pendant un épisode de crue EDF Hydro met en œuvre les principes

d’exploitation prioritaires suivants :
• empêcher la submersion de l’ouvrage afin de maintenir son intégrité,

• ne pas aggraver les conséquences de la crue par rapport à celles qui auraient
été constatées en l’absence d’ouvrages.
Plusieurs modalités d’organisation concourent à atteindre ces objectifs :
• Prévisions
L’anticipation des crues est un élément clé pour l’exploitant. Pour cela il peut
s’appuyer sur un service de prévisions hydrométéorologiques interne à EDF
Hydro, en charge d’alerter et informer les exploitants en continu afin de leur
permettre d’optimiser la gestion des aménagements. Ils peuvent ainsi disposer
de prévisions de débits à différents horizons de temps, y compris au pas
horaire durant les épisodes de crues, et de bulletins quotidiens et d’alertes.
• Consignes de crues
Le chargé d’exploitation dispose d’un référentiel documentaire pour chaque
aménagement, constitué de consignes d’exploitation, instructions et modes
opératoires. Ce référentiel comprend une loi de conduite en crue décrivant
périodiquement le débit à évacuer en fonction des caractéristiques du barrage
(capacitif, barrage mobile en rivière…).
• Retour d’expérience
Les crues étant parfois rares dans certaines vallées, il est important d’en
constituer un historique qui pourra être exploité ultérieurement. Pour cela
l’exploitant consigne dans un rapport de crue les principales informations rel-
atives à la crue et ses conséquences : débits, présence de corps flottants,
manœuvres réalisées, dégradations sur les ouvrages et matériels…
391
Q.105 - R.22
2.7. GESTION DU RISQUE RUPTURE D’OUVRAGE
Il s’agit du risque majeur pour tout exploitant hydroélectrique, en termes de

conséquences potentielles pour les biens et les personnes.
Ce sont les accidents emblématiques en France, notamment la rupture du

barrage de Malpasset en 1959, qui ont contribué à faire évoluer les référentiels des
opérateurs et la réglementation française.
Trois champs d’intervention sont mis en œuvre en exploitation pour maîtriser

ce risque : les activités de surveillance des ouvrages et matériels, leur gestion
patrimoniale (traité au §4 avec l’exemple de la CNR) et leur requalification après
intervention.
2.7.1. Surveillance des ouvrages et matériels
La surveillance des ouvrages et matériels est un des facteurs clé pour maitriser
ce risque, dont l’exploitant est un des acteurs essentiels. Elle concerne les prin-
cipaux ouvrages et matériels des aménagements hydroélectriques : barrages,
canaux, galeries, écluses, conduites forcées et vannes.
Elle se décline suivant quatre axes principaux :
• examens visuels des exploitants,

• mesures d’auscultation,
• essais périodiques des organes de manœuvres,
• maintenance de premier niveau des équipements de surveillance/protection.
Les modalités de cette surveillance sont adaptées aux spécificités techniques

et fonctionnelles des ouvrages et matériels.
Des dispositifs automatiques d’alerte existent pour détecter les surpressions

dans les conduites forcées et les débits de fuite excessifs dans certains remblais
sensibles.
Les ingénieries intégrées d’EDF Hydro produisent ou pilotent la production

des principaux documents de surveillance règlementaires (rapports d’auscultation,
visites techniques approfondies) suivant les périodicités en vigueur.
Les séismes sont surveillés en permanence grâce à un partenariat avec un

laboratoire de géophysique national. Tout séisme de magnitude supérieure ou égale
à 4 peut générer un plan d’action en fonction de la distance de l’ouvrage par rapport
à l’épicentre.
392
Q.105 - R.22
2.7.2. Requalification
A l’issue de toute activité de maintenance ou intervention sur un ouvrage

ou matériel, une série de contrôles et essais sont réalisés avant leur remise en
service.
La requalification se décompose en trois parties :
• requalification intrinsèque pour vérifier les caractéristiques propres,

• requalification fonctionnelle de sous-ensemble afin de vérifier le fonction-
nement du sous-ensemble,
• requalification fonctionnelle d’ensemble, qui regroupe le fonctionnement du
matériel et ses interactions avec le système global.
Pour assurer correctement ces différentes étapes les rôles sont clairement
définis (chargé d’essai, coordinateur des essais…) et les taches pilotées au travers
d’un plan global d’essais.
2.8. TRAITEMENT DES ÉCARTS
Les processus de traitement des écarts et de vérification de la confor-

mité des barrages aux exigences réglementaires applicables sont conçus pour
responsabiliser les exploitants régionaux, sous le contrôle du niveau national.
Tout évènement concernant la sécurité hydraulique est tracé et fait l’objet

d’un traitement systématique dans les 15 jours suivant l’évènement : tracé des faits
et actions immédiates, analyse sommaire si nécessaire. Après validation par la
Direction, les évènements importants sont déclarés à la Préfecture, conformément
à la réglementation française.
Lorsqu’en raison d’une anomalie, il est nécessaire de modifier temporairement

l’état fonctionnel de l’installation pour préserver son niveau de sécurité, on procède à
la mise en place de Dispositifs et Moyens Particuliers (DMP), suivant deux catégories
principales :
• DMP Exploitation correspondant à une modification de l’état fonctionnel de

l’exploitation, sous la responsabilité de l’exploitant,
• DMP Patrimoine correspondant à une dérogation au référentiel technique,
sous la responsabilité du gestionnaire d’actifs.
Ces DMP font l’objet d’une analyse de risques systématique et d’une revue
périodique.
393
Q.105 - R.22
2.9. PRISE EN COMPTE DU RETOUR D’EXPÉRIENCE DES INCIDENTS
La veille et la prise en compte du retour d’expérience des incidents au

niveau interne, national et international, permettent de maintenir un haut niveau
d’exigences en matière de sécurité sur les installations.
Chaque incident estimé significatif, en France ou à l’étranger, fait l’objet d’une

analyse sur la base des informations disponibles, et d’un partage en interne.
3. ETUDES DE DANGER A EDF HYDRO
Un élément essentiel du système de gestion de la sécurité des barrages à EDF

Hydro est constitué par la réalisation périodique des Etudes De Dangers (EDD) de
barrages. Introduite par la réglementation française en 2007 puis amendée en 2015,
l’EDD constitue une revue de sûreté du barrage basée sur une approche de type
analyse de risque.
EDF Hydro a réalisé depuis 2007 plus de 300 EDD de barrages pour un parc
d’environ 240 ouvrages classés concernés par cet exercice.
Pour la réalisation de ses EDD, EDF Hydro a choisi de retenir une approche
dite semi-quantitative basée sur une représentation d’arbres de défaillance de type
“ nœud papillon ”.
Fig. 1
Représentation des risques par nœud papillon.
Bow-tie diagram used for dam risk analysis.
Le choix de cette méthode a présenté certains avantages :
– Une relative simplicité de mise en œuvre.

Les occurrences des évènements sont classées par des classes de probabilité
définies de manière qualitative correspondant à des probabilités d’occurrence
394
Q.105 - R.22
couvrant un ordre de grandeur de puissance de 10 (par exemple entre 10−2

et 10−3 pour la classe B).
– Une compréhension facilitée pour les spécialistes de l’analyse de
risques.
Les opérations mathématiques d’agrégation des évènements se font par le
biais d’opérations booléennes simples de type “ ET ” et “ OU ” à partir de
règles simples.
– Une démarche qui s’appuie pour bonne part sur les dires d’experts pour
la cotation des événements initiateurs et des barrières.
La cotation des évènements relatifs aux ouvrages et matériels se base notam-
ment sur les principes de conception initiale, le comportement historique, l’état
actuel, les études et notes de calcul existantes.
– Un outil relativement “rustique” qui se révèle bien adapté à la réalité du
domaine des barrages où les ouvrages présentent des caractères uniques.
La méthodologie retenue fournit des approches par “ordre de grandeur”,
que cela soit pour la cotation des évènements ou l’évaluation des con-
séquences aval. Ce grain permet tout à fait d’identifier les scénarios critiques
de défaillance du barrage et de proposer les mesures de réduction de risque
appropriées.
Les EDD en France sont réalisés par un Organisme Agréé structurellement

indépendant. A EDF Hydro, elles sont réalisées par son ingénierie intégrée.
Les analyses de risques menées à travers les EDD à EDF Hydro ont eu des
impacts positifs notables pour l’amélioration de la sécurité des barrages, notamment:
(i) Une nouvelle approche fonctionnelle multi-métier qui va au-delà de la

traditionnelle approche “Génie civil” ;
L’incident de la Croux en 1984 [2] puis l’accident du Drac en 1995 ont
largement contribué à faire prendre conscience que les risques liés à la
présence des barrages dépassent largement le cadre des grands ouvrages
de génie civil.
Le retour d’expérience international des incidents et accidents de barrages
montre qu’une bonne partie de ceux-ci ont des causes multifactorielles
incluant des facteurs techniques, organisationnels et humains, les difficultés
se situant souvent aux interfaces entre “métiers”. Les EDD permettent
d’apprécier la sécurité des barrages de manière globale, fonctionnelle et
“décloisonnée”.
(ii) Le développement d’une culture sécurité partagée par tous.
Les EDD sont réalisées par un collectif multi-métier réunissant des acteurs
de l’ingénierie intégrée et de l’exploitation. La contribution à ce groupe de
travail multi-métier est l’occasion d’un échange et partage d’expérience
unique qui permet à tous les acteurs de s’enrichir des compétences
des autres. Cette contribution nécessite également l’appropriation de la
méthodologie interne de réalisation des EDD, qui constitue un référentiel
commun d’évaluation et d’appréciation du niveau de sûreté.
395
Q.105 - R.22
(iii) Une amélioration notable de la connaissance générale et exhaustive

des ouvrages ;
Les barrages du parc EDF Hydro sont relativement anciens avec une
moyenne d’âge supérieure à 65 ans. Les difficultés de maîtrise docu-
mentaire et de pérennité de l’information pour des ouvrages ayant une
telle durée de vie n’est pas à négliger. Les générations passent, les
structures des entreprises changent. A ce titre, l’EDD constitue une excel-
lente synthèse documentaire et descriptive de la connaissance du barrage.
Les données s’y trouvant ont fait l’objet d‘une critique et d’une validation
formelle. Le travail associé à cette collecte et critique de document est
important. Il est complété par la réalisation d’examens exhaustifs de toutes
les parties du barrage qui viennent compléter et mettre à jour les données
historiques sur la base de l’état réel actuel des ouvrages et matériel.
L’amélioration de la connaissance du barrage concerne aussi une meilleure
compréhension de son schéma fonctionnel avec l’identification de tous les
ouvrages et matériels pouvant impacter la sûreté : cela peut être une poire
de niveau, un câble de fibre optique, un groupe électrogène de secours,
les bouchons béton des galeries de dérivation provisoire ou un fond plein
métallique ;
Fig. 2
Barrages supérieurs à 20 m mis en eau par décennie [3].
Dams above 20 m commissioned by decade [3].
(iv) Une identification et un interclassement des scénarios de défaillance et

des risques sûreté majeurs ;
Pour chaque barrage, l’EDD a permis d’interclasser les risques et d’identifier
les “maillons faibles” des scénarios critiques de défaillance.
L’EDD fournit une “métrique” normalisée permettant d’apprécier, sur une
même échelle des risques aussi différents que le risque sismique, le risque
crue, ou le risque d’érosion interne en situation normale d’exploitation.
Pour certains barrages vannés, les EDD ont pu mettre en évidence
l’importance potentielle de certains modes communs pouvant affecter la
fiabilité d’ouverture des évacuateurs de crue.
396
Q.105 - R.22
(v) La réduction effective de ces risques par la mise œuvre d’actions

éventuellement entreprises à l’issue de l’EDD avec l’adoption d’une posture
préventive.
Pour les EDD de première génération, plus de 1000 mesures ont été iden-
tifiées permettant d’améliorer la connaissance des ouvrages et leur niveau
de risque. Ces mesures couvrent tous les domaines de l’ingénierie (génie
civil, hydraulique, électromécanique, contrôle commande), ainsi que les
modalités d’exploitation, de surveillance et d’auscultation.
Fig. 3
Répartition des mesures de réduction de risque par métier [3].
Risk reduction distribution by field [3].
(vi) Fournir une vision globale du niveau de sécurité d’un grand parc
d’ouvrages.
La vision globale des EDD permet l’inter-comparaison entre barrages et
une priorisation raisonnée des actions à entreprendre à l’échelle d’un grand
parc d’ouvrages de tout type. Les EDD contribuent ainsi à une gestion d’actif
prenant en compte la sûreté.
(vii) Un outil de communication interne et externe.
En interne EDF Hydro, les EDD constituent un outil permettant aux nou-
veaux arrivants de s’approprier rapidement les problématiques et enjeux
sûreté principaux d’un barrage. Il s’agit d’un outil commun de partage et de
compréhension des enjeux et risques des barrages, partagé par tous les
niveaux hiérarchiques de l’entreprise depuis le directeur d’unité jusqu’au
technicien d’exploitation.
La partie descriptive de l’EDD devient la référence documentaire sur les
données de l’ouvrage.
A l’externe, l’EDD est devenue le vecteur d’échange majeur avec
l’Administration pour les questions relatives à la sécurité. L’EDD constitue
le socle sur lesquels se bâtissent les échanges et la communication. Les
EDD permettent de partager les enjeux principaux et l’analyse de risque
du barrage. Vis-à-vis du public, le résumé non technique de l’EDD, livrable
public, permet également à EDF Hydro de communiquer sur les risques
397
Q.105 - R.22
Fig. 4
Répartition des types de barrages supérieurs à 20 m [3].
Dam type distribution above 20 m [3].
afférents à ses barrages, avec les avalisant, les riverains, les collectivités
locales, associations et parties prenantes.
4. MAITRISE DES RISQUES PAR LA GESTION D’ACTIFS - L’EXEMPLE DE

LA CNR : PROTECTIONS EN AVAL DES BARRAGES
Afin de respecter ses obligations règlementaires, en particulier la réalisation

des EDD, et assurer la disponibilité de ses actifs industriels de production électrique,
de navigation, et de ses organes de sécurité, CNR assure régulièrement des anal-
yses de risques et fonctionnelles globales ou particulières, notamment suite à la
survenue d’évènements ou la détection de précurseurs.
Ces analyses sont une des composantes permettant d’alimenter les plans
d’actions et le schéma directeur de maintenance de ses actifs. Les coûts et les
gains obtenus du point de vue technico-économique sont régulièrement évalués
par famille d’actifs ou à travers l’analyse de scénarii de travaux et ceci sous forme
de risques “monétisés”.
Après une présentation globale de la Gestion d’Actifs et sa déclinaison à CNR,

la démarche est illustrée au travers de l’exemple des protections à l’érosion en aval
des barrages.
4.1. QU’EST-CE QUE LA GESTION D’ACTIFS ?
Un “actif” (Asset) représente quelque chose qui a une valeur potentielle ou

réelle pour une organisation. L’Asset Management s’inscrit dans la même logique
de compétitivité des entreprises qu’un système de management des risques.
398
Q.105 - R.22
Etabli en particulier pour les industriels et les gestionnaires d’infrastructures,

l’objectif est de mettre en œuvre des systèmes de gestion des actifs permettant
d’optimiser les budgets d’investissement et d’exploitation afin d’assurer la meilleure
adéquation entre disponibilité et rentabilité des équipements sur leur cycle de vie.
Pour mettre en place un tel système, les normes notamment ISO 55000
pour la mise en œuvre d’un système de gestion d’actifs efficace sont des outils
indispensables.
Dans le cadre de sa gouvernance opérationnelle, CNR s’inscrit actuellement

dans cette démarche de certification ISO et développe une politique de gestion
d’actifs, une méthodologie et des outils en phase avec les domaines d’application
exposés ci-dessus.
Ses lignes directrices sont notamment relatées dans les bulletins 154 et 130
[4] de la CIGB.
4.2. DECLINAISON DE LA GESTION D’ACTIFS A CNR
Parmi les outils de Gestion d’actifs que développe CNR, nous pouvons citer
les deux exemples suivants :
• inventaire de l’ensemble des actifs et mise à jour régulière du registre cor-

respondant. Les éléments y figurant sont aussi bien des éléments d’ordre
technique, concernant la sécurité, et d’ordre financier ;
• analyse de l’impact des projets sur les objectifs de l’entreprise afin d’optimiser
les budgets à court, moyen ou long terme à travers les Plans et Sché-
mas Directeurs de Maintenance des équipements Electro-Mécaniques et des
Ouvrages Génie Civil. Les impacts analysés concernent :
◦ les règlementations en vigueur et les obligations de concessionnaire,
◦ la disponibilité des moyens de production hydroélectriques, des
ouvrages de navigation ainsi que des organes et ouvrages permettant
la gestion des crues.
La Gestion d’Actifs intègre les conclusions des analyses périodiques règle-

mentaires d’intégrité et de suivi du vieillissement des ouvrages, déclinées notam-
ment pour les barrages classés au titre de la règlementation française. Ces analyses
sont présentées à travers les inspections détaillées et les bilans de l’état, du com-
portement et de la conception, produits dans le cadre des Etudes De Dangers
(EDD).
La sécurité des ouvrages et la gestion d’actifs à CNR se déclinent à travers

les grandes lignes organisationnelles et les outils de vision globale suivants :
– gestion des données favorisant l’accès et le partage de la connaissance,
399
Q.105 - R.22
– système de management de la qualité intégrant notamment les référentiels

suivants :
◦ les consignes d’exploitation et de surveillance, les doctrines pour la
gestion des ouvrages,
◦ les procédures en termes de maintenance,
– planification et ordonnancement des projets :
◦ plans pluriannuels et schémas directeurs de surveillance et de main-
tenance, avec utilisation d’outils de Gestion de Maintenance Assistée
par Ordinateur (GMAO),
– évaluation et suivi de la gestion des ouvrages et des incidents en lien avec
la sécurité :
◦ bases de données des rapports périodiques de surveillance,
d’auscultation et de visite, des analyses comportementales des
ouvrages, de l’évaluation de leur état technique,
◦ gestion des données en lien avec la surveillance et l’auscultation des
barrages,
◦ base de données et suivi général du processus de traitement des
évènements,
– développement et maintien des compétences :
◦ plans des formations internes et externes.
4.3. L’EXEMPLE DES PROTECTIONS A L’EROSION EN AVAL DES BARRAGES
4.3.1. Périmètre des actifs concernés
Le parc des barrages de retenue CNR est composé de 19 ouvrages (hors

Génissiat).
Les protections à l’érosion situées en aval des barrages correspondent à

l’ensemble des dispositions constructives mises en place en aval et aux interfaces
avec la partie d’ouvrage rigide (radier). Leur fonction est de protéger la fondation du
radier des effets des sollicitations hydrauliques notamment en période de crue.
Les éléments du registre des actifs composant les dispositions constructives

de protection en aval des barrages CNR au fil de l’eau sont les suivants :
– Protections de fond aménagées :

– Enrochements (blocs) lourds percolés ou non (béton),
– Le cas échéant, petits enrochements autoplaçants,
– Berges :
– Murs en retour,
– Talus avec enrochements (blocs) lourds percolés ou non (béton),
– Fonds naturels situés plus en aval (substratum rocheux et / ou recouvrement
alluvionnaire).
400
Q.105 - R.22
Les actifs concernés qui correspondent à chacune de ces parties ont été
intégrés dans l’outil CNR de gestion des équipements et des ouvrages (GMAO)
qui permet d’assurer le suivi des plans d’actions en termes de maintenance et de
surveillance.
4.3.2. Présentation de la méthode utilisée
Le maintien en état de fonctionnement des protections aval barrage est en

général fortement impactant sur les objectifs de la gestion d’actifs, en particulier
au regard de la disponibilité des organes de sécurité permettant la retenue et/ou le
passage de l’eau notamment en crue.
L’engagement des opérations de maintenance lourde se fait sur la base de

seuils fixés et au-delà desquels les experts jugent que les marges de sécurité en
termes de stabilité du barrage ne sont plus suffisantes.
La méthode d’analyse multicritères utilisée pour le parc des barrages CNR

aboutissant aux plans et à la stratégie de maintenance des protections se décom-
pose dans le détail de la manière suivante :
– évaluation des niveaux d’impacts (forts) sur les objectifs de la Gestion

d’Actifs :
– obligations règlementaires au titre de la sûreté notamment dans le
cas d’évolution bathymétrique observée avec incidence avérée sur la
sûreté : décla ration en Préfecture, détermination et déclinaison d’un
plan d’actions correctives,
– le cas échéant au regard de la règlementation environnementale (ex. :
présence d’une frayère à poissons au droit des travaux),
– en rapport avec les conditions d’exploitation et de maintenance des
organes de sécurité des barrages et des aménagements,
– analyse des conditions hydrologiques et du niveau de sollicitations en crue,
– analyse de l’historique des évènements hydrologiques et des principales
manœuvres depuis la mise en eau,
– évaluation de la remise en cause des protections d’origine – Niveau
“d’Importance des travaux correctifs mis en œuvre”,
– évaluation de l’état technique des protections : état des lieux “0” (diagnostic
2001), analyses bathymétriques par des experts de l’ingénierie intégrée,
mise à jour en 2018. Prochaine mise à jour : a minima au gré des inspections
et analyses règlementaires en cours et à venir dans le cadre des EDD,
– évaluation des risques en termes de vieillissement ou d’évolutions ainsi que
des durées de vie espérées,
– détermination à partir des seuils fixés de la phase de vie des actifs :
– protections en phase de “vie utile” - surveillance courante, entretiens
courant ou spécifique : opérations inscrites dans les plans des travaux
génériques (OPerationnal EXpenditure - OPEX),
401
Q.105 - R.22
– ou protections en phase de “vie mature” nécessitant l’exécution

d’opérations de maintenance lourdes à plus ou moins long terme
(projets identifiés d’investissement : CApital EXpenditure - CAPEX),
– évaluation des montants des principaux travaux historiques, montants
programmés et projections jusqu’en 2041,
– classement des travaux en générique ou en projet identifié et leur inter-
classement sur du court-moyen terme à travers les Plans Pluriannuels de
Maintenance à 3-5 ans,
– évaluation du budget annuel moyen sur le parc de barrages, vérification
régulière des indicateurs en termes de Gestion d’Actifs et de Politique de
Maintenance des Ouvrages, à travers le Schéma Directeur à 20 ans (♦ 2041).
Ce schéma mis à jour en 2018 l’est tous les 5 ans (♦ 2023). Les indicateurs
sont mis à jour suivant une fréquence annuelle à quinquennale.
4.3.3. Exemple des protections en aval du barrage du POUZIN
Cet exemple de gestion d’actifs est intéressant au regard du maintien en

exploitation des protections avals de ce barrage. Les protections ont un niveau
d’impact fort sur les objectifs de la Gestion d’Actifs.
Les sollicitations hydrauliques au droit de ce barrage sont maximales en crue

pour une période de retour faible ou lors de manœuvres exceptionnelles conduisant
à déverser avec une vanne ouverte sur un tapis d’eau de faible épaisseur en
particulier pour chasser les bois.
Du fait du contexte hydrologique, des conditions d’exploitation singulières

exposées ci-dessus et des dispositions constructives pour les protections aval ren-
forcées à plusieurs reprises depuis la mise en eau du barrage, le diagnostic réalisé
par les experts CNR en 2020 a permis de statuer sur la nécessité d’engager une
opération de reprise des protections aval.
Du point de vue de la Gestion d’Actifs, ce diagnostic nous permet de nous

projeter sur les travaux à envisager. En l’état actuel il apparaît qu’aucune opération
de maintenance lourde (nouvelle reprise des enrochements de protection) ne sera
nécessaire à court-moyen terme. Ceci nous permet aujourd’hui d’envisager sere-
inement l’éventualité de travaux lourds à plus ou moins long terme jusqu’en 2041,
projection inscrite dans le Schéma Directeur de Maintenance à 20 ans du fait du
contexte sensible exposé sur ce barrage.
Les inspections et les analyses règlementaires qui seront notamment réal-

isées dans le cadre de la prochaine EDD programmée en 2025 permettront
d’actualiser l’évaluation de l’état technique des protections et statuera sur le com-
portement du barrage et les risques associés en termes de sûreté, le cas échéant
les études et travaux à envisager.
402
Q.105 - R.22
Par ailleurs CNR a mis en place des indicateurs permettant le contrôle des
objectifs en termes de gestion des actifs et de maintenance des ouvrages génie civil,
notamment pour les barrages dont les principales opérations pour les années à venir
seront les reprises des protections aval. Le cas du barrage du POUZIN s’inscrit dans
cette démarche de vision et de contrôle global de la stratégie d’entreprise.
5. CONCLUSION
Ce rapport met en avant quelques facteurs de réussite pour une gestion

efficace de la sécurité des barrages hydroélectriques et termes de politique,
d’organisation, de revues périodiques de sécurité, de formation et de retour
d’expérience. Il indique comment les Etudes De Dangers requises par la règle-
mentation française s’avèrent être de précieux outils pour améliorer la maitrise des
risques. Il présente aussi un exemple de déclinaison opérationnelle d’une politique
de sécurité dans la gestion et priorisation d’opérations de maintenance.
Ces éléments, qui ont pu montrer leur efficacité chez ces opérateurs
français, doivent cependant être pris avec modestie et dans une démarche assidue
d’amélioration continue car le retour d’expérience mondial montre encore et tou-
jours qu’aucun pays n’est à l’abri de subir un accident grave sur ses installations
hydroélectriques.
REFERENCES
[1] CIGB/ICOLD (2017). Gestion de la sécurité des barrages en exploitation.

Bulletin 154.
[2] Reverchon B. et al. (2015). Conception, exploitation et rénovation des évac-

uateurs de crue vannés des barrages exploités par EDF. 25ème congrès des
grands barrages CIGB 2015, Stavanger.
[3] Guilloteau T. et al. (2016). Contribution des revues de sûreté à la sécurité des
barrages. Colloque CFBR : “Sûreté des barrages et enjeux”, 23 – 24 novembre
2016.
[4] CIGB/ICOLD (2005). Evaluation du risque dans la gestion de la sécurité du

barrage. Examen des bénéfices, des méthodes et des dernières applications.
Bulletin 130.
[5] Peyras L. et al (2021). Evaluation de la sûreté des barrages en France. Retour

d’expérience et développement méthodologique. 27ème congrès des grands
barrages CIGB 2021, Marseille. Q105.
403
Q.105 - R.23
-------
-------
EVOLUTION OF DAM SAFETY AND EMERGENCY MANAGEMENT PRACTICE

FOR TRANSBOUNDARY DAMS – A GLOBAL PERSPECTIVE WITH A
NEPALESE CONTEXT (∗)
M. Acharya PMP, P.Eng.

Senior Dam Safety Engineer,
Environment and Parks
C. Richard DONNELLY P.Eng.

Global Principal Consultant and Geotechnical Specialist – Dams and Waterpower,
HATCH LTD.
Przemyslaw A. ZIELINSKI, PhD, P.Eng.

Principal,
HYDROSMS INC.
CANADA
SUMMARY
While many countries have transboundary agreements to promote coopera-

tion around the management of transboundary water resources, these agreements
often do not adequately address dam safety considerations. Recent dam incidents
around the world have highlighted the importance of maintaining a robust dam
safety and emergency management practice and the need for adopting proactive
measures to implement the lessons learned from these incidents. Unfortunately, in
many parts of the world, the unique challenges that transboundary dams present
are largely overlooked. This paper, outlines current trends and best practices in
dam safety and emergency management, emphasizing the need for an integrated
regional approach, sound legal and institutional arrangements, collaborative and
∗ Évolution de la sécurité des barrages et des pratiques de gestion des situations d’urgence
pour les barrages transfrontaliers – une perspective mondiale dans un contexte népalais
404 DOI: 10.1201/9781003211808-69

Q.105 - R.23
diplomatic frameworks, and effective communication protocols for transboundary

dams to manage the residual risks and public safety. Special attention is given to
the issues around transboundary dams along the Indo-Nepal border.
RÉSUMÉ
Alors que de nombreux pays ont des accords transfrontaliers pour promou-
voir la coopération autour de la gestion des ressources en eau transfrontalières,
ces accords ne traitent souvent pas de manière adéquate les considérations de
sécurité des barrages. Les récents incidents de barrages dans le monde ont mis
en évidence l’importance de maintenir une solide pratique de sécurité des barrages
et de gestion des urgences et la nécessité d’adopter des mesures proactives pour
mettre en œuvre les leçons tirées de ces incidents. Malheureusement, dans de
nombreuses régions du monde, les défis uniques que présentent les barrages trans-
frontaliers sont largement négligés. Ce document décrit les tendances actuelles et
les meilleures pratiques en matière de sécurité des barrages et de gestion des
urgences, en soulignant la nécessité d’une approche régionale, d’arrangements
juridiques et institutionnels solides, de cadres collaboratifs et diplomatiques et de
protocoles de communication efficaces pour les barrages transfrontaliers afin de
gérer les risques résiduels et la sécurité publique. Une attention particulière est
accordée aux problèmes liés aux barrages transfrontaliers le long de la frontière
indo-népalaise.
1. INTRODUCTION
According to ICOLD Bulletin 167 [17] almost half of the world lies within river
basins that cover territory in two or more countries. When these transboundary
rivers are dammed for energy or irrigation purposes, often conflicts arise [13]. As
shown in Fig. 1 there are thousands of dams under construction or in the planning
phase worldwide and several hundred located in transboundary river basins. In the
Ganges-Brahmaputra-Meghna-Nepal basin alone, in South-East Asia, there are 183
such projects [30].
The benefits dams can provide has resulted in expanded hydropower devel-
opment in many countries [1] that is also resulting in increasing numbers of
trans-boundary dams and the conflicts that this can cause. Many nations have
shared rivers and transboundary agreements. However, the requirements for robust
dam safety management practices are largely overlooked. This issue is more
pronounced in the Himalayan regions such as Nepal, India, China, Pakistan,
Bangladesh, Bhutan, where natural hazards are more prevalent and dam safety
management practices often relatively weak. For example, in Nepal, many river
system originate in Tibet, China, crosses the Himalayan Mountains in Nepal. Most
of them flow to the Ganges in India with some flowing into Bangladesh and eventually
draining into the Bay of Bengal. Although there are four transboundary water
405
Q.105 - R.23
Fig. 1
A Global Boom in Dam construction [13].
Un boom mondial dans la construction de barrages [13].
management treaties between Nepal and India on the larger river basins [20], these
agreements rarely address safety management issues focusing instead on bi-lateral
agreements for the use of the transboundary waters. [14] notes that the agreements
often do not fully deal with the legal, socio-economic, and political ramifications
involved in the shared water use. Because dam safety and public safety issues are
typically a much lower priority when the agreements were formulated, issues asso-
ciated with the potential flooding, public safety and other impacts are often not well
addressed or, in some cases even considered.
The development and implementation of a modern dam safety management

system that is specifically tailored to the needs of the transboundary dams, should
include a legal and institutional framework, regional approach, integrated river
basin management and hazard assessments, consideration of societal risk toler-
ance, risk-informed decision making and improved risk communication throughout
a project’s life cycle. Increased cooperation, coordination, improved diplomatic com-
munication, and the unobstructed exchange of information are key factors needed
to ensure the safe and sustainable development and operations of transboundary
dams, and the safety of the communities and infrastructure upstream and down-
stream of such dams. Collaboratively facing common external threats can lead
to enhanced financial results, accelerated economic development and improved
political stability [30].
This paper presents some thoughts and recommendations on how owners

of transboundary dams should work with their neighboring counterparts to further
strengthen their capacity to develop the nation’s dam safety management practices
and ensure they are in line with international best practices. Special attention is
given to the issues around transboundary dams along the Indo-Nepal border with
some recommendations provided for the path forward.
406
Q.105 - R.23
2. TRANSBOUNDARY CONSIDERATIONS FOR DAM SAFETY

MANAGEMENT
Dams located in internationally shared transboundary river basins or sub-

national jurisdictions within a country can create complex interdependencies and
present a unique set of considerations to the safety of dams and downstream com-
munities. These include different and, sometimes, conflicting legal, cultural, and
political issues [30]. [13] notes that transboundary dam projects should be consid-
ered as a wider institutional setting by focusing on historical perspectives, rules
of water allocation, existing and evolving intuitional governance architecture in the
basin. As such, special provisions related to the safety of the transboundary dams
and the public need to be established at a regional level in order to reduce potential
risks and to realize mutual benefits. This is even more important where there is a
possibility of cascade effects of a dam failure.
2.1. PRINCIPLES FOR SHARED RIVER BASIN AGREEMENTS AND TRANSBOUNDARY

DAMS
The overall objective of international conventions/protocols/treaties on shared

watercourses is to foster closer co-operation for judicious, sustainable and coordi-
nated management, protection and utilization of shared watercourses [17]. Although
the international and regional treaties may differ substantially in their details, there
are some common principles around them as noted below:
• Equitable and reasonable utilization and participation.

• Obligation not to cause significant harm.
• Co-operation.
• Exchange of data and information.
• Priority of different uses.
• Planned measures.
• Protection and preservation of ecosystems.
• Dealing with harmful conditions and emergency situations.
• Settlement of disputes.
There is no doubt that well-thought-out international agreements can provide

a mechanism to facilitate improved cooperation around the safety of dams and
promote joint decision making on the allocation of resources to address the most
pressing dam safety challenges.
2.2. THE NEED FOR LEGAL CONSIDERATIONS FOR TRANSBOUNDARY DAMS
A significant amount of work has been done by the international legal commu-
nity to establish and promote laws that help to facilitate water resource management
407
Q.105 - R.23
for internationally shared rivers [17]. There is, however, no universally accepted
international law that is applicable to all shared rivers, and very little work has been
done to deal with the issues specific to the safe management of transboundary
dams. The issue is complex, because several large transboundary river basins
have riparian states with different legal jurisdictions. Under international law, the
interests of all riparian states need to be taken into account and coordination mech-
anisms put in place, including institutions and alignment in the equivalence of the
national legal frameworks, to ensure the safety of all dams and their downstream
communities [30].
The complexity of dam safety issues for transboundary dams is discussed in

detail in [32]. It was observed that the International Law Association proposed the
“Guidelines for the Establishment of an International Water Resources Administra-
tion” as early as in 1976. The Guidelines defined the elements to be considered
when establishing an international water resources administration and address the
form, constitution and duration, procedures for decision-making, legal status, ter-
ritorial competence, functions and powers, objects and purposes, financial and
economic matters and dispute settlement mechanisms required for an international
water resources institution. The functions considered in the Guidelines include the
following:
• Advisory functions that also extend to consultative, coordinating, policy-

making functions,
• Executive functions which may include carrying out of studies, exploration,
investigation and surveys, preparation of feasibility reports, inspection and
control, construction, operation, maintenance or financing,
• Regulatory functions which may include the implementation of the decisions
of the administration, as well as law making, in which decisions may take effect
directly or after acceptance by Member States,
• Judicial functions which may include arbitration, mediation, fact-finding or
final dispute settlement; and settlement of disputes arising out from the
interpretation and implementation of the founding Treaty.
There are many examples around the world of transboundary basins that span
countries with different legal systems. For example, the Saint Lawrence River basin,
includes common law jurisdictions in Canada and the United States, and the Civil
Law system of the Province of Quebec (in Canada). In the Middle East, the Tigris
River flows through a range of civil, mixed, and religious legal systems as it moves
through Turkey, Iran, and Iraq, among others. In Asia, the Brahmaputra - Yarlung
Tsangpo River shared between Bangladesh, China, and India includes Civil Law
systems in China and Common Law systems in Bangladesh and India [30].
These different legal regimes, and their associated geopolitical issues, add a
layer of additional complexity to international efforts focused on improving cooper-
ation in transboundary dam safety and emergency management issues [30]. [30]
further notes that the coexistence of different legal regimes for dam safety assur-
ance among riparian states or subnational jurisdictions can create new hazards if
408
Q.105 - R.23
there is not a consistent level of care in managing these facilities. Very few of the
national legal frameworks for dam safety and very few of the transboundary river
basin agreements include explicit provisions for addressing obligations to the safety
of the dams and the surrounding communities.
Construction and management of dams located in transboundary river basins

are not always regulated by one single legal system. They are often guided by a
set of binding and non-binding standards, norms, principles, guidelines, recom-
mendations and customary rules [13]. When riparian or subnational states in the
same transboundary river basin have different legal systems, portfolio characteris-
tics or institutional capacity, efforts are needed to ensure a minimum level of safety
assurance across the basin [30].
2.3. THE NEED FOR AN INTEGRATED REGIONAL APPROACH FOR TRANSBOUNDARY

DAMS
Integrated water resources management has become an accepted interna-

tional approach, promoting co-operation among stakeholders within international
river basins. The United Nations encouraged the development of integrated water
resources management and water efficiency in its Millennium Development Goals
[17]. None of these plans explicitly discussing the safety considerations of dams
located on shared river basins. However, [31] promotes a basin-wide perspective
for open discussion of the issues, negotiation on sharing the benefits, and the miti-
gation of any adverse impacts. A regional dam safety program can play a meaningful
role in promoting improved safety of dams and downstream communities, while also
contributing to broader co-operative ventures [28]. Such programs can help facilitate
equivalence across frameworks within the basin and strengthen the regional network
of technical professionals. Regional dam safety units or peer-to-peer arrangements
can also be useful in addressing capacity constraints by capitalizing on economies
of scale, expertise, and portfolio differences. These regional programs need to facil-
itate the joint development of tools that contribute to improved safety of dams such
as flood mapping simulations, dam break analyses, development of emergency
action plans, and community safety nets to be deployed in case of dam failure or
misoperation.
Among many world-wide examples, Spain and Portugal signed The Albufeira
Convention in 1998 to govern cooperation around five shared river basins, estab-
lishing the flow regime between both countries and containing a specific provision
relating to Dam Safety [30]. It states, “The parties will jointly develop specific pro-
grams about the safety of hydraulic infrastructure and assessment of actual and
potential risks that could lead to significant adverse effects upon any of the riparian’s
in case of dam break or serious accident.” The Convention also requires transbound-
ary impact assessments for those projects or activities less than 100 kilometers
upstream or downstream of the border that could significantly alter the flow regime
or cause the discharge of pollutants [30]. More recently (2018), in an attempt to
409
Q.105 - R.23
improve the transboundary cooperation, a MOU between the Ministry of Energy,

Water Resources and Irrigation, Nepal and the National Administration of Energy of
the People’s Republic of China on Energy Cooperation was signed in Beijing. One
of the areas of cooperation is to undertake river basin planning and hydropower
resource investigation in Nepal in order to improve the water resources manage-
ment and sustainable development of hydropower of Nepal. A Joint Implementation
Mechanism (JIM) to execute, review and monitor implementation has been functional
under the MOU.
2.4. THE NEED FOR INSTITUTIONAL ARRANGEMENTS FOR TRANSBOUNDARY DAMS
According to Bulletin 167 [17], an institutional arrangement is normally required

to achieve effective co-operation in international water matters. These arrangements
are normally structured in terms of international treaties or agreements between two
or more countries for one or more of the following objectives:
1. Gathering and sharing of information.

2. Undertaking joint studies.
3. Development project implementation and operations.
4. Development of operational and management approaches for the specific
resources.
5. Management of the common water resources.
Institutional arrangements need to consider the specific regional context to

best facilitate co-operation around the improved safety of dams. These important
considerations may include the size of the portfolio in the different riparian states or
subnational jurisdictions, the population distribution and growth projections, the level
of human and financial capacity, as well as the prevailing national arrangements.
Where basin organizations already exist, they can play an important regulatory
function in facilitating basin level dam safety assurance, particularly where exec-
utive functions reside within national jurisdictions or special purpose vehicles for
infrastructure ownership and operations [30]. Currently there are over 80 interna-
tional commissions/committees working on transboundary river basins [17] and [30].
Established in 2014, the Zambezi River Dam Operators Joint Operations Technical
Committee (JOTC) is one example of such a committee [17] and [30]. It is designed
to promote equitable and reasonable utilization, as well as the efficient management
and sustainable development of the basin’s water resources. The JOTC developed
a framework for collaboration, and to facilitate the exchange of hydrometeorologi-
cal and dam related information to improve and inform management of the water
resource usage in the basin. For example, a joint dam break analysis was performed
to provide a comprehensive assessment of the potential consequences of a partial
or full dam breach at one or several dams in the Zambezi and Kafue rivers [30].
Transboundary organizations responsible for the ownership and operations

of large dams are typically established through international agreements. As such,
410
Q.105 - R.23
they are typically recognized as international public law corporations subject to the
founding agreements and established governing bodies with their own legal per-
sonality different from that of the individual shareholders. Therefore, they are not
usually subject to the national dam safety regimes of the states that might own
them or other riparian or subnational states within the transboundary basin. In the
absence of prevailing legislative provisions, such organizations typically establish
their own dam safety requirements and standards with appropriate references to
other jurisdictions [30].
2.5. THE NEED FOR DISPUTE RESOLUTION MECHANISM
Disputes and conflicts can be expected, particularly if international best prac-

tices are not followed. [31] promotes the development of a procedure to ensure
equitable and reasonable utilization of the water resource, no significant harm,
prior notification, impact assessment, and dispute resolution. This builds on the
provision of the UN Convention on the Law of the Non-Navigational Uses of Inter-
national Watercourses and other agreements. Dispute with respect to dam safety
and emergency management for transboundary dams that cannot be resolved by
diplomatic dialogue and negotiation are not uncommon. Therefore, the dispute res-
olution mechanism becomes an important element of trans-boundary projects. This
could include 1) mediation by a third party, 2) arbitration, or 3) reference to the Inter-
national Court of Justice or a Regional forum [17]. In any agreement, the dispute
resolution and cost-sharing mechanisms and timeframes shall be clearly included
on all transboundary agreements.
3. TRANSBOUNDARY DAMS IN NEPAL
As illustrated in Fig. 2, many of the high dams proposed or planned in Nepal

(e.g., Budhi Gandaki, Arun III, Pancheshwar, Upper Karnali, West Seti, Koshi High
Dam) are located in internationally shared transboundary river basins or sub-national
jurisdictions within the nation. All of Nepal’s dams are located in a seismic region
of the world where this and other natural hazards can present a significant risk [1].
The impact of the failure of major dams may extend beyond national borders to
India and Bangladesh. Also, as shown in Fig. 2, there are over dozens of dams,
barrages, and embankments located in the transboundary river basins along the
Indo-Nepal border. Every year, bordering peoples of Nepal and India suffer from
monsoon floods. For example, in 2017, hundreds of people’s lives and many millions
of properties were lost [3]. Despite this reality there are currently no established
approaches to transboundary water issues in Nepal. Prior notification and approval
for hydropower projects from upstream and downstream neighbors are not generally
practiced and the exchange of data between China and Nepal and between India
and Nepal is generally lacking.
411
Q.105 - R.23
Fig. 2
Dams, Barrages, and Embankment Constructed along the Indo-Nepal Border.
Barrages, barrages et remblais construits le long de la frontière indo-népalaise.
To reduce the flood risk created by some of these transboundary dams,

embankments are also constructed along the border, which in some instances,
has led to flooding, leading both governments to blame each other for the dam-
ages [20]. In Nepal, formal dam safety management programs designed to address
the full range of potential risks that these existing and new dams may pose do not
yet exist [1], resulting in inconsistent dam safety design and management practices
and, therefore, increased transboundary risks. Sub-optimal operational practices, as
noted in Koshi Barrage in Section 3.1, impact people and property on both sides of
the border. These types of issues need to be considered while developing the safety
management policy for the dams located in internationally shared transboundary
river basins or sub-national jurisdictions.
3.1. TRANSBOUNDARY AGREEMENTS, ISSUES AND CHALLENGES BETWEEN INDIA

AND NEPAL
Located in the heart of Himalayas with water as the greatest natural resource
of Nepal, significant hydropower development is under construction in recent years,
or in the planning stage (Fig. 3). Given that Nepal has more than 6 000 rivers and
rivulets that flow into the Ganges river system [20], hydropower projects in Nepal
can have impacts beyond its national borders.
There are four treaties between Nepal and India 1) The Sharada Agreement
(1927) on Mahakali River; 2) The Koshi Agreement (1954) on Koshi River 3) The
412
Q.105 - R.23
Fig. 3
Major Hydropower Projects in Nepal [25].
Grands projets hydroélectriques au Népal [25].
Gandak Agreement (1959), on Gandak River, and 4) The Mahakali Treaty of 1996
on Mahakali River [20]. The treaties mostly have bi-lateral agreements without any
regional coordination. In South Asia where rivers cross many countries, integrated
river basin planning, and regional approaches would be better than just the bilateral
agreements for the management of transboundary dams to incorporate broader dam
safety and emergency management issues and to maximize mutual benefits.
The Koshi transboundary water management agreement is an example that

demonstrates the dam safety and emergency management challenges in the region.
The agreement covers a 56-gate dam barrage known as the Koshi Barrage. The
dam was primarily conceived by India to meet its irrigation requirements and solve
flooding problems [20]. It was constructed inside Nepalese territory, but construction,
control, and maintenance are governed by India with Nepal receiving an irrigation
canal and a bridge across the East-West highway as an economic concession. This
multi-purpose project has suffered from issues such as the inefficient implementation
of treaty terms and operational practices that result in the flooding of a large area of
Nepalese land annually. As a result, this agreement is frequently criticized in Nepal
for the perceived disproportionate benefits that India receives, and the permanent
loss of large tracts of fertile land in Nepal that was flooded by the reservoir. While the
treaty was amended, the sense of grievance has continued with the perception that
the ongoing impacts have not been fairly addressed and compensated [20]. Such
ongoing impacts point to the lack of consistent practice in dam safety and emergency
management when formulating the agreements reinforcing the need for integrated
river basin planning, regional cooperation and development for transboundary water
413
Q.105 - R.23
management in South Asia [8]. Unfortunately, the special considerations required

for the safe management of transboundary dams are typically not considered when
formulating these agreements.
3.2. RISKS OF UPSTREAM FLOODING FOR DAMS ALONG INDO-NEPAL

TRANSBOUNDARY RIVER BASINS
Upstream flooding along the transboundary water basins of the Indo-Nepal

boarder is also a significant concern, generally as a result of sub-optimal project
location, design issues and poor operational practices [2]. To minimize the impact
of these flooding events, NGOs, development organizations and the Government
of Nepal have been working to develop a national emergency preparedness and
response system as well as first responder training packages for the use of emer-
gency response teams after natural disasters and other major emergencies [2].
However, these emergency management systems and training packages do not
specifically address emergencies related to dams and hydropower projects with
transboundary implications. Under international law, the interests of all riparian
states need to be taken into account and coordination mechanisms put in place
to ensure the safety of the public. Riparian and subnational states need to opti-
mize the location of the project, enhance the exchange of information relating to
operations and emergency planning in order to reduce the potential for harm to all
stakeholders in the basin [30]. In North America, for example, all affected parties are
proactively engaged to carefully review a project’s potential risk before it is approved.
This process helps to ensure that public concerns are addressed and that measures
are put in place to minimize the risks. Unfortunately, these best practices are not
universal.
3.3. CONSIDERATION OF THE EFFECTS OF CLIMATE CHANGE
Climate change science indicates that there is a strong likelihood of an

increased occurrence of severe events over the next century [26]. The increased
severity and frequency of storms increases the potential for heavy rainfalls that can
result in overtopping induced failures affecting people and increased potential for
cascade failures [11]. Therefore, the development of a reliable means to estimate
these impacts should, in the long-term, represent a critical component of planning
and management of transboundary infrastructure. However, the present projected
impacts of climate change on water resources and flows are highly uncertain and the
extent of this uncertainty remains still unknown. [19] points out that for this reason,
exact information on the rate of future changes and potential change in magnitudes
of hydrological hazards remain and for some time will remain unavailable. However,
this does not mean that nothing can be done in addressing the safety of dams in a
changing climate. A realistic and practical approach proposed in [21] is based on
414
Q.105 - R.23
a set of comprehensive adaptation procedures which do not need precise, quan-

titative estimates of relevant climate and hydrological variables. Such procedures
would include adaptive and resilient dam design and dam system management,
adaptive operation of dams and enhancement of downstream community resilience
[21] and [26].
The Himalayan belt, including Nepal, is considered as one of the most vulnera-
ble places in the world facing the effects of climate change [2], and the nation needs
to consider it seriously in its water resource development plan. As noted in [6], the
impact of climate change includes projected intensification of monsoons, acceler-
ated retreat of glaciers, flow unpredictability in the glacier-fed rivers, and increased
potential for debris flows, all of which pose increased risk to the future of hydropower
and related infrastructures in Nepal. While any one country can invest its resources
to assess as to how the climate change can impact the nation’s water resources,
it is the coordinated approach, collaborative initiatives, and partnership efforts at a
regional level across the Himalayas, and globally that makes the real difference to
combat the effect of climate change. Transboundary coordination is, or should be
part of the solution to manage this evolving risk.
3.4. SEDIMENTATION CONSIDERATIONS
In Nepal, natural hazards such as earthquake, landslides, heavy monsoon,

debris flows, glacial lake outburst floods and landslide dam outburst floods can
yield a significant amount of sediment in transboundary rivers since many of the
rivers originate on steep mountain slopes and eventually flow into the Ganges
river system in India. [25] identified sedimentation control and management as
one of the most overlooked areas in the planning and design of hydropower
projects in Nepal. As shown in Fig. 4, developing regions of the world (e.g.
China, South America, Northern India and Nepal), that stand to benefit most from
the production of hydroelectricity, are often those that have the highest sediment
yields [15].
Sedimentation can alter reservoir routings, complicate the management of

seasonal inflows, reduce spillway discharge capacity, alter reservoir ice formation
and increase loads on the dam. For example, at the Koshi Barrage, there has been
more than five meters of deposition [10] which led to the overtopping and failure
of one of the irrigation canals, affecting thousands of people both upstream and
downstream in both countries [20]. In the case of Upper Arun hydropower project, the
challenge is the planning and management of flows, sedimentation and geohazards
for the entire river system from the headwaters in Tibet, through Nepal, northern
India into Bangladesh and ultimately into the Bay of Bengal [6]. Therefore, estimating
and managing the effects of sedimentation on a regional basis is an important, but
often overlooked part of the planning process for transboundary dams. Practical
guidance on these matters can be found in [4], [5], [18], [24] and [29].
415
Q.105 - R.23
Fig. 4
Comparison of Hydroelectric Potential and Sediment Production [15].
Comparaison du potentiel hydroélectrique et de la production de sédiments [15].
3.5. INTEGRATED HAZARD MANAGEMENT CONSIDERATIONS
Many rivers in the Himalayan region cross numerous countries, serving impor-
tant functions for various purposes for downstream users in Nepal, Indian and
Bangladesh [6]. Considering the elevated levels of regional geohazards and sedi-
mentation challenges, the need for integrated river basin planning and management
is essential for sustainable management of available water resource.
In the Himalayan region, the commonly studied natural hazards, flood and
earthquake, are typically not the most serious hazard with respect to the safety of
dams. During the 2015 Gorkha Earthquake and aftershocks it was debris flows and
landslides that caused the most damage, resulted in failure at some dam, such as
the Upper Bhote Kopshi headworks [12]. A survey of small hydroelectric facilities
conducted after the event confirmed that the earthquake was not the most significant
hazard. Only 13% of the facilities surveyed showed structural damage while almost
half suffered damage from landslides.
In assessing the risks that natural hazards present to transboundary dams in

this region, it is important to recognize that these events can often occur in com-
bination [27] potentially impacting multiple jurisdictions. In the case of the severe
damage to the Upper Bhote Koshi facility, a small Glacial Lake Outburst Flood
(GLOF) occurred in northern Nepal, as a result of a cloudburst in the southern part of
Tibet [6]. The GLOF, then, triggered a series of landslides and Landslide Dam Out-
burst Flood (LDOF) in Nepal that greatly increased the destructive power of the event
416
Q.105 - R.23
[12] to the Bhote Koshi facility. As discussed in [27], a process to assess the potential
for natural hazards to occur and the potential for them to occur in combination is one
of the most important considerations for dams in Nepal and the Himalayan region in
general. This points to the need for integrated hazard management considerations
for the region.
4. CURRENT DAM SAFETY MANAGEMENT PRACTICE IN NEPAL
While several treaties exist around the management of transboundary water

resources, they rarely address dam safety considerations. Currently, Nepal does
not have a national dam safety regulation or integrated emergency management
programs, although the dam safety hazards that exist in Nepal are unique and sig-
nificant, compared to many other jurisdictions around the world. Instead, most of
the major dams in Nepal are designed based on the international standards or stan-
dards of the consultant’s country of origin. As such, the hazards unique to Nepal may
not be fully accounted for and designs can vary significantly depending on the judg-
ment and experience of the practitioner, the standards with which they are familiar,
and the risk tolerance of the owner of the facility [1]. Risks are further compounded
by the aging of dams, inadequate maintenance, climate change and inconsistent
transboundary practices.
4.1. EMERGENCY MANAGEMENT PROGRAMS FOR TRANSBOUNDARY DAMS
Recent dam safety incidents, the evolution of dam safety management prac-
tice and its current global trend is discussed in [1] and [2]. Despite continuous
gain of knowledge and experience in design, construction, and operation of dams,
incidents and failures continue to occur. Aging of infrastructures, extreme events,
inadequate maintenance, improper standard of care, insufficient level of safety mea-
sures, increase vulnerability of infrastructures. Risk is continuously increasing due
to factors such as age, the impact of climate change, downstream development, and
inconsistent transboundary practices in dam safety.
A well-structured dam safety management program reduces the likelihood of

dam safety incidents and their consequences. The risks can be minimized with
sound dam safety programs but cannot be completely eliminated. Therefore, an
effective emergency management plan should be an important element of a sound
dam safety management program to reduce residual risks of transboundary dams.
An emergency management program that is shared with first responders and trans-
boundary stakeholders [7] helps to ensure that practicable measures are taken
to minimize life safety hazards from reasonably foreseeable incidents [16]. These
programs represent a low cost, but very effective, strategy to manage residual dam
safety risks for all types of dams.
417
Q.105 - R.23
Collaboration between private-sector business continuity programs, national

critical infra-structure protection programs, transboundary partners and community
emergency management officials, and the community itself, yield lasting benefits
for managing the residual risk posed by dams by enhancing resiliency of potentially
impacted stakeholders and communities. In the case of dams in Nepal, or along the
transboundary rivers basins, community, local leaders, and transboundary partners
will have an important role in emergency preparedness, information sharing, and the
coordination and implementation of both Emergency Preparedness and Response
Plans.
Elements of sound dam safety management system (e.g. policy, planning,

implementation, checking and reviewing, corrective action, reporting, continuous
improvement) and emergency management programs (e.g. prevention and mitiga-
tion, preparedness, response, recovery, implementation of lessons learned) that
might be applicable to Nepalese context are further discussed in [1] and [2].
4.2. THE NEED FOR IMPROVED DAM SAFETY RISK COMMUNICATION FOR
TRANSBOUNDARY DAMS
In [9], risk communication was identified as a major gap in modern dam safety
management practice. It was noted that the recent high-profile dam failures have
demonstrated the need for clearly communicating dam safety risks, which can only
be effectively achieved by a collaborative approach involving all stakeholders. Risk
communication is even more important for transboundary dams with potential cas-
cade effects beyond national borders. To manage risk effectively, all stakeholders
must be included in risk management and decision-making processes during the
planning, design, construction, and subsequent operational phases of the project.
Risk communication needs to include the possible emergencies that could occur,
and the mitigation strategies being employed [22]. Involving stakeholders early and
often helps to establish and maintain the owners’ social license to operate the facility
and gets ahead of potentially misleading statements and perceptions that can be
communicated through social media.
In 2018, the Xe-Pian Xe-Nam dam collapse resulted in dozens of fatalities

and over 8000 homeless across Laos and Cambodia [2]. Although the reasons for
dam failure was a result of a combination of factors such as lack of treatment mea-
sures of the foundation, inadequate follow-up during first filling, poor site access
during emergency [28], the impacts of the failure in the two countries were exac-
erbated because of the lack of sound emergency management programs, weak
transboundary emergency coordination, and poor risk communication. Improved
risk communication alone may have saved many lives [2].
In the case of transboundary dams, institutional arrangements need to con-

sider the regional context to best facilitate cooperation around the improved safety
of dams and both upstream and downstream communities. Considerations include
418
Q.105 - R.23
the size of the portfolio in the different riparian states or subnational jurisdictions,
population distribution and growth projections, the level of human and financial
capacity, as well as the prevailing national arrangements for ensuring the safety
of dams and both upstream and downstream communities. Where basin organi-
zations exist, they can play an important regulatory function in facilitating basin
level dam safety assurance, particularly where executive functions reside within
national jurisdictions or special purpose vehicles for infrastructure ownership and
operations [30].
5. CONCLUSIONS
Construction and management of dams located in transboundary river basins

are often guided by a set of binding and non-binding standards, norms, principles,
guidelines, recommendations, historical and cultural perspectives, and customary
rules. Typically, ensuring the safety of dams and downstream communities within
the context of transboundary river basins requires a unique set of considerations
that have often been not well accounted for. The number of planned dam projects,
increasing uncertain climatic conditions, and changing patterns of human settle-
ment, demonstrate that this will be an increasingly important issue. Transboundary
best practices such as integrated river basin planning and management, coopera-
tion in good faith, regular exchange of data and information is more important than
ever before for the mutual benefit, sovereign equality, territorial peace and integrity.
Recent dam incidents have highlighted the importance of robust dam safety and
emergency management practice and the need to proactively implement measures
learned from these incidents. Unfortunately, in many parts of the world, the unique
challenges that transboundary dams present are largely overlooked. Therefore, the
implementation of modern dam safety management programs, specifically tailored
to the needs of the transboundary dams that include regionally integrated hazard
assessments, societal risk tolerance, risk-informed decision making, and improved
risk communication, represents a practical vehicle to minimize the likelihood and
consequence of dam safety incidents now and into the future.
A robust dam safety program specifically tailored to the needs of the trans-
boundary dams will increase public awareness and safety and would strengthen
societal and professional confidence that the inherent risks of these large engineered
structures are appropriately addressed and managed during all phases of their life
cycle. Such programs will improve sub-national, national and international coop-
eration as well as awareness on dam safety and related issues; improve national
legislation and regulatory frameworks related to dam safety; increase the technical
and legal capacity of experts and officials on dam safety issues and improve safety
and transboundary cooperation on individual dams. This will reduce the risk related
to major dams located in transboundary water basins and, ultimately, help to create
an attractive investment climate in the Himalayan region for national and international
investors on major hydropower development projects safely and sustainably.
419
Q.105 - R.23
6. RECOMMENDATIONS
Nepal would significantly benefit from:
1. Supporting the development and implementation of a modern dam safety

program, regulations, and policy tailored to the unique needs of Nepal.
2. Developing a transboundary dam safety management policy that includes:
(a) A legal and institutional framework ensuring the safe management of
transboundary dams.
(b) transboundary considerations and regional approaches such as joint
studies on assessment of failure risks and mitigation measures.
(c) Institutional arrangements that establish mechanisms to share infor-
mation such as internal and external emergency plans,
(d) Funding for universities, research institutes and professional orga-
nizations to promote research and development, capacity building,
and information sharing for the management of issues on dams on
transboundary water basins.
3. Establishment of consistent dam safety management requirements for trans-
boundary dams.
4. Adequate public participation and consultation during planning, site selection.
5. Development of effective operational strategies for dams located in trans-
boundary water basins.
6. Mobilizing professionals for collaboration with other governments, international
dam safety organizations and regulatory agencies to accelerate the learning
regarding the best practices in transboundary dams.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the review of J. Groeneveld, Senior

Hydrotechnical Engineer and Western Regional Manager, Waterpower, Hatch Ltd.,
Alberta, Canada.
REFERENCES
[1] Acharya, M., and Donnelly, C.R., (2020) “The need for Dam Safety Man-
agement in Nepal” 88th Annual Meeting and ICOLD-CIGB 2020 Symposium:
Sustainable Development of Dams and River Basins, New Delhi, India.
[2] Acharya, M., Donnelly, C.R., Groeneveld, J., Rutherford J.H., Bennet T., 2019.
Importance of Emergency Management Program for Dams and Hydropower
Projects – Canadian Perspective and Nepalese context. ICOLD-CIGB 2019
420
Q.105 - R.23
Symposium: Sustainable and Safe Dams around the World, June 9–14
Ottawa, Canada.
[3] Aljazeera (2017). Floods kill over 1,200 in India, Nepal and Bangladesh,
Aljazeera.com.
[4] ANEEL (2000). Reservoir Sedimentation Assessment Guideline. Brazilian

Electricity Regulatory Agency, Hydrological Studies and Information Depart-
ment – SIH.
[5] Annandale, G.W., Morris, G.L. and Karki, P. (2016). Extending the Life of
Reservoirs: Sustainable Sediment Management for Dams and Run-of-River
Hydropower. Directions in Development. Washington, DC: World Bank. doi:
10.1596/978-1-4648-0838-8. License: Creative Commons Attribution CC BY
3.0 IGO.
[6] Basynat D.B., Ray P., Reynold J., Immerzeel W., and Harshadeep N.R. (2020).
Addressing Misconceptions about Climate Change in the Himalayas. Webinar
organized by the World Bank on March 25, 2020, moderated, summarized and
edited by Martin Burdett. The International Journal of Hydropower and Dams,
Volume 27, Issue 3, 2020.
[7] Bennett T., and Spektor T., 2019. Examining Dam Safety Guidelines for Emer-
gency Management in the Context of Industry Standards and Practices. Proc.
CDA 2013 Annual Conferences, Montreal, QC, Canada.
[8] Bishwas A.K. 2011. Cooperation or conflict in transboundary water man-

agement: case study of South Asia. Hydrological Science Journal 56(4),
662–670.
[9] Boswell J., Cavanagh P., Staples L., Lefsrud L., Rentz A., 2019. Preliminary
Work by DIAC towards Thinking Clearly and Communicating Effectively about
Risks. Proc. CDA 2019 Annual Conferences, Calgary, AB, Canada.
[10] Devkota L., Giri S., Crosato A., Baral B. (2018), Impact of Koshi barrage and
embankments on river morphology and dynamics, 7th International Confer-
ence and Exhibition on Water Resource and Renewable Energy Development
in Asia, 13–15 March 2018, Ariyana Convention Centre, Danang, Vietnam.
[11] Dick W., Woodvine R., Abrahamson B.,& Acharya M., 2019. PMP and PMF
Estimates in Alberta n Saskatchewan: Initial Review. CDA 2019 Annual
Conferences, Calgary, AB, Canada.
[12] Donnelly, C.R., Rutherford, J.H., Cadou, C., Moler, W., Taylor, S. & Karki, P.
2018. Lessons Learned from Five Hydroelectric Power Projects in Nepal in
the Aftermath of the 2015 Gorkha Earthquake. Proc. 26th ICOLD Congress &
86th Annual Meeting, ICOLD - Question 101, Vienna, June 2018.
[13] Geneva Water Hub, 2019. The Role of Large Dams in Transboundary Water
Negotiations. https://www.genevawaterhub.org/sites/default/files/atoms/files/
roundtablelargedams_20191002.pdf
421
Q.105 - R.23
[14] Gyawali, D. (1999). "Institutional forces behind water conflict in the Ganga
plains." GeoJournal 47(3): 443–452.
[15] Grummer,J., Combating Silt Erosion in Hydraulic Turbines, HRW-Hydro

Review Worldwide, Volume 17, No. 1, March 2009.
[16] ICOLD (2017). Dam Safety Management: Operational Phase of the Dam Life
Cycle. International Commission on Large Dams (ICOLD), Bulletin 154, Paris.
[17] ICOLD. 2007. Shared Rivers: Principles and Practices. International Commis-
sion on Large Dams (ICOLD), Bulletin 132, Paris.
[18] ICOLD. 2007. Mathematical Modelling of Sediment Transport and Deposition

in Reservoirs: Guidelines and Case Studies. International Commission on
Large Dams (ICOLD), Bulletin 140, Paris.
[19] ICOLD. 2016. Global Climate Change, Dams, Reservoirs and Related Water
Resources. International Commission on Large Dams (ICOLD), draft Bulletin
169, Paris.
[20] Jha, P, 2018,. An Analysis of Transboundary Water Resources: A Case Study

of South Asian Rivers. M. Sc. Thesis, McMaster University, USA.
[21] Kundzewicz, Z.W., Krysanovad, V., Benestad, R.E., Hov, Ø., Piniewskic, M.,
Otto, I.M. (2018). ‘Uncertainty in climate change impacts on water resources’,
Environmental Science and Policy 79 1–8.
[22] MAC (Mining Association of Canada). 2016. “Crisis Management and

Communications Planning Reference Guide 2016.” Internet accessed May
2019. https://mining.ca/sites/default/files/documents/Crisis-Management-and-
Communications-Planning-Reference-Guide-2016_0.pdf
[23] International Water Law Project Website. https://www.internationalwaterlaw.org/

institutions/ transboundary_wmos.html
[24] Palmieri, A., Shah, F., Annandale, G.W., and Dinar, A. (2003). Reservoir
Conservation Volume I: The RESCON Approach - Economic and engineer-
ing evaluation of alternative strategies for managing sedimentation in storage
reservoirs. World Bank, Washington DC.
[25] Kaini, P. and Annandale G. 2019. Hydro Review: The way forward for Nepal’s
Hydropower Development. https://www.hydroworld.com/articles/2019/01/
hydro-review-the-way-forward-for-nepal-s-hydropower-development.html.
[26] Ouranous 2015. Probable Maximum Floods and Dam Safety in the 21st Cen-
tury Climate. Report submitted to Climate Change Impacts and Adaptation
Division, Natural Resources Canada, 39 p.
422
Q.105 - R.23
[27] Reynolds, J.M., Donnelly, C.R., Rutherford, J.H., Taylor S., Karki, P. & Sulpya,
R.M. 2018. Accounting for Natural Disasters in Dam Design and Dam Safety.
Proc. Annual Conferences CDA, Quebec City, Canada, 2018.
[28] Schleiss, A.J., Tournier J-P., and Chraibi A. F., 2019. Report of Independent
Expert Panel, Xe-Pian Xe-Namnoy Project, Failure of Saddle Dam D, Final
Report, March 20, 2019.
[29] Schellenberg, G-., - Donnelly C.R., -Holder C., and - Ahsan R.,. 2017. Dealing
with Sediment. Effects on Dams and Hydropower Generation. Hydro review
Worldwide. January - February 2017.
[30] The World Bank, 2019. “Legal and Institutional Frameworks for Dam Safety
Assurance: A Comparative Global Assessment” Volume 1, Main Report
(Draft). World Bank, Washington, D.C.
[31] The World Commission on Dams, 2000. Dams and Development: A New
Framework for Decision-Making, Easthscan Publications Ltd., November
2000.
[32] Wishart, M.J., Ueda, S., Pisaniello, J.D., Tingey-Holyoak, J.L., Lyon, K.N.,
and Boj García, E. (2020) “Laying the Foundations: A Global Analysis of Reg-
ulatory Frameworks for the Safety of Dams and Downstream Communities.”
Washington, D.C.: World Bank Group.
423
Q.105 - R.24
-------
-------
CHUTE-BELL DAM EMERGENCY MEASURES AND REMEDIAL WORKS (∗)
Annick BIGRAS
Chef Expertise Barrages, HYDRO-QUÉBEC
Eric PÉLOQUIN
Chef Études de sécurité des barrages, HYDRO-QUÉBEC
Simon-Nicolas ROTH
Chef Expertise Géotechnique, géologie et structures, HYDRO-QUÉBEC
Éric MAINVILLE
Ingénieur en hydraulique, HYDRO-QUÉBEC
Mathieu ROY
Ingénieur en hydraulique, HYDRO-QUÉBEC
CANADA
SUMMARY
In April 2019, more than 70 residents had to be evacuated downstream from

the Chute-Bell dam in the south of the province of Québec in Canada because of the
precipitation forecasts which announced the exceedance of the safety check flood.
The stability of the dam could have been compromised according to the last dam
safety review. Hydro-Québec has therefore initiated its emergency procedures and
informed Québec government authorities in accordance with the Dam Safety Act.
The Chute-Bell hydropower development, built in the 1910s and modified in 1942
∗ Évacuation de la population et travaux correcteurs au barrage de la Chute-Bell
424 DOI: 10.1201/9781003211808-70

Q.105 - R.24
and 1999, includes, among other things, an overflow spillway dam and a currently
dormant power station with a capacity of 10.6 MW. Corrective works were carried
out in record time in winter conditions to be ready to face the next spring flood. A
spillway channel 4.6 m wide and 25 m long was excavated in the rock. The spillway
dam was stabilized with anchors and the inflatable rubber dam was removed. An
unusual solution has also been implemented to avoid the installation of a cofferdam
and also to increase the total discharge capacity of the system by modifying the
turbines.
RÉSUMÉ
En avril 2019, plus de 70 résidents ont dû être évacués en aval du barrage

de la Chute-Bell dans le sud de la province de Québec, au Canada, en raison des
prévisions de précipitation qui annonçaient le dépassement de la crue de sécurité.
La stabilité de l’ouvrage de retenue aurait pu être compromise selon la dernière éval-
uation de la sécurité. Hydro-Québec a donc enclenché ses procédures d’urgence et
informé les autorités gouvernementales québécoises conformément à la Loi sur la
sécurité des barrages. L’aménagement de la Chute-Bell construit dans les années
1910 et modifié en 1942 et 1999, comprend entre autres un barrage à crête déver-
sante et une centrale actuellement en dormance avec une capacité de 10,6 MW.
Les travaux correctifs ont été réalisés en un temps record en conditions hivernales
afin d’être prêt à affronter la prochaine crue printanière. Un canal d’évacuation de
4,6 m de largeur et de 25 m de longueur a été excavé dans le roc. Le déversoir a
été stabilisé avec des ancrages et la vanne gonflable a été enlevée. Une solution
inusitée a aussi été mise en œuvre afin d’éviter la mise en place d’un batardeau
tout en augmentant la capacité d’évacuation en modifiant les turbines.
1. INTRODUCTION
The Chute-Bell hydropower development is located on the Rouge River, near

the town of Grenville-sur-la-Rouge, Québec, Canada. Commissioned in 1915 by
Hawkesbury Electric Light and Power, the development includes an existing run-of-
river hydropower plant with an installed capacity of 10.6 MW that is out of service
since 2011 and a spillway with an inflatable rubber dam mounted on it.
During the 2019 spring flood, the Rouge River reached the maximum hourly
discharge of a 100 year time series, with a flow close to 1000 m3 /s. At that time,
inflow forecasts indicated that the development would be facing or exceeding its
safety check flood.
To ensure the safety of residents and as part of the emergency preparedness

action plan, the decision was taken to evacuate more than 70 residents downstream
425
Q.105 - R.24
of the dam as a preventive measure. To avoid a similar situation in the future, Hydro-
Québec carried out major remedial work in fall 2019 and winter 2019-2020 to improve
the discharge capacity of Chute-Bell development. This paper describes the 2019
flood event at Chute-Bell as well as the fast-track remedial work that was successfully
carried out before the 2020 spring flood.
2. CHUTE-BELL DEVELOPMENT LAYOUT
Chute-Bell is located on the Rouge River approximately 18 km upstream of the

confluence with Ottawa River and 44 km upstream of Carillon Hydroelectric Power
Station and Dam. Its location is presented in Fig. 1.
Fig. 1
Chute-Bell Dam location.
Localisation du barrage de la Chute-Bell.
The Rouge River watershed is approximately 5 500 km2 with an average slope
of 3.7%. The average annual flow is 100 m3 /s. Fig. 2 presents the different compo-
nents of the Chute-Bell development before and after the remedial work. A spillway
dam was built in 1942, downstream of an existing dam built in the 1915. This spill-
way consists of an overflow Creager type concrete spillway composed of 10 blocks
where block 1 is on the right bank and block 10 on the left bank. Each block has
426
Q.105 - R.24
Fig. 2
Chute-Bell layout before and after the remedial works.
Schéma d’aménagement de la Chute-Bell avant et après les travaux.
its own profile. The water intake and the powerhouse, which are not in operation
since 2011, are located on the left bank. When the powerhouse was generating, the
maximal operating level was 141.9 m. Cross-sectional views US/DS and along the
axis of the dam after the remedial work are presented in Fig. 3.
427
Q.105 - R.24
Fig. 3
Chute-Bell Dam cross-section views US/DS and along the axis after the remedial
work (remedial work shown in red).
Sections transversales et longitudinales du barrage de la Chute-Bell.
3. DESIGN, CONSTRUCTION AND REPAIRS
The following list of major events for the Chute-Bell development includes all
the rehabilitation works and repairs that occurred since its construction:
• 1913–1915 Design and construction

• 1939–1940 Wood crib repair works
• 1942 Construction of a 15 m concrete overflow spillway dam downstream of
the wood crib
• 1962 Hydro-Québec becomes the owner
• 1968 Log chute closure
• 1984 First interruption of power generation
• 1999 Major rehabilitation
(a) Installation of an inflatable rubber dam with a diameter of 1.4 m
(b) Turbines modification (2 × 5.3 MW)
(c) Rehabilitation of the powerhouse left wall
• 2011 Dam Safety Review
• 2011 Left wing wall rehabilitation and second interruption of power generation
• 2014 Outage of the upstream water level gauge in the month of July
• 2019 Outage of the downstream water level gauge in the month of April just
before the flood event
428
Q.105 - R.24
4. DAM SAFETY REVIEW
The latest Chute-Bell Dam Safety Review (DSR) was produced in 2011 and
the main conclusion was the following:
• The overall condition of the spillway was rated in good conditions with no signs
of deterioration;
• The dam was stable for different loading conditions (including normal, unusual
and extreme operation conditions) according to Hydro-Québec internal guide-
lines;
• The rock support on left bank was found to be fractured. It was recommended
to closely monitor it to detect any evolving rock fissure pattern;
• A dam break analysis was conducted and it was found that the development
had sufficient discharge capacity to pass the safety check flood (1:1000 year
flood) at a maximum water level of 14288 m.
5. 2019 FLOOD EVENT CHRONOLOGY
The 2019 spring flood was an exceptional event for many watersheds in the
province of Québec. The Rouge River Basin was no exception. Prior to the event,
wet spring conditions combined with a snow pack volume above the average were
prevailing on the watershed by mid-April. On April 17th , weather forecast showed
a strong low-pressure system approaching the Chute-Bell watershed and the river
discharge was expected to reach extreme values.
On April 24th , the regional staff observed an unusually high-water level. The
water was almost reaching the water intake concrete deck and water overflowing
on the adjacent river’s banks and over the wing walls was observed. At that time,
the water level was estimated to be 142.6 m, about 0.3 m below the water level
defined as the safety check flood (142.88 m) stated in the latest DSR. Considering
the situation and its unpredictable evolution, immediate actions were taken including
a close monitoring of water level by installing water gauges, more frequent inflow
forecast, spillway rating curve assessment and structural stability analysis.
5.1. INFLOW FORECAST
On April 25th at 11:30, Hydro-Québec’s average flood forecasts scenario indi-

cated that upstream water level at the dam would exceed the safety check flood
water level (142.88 m) on April 27th . However, if the high flood forecasts scenario
was considered, the safety check flood water level could be exceeded on this very
same day of April 25th .
429
Q.105 - R.24
Hydro-Québec forecasters also predicted that inflows could continue to rise

until April 28th . Hydro-Québec’s forecast was compared to an independent forecast
issued by the Government of Québec based on data from a hydrometric station
located 4 km downstream of Chute-Bell. These independent forecasts led to similar
conclusions confirming that Chute-Bell Dam was facing a critical situation potentially
requiring the deployment of emergency measures.
5.2. SPILLWAY RATING CURVE VERIFICATION
While the flood forecasts were being conducted, the discharge capacity of the
spillway was verified using the most recent data acquired with the newly installed
gauges. It was found that at very high flow rates, such as those of the 2019 flood
conditions, the applicable discharge capacity curve for the spillway overestimated
the actual onsite discharge capacity. More precisely, during the dam design and
rehabilitation in the 90’s, a theoretical discharge coefficient was used to produce
the applicable discharge rating curve based on the spillway shape with the rubber
dam deflated as was the case at the moment. During the flood event, by correlating
the flow rates calculated during the 2019 spring flood and the upstream water level
measured at the dam, it was found that the discharge coefficient was lower than
expected. It was put forward that the reduced efficiency was caused by the deflated
rubber dam causing a head loss, hence reducing the discharge capacity.
Considering the recalculated coefficient, the spillway efficiency had to be

reduced by 15% for high flood conditions causing that the discharge capacity was
more likely in between the 1:100 year and 1:1000 year flood. This reduction was less
than favourable in the current situation by adding to the uncertainties and enforcing
that critical condition could be reached.
5.3. DAM STABILITY ASSESSMENT
The inflow forecast revealed that safety check flood water level could be
exceeded. Adding to the fact that the discharge capacity curve had some uncertain-
ties, it was imperative to revisit the latest DSR to verify the dam stability calculations.
The following observations were made:
• Stability analysis in extreme flood conditions were conducted using the safety
check flood maximum water level of 14288 m;
• At this water level, considering the assumptions made, the concrete sec-
tion was at risk of cracking based on a residual sliding safety factor at the
rock/concrete interface of 1.15.
• There was no stability analysis that was conducted for a water level exceeding
14288 m.
430
Q.105 - R.24
5.4. DECISION TO EVACUATE DOWNSTREAM POPULATION
On April 25th at 13:00, based on all the verification described above, Hydro-
Québec asked the “Ministère de la Sécurité Publique” to initiate the evacuation
procedures as a preventive measure. As discussed previously, both high flood fore-
casts and average forecasts indicated that safety check flood water level would be
exceeded between April 25th to 27th . The very same day, the “Organisation de la
Sécurité Civile du Québec” (OSCQ) organised the evacuation of more than 70 resi-
dents located downstream of the dam, in a zone defined by the dam break analysis
carried out during 2011 DSR and presented in the Chute-Bell Dam Emergency
Measures Plans.
6. EVENTS FOLLOWING EVACUATION
The day following resident evacuation, Hydro-Québec took numerous actions

to ensure safety including a close monitoring to assess the dam behaviour and
define a strategy to allow a safe return for the evacuated population. These actions
involved daily flood forecasts update, aerial and terrestrial inspections with upstream
dam inspection.
The strategy to safely reintegrate the evacuated population included the

followings steps:
• Obtain multi-day forecasts for the Rouge River;

• Stability assessment of the concrete dam and verification of its structural
condition by helicopter and then by ground inspection;
• Assessment, by helicopter then by ground visit of the rock conditions of the
left support after peak discharge that had just occurred over the last days;
• Obtaining and reviewing data on the existing dams upstream of Chute-Bell to
ensure that the possible risk of one of these dams breaking could not have
an impact on the Chute-Bell development, including a helicopter inspection of
structures with a higher storage capacity;
• Assessment of the condition of the freeboard dike after peak discharge on
April 25th by helicopter and then by a ground visit.
6.1. HELICOPTER AND DRONE INSPECTION
From April 26th to April 29th , multiple aerial inspections of the dam site were
conducted with a drone. High-definition pictures and videos were collected to verify
the dam state and evolution. Fig. 4 shows comparison of pictures taken in normal
hydraulic conditions during the inspections related to the 2011 DSR with pictures
taken by the drone during extreme hydraulic conditions.
431
Q.105 - R.24
Fig. 4
Drone pictures during flood event compared to normal conditions.
Photos en conditions normales et pendant la crue extrême de 2019.
The drone inspection was used to verify that there were no irregularities in the
water flowing on the spillway crest as an indication of the absence of local damage
to the structure. This allowed at that time to exclude the possibility of imminent dam
break. With the drone, special attention was given to the left abutment which had
signs of rock erosion in the past. However, as shown in Fig. 4, there was an excess
of water flowing over the left abutment and it was not possible to conclude to any
additional damages to this abutment at the time of inspection.
On April 29th , an aerial inspection was conducted to verify the condition of

six dams, with significant reservoir storage capacity (>5 hm3 ), located upstream
Chute-Bell Dam. These dams are not owned by Hydro-Québec, but there was a
concern that a dam break of one or more of these six sites would cause a cascading
breach flood at Chute-Bell. The upstream dams appeared to be in good condition
even though hydraulic conditions were extreme. Additionally, a review of the dam
break studies for these six dams confirmed that a wave front related to a breach at
432
Q.105 - R.24
these dams would be almost completely dampened when reaching Chute-Bell, or

for the worst, producing only a few centimetres increase of the water level.
On May 1st , as the water levels started to decrease, another helicopter inspec-
tion was conducted at the Chute-Bell site. It was possible to visualize that the rock on
left abutment was not damaged and ensured rim closure. The helicopter inspection
was a requirement to give the authorization for a terrestrial inspection to ensure the
safety of the workforce to be deployed on site.
6.2. SITE INSPECTION PRIOR TO POPULATION REINSTATEMENT
On May 2nd upstream water level was still decreasing, therefore Hydro-Québec
conducted a terrestrial inspection. The concrete structure and left abutment were
inspected and were found to be in satisfactory conditions despite the extreme event
that had just occurred.
6.3. POPULATION REINSTATEMENT
On May 2nd , after the terrestrial inspection, the forecasts indicated that the
water level would remain below 142 m for the coming week (Fig. 5). Based on all
the activities from the defined strategy which include investigations combined with
post-flood stability analysis of the dam, the emergency measures were lifted by
Hydro-Québec on the evening of May 2nd . The population that was evacuated since
April 25th could now safely return to their home.
Fig. 5
Upstream Water Level at Chute-Bell (measurements and forecasts) on May 2nd .
Prévisions et observations du niveau d’eau amont le 2 mai
433
Q.105 - R.24
7. POST EVENT INVESTIGATIONS AND DAM SAFETY ASSESSMENT
The post-event investigations were very important to verify whether the

extreme flood damaged the Chute-Bell development. These investigations included
geotechnical (dike), geology (rock support and scour) and structural (concrete)
aspects. Other investigations, not detailed in the present paper, include upstream
submarine inspection, reservoir bathymetry, installation of a new upstream water
level gauges, visual inspection and scan in dry condition of the downstream face of
the spillway.
7.1. GEOTECHNICAL INVESTIGATION
The freeboard dike located on the right bank (Fig. 2), where there is a low
natural topographic point, ensures rim closure during flood periods. The design
of this dike includes a concrete wall. Based on the limited information available
prior to the investigations, the drawings of the backfill under the wall (waterproof
element) showed that it contained pervious materials, especially a gravel layer. The
waterproofing at the base of the wall was understood to be provided by the backfill
on either side of the concrete wall. Therefore, geotechnical investigations including
trenches and land surveying were required to better understand the design of the
dike and to assess the closure of the reservoir. Moreover, removal of the vegetation
that grows over the year on the dike area was required to allow a thorough inspection
of the structure to confirm that there was no damage caused by the flood.
The geotechnical investigations revealed that the concrete wall is founded on

rock, as opposed to what was indicated on the construction drawings. In addition, in
the upstream upper sections of the dike, a till layer was found to be present at the rock
interface. Thus, the investigations allowed to conclude that the dike is impervious.
7.2. GEOLOGICAL INVESTIGATION
Prior to the flood, the erodibility index of the rock mass at the base of the
spillway was evaluated using the Annandale’s scour threshold method [1] and it was
concluded that it is very unlikely that the sound rock would erode significantly in the
long term. It was, however, recommended to monitor the evolution of a fractured zone
adjacent to the inflatable rubber dam anchor block on the left bank. As mentioned
previously, during the flood event, specific daily inspections by drones and helicopter
were carried out to assess the condition of this fractured zone especially because
water was flowing over it (Fig. 4 – left bank). Post-event inspection revealed that
there was no scour under the concrete structure. On the right bank, the rock mass
is of good quality and is not subject to erosion.
434
Q.105 - R.24
7.3. STRUCTURAL INVESTIGATION
Prior to the investigation, structural safety assessment of the spillway was

performed based on the geometry taken from available pictures and images. This
because there were no construction drawings or geometric survey available. For the
log chute section on block 9, there was an uncertainty in the dam stability calcula-
tions regarding the geometry as there was no information whether this section was
stabilized with concrete during the repair works performed at the end of the 90’s.
Therefore, conservative assumptions were considered on the geometry. Hence, it
was required to validate the geometry and to perform a visual inspection of the spill-
way in dry conditions. Besides inflating the rubber dam, there was no alternative to
bypass or stop the inflows for short time periods to allow an inspection. A trial using
a high-flow compressor demonstrated that it was not possible to inflate it up to more
than 50% which was not sufficient to perform the visual inspection. The inflatable
rubber dam was therefore considered unusable.
Making core samples to obtain concrete, rock-concrete and concrete-concrete

interfaces properties was discussed. The deterministic safety assessment per-
formed using the gravity method, coupled with the Mohr-Coulomb criterion, defined
by a friction angle and cohesion parameters, is typically used and recommended
by many guidelines for geometries similar to that of the Chute-Bell Spillway. These
two parameters have however significant uncertainties especially those at the rock-
concrete interface. The revised parameters, reduced compared to the previous DSR
as the dam had just sustained an extreme flood event, were consistent with state-of-
the-art engineering practices. Therefore, it was concluded that the revised stability
safety factors of blocks 1 and 2 were close to being unacceptable. Furthermore,
even with core samples data, these conclusions would have remained the same con-
sidering the uncertainties associated with these two modelling parameters as well
as their spatial variability [1]. The remedial works would therefore need to include
improvement of the spillway stability safety factors.
7.4. FAST-TRACK ENGINEERING AND REMEDIAL WORKS BEFORE 2020 SPRING

FLOOD
After a post-mortem review of the situation experienced and the results of the
additional investigations performed, it was decided that corrective works had to be
completed on the Chute-Bell development, prior to the 2020 spring flood, to ensure
safety of the dam. The objectives of the remedial works were to:
• Increase the discharge capacity of the development in order to pass the safety
check flood at a maximum upstream water level of 142.88 m;
• Improve the stability of blocks 1 and 2 for a maximum water level of 142.88 m.
Taking into account the paradigm of sustainable development [2] [3], engi-
neering requirements considered many design criteria involving technical, legal,
435
Q.105 - R.24
Table 1
Design criteria considered
Technical • Considerations of health and safety: no motive justifies endangering the health and
Aspects lives of public and workers
• Reduced time frame for planning engineering, awarding contracts and carrying out
repair works
• Challenges related to the northern climate of Québec for the works done during
winter conditions
• Consideration of uneven site topography
Social Aspects • Preoccupation of the local community during the project, including the work
schedules, the noise generated as well as the local road traffic
• Compliance with the restrictive Hydro-Quebec real estate ownership limits
• Consideration of the impact that would have a change in upstream water level on
the navigability and the residents’ habits
• Preservation of the waterfall appearance
• Compliance with the aesthetic flow required from June to September as agreed with
the residents
• Management of communications with the nearby residents for social acceptability
and keeping the confidence level in the hydropower development
Environnemental • Analysis of the natural environmental impacts on parts of the biosystem that are
Aspects affected by the remedial and temporary construction works
• Analysis of the impact that would have a change in upstream water level on animal
species, flora and fauna
Economic • Estimation of the order of -40%/+50% for the solutions for that had been experi-
Aspects mented beforehand at other sites and -50%/+100% for the solutions that presented
uncertainties
• Inclusion of the operating expenses (OPEX) in the economic analysis
environmental, social and economic aspects. These aspects, given in more detail in
Table 1, were considered in the analysis of the different scenarios for the remedial
works.
7.5. REMEDIAL WORKS OPTION
Finding a solution was a puzzle. One of the reasons was that the work per-
formed on the spillway to increase its stability safety factor had to be carried out
in dry conditions to guarantee with health and safety measures. The risk of being
able to deliver the work on time before the 2020 flood was assessed for each of the
proposed variant.
To achieve the remedial work objectives, numerous variants were identified,

discussed and analyzed as there was a risk that a single variant would not entirely
meet the desired objectives. In addition, with the work required to increase the
spillway stability safety factors, a total of four variants were examined in more detail.
1. Excavation of a diversion channel on the left or right banks;

2. Removal of blocks 1 and 2 from the spillway;
3. Removal of the inflatable rubber dam;
436
Q.105 - R.24
4. Conversion of two unconnected turbine-generator (TG) units and their pen-

stocks to discharge pipes.
To meet the first objective of the remedial works, the variants considered will
be presented in the following sections. First, the solution chosen to increase the
stability safety factors of spillway blocks 1 and 2 will be presented.
7.5.1. Stability safety factors increase of spillway blocks 1 and 2
To comply with the second objective of the rehabilitation works, the solution
of adding active anchors to blocks 1 and 2 was adopted. For this purpose, three
post-tensioned anchors were added per block, each block has a 6.10 m (20 ft)
width. Hollow Deformed Rebar (HDR) with a diameter of 35 mm had to be installed.
These anchors were positioned 650 mm from the upstream face of the structure
and anchored at a depth of at least 5.5 m in the rock. These bars are mechanical
anchors and they are grouted after tensioning to protect them against corrosion. To
ensure that they are anchored in sound rock mass, a depth of 5.5 m was considered
sufficient considering this crystalline type of rock mass. In order to not interfere with
the flow and to protect the anchor heads, the concrete surface had to be reamed to
create a circular opening to receive the anchors’ head. The circular openings were
grouted following the anchor installation, therefore the anchor heads are encapsu-
lated in the spillway. Note that this stabilization work may not be necessary for the
second variant proposed.
7.5.2. Excavation of a diversion channel
The excavation of a diversion channel is a fairly classic idea and appears

simple at first glance to increase the onsite discharge capacity. However, several
constraints turned this simple solution into a challenging one.
On the left bank, the channel would need to be excavated in a topographically

unfavourable location. The estimated rock excavation volume for a 10 metres-wide
channel was too important considering the tight schedule of the project. The increase
in local road traffic to remove excavated materials, the vibrations generated by blast-
ing near existing structures and the post-construction site aesthetics were important
issues. Moreover, as specified in the previous sections, the rock quality on the left
bank would probably require stabilization work in the excavated area and possibly
causing additional delays. A new flow regulation structure would have been required
as there would have been impacts to the upstream water level. The environmen-
tal and social impacts linked to navigability and residents’ habits would need to be
studied. Also, authorizations would be required from governmental agencies to build
this channel. For all these reasons, any proposed variants that included left bank
diversion channel were put aside while other options were investigated.
437
Q.105 - R.24
On the right bank, the restrictive Hydro-Québec real estate ownership would
limit the size of the channel, making it impossible to achieve the required increase in
discharge capacity without strongly affecting the water level and the waterbody users
located upstream. However, a decent discharge capacity increase could be achieved
by excavating a 4.6 metres wide channel within the Hydro-Québec ownership limits.
This would represent a gain of near 3% in total discharge capacity in safety check
flood conditions. To comply with a minimum aesthetic flow requirement, a 10 cm
sheet of water overflowing on the spillway is required at all time. It was therefore
important that the proposed diversion channel conforms to this requirement, hence
the headrace’s invert elevation should be 10 cm above the spillway sill.
7.5.3. Removal of blocks 1 and 2 from the weir
Complete removal and reconstruction of blocks 1 and 2 was performed in

1998 to install an inflatable rubber dam (Fig. 6, left). An important construction
phase required for its installation was the construction of a cofferdam. This coffer-
dam extends from the right bank to block 3 to create a dry working environment
upstream of the two blocks involved in the work (Fig. 6, right). Based on this past
experience, a similar solution was considered. However, depending on the site con-
ditions (river flow rate and weather conditions), the construction of a cofferdam may
have important impacts on the schedule. Also, to limit the consequences of the
removal of these blocks on the upstream water level, the installation of a regulating
structure would be required. It is important to note that partial removal of these blocks
is not possible as the weight removal would cause structural stability issues. Instal-
lation of anchors on the remaining parts of the blocks would therefore be necessary.
This solution is quite complex considering the need to add a regulating structure in
the project. Also, the cost of operating the facility would increase. This solution was
thus discarded.
Fig. 6
Photo taken from the 1998 project (left), sketch of the regulating structure with
cofferdam (right).
Photo prise en 1998 lors du projet de refection (gauche) avec le schéma d’un
ouvrage régulateur et batardeau nécessaire (droite).
438
Q.105 - R.24
7.5.4. Removal of the inflatable rubber dam
As explained previously, the expected discharge coefficient was lower than the
one used for the design of this Creager spillway. As already mentioned, it was sus-
pected that it was the presence of the inflatable rubber dam in deflated state that was
creating a head loss as well as a reduction of the discharge coefficient. Such an issue
was studied by many authors [4, 5] and they have highlighted that potential hydraulic
jump, above or next to the deflated membrane, the potential vortex shedding in wake
regions caused by the membrane protuberance into the flow and downstream end
of the membrane, could reduce the discharge coefficient. Therefore, removing it
could provide an increase in the discharge capacity of the Chute-Bell development.
However, the estimation of capacity increase had many uncertainties. The main
benefit of this variant is that it minimizes the impacts on social, environmental and
economic aspects. The main drawback was that the additional discharge capacity
obtained with this variant was uncertain.
Additionally, to achieve this variant in accordance with health and safety

requirements, the total river flow would need to be diverted. Consequently, the
combination with one of the other variants was deemed essential to put this solution
forward.
7.5.5. Conversion of two unconnected turbine-generator (tg) units and their

penstocks into discharge outlets
It is quite common in a development to use the TG units in spin-no-load

conditions to avoid frequent start/stop sequences, for example. This is done to
be ready to be reconnected to the power grid and provide short time produc-
tion to fulfill spot-market contracts. Typically, this operating condition is associated
with low discharge and small guide vanes opening angles. The proposed variant
seeks to optimize the guide-vane opening angles to provide the highest possi-
ble discharge using the unconnected TG units which therefore must not have a
particular synchronization for this particular and uncommon use. The maximum
flowrate is therefore linked to the maximum water velocities leaving the distrib-
utor outlet to the inlet of the draft tube that could lead to vaporization of the
water.
The Chute-Bell power plant has two Saxo-type turbines (S axial turbine) whose
section is shown in Fig. 7. These turbines are similar to Kaplan-type turbines, how-
ever, with a conical distributor and without a spiral case. The spiral case is replaced
by a compact elbow with deflector vanes, which predetermines and unifies the flow
downstream to the stay vanes, guide vanes and the runner [6]. A vertical shaft
passes through the elbow to connect with the generator.
439
Q.105 - R.24
To avoid rotation of the unit that could revive the problems of vibrations and
wear of the turbine bearing which was experienced in the past, the runner blades
would need to be replaced with plugs to prevent cavitation on the hub. This solution
is reversible, which means that the group could be put back into service for future
use, meeting the criteria of sustainable development.
Preliminary computational fluid dynamic (CFD) analyzes were used to define

the optimal guide vane angle (Fig. 7). The optimal guide vane position that was
found to be feasible was a 70% opening. However, the resulting discharge capacity
increase was still insufficient to what Hydro-Québec was aiming for. This variant
would require a combination with another one to fulfill the capacity Hydro-Québec
was aiming for. Moreover, other known considerations in the flow could possibly
reduce the discharge capacity computed. It was studied that turbine operating in
discharge mode have a swirl at the runner trailing edge that leads to a complex and
unstable vorticity distribution. This unstable flow may result in the appearance of
cavitating or non-cavitating vortex rope of helical shape [7] reducing the discharge
capacity. The numerical simulation of these phenomena requires complex turbu-
lence models that are computationally intensive which were not used considering the
reduced engineering time available. Therefore, appropriate instrumentation would
need to be installed in order to see if these phenomena occur.
Fig. 7
Cross-section of the Saxo turbine and CFD analysis.
440
Q.105 - R.24
8. CONCLUSIONS
A combination of the previously mentioned variants was required to reach

the discharge capacity needed. The remedial works needed to fulfill two
objectives:
1. increase the discharge capacity of the development;

2. increase the spillway stability safety factors of blocks 1 and 2
The final solution consisted to convert first the two unconnected TG units and their
penstocks into discharge outlets. This allowed to divert the flow and provide dry con-
ditions for additional works on the spillway. These works consisted in the excavation
of a diversion channel on the right bank, the installation of rock anchors on blocks 1
and 2 and the removal of the inflatable rubber dam (Fig. 8). The remedial works
were conducted in harsh winter conditions according to schedule and to what was
planned.
Fig. 8
Removal of the inflatable rubber dam.
441
Q.105 - R.24
After the works, the population was advised that the Chute-Bell Dam was now
in order to safely resist the coming 2020 spring flood at least to its safety check flood
value. This 2020 spring flood was monitored closely by Hydro-Québec to confirm
some of the assumptions and the exact values for the discharge capacity. It was
possible to correlate the upstream water levels with the measured flow rates in
order to calculate a revised discharge coefficient for the spillway. The increased
discharge capacity could be assessed as well. It was found that the remedial work
would provide sufficient discharge capacity in order to prevent the upstream water
level to exceed 142.88 m under safety check flood conditions. Finally, analysis of the
data collected by the pressure sensors installed along the turbine section, starting
upstream of the guide vanes up to the runner cone confirmed that vortex breakdown
occurs. Mitigation technique were proposed in the literature such as air or water
injection under the runner hub to reduce or remove the vortex breakdown [8, 9].
The aim of this paper was to describe the different actions and decisions
before, during and after the 2019 flood event at Chute-Bell development. Since
the Quebec Dam Safety Act, it was the first time in Hydro-Québec history that
emergency measures were initiated and that population evacuation was required.
Despite the circumstances, Hydro-Québec did everything possible to fulfill its civic
duty and showed interdisciplinary expertise to successfully deliver a fast-track engi-
neering and construction project to prevent the occurrence of a similar situation in
the future. Finally, this type of flood event allows to give afterthoughts on activation
of emergency plan and how to improve current practices.
REFERENCES
[1] G. Annandale, Scour technology, Mc-Graw-Hill, 2006.
[2] S. Renaud, N. Bouaanani et B. Miquel, “Numerical simulation of experimen-

tally shear-tested contact specimens from existing dam joints,” Computers and
Geotechnics, vol. 125, 2020.
[3] Brundtland Commission, Our Common Future, Oxford: Oxford University

Press, 1987.
[4] M. A. Khan, “Sustainable development: The key concepts, issues and

implications,” Sustainable Development, vol. 3, n◦ %12, pp. 63–69, 1995.
[5] A. Chervet, “Model-prototype comparison of the defective behaviour of an

inflatable dam,” chez Symposium on Scale Effects in Modeling Hydraulic
Structures, 1984.
[6] H. Chanson, “Some Hydraulic Aspects during Overflow above Inflatable Flex-
ible Membrane Dam,” Report CH47/96, Dept. of Civil Engineering, University
of Queensland, Australia, 1996.
442
Q.105 - R.24
[7] E. Höfler, J. Gale et A. Bergant, “Hydraulic Design And Analysis Of The

Saxo-Type Vertical Axial Turbine,” Transactions of the Canadian Society for
Mechanical Engineering, vol. 35, n◦ %11, pp. 119–143, 2011.
[8] S. Houde, M. S. Iliescu, S. L. Richard Fraser, G. D. Ciocan et C. Deschênes,

“Experimental and numerical analysis of the cavitating part load vortex dynam-
ics of low-head hydraulic turbines,” chez ASME-JSME-KSME Joint Fluids
Engineering Conference, Japan, 2011.
[9] R. W.J., “Power swings in hydroelectric power plants,” Transaction of the

ASME, vol. 62, n◦ %13, 1940.
443
Q.105 - R.25
-------
-------
ASSESSING THE SAFETY OF THE ITUANGO CEMENT-BENTONITE CUTOFF

WALL (∗)
C. Richard DONNELLY
Global Principal Consultant,
Dams and Waterpower, HATCH, Niagara Falls
Eric TIEDJE
Geotechnical Engineer,
HATCH, Mississauga
Ing Jesús ARANGO

Project Manager,
EPM, Medallion, Colombia
Daniel FLORES
Waterpower lead,
Waterpower lead, South America, HATCH, Niagara Falls
CANADA
SUMMARY
During the construction of the 235 m high Ituango rockfill dam in Colombia, a
landslide blocked the diversion tunnel resulting in an uncontrolled rise of the reservoir
while the dam and spillway were still under construction. Attempts to unseal the
diversion tunnels ultimately were unsuccessful leading to the very real possibility
that the dam could be overtopped. Therefore, a fast-track priority raise scheme was
developed to ensure that the dam crest remained above the rising reservoir level. The
∗ Évaluation de la sécurité du mur parafouille ciment-bentonite d’Ituango
444 DOI: 10.1201/9781003211808-71

Q.105 - R.25
scheme involved the construction of a 35 m high cement bentonite cutoff in the upper
portion of the dam and diverting water through the partially completed powerhouse
complex. The solution proved effective with the dam and spillway completed without
incident, although extensive damage to the powerhouse complex occurred which is
being addressed at the time of the preparation of this paper. Given the somewhat
unique nature of the priority raise solution that was ted, a detailed assessment of the
integrity of the wall under conditions of future settlement was commissioned. This
paper outlines the assessments performed to assess the safety of the Ituango dam
under static conditions. Dynamic assessments are currently being finalized and will
form the basis of a future discussion.
RÉSUMÉ
Pendant la construction du barrage en enrochement d’Ituango de 235 m

de haut en Colombie, un glissement de terrain a bloqué le tunnel de dérivation,
entraînant une augmentation noncontrôlée du réservoir alors que le barrage et
l’évacuateur de crues étaient encore en construction. Les tentatives d’ouverture
des tunnels de dérivation ont finalement échoué, ce qui a conduit à la possibil-
ité très réelle que la crête du barrage puisse être dépassée. Par conséquent, un
programme prioritaire de rehaussement de la crête du barrage a été mis au point
pour s’assurer que le barrage demeure plus haut que le niveau amont du réservoir.
Le projet comprenait la construction d’un mur parafouille en ciment bentonite de
35 m de hauteur au sommet du barrage et le détournement de l’eau par la centrale
hydroélectrique partiellement complétée. La solution s’est avérée efficace puisque
le barrage et l’évacuateur de crues ont été complétés sans incident, bien qu’il y ait
eu des dommages importants à la centrale hydroélectrique, dommages en cours
de revue au moment de la préparation du présent document. Compte tenu de la
nature quelque peu unique du programme prioritaire de rehaussement de la crête
du barrage, une évaluation détaillée de l’intégrité du mur dans les conditions futures
de règlement a été demandée. Cet article décrit les analyses effectuées pour éval-
uer la sécurité du barrage d’Ituango dans des conditions statiques. Des analyses
dynamiques sont en cours et constitueront la base pour de futures.
1. INTRODUCTION
The 2456 MW Ituango hydroelectric station is currently nearing completion on

the Cauca River near Ituango in Antioquia Department, Colombia. The hydroelectric
scheme is a relatively compact layout consisting of diversion tunnels, an intake, an
eight-unit underground powerhouse, a power tunnel, a tailrace tunnel, a spillway
and a stilling basin as illustrated in Figure 1.
When completed, Ituango will be the largest power station in Colombia with
eight 307 MW (412,000 hp) Francis turbines operating under a head of 197 meters.
The initial feasibility study was completed in 1983, but the project was shelved in
445
Q.105 - R.25
Fig. 1
The Ituango Hydroelectric Project Site.
Le site du projet hydroélectrique Ituango.
the 1990s due to an economic crisis. The final designs for the project were finished
in 2008 and on July 8th , 2011, the project management contract was awarded.
Preliminary construction (surveying, roads, bridges, diversion tunnels) began in
September 2011. The Ituango dam is a 235-metre high zoned rockfill dam with
a clay core, filters, transition zones and riprap involving about 19 million cubic
meters of fill (Figure 2). It creates a 27 km long reservoir that covers an area of 38
square kilometers with a capacity of about 2,720-million-cubic-metres of which about
980-million-cubic-metres is active. The spillway discharge has a design discharge
of 22,600 m3 /s achieved by means of four radial gates.
In this paper, a discussion of the assessment of the safety of a 37 m high

cement-bentonite wall, installed in the upper portion of the dam. The wall was used
as part of a fast-track raise of the Ituango dam, which was needed as a result of
an uncontrolled rise in reservoir levels that occurred when the diversion tunnel was
blocked prior to the completion of the dam and spillway. Given the rather unique
configuration of a cement-bentonite cutoff being constructed within the body of a
dam, detailed static∗ analysis of the wall was undertaken to establish the impact
that ongoing settlement would have on the cutoff in the years following construc-
tion using the PLAXIS2D and PLAXIS3D programs for static loadings. The analysis
∗A review of the project seismic hazard assessment in accordance with current ICOLD
guidelines (ICOLD 2014) is ongoing
446
Q.105 - R.25
Fig. 2
The 235 m high Ituango Embankment Dam and Spillway.
Ituango – Barrage de 235 m de hauteur et évacuateur de crues.
included derivation of representative material models and their associated param-

eters for each of the major components of the dam based on information obtained
from construction records, monitoring data, construction quality control testing and
material specifications. The work also involved a unique series of stress-permeability
tests to establish strain thresholds where strain induced cracking could materially
impact the functioning of the cement-bentonite wall that are currently ongoing. The
447
Q.105 - R.25
results of these assessments as well as recommendations for ongoing monitoring

and assessments to ensure the dam continues to be safe are presented.
2. GEOLOGY AND GEOMORPHOLOGY
The Cauca river at the location of the dam is a typical, slightly asymmetric "V"
shaped canyon (Figure 3) composed of metamorphic rocks with a right abutment
that is steeper than the left as a result of the orientation of the foliation.
Fig. 3
Geology of the Ituango Dam Foundation.
Géologie de la fondation du barrage d’Ituango.
There is more than 1000 m of relief. Bedrock consists of generally good quality
gneisses composed of quartz feldspars that are locally schistose due to the rocks
sedimentary origins which makes the rock mass highly variable texturally over short
distances, ranging from thick layers of gneiss interspersed quartz and feldspar to
layers of biotite and amphiboles. Fault systems of regional importance surround the
project;
• Santa Rita West and East fault system to the east which cross the Cauca river;
• Cauca, Romeral, Sabanalarga, Ituango and Sardinas fault systems to the
west;
• Espíritu Santo system in the northeast and
448
Q.105 - R.25
• Murri – Mutatá to the north-west
3. THE PRIORITY RAISE INCIDENT
Diversion of the Cauca River around the construction site was achieved using
three tunnels. Towards the end of the construction program, two of these tunnels
were sealed, as part of the impounding procedure, leaving a third to divert the
river. Unfortunately, between April 28 and May 7, 2018, three landslides blocked
the remaining tunnel, resulting in an uncontrolled filling of the reservoir, while the
spillway and the dam were still under construction. Attempts to re-open the two
sealed diversion tunnels failed resulting in a continued rise of the reservoir. On
May 10, when the reservoir reached the power station intake, water was released
through the unfinished powerhouse to slow the rate of rise of the reservoir. This set
off a series of events that are summarized in Figure 4.
Fig. 4
The Ituango Dam Safety Incident – A Timeline.
L’incident du barrage d’Ituango – Chronologie.
On May 12, 2018, one of the previous sealed tunnels partially reopened send-
ing a surge of water downstream that resulting in the need to evacuate about 25,000
residents from the township of Puerto Valdivia. The limited additional discharge
offered by the partially re-opened diversion tunnel did little to stem the reservoir rise
and, on May 16, the discharge through the tailrace was partially obstructed with con-
tinued heavy rainfall was expected in the Cauca River basin through the end of May.
Immediately following the occurrence of the landslide that blocked the discharge tun-
nel on April 30, a 33 m high priority dam raise was instituted to ensure that the rising
reservoir levels would not overtop the dam. A key feature of the priority raise was a
unique 37 m high cement-bentonite cutoff wall designed and constructed on a fast
track basis within the upper portion of the dam. This unique design made use of the
449
Q.105 - R.25
available construction materials and allowed for a rapid construction of the rockfill
embankment followed by the installation of a two-pass, 1 m thick, cement-bentonite
wall. The design proved to be successful with the dam completed ahead of the rising
reservoir on May 24, less than 30 days after the start of construction of the priority
raise. The spillway was completed on June 2 mitigating the risk of overtopping.
4. CONSTRUCTION
The Ituango dam was, for the most part, constructed as a conventional zoned
rockfill embankment using modern placement and compaction methods. An example
of the quality of the work is illustrated in Figure 5.
Fig. 5
Placement of Fill Materials above the Cement-bentonite Wall.
Placement des matériaux de remplissage au-dessus du mur ciment-bentonite.
The regular construction of the dam was suspended when the core reached
385 meters above sea level (masl), corresponding to a dam height of 175 m, due
to the unscheduled filling of the reservoir caused by the blockage of the Auxiliary
Deviation Tunnel. From this level, an innovative fast track design was formulated
involving the construction of a 37 m high cement bentonite wall tied into the core in
order to reach a height 418 masl ahead of the rising reservoir level. The construction
of this portion of the dam was performed in three stages (Figure 6). The first stage
was designed to reach a height of 410 masl which was a level that would mitigate
most of the risk associated with the rising reservoir, the second stage involved raising
the embankment an additional 5 m to 415 meters above sea level followed by the last
450
Q.105 - R.25
Fig. 6
Details of the Construction of the Ituango Dam in the Priority Zone.
Détails de la construction du barrage d’Ituango dans la zone prioritaire.
Fig. 7
Construction of the Cement-bentonite cut off in the priority raise zone.
Construction du mur parafouille ciment-bentonite dans la zone de rehaussement
prioritaire.
stage to achieve the 418 masl. During this portion of the raise an inclined impervious
blanket tied into the underlying core with an upstream blanket forming the temporary
water retaining component of the dam.
The permanent cement-bentonite cut off was installed following a two-stage

excavation and tremie-replacement methodology from the top of the priority raise
451
Q.105 - R.25
Fig. 8
Abutment treatment details.
Détails du traitement de l’appui.
zone using five separate plants starting on September 5, 2018 (Figure 7). The
primary panels were 2.8 m long and 38 m deep, keyed 5 m into the central clay
core. The secondary panels were 1.8 m long and 38 m deep. All of the panels
were determined to be within 0.5 % of vertical. The wall was constructed generally
without incident but at six locations the clam-shell excavator tool was jammed which
required enlargement of the wall to remove equipment. The reservoir levels rose
to 402 masl during phase 1 construction and reached a maximum of el 410 m by
the end of construction. Instrumentation for the wall included seven inclinometers
at 75 m spacing installed just downstream of the wall in perpendicular panels, 13
piezometers installed downstream of wall and 3 installed upstream of wall, and 6
magnetic extensometers installed in dedicated cement-bentonite panel adjacent to
the wall.
The bedrock abutments were treated with formed reinforced concrete to

provide a suitable seal for the core materials and the cut off wall (Figure 8).
452
Q.105 - R.25
The construction of the cut off wall was complicated by the fact that only rel-
atively coarse rockfill was available due to the loss of many of the project borrow
areas. While this resulted in up to 400% losses in the bentonite slurry, the wall
was successfully completed on January 30, 2019 about 21 weeks after starting of
construction. At the right and left abutments, two additional cement-bentonite walls
were installed, 21.7 m and 54.8 m long respectively, parallel and each offset from
each other by 1 m, in order to widen the cut off in these critical areas, with plastic
cement grouting performed between the walls.
5. ASSESSMENT OF THE SAFETY OF THE CEMENT-BENTONITE WALL
The configuration of the Cement-Bentonite wall within the dam is relatively

unique in that it is located solely in the upper region of the dam. The bottom of the
wall is not fixed and will settle and deform with the rest of the dam. An analysis
was undertaken to assess the potential for cracking of the wall to occur under two
conditions:
1. The development of localized stresses and strains within the wall due to the
placement of material overtop and impacted by the relative differences in
stiffness of the surrounding materials.
2. The three-dimensional deformation of the wall caused by deformation at the
center of the dam relative the abutment connections.
The dam was assessed using 2D finite element methods to determine the
anticipated settlements under static loadings considering the staged construction
of the dam. These predicted settlements were then calibrated with the results of
the monitoring program. 3D analyses were then performed using the calibrated
parameters to assess the behaviour of the dam and the cement-bentonite wall under
static loading conditions.
5.1. FAILURE MECHANISMS
The Ituango cement-bentonite wall is not a structural element. Therefore, it

does not contribute to the global stability of the embankment. Its purpose is to
control internal seepage forces in order to avoid excessive seepage losses and
internal erosion. From that perspective, a “failure” of the cutoff can be defined as
stress and strain induced cracking of the cement-bentonite wall to such an extent
that the wall fails to meet its intended purpose in terms of serving as a hydraulic
barrier. Two main failure mechanisms were considered.
• Overstressing of the cement-bentonite material due to the variation in the

stiffnesses between the cutoff and the surrounding fill materials.
453
Q.105 - R.25
• Non-uniform deformation of the cutoff due to ongoing settlement of the lower

portion of the embankment.
The height of the embankment above the wall is only about 18 m and would not
be expected to result in significant settlement induced shear stress at the wall/rockfill
interface. In addition, the abutment treatment measures place the wall and fill mate-
rials in compression which is favourable for the long-term performance of the dam
and the cement-bentonite wall.
5.2. MATERIAL PROPERTIES
The stress-strain characteristics of the dam fill materials were defined using a
hardening soil with small strain (HSsmall) constitutive model. Due to the significantly
higher strength and stiffness of the bedrock foundation relative to dam fill, it was
considered that the behaviour and deformation of the dam would not be materially
affected by the characteristics of the bedrock foundation. At the outset of the analysis,
it was not clear how the variation in material stiffness between each zone would
impact the strain level within the cutoff wall.
Three sets of parameters were developed to reflect the anticipated variation

in density, strength, and stiffness in the different fill zones. The three parameter sets
corresponded to the minimum, mean, and maximum anticipated values. Likewise,
for the cement-bentonite material three sets of parameters were considered to reflect
the observed variation in strength and stiffness during construction. The key material
properties were derived from the following sources of information, ranked by their
order preference:
• Direct measurement of properties from as-constructed conditions, such as

quality control testing, including plate load testing.
• Correlation of properties from construction records, such as field density.
• Direct measurement of properties from laboratory testing of reconstituted
samples.
• Properties of similar materials reported in the reference literature.
Figure 9 illustrates an example of determining the stiffness parameters from

rockfill based on as-constructed plate load test performed as part of the con-
struction quality program. In this case the plate load test was reproduced with a
simple, 2D axial symmetric Finite Element Analysis, and the material properties
were selected in order to reproduce observed behavior. The validity of the model
and input parameters was confirmed where possible through comparisons to the
available instrumentation data, particularly the settlement observed during the var-
ious stages of construction. Where suitable data was available, various in-situ and
laboratory test results were compared to simulated tests to demonstrate the validity
of the proposed models and parameters.
454
Q.105 - R.25
Fig. 9
Simulated Plate Load Tests Versus Construction Quality Records.
Essais simulés de charge de plaque par rapport aux enregistrements de suivi de
construction.
For the purposes of the static analysis of the safety of the wall, failure was
defined as a strain level in excess of 2% which, on the basis of laboratory testing,
was conservatively determined to be a level of strain in which no increase in hydraulic
conductivity was measured† . This damage criterion was selected considering the
performance requirements of the cut off wall as a hydraulic barrier. At these strain
levels structural weakening is not anticipated to represent a material concern. For
the purposes of the static safety assessment, material cracking was defined as
any significant increase in the hydraulic conductivity of the material, such as from
shear or tension fractures. It was assumed that a ten-fold increase in the specified
hydraulic conductivity of the Cement-Bentonite would represent the onset of damage
that would require remedial attention. It is also noted that Cement-Bentonite material
within the wall may experience local cracking in a limited area without impacting the
performance of the wall as a hydraulic barrier. In this sense, cracking of the Cement-
Bentonite does not necessarily mean that the wall will not function as an adequate
hydraulic barrier.
† a more extensive investigation is currently underway to assess the behaviour of the cement-
bentonite samples at greater strains to better define the threshold for material cracking of
the wall where material cracking is defined as a tenfold increase in the hydraulic conductivity
of the samples.
455
Q.105 - R.25
Fig. 10
Staged construction of the Ituango Dam.
Étapes de construction du barrage d’Ituango.
5.3. MODELLING METHODOLOGY
One of the unique features of the configuration of this wall is that it is essentially
floating within the upper portion of a large rockfill embankment. As such, it will
deform in accordance with the fill materials surrounding it. To assess the future
performance and safety of the wall, a deformation analysis was performed modelling
the actual construction stages (Figure 10) and the performance of the fills during
each of the phases, calibrated to monitoring data and construction records to assess
the interaction between the fill and the Cement-Bentonite cut off wall within the upper
section of the Ituango Dam.
Modelling was performed assuming the dam could be characterized by three

broad sections:
• Section 1: below elevation El 385.0, is a standard zoned embankment dam

with vertical clay core, downstream filter protection and rockfill shells.
• Section 2: the full priority zone (FPZ), between elevations El 385.0 and El 418.0
(33.0 m high) is a rockfill dam with a 6.0 m wide inclined clay core sloped at
1V:1.65H and a 1 m thick, 37.1 m high cement-bentonite cutoff wall. The cutoff
wall is embedded 5 m into the bottom part of the dam El 380.0.
• Section 3: 17.9 m above the full priority zone (FPZ), between elevations El
418.0 and El 435.0 is a zoned dam with 8 m thick vertical core, downstream
filter protection and upstream and downstream supporting rockfill shells.
Deformations were estimated by means of the Finite Element Method within a

two-dimensional and three-dimensional framework using PLAXIS 2D (Version 2018)
and PLAXIS 3D (Version 20.0.0.119) respectively for normal reservoir level loadings
immediately after the construction, rapid draw-down after the construction, rapid
draw-down 10 years after the construction - static/creep. The analysis was performed
under drained, or effective-stress, conditions for all materials. No time dependency
456
Q.105 - R.25
(i.e. consolidation) was considered in the deformation analysis. Pore pressures were
defined using a prescribed phreatic surface, including for the rapid drawdown cases.
The modelling program consisted of three main components.
1. A two-dimensional, plain-strain model of the full dam and foundation with

careful consideration for the potential variations in the stiffness of the vari-
ous materials at a location corresponding to the highest section of the dam
fill at an approximate chainage of 0+480. This included modelling the staged
construction of the dam (Figure 8) to compare measured deformations with
predicted deformations in order to calibrate the model.
2. A two-dimensional, plain-strain model of the dam above El. 360 m and cor-
responding to an approximate chainage of 0+480 where instrumentation was
available for comparing the actual and simulated deformations.
3. A three-dimensional model of the full width of the dam above El. 360 m to
assess deformation of the wall caused by deformation at the center of the
dam relative the abutment connections.
5.4. TWO-DIMENSIONAL ANALYSIS
The stress-strain response (i.e. stiffness) of the various materials is sensi-

tive to the local stress condition. Consequently, it was important to properly define
the stresses within the dam before and after the construction of the cutoff wall
and an accurate reproduction of the construction sequence was required. A total
of 19 phases were modelled starting with the pre-construction conditions the final
construction of the dam, impounding and a rapid drawdown simulation (Table 1).
Table 1
Construction and Model Stages.break Étapes de construction et de modélisation
Model Construction Stages Crest Reservoir Remarks
Phase(s) Elevation (m) Elevation (m)
1 Initial Phase N/A 200 Model Initiation.
2 Pre-Construction 240 200 Pre-construction.
3–12 Stages 1-11 285–385 220–240 Planned construction staging.
13, 14 Priority Zone 1, 2 410 393 Contingency fill placement.
15 Priority Zone 3 418 393 Cutoff wall installation.
16 Raise Sides to 418 418 393 Contingency fill placement.
17 Raise Crest to El. 435 435 393 Complete dam.
18 Impounding 435 420 Full reservoir level.
19 Rapid Drawdown 435 415 Maximum potential drawdown
The model setup is illustrated in Figure 11.
As illustrated in Figure 12 the predicted deformations at EL 360 m matched

the measured deformations reasonably well.
457
Q.105 - R.25
Fig. 11
Results of Two-Dimensional Analyses.
Résultats des analyses bidimensionnelles.
Fig. 12
Comparison of Measured and Predicted Settlement Within the Priority Fill Zone
During Construction from El. 418 to 435.
Comparaison des tassements mesurés et prévus dans la zone de remplissage
prioritaire pendant la construction entre les élévations 418 et 435.
5.5. POST CONSTRUCTION CREEP SETTLEMENT
After the completion of construction, the dam will continue to undergo settle-
ment due to creep in the rockfill and consolidation in the core. The magnitude of
this settlement is generally related to the rockfill type, degree of compaction and
the dam height. For well compacted and medium to high strength rockfill, the total
crest settlement ten years following construction is expected to vary from 0.15% to
0.40% of the dam height. For the purposes of analyzing the cement-Bentonite cutoff
wall, a total 10-year post construction vertical settlement of 0.30% H was assumed,
where H is the section height of the dam. As this estimate is approximate in nature,
458
Q.105 - R.25
with factors such as the size and consolidation of the clay core not accounted for, a
conservative assessment was performed to attempt to account for the influence of
these other factors.
5.6. SEISMIC SETTLEMENT ANALYSIS
At the time of preperation of this report, an updated seismic hazard assess-

ment in accordance with ICOLD Bulletin 148 was underway. For this reason, the
deformation analysis of the cutoff wall under seismic conditions was based on a
low-order assessment of the maximum crest settlement of the dam using an empir-
ical method proposed by Swaisgood (Figure 13). The objective of this analysis was
to provide an initial screening of the potential risk of cracking
Fig. 13
Reported Settlement as a Function of Height and Earthquake Magnitude for
Rockfill Dams [1].
Tassement observé en fonction de la hauteur et de l’ampleur du tremblement de
terre pour les barrages de remplissage des roches.
This method relates the maximum crest settlement of an embankment to the

peak ground acceleration (PGA) and magnitude (M) of the associated earthquake.
While this method is empirical, in the case of the Ituango Dam the method is consid-
ered to be conservative. For example, the precedent dams considered are generally
at least 50 years in age and were constructed at a time when rockfill compaction
459
Q.105 - R.25
was not performed to the extent it was done for the Ituango Dam. In addition, the
dams were constructed in wider valleys than Ituango and some of the dams were
founded on overburden. All of these factors would result in higher seismic induced
settlement than is likely to occur at Ituango.
5.7. THREE-DIMENSIONAL ANALYSIS
The stability of the dam has been well established by others during the orig-
inal design. The three-dimensional analysis was performed to assess the effects
of differential movement of the cut off wall relative to the abutments. For the pur-
poses of this analysis, any failure-plane within the embankment would be assumed
to be associated with a failure of the dam that was unrelated to the cutoff wall
function. Accordingly, no shear planes through the cutoff wall were considered for
the strain analysis. It is determined that, due to the flexibility about its weak axis,
upstream-downstream movement of the wall causes very little strain material. Ulti-
mately, differential vertical movement was determined to be the primary damage
mechanism for the cutoff wall
Based on these assumptions, the three-dimensional analysis was limited to the

upper portion of the dam in the priority raise zone. Figure 14 illustrates the conception
approach for modelling the differential settlement of the cutoff wall and the upper
portion of the dam. The lower bound of the 3D model includes prescribed values
of displacements, determined based on the findings outlined in Section 5.4 through
5.6. A comprehensive 3D deformation analysis of the entire dam and surrounding
features had been performed by others and continues to be examined . However,
a comprehensive model is not well suited for analysing the localized strain levels
within such a thin element as the cutoff wall, relatively to the dimensions of the dam
and surrounding features.
6. RESULTS OF ANALYSES
The results of the seismic and static analyses, as summarized in Table 2,

indicated the following:
1. No material cracking of the wall immediately following construction under

normal reservoir conditions is anticipated.
2. No material cracking of the wall immediately following construction under a
rapid draw-down conditions is anticipated.
3. No material cracking of the wall 10 years following construction and thereafter
under normal loading conditions is anticipated.
4. No material cracking of the wall under an Operating Basis Earthquake (OBE)
is anticipated.
460
Q.105 - R.25
Fig. 14
Conceptual Illustration of the Three-Dimensional Analysis.
Illustration conceptuelle de l’analyse tridimensionnelle.
5. Some material cracking of the wall under a Safety Evaluation Earthquake

(SEE) is anticipated. This could result in an increase in seepage but would
be classified as repairable damage. As such the wall complies with current
ICOLD standards.
Table 2
Summary of the Results of Assessments.
Résumé des résultats des évaluations.
Description Peak Maximum Remarks on Tension
Ground Strain Cracks
Acceleration (ε1 ), %
Post Construction NA 0.59 Very limited area impacted, only at
the top edge of the cutoff wall.
Post Construction, with Rapid NA 0.44 Very limited area impacted, only at
Drawdown the top edge of the cutoff wall.
10-Years Post Construction (0.30%H) NA 1.10 Very limited area impacted, only at
OBE Seismic (0.07%H) 0.15 1.30 Very limited area impacted, only at
SEE Seismic (1.03%H) 0.61 5.21 Areas of potential cracking near
the left and right abutments.
As is illustrated in Figure 15, the potential for material cracking of the wall was
determined to be highest at the right abutment contact. At both abutments, a triple
cutoff wall sealed against a concrete abutment wall was installed which helps to
reduce the risk.
461
Q.105 - R.25
Fig. 15
Estimated locations within the cutoff wall in tension under SEE induced dam
settlement.
Estimation des emplacements en tension à l’intérieur du parafouille dû au
tassement du barrage induit par l’ESE.
7. CONCLUSIONS
The decisive actions of the owner, designer, and contractor to divert water
through the unfinished powerhouse and to change the design of the dam to allow
for fast track construction of the upper portion of the Ituango dam prevented a
catastrophic over topping event. Since construction, the modified design of the
embankment has performed successfully.
To assess the future safety of the dam a deformation assessment was

undertaken for the cement-bentonite cutoff wall using the Finite Element Method
within both a two-dimensional and three-dimensional framework. The results of this
analysis indicate that:
1. In general, the localized strain levels within the cutoff wall are highest near the
abutments, and particularly near the right abutment. This is due to asymmetric
cross section of the valley at the dam location.
2. Material cracking of the wall under static loadings now and under the effects
of future settlement is not expected.
3. Cracking of the wall due to strain compatibility between adjacent materials of
different stiffnesses is not expected.
4. Material cracking under the effects of the safety evaluation earthquake (SEE)
would be expected.
Overall, the results of the analyses show that cracking of the cutoff wall will
likely be associated with differential settlement of the wall at the center of the dam,
relative to the abutments. Due to the asymmetric shape of the valley cross section,
any cracking of the cutoff wall that might occur is most likely to be located near the
right abutment which allows for any remedial treatments that may be required to be
focused in this location. While cracking of the cement-bentonite wall was determined
to be a possibility under extreme earthquake loadings, this would not be expected
to lead to an immediate catastrophic failure of the dam. The increased seepage,
if left untreated, could, over time, lead to erosion of the core materials beneath or
above the wall which could in turn result in a relatively slowly developing failure
mode associated with internal erosion. At present, there is insufficient information
462
Q.105 - R.25
to quantify the extent of cracking and quantify the increase of conductivity in order
to determine if this increased seepage represents a material risk‡ . However, any
risk that may exist can be mitigated by lowering the reservoir, at a controlled rate,
and undertaking remedial measures, such as cement-bentonite grouting, to re-seal
the area following a major seismic event.
Overall, the results of the analyses indicate that the Ituango dam is safe. The
wall is currently functioning as intended and would only be expected to exhibit any
material cracking under the effects of the SEE earthquake. This damage would not
be expected to lead to sudden failure of the dam allowing repairs to be performed.
Ongoing assessments include;
1. Establishment of alarm levels for the inclinometers using the project Triple
“A” alarm system (alert, alarm and action levels) and all other current instru-
mentation within the dam that are related to the potential for damage of the
Cement-Bentonite wall.
2. Assessment the strain-permeability characteristics of the cement-bentonite
materials to verify threshold strain levels where unacceptable increases in
permeability will occur .
3. Performance of a Probable Failure Modes Analysis in accordance with FERC’s
standards to verify instrumentation requirements for future monitoring and
contingency plans.
4. Performance of dynamic triaxial tests
5. Performance of dynamic analyses un the effects of the SEE event
REFERENCES
[1] Swaisgood, J.R. (1998) Seismically-induced deformation of embankment

dams, 6th US National Conference on Earthquake Engineering, Seattle,
Washington.
[2] Swaisgood, J.R. (2003) Embankment dam deformations caused by earth-

quakes. Pacific Conference on Earthquake Engineering, paper No. 14.
[3] Swaisgood, J.R.(2014) Behavior of embankment dams during earthquake,

Journal of Dam Safety, ASDSO, 12(2), 35–44.
[4] Fell, R., MacGregor, P., Stapledon, D., Bell, G., Foster, M. (2015) Geotechnical
Engineering of Dams, 2nd Ed. Taylor and Frances Group, London, UK.
[5] Gazetas, G., Dakoulas, P. (1992) Seismic analysis and design of rockfill dams:
state-of-the-art., Soil Dynamics and Earthquake Engineering, 11, pp 27-61.
‡A comprehensive dynamic assessment of the dam under SSE loadings and strain-
permeability testing of the cement-bentonite material is underway.
463
Q.105 - R.26
-------
-------
LESSONS LEARNT FROM THE SADDLE DAM D FAILURE OF XE-PIAN

XE-NAMNOY PROJECT IN LAOS PDR (∗)
Jean-Pierre TOURNIER
Expert Advisor,
HYDRO-QUÉBEC
Ahmed F. CHRAIBI
Independent dam consultant, Morocco
Anton J. SCHLEISS
Honorary professor and expert consultant,
ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE (EPFL), Switzerland
CANADA
SUMMARY
The failure of the Saddle Dam D of the hydroelectric project named Xe-Pian
Xe-Namnoy in Laos PDR, occurred on July 23, 2018 during the first filling of the
reservoir. The Independent Expert Panel (IEP) mobilised by the Government of Laos
PDR, found that the root cause of the failure is related to the high permeability of the
foundation combined with the presence of erodible horizons in connection with the
existence of canaliculus interconnected paths. The foundation was without doubt
involved in the failure, which could have been prevented by adequate treatment
measures.
Based on the report of the IEP the paper will first outline the result of the
analysis of failure incident and generalize it for other projects. Then the lessons
learnt will be addressed in detail by discussing the design and construction of the
∗ Leçons tirées de la rupture de la digue de col D du projet Xe-Pian Xe-Namnoy au Laos
464 DOI: 10.1201/9781003211808-72

Q.105 - R.26
saddle dam; the adequacy of the monitoring system and the follow-up during the
first filling; the geotechnical site and laboratory investigations; the urgency planning
with site accessibility and finally the provisions for the reservoir level control. The
lessons learnt will be presented in a way that such failure incidents can be avoided.
RÉSUMÉ
La rupture de la digue de col D du projet hydroélectrique Xe-Pian Xe-Namnoy

au Laos s’est produite le 23 juillet 2018, lors du premier remplissage du réservoir.
Le panel d’experts indépendants (IEP) mobilisé par le gouvernement du Laos pour
l’analyse technique de cet incident, a considéré que la cause première de la rupture
est liée à la forte perméabilité de la fondation de la digue, associée à la présence
d’horizons érodables en relation avec l’existence de canalicules interconnectés. La
fondation a sans aucun doute été impliquée dans la rupture, laquelle aurait pu être
évitée si des mesures adéquates de traitement de la fondation avaient été prises.
Sur la base du rapport de l’IEP, le présent article présente dans un premier

temps les résultats de l’analyse de la rupture et la possibilité de leur généralisation
à d’autres projets. Ensuite, les leçons tirées de l’incident sont abordées en détail
tant en ce qui concerne la conception et la construction de la digue, l’adéquation du
dispositif d’auscultation, le suivi pendant le premier remplissage, les investigations
géotechniques in-situ et au laboratoire, le plan d’urgence y compris l’accessibilité
des sites et enfin les dispositions pour le contrôle du niveau du réservoir. Les leçons
tirées sont présentées de manière que de tels incidents de rupture puissent être
évités à l’avenir pour des projets réalisés dans un contexte comparable.
1. INTRODUCTION
1.1. PROJECT OVERVIEW
The Xe-Pian Xe-Namnoy Hydropower Project (XPXN HPP) is located in south-

ern Laos P.D.R. on the Bolaven Plateau, approximately 550 km to the southeast of
the capital Vientiane, 80 km to the east of Pakse on the Mekong River, and 35 km
to the west of Attapeu, a town located on the Xe-Kong River below the plateau. As
shown in Fig. 1 XPXN HPP consists of a main large storage reservoir impounded by
a dam on the Xe-Namnoy River. From that reservoir, the water is transported by a
long underground waterway in order to gain a head of some 650 m, used for hydro-
electric generation in an open-air powerhouse at the foot of the Bolaven Plateau.
From the powerhouse, the water is released into the Xe-Kong River by a tailrace
channel.
465
Q.105 - R.26
In order to increase the inflow to the Xe-Namnoy reservoir and consequently

the hydropower generation in the powerhouse, the runoff from two neighboring
watersheds, Houay Makchan and Xe-Pian, is diverted into the main reservoir. The
diversion is made in cascade, namely from the Houay Makchan catchment to the
Xe-Pian reservoir, which also stores the water from the Houay Namliang River and
from there to the Xe-Namnoy reservoir. This diversion system comprises a bottom
intake on the Houay Makchan and a dam and intake on the Xe-Pian River. Two
transfer conduits, one working under free flow from Houay Makchan to Xe-Pian and
the second one under both pressurized and free flow conditions from Xe-Pian to
Xe-Namnoy ensures the transfer of the water to Xe-Namnoy reservoir.
From the Xe-Namnoy reservoir the water is conveyed to the powerhouse units
by means of a power intake, 13.6 km of low-pressure headrace concrete lined tunnel,
475 m of vertical shaft, 1,590 m of high-pressure tunnel, 810 m of penstock including
the units flow distribution steel manifold.
The powerhouse with three Francis units and one Pelton unit has an installed
capacity of 410 MW. From the powerhouse, the water is released into the tailrace
channel which restitutes the flow some 6.33 km downstream in the Xe-Kong River.
Fig. 1
Overview of XPXN HPP showing the main features of the project.
Situation de l’aménagement XPXN avec les composantes principales du projet.
466
Q.105 - R.26
The Xe-Namnoy Dam and reservoir stores water from Xe-Namnoy River as
well as the diverted water from the Xe-Pian reservoir with total storage capacity
of 1,043 million m3 with Full supply level (FSL) and Maximum water level (MWL)
respectively at 788.5 m and 791.70 m. The Xe-Namnoy Dam’s crest level is at
elevation 794.0 m resulting in a maximum height above foundations of 73.7 m. The
crest length is ∼1,600 m. The spillway designed for a PMF of 4,000 m3 /s consists
of an arched labyrinth weir with crest level at 788.5 m followed by chute and a flip
bucket.
In addition to the main Xe-Namnoy Dam, five saddle dams are necessary to
keep the reservoir within the full supply level of EL. 788.50 m. The crest level of
the saddle dams corresponds with the crest level of the main Xe-Namnoy Dam,
EL. 794.0 m. This leads to maximum saddle dam heights of 17 m above natural soil
ground.
1.2. SADDLE DAM D – AS BUILT DESIGN
For Saddle Dam D (SDD) a 1:2.5 sloping homogeneous earth-fill dam with an
increased crest elevation at 794.0 m was adopted as shown in Fig. 2 and Fig. 3.
After removing 1 m of the top natural soil ground, SDD is founded with a maximum
height of some 17 m directly on laterite soil-like formation reaching almost 20 m
thickness.
Fig. 2
Plan view of Saddle Dam D as Built.
Vue en plan de la digue de col D conforme à l’exécution.
As the safety factor against slope instability didn’t reach the required values
(1.5), the upstream and downstream toe at their highest section were loaded by a
small counterweight fills (see Figs. 2 and 3).
467
Q.105 - R.26
Fig. 3
Section A-A and Detail A of Saddle Dam D as built.
Section A-A et détail de la digue de col conforme à l’exécution.
The homogeneous earth-fill SDD comprises the following zones:
• Zone 1 consists of selected impervious lateritic material as used for the core
material of main Xe-Namnoy Dam.
• Zone 2C is a 0.5 m thick transition zone between the riprap (Zone 3C) and the
pervious fill (Zone 3D) respectively and the dam material (Zone 1).
• Zone 2A is a 0.5 m thick fine filter between laterite soil and the toe drain.
• Zone 3C protects the upstream face of dam by 1.5 m thick riprap against
erosion by wave action in the reservoir, rain and surface runoff.
• Zone 3D protects the downstream face by a 1.5 m thick pervious fill, “almost
the same as a rockfill”, against rain and surface runoff. The pervious fill is a
dumped fill with generally required but not specified compaction.
• Zone 3B at the downstream toe of the embankment was provided to control
the seepage through dam body and foundation. Toe drain consisted of the
same rockfill material as used for main Xe-Namnoy Dam.
2. ANALYSIS OF FAILURE INCIDENT
2.1. RESERVOIR LEVEL
The reservoir filling begun in April 2017 and went according to the follow-
ing graph (Fig. 4). The reservoir stabilized at el. ∼781 m corresponding to the
upstream toe of SDD, during the 2017/2018 dry season. It restarted raising in
May up to July 2018. The reservoir water level rapid drawdown corresponds to the
SDD failure and occurred on July 23, 2018 under reservoir elevation of 789.35 m
468
Q.105 - R.26
Fig. 4
Evolution of rainfall and Xe-Namnoy reservoir elevation from April 2017 to July
2018 together with the intended impounding curve.
Évolution des précipitations et du niveau du réservoir Xe-Namnoy d’avril 2017 à
juillet 2018. En bleu la courbe prévisionnelle de remplissage.
(0.85 m above the FSL) and 4.65 m below SDD original crest level. This observa-
tion allows for the statement that the failure was not triggered by the embankment
overtopping.
2.2. SEQUENCE OF FAILURE
The sequence of failure presented here is based on a careful analysis of avail-

able photos of the incident. According to the neighboring villagers, the first cracks
were observed on July 22, 2018 at ∼12:00, and the total breaching of the SDD
occurred on July 23, 2018 at 17:30. No relevant information related to these cracks,
providing information on their location, development with time, opening or offset is
available (photos, rough survey or even a hand sketched drawing).
Derived from available information, i.e. photographs, together with piezometric

monitoring data and foundation quality appraisal, the IEP concluded that the most
likely scenario of the failure went according to 3 main stages as presented in the
following sketches (Figs. 5 to 7 below):
469
Q.105 - R.26
Fig. 5
Stage one: resurgence development downstream and regressive erosion starting.
Phase 1: apparition de résurgences à l’aval et développement de l’érosion
régressive.
Fig. 6
Second stage: sliding triggering as the foundation beneath the downstream toe
became too loose and so weak that it couldn’t mobilize any shearing strength.
Phase 2: déclenchement du glissement dès que la fondation dans l’emprise du
pied aval est devenue meuble et si fragile qu’aucune résistance au cisaillement ne
pourrait être mobilisée.
470
Q.105 - R.26
Fig. 7
Final stage: acceleration of the sliding movements and breaching of the upstream
small remaining edge made of protections and clayey material. Embankment
washing out took over sliding.
Phase finale : accélération des mouvements de glissement et rupture du coin en
matériau argileux subsistant au couronnement. Le déversement qui s’en est suivi
a emporté le remblai.
Fig. 8
Synthetic representation of the failure scenario.
Représentation synthétique du processus de rupture.
Of course, the reality is much more complex than the 2D sketches presented
hereinabove. Indeed, the instability involved large volumes of embankment and
foundation that should be analyzed in a 3D view as they are not moving all in the
same direction. A tentative of visualization of the numerous horizontal and vertical
movements preceding the SDD breaching is presented in Fig. 8 above.
471
Q.105 - R.26
2.3. ROOT CAUSES OF FAILURE
The IEP inspection of the foundation of the Saddle Dam D led to the following
observations:
• The foundation is made of residual formation (Laterite) sitting on a basaltic

bedrock which is sound but fractured. In its interface, weathered bedrock in
the stage of becoming lateritic soils is observed. Such geological profile is
quite common in tropical regions. The boundaries between these formations
are very variable depending on the erratic development of the weathering.
• At the scale of the dam site, the laterite is very heterogeneous which takes
various facies:
◦ Clayey silt to silty clay very often with canaliculus and vegetal roots up
to 10 m deep. Canaliculus are prone to convey water and lead to per-
meability values much higher than those obtained in boreholes as the
walls of the hole are smoothed by the drilling equipment.
◦ Coarse to fine sandy silt in continuous layers of several meters thick-
ness. These facies, uncovered in the left abutment trench, seems to
have relatively high permeability. It exhibits a cohesion that decreased
dramatically after saturation and should be very sensitive to erosion.
◦ Open material made of millimetric to centimetric elements with voids and
silt in between, along with evidence of water seepage. Its permeability
can be high to extremely high. Suffusion is likely to occur in these facies.
◦ Passages containing large boulders or colluvium like zones are present.
◦ Shearing surfaces are present either in the clayey formation or in the
facies under weathering.
The analysis of available monitoring data together with observations made dur-
ing the IEP site visit show that the horizontal permeability of the foundation is much
higher than that considered in the detail design. With saturation line almost outcrop-
ping at the downstream toe and a relatively steep declining topography downstream,
water resurgences are highly likely to occur. As an immediate consequence, the
silty-sand to sandy-silt identified in the foundation is prone to erosion. This results
in concentrated seepage path which developed and collapsed from time to time.
The distribution and the dimensions of the seepage paths may vary depending
on the sensitivity of the foundation to erosion. There is an evidence that a similar
phenomenon initiated in Saddle Dam E having a similar foundation and design.
The stability of the embankment is jeopardized when the erosion of the foun-
dation spreads upstream (backward erosion) combined with a significant softening
of the lateritic formation never exposed before to water submersion, seepage or
embankment loading.
In these conditions, it has been considered that the root cause of the incident
is the sensitivity to erosion of the foundation together with its high permeability
attributed to the presence of sandy horizons and canaliculus interconnected paths.
472
Q.105 - R.26
3. LESSONS LEARNT
3.1. DESIGN AND CONSTRUCTION OF SADDLE DAM D
The Basic Design took into account the uncertainty of the foundation by imple-
menting a cut-off trench and systematic consolidation grouting which was then
omitted in the detail design based on a more optimistic evaluation of the perme-
ability of the laterite. However, such a cut-off trench without connection to the sound
rock would not have been sufficient to reduce the risk of preferential seepage paths
in the lateritic foundation. On the other hand, grouting, in the proper sense of the
word “grouting” is not adapted to the soil like lateritic foundation. The feasible tech-
nical way to connect the cut-off trench safely with the rock foundation would have
been a combination of a positive cutoff with a grouting of the fractured bedrock.
The positive cutoff could have been obtained either by excavation and backfill using
impervious compacted material or by a diaphragm wall made of plastic concrete.
The slope stability analysis during the detailed design were performed using
SEEP/W for seepage calculations and SLOPE/W for slope stability calculations. Four
load cases were considered, namely: (1) End of the construction and during con-
struction; (2) Impoundment and Full reservoir (Steady seepage); (3) Full reservoir
with Operation Basis Earthquake; (4) Rapid drawdown. Furthermore, a traditional
seepage analysis was carried out in order to check the potential of piping i.e. internal
erosion. Regarding the design the following observations may be highlighted:
• In the lateritic foundation or embankment, the possibility of pore pressure

buildup during construction has not been discussed.
• Input parameters, especially internal friction, have been considered without
addressing the range of uncertainties.
• The apparent cohesion of the foundation based on SPT testing should be
considered with caution as it is known that lateritic soil-like formation soft-
ens when saturated during a long period of time. This was revealed by the
significant lower N-values obtained by post-failure geotechnical investigations
carried out under higher moistening conditions.
• Permeability tests in boreholes in laterite give misleading results since the walls
of boreholes are smeared and becomes tighter. More reliable permeability
values are obtained from soaking tests in-situ pits, particularly when dealing
with coarse elements or low cohesion sandy passages as mentioned hereafter.
• More reliable insight on laterite characteristics can be obtained only from
observation in open trenches and shafts.
• It is known that compacted embankment has a significant higher horizon-
tal permeability than vertical permeability (typically 10 times). This may also
be applied to the natural laterite soils provided that more permeable hori-
zons are identified. Thus, an anisotropic permeability in the embankment as
well as in the laterite foundation should have been considered in the design.
Furthermore in view of its natural heterogeneity, it was not reasonable to take
473
Q.105 - R.26
for the foundation a permeability value which is lower than for the compacted
embankment.
• In order to take into account the uncertainty of the material characteristics and
also to consider the possibility of pore pressure buildup during construction,
sensitivity analysis would have been required in the detailed design.
According to the available information, SDD embankment seems to have

been adequately built, that means compacted and supervised during construction,
according to the common practice. On the remaining part of the embankment on the
left bank, clear compaction layers with different composition are visible as observed
in the open trenches on the left and right sides (Fig. 9). This indicates that the laterite
Fig. 9
Trench excavated in the left abutment through the remaining embankment. Top
left: the interface between the embankment and the foundation is well materialized
by the horizontal crack; Top right and bottom: the layers of the embankment may
be recognized by their different colors and/or horizontal cracks development.
Tranchée excavée dans la partie restante de la digue en rive gauche. En haut à
gauche : l’interface entre le remblai et la fondation est bien marquée par la fissure
horizontale; en haut à droite et en bas : les couches du remblai peuvent être
reconnues par les couleurs différentes et/ou le développement des fissures
horizontales.
474
Q.105 - R.26
material has been transported from the borough area, without mixing the fine and
coarse nuances, directly to the construction site.
3.2. MONITORING SYSTEM AND FOLLOW-UP DURING FIRST FILLING – DATA

INTERPRETATION
The monitoring system installed on the saddle dams is deemed insufficient for
embankments of this importance (Fig. 10). It is made of 3 geodesic points and 4
stand pipe piezometers. It is reminded that according to ICOLD, any dam of more
than 15 m height or controlling more than 1 million m3 , should be classified as a
large dam, which requires a certain level of surveillance (ICOLD Bulletin 158).
Fig. 10
Monitoring system installed on SDD: 3 geodesic points (triangles on the crest, at
ST0+200, +400 and +650) and 4 stand pipe piezometers (GW1 and GW3 on the
highest section and GW2 and GW3 in the left abutment). No geodesic survey was
installed on the highest section of the dam.
Dispositif d’auscultation installée sur SDD : 3 repères géodésiques (triangles sur le
couronnement, au PM 0+200, 0+400 et 0+650) et 4 piézomètres (GW1 et GW3
sur la section de plus grande hauteur et GW2 et GW3 sur l’aile gauche). Aucun
repère géodésique n’était installé sur la section de plus grande hauteur.
The geodesic survey at the crest of SDD, station 0+400, where the first signs of
the failure were observed (Fig. 11) shows that during the first phase of impoundment
(2017), the settlements increased almost uniformly with the rising of the reservoir
level, reaching 0.14 m, and then stabilized with the reservoir level. Horizontal move-
ments were very small, in the range of the measurements’ accuracy. During the
second phase of the impoundment in 2018, the movements started again, imply-
ing this time also the horizontal ones. On July 18, a movement acceleration was
observed in all directions, with a settlement exceeding the “warning threshold”, as
set by the designer. Both other geodesic survey points exhibited of suspicious move-
ments of the crest few days before the failure. No specific action was taken although
a dramatic acceleration of the settlement was observed.
475
Q.105 - R.26
Fig. 11
Geodetic movements recorded at Saddle Dam D at station 0+400.
Mouvements géodésiques enregistrés à la digue de col D au PM 0+400.
The stand pipe piezometers GW1 and GW2, installed on SDD, are crossing
the soil foundation on its entire depth with their bottom located on the bedrock. At
their installation in December 2016, they indicated a water level which correspond
likely to the natural water table elevation as for both the water level was at a similar
depth compared to the natural ground.
After the stabilization of the reservoir water level following the first impound-
ment, during the dry season between October 2017 and May 2018, both piezometers
stabilized but with an increase in the water table elevation higher by ∼ 6m for GW1
and ∼ 4m for GW2, compared to the situation before impoundment. During the sec-
ond phase of the reservoir filling beginning in June 2018, GW1 and GW2 reacted
probably to the rainfall but also to the reservoir rising. However, GW1 and to a lesser
extent GW2, exhibited a decline of few meters in the beginning of July and on July
20th . While the first drop of the piezometer may be related to the lack of rain during
the last days of June, the second one, corresponding to the latest available reading,
occurred after 9 rainy days, totalizing more than 350 mm. This type of behavior may
be attributed to the development, by regressive erosion, of open flow paths allowing
for free water release and resulting in a drop of hydrostatic pressure (Fig. 12).
However, on July 18th , 2018, the ground water level in the observation wells
reached or even exceeded the ground level, along with the acceleration of the set-
tlements, not only for SDD but also for saddle dams E and F, and no action was
taken. At least a thorough inspection by a specialist of the downstream area should
have been conducted on each saddle dam. Normally, it is a requirement to inspect
regularly (at least every day) above all during the first impoundment the structure
and particularly the toe of the dam. Vegetation should have been removed to access
476
Q.105 - R.26
easily the downstream zone subject to potential resurgences development. Indeed,

at the downstream toe of Saddle Dam E, seeping flow along with an initiation of
erosion was observed.
Fig. 12
Piezometers water level variations with time at Saddle Dam D.
Variations des niveaux d’eau dans les piézomètres de la digue de col D.
3.3. GEOTECHNICAL SITE AND LABORATORY INVESTIGATIONS
As far as the geotechnical issues are concerned, the lessons derived from the
Saddle Dam D failure may be summarized as follows:
• Foundation investigations in this type of material should include digging deep

trenches, allowing for a direct observation of the foundation quality, particularly
its homogeneity and the presence and shape of canaliculus. Boreholes may
provide a misleading information on the permeability as drilling tool smear the
hole plugging any eventual canaliculi.
• Foundation permeability appraisal should rely on soakage tests conducted at
different levels, while digging investigation trenches.
• In-situ testing (SPT or CPT for example) should consider the future saturation
of the foundation when tests are conducted on unsaturated material. Corre-
sponding interpretation tools used for transported soils are not necessarily
applicable to tropical residual soils.
• Geophysical investigations provide a valuable information to be considered in
a different way compared to transported soils.
Indeed, engineering references related to tropical residual soils’ geotechnical

investigations and interpretation are extremely limited and deserves to be further
477
Q.105 - R.26
investigated. ICOLD Bulletin 151 published in 2009, dedicated to this topic is remark-
ably interesting and has filled a large gap in ICOLD publications. However, it has
been decided to set a new Technical Committee to update and complete this bul-
letin by the more recent experiences in dam construction and operation over the last
decade, founded and/or constructed using tropical residual soils.
3.4. URGENCY PLANNING – ACCESSIBILITY
All the saddle dams were not considered in the Emergency Action Plan (August
2017) of the XPXN project prepared by the Contractor. This document focused only
on main dams Xe-Pian and Xe-Namnoy and the Houay Makchan Weir. In principal
any emergency action plan has to consider all critical structures of a project.
After the observation of the first cracks in the dam crest, or even before when
the settlements’ speeding up occurred (July 13, 2018), indicating that the situation
is concern to the dam safety, there was no attempt to investigate the reasons of this
behaviour and take appropriate actions.
A trial action would have been to start immediately to reinforce the deepest
toe region by dumping of permeable rock material. It seems that access during
heavy rain was difficult, but any embankment dam should be accessible even during
extreme conditions at critical location as toe regions not only for visual inspection
but also for emergency measures. Normally during impoundment, the contractor as
well as the impoundment surveillance team has to be prepared for such emergency
measures.
3.5. RESERVOIR LEVEL CONTROL
There was no possibility to control or to lower the level of the reservoir when
the failure of Saddle Dam D was announced by the cracks. The bottom outlet has
only a small capacity mainly for release of environmental flow. In principle, the size
of the diversion tunnel (7 m Dia.), together with a gate chamber, would have allowed
to install a bottom outlet with two gates having a total surface of about 15 m2 .
Such a bottom outlet would have had a significant higher capacity (at least about 5
times) than the actual outlet. Nevertheless, even such a bottom outlet capacity would
probably not have had the capability to lower the reservoir sufficiently to prevent the
failure. The spillway is an ungated weir i.e. labyrinth weir. In the case of embankment
dams, especially a reservoir contained by a main dam and several saddle dams, an
ungated spillway is a risky conception considering the consequences after failure
not only of the main dam but also of one of the saddle dams. Besides the option
of a bottom outlet with sufficient capacity, an at least partially gated spillway would
have allowed to control and lower the reservoir level in the most upper critical part
and therefore to prevent the Saddle Dam D failure.
478
Q.105 - R.26
4. CONCLUSIONS
The lessons learnt from the failure incident of the Saddle Dam D of the XPXN
HPP project can be summarized as follows:
1) Laterite foundations especially on saddle topography are delicate and are

likely to be very heterogeneous and contain canaliculus involving high per-
meability with occurrence of water resurgence and concentrated leaks,
which may trigger internal erosion. Furthermore, Laterite formations are
residual soils, which may have a potential of collapse if they are not
permanently saturated.
2) In Laterite formations, observations and tests from boreholes alone can-
not give the full picture of the material characteristics required for a safe
design, particularly when the volume of investigations is insufficient and/or of
non-adequate quality. Observations and tests in open excavated exploratory
trenches and shafts, along with in-situ testing are mandatory.
3) When interpreting in-situ test results, the level of soil saturation should be
considered in comparison with the situation after impoundment. The problem
arises especially in the saddles where the ground is often kept dry for a long
period during the year.
4) The most adapted seepage control in this type of formation is a positive
cutoff along with efficient downstream drainage system.
5) Slope stability assessment has to take into account the uncertainties involved
with materials’ characterization and sensitivity analyses are required in the
design calculations.
6) The first filling of the reservoir is among the most critical phases. Experi-
enced dam monitoring specialist have to be permanently on site, aiming
at detecting and making a correct interpretation of any anomaly of dam
behavior during first impoundment and take timely action in order to avoid
catastrophic incidents.
7) Threshold values, which are critical in the assessment of the dam safety
during its operation, should be evaluated carefully taking into account all
relevant parameters governing the dam and its foundation behavior.
8) All scenarios of failure should be considered before the reservoir filling and
necessary actions, in case this eventuality became probable, should be
listed. Among them, threated areas in case of failure of any component
of the project should be determined.
9) The access to critical infrastructures like saddle dams retaining a major part
of the reservoir has to be ensured even during extreme meteorological events
not only for visual inspection but also technical intervention.
10) Saddle dams, which failure have catastrophic consequences by releasing
uncontrolled significant volume of the reservoir, have to be treated with the
same level of safety requirement from design to operational phase like the
main dam. Saddle dams are even more critical than the main dam as a
479
Q.105 - R.26
failure would drain in areas with no stream, highly endangering the unaware
population located downstream the saddle.
11) Any reservoir, especially contained by several embankment dams, must
have water release structure which is not only able to control the level of
the reservoir during first impoundment but also has the capacity to lower the
level in case of subsequent danger.
12) Large hydropower and dam projects have to be accompanied by an inde-
pendent panel of experts, which ensures the required high safety standards
according to international best practice from early design stages to the
construction project.
ACKNOWLEDGEMENTS
The IEP thanks the Government of Laos for the strong support through the
National Investigations Committee (NIC) and the excellent organization of the site
visits as well as for the fruitful collaboration. The geological and geotechnical site and
laboratory investigations as recommended by the IEP have been planned, organized
and supervised by TEPCO supported by Pöyry - AFRY, which was involved also in
the preparation of the IEP framework and the first mission of Prof. Anton Schleiss.
The IEP appreciates the efficient and timely collaboration of the EPC Contractor (SK)
with the Panel and the cooperative communication towards the requests of IEP.
The IEP report is available on ICOLD website in the “MEMBERS SECTION”

under Circular Letter No. 1953 with link “Circ. 1953”.
480
Q.105 - R.27
-------
-------
MUSKRAT FALLS PROJECT, REPAIRING THE UPSTREAM COFFERDAM (∗)
G. SNYDER & Regis BOUCHARD

SNC-LAVALIN INC., St. John’s, Newfoundland and Labrador
A. RATTUE
SNC-LAVALIN INC., Montreal, Quebec
S. O’Brien
LOWER CHURCHILL MANAGEMENT CO., St. John’s, Newfoundland and
Labrador
CANADA
SUMMARY
A 25 m high upstream cofferdam was constructed to permit the diversion of

the lower Churchill River for construction of the North Dam. The ice management
program also required raising of the head pond prior to the winter 2016-2017. During
the impoundment in fall 2016, internal erosion manifested before reaching the target
water level (25m). This necessitated lowering the reservoir level back to natural levels
to undertake repairs.
This paper describes the repair works, primarily various techniques of grout-
ing. The work was still carried out with a hydraulic head across the structure and in
extreme winter conditions but was successful in allowing the impoundment to con-
tinue to the required interim level. Techniques included the GIN method for pressure
grouting in rock, high-mobility grout at the till-rock interface and low-mobility grout in
areas of low compaction or voids in the fill.
∗ Projet de Muskrat Falls – réparation du batardeau amont
DOI: 10.1201/9781003211808-73 481

Q.105 - R.27
RÉSUMÉ
Un batardeau amont d’une hauteur de 25 m fut construit afin de permettre la

dérivation de la rivière et la construction du barrage nord. La gestion des glaces a
nécessité le rehaussement du bief amont à l’avance de la saison hivernale 2016-
2017. Pendant la mise en eau à l’automne 2016, un épisode d’érosion interne s’est
manifesté avant d’atteindre le niveau d’eau ciblé. Un rabattement immédiat du bief
amont a été entrepris afin d’effectuer des réparations.
Cet rapport décrit les travaux de confortement, surtout les diverses techniques
d’injection. Les travaux ont été réalisés en présence d’une charge hydraulique et
en hiver. Néanmoins, les réparations ont réussi et la mise en eau a pu être reprise.
Les techniques d’injection comprenaient la méthode GIN pour l’injection du rocher,
l’injection de l’interface aussi avec un coulis de haute mobilité. Quant au remblai
affecté par l’érosion interne, une injection de mortier à mobilité réduite a été favorisé
pour le remplissage des vides et la compaction du matériau.
1. INTRODUCTION
The Muskrat Falls Hydroelectric Development forms part of the Lower Churchill
Project (LCP) in Newfoundland and Labrador, Canada, and is located on the
Churchill River (Fig. 1), about 291 km downstream of the Churchill Falls Hydroelectric
Development. The installed capacity of the Muskrat Falls facility will be 824 MW –
four units of 206 MW each.
Fig. 1
Location map
Carte de situation.
482
Q.105 - R.27
Construction of the roller compacted concrete (RCC) North Dam required

diversion of the river from its natural channel. Upstream and downstream coffer-
dams were constructed. Shortly after partial raising of the head pond was initiated,
incidents of high-volume seepage and material erosion occurred in the upstream
cofferdam. As such incidents could be a premonition of more serious erosion poten-
tially jeopardizing the integrity of the structure, the head pond was lowered, the
situation investigated, and remedial works undertaken.
This paper presents the history of the design, construction and remedial works,
and the eventual return to service of the cofferdam facilitating the construction and
commissioning of the North Dam.
2. GENERAL PROJECT DESCRIPTION
At the site, the river flows in a rock-controlled channel to the south of the Rock
Knoll. Prior to river diversion, the two falls (Muskrat Falls), about 1 km apart, caused
a drop of about 14 m in the Churchill River level from el. 17 m on the upstream side
to el. 3 m on the downstream side of the falls (Fig. 2). The development comprises
an RCC North Dam, a gated overflow spillway, the powerhouse and a rockfill dam
which closes the reservoir on the right or south bank just above the lower falls. The
switchyard and the converter station are located southeast of the powerhouse. On
the opposite (north) side of the Rock Knoll, a wedge-shaped piece of land forms what
is referred to as the North Spur; a natural dam which contributes to the reservoir
retention.
Fig. 2
Site view
Aménagement général.
483
Q.105 - R.27
3. GEOLOGICAL CONTEXT
The region of the site is underlain by an eroded Precambrian gneiss. At some

time in the geologic history, the river flowed through a deep gorge to the north of the
Rock Knoll. The main landforms date from the last glaciation. With the final retreat
of the ice and during a subsequent period of rising sea levels, the gorge was filled
by a series of lower sands and gravels, marine and alluvial deposits. As the land
rose due to glacial rebound, the river cut into the overburden, sometimes to bedrock
but, flow and erosion patterns guided the river to form the present valley alignment
passing to the south of the Rock Knoll.
The lower Churchill River hydro potential was identified in the early 1960’s. Site
reconnaissance and studies began in 1965 but were of limited scope. A comprehen-
sive field investigation program was undertaken in 1979 including geological map-
ping, seismic survey lines, and geotechnical drilling. Seismic lines crossing the river
indicated a low compression wave velocity band along the center of the river indi-
cating the potential for weak rock or a sediment filled valley. Inclined boreholes were
drilled from the left and right banks near the lower falls but not at the cofferdam site.
Little work was carried out until 2010, when another significant program was
undertaken, but only two inclined boreholes were oriented to give additional infor-
mation on the river bed. Weathered rock and brecciated zones were encountered
at depth, potentially in the mid-river fault zone. No information on overburden was
obtained as the holes were started on rock outcrops.
In 2011, and again in 2012, in-river work was contemplated but, though the
technology was available, safety concerns prevailed, and these elements were not
executed.
4. COFFERDAM DESIGN
Given the sparse geotechnical database, the cofferdam design had to be

adaptable to various possible hypotheses of foundation conditions.
The Upstream Cofferdam (Fig. 3), constructed from July 18th to November
4th, 2016, is a 25 m high and 400 m long earth and rockfill embankment built to
divert the river through the Spillway during the construction of the RCC North Dam.
Below elevation 14 m, the cofferdam is comprised of a central impervious

zone of glacial till (Zone 1A) dumped in water between two rockfill groins. The till
is protected on the downstream side with a granular filter (Zone 2C) itself resting
on random rockfill maximum 900 mm (Zone 3) which constituted the rock groin.
Closure was achieved by a double groin advance from the right to the left bank.
Suspended sediments in the river water had to be avoided for environmental reasons
and consequently, no initial sealing could be made by dumping upstream of the
upstream groin as erosion of the material would quickly ensue. The solution adopted
484
Q.105 - R.27
Fig. 3
Cross-section
Coupe type
1A Dumped till Till déversé
1 Compacted till Till compacté
2A Sand and gravel filter Filtre en sable et gravier
2E Dumped sand and gravel Sable et gravier déversé
3 Dumped rockfill Enrochement déversé
3A/3C crushed rock transitions Transitions en pierre concassée
3D Compacted rockfill Enrochement compacté
was to dump material only between the two groins to bring the fill above water level.
Thereafter, a zoned fill, placed and compacted in layers was added to bring the
structure to the design height.
A symmetrical design was chosen to permit the insertion, if required, of a cut-

off through the fill and into the underlying riverbed overburden. As mentioned above,
little was known of the nature of the overburden, but swift currents gave credence to
the idea that coarse alluvium could exist in the central channel. From aerial photo
shots, it was clear that sand banks existed in the northern (left bank) part of the
channel and their extent varied according to flow conditions. Possible options for
a positive cut-off included a diaphragm wall or jet grouting. Similarly, treatment of
the bedrock was a potential need and pressure grouting and/or a continuation of jet
grouting were envisaged.
5. CONSTRUCTION SEQUENCE
The spillway, constructed on the right bank of the river, served as the diversion
channel.
Physical and numerical models were used to establish the weight of rock frag-
ments necessary for the advance of the groins, with larger sizes, up to 1.5 m, being
required for closure.
It had been determined that, under the prevailing flow velocities, it was not nec-
essary to advance simultaneously the two groins. The sequence of the downstream
groin followed by the upstream (Fig. 4) was chosen to minimize the accumulation of
rock fragments deposited by the current in the area of the core between the groins.
485
Q.105 - R.27
Fig. 4
Advance of rock groins
Avancée des jetées d’enrochement.
However, placing the downstream groin first had the effect of not only requiring
larger rock fragments to effect closure in this groin but also exposing the rockfill to
potentially erosive through-flow.
Prior to the placement of till between the two groins, dredging was carried out
using a backhoe equipped with a long-reach boom. To facilitate dredging and to
reduce the water depth in the pool, a breach was made in the downstream groin.
The surface of the bedrock was suspected to be very irregular and, due to
underwater working conditions, the submerged bedrock was possibly not completely
cleaned. A wide till core was planned to help compensate for that possibility. In
order to minimize the depth of dumped fill, the contractor maintained the breach in
the downstream groin to allow water to escape as the fill advanced. Though initially
seen as a beneficial approach, it was later surmised that some erosion of the till
may have taken place due to the continued flow through the upstream groin and out
through the breach.
Till placing began from the downstream filter crossing from downstream to
upstream and subsequently extended to the North and to the South. Meanwhile,
486
Q.105 - R.27
Fig. 5
Dredging and till placement
Dragage et déversement du till.
clean-up with the long stick backhoe continued as shown in Fig 5. Raising of the plat-
form was done in lifts of 1.5 m to 2 m with compaction by dozer and truck traffic only.
Seepage flows through the structure (with a head of around 10-12 m) were
deemed to be manageable by pumping, and due to schedule pressures, jet grouting
was postponed to be done at a higher elevation, if required.
The placement of the compacted fill in the upper section proceeded without
incident.
6. INSTRUMENTATION AND MONITORING
Instrumentation installation prior to impounding was limited to a series of sur-

vey monuments along the crest of the cofferdam. Regular inspections were carried
out prior to partial impoundment, and the plan for impoundment was a raise of 0.5 m
per day. Once impoundment started, two inspections were conducted per day (start
and end of day shift).
Seepage was monitored twice daily and plotted against head. This indicated
a straight-line relation with an anticipated flow of 10 or 12 m3 /min for the planned
partial impounding to el. 25 m. No sediment was observed.
487
Q.105 - R.27
7. OBSERVATIONS AND INITIAL ACTIONS
Partial impounding to elevation 25 m for the head-pond was required to be

achieved at this stage of the project to drown out the upstream rapids and prevent
the formation of frazil ice. Prior to the development of the Muskrat Falls project, the
accumulation of frazil ice in the pool downstream of the falls was an annual occur-
rence. Ice thicknesses of up to 30 m had been observed in the ice jam and river
levels could rise by 16 m to el 20 m – enough to completely drown out the upper falls.
Excavation of the powerhouse tailrace required that this ice jam occurrence be pre-
vented for the winter of 2016-2017. Cofferdam construction and partial impounding
to elevation 25 m were therefore on the critical path.
River flow rates and upstream water levels were carefully controlled by oper-
ation of the spillway gates. It is important to mention that outflow from the Churchill
Falls hydroelectric generating station situated some 291 km upstream, is very steady
due to a Guaranteed Water Availability Contract (GWAC) with Hydro-Quebec which
stipulated that a minimum of 10 of the 11 units must always be producing at full capac-
ity. The flow would therefore normally only vary during the winter period between
2000 m3 /s and 2200 m3 /s at the Muskrat Falls site.
From November 5th to November 14th , the water level of the reservoir was
raised from el. 14.7 to el. 21.5, at an average rate of about 0.6 m per day and
held at this elevation for observation prior to continuing. During this period, the total
measured seepage increased regularly from 4.5 to 8.5 m3 /min. With the exception of
the locations of small seeps at the downstream toe of the upper part of the cofferdam,
most of the seepage occurred around 60 m downstream of the axis of the cofferdam
at or near the surface of the bedrock, which is at el. 3 or 4 m. The water was quite
clear.
About a day after impoundment was halted, early in the morning of November
15th , a sudden and significant flow of dirty water occurred at the toe of the rock groin
some 60 m downstream of the axis at or near the surface of the bedrock (Fig. 6).
After about two and a half hours, this seepage stopped as quickly as it began. This
event was accompanied by settlement and cracks on the upstream slope and crest
of the cofferdam in the same area.
Various scenarios were identified and evaluated. The erosion could have been
initiated either through:
1. an open joint in bedrock by-passing the filter;

2. a permeable soil at the base of the filter Zone 2C;
3. large boulders at the toe of the downstream groin incompatible with filter Zone
2C
4. a segregated zone within the filter Zone 2C.
488
Q.105 - R.27
Fig. 6
Location of observations
Localisation des observations
An open joint in the bedrock would have required a seepage section great enough
for a flow of several m3 /min combined with a continuity on a length of many meters
in the downstream direction. Examination of the bedrock surface in the footprint of
the future RCC North Dam did not reveal such joints in the northern sector.
The other possibility is that till erosion had been initiated at the base of the filter
of Zone 2C due to the presence of fine sand from the sand bar or left in place during
the construction. The high current velocities at closure, created as the flow section
diminished, would have removed much of the sand bar adjacent to the downstream
groin. However, during the period that the breach was open (one month) more sand
could have accumulated.
The location of the event aligned with the start of the use of large rock
fragments for groin closure and where a breach was excavated.
All these hypotheses were plausible but wide-open joints were not a feature
of the neighbouring rock mass and, more importantly, no seepage was observed
prior to and after the event at that location. Incompatibility between fill zones and/or
deposited material appeared to be the more likely cause.
With the collapse of the embankment fill into the eroded space, it is not unlikely
that much of the sandy material present in the lower part of the Zone 1A and
which was removed during the high-flow period of the incident was replaced with
till considering that the volume of settlement estimated to be in the range of 200 to
250 m3 .
489
Q.105 - R.27
The event which occurred on November 15th was attributed to erosion of till
in Zone 1A through the base of Zones 2C and 3 where permeable materials where
left in place during construction due the method of placement of till of Zone 1A. As
erosion progressed, transporting materials in the downstream direction, a process
of self-healing developed resulting in the development of a plug which finally stopped
erosion and seepage.
Two days later, on the night of November 17th , another significant seepage
accompanied with erosion of till occurred at the toe of the upper part of the cofferdam
near the North Abutment. Outflow occurred at around el. 15 m through rockfill placed
above the till of Zone 1A. The event could have been a delayed reaction to the earlier
one or an independent but similar issue of compatibility of fill zones and/or foundation
bedrock or overburden materials. Continuation of the erosion of till from the central
impervious zone could have affected the integrity of the cofferdam and it was then
decided to lower the reservoir to el. 14 m as quickly as possible during the following
hours.
8. OPTIONS FOR REMEDIAL WORKS
Evaluation of the situation led to the conclusion that immediate action was
required, to prevent uncontrollable deterioration, followed by a systematic approach
to establish the nature of defects and repair the internal core of the cofferdam.
A rockfill buttress was added on the upstream slope and crest to fill the
depressions where settlements and cracks occurred, and to cover the portion of
the cofferdam which was considered more at risk due to the method of placement
of the till of Zone 1A (Fig. 7).
Fig. 7
Remedial works
Travaux remédiateurs
To prevent another piping incident, a downstream rockfill berm was added at

the toe of the upper part of the cofferdam, between el. 14 m and el. 19 m in the area
490
Q.105 - R.27
of concern near the North abutment. This rockfill berm included an inverted filter of
Zone 2C material, abutted and covered by Zone 3 rockfill.
The observations made during the period immediately following the incidents
indicated the potential presence of heterogeneities in the embankment fill that was
placed to construct the cofferdam platform to el. 14 m. Moreover, the incidents of
internal erosion and piping likely led to the presence of disturbed zones and possibly
voids. Remedial measures were required to treat such zones to prevent a worsening
condition upon resumption of the impoundment.
Given the uncertainties in the cofferdam condition, the remedial works were
also exploratory in nature. Construction techniques had to be adaptable to site
conditions including the imminent onset of winter temperatures.
Several concepts were put forward for preliminary evaluation as shown in

Table 1. Given the possibility of flow pathways through the bedrock, bedrock grouting
would also form part of the program.
Table 1
Concepts for Remediation
Description Pro Con
Steel sheet Continuous cut-off Driving difficulty due to cobbles and
piles boulders.
Windows at base of piles due to irregular
bedrock surface.
Diaphragm Continuous cut-off Risk of panel instability due to slurry
slurry wall loss.
Tube-à- Some of equipment already on site Risk of hydraulic fracture and loss of
manchette grout.
grouting Does not compact soft ground. Creates
reticulate ill-defined structure.
Difficult to monitor and control.
Jet grouting Flexible method. Pressure in soil at least equal to head
at crest of cofferdam.
Treats material types of interest. Risk of hydraulic fracture and loss of
grout.
Low Mobility Reduced risk of hydraulic fracture. Less control over cut-off continuity than
Grout (LMG) panel wall.
PREFERRED Simpler equipment.
OPTION
Backfilling of voids and compaction of
existing material. Reasonable capa-
bility of winter work.
Likely to be quicker.
TRD Trenching Continuous wall. Presence of boulders.
System
Stable trench walls. Limited width of treatment.
Given that there was still no consensus on the root cause, grouting of the
embankment, the foundation overburden and the rock was viewed as offering the
most practical option to rehabilitate the cofferdam.
491
Q.105 - R.27
9. EXECUTION OF THE WORKS
Considerable thought was given to the methods for drilling and grouting per-
mitting the treatment as required of the rock by traditional grouting procedures and
filling and compaction of the embankment fill by the use of the Low Mobility Grout
injection.
The criteria for determining the details and parameters were the following:
• Execute drilling and grouting from the crest of the cofferdam in the center of
the core;
• Select drilling method to minimize damage to core. Preferably avoiding the use
of compressed air;
• Minimize pressures that could re-initiate erosion and piping;
• Select mix designs to control and limit penetration in the first instance until
integrity has been assured. The primary objective was erosion resistance and
not strength of grout columns;
• Carry out drilling and grouting on a primary hole sequence of 6 m c/c,
secondary holes at 3 m and tertiary holes at 1.5 m;
• Inject grout using a controlled pressure/volume relationship;
• Aim to create continuous columns in a secant pile configuration to build a
cut-off, while minimizing the contrast in stiffness between columns and un-
compacted in-situ fill.
The grouting operations would begin in the vicinity of the observed seepage (see
Fig. 6) proceeding towards the north abutment followed by work in the opposite direc-
tion to cover the area potentially affected by the piping evidence. Other exploratory
work along the cofferdam would follow.
The grouting works were performed between the end of December 2016 and
end of March 2017. Given the late period of the year, a decision was made to forego
a separate exploration program to characterize the variability of soil density and to
proceed directly with the grouting works which would become exploratory in nature.
10. GROUTING
10.1. GENERAL
The work covered a total length of 325 m divided into four segments to prioritize
and organize the activities. The segments were denoted as A to D, from area most
at risk to least (Table 2 and Fig. 8).
492
Q.105 - R.27
Table 2
Holes quantity number section
Section Primary Secondary Tertiary Total
SECTION A 12 10 8 30
SECTION B 13 5 - 18
SECTION C 8 5 - 13
SECTION D 11 5 - 16
Total 44 25 8 77
Fig. 8
Grout hole layout
Disposition des trous d’injection
10.2. DRILLING EQUIPMENT AND PROCEDURE
With a view to minimizing further damage to the embankment fill, drilling was
carried out with SONIC drills. The holes were drilled with 150 mm (6 in) O.D. casing
and a 121 mm (4.75 in) core barrel. The drill hole was advanced as follows:
• The core barrel was advanced in 3.04 m (10 ft) increments with recovery
of samples extruded into plastic sleeves. This procedure was used both in
overburden and in rock.
• After reaching the designated depth for the end of hole, washing the hole was
carried out until water was clear, the drill string and core barrel were withdrawn,
and the final rock sample extruded into a plastic sleeve.
At this point the hole was ready for the geo-camera survey or rock grouting.
493
Q.105 - R.27
10.3. GROUT MIXES AND GROUTING CRITERIA
10.3.1. Bedrock and Interface, High Mobility Grout
The bedrock and the interface were injected with high-mobility grout (HMG)
for good penetration. Essentially one base HMG mix was used for bedrock and
interface grouting with a water-cement ratio of 1.25 and bentonite having a bentonite-
cement ration of 0.06. Rheological characteristics were varied by the use of three
different doses of superplasticizer, to give Marsh Cone viscosities between 40 and
90 seconds. A fourth mix was obtained by the addition of anti-washout additive.
In the case of excessive grout take, the viscosity was progressively increased
through the three steps and finally anti-washout additive was used.
The injection of the HMG was stopped according to the following criteria:
Effective pressure reached maximum values:
Rock: 500 kPa
Interface: 250 kPa
Volume injected reached 1500 liters
Flow of less than 1 liter per minute at the allowable effective maximum pressure
for 5 minutes.
10.3.2. Compaction, Low Mobility Grout
As the objectives were to fill voids and enhance the physical characteristics
of the fill, the low mobility grout (LMG) was used for the embankment because of
several inherent advantages as compared to other remedial techniques:
• Reduced risk of hydraulic fracture;

• Simpler equipment;
• Backfilling of voids and compaction of existing material;
• Reasonable capability of winter work;
• Likely to be quicker.
On other hand, LMG grouting methods offer less control over cut-off continuity
than, for example, a panel wall. The LMG consisted of glacial till mixed with cement
and bentonite. The LMG mix was not intended to be high strength. The water-cement
ratio was 12 with the bulk of the mass being obtained with screened till and sand.
Slump was 40-90 mm and maximum strength 250 kPa at 28 days.
494
Q.105 - R.27
The injection of the LMG was stopped according to the following criteria:
• Effective pressure reached maximum values from pressure chart (250 and
700 kPa effective pressure)
• Volume injected reached 1500 liters
10.4. GROUTING OPERATIONS
10.4.1. Bedrock grouting
The bedrock grouting used both HE (30) or GU Portland cement, depending

on the availability at site, to produce a stable HMG and included superplasticizer
additive (Rheobuild 1000), and anti-washout (Celbex or UW450). All grout mixes
were prepared in a high shear (colloidal) mixer.
Control of grouting following the Grouting Index Number (GIN) method with
criteria for Pmax, Vmax and a GIN number.
The HMG monitoring station was equipped with flow meters, pressure
transducers and computerized real time data recording capability.
10.4.2. Interface grouting
Initially, the interface grouting was intended to treat the interface between the
bedrock and the dumped till. As the program developed, it became apparent that
a native till, in close contact with the rock and essentially impervious, was present.
Thus, in many grout holes where native till existed, interface grouting was applied
to the interface between native till and dumped material or granular materials.
Interface grouting followed immediately after completion of rock grouting. The

interface grout is a stable, high-mobility grout (HMG-1) with the intent of permeating
the soils at the interface.
The grout components were batched in the same batch plant as used for the
rock grouting.
10.4.3. Compaction Low Mobility Grout (LMG)
In order to have a reasonable understanding of the effective pressures dur-

ing LMG grouting, the pressure losses associated with pushing LMG through the
495
Q.105 - R.27
lines and casing was determined by measurements made on a surface set-up of a

Putzmeister pump with a 15 m hose line and 10 m of casing.
The LMG was produced in the following manner:
• Pre-screening of till to 20 mm maximum particle size;

• Pre-weighed cement and bentonite delivered to the LMG grout shack;
• Cement, bentonite and water mixed;
• Till added and water adjustment made if necessary to give a 50 mm slump.
LMG grouting was commenced with an effective pressure of 250 kPa. On recom-
mendation of the grouting consultant, the effective pressure for the LMG grouting
was raised by 700 kPa for all holes after January 20th , 2017.
The LMG followed the interface grouting. The sonic drill casing, which has an
I.D. of 125 mm, was used as the delivery pipe for LMG. The grouting progressed in
1.5 m stage lengths, or less, by pumping grout out of the end of the casing.
The LMG grouting was generally a lengthy process from the time the grout
began to enter the lines to final completion of the compaction stages from top
of bedrock to the crest of the dam. There were generally 9 or 10 compaction
grouting stages within the lower embankment and 6 backfilling stages in the upper
compacted fill.
10.5. TIME FRAME
The drilling and grouting program was performed from December 16th , 2016
to March 26th , 2017 with 4 days off during Christmas.
The drilling and grouting were executed during the most difficult part of the win-
ter season. Temperatures down to minus 30˚ C and strong winds were encountered
during the fieldwork. Such conditions dictated insulated lines to eliminate freeze-up,
the creation of wind breaks on the west side of the cofferdam, the use of heating units
to heat the grout shacks, and hoarding around the drill head when not in operation.
11. RESULTS
11.1. GENERAL
Acoustic tele-viewer images and anecdotal drilling observations indicated, in

general, highly fractured bedrock. Fracture apertures up to 5-10 cm were observed
496
Q.105 - R.27
by the camera survey. But, these fractures were not always associated with water
return loss or significant grout takes, as some of the joints were filled with silt or till.
The conductivity of the fractures or joints was therefore highly variable.
11.2. SEQUENCE A
Sequence A comprised a total of 30 grout holes from 0+323 to 0+400, a dis-

tance of 77 m. Twelve primary, 10 secondary and 8 tertiary holes were drilled during
the remediation works. This section is located at northern end of the cofferdam,
where the seepage and settlement event occurred in November 2016 during the
first impoundment.
As expected, compacted and dumped embankment were encountered. Some

thin sandy layers were found in Zone 1A, and more important sand and gravel layers,
up to 5 m in length, were encountered in some boreholes
A dense native till deposit was encountered above the bedrock along most of
the section A. Thickness varied from a few decimeters to more than 5 m, in a rock
depression. Highly fractured bedrock was encountered with variable permeability as
discuss earlier.
11.3. SEQUENCE B
Sequence B included grout holes from 0+240 to 0+323, a distance of 83 m. A

total of 18 holes were drilled of which 13 were primary holes and 5 were secondary
holes. No tertiary holes were drilled.
Sand layers were encountered in every borehole, within the embankment

dumped till, with up to 7 m in thickness being observed in one borehole.
The boreholes revealed the presence of a continuous native till deposit under
the sequence B of the cofferdam, varying from 2 to 4 m at the lowest point of the
bedrock profile.
11.4. SEQUENCE C
Sequence C included grout holes from 0+040 to 0+086, a distance of 46 m. A

497
Q.105 - R.27
The stratigraphy revealed by the boreholes showed some relatively thin widely
distributed sand layers and locally up to 2 m.
Native till was not wide-spread.
Fractured bedrock was observed in some boreholes.
11.5. SEQUENCE D
Sequence D included grout holes from 0+114 to 0+232, a distance of 118 m. A

11.6. GROUTING TAKES
Grout takes per hole were computed and the total volume of HMG (bedrock
and interface) and LMG injected in each hole is summarized in Table 3.
Table 3
Grout Takes Summary
LMG Compaction Grout HMG Interface HMG Bedrock
Section Total m3 Average of L/m Total m3 Average of L/m Total m3 Average of L/m
A 68.4 152 5.3 151 14.7 223
B 49.1 139 1.8 48 2.7 69
C 20.8 105 1.2 46 6.9 144
D 42.6 124 4.3 191 22.4 233
Grand Total 181 135 12.6 112 46.7 186
12. COMMENTS
The principal findings from the drilling and grouting works can be summarized
as follows:
The rock was encountered in almost all boreholes. The rock grouting showed
a large range of takes, confirming the presence if not the ubiquity of open joints and
fractured zones in the rock as could be seen in the footprint of the main dam located
downstream of the cofferdam.
The large takes were more frequent in the middle of the cofferdam
(sequence D). Furthermore, during the work done in that section, the pH in the
downstream seepage increase markedly. The location of the weak rock zone at the
low point in the profile, from where seepage was collected and pumped, was in
498
Q.105 - R.27
alignment with the observed downstream open joint location. It may be recalled that
a zone of low seismic velocity had been noted and that brecciated rock had been
encountered in the early inclined investigation holes.
In sections A and B, an unexpected native till layer was encountered. This was
a favourable situation, as the dense native till provided a natural protective barrier
between an open joint in the rock and the dumped till, thereby reducing the risk of
erosion.
The interface between the native till or the rock and the dumped till did not
take a large volume of grout except in sequence A near the North abutment. The
interface grout probably penetrated the open structure of rockfill or into an open joint
but not significantly into voids in the washed till.
Compaction grout was used with success and the grout, depending on the soil
properties, may have filled voids or, by cavity expansion in the dumped till, achieved
additional compaction and lowered the permeability of the till core.
Soft soil and more pervious material were encountered in a variety of locations,
at different elevations, and over different thicknesses. Based on the borehole results,
there were potential deficiencies in the entire dumped till and also in the rock.
Interface deficiency was encountered mainly in the sequence A near the North
abutment.
13. NEW INSTRUMENTATION
Given the criticality of the cofferdam in the overall project planning, it was
imperative to provide means to monitor the effect of the grouting program and the
subsequent return to service.
A series of piezometers were installed as standpipes with Van Essen Micro

Diver insertions to record the piezometric levels.
A total of 13 standpipe piezometers were installed between 0+060 and 0+380.

A track mounted CME 50 drill rig was used to install the piezometers. HW casing
(100 mm I.D.) either for single or nested piezometers was used to advance the hole
to the target zone.
Nine settlement monument pins were installed along the crest from 0+240 to
0+380. The maximum vertical displacement on the pins was negative (settlement)
and ranged from 10 to 42 mm but was generally around 15 mm. During the remedial
works and the subsequent operation, there were no visible signs of cracking or
settlement on the crest or slopes.
499
Q.105 - R.27
In general, piezometric levels rose in parallel with the increasing head

pond elevation but the head dissipation across the structure indicated acceptable
conditions.
14. RETURN TO SERVICE
Progress of the grouting and its effect were judged by the grout take. The only
way to really test the rehabilitated structure was to re-start the impounding procedure
in a progressive manner while monitoring, by piezometers and flow measurements,
the behavior of the structure.
By January 25th it was determined that sufficient work had been done to
address the high-risk areas and that impoundment could be restarted to prevent
the formation of frazil ice and the creation of the downstream ice jam. The ice jam
had been continuing to build and was beginning to threaten to flood the project from
downstream. On February 14th , 2017, with the water level at 20.5 m, an upstream
ice cover did form and the frazil supply was immediately cut off. The growth of the
ice jam ceased, and the downstream water level began to recede. Total seepage
volumes were in the range of 9 to 11 m3 /min which, while not insignificant, was in
the manageable range.
15. CONCLUSIONS
The Muskrat Falls upstream cofferdam was designed with limited knowledge
of the riverbed conditions and constructed largely in water. An unfortunate com-
bination of bedrock discontinuities, overburden deposits, and certain features of
the construction methodology led to the initiation of internal erosion during the ini-
tial partial impounding behind the cofferdam. Immediate precautionary measures
were taken to lower the reservoir, buttress the embankment and add protective fil-
ter zones at strategic locations. These actions allowed more time to plan additional
remedial works, to mobilize crews, equipment and materials and to execute a grout-
ing program. The work was completed, and the cofferdam returned to service thus
permitting the impounding to be re-started and upstream water levels to be reached
so as to prevent the continued accumulation of frazil ice and avert the threat to
flooding of the powerhouse construction from downstream. A critical milestone in
the project was therefore achieved as intended. Construction of the North Dam was
able to start in spring 2017 and the dam was completed in October 2018. The cof-
ferdam was decommissioned with the final impoundment to el. 39 m in August –
September 2019.
500
Q.105 - R.28
-------
-------
INUNDATION MODELLING OF NON-NEWTONIAN TAILINGS DAM BREACH

OUTFLOWS (∗)
Violeta MARTIN
Specialist Hydrotechnical Engineer,
KNIGHT PIÉSOLD LTD.
Daniel ADRIA & Heleman WONG

Junior Engineers,
KNIGHT PIÉSOLD LTD.
CANADA
SUMMARY
The Mount Polley, Fundão, Feijão and other tailings dam failures have had
catastrophic environmental and social impacts. Tailings dam owners, operators,
regulators, and stakeholders rely on dam breach studies to evaluate potential
consequences of a failure and make informed decisions and ultimately safeguard
the public. However, scientific understanding of the breach processes and phys-
ical phenomenon of tailings outflows is still developing. Uncertainties in model
inputs combined with lack of standardized methodologies for completing the var-
ious analyses required for these complex studies present considerable challenges
to professionals. Experience and judgement must be relied on to construct tailings
dam breach models and assess the results.
Tailings dam breaches typically result in hyperconcentrated flows due to the

mobilization of stored tailings solids. After determining the hypothetical outflow vol-
ume and the breach hydrograph, the breach outflow is routed downstream using
hydrodynamic or geomechanical modelling tools to estimate the inundation. Site-
specific tailings rheology is required to characterize these flows, which is rarely
∗ Modélisation des écoulements non newtoniens dans les brèches des barrages de stériles
DOI: 10.1201/9781003211808-74 501

Q.105 - R.28
available. This paper describes recent experiences with modelling the flood wave
propagation of non-Newtonian tailings dam breach outflows using hydrodynamic
modelling tools. Challenges encountered and lessons learned from various case
studies are discussed, including availability and quality of input data and selection
of appropriate models for the given conditions. All these factors have profound.
RÉSUMÉ
Les ruptures du Mont Polley, de Fundão, de Feijão et d’autres barrages de

stériles ont eu des impacts environnementaux et sociaux catastrophiques. Les
propriétaires, opérateurs, régulateurs et parties prenantes sont dépendants des
études des brèches de barrages pour prendre des décisions et sauvegarder le pub-
lic. Cependant, la compréhension scientifique des processus et du phénomène
physique des écoulements de résidus se développe encore. Les incertitudes dans
chaque entrée de modèle combiné avec le manque de méthodologies normalisées
pour accomplir les diverses analyses requises pour ces études complexes présen-
tent des défis considérables pour les professionnels. L’expérience et le jugement
doit être fondée pour construire des modèles de brèches de barrages de stériles et
évaluer les résultats.
Les brèches de barrages de stériles entraînent généralement des écoule-

ments hyper-concentrés en raison de la mobilisation des résidus solides stockés.
Après avoir déterminé le volume d’écoulement hypothétique et l’hydrographe de la
brèche, l’écoulement de la brèche est acheminé vers l’aval à l’aide d’outils de mod-
élisation hydrodynamique ou géomécanique pour estimer l’inondation. La rhéologie
des résidus propres au site est nécessaire pour caractériser ces écoulements, ce
qui est rarement disponible. Ce rapport décrit les expériences récentes de modéli-
sation de la propagation des ondes de crue des débits de rupture de barrages de
stériles à l’aide d’outils de modélisation hydrodynamique non-Newtoniens. Les défis
rencontrés et les leçons apprises pour une sélection diverse d’études de cas sont
discutés, y compris la disponibilité et la qualité des données d’entrée et la sélec-
tion de modèles appropriés pour les conditions données. Tous ces facteurs ont des
impacts profonds sur les résultats.
1. INTRODUCTION
Dam breach inundation studies are required to evaluate the potential impacts
and hazards associated with a tailings facility at all stages of design, whether
the facility is proposed, operating, or closed. The results from these studies can
inform the dam consequence classification, emergency preparedness and response
planning, and can also be used in environmental impact assessments, alterna-
tive assessments, or for other purposes that are critical for the design and safety
management of tailings facilities.
502
Q.105 - R.28
Dam breach studies are an important requirement and an integral part of the
Global Industry Standard on Tailings Management (GISTM) released on August 4,
2020, which provides a framework for safe tailings facility management. The goal
of the GISTM is to work towards achieving zero harm to people and environment
with zero tolerance for human fatality [1]. Furthermore, mining companies and gov-
ernments around the world are also introducing requirements and regulations for
tailings dam breach studies (e.g. Resolution 13 issued on August 18, 2019, or Res-
olution 32 issued on May 11, 2020, both by the Agência Nacional de Mineração
(ANM) in Brazil, [2, 3]).
Recent work to standardize approaches for tailings dam breach studies has
been completed by the Canadian Dam Association (CDA) in a Technical Bulletin
on Tailing Dam Breach Analyses [4] (CDA TDBA Bulletin for short), which were
previously summarized in Martin et al. [5]. The science and understanding of the
physical processes driving tailings dam breach events and the hyperconcentrated
flows released in breach outflows is still evolving. As such, many uncertainties in
the required inputs and in appropriate modelling methodologies still exist.
Research efforts are increasing to cater to the mining industry’s need to better
predict tailings dam breach outflows. The Canadian Tailings Dam Breach Research
Project (CanBreach) is one example of a joint research and technology develop-
ment effort between several Canadian Universities (University of Waterloo, Queen’s
University, and the University of British Columbia), the mining industry, and the pro-
fessional community. To advance the understanding and modelling capabilities of
breaching and runout processes, CanBreach is combining forensic analysis and field
observations of past failures with laboratory-scale experimentation and numerical
modelling.
In addition to various research efforts, numerical tools for modelling the breach
outflow hydrograph (e.g., BREACH, DL Breach, FLDWAV, FLO-2D, HEC-RAS,
HEC-HMS, or WinDAM), and hydrodynamic tools for modelling the downstream
propagation of non-Newtonian tailings breach outflows (e.g., FLO-2D, FLOW-3D,
and RiverFlow2D, or tools in development at the time this paper was prepared
like HEC-RAS Version 5.1, TELEMAC 2D, and TUFLOW) are constantly evolving.
These modelling tools are becoming increasingly complex as they aim to incorpo-
rate additional processes, such as the mixing and dilution with incoming flows from
downstream tributaries, or the erosional and depositional processes caused by the
passage of a breach flood wave.
There are two high level steps in a dam breach study related to modelling: (1)
determining the critical failure modes, the outflow volumes and the breach outflow
hydrographs; and (2) developing a hydrodynamic model to route the flood wave and
predict how the breach outflow impacts the downstream environment. While the
former step is briefly addressed, the latter step is the focus of this paper. Specific
challenges involve the availability and quality of input data, followed by the selection
of an appropriate hydrodynamic model while considering the limitations of each
model. Inputs to the model are discussed in general, and specific applications and
503
Q.105 - R.28
difficulties are demonstrated using case studies. It is acknowledged, however, that

some of the findings in this paper may be outdated by the time of final publication
considering the fast pace of research and development of the numerical tools.
2. DAM BREACH STUDY PROCESS
2.1. GENERAL
The physical processes of a dam breach event may vary substantially depend-
ing on the presence of a supernatant pond and on susceptibility of the dam fill and/or
the tailings to liquefy and flow [4, 5]. The breaching process may be erosional, driven
by overtopping or piping with the discharge of the supernatant pond carrying eroded
tailings and dam fill materials. Due to the loss of containment, the tailings mass may
undergo flow liquefaction resulting in additional discharge of liquefied tailings, or
it may undergo progressive slumping of unsupported non-liquefiable tailings until
the tailings slopes reach equilibrium. The runout from such a failure may range
from water floods to mud floods to mudflows, as characterized by O’Brien [6] and
discussed further in the CDA TDBA Bulletin [4].
The breaching process may also be nearly instantaneous if triggered by a

different failure mechanism such as mass liquefaction of the dam fill and/or tailings
mass, similar to what occurred at the Feijão dam near Brumadinho, Brazil, in January
2019. In this scenario, the outflow would comprise liquefied dam fill and tailings
materials, as well as any water contained within an impacted supernatant pond,
if present. The outflow of liquefied tailings and water would also cause additional
erosion of tailings and dam fill materials as it leaves the facility. The runout from
such failures may range from mudflows, to flow slides or tailings flows, or it may be
a mixture of water floods, mud floods and mudflows for facilities with a pond that
discharges in a breach [4].
2.2. BREACH HYDROGRAPH
For a hypothetical dam breach, the professional must determine a reasonable

and critical failure mode, the breach outflow volume, the breach characteristics,
and the resulting breach outflow hydrograph. The breach outflow volume for water
retaining dams is typically considered to be the entire stored volume above the
breach invert at the time of a hypothetical failure. The breach volume for tailings
dams, however, most often includes the supernatant pond and a volume of stored
tailings that are released as a result of erosion or static liquefaction. It is expected
that the tailings would continue to mobilize until a stable residual slope is formed in
the facility.
504
Q.105 - R.28
Breach parameters (size, shape, and time to fail) and peak outflows can be
estimated using several existing empirical equations and guidelines that were sum-
marized by Wahl [7, 8], West et al. [9], and Brunner [10]. These include the works
of: MacDonald and Langridge-Monopolis (1984), Costa (1985), Bureau of Recla-
mation (1988), Von Thun and Gillette (1990), FERC (2015), Froehlich (1995, 2008,
2016), Walder and O’Connor (1997), Xu and Zhang (2009), Pierce et al. (2010), and
others. The limitations regarding the applicability of empirical equations to tailings
dams include the following:
• The equations were developed for water retaining dams;

• Most equations were developed with limited data sets for failures of relatively
small dams (with dam breach studies for larger dams must rely on extrapolated
results);
• The various equations often generate a wide range of possible breach
parameters and peak outflows, as was illustrated in Martin and Akkerman
[11];
• Most equations would not be applicable for liquefaction type failures of
upstream constructed tailings dams.
In consideration of these limitations, sensitivity analysis using the possible

ranges of breach parameters are required in order to evaluate the impact on peak
outflows and downstream consequences. Monte Carlo analysis can be used to
provide a meaningful stochastic analysis of peak breach outflows for various combi-
nations of breach parameters, as described by Goodell [12]. Assuming that a breach
occurs, the final breach outflow hydrograph can be developed for the combination
of parameters that resulted in a selected peak discharge (e.g., the 95th percentile
peak discharge, or the peak discharge that has a probability of exceedance of 5%
if a dam breach occurred).
Physically-based numerical breach models attempt to model the erosional pro-

cesses using a combination of sediment transport, soil mechanics, and hydraulics
principles. Models based on typical water retaining dam configurations currently
exist, however, the applicability of these physically-based models for tailings dam
breaches is still uncertain. These models are not well suited for upstream con-
structed tailings dams that are relying on the strength of potentially liquefiable
tailings. Similarly, the impact a tailings beach commonly adjacent to the dam would
have on a breach development is not well understood.
2.3. HYDRODYNAMIC MODELLING AND FLOOD WAVE ROUTING
Determining the breach outflow hydrograph, while complex and highly uncer-
tain, is only part of the process. Hydrodynamic models are commonly used to predict
the potential downstream impacts following a breach event. Generally speaking,
hydrodynamic models compute the depth and velocity of a flood wave incorporating
the equations for conservation of continuity, energy, and momentum. All numerical
505
Q.105 - R.28
models are simplifications of real-world phenomena that calculate results discretely,

both temporally and spatially, in addition to including parameterizations to describe
physical processes not included in the basic conservation equations. The temporal
and spatial discretization may result in discretization errors, arising from the fact a
fixed and finite number of elements is used to represent a continuous variable (fluid
flow). The discretization error may be reduced with finer resolution (or mesh size) in
the hydrodynamic model, at the expense of greater computational requirements.
The discretization and various parameterizations are treated slightly differently

in various software packages. Knowledge of these differences and their limitations is
important when selecting and constructing a model to best estimate the downstream
inundation. One common simplification used in dam breach modelling is using two-
dimensional (2D) instead of 3D models, in which the vertical velocity variations
are removed and depth-average calculations (i.e. shallow water approximation) are
used instead. This assumes that the variation in the vertical velocity component
is negligible compared to the horizontal velocity components. The computational
requirements are considerably reduced when using a 2D instead of a 3D model. Sim-
ilarly, the velocity variation across the channel may also be averaged, resulting in 1D
models that are correspondingly faster than 2D models. Further details on selecting
steady vs. unsteady, or 1D vs. 2D. vs. 3D models are discussed in USACE [13].
Dam breach events are typically highly turbulent and unsteady, and the ver-
tical velocity component may not be negligible compared to the horizontal velocity
components, especially immediately downstream of a breach. The required down-
stream extent of a dam breach model, however, almost always limits the study to
depth-averaged or 2D models rather than 3D models due to computational require-
ments. With highly dynamic and complex flow directions typically associated with
dam breach and overland flooding, 2D is preferable to 1D modelling. A 1D model may
suffice if the flow is expected to be relatively unidirectional without much overland
flooding in canyon-like terrains, or when the model domain is very large (>100 km)
such that the computation requirements prohibit a 2D model. Hybrid 1D/2D models
can be applied to utilize the advantages of both approaches, however, these may
result in model stability issues at the 1D/2D interface, which occasionally negate the
possible advantages.
Given the flood wave dynamics and their large area of impact, 2D models
often strike a reasonable balance between improved complexity and computational
requirements. Advancements in processing power and cloud computing may change
this in the coming decade allowing full 3D models to eventually become standard.
2D models will remain in use for tailings dam breach inundation modelling in the
near future, and as such are the focus of this paper.
Another common simplification in tailings dam breach models is an assump-

tion of a fixed-bed model, where erosion and tailings deposition due to the passage
of a flood wave are neglected. This simplification is often needed due to lack of
available data to support erosion modelling over large downstream extents, or due
to hydrodynamic models not being capable to perform both sediment transport and
506
Q.105 - R.28
non-Newtonian flow calculations concurrently. Some of the current software devel-

opments are starting to overcome this deficiency but are yet to be sufficiently proved
in practice.
3. HYDRODYNAMIC INPUT DATA
3.1. GENERAL
Inputs for the hydrodynamic model include the breach outflow hydrograph, the
representation of the downstream terrain, the natural flows occurring in the down-
stream watercourses at the time of the breach, and the rheology or non-Newtonian
flow properties of hyperconcentrated flows that are typical for tailings dam breaches.
The quality of the results of any computer model are highly dependent on
the quality and representativeness of the input data. Regardless of the skill and
knowledge of the modeller or the sophistication of the model, low quality inputs will
result in poor quality and potentially inaccurate outputs, and therefore, an extremely
important element for all dam breach studies is careful consideration of the data
used to build these models.
3.2. TERRAIN DATA AND SURFACE ROUGHNESS
The terrain downstream of the facility is integral to determining the inundation

and impact of a hypothetical dam breach. Gridded elevation data sets, commonly
referred to as Digital Elevation Models (DEMs), are used to represent the terrain in
the hydrodynamic models. These data can come from various technologies, such
are drone measurements (tens of hectares), aircraft surveys (few km2 to few tens of
km2 ), or satellite measurements for larger areas (few tens of km2 to few hundreds
of km2 ).
There are two main measures of the quality of a DEM: (1) the horizontal res-
olution, or the size of each gridded elevation cell; and (2) the vertical accuracy,
represented by the vertical distance between the modelled and the true elevation
due to systematic and random errors. Fine horizontal resolution with higher accuracy
is required to adequately describe the topography. Publicly available satellite data
typically range from 30 m to 90 m horizontal resolution for various locations around
the world. Site-specific DEMs from drone or aircraft data can reach resolutions of
0.5 m but may be costly to procure for large model extents.
The scale and scope of a dam breach study, the complexity of the downstream
area, and the potential impact of a hypothetical dam breach should be taken into
507
Q.105 - R.28
account when assessing the quality of a DEM. It is frequently up to the professional

to determine if a publicly available coarser resolution data would be acceptable
compared to a more expensive higher resolution data developed specifically for the
dam breach study. This decision is often subjective and comes with professional
judgement and experience, but in general, it is governed by the type of terrain
and potential risk to the downstream population. Flat terrain or heavily populated
areas would typically require higher resolution topographic data in order to define
the potential inundation with higher accuracy.
Further complication with terrain data is that most of the measurement meth-
ods do not include sub-aqueous terrain data (bathymetry) for the water bodies (e.g.,
rivers, lakes, etc.). A separate hydrographic survey is required to provide this infor-
mation. For larger river systems, the bathymetry can have a profound effect on the
possible flood wave attenuation. If no bathymetry data is available, the professional
must either use the terrain “as is” and assume/accept the attenuation is negligible,
estimate the shape and depth of the river channel and “burn” in channels to the
terrain data prior to modelling, or recommend that bathymetry data be collected.
An example of the impact the terrain can have on the inundation results is
shown on Figure 1. A tailings dam breach study was updated six years after the ini-
tial study was completed, following material changes at the mine site. The publicly
available topographic data improved in resolution and accuracy in the meantime.
The newer terrain has a clearly defined channel while the older terrain does not,
as shown in the cross section on Figure 1. Although the updated analysis indicated
a larger breach volume and higher peak flow, the inundation area in the down-
stream watercourse was smaller due to a better topographic definition indicating a
confinement in the terrain.
Fig. 1
Example of the impact of topographic resolution on the inundation extent.
Exemple de l’impact de la résolution topographique sur l’étendue de l’inondation.
Lastly, surface roughness that defines the resistance to flow must be assigned
to the terrain. Most 2D models describe the surface roughness with a Manning’s
n value, however some may define it using roughness height. Regardless of the
508
Q.105 - R.28
definition, various sources exist and can be referred to when selecting these values.
Publicly available digital land use maps are becoming more prominent, allowing
for easier classification and spatial variation of the roughness parameter within the
hydrodynamic model domain. Professional experience should be used to adjust the
parameter in some areas based on satellite imagery. Sensitivity of the inundation
extent and flood wave arrival time to the selected surface roughness should be
evaluated.
3.3. HYDROLOGY AND NATURAL FLOWS
While the severity of the inundation is directly related to the breach volume
and peak outflow, the background or natural flows in the downstream watercourses
can also have a large impact on the inundation and attenuation of the breach flood
wave. Two hydrologic conditions are typically considered for dam breach studies:
(1) flood induced or rainy day, where the failure is coincidental with extreme flood
conditions (e.g. the Inflow Design Flood or IDF); and (2) fair weather or sunny day,
where the failure occurs due to any trigger under normal operating conditions.
The downstream watercourses in flood induced scenarios should be modelled

at flood levels expected to occur concurrently with the assumed flood (e.g. IDF) at
the facility. The larger the downstream drainage, the more difficult determining the
downstream flood magnitude becomes. A heavy storm centered at the facility would
decrease in intensity over large areas. Catchments that are farther downstream,
which contribute to the watercourses along the main flood wave path would not
concurrently experience the same level of storm intensity. Furthermore, different
catchments would have different response times, and while one catchment could be
contributing its peak flow, the others could either be receding or intensifying. This
means that while the facility is experiencing the IDF, the downstream catchments
could be experiencing anywhere from bankfull flows (often represented by a 2-year
flood) to a flood equivalent to the IDF. The local climate, catchment characteristics,
and distance to the breached facility should guide the selection of the return period
flood event for each contributing catchment. Selecting high return period flow values
results in larger overall inundation and faster flood wave arrival time, but potentially
lower incremental inundation. Conversely, low return period flow values would result
in the opposite.
There is substantially less subjectivity in selecting flow conditions for a fair

weather scenario. Mean Annual Discharge (MAD) is typically used for downstream
watercourses, which is often easier to estimate and requires less professional judge-
ment and subjectivity. If the MAD is small compared to the breach flow (approximately
two times or more lower, as noted in FEMA [14]), it can be neglected in the hydro-
dynamic model as the flow would not materially increase the inundation, nor would
the watercourse storage materially change the attenuation of the breach flood wave.
The MAD should be included in the model for larger rivers where the impact of the
flow and storage is not negligible.
509
Q.105 - R.28
3.4. RHEOLOGY
Input data specific to tailings dam breaches consider the rheology of the hyper-
concentrated flows, often represented by the yield stress and the dynamic viscosity
at different volumetric solids concentrations. With increasing solids concentration in
the flood wave, the flow regime transitions from Newtonian flows (such as water)
to non-Newtonian flows (mud floods to mudflows to debris flows and landslides),
as classified by O’Brien [6] and applied to tailings flows in CDA TDBA Bulletin
[4]. The flow properties of non-Newtonian fluids are an active area of research,
which includes uncertainties associated with measuring and implementing these
properties into the flood wave routing of tailings dam breaches.
There are typically three main parameters that define the flow characteristics in
a hydrodynamic model: the solids concentration, the yield stress, and the viscosity.
The viscosity is a measure of the fluids flowability (e.g., fluids like honey with high
viscosity are less flowable), while the yield stress is a measure of the stress required
to mobilize the tailings mass at low confining pressures. There is limited published
data on viscosity and yield stress. Figure 2 represents the compilation of available
data from various sources.
These data include mudflow samples in Colorado (Glenwood, Aspen and Nat-
ural Soil) presented in O’Brien and Julien [15], various samples (Bentonite, Coussot,
Kaolinite, St. Helens and Quick Clays) presented in Julien [16], NA Tailings samples
from a copper / gold mine (yield stress only) presented in Adams et al. [17], tailings
slurry samples from a gold mine (Project X data, confidential), and samples from the
deposited tailings runout of the Fundão failure in Brazil in 2015 presented in Días
[18] (yield stress only) and Machado [19]. Generally, more yield stress than viscosity
data are available in various literature.
Tailings from hard rock mines (e.g., copper, nickel) behave differently at similar
solids concentrations than tailings from soft rock mines (e.g., coal), or tailings with
high clay content. For example, NA Tailings with higher clay content (>25%) and
Quick Clays shown on Figure 2, have a higher yield stress at lower solids concen-
trations than NA Tailings with a lower clay content (<25%), or Project X samples that
also had a very low clay content. Similarly, the Aspen samples had a higher clay
content and a higher yield stress than the Glenwood samples, as discussed in [15].
The iron tailings samples from the Fundão failure had a very high specific gravity
of about 4, which likely contributed to high flowability with low yield stress and low
viscosity at high solids contents.
Extrapolation is often required for tailings samples for volumetric solids con-
centrations above 45% - 55%, as these are difficult to measure (see example trend
for Project X in Fig. 2). Site-specific rheology data that would be representative of
the spatial variation and variation through the depth of the tailings deposit are dif-
ficult to procure and are rarely available. Furthermore, rheological trends may vary
510
Q.105 - R.28
Fig. 2
Yield Stress and Viscosity Variation with Volumetric Solids Concentration.
Variation de la limite d’élasticité apparente et viscosité avec concentration
volumétrique de solides.
temporary and spatially as the mine life progresses and the stored tailings settle and
densify within the facility.
Although viscosity and yield stress are commonly used in conventional

fluid mechanics to characterize the shear properties of a fluid, these can be
inadequate to describe some granular flows with low moisture content in which
511
Q.105 - R.28
particle-to-particle contacts become dominant and control the flow [4]. Additional
rheological constitutive relations from the field of continuum mechanics need to be
considered to characterize tailings flows for these cases; however, such tailings
flows are beyond the scope of this paper.
4. CASE STUDY 1
4.1. PROJECT BACKGROUND AND AVAILABLE DATA
The tailings facility in Case Study 1 is part of an active gold mine in South
America. The facility has a ring dyke configuration on a divide between two water-
sheds. The downstream area is rural with 20 to 30 permanent residences within
the first 20 km of the facility on one side, and mine facilities, stockpiles and open
pits on the other side of the facility. Multiple breach locations were considered for
this facility, but only the breach runout towards the populated rural farmlands is dis-
cussed in Case Study 1. The study area followed the downstream watercourses for
110 km and the total contributing watershed area was approximately 8,500 km2 .
The downstream watercourses are at a steady slope of 0.5%.
High resolution aerial surveys were available for the mine site. A satellite DEM
based on Shuttle Radar Topography Mission Version 3.0 (SRTM) was obtained for
the downstream watercourses and rural areas from the United States Geologic Sur-
vey (USGS) website. The DEM has a horizontal resolution of 30 m and a reported
accuracy of around 5 m, or better for flat terrains similar to the area downstream of
the facility.
The nearest hydrologic station on the breach flood wave path is 70 km down-
stream of the project with a relatively short flow record of 20 years. To confirm
appropriate flood flow values for the downstream watercourses, a regional hydrologic
analysis was conducted. Flood frequency analysis was completed on 30 hydrology
stations within 400 km of the project location with periods of record from 30 to 50
years. The IDF for the facility is a 10,000-year event, which was also adopted for the
immediate downstream watershed. Catchments further downstream were assigned
a 1,000-year event to reflect the intense monsoons experienced in the region.
Site-specific rheology information was not available, and as such, the rheol-
ogy was based on data found in literature (O’Brien and Julien [15], and Adams et
al. [17]). Based on comparison of site-specific tailings characteristics to those from
the literature, a strong and weak rheology (representative of low and high flowa-
bility, respectively) was assessed as part of the sensitivity analysis to evaluate the
impact of different rheologic parameters on breach runout progression and inunda-
tion. The strong and weak rheologic parameters that were used in this study were
approximately one order of magnitude higher and lower than the baseline rheology
(viscosity and yield stress).
512
Q.105 - R.28
4.2. SELECTED HYDRODYNAMIC MODEL
The selected model in this study was FLOW-3D, developed by Flow Science,
Inc. [20]. The Shallow Water Module in FLOW-3D was used in conjunction with
the non-Newtonian fluid properties. FLOW-3D models non-Newtonian flows using a
laminar flow approximation and either the Bingham or the Herschel-Bulkley general-
ized equation. The laminar flow approximation does not consider turbulent stresses
that could be significant for dam breach flows, flows on steep slopes, flows with low
solid concentrations (20% to 40% by volume), and flows on rough surfaces. Further-
more, at the time of the study and writing of this paper, FLOW-3D did not have the
capacity to simulate the dilution of the tailings laden breach flood wave with incoming
flows from downstream tributaries, and as such, the non-Newtonian fluid properties
(yield stress and viscosity) had to be set at a constant value that corresponded to
the solids concentration specific to a given model domain. The downstream model
extent was subdivided into smaller domains to capture the dilution due to incoming
larger tributaries.
4.3. MODEL RESULTS AND LESSONS LEARNED
During normal operations, the supernatant pond in this facility is more than
a kilometer away from the dam, and as such, it would not discharge during a fair
weather breach. The runout would constitute tailings solids and interstitial water only.
It would have a volumetric solids concentration of approximately 50% and would
behave like a mudflow, as discussed in [4], [5], and [6]. In consideration of the low
roughness associated with the grassy farmlands and shallow slopes downstream of
the facility, the turbulent stresses generated by the breach outflows were estimated
to be less dominant than the viscous forces governed by the rheological properties.
As such, the laminar approach in FLOW-3D was considered valid for modelling the
mudflow runout.
The total volumetric solids concentration was estimated to be between 20%

and 30% for the flood induced scenario. This scenario was modelled as a Newtonian
flow, as the breach flood wave would be diluted to below 20% solids concentration
upon reaching a larger watercourse located a short distance downstream of the
facility. The rheologic parameters would have a much smaller effect in this case than
the turbulence associated with the breach outflows.
Lack of site-specific rheology data proved to introduce a large variation in

the fair weather results, increasing the uncertainty of the study. The impact of the
rheology was evaluated as part of the sensitivity analysis, with the results shown on
Figure 3. The mudflow was predicted to travel more than twice as far when using
the weaker rheologic parameters (blue outline) compared to the stronger rheologic
parameters (red outline). This result demonstrates how sensitive the results could
be to these inputs, and highlights the need for the careful choices a professional
513
Q.105 - R.28
Fig. 3
Inundation extent under different assumed rheology trends.
Étendue de l’inondation avec différentes tendances rhéologiques supposées.
must make when selecting inputs in the absence of site-specific data. The results
of the assessment prompted the dam owner to collect rheology data to reduce
uncertainties in future studies.
5. CASE STUDY 2
The tailings facility in Case Study 2 is an inactive, reclaimed tailings impound-

ment located in a mountainous area in coastal North America. Reclamation work
included reshaping and capping the facility to prevent ponding of water, constructing
spillways and diversion channels around the facility that pass the IDF, and construct-
ing stabilization buttresses along the embankments. The slope immediately down-
stream of the facility is 50%. The average slope for the downstream watercourse is
10%, reducing to 2% on an alluvial fan approximately 3.5 km downstream where a
major river is intercepted. The total contributing watershed area is about 20 km2 .
The area of impact includes narrow canyons surrounded with dense mixed forests,
and a few farmlands with a small permanent population located adjacent to the major
river at the downstream end of the model. A DEM with a horizontal resolution of 1 m
and a vertical accuracy of around 15 cm was available for the dam breach study.
The IDF for the facility is the Probable Maximum Flood (PMF). Based on the
small total watershed area in this study, the downstream watercourse was also
assigned a PMF flow for the flood induced scenario. The MAD in the watercourse
was considered negligible and was not included in the fair weather scenario. The
flow in the major river was modelled at bankfull conditions in both scenarios to
assess the impact of potential backwatering on the breach flood wave propagation
through the populated area. Due to the geographic setting for the project and the
large watershed size for the major river, a large flood in the river coincidental with a
514
Q.105 - R.28
PMF at the facility is not likely to occur, which governed the selection of bankfull flows
for the flood induced scenario. Near-bankfull flows in the major river that last over
several weeks do occur regularly during the freshet period in the spring, which may
be coincidental with normal hydrologic conditions at the project site. This governed
the selection of bankfull flows for the fair weather scenario.
Considering that ponding is not possible in the tailings facility, the breach
outflows in this study would have a very high volumetric solids concentration of
about 54% in both the flood induced and the fair weather scenarios, resembling a
dense mudflow. Site-specific rheology information was not available, and as such,
the rheology was based on data found in literature (O’Brien and Julien [15], and
Adams et al. [17]). Similar to Case Study 1, a reasonable range in rheological
values was assessed as part of the sensitivity analysis.
The FLO-2D hydrodynamic model, developed by FLO-2D Software Inc. [21],

was selected for this study. FLO-2D uses a quadratic rheological model developed by
O’Brien and Julien [15] that describes the continuum of flow regimes from turbulent
to viscous flows as the solids concentration increases.
FLO-2D has limited options for modelling the initial/pre-breach conditions. The
model starts entirely dry in every simulation and the background flows must be routed
through the model domain for a period sufficiently long for these flows to reach a
steady state prior to starting the routing of the breach flood wave. The domain wetting
period is required for all scenarios in which the failure is coincidental with a flood
event. This can significantly increase the computational requirements, which is not
feasible for projects with large model domains such is the domain in Case Study
1. Case Study 2 had a relatively short model domain, and therefore, the additional
computation time was acceptable. In comparison, the hydrodynamic models used
in Case Studies 1 and 3 have the capability to set initial flow conditions for the
downstream drainage based on prior model results, and are therefore not limited
similarly to FLO-2D in this regard.
FLOW-3D was initially used in this case study; however, the conditions for
this project proved to be an example where the laminar flow approximation for
non-Newtonian flows in FLOW-3D was inappropriate. The flood induced conditions
encountered in this project include turbulent stresses developed due to steep slopes,
high roughness associated with flows through the densely forested narrow canyons,
and solids concentrations diluted with the background PMF flows, which would not
be accounted for in FLOW-3D.
515
Q.105 - R.28
Fig. 4
Comparison between FLO-2D and FLOW-3D for Case Study 2.
Comparaison entre FLO-2D et FLOW-3D pour l’étude de cas 2.
The non-Newtonian flood wave propagation in FLOW-3D was unstable and

faster than when the outflows were modelled as Newtonian fluids, which was not
considered realistic for this project. This led to selecting the FLO-2D platform for
routing the breach flood wave and modelling the inundation. A comparison between
FLO-2D and FLOW-3D is shown on Figure 4, which includes Newtonian and non-
Newtonian simulations in both models.
Sensitivity analysis on the rheologic parameters in FLO-2D indicated a similar

trend as in Case Study 1, where the flood wave travelled farther and faster when
using the weaker rheologic parameters (lower yield stress and viscosity) compared to
the flood wave with stronger rheologic parameters (higher yield stress and viscosity).
The resulting inundation was sensitive to both the selection of the hydrodynamic
model and to the rheology input data.
6. CASE STUDY 3
The facility in Case Study 3 is the same as in Case Study 1. The project
background and available data are detailed in Section 4.1. This case study focuses
on a breach scenario on the opposite side of the facility to Case Study 1, which would
outflow towards the mine facilities. The supernatant pond is located against a water
retaining dam on this side of the facility, and there is no tailings deposition adjacent
to this dam. The breach flood wave would contain eroded tailings that would get
mobilized from deeper within the facility as the pond discharges.
The breach flood wave was found to be contained by the open pit approxi-
mately 500 m downstream of the dam. The terrain between the dam and the open
pit is higher than the toe of the dam. The adjacent stockpiles and the elevated ter-
rain would interact with the breach outflows creating backwatering and impacting
the breach development. The breach outflow hydrographs and the peak discharges
516
Q.105 - R.28
were consequently modelled in conjunction with the downstream flood routing to cap-
ture the flow dynamics caused by the higher topography that would limit the outflow
volume. The modelling approach was therefore different than for Case Study 1.
The selected model was HEC-RAS Version 5.0.7 (2D), developed by the US
Army Corps of Engineers [22]. This hydrodynamic model was selected for its dam
breach tools and capability to account for backwatering as it calculates the out-
flow from the supernatant pond. HEC-RAS is currently limited to Newtonian flows,
and rheology cannot be included in the model. The breach outflow contained the
supernatant pond with eroded tailings and the volumetric solids concentration was
estimated to be between 20% and 30% using the methodology of Fontaine and
Martin [23]. Even though this flow would be non-Newtonian based on the solids
concentration, Newtonian flow was modelled in this case. It was anticipated that the
backwatering caused by the elevated terrain would have a greater impact on the
results compared to the effects of the rheology.
The modelling results demonstrated the considerable impact of the down-

stream terrain on the breach development and outflow, which in turn affected the
downstream inundation. The backwatering from the downstream terrain limited the
progression of the breach outflows within minutes, as shown on Figure 5.
Fig. 5
Breach hydrographs with and without backwatering effects.
Hydrogrammes de brèches avec effets de remous d’eau et sans effets de
remous d’eau.
517
Q.105 - R.28
With the downstream elevated terrain essentially acting as a weir, the outflow
extended over several days. The outflow volume in this case was roughly two thirds
of the volume that would potentially discharge if backwatering was not present. The
model selection in combination with the physiographic setting proved to be very
important in estimating the outflow hydrograph and the downstream impacts of this
hypothetical breach.
7. CONCLUSIONS
Tailings dam breach inundation estimates must be credible in order to be use-

ful, yet they must also be prudently conservative to guard against potential loss of
life and severe environmental, cultural, and economic consequences. These esti-
mates are based on multiple sources of data in combination with subjective model
parameter evaluations and assumptions, which all have different levels of impact on
the model results and final inundation extents.
1The results from the case studies underscore the importance of high qual-
ity input data requirements, as well as the professional experience and knowledge
required to select appropriate hydrodynamic models with consideration of the var-
ious simplifications inherent in these models. The case studies discussed in this
paper were recently completed, yet some of the approaches taken are already
becoming outdated due to the ongoing rapid advancement in both the development
of modelling tools and the quality of model inputs. Non-Newtonian flow modelling
tools are evolving at a rapid pace as research efforts and computational power
increase.
One of the major difficulties unique to tailings dam breach studies is the
availability of rheology data and the implementation of the non-Newtonian flow char-
acteristics within hydrodynamic models. Rheology data for stored tailings are rarely
available; however, dam owners are starting to collect these data to improve the
input requirements for dam breach studies.
The British statistician George Box wrote that “all models are wrong, but some
are useful,” and this statement is very applicable to tailings dam breach studies
and the associated hydrodynamic models. Dam breach assessments are never
simple and will always carry uncertainties, but they can provide useful informa-
tion when combined with good professional judgement and experience. The quality
of tailings dam breach studies will continue to improve with advances in data col-
lection technologies, improvements in the scientific understanding of breaching
processes and non-Newtonian flows, and progresses in the engineering state of
practice.
518
Q.105 - R.28
8. ACKNOWLEDGEMENTS
The authors gratefully acknowledge the opportunity to conduct these tailings

dam breach assessments and learn from these experiences. They wish to thank
their clients and colleagues for their valuable advice and support during the actual
dam breach assessments and preparation of this manuscript.
REFERENCES
[1] Global Tailings Review, Global Industry Standard on Tailings Man-

agement. https://globaltailingsreview.org/wp-content/uploads/2020/08/global-
industry-standard_EN.pdf (2020).
[2] Agência Nacional de Mineração, Resolution No. 13, of August 8, 2019.

https://www.in.gov.br/web/dou/-/resolucao-n-13-de-8-de-agosto-de-2019-210
037027
[3] Agência Nacional de Mineração, Resolution No. 32, of May 11, 2020.
https://www.in.gov.br/web/dou/-/resolucao-n-32-de-11-de-maio-de-2020-2572
01163
[4] Canadian Dam Association, 2020. Technical Bulletin: Tailings Dam Breach
Analyses. (Draft posted on CDA Member Portal, August 2020).
[5] Martin, V., Al-Mamun, M., Small, A., CDA technical bulletin on tailings dam
breach analyses, in Proceedings of the International Commission on Large
Dams 2019 Annual Meeting, ICOLD 2019, June 9-14, Ottawa, Canada (2019).
[6] O’Brien, J.S., Physical Processes, Rheology, and Modeling of Mud Flows.
Ph.D. Dissertation, Colorado State University, Fort Collins, USA (1986).
[7] Wahl, T.L., Prediction of the Embankment Dam Breach Parameters – A Lit-
erature Review and Needs Assessment. Dam Safety Rep. No. DSO-98-004,
U.S. Dept. of the Interior, Bureau of Reclamation, Denver, USA (1998).
[8] Wahl, T.L., Evaluation of Erodibility-Based Embankment Dam Breach Equa-

tions. Hydraulic Laboratory Report HL-2014-02U.S. Dept. of the Interior,
Bureau of Reclamation, Denver, USA (2014).
[9] West, M., Morris, M., Hassan, M., A Guide to Breach Prediction, HR
Wallingford (2018).
[10] Brunner, G., Using HEC-RAS for Dam Break Studies. Training Document no.
39, U.S. Army Corps of Engineers, Institute for Water Resources, Hydrologic
Engineering Center, Davis, USA (2014).
519
Q.105 - R.28
[11] Martin, V. and Akkerman, A., Challenges with conducting tailings dam breach
assessments, in Proceedings of the International Commission on Large Dams
2017 Annual Meeting, ICOLD, July 3-7, Prague, Czech Republic (2017).
[12] Goodell, C., Moving Towards Risk-Based Dam Breach Modelling, in Pro-
ceedings of the Association of State Dam Safety Officials 2013 Dam Safety
Conference, ASDSO 2013, September 8-12, Providence, USA (2013).
[13] U.S. Army Corps of Engineers (USACE), Institute for Water Resources,
Hydrologic Engineering Center, Modeler Application Guidance for Steady ver-
sus Unsteady, and 1D versus 2D versus 3D Hydraulic Modeling. Training
Document no. 41, Davis, USA (2020).
[14] Federal Emergency Management Agency (FEMA), U.S. Dept. of Homeland

Security, Federal Guidelines for Inundation Mapping of Flood Risks Associated
with Dam Incidents and Failures, FEMA P-946, (2013).
[15] O’Brien, J.S., and Julien, P.Y., Laboratory Analysis of Mudflow Properties, in
the Journal of Hydraulic Engineering, Vol. 114, No. 8, August (1988).
[16] Julien, P.Y., Erosion and Sedimentation (2nd ed.), Cambridge University Press
(2010).
[17] Adams, A., Friedman, D., Davidson, S., Characterizing and Stabilizing a His-
torical Tailings Facility: The Rheology to Soil Mechanics Continuum, in the
Proceedings of the Tailings and Mine Waste 2017 Conference, November 5-8,
Banff, Canada (2017).
[18] Días, N. A., Determinação de Propriedades Reológicas de Rejeito de Min-

eração por meio de Reômetro Rotacional. Universidade Federal do Espírito
Santo Centro Tecnológico Programa de Pós Graduação em Engenharia Civil,
(2017).
[19] Machado, N. C., Retroanálise da Propagação Decorrente da Ruptura da

Barragem do Fundão com Diferentes Modelos Numéricos e Hipóteses de Sim-
ulação. Universidade Federal de Minas Gerais, Programa de Pós-Graduação
em Sameamento, Meio Ambiente e Recursos Hídricos, (2017).
[20] Flow Sciences, Inc., FLOW-3D Version 12.0. Retrieved on November 22, 2019
from: https://www.flow3d.com/
[21] FLO-2D Software Inc., FLO-2D Build No. 18. Retrieved on April 6, 2020 from
https://flo-2d.com/ https://flo-2d.com/
[22] U.S. Army Corps of Engineers Hydrologic Engineering Center (USACE

HEC), HEC-RAS Version 5.0.7. Retrieved on February 14, 2020 from
https://www.hec.usace.army.mil/software/hec-ras/
[23] Fontaine D.D. and V. Martin. 2015. Tailings Mobilization Estimates for Dam
Breach Studies. Proceedings of the 2015 Tailings and Mine Waste Conference.
Vancouver, BC, October 26-28, 2015.
520
Q.105 - R.29
-------
-------
INCIDENTS AND FAILURES OF SMALL EARTHFILL DAMS IN

SOUTH AFRICA: LESSONS LEARNED (∗)
D.J. HAGEN
Director,
DJ HAGEN & ASSOCIATES (PTY) LTD, Cape Town
Louis C. HATTINGH
Chief Executive Officer, HATTINGH ANDERSON ASSOCIATES CC, Pretoria
SOUTH AFRICA
SUMMARY
This paper presents practical lessons learned from incidents and failures
of smaller earthfill dams in South Africa. The most important of these can be
summarized as follows:
• Internal erosion and slope failure:

◦ Proper and appropriate clay core and core trench compaction.
◦ Proper core cut-off trench excavation down to foundation material with
permeability less than that of the core or deeper into a rock foundation
which may degrade over time as well as across the full length of the
embankment.
◦ Overlapping of core trench excavation sections, proper tie-in with the
original embankment core and ensuring that the core extend vertically
far enough to limit downstream inundation, is very important aspects to
attend to during construction as well as dam raisings.
∗ Les enseignements tirés des incidents et des défaillances des petits barrages en terre en
Afrique du Sud
DOI: 10.1201/9781003211808-75 521

Q.105 - R.29
◦ The use of welded HDPE pipes with stainless 316 steel specials (at either
end outside the earthfill) encased in appropriately designed reinforced
concrete is considered a good long-term investment for smaller dams
over rigid or mild steel pipes.
• Overtopping:
◦ Allow for long-term settlement of the embankment and its foundation no
matter the size of a dam by providing sufficient camber.
• Spillway erosion:
◦ Lining of the vulnerable part or the complete spillway of a small dam to
prevent undercutting or erosion of rock can be a costly, but necessary
expense.
RÉSUMÉ
Cet article présente les enseignements pratiques tirés des incidents et des
défaillances de petits barrages en terre en Afrique du Sud. Les plus importants
d’entre eux peuvent être résumés comme suit:
• Érosion interne et glissement de talus :

◦ Compactage correct et approprié du noyau d’argile et du noyau de la
tranchée.
◦ Excavation de tranchée de coupure appropriée jusqu’au matériau de
fondation avec une perméabilité inférieure à celle du noyau ou plus pro-
fonde dans une fondation rocheuse qui peut se dégrader au fil du temps
ainsi que sur toute la longueur du remblai.
◦ Le chevauchement des sections d’excavation de la tranchée principale,
le raccordement approprié avec le noyau du talus d’origine et la garantie
que le noyau s’étende suffisamment à la verticale pour limiter les inon-
dations en aval, sont des aspects très importants à prendre en compte
pendant la construction ainsi que le soulèvement des barrages.
◦ L’utilisation de tuyaux en PEHD soudés avec des matériaux spéciaux
en acier inoxydable 316 (à chaque extrémité à l’extérieur du talus)
enrobés de béton armé correctement conçu est considérée comme un
bon investissement à long terme pour les petits barrages sur des tuyaux
en acier rigide ou doux.
• Dépassement:
◦ Permettre le tassement à long terme du talus et de ses fondations, quelle
que soit la taille du barrage, en assurant une cambrure suffisante.
• Érosion de l’évacuateur de crue :
◦ Le revêtement de la partie vulnérable ou de l’évacuateur de crue d’un
petit barrage pour empêcher le sapement ou l’érosion de la roche peut
être une dépense coûteuse, mais nécessaire.
522
Q.105 - R.29
1. INTRODUCTION
The latest update of the ICOLD incidents database [1] highlighted the most
prevalent failure modes for the different types of dams. For earthfill dams overtopping
(44%) is the most prevalent failure mode followed by internal erosion (35%). All
other structural failures including slope failures and spillway erosion only represents
20% of all failures. These values are similar to the percentages observed in South
Africa for incidents and failures [2]. Unfortunately, in many cases the design and
construction of smaller dams earthfill do not take cognizance of the importance of
considering especially overtopping and internal erosion as potential failure modes
resulting in unnecessary incidents and/or failures. Also, for smaller dams the failure
mechanisms causing incidents and failures are not always investigated in detail due
to the lack of funding and limited time available for repair resulting in a lost opportunity
for lessons to be learnt. This paper discusses practical lessons learnt from 9 case
histories of incidents or failures that happened in smaller earthfill dams in South
Africa focusing not only on internal erosion and overtopping potential failure modes
but also spillway erosion and slope failure.
2. INTERNAL EROSION
2.1. FOUNDATION CORE CUT-OFF TRENCH
The importance of excavating the core trench down to foundation material with
permeability less than that of the core or deeper into a rock foundation which may
degrade over time as well as across the full length of an embankment dam, are
emphasised by these case studies.
In the case of the 23 m high Houtkaprug Dam, originally completed in 1997,

the core trench on the right flank was founded on an in-situ permeable alluvial layer.
Slightly turbid foundation seepage of up to 6 /s occurred on the right flank at higher
reservoir levels. The dam was repaired in 2013 by excavating the core trench in the
upstream slope of the embankment down to an impermeable foundation consisting
of clayey material (in-situ weathered siltstone). The complete core zone of the right
flank section of the embankment was also re-constructed. The seepage flow was
reduced to less than 1 /s greatly reducing the potential for internal erosion.
Another example is the 22 m high Betko Dam, a homogeneous earthfill dam,

completed in 1992. Although some foundation seepage was recorded earlier, a sud-
den increase in seepage was reported in 2011 peaking to 20 /s in 2012. Sinkholes
were reported on the upstream face on the right flank located at the interface of the
embankment and the in-situ soil foundation. During July to August 2013 the dam was
523
Q.105 - R.29
repaired by excavating a new deeper and wider core trench into the shale founda-
tion at the upstream heel connecting to the impervious homogeneous embankment.
Following the repair work, no seepage from the right flank has been reported.
In the case where the core trench is excavated in sections, it is important

to adequately back excavate into the previously constructed section of the core
trench. This is particular important in areas around the outlet pipe trench and on the
abutments during dam raisings. This is highlighted by the following case.
The 19 m high Rondekop Dam breached during the first filling after raising in
June 2009 due to piping along the new outlet pipe. The original dam was completed
in 1979. Investigations after the failure revealed that the core trench excavations
either side of the pipe trench was not adequately back excavated into the pipe
core trench section which was constructed first. Foundation seepage initiated in the
resulting “window” and continued along the outlet pipe encasement. Unfortunately,
adequate filter protection (a collar) around the outlet pipe encasement was also not
in place, which would have stopped the migration of core material. The dam was
repaired immediately after the event with a new central core trench and gypsum
stabilized earthfill core both protected on the downstream side by an appropriate
sand and rock filter. The second dam safety evaluation of the dam completed in
2015, confirmed that the dam was performing well since the repair.
2.2. CLAY CORE COMPACTION
The importance of the correct clay core compaction as first line of defense
against internal erosion cannot be overstated. Experience has shown that com-
paction to a minimum of 98% of Standard Proctor density at a moisture content
between Optimum Moisture Content (OMC) and +3% OMC is in most cases an
acceptable standard for use with typical clay core earthfill material. This is highlighted
by the following case.
Construction for raising the De Eike Dam to a height of 12 m wall commenced in

June 2014 just as the annual winter rains started in the Western Cape. Unfortunately,
clay core compaction too wet of OMC occurred. This resulted in serious desiccation,
with transverse as well as longitudinal cracking. Fortunately, this was picked up
during a construction site inspection by the designer. The dam was repaired by re-
construction of the core zone with the core material at the correct OMC related to
the appropriate Standard Proctor density.
Field density test results from the nuclear density testing on the Cape Granite
derived clayey sand, indicated that although higher compaction was achieved dry of
OMC and wet of OMC, the specified density of 98% of Standard Proctor density was
not achieved. The results highlighted the principle that field moisture of core material
must first be corrected to OMC to attain the specified Standard Proctor density.
524
Q.105 - R.29
2.3. FAILURE OF OUTLET PIPE MATERIALS
The improper design and construction of outlet pipes is also one of the major
causes of internal erosion incidents or failures of smaller earthfill dams. Many inci-
dents have been reported where rigid outlet pipes (asbestos-cement, concrete,
etc.) have failed structurally due to differential settlement including those previ-
ously reported by Hagen et al in 2012 [3]. One such case is the 32 m high Oubos
Dam whose original 250 mm diameter asbestos-cement (AC) pipe was repaired by
inserting a new 225 mm diameter High Density Polyethylene (HDPE) pipe inside the
original pipe. The cavity between the two pipes was grouted with a non-shrink flow-
able grout at low pressure to limit the potential for hydraulic fracturing of the earthfill
material around the pipe. The repair followed cracking of the original pipe shortly
after first filling in 1996, due to differential settlement of the outlet pipe foundation.
Another issue that are regularly observed, is internal pitting corrosion of gal-
vanized or painted mild steel welded outlet pipes. In the past, specifying a steel pipe
wall thickness of 8 mm, was considered adequate. In the case of the 11 m high
Green Oaks Dam, leakage was reported near the downstream end of the 165 mm
outside diameter galvanized steel (5 mm wall thickness) concrete encased outlet
pipe. Rehabilitation of the outlet pipe was recommended. A camera inspection of
the inside of the outlet pipe revealed severe corrosion on the inside of the pipe with
the corrosion product (“rust’) at surface. The pipe was similarly repaired as above.
To limit both the structural failure rigid outlet pipes as well as the potential
for internal corrosion pitting it was found that Class 10 welded HDPE pipes with
stainless 316 steel specials at either end outside the earthfill encased in appropri-
ately designed reinforced concrete encasement, is considered a good long-term
investment for small dams.
3. OVERTOPPING
3.1. EMBANKMENT SETTLEMENT
A simple, but important design detail that unfortunately is ignored in many

smaller earthfill dams, is to make allowance for long-term settlement of the embank-
ment and its foundation. An allowance of a camber of 2% of the wall height for the
embankment settlement is typically specified no matter the height of the dam.
Klipfontein Dam is a 10 m high dam which was constructed many years ago.
The dam failed in January 2014 due to overtopping. From site observations it was
clear that the embankment overtopped across its full length as erosion gullies were
present all along the downstream face. The overtopping head was the greatest at
525
Q.105 - R.29
the maximum embankment height where the largest long-term settlement of the
embankment occurred and therefore the lowest freeboard. The breach occurred at
this low point in the embankment.
4. SLOPE FAILURE
4.1. EXTENDING THE CORE DURING RAISING
Dam embankment raisings require careful attention especially for limited rais-
ings, which are regarded as a quick solution to attain more storage capacity.
Adequate overlapping of core trench excavation sections is very important (see
above), so is adequate tie-in with the original core of the dam. In the case where
only the spillway is raised and not the non-overspill crest of the embankment, it is
important to check if the embankment core extends vertically high enough. Expe-
rience have shown that attention should be given to the core extension contact
interface as well as to extend the core to 300 mm below the crest with a gravel
capping layer to seal-off the core.
The original Vrede Dam B was constructed before some time before 2000.
The dam was raised by 2 m to a maximum wall height of 12 m in 2008. High level
seepage occurred on the original/raised core contact interface, which lead to the
downstream slope becoming saturated. This in turn resulted in a downstream slope
failure. The dam was operated at lower levels for years after this to prevent further
deterioration of the embankment. It is currently under rehabilitation by constructing
a new core trench and core on the upstream side to embankment crest level.
5. SPILLWAY EROSION
Spillway erosion is another failure mechanism that makes up the significant

part of incidents at smaller earthfill dams. Although costly, careful consideration
should be given to the lining of the vulnerable part or the complete spillway of a
small dam to limit the risk of this potential failure mode.
The spillway of the 24 m high Rocklands-Berg Dam (constructed in 1976/1977

and raised in 1999) consists of a concrete labyrinth overflow, situated on the right
flank with an unlined sandstone rock downstream spillway channel. A 35 m section
in the middle of the 90 m long channel, which consisted of soft weathered sandstone,
was expected to erode over time. The dam owner elected not to incur the cost of
lining the vulnerable section of the spillway at the time. A 10 m deep hole eroded
in this section. Water in this large hole discharges through eroded tunnels into the
526
Q.105 - R.29
surrounding sandstone rock. Some of the water daylights close to the toe of the dam,
resulting in the concern of undermining of the toe of the embankment. Furthermore,
there is a risk that left flank of the spillway control structure could also be undercut.
Design of the repair of the spillway is currently in progress.
6. MAIN STATISTICS OF CASE HISTORY DAMS
The main statistics of the dams listed in this paper are provided in the table
below.
Table 1
Main statistics of case history dams
Principales statistiques de cas de barrages
DAM NAME MAXIMUM WALL STORAGE COMPLETION DATE OF
HEIGHT (M) CAPACITY (M3 ) DATE INCIDENT
Houtkaprug 23 1 600 000 1997 2013
Betko 22 115 000 1992 2011/2012
Rondekop 19 323 000 2009 2009
De Eike 12 100 000 2014 2014
Oubos 32 276 000 1996 1996
Green Oaks 11 90 000 before 2000 2019
Klipfontein 10 300 000 before 2000 2014
Vrede Dam B 12 100 000 2008 (raising) 2008
Rocklands-Berg 24 900 000 1999 (raising) 2007 to date
7. CONCLUDING REMARKS
From the lessons learnt from the case histories presented as well as other
experience of similar cases for smaller embankment dams, the following is of
importance:
• Internal erosion and slope failure:

◦ Proper and appropriate clay core and core trench compaction.
◦ Proper core cut-off trench excavation down to foundation material with
permeability less than that of the core or deeper into a rock foundation
which may degrade over time as well as across the full length of the
embankment.
◦ Overlapping of core trench excavation sections, proper tie-in with the
original embankment core and ensuring that the core extend vertically
527
Q.105 - R.29
far enough to limit downstream inundation, is very important aspects to

attend to during construction as well as dam raisings.
◦ The use of welded HDPE pipes with stainless 316 steel specials (at either
end outside the earthfill) encased in appropriately designed reinforced
concrete is considered a good long-term investment for smaller dams
over rigid or mild steel pipes.
• Overtopping:
◦ Allow for long-term settlement of the embankment and its foundation no
matter the size of a dam by providing sufficient camber.
• Spillway erosion:
◦ Lining of the vulnerable part or the complete spillway of a small dam to
prevent undercutting or erosion of rock can be a costly, but necessary
expense.
ACKNOWLEDGEMENTS
The authors wish to express their gratitude to the dam owners for their permis-
sion to report on the incidents/failures. The opinions expressed are, however, those
of the authors and do not necessarily reflect the views of the dam owners.
REFERENCES
[1] ICOLD. ICOLD Incident database – Bulletin 99 update: Statistical analysis of

dam failures. Draft bulletin, Paris, 2020.
[2] HATTINGH, L, NORTJE, J, VAN DEN BERG, L & OOSTHUIZEN C. Lessons

learnt from Dam Safety Incidents in South Africa During the Past 100 + Years.
Q93/R31. ICOLD 24th Congress, 2012.
[3] HAGEN, D.J. & SHELLY, A.J. Practical lessons learned during statutory dam
safety evaluations since 1986. Q93/R18. ICOLD 24th Congress, 2012.
528
Q.105 - R.30
-------
-------
MIDDLE LAKE DAM SPILLWAY SAFETY INCIDENT AND

LESSONS LEARNT (∗)
Henriette ANDERSON
Dam Engineer,
Verve Civil Eng (Pty) Ltd, Pretoria
Philip NICE
Regional Executive Manager,
Roads and Storm Water, City of Ekurhuleni
Louis C HATTINGH
Chief Executive Officer,
Hattingh Anderson Associates CC, Pretoria
SOUTH AFRICA
SUMMARY
This paper discusses the lessons learnt from the emergency caused by the
dam safety incident that occurred with the spillway of the Middle Lake Dam in South
Africa, when two panels failed on the downstream slope of the spillway during
December 2018. The following is of importance:
• When dealing with spillways using a similar layout of concrete panels on top
of earthfill material, attention during planning and design should be given
to the long-term settlement of the earthfill material, the potential for erosion
through the joints and long-term operation and maintenance considerations.
Using properly designed structural concrete slabs with appropriate water
∗ Enseignements tirés de l’incident de sécurité de l’évacuateur de crues du barrage de Middle
Lake
DOI: 10.1201/9781003211808-76 529

Q.105 - R.30
stops will mitigate the initiation of the potential failure mode experienced dur-
ing this incident and greatly reduce the need for continuous monitoring and
maintenance.
• In the case where a spillway however consists of non-structural concrete slabs
without any proper water stop system (of which many exist unfortunately), the
following are relevant:
◦ The regular proper inspection and monitoring of the condition of the
spillway including the joints is essential.
◦ Proper regular maintenance of especially the joints, is also essential.
With this in mind, dam owners should allocate funds every year for
normal maintenance and plan and budget for major maintenance and
rehabilitation. This should also be clearly spelled out in the Operational
and Maintenance Manuals.
• Emergency Preparedness Plans must be made available to Disaster Manage-
ment authorities at all times.
• The use of social media platforms for the public to report suspicious activities
or possible incidents. The importance of effective communication during such
an incident.
• The officials responsible for dams in a municipal environment should be men-
tored/trained to properly undertake routine inspections (including identifying
potential failure modes) and should not hesitate to communicate with the APP
when they have identified areas of concern.
• Siphons proved to be a cost-effective way to draw down water levels in dams
where the outlet works are no longer operational or effective.
RÉSUMÉ
Ce rapport met en lumière les enseignements tirés de l’incident survenu sur

l’évacuateur de crue du barrage du Middle Lake Dam en Afrique du Sud, lorsque
deux dalles se sont rompues sur la pente en aval de l’évacuateur de crue en
décembre 2018. Il est à noter que :
• Dans les cas où les évacuateurs de crue utilisent une disposition similaire de
dalles de béton sur le matériau de remblai, il convient, lors de la planifica-
tion et de la conception, de prêter attention au tassement à long terme du
matériau de remblai, au potentiel d’érosion par les joints et aux considéra-
tions d’exploitation et d’entretien à long terme. L’utilisation de dalles de béton
structurelles correctement conçues et dotées d’arrêts d’eau appropriés permet
d’atténuer le déclenchement du mode de défaillance potentiel rencontré lors
de cet incident et de réduire considérablement la nécessité d’une surveillance
et d’un entretien continus.
• Dans les cas fréquents où un évacuateur de crue est constitué de dalles de
béton non structurelles sans système d’arrêt d’eau approprié, les éléments
suivants sont pertinents :
530
Q.105 - R.30
◦ L’inspection et la surveillance régulières et appropriées de l’état de

l’évacuateur de crue, y compris des joints, sont essentielles.
◦ Un bon entretien régulier, notamment des joints, est également essen-
tiel. Dans cette optique, les propriétaires de barrages doivent allouer
chaque année des fonds pour l’entretien normal et planifier et budgé-
tiser les gros travaux d’entretien et de réhabilitation. Cela doit également
être clairement énoncé dans les manuels d’exploitation et d’entretien.
• Les plans d’interventions d’urgence doivent être mis à la disposition des
autorités de gestion des catastrophes à tout moment.
• L’importance d’une communication efficace pendant un tel incident est essen-
tielle et les plateformes de médias sociaux pourraient être un moyen pour le
public de signaler des activités suspectes ou des incidents éventuels.
• Les fonctionnaires responsables des barrages dans un environnement munic-
ipal devraient être formés pour leur permettre d’effectuer correctement les
inspections de routine (y compris l’identification des modes de défaillance)
1. INTRODUCTION
In general, the perception exists that small dams don’t require as much design,
operation and maintenance effort as large dams. As a result, the design solutions
for small dams don’t always take into account the various long-term risks especially
from an operation and maintenance point of view. Some small dams especially those
in urban environments, also poses a significant threat to the environment (loss of
life, damage to property, loss of function etc.). In addition, the typical organiza-
tion hierarchy in the local government environment in South Africa normally places
dams under the auspices of the roads department with an ensuing general lack of
knowledge with regards to the requirements for proper operation and maintenance
of these structures.
To highlights these, this paper discusses the lessons learnt from the emer-
gency caused by the dam safety incident that occurred with the spillway of the Middle
Lake Dam in South Africa, when two panels failed on the downstream slope of the
spillway during December 2018. The lessons learnt in this case study should assist
small dam owners with their dam safety, to make decisions and to take corrective
actions to deal with dam safety shortcomings or deficiencies identified at their dams,
in order to safeguard the public and associated resources.
2. BACKGROUND
The City of Ekurhuleni in South Africa owns a number of small dams, some of
which have been registered under the Dam Safety legislation as dams with a safety
risk. In addition, a number of these small dams have a significant hazard potential.
531
Q.105 - R.30
As is mostly the case in the municipal environment, these dams are considered as
being part of storm water management which falls under the auspices of the Roads
and Stormwater Department at the City of Ekurhuleni.
The dam infrastructure of the City of Ekurhuleni is closely link to the discov-
ery of gold in the area in 1897. A chain of dams had been built by 1895, which
included the Homestead and Kleinfontein Dams for the water supply to the mines
[1]. Middle Lake Dam was constructed between Homestead and Kleinfontein Dams,
somewhere between 1895 and 1912. According to Gardiner [1], the original Middle
Lake Dam wall was lowered, including the introduction of a concrete lined spillway
over part of the main embankment, to allow for the construction of the N12 highway
in the late 1960s. All these dams are currently owned by the City of Ekurhuleni.
3. MIDDLE LAKE DAM
The main purpose of Middle Lake Dam, a homogeneous earthfill embankment

dam with a maximum height of 8.3 m, is flood control. This dam is also extensively
used for recreational purposes and to a lesser degree to provide water for irrigation
of a nearby golf course. Although the dam is only 8.3 m high, it has a significant
storage capacity of 513 098 m3 and directly downstream of the dam is a number
of major roads leading to the city center, a functioning railway, a private hospital as
well as regional shopping centre. As a result, the dam is registered and classified
according to the South Africa dam safety regulations as a category II dam with a
significant hazard potential. The existing spillway, located in the river section, is a
broad-crested type spillway which is lined on the crest and downstream slope with
non-structural concrete panels supported by the original earthfill embankment. The
spillway also has a downstream stilling basin which is also lined with concrete panels.
The joints between the concrete panels were originally designed to only contain a
joint filler that requires constant monitoring and regular maintenance. It seems that
the original outlet of the dam was never moved during the relocation of the spillway
and is as a result situated in the spillway section. The outlet, unfortunately, is no
longer functional resulting in no capability to draw the dam down during emergencies.
Cavities underneath the spillway panels due to erosion of the earthfill material
as a result of open joints was already evident during the second compulsory dam
safety evaluation done in 2002 [2]. Subsequently extensive repair work to the spillway
was completed at the end of 2004, which included grouting of the cavities beneath
the concrete panels, sealing of the joints between the concrete panels with joint
filler and the installation of a cut-off beam on the upstream side of the spillway
[3]. Unfortunately, no monitoring or maintenance of the joints between the spillway
slabs were done subsequent to the rehabilitation work to negate any further erosion
of the earthfill material. During the fourth compulsory dam safety evaluation [4], it
was observed that extensive vegetation growth was evident in the joints between
the spillway panels (see Fig. 1). The concrete panels were also uneven indicating
532
Q.105 - R.30
Fig. 1
Middle Lake Dam: Spillway December 2018
Barrage de Middle Lake: Evacuateur de crue Décembre 2018
settlement of the panels where erosion of the soil has taken place at the interface of
the soil with the panels. Other dam safety deficiencies were also identified including
insufficient freeboard and the non-functioning outlet.
3.1. SPILLWAY INCIDENT DURING DECEMBER 2018
During the latter part of December 2018, two of the concrete panels collapsed
on the downstream slope (see Fig. 1). This was first noticed by a member of the
public and reported to a local newspaper (Benoni City Times) on 21 December 2018.
The dam owner became aware of the problem through social media reporting by the
newspaper reporter 2 days later on 23 December 2018. After a visit to the dam by
the Approved Professional Person (APP) on the same day together with officials of
the City of Ekurhuleni, emergency measures were implemented to reduce the threat
of the potential failure of the spillway. It is important to note that the dam was spilling
at the time of this visit.
During the site visit it become apparent that the Disaster and Emergency Man-
agement Services of Ekurhuleni (DEMS) did not know about the existence of the
533
Q.105 - R.30
Emergency Preparedness Plan (EPP) for the dam. The potential risk as a result
was verbally communicated to DEMS by the APP and the EPP was made avail-
able electronically. Once made aware of the risk the DEMS immediately alerted all
the relevant stakeholders should the need to evacuate arise. The following emer-
gency measures were implemented immediately by the owner with support from the
national Department of Water and Sanitation (DWS):
• Construction of a sandbag cofferdam wall upstream of the affected area to

divert the water away during spilling.
• Drawdown of the reservoir water level to prevent water spilling over the dam-
aged spillway and to create a buffer to mitigate flood events. As a result of
the non-functionality of the outlet works, this was initially attempted using
pumps which proved not to be that successful. The pumping was subsequently
replaced by the installation of six siphon pipes. In addition, the water levels
at Homestead Dam, Middle Lake Dam and at Civic Lake were daily manually
monitored together with rainfall predictions on an ongoing basis.
• During the effort to draw the water level down, it also became evident that the
downstream water course was partially blocked by sediment and vegetation
growth raising the risk of overtopping to the major downstream roads. The
downstream water course was cleaned to mitigate the overtopping risk.
3.2. CAUSE OF THE SPILLWAY INCIDENT
The cause of the spillway incident was investigated by the APP that amongst
other things included a detailed survey, including an underwater survey of the dam
basin, as well as ground-penetrating radar scanning of the spillway and the down-
stream stilling basin area to determine the location and extent of anomalies (cavities)
underneath the concrete panels. From the investigation the following conclusions
were made:
• The design of the current layout of the spillway did not allow for the potential
long-term differential settlement of the earthfill embankment and foundation.
As a result, the concrete panels were not designed as structural members
but rather as erosion protection with minimum steel. The joints were also just
sealed with standard joint filler with no provision for proper water stops that
would be able to handle joint movement and hydraulic forces.
• No maintenance was done on the joints since the completion of construction
in the 1960s except for the rehabilitation effort in 2004.
• The initiator of the failure of the concrete panels was as a result of the ingress
of water over many years through the open joints also on the upstream side
of the panels, causing erosion of the material at the interface of the concrete
panels and the supporting earthfill material and/or foundation material, mainly
on the downstream slope of the spillway and in the stilling basin.
• The repair work done in 2004 appeared not to have effectively addressed the
erosion in the area where the concrete panels failed.
534
Q.105 - R.30
• The long-term erosion of the earthfill material left some of the concrete panels
suspended in air. As a result, two of them subsequently failed due to their
structural frailties. It was postulated that any substantial flows over the spillway
in the vicinity of the failed panels will result in water coming into direct contact
with the earthfill under super critical conditions as well as the potential of the
removal of the panels downstream of the failed panels as well as to their sides
due to the dynamic hydraulic action of the super critical flows leading to the
rapid failure of the spillway section due to backward erosion (similar to the
mechanism of overtopping failure).
• The erosion damage below the slabs was also evident for the rest of the
spillway and stilling basin.
3.3. REHABILITATION
The remedial works were planned in three phases of which Phase 1 represents
the emergency repair in the area affected by the failed spillway slabs that took place
immediately following the incident. The other two phases are planned as soon as
regulatory approval is obtained.
3.3.1. Phase 1 (completed)
The spillway incident happened during the annual builders’ break when it is
notoriously difficult to get contractors and building materials to name a few chal-
lenges during this time. However, the damage to the spillway in the localized area
was assessed and an engineering solution to repair the damaged area of the spill-
way was proposed taking into account available materials and equipment supplied
to be by the contractor Rostruct (previously known as Rodio), the City of Ekurhuleni
and DWS Construction. The remedial works involved the following:
• Initial risk mitigation measures including building of a sandbag wall on the

spillway crest and the installation of six siphons to draw down the water level –
see above.
• In the immediate area where the two panels failed the following rehabilitation
methodology was followed:
◦ The pumping of around 80 m3 of cementitious grout material into the cav-
ities downstream of the failed concrete panels including the stilling basin
area to mitigate the potential failure mechanism described above. Woven
geotextile layers and bags were used to make allowance for drainage of
the phreatic water on the downstream slope before grouting took place
as well as to contain the grout material in the relevant areas as there
was also erosion damage (cavity or tunnel) evident underneath the still-
ing basin which had to be repaired to ensure the successful repair of
535
Q.105 - R.30
the localized area. This grouting also had to be done in phases to allow
the grout to attain strength to limit potential downstream failure of the
concrete panels due to uplift.
◦ This was followed by the demolition and removal of the failed concrete
panels. Once removed the cavity below the failed concrete panels were
filled using a pumped high slump plasticized concrete mix followed by
the reinstatement of the concrete panels. The grouting of the cavities
under the adjoining concrete panels followed using the same methodol-
ogy described above. The concrete work was completed on 4 February
2019.
◦ Subsequently, the joints on the spillway have been sealed and minor
concrete repairs done to some concrete panels.
◦ Automatic digital water level recorders were installed to replace the
manual water level monitoring.
• In addition, the responsible officials in control of the dams at the City of
Ekurhuleni were trained to undertake routine inspections and introduced to
the requirements of the dam safety legislation including regular monitoring
and maintenance.
3.3.2. Planned phases 2 and 3
Phase 2 will consist of the repair of the erosion damage on the rest of the
spillway and stilling basin which has not been done during Phase 1. This phase will
continue from the works completed during Phase 1, using the results of ground-
penetrating radar survey. This phase will also include the grouting of the outlet pipe
which is no longer operational to eliminate the potential for internal erosion.
The rehabilitation of Phase 3 will be undertaken sometime in the future and will
include the repair of erosion damage to and reinstatement of the earthfill embank-
ments non-overspill crest, concrete repair to guide walls and the stilling basin end
wall, and the removal of major trees and stumps. This will also include the repair
of the upstream riprap slope protection and installation of permanent outlet pipes in
the right flank embankment.
A number of important conclusions can be made from this particular incident

not only pertaining to the importance of proper implementation of dam safety but also
with regards to taking cognizance of potential failure mechanisms during design as
well as the difficulties experienced in the local government environment to properly
operate and maintain infrastructure. The following is of importance:
536
Q.105 - R.30
• When dealing with spillways using a similar layout of concrete panels on top of
earthfill material, attention during planning and design should be given to the
long-term settlement of the earthfill material, the potential for erosion through
the joints and long-term operation and maintenance considerations. Using
properly designed structural concrete slabs (based on the beams on elastic
foundations principles) with appropriate water stops will mitigate the initiation of
the potential failure mode experienced during this incident and greatly reduce
the need for continuous monitoring and maintenance.
• In the case where a spillway however consists of non-structural concrete slabs
without any proper water stop system (of which many exist unfortunately), the
following are relevant:
◦ The regular proper inspection and monitoring of the condition of the
spillway including the joints is essential. This could also include the use
of ground penetrating radar should any erosion be suspected.
◦ Proper regular maintenance of such a layout, especially the joints, is also
essential. With this in mind, dam owners should allocate funds every year
for normal maintenance and plan and budget for major maintenance and
rehabilitation. This should also be clearly spelled out in the Operational
and Maintenance Manuals.
• Emergency Preparedness Plans must be made available to Disaster Manage-
ment authorities at all times.
• The use of social media platform for the public to report suspicious activities
or possible incidents.
• Proper training for the officials responsible for dams in a municipal environment
including timeous communication with the APP when they have identified areas
of concern.
• Siphons proved to be a cost-effective way to draw down water levels in dams
where the outlet works are no longer operational or effective.
ACKNOWLEDGEMENTS
The authors thank the City of Ekurhuleni for permission to publish this paper.
The opinions and views presented in this paper are, however, those of the authors
and do not necessarily reflect those of the City of Ekurhuleni. The authors also wish
to express their gratitude to everybody involved with the first phase of emergency
rehabilitation of Middle Lake Dam.
REFERENCES
[1] Gardiner, K., ed. Benoni – A Golden Anniversary for a Golden City.
https://www.norval.co.za/benoni/Benoni-History-23.pdf, Accessed 3 February
2019.
537
Q.105 - R.30
[2] Mercer, KG & Van Bladeren, D. Dam safety inspections in Greater Benoni:
Stability analysis for Homestead Lake, Rynfield Dam and Middle Lake. Report
no. 271204/1. SRK Consulting, Illovo. South Africa, 2002.
[3] Van Bladeren, D. Letter dated 25 February 2004 addressed to the Dam Safety
Office with reference to Raising of Spillway Flank Walls and NOC: Middle Lake
(File no. 12/2/C121/44) including SRK plan 271204/113 of proposed works
and Form 695E application to alter dam. SRK Consulting, Illovo. South Africa,
2004.
[4] Anderson, H. Dam Safety Evaluation of Middle Lake Dam (12/2/C11/44).

Report No. P17047/R. DeltaBEC, Pretoria. South Africa, 2018.
538
Q.105 - R.31
-------
-------
MALPASO DAM INCIDENT AND POST-EVENT ACTIONS (∗)
Anibal MAITA
Deputy Manager Hydropower Planning & Dam Safety Officer,
STATKRAFT PERU
Dr Eduardo Martins BRETAS

Dam Safety Specialist
STATKRAFT ENERGI AS and SINTEF Narvik
PERU
SUMMARY
The Malpaso dam, which started operations in 1936, is a concrete-face rockfill

dam (CFRD), about 78 m high and 156 m long. The dam is in the central region
of Peru and serves the purpose of regulating the water from the Mantaro river to
obtain the necessary head for the Malpaso hydropower plant. in October 2019, small
openings appeared on the dam crest, which gradually formed a longitudinal crack,
parallel to the upstream face. In response to this small incident, immediate actions
included (i) Lowering the reservoir level, (ii) Having external and Statkraft’s experts
inspect the dam, and (iii) Performing topographical and piezometric surveys, and
visual daily inspections. Geophysical tests were then performed, and based of the
test results, in addition to the opinion of experts, a decision was made to seal up the
crack and establish a monitoring system with the installation of inclinometers and
piezometers. As a result, a comprehensive dam monitoring will be conducted in real
time. The dam is currently operating under normal conditions.
∗ Incident du barrage de Malpaso et actions post évènement
DOI: 10.1201/9781003211808-77 539

Q.105 - R.31
RÉSUMÉ
Le barrage de Malpaso, qui a été mis en service en 1936, est un barrage

à masque amont en béton (CFRD), d’environ 78 m de haut et 156 m de long. Le
barrage se trouve dans la région centrale du Pérou et sert à réguler l’eau de la
rivière Mantaro pour obtenir la chute nécessaire pour la centrale hydroélectrique
de Malpaso. En octobre 2019, de petites fissures sont apparues sur la crête du
barrage, et elles ont progressivement formé une fissure longitudinale, parallèle
au parement amont. En réponse à cet incident, les mesures immédiates compre-
naient (i) l’abaissement du niveau du réservoir, (ii) l’inspection du barrage par des
experts externes et de Statkraft, (iii) l’exécution de relevés topographiques et pié-
zométriques, et des inspections visuelles quotidiennes. Des tests géophysiques
ont ensuite été effectués ; sur la base des résultats des tests, en plus de l’avis
d’experts, il a été décidé de sceller la fissure et d’établir un système de surveillance
avec l’installation d’inclinomètres et de piézomètres. Par conséquent, une surveil-
lance complète des barrages sera effectuée en temps réel. Le barrage fonctionne
actuellement dans des conditions normales.
1. OVERVIEW
All dam-related incidents, however small or insignificant they appear to be,

shall be addressed diligently, especially in connection with old dams, as they may
cause irreparable damages, or even worse, dam failure.
Specifically, the Malpaso dam has been operating for more than 84 years
and an incident occurred in October 2019, which consisted in a crack suddenly
appearing on the dam crest. The incident was timely addressed, which permitted
taking appropriate actions, and implementing follow-up and corrective measures,
which are detailed in this paper.
As part of the learning and lessons learned from this incident, we underline
the importance of: i.- having a simple and reliable control system, ii.- working in
a multidisciplinary team, iii.- having the opinion and judgment from experts, iv.-
appropriately applying state-of-the-art technology, among other things.
Last but not least, a decision was made to install new instrumentation to
improve the current control system and put forth the planning of future actions aimed
at guaranteeing the safe operation of the dam.
2. DAM DESCRIPTION
The Malpaso Dam is a rockfill dam located on the upper reaches of the Mantaro
river, in the central region of Peru, which started operations in October 1936. The
540
Q.105 - R.31
purpose of the dam is to regulate the Mantaro river water and obtain the necessary
head for the eponymous hydropower plant (ref. 1). This plant is currently owned by
Statkraft Perú S.A.
The slope of the upstream shell is 1:2 (H:V) and is slightly convex towards the
upstream side, with an initial radius of 762,0 m, built with manually and mechanically
shaped rock, whose gaps were filled with fine materials from quarry waste and
coarse sand free of clays and any type of soil through water jets at the maximum
possible pressure. In order to obtain a uniform and compacted bed for the impervious
screen, a masonry layer was built, whose thickness varies from 0,60 m on the crest
up to 1,20 on the base.
The impervious screen is composed of reinforced concrete slabs of 24,0 m in

width in the lower part and slabs of 12,0 m in width in the upper part. In-between
there are vertical joints of 5 cm of thickness filled with asphalt and covered all along
the joint with steel plating of t = 1/4” and 30 cm in thickness. The thickness of the
screen varies from 0,255 m on the crest up to 0,90 m on the base.
The slope of the downstream shell is 3:4 and 3:5 (V:H), built with quarry mate-
rial, including rock, rubble and unsifted coarse sand, dumped from the necessary
height, whose maximum density was achieved with high-pressure water jets. The
first years of operation, the central area of the dam crest was displaced downstream
as much as 54,0 cm, a figure which was deemed high back then by the specialists,
who decided, as a corrective measure, to increase the downstream shell, complet-
ing construction works in December 1939 (see Figure 1). Subsequent controls show
an adequate dam behavior.
In order to obtain an impervious closure and prevent seepages underneath

the dam, a concrete cut off was built that extends from the foot of the impervious
Fig. 1
Dam cross section
541
Q.105 - R.31
screen up to solid bedrock. The cutoff is also convex towards the upstream side with
a radius of 61,0 m.
For maximum floods, a spillway is located on the left bank of the dam, which
is fitted with 5 radial gates and a free weir of 40 m in length.
Other major dam characteristics are described below:
Initial height (dam centerline): 77,54 m

Current height (dam centerline): 77,08 m
Crest length: 156,00 m
(slightly convex towards the upstream side)
Maximum cross section: 182,3 m
Crest width: 8,50 m
Maximum operating water level: 3 875,53 masl
Minimum operating water level: 3 863,64 masl
Gate invert level: 3 872,48 masl
Useful volume: 23,57 Hm3
Dead volume: 22,0 Hm3
Total volume: 45,57 Hm3
Reservoir length: 13,0 km
Maximum water surface area: 2,9 km2
3. DAM BEHAVIOR BEFORE THE INCIDENT
3.1. MONITORING PARAMETERS
The instrumentation is simple and has been used to control the Malpaso
dam behavior, through monitoring programs from the start of operations to date,
considering the following parameters:
3.1.1. Deformations and Displacements at the Dam and Screen Joints
The deformations of rockfill-type dams typically depend on two factors, one of

which derives from its own weight and the other from the water pressure acting on the
impervious dam screen after the reservoir has been filled. When filling the reservoir,
the pressure acts on the rockfill causing some rocks and fine material to be displaced
to more stable positions, which changes conditions corresponding to the balance
542
Q.105 - R.31
of all other rocks and fine material. As the rearrangement of the rockfill decreased
over time, the progress and velocity of deformations also decreases (ref. 2).
In this case, the subsidence (vertical deformations), displacements (horizontal

deformations) on the dam crest have been considered; these parameters permit to
analyze whether the progress of deformations is detrimental to the dam stability.
The vertical joints in the screen allow for relative movement between the slabs
due to horizontal and vertical deformations; in this sense, this monitoring provides
information on the expansion and contraction at the joints, as well as the incidence
on such movement on the impervious screen.
For displacement and subsidence, and displacement on the screen joints,

the control is carried out at the dam crest. To this end, twelve points have been
installed on the impervious screen, symmetrically distributed along the crest, as
shown in Figure 2. Each point is composed of two steel rods located on both ends
of the vertical joint, at an initial distance of 457 mm. For subsidence, we start from
a benchmark on the solid rock at the right end of the dam and measurements are
performed at each point through precision leveling. For displacements, a control
axis is used, the ends of which are placed on solid rock. With a theodolite (currently
total station), the axis is located, and measurements are taken between the control
points and control axis.
Fig. 2
Control points – Dam crest
For displacement in the screen joints (see figure 3), the horizontal control is
carried out by measuring the distance between the rods located on both ends of the
vertical joints with a duly calibrated metal ruler used exclusively for this purpose.
The vertical control is carried out through the leveling between the heads of the
abovementioned rods.
543
Q.105 - R.31
Fig. 3
Control at the screen joints
3.1.2. Seepages
Seepages in rockfill-type dams may occur through the screen or the cut off
and are rarely caused by rains, in which case, due to the dam material, seepages
are drained forming a hydraulic gradient. Monitoring is carried out downstream of
the dam, which allows to find the piezometric levels, hydraulic gradient, as well as
to indirectly confirm the performance of the impervious screen and concrete cut off.
For seepage control, three Casagrande-type piezometers have been installed

at 16,40 m, 21,20 m and 78,90 m, from the foot of the dam. (See figure 4).
Fig. 4
Piezometers
544
Q.105 - R.31
3.1.3. Reservoir levels
The reservoir level variations are frequent and occur hourly or daily due to the
regulation for the Malpaso hydropower plant operation. In order to analyze whether
the level variation, especially during the first years of operation, had any influence
on the progress of deformations at different water surface levels, the corresponding
monitoring has been carried out.
3.1.4. Maximum floods
Controlling maximum floods is of paramount importance as the spillway may

behave as a bottleneck section which would raise the water level and therefore cause
a probable overflow.
3.1.5. Sedimentation
It permits having control of the sediments in the reservoir and its impact on
the dam behavior.
3.2. MONITORING
The data has been processed, and the values obtained from an 83-year control
period (ref. 3) until before the incident, have been analyzed and interpreted.
3.2.1. Subsidence
The expected subsidence for extended monitoring periods as per LANTON-

LESTER is given by:
S = 0, 01 ∗ H 3/2 (1)
Where:
S = Subsidence (m)
H = Initial Dam Height (m)
545
Q.105 - R.31
Fig. 5
Subsidence on the dam crest
Table 1
YEARS SUBSIDENCE “E” SUBSIDENCE SPEED RELATIVE SUBSIDENCE
(CM) (CM/YEAR) (MM/M)
2 26.0 13.0 3.4
5 36.0 3.3 4.6
10 38.0 0.4 4.9
30 42.0 0.2 5.4
61 46.0 0.1 5.9
81 (∗ ) 46.0 0.0 5.9
∗ The entire registration period
For Malpaso, according to formula [1], S = 68,3 cm and the last control, the
subsidence was recorded at 46,0 cm, equivalent to 67% of the expected value in
accordance with Lanton-Lester.
Figure 5 shows the subsidence at control points on the dam crest. Subsidence
is low at both ends and high at points around the centerline as the ends are in
contact with the bedrock. During the first years of operation, subsidence has been
several centimeters and has decreased over time, as expected from the behavior of
rockfill-type dams.
Table 1 shows a summary of three variables at point “E” (dam centerline).

The relative subsidence is the relation between the subsidence and the dam height.
These results indicate that the dam has basically consolidated.
By way of illustration, figure 6 shows the relation between the dam opera-
tion times (days) and the subsidence/height factor, for which information on four
Hungarian dams has been obtained and compared with the Malpaso dam. For the
first 3 000 days of operation, Malpaso had 4,8 mm/m, a value similar to that of the
Haznos dam and higher than the other three dams (ref. 4 & 5).
546
Q.105 - R.31
Fig. 6
Comparison of subsidence with other dams
Fig. 7
Displacement on the dam crest
3.2.2. Displacement
Figure 7 shows the displacements with respect to the control axis and the
results have the same behavior as those obtained for subsidence, i.e., they are low
at the ends and high at the dam centerline.
Table 2 shows a summary of three variables at point “E”. The relative displace-
ment is the relation between the displacement and the crest length. These results
confirm the statement above that the dam is basically consolidated.
547
Q.105 - R.31
Table 2
YEARS DISPLACEMENT DISPLACEMENT RELATIVE
“E” SPEED DISPLACEMENT
(CM) (CM/YEAR) (MM/M)
2 54.0 27.0 3.5
5 73.0 6.3 4.7
10 77.0 0.8 4.9
30 86.0 0.5 5.5
61 92.0 0.2 5.9
81(∗ ) 92.0 0.0 5.9
∗ The entire registration period
Fig. 8
Displacement at the screen joints
3.2.3. Displacement at the screen joints
Figure 8 shows the results of controls in the three most representative points
of the impervious screen. For the dam end points 200L and 200R, the joints have
expanded up to 5 mm with respect to the first control, whereas in point “E”, the joint
has contracted up to 7 mm. The expansion of the ends and the contraction at the
center is explained to some extent by the convex shape of the dam crest.
Regarding the vertical control, a maximum level difference of 6 mm was found

in point “E” between the two control rods.
548
Q.105 - R.31
Fig. 9
Levels in piezometers
3.2.4. Seepages
Figure 9 shows the levels in piezometers No.1 and No.3, which increase after
the seismic event that occurred on October 17, 1966 in an area close to the dam.
From April 1975 onwards, the levels increase – in average – 2,4 m compared to
October 1966 and these levels are kept around this figure to date.
The maximum hydraulic gradient between piezometers No.1 and No.3 was 7%
as logged in September 1959. In all the monitoring period, the hydraulic gradient is
1,8 % in average. Both the hydraulic gradients and the levels logged in the piezome-
ters confirm that the dam body was built with rocks and gravel material (free of fine
material) and the behavior is as expected.
3.2.5. Water surface levels and discharge flows
The maximum historical level was 3 875,98 masl and it has been estimated
that the level, due to waves, will reach 3 877,10 masl, 1,48 m below the dam crest.
The maximum historical inflow was 392,5 m3 /s.
In 2019, the study “Breakage Analysis and Supplementary Studies – Malpaso

Dam” was performed, where the flood routing was estimated for a 10 000 year return
period. The determined outflow was 626,0 m3 /s, and the gates fully opened plus the
549
Q.105 - R.31
overflow through the spillway successfully satisfy the function of discharging this
peak flow and therefore no overtopping is expected to occur. This peak flow will
reach a level of 1,42 m below the dam crest (ref. 6).
3.2.6. Sedimentation
Three bathymetries have been performed in the Malpaso reservoir, the last
of which was performed in July 1998, and according to its results, the total
sedimentation volume added up to 4 Hm3 in 62 years of operation, which is equiva-
lent to approximately 64 500 m3 per year, a very low value, taking into account that
the dead volume of the reservoir is 22 Hm3 .
In such study from 2019, it was estimated that the sediment level at the foot
of the dam is 11,7 m over the original level and 2,4 m over the level from 1998.
3.3. CONDITION ASSESSMENT
The condition assessment is a way to analyze the state of a facility through

inspections and/or direct measurements of the facility’s components. The results of
each inspection lead to a classification as shown in table 3, where the state scale
ranges from 1 to 5, where 4 is the poorest condition and 5 is the state of failure.
In Statkraft Peru, we have adapted this methodology for the scheduled inspec-
tions of dams, both to assess the conditions of components and to manage the
risks.
Table 3
STATE DESCRIPTION ACTION
1 No indication of degradation (“as No action
good as new”)
2 Some indication of degradation. The No specific action is necessary for further operation, but it
condition is noticeably worse than “as may be necessary to increase the frequency of inspection
good as new” to secure necessary control of the development
3 Serious degradation. The condition is Corrective action will normally be technically necessary
considerably worse than “as good as and may be also economically profitable but need not be
new” carried out immediately. The operation is maintained, and
any actions are carried out at the earliest opportunity
4 The condition is critical Action should be taken prior to further operation
5 Fault state
550
Q.105 - R.31
The latest condition assessment (scheduled inspection) at the Malpaso dam

prior to the incident was on February 10, 2019, when no finding or anomaly was
detected at the dam.
3.4. DAM BEHAVIOR RESULTS
The dam monitoring is performed through a simple and reliable system, pro-
viding the necessary and sufficient information that has been useful to know the
dam behavior over the 84 years it has been operating.
The monitoring results have demonstrated its effectiveness in dam control

ensuring its safety. It also permitted to detect an unusual displacement of the dam
crest in the first two years of operation, as a result of which the downstream shell
was raised as a corrective measure.
4. INCIDENT AT THE DAM
On October 16, 2019, personnel from the Generation area detected 2 holes of
approximately 15 cm in diameter on the dam crest, at approximately 40,0 m from the
right abutment and 0,90 m from the upstream border, right in the direction of control
point 120R (see figure 10). The reservoir level was 4,69 m below the dam crest.
Subsequently, on October 18, the holes were found to be approximately 20 cm in
diameter and 1,2 and 1,6 m in depth, respectively.
During the inspection conducted on October 20, 03 additional holes were

detected with depths between 0,6 and 1,3 m. Subsequently, on October 25, the
crack of approximately 3,0 in length, 0,6 m in width and 2,0 m in depth was detected.
As mentioned previously, during the last inspection (February 10, 2019) prior
to the incident, no findings or anomalies on the dam were reported. Moreover, in
the technical dossier of the dam no information was found as to holes or cracks
appearing on the dam in the past.
The evolution of the incident is shown in figure 10.
Due to the relatively quick progress of the crack formation, the actions detailed
in the following section were implemented.
551
Q.105 - R.31
Fig. 10
Incident Location and Evolution
4.1. IMMEDIATE AND FOLLOW-UP MEASURES
The following immediate and follow-up actions were undertaken:
i.- The dam does not have undersluice gates and the only way to lower the
water level is through the Malpaso HPP headrace tunnel, which only permits to
552
Q.105 - R.31
Fig. 11
Levels of control points after the incident
reach 11,0 m below the dam crest. The reservoir level was therefore lowered from
the day the holes were detected (October 16, 2019), until the abovementioned level
was reach in approximately 6 days. This level was kept until after the corrective
measures were implemented.
ii.- On the third day after the incident, an in-depth inspection of the entire dam
was led by the Dam Safety Officer of Statkraft Peru. In this inspection, no other
findings were found in addition to the holes.
iii.- Report of the incident area every other day by the engineer in charge of the
dam operations. Subsequently, on October 25, the crack was stable without further
progress.
iv.- High-precision leveling at the control points on the dam crest and
measurements at the piezometers.
For reference purposes, the level measurements were compared with those
taken on October 13, 2017 and as observed in figure 11, no significant differences
were found with the controls after the incident.
The levels in the piezometers were kept within normal ranges, as shown in
figure 12.
v.- Inspections with drones equipped with high-resolution cameras, which per-
mitted to observe the condition of the impervious screen and showing only normal
wear due to the dam age. On the downstream slope, mainly the contacts between
the dam and the abutments were inspected, without finding any anomalies. (See
figure 13).
553
Q.105 - R.31
Fig. 12
Piezometric levels after the incident
Fig. 13
Inspections with drones
vi.- Dam inspection by three team of experts: 1.- National consultant,

2.- International consultant and 3.- Statkraft. The main conclusion of the three teams
of experts was that the incident was localized and would not affect the structural dam
behavior.
vii.- The Emergency Response Plan (ERP) of the dam was ready, but it was not
required to be activated for the reasons explained. It should be noted that Statkraft
Peru conducted an internal drill mainly to reconfirm the communication channels
and responsibilities of each member of the ERP.
554
Q.105 - R.31
4.2. CORRECTIVE MEASURES
Two corrective measures were undertaken, as detailed below.
4.2.1. Geophysical dam assessment
In order to characterize the dam structure, obtain more information and detect
a possible anomaly in the dam body, a consultant specialized on seismic refraction
tests was engaged, and 4 types of exploration were used: i).- 12 seismic refraction
tests, ii).- 14 Multichannel Analysis of Surface Waves (MASW), iii).- 12 Microtremor
Array Measurements (MAM) and iv).- 8 electrical tomography lines (ref. 7).
These explorations were also used to determine the shear strength of the
soil, determine the seismic-resistant parameters, characterize the materials of
the dam structure and identify the geotechnical issues based on the geophysical
investigation.
Through the interpretation of the seismic refraction tests, MASW and MAM,
three representative horizons on the dam body were obtained, the first as a dense
soil from the surface up to approximately 8,0 m, then a soil of a lower density than the
first horizon up to 20,0 m and the last horizon up to the bottom of the dam composed
of very dense soil. Given the shear wave velocity, the presence of weak areas would
be ruled out and, therefore, soil liquefaction issues are ruled out.
With the electrical tomography tests, a ground water profile was inferred as
shown in figure 14. It is observed that water seeps towards the bottom, an expected
Fig. 14
Inferred ground water profile based on the electrical tomography
Source: Geophysical Assessment of the Malpaso Dam – JCI Cosultores.
555
Q.105 - R.31
behavior given the dam material; however, there is no certainty as to the water levels,
especially in areas close to the impervious screen.
Regarding the crack, the consultant concludes that, due to the location in
parallel to the slope, it does not continue and does not reach the slope face; it
cannot be correlated to a deep sliding.
4.2.2. Crack filling
Taking into account the thorough follow-up of the incident, the conclusions
from experts, the completion of corrective studies and especially that the incident
was localized, a decision was made to fill the crack before the heavy rainy season,
which starts in January.
The works were carried out between December 3 and 11, 2019, and first,
the affected area of 2,0 m in width and 8,0 m in length was uncovered. At a depth
of 0,3 and 0,6 m, rocks of 0,8 to 2,0 m in equivalent diameter were found, and the
space in-between was filled with non-classified and fines-free granular material. It
should be noted that in the crack area, the rock conditions were similar to those of
the intervened area.
The filling was carried out with material similar to the material found, i.e.
medium-sized stones, gravel and sand free of fine material. Figure 15 shows a
sequence for the crack filling.
Fig. 15
Sequence for filling the crack
556
Q.105 - R.31
4.3. MEASURES TO IMPROVE THE CONTROL SYSTEM
While the control of the dam crest for the last 30 years shows that there are
basically no further deformation, the results of the geophysical assessment indicate
that the dam body material is dense or very dense; it should be noted that the
material is heterogeneous and there is no geotechnical instrumentation to know the
condition of stresses and deformations inside the dam.
The ground water levels are measured downstream of the foot of the slope,
and these have increased compared to the measurements from the initial period. On
the other hand, the electrical tomography permitted to infer a ground water profile in
the dam body, the levels of which have a certain degree of uncertainty, in addition
to being specific, requiring to keep a log of ground water levels in the dam body.
In view of the previous paragraphs, a decision was made to improve the control
system through the installation of the following instrumentation, which is currently
being implemented:
• Three inclinometers on the dam crest, one in the centerline and the other in
the direction of control points 120 L and 120R. The latter corresponds to the
area where the incident occurred.
• Fifteen vibrating wire piezometers, the logs of which will be available almost
in real time. On the crest, 3 boreholes will be made with the same layout of
inclinometers and 3 piezometers will be installed in each borehole. 2 boreholes
in the dam body each with 2 piezometers. Then, another piezometer close to
the foot of the dam and the last in the same platform of the existing piezometers.
• Three extensometers at the screen joints, one in control point “E” – dam
centerline and the other two in points 120L and 120R.
Figure 16 shows the plan and profile layout of the new instrumentation that
will start to operate in November 2020.
Fig. 16
New dam instrumentation
557
Q.105 - R.31
5. FUTURE ACTIONS
The concrete slabs of the impervious screen on the upstream dam face are
subject to tensile and compressive stresses due to the hydrostatic and hydrodynamic
loads of the reservoirs, under static and dynamic conditions (seismic events), and
have been in operation for more than 84 years and it is likely that they have lost
part of their cementing properties, which affects their structural and permeability
properties (ref. 8).
The current ground water levels in piezometers are approximately 2,3 m com-
pared to the levels in the initial period, which - among other reasons – may be
explained by the fact that the joints in the screen that are protected by metal plates
have deteriorated by age, which would be causing seepages flowing into the dam
body. This situation of ground water levels would indicate that the water inflow would
be concentrated in the lower portion of the dam.
In view of these aspects, future dam rehabilitation and, if required, dam rein-
forcement actions are proposed, mainly aimed at waterproofing the impervious
screen and ensuring the dam operation and safety.
6. CONCLUSIONS
The Malpaso dam has been operating non-stop for more than 84 years in an
optimal and safe manner.
The incident that occurred in October 2019 consisted in a crack suddenly

appearing on the dam crest, and, after the analysis and judgment of three teams
of experts, and the results of the seismic refraction tests, it was concluded that
the crack was localized and did not affect the structural dam behavior. The dam is
currently operating without restrictions.
The dam monitoring is conducted with a simple and reliable system which has
been useful to know the dam behavior. Its effectiveness has been demonstrated
guaranteeing the dam safety.
The – multidisciplinary – team work permitted to implement immediate, follow-

up and corrective actions in a diligent and effective manner.
State-of-the-art technology, such as drones, HoloLens, IMAR, real wear,

among others, are great support for dam inspections, which has been made evident
in the case of Malpaso.
558
Q.105 - R.31
As part of the continuous improvement on dam safety, a decision was made

to improve the current control system with the installation of new instrumentation
composed of inclinometers, piezometers and extensometers. Along these lines,
future dam rehabilitation actions and, if required, dam reinforcement actions are
proposed to continue to ensure safe dam operation.
REFERENCES
[1] MAITA ANIBAL. Malpaso Dam Assessment and behavior in case of a

contingency. Centromin Perú S.A. La Oroya – Peru, 1994.
[2] KARL TERZAGHI & RALPH PECK. Soil mechanics in practical engineering.
2nd Edition, 1986.
[3] MAITA ANIBAL & ÑAUPARI HILDEBRANDO. Malpaso Dam Monitoring. XI

National Congress of Civil Engineering. Trujillo – Peru, 1997.
[4] Hydro Review Worldwide. Volume 1, Number 2, Fall 1993.
[5] Hydro Review Worldwide. Volume 2, Number 3, Fall 1994.
[6] Poyry. Breakage Analysis and Supplementary Studies – Malpaso Dam. Peru,
2019.
[7] JCI CONSULTORES. Geophysical Assessment of the Malpaso Dam. Peru,

2019.
[8] MIGUEL DE LA TORRE SOBREVILLA – SZ DESSAU. Assessment of Physical

Conditions of Malpaso Dam. Peru, 2019.
559
Q.105 - R.32
-------
-------
REHABILITATION OF THE OPERATING COMOE DAM LATERITIC

FOUNDATION (∗)
A.F. CHRAIBI
Independent dam consultant,
Morocco
A. NOMBRE
Dam Expert,
Burkina Faso
S. RIHI
Engineering Geology expert,
Novec Morocco
MOROCCO
SUMMARY
Comoé dam in Burkina Faso is an embankment dam of 24 m maximum height

above the riverbed and around 1170 m crest length. The dam foundation on the
abutments is lateritic ground made of a superficial hard duricrust covering a lateritic
carapace, soft to firm layer of 6 to 10 m thickness with developed canaliculus and
small cavities. This facies is sitting on a variable thickness highly weathered dolerite
at the top, sound and watertight at depth. Since the first impoundment of the dam in
1991, uncontrolled seepage appeared at the downstream toe of the embankment
on both abutments. The seepage flow was more important in the left bank. This flow
gradually increased with time. In 2010, the situation became critical with develop-
ment of subsidence downstream in the left abutment. Based on extensive investiga-
tions, remedial works were launched in 2013 and achieved in 2018, making a stop
∗ Réhabilitation de la fondation latéritique du barrage de la Comoé en exploitation.
560 DOI: 10.1201/9781003211808-78

Q.105 - R.32
from June 2015 to July 2017, due to local critical situation. The treatment included a
positive sheet piles cutoff of 40 m maximum depth, tubes-à-manchettes grouting and
compaction grouting. It had to be conducted together with the normal reservoir oper-
ation, except lowering the reservoir by 1m below the normal elevation. At the end
of the work, a significant reduction in seepage flow was observed, which has been
maintained over the last 3 years. No resurgences and no new subsidence occurred.
RÉSUMÉ
Le barrage de la Comoé au Burkina Faso est un barrage en remblai d’une

hauteur maximale de 24 m au-dessus du lit du cours d’eau et d’une longueur en
crête d’environ 1170 m. Étant donné sa situation en zone tropicale, la fondation du
barrage sur les rives est constituée, de haut en bas, d’une cuirasse latéritique, de la
carapace de dureté variable de 6 à 10 m d’épaisseur avec des canalicules et petites
cavités, puis d’une dolérite d’épaisseur variable, très altérée au sommet, saine et
étanche en profondeur. Depuis la première mise en eau du barrage, intervenue
en 1991, des fuites incontrôlées sont apparues au pied aval du barrage aux deux
appuis. Le débit des résurgences était plus important sur la rive gauche. Celui-
ci a progressivement augmenté avec le temps. En 2010, la situation est devenue
critique avec le développement d’affaissements dans la zone de résurgences en rive
gauche. Sur la base d’investigations approfondies, des travaux de confortement
ont été lancés en 2013 et achevés en 2018, avec un arrêt de juin 2015 à juillet
2017, en raison d’une situation locale critique. Le traitement comprenait une coupure
étanche en palplanches d’une profondeur maximale de 40 m, une injection de coulis
à l’aide de tubes-à-manchettes et une injection dite de compactage par mortier. Les
travaux ont été menés tout en assurant l’exploitation normale du réservoir. Seul un
abaissement de la retenue de 1m a été consenti. À la fin des travaux, une réduction
significative du débit des fuites a été observée, laquelle s’est maintenue au cours
des 3 dernières années. Aucune résurgence et aucun nouvel affaissement ne se
sont produits.
1. INTRODUCTION
Comoé dam, is located close to Moussodougou village in the South-Western

part of Burkina Faso upstream of Banfora city, the fourth largest city of Burkina
Faso (fig. 1). The dam, built during the period 1989 to 1991, was commissioned
and impounded in July 1991. Comoé dam creates the main reservoir regulating
the Comoé river for the development of sugar cane irrigation up to 4 000 ha and
water supply to the city of Banfora. The reservoir has a catchment area of 500 Km2
and 38 Mm3 capacity. Due to its location upstream of Banfora city on the Comoe
561
Q.105 - R.32
Fig. 1
Dam location and reservoir Satellite view
Localisation du barrage avec l’image satellitaire de la retenue
River which cross Ivory Coast country reaching the Atlantic Ocean, Comoé dam is
classified as a high hazards dam.
During the first filling of the reservoir, seepage started just at the downstream
toe of the dam on both abutments. The seepage flow increased very rapidly during
the early years (1991 to 1995). The increase in the monitored seepage flow slowed
down but without stabilization. As only a part of the seepage was resurging (Photo 1),
Photo 1
Downstream toe of the left abutment: main seepage area with
local subsidence (left), part of seepage flow (right)
Pied aval du barrage en rive gauche: Zone des principales venues
d’eau avec affaissement local (à gauche), une partie de débit
des venues d’eau (à droite)
562
Q.105 - R.32
the hidden (underground) flow and some emerging seepage on the right abutment
and further downstream reach of the river, escaped being measured so that its
progress with time could not be assessed.
Statistical modeling of the seepage flow collected in the left bank up to 2010
is represented in figure 2. After 2010 the model was not any more reliable due
to unexplained variations in outcropping seepage flow, likely in connection with a
transfer of seeping paths from the surface to the ground.
Fig. 2
Statistical modeling of the seepage flow on the left bank
(correlation coefficient = 0.98), X = years and Y = flow (l/s)
Modèle statistique du débit des venues d’eau en rive gauche
(coefficient de corrélation = 0,98), X = années et Y = Débit (en l/s)
Since 2005, worsening signs of the situation appeared as subsidence spots

developed further in the ground downstream of the dam on the left abutment. Con-
currently, one longitudinal depression of 10 to 30 cm depth, locally 50 cm, in line
with the emerging seepage zone, was observed on the upstream face of the dam
at el. 451asl, 3 m below the normal water level. Except for the upstream, most
of depressions’ magnitude increased rapidly during wet years particularly under
high reservoir level. The situation was therefore considered critical, jeopardizing the
dam’s safety. It was then decided to limit the maximum reservoir level to 1 m below
the normal storage level and to carry out necessary investigations to obtain a real-
istic view of the main flow paths and an objective assessment of the gravity of the
situation.
563
Q.105 - R.32
2. MAIN FEATURES OF THE DAM
Comoé dam is a pseudo homogeneous embankment with a vertical chimney

filter-drain, having the following main features:
• The dam crest has 1 170 m length and 6m width, a maximum height of 31 m
above the foundation (24 m above natural ground) and total embankment
volume of 800 000 m3 .
• The main spillway, located in the left abutment, is made of a free over-
flow concrete gravity structure with an ogee crest and a total length
of 115 m.
• A service spillway, combined with the outlet work, located in the center right
of the valley, with a capacity of 70 m3 /s for the 10 years return period floods
management. They are installed in a tower, connected to a culvert ending with
a water energy dissipation structure downstream.
• The reservoir capacity at normal water operation level (454 asl) is 38 Mm3 .
The dam body is made of fine clayey lateritic soils in the upstream shell
(upstream of the filter) and gravelly lateritic soils downstream. Both materials were
obtained from borrow areas close to the dam site corresponding to silty clayey resid-
ual soil. The vertical chimney and a blanket filter-drain is made of crushed and
screened sound dolerite.
Figure 3 provides schematic typical cross sections of the dam and the foun-
dation at the abutments and the valley bottom. The top layer of the foundation is
made of “duricrust”, very hard cemented material with variable thickness generally
from 1 to 2.5 m. It overlies the lateritic carapace made of soft to firm clayey soil
of 4 to 9 m thickness with canaliculus and vugs, some of them reaching several
centimeters’ diameter and are of large extent. The bedrock is made of dolerite,
sound and watertight at depth, and weathered to highly weathered clay materials
on top (9 to 17 m). It is often not easy to differentiate between lateritic carapace
and highly weathered dolerite which is quite common in tropical residual soils’
profile.
At the valley bottom, where the ground is permanently saturated, duricrust

is missing and the lateritic carapace is thin and free of canaliculus. The upstream
shell is anchored, along the centerline, in the weathered bedrock at the valley bot-
tom and in the carapace at the abutments, where it is in contact with a horizon
containing canaliculus which cross the foundation from upstream to downstream.
During the dam construction, the foundation treatment by tubes-à-manchettes
grouting was limited to the valley bottom subject to the highest hydraulic gradi-
ent. During the excavation of the cut off trench on the left bank concentrations of
564
Q.105 - R.32
Fig. 3
Typical cross section of the dam and foundation
Profil type du barrage et de la fondation:
en rives (à droite)/fond de vallée (à gauche)
vugs and canaliculus were observed, and a very localized attempt of grouting was
undertaken.
3. INVESTIGATIONS CARRIED OUT
It was established since 2010 that the dam body was sound. The seepage was
mainly affecting the foundation in the abutments, particularly the lateritic carapace
and, to a lesser extent, the highly weathered bedrock. This assessment was based
on the site visual inspection and monitoring data analysis. The latter was related to
pore pressure cells installed in the embankment, relief wells at the downstream toe,
standpipe piezometers and seepage flow measurements.
Furthermore, the owner required maintaining the reservoir operation to remain

almost unchanged during repair works, as the dam is vital for irrigation and domestic
water supply of Banfora city. Emptying the reservoir to carry out, for example, a
positive cutoff at the upstream toe, as envisaged at the beginning, was therefore
rejected.
Other solutions had to be considered. They had to focus mainly on (1) the
treatment of the main flow paths from the dam crest to be in continuity of the cutoff
trench in the abutments and (2) the consolidation of the loose zones in the foundation
aiming at preventing any subsidence in the dam itself.
565
Q.105 - R.32
Whatever the final solution, a good knowledge of the conditions in the foun-
dation was necessary. For this purpose, the following investigations program was
conducted:
• Boreholes with partial or total recovery and drilling parameters monitoring

carried out from the crest and the downstream toe.
• CPTU soundings from the crest and downstream.
• Geophysical survey either by electrical method (performed by Fugro using
self-potential method) or magnetic field measurement (Willowstick patented
method).
• Laboratory testing.
• Injection of tracers (fluorescein, Rhodamine, brine), temperature measure-
ments at different depths in the reservoir and downstream, electric conductivity
measurements.
In order to minimize disturbance of the dam and prevent any adverse incident
while the reservoir remained in normal operation, boring trough embankment was
carried out using steel casing protection of Ø128mm, enabling the passage of a
double or triple tube core barrel for undisturbed sampling. Using air as the only
drilling medium was specified in order to prevent any hydro-fracturing particularly
in the embankment (ICOLD bulletin 158, §7.3-[1]). This requirement was respected
for core drilling but could not be adhered to when using a tri-blade (a tool developed
for destructive drilling), as it clogged. Only water under a gravity head prevented
clogging. It has been observed that the embankment became wetter with depth and
the foundation was waterlogged as the reservoir was never emptied.
Investigations were carried out in several stages with adjustments as these

progressed (Figure 4). At their completion, most boreholes were equipped with
standpipe piezometers. Almost two years were necessary to complete these inves-
tigations as all of them required mobilization of foreign companies. They were
completed in August 2013. Corresponding main findings may be summarized as
follows:
1. The embankment of the dam showed nothing unusual except at contact with
the foundation occasionally on the left bank. CPTU soundings, conducted
from the crest of the dam, showed a tip resistance systematically higher in the
dam body (around 5 MPa), compared to the foundation (1 to 2 MPa), except in
the valley center where no difference was noticed between embankment and
foundation.
The foundation is as lost residual tropical soils very heterogeneous. Four
facies could be distinguished as follows from top to bottom:
A. Silty clay, fine red to yellowish becoming sandy downward. A process of
laterization is underway at the top of this clayey material
B. Highly weathered clayey dolerite sandier at the base. This horizon is not
continuous.
566
Q.105 - R.32
Fig. 4
Longitudinal cross sections of the dam and foundation:
Geological model and Site investigations
Profil longitudinal du barrage et de sa fondation:
Modèle géologique et reconnaissances
C. Grayish dolerite sandy-gritty, fractured and permeable. This horizon is

also not continuous.
D. Dark, sound and watertight dolerite.
2. Both geophysical surveys did not lead to accurate location of the main seep-
age paths in the foundation. However, they provide their approximate position,
along with the presence of bypassing flow far in the left abutment. All signif-
icant seepage paths are located in the silty clay horizon in contact with the
dam. Willowstick method allowed deeper investigation and demonstrated the
absence of any significant seepage at the valley bottom.
3. On both abutments, some boreholes penetrated the foundation so quickly that
drilling advancement was considered as a tool drop. These very loose horizons
reached almost 11m height in the left bank. Additional investigations showed
that these singularities were of very small extension laterally, not exceeding
a few meters along the dam axis. Their extension in upstream-downstream
direction could not be determined.
4. The main geotechnical parameters of (disturbed) silty clay, representing the
horizon where the most significant seepage paths were located, are as follows:
A. Natural water content 35% and dry density 1.4 same range as during
the design. There is almost no variation of these parameters with depth.
B. Dmax = 2 mm, 0.076 mm content between 70 and 80%.
C. Plasticity Index and Liquid Limit average respectively 18.5% (16 to 24%)
and 53% (45 to 64%), corresponding to silty clay moderately plastic.
D. Undrained cohesion, (“indisturbed” samples) was evaluated between 20
to 60 KPa.
567
Q.105 - R.32
5. Permeability of the silty clay, according to CPTU and constant head tests
was estimated to 10-7 to 10-5 m/s depending on the in situ structure of the
materials (pumping tests couldn’t be carried out). It could be higher where this
cohesive material has been hydro-fractured, or within the lateritic carapace,
rich in canaliculus, vugs and cavities, far in the left abutment. Localized water
losses and communications between boreholes were observed. The presence
of thin horizons of silty sand to sandy gravel have been identified in the silty
clay and may participate in conveying water and fine particles (suffusion).
6. Subsidence area observed on the upstream slope of the dam around el. 451 is
located close to the vertical of the main seepage path highlighted by Willowstick
method on the left bank. This might be a sign of erosion that extends to the
upstream face of the dam.
4. REMEDIAL WORKS DESIGN
Remedial works focused on the following targets:
1. Stoppage of the erosion causing the downstream subsidence.

2. Elimination of any risk of subsidence affecting the embankment dam, where it
may be the most critical.
3. As a corollary, a significant reduction of leakage rates.
As it was not permitted to empty the reservoir or even to limit water storage
elevation less than 453 asl, i.e. 1 m below normal water storage (representing 15%
of the reservoir volume), remedial works could only be located on the crest and
downstream.
Under these conditions, the first solution considered was a positive diaphragm
wall located along the dam axis, keyed in the sound bedrock. It was to be carried out
from the crest of the dam and should have reached a maximum depth of 40 to 45 m.
This solution is indeed expensive, but highly effective in providing the necessary
water tightness. However, the presence in the foundation of very soft and loose
material or cavities along with free water passages highlighted by communications
between boreholes and by water losses, led the designer to discard this solution
as it presented the risk of a sudden drawdown in the wall trench, during either
excavation or concreting. This situation was not acceptable as the trench might
have collapsed causing unpredictable damage, including large flow release as the
reservoir remained operational.
Grouting, using Tubes à Manchettes (TAM) is considered unsuitable for silty

clay material where it is likely to generate useless and potentially dangerous
hydraulic fracturing mainly sub-vertical in upstream-downstream direction. Further-
more, at the exit of the manchette, it is not possible to monitor the grouting pressure
applied at the ground contact. However, as the silty clay is cohesive and contains
568
Q.105 - R.32
void like passages, TAM grouting can be somehow relevant for foundation treatment
aiming at filling these voids.
The solution adopted finally was as follows:
• A positive cutoff made of sheet piles (SP) located along the dam axis and
reaching a maximum depth of 40m.
• Grouting using TAM immediately upstream of the SP to (1) plug any defect in
the SP keys and (2) complete foundation treatment in depth beyond the limit
of the SP.
The main potential issue of this solution was the difficulty to drive the SP to the
required depth, due to the eventual presence of coarse elements in the foundation.
Furthermore, it was not possible to carry out any full-scale trial before corresponding
call for bids, as neither sheet piles nor driving equipment were available in Burkina
Faso or even in neighboring countries.
The reassessment of investigations’ findings, boreholes and CPTU, confirmed

the absence of coarse elements likely to stop SP driving. This was discussed with
SP manufacturers who confirmed the low risk of driving problems and recommended
using high inertia profiles, i.e. AZ-36-700N made by Arcelor Mittal. The surface in the
foundation to be covered by SP wall was fixed in connection with the main seepage
paths identified during investigations.
Where the required depth could not be reached and outside the limits of the SP
wall, TAM grouting remains the only available alternative. This grouting is therefore
extended vertically to the sound bedrock in continuation of the SP wall and laterally
outside its limits.
Regarding the risk of subsidence of the dam, due to the development of erosion
cavities and loose soils under the duricrust, leading to pseudo-karstic behavior of
the foundations, the compaction grouting was considered the most suitable. This
treatment was limited to the clayey silt horizon considered as the weakest in the
foundation. It would have been interesting to perform some in situ testing before
launching the corresponding tender, but this was not possible. The design was
therefore set on the most realistic assumptions regarding mortar consumption and
holes’ spacing. Call for bids’ documents offered, however, flexibility in the design
adaptation during the course of actual works.
Figure 5 represents the position of each component of the remedial works.

The SP cutoff placed on the dam axis along with the TAM grouting holes, located
upstream, were far enough from the chimney filter to have no impact on its
integrity. Compaction grouting holes may cross the filter but under systematic casing
protection. Each hole was filled under very low pressure during the casing removal.
569
Q.105 - R.32
Fig. 5
Remedial Works Definition
Définition des travaux de réhabilitation
The durability of the SP is considered satisfactory as they are totally buried,

having no contact with air. For its part, compaction grouting is performed after
completion of SP driving and TAM grouting, to minimize mortar leaching.
5. WORKS CARRIED OUT
Remedial works consisted of three main activities conducted in the following

order:
• Driving SP through the embankment and the foundation to a maximum depth

of 40 m.
• TAM grouting to seal SP joints and, as much as possible, improve the water
tightness of the foundation at depth and laterally beyond SP wall limits.
• Densifying the foundation by compaction grouting targeting the very loose silty
clay horizons in the dam foundation.
No grouting was carried out in the body of the dam, except sleeve grout of
TAM and filling under minimum possible pressure the upper part of the holes used
for compaction grouting. Figure 6 provides the as-build configuration of the SP wall
and TAM grouting on the longitudinal cross section.
SP were paired by crimping in the factory (Photo 2) to provide high buckling

resistance. Installation was conducted using heavy impact hammers. Vibratory ham-
mers were abandoned because they could not drive SP through the embankment,
confirming its good compactness.
570
Q.105 - R.32
Fig. 6
Longitudinal section, presenting the as build SP, TAM grouting and
compaction grouting on both abutments
Profil longitudinal présentant le récolement des travaux de réhabilitation en
palplanches, injection en tubes à manchettes et injection de compactage.
Photo 2
Paired sheet piles delivered on the jobsite (left),
driving on the dam crest (right)
Palplanches appairées approvisionnée (à gauche) et rassemblées pour
fonçage à partir de la crête du barrage (à droite)
The total surface of SP wall installed is almost 12,000 m2 , 7,650 of which in the
left abutment reaching a maximum depth of 36 m and 4,350 m2 in the right abutment,
reaching 34m maximum depth.
The driving stop criterion of the piles was 10 blows of the hammer for a max-
imum penetration of 2 cm, i.e. 500 blows per meter. Actual refusal systematically
extended to 800 blows per meter, where the design depth was not reached. The
number of blows was accurately monitored and used for characterization of the
foundation.
571
Q.105 - R.32
Fig. 7
Sheet piles-Characterization of the foundation using monitored number of blows
Palplanches-Caractérisation de la fondation en utilisant l’enregistrement du
nombre des coups/m d’enfoncement
Table 1
Requirements for sleeve and injection grout used with TAM
Exigences pour le coulis d’injection et le coulis de
gaine utilisé en injection en tubes à manchettes
TYPE OF GROUT BLEEDING AT 2 HOURS VISCOSITY MARSH (S) UCS AT 28 DAYS
Injection <3% <40 s >10 MPa
Sleeve filling <5% <40 s 0.5 to 1 MPa
During SP driving, particle velocity was monitored to check the incidence of

impacts’ vibration on the dam and appurtenant structures. No problem was recorded,
except that a few localized sinkholes appeared due to the shearing of the embank-
ment material during several SP drives. These zones were excavated and filled with
compacted lateritic clayey material.
The TAM grouting holes (Ø 101 mm) were located 0.55 m upstream of the
dam axis on the crest as shown in figure 5. They are enough close to the SP wall
to grout corresponding key joints where needed. They included primary, secondary
and tertiary holes. Primaries were spaced 5.60 m where SP are installed and 8m
elsewhere. Thus, in the former case for the tertiaries, the final spacing is 1.40 m
corresponding to the distance between each pair of sheet piles. Out of SP walls,
the final spacing for the tertiaries is of 2 m. No grouting was carried out in the dam
embankment.
Two types of grout were specified, (1) sleeve filling and (2) foundation injection.
Table 1 provides the main requirements and Table 2 corresponding proportioning.
572
Q.105 - R.32
Table 2
Sleeve and injection grouting proportioning
Formulations des coulis d’injection et du coulis de gaine
TYPE OF GROUT W/C BENTONITE/CEMENT (%) ADMIXTURE/CEMENT (%)
Injection 0.75 0.7 0.5 to 0.9
Sleeve filling 3 10 0.5
Table 3
Pmax specified at the top of the hole
Pmax spécifiée en tête de forage
Depth (m) 0 to 4 4 to 7 7 to 10 10 to 13 13 to 16 16 to 20 20 to 25 25 to 45
Pmax at the top (MPa) 0 0.4 0.6 0.8 1 1.2 1.5 2
TAM grouting was conducted according to GIN method (Lombardi & al 1993-
[2]) in order to reduce the risk of grout wastage due to hydro-fracturing. The GIN
number was low, limited to 100 MPa.liter/m, with a Vmax of 300 l/m and Pmax as
fixed in Table 3.
The main conclusions derived from the TAM grouting were as follows:
• GIN method, even considered unsuitable for this kind of soil, turned out to be
somehow adapted to the silty clay grouting.
• Regarding the very low hydro-fracturing occurrence in the left bank grouted
first, the GIN number was increased in the right abutment TAM grouting to
300 MPa.l/m and Vmax to 1000 l/m. These parameters turned out to be well
adapted to this foundation. A local retreatment of the other abutment using
these parameters was then carried out.
• There was no reduction in average grout consumption from primaries to ter-
tiaries. However, at the tertiaries, almost no hydro-fracturing was observed,
and no grouting was stopped at Vmax.
• Sleeve grout consumption was much higher near the SP wall due to the
disturbance of the ground during SP driving.
• No significant resurgence of grout was observed, except a single resurgence
at the crest where the grout climbed up along a sheet pile.
• The total absorption of dry matter (cement and bentonite) reached 1,540 tons,
spread on 23,300 m2 of the vertical treated surface (66 kg/m2 ).
• The large number of injection holes drilled was also considered as a system-
atic direct investigation of the foundation, providing a very precise location of
areas of weakness (colored in red in figure 8). They are in continuity with the
downstream resurge areas on both sides.
573
Q.105 - R.32
Fig. 8
TAM grouting – Imaging of sleeve filling and injection consumption grouting
Injection en tubes à manchettes – Représentation en image de la
consommation en coulis de gaine et coulis d’injection
Table 4
Compaction grouting mortar proportioning
Formulation du mortier d’injection de compactage
SAND (KG) OPC CEMENT (KG) WATER (KG) ADMIXTURE (HRWR) (KG)
1582 300 205 3
Compaction grouting was carried out where SP and TAM works were com-
pleted. It was conducted according to ASCE “Compaction Grouting, Consensus
Guide (2010)-[3]”.
Requirements related to the mortar components were fixed for sand (1) dmax
of 20 mm, (2) 10% to 20% of minus 200 and (3) fines plasticity index smaller than
10%. The slump required was 50 mm at the injection pump delivery and UCS at 28
days greater than 10 MPa.
The final proportioning of the mortar mix fulfilling these requirements is given
in Table 4
Sand produced from natural alluvium was scalped to 12.5 mm. It contained
90% minus 5 mm and 16% of non-plastic minus 200. The 28 days UCS exceeded
13 MPa. Due to the high ambient temperature at the site, the slump at the batching
plant was fixed at 80 mm so as to respect the 60 to 50 mm during injection.
574
Q.105 - R.32
Photo 3
Appearance of the mortar
Apparence du Mortier
A regular piston concrete pump was used for the injection. It is not designed
to control grouting pressure accurately. Grouting pressure was fixed to a maximum
of 6 MPa in the beginning and reduced to 4 MPa, monitored by a manometer at the
exit of the pump. Unlike TAM there was no possibility to get automatic recording and
monitoring of injection parameters. Corresponding results may be summarized as
follows:
• Compaction grouting highlighted the presence of very loose areas (muddy

zones), so they may be considered as voids. It is called cavernous soil.
Indeed, grout takes exceeding 5 m3 /m at low pressures are numerous. Near
the above-mentioned main tool drop area, 60 m3 of mortar was injected. Cor-
ing conducted near big takes crossed almost intact mortar sometimes for a
few meters. There was no permeation of mortar away from the injection point.
• The final average mortar consumption reached 2.2 m3 /m (2.43 m3 /m in the left
abutment and 1.88 m3 /m in the right abutment). These large takes confirm the
relevance of this treatment.
• Due to the large takes observed at each 0.5 to 1 m grouting stage, jamming
of casing tubes occurred very often. As it had a significant negative impact
on the progress of work, it was decided to reduce the casing diameter from
100 mm to 70 mm, to install hydraulic jacks to pull out the tube and to move
it slightly every 15 to 30 minutes. Depth of grouting, reaching a maximum of
30 m, had minor impact on casings’ jamming.
6. DAM SURVEILLANCE
Dam surveillance through regular visual inspections and monitoring data-

gathering and interpretation was set since completion of the dam. It helped in
575
Q.105 - R.32
detecting early signs of seepage problems and led to carrying out remedial works
described in this paper.
During these remedial works, the dam surveillance team permanently mon-
itored seepage flow at several strategic points and hydrostatic pressure along the
dam axis and downstream. The team also inspected the dam and adjoining areas,
during working hours, looking for any eventual suspect new element such as crack-
ing, subsidence, heave, or grout/mortar resurgence. Furthermore, outlet gates were
checked to ensure they could be operated in case the reservoir level should be
lowered.
During remedial works, no damage or any suspect events occurred either on

the dam or the foundation. Seepage flow never showed any dirty water or cement
particles content. The following harmless observations were reported:
• During compaction grouting, clean water came out momentarily from some
piezometers, confirming the densification action of these works.
• Resurgence of mortar from insufficiently plugged holes during grouting of an
adjacent hole.
7. WORKS EFFICIENCY
The total volume of works carried out reaches 12 000 m2 sheet piles, 40 km
drilling holes and 38 000 tons dry materials (sand, cement & bentonite). Since the
beginning of the works, a significant decrease in measured seepage flow and a
reduction of the hydrostatic pressures downstream of the dam was established.
Large wet and seepage resurging zones became more and more dry as work
progressed.
During wet periods in 2018 and 2019 when the dam spilled more than one
month:
• Maximum total seepage flow reached 5 liters/s, instead of 40 to 60 l/s before

remedial works (see figure 9 below).
• Hydrostatic pressures downstream in relief wells decreased significantly
particularly in the right abutment (see figure 10 below).
8. CONCLUSION/LESSONS LEARNED
1. The remedial works carried out confirmed that the seepage through the foun-
dation of Comoé dam have weakened it considerably and that it was indeed
urgent to carry out these works to preserve to safety of the dam.
576
Q.105 - R.32
Fig. 9
Total downstream seepage flow
Débit total des fuites l’aval du barrage
Fig. 10
Hydrostatic pressures in downstream relief wells
Pression hydrostatique dans les puits de décharge aval
2. Although it is not common, the combination of the 3 foundation treatment

technics (sheet pilling, tube-à-manchette and compaction grouting) was the
most appropriate as it was not possible to empty the reservoir.
3. Tropical soils, such as laterites present in the abutments of the Comoé dam,
cannot generally provide a satisfactory watertightness as a dam foundation,
577
Q.105 - R.32
especially if they contain canaliculi, which are not intersected by a positive

cutoff.
4. Investigations by core drilling often prove insufficient for a proper assessment
of their permeability due to the smoothing of the borehole walls during the
perforation process. Canaliculi are often difficult to detect in cores, hence
the importance of large reconnaissance pits and trenches, with water testing
by filling small pits excavated in this foundation. Pumping tests may provide
valuable information, but only under the water table.
5. Efficient drainage including relief wells is necessary in soft or highly altered
soils over their entire thickness, to prevent erosion. In Comoé, relief wells
were note enough deep and enough extended far un the abutments to prevent
erosion.
6. The example of the Comoé dam confirms, if needed, the aptitude of laterites
to constitute a good quality watertight fill. Well humidified and compacted this
material exhibits a good resistance to erosion.
REFERENCES
[1] ICOLD (International Commission on Large Dams), Dam Surveillance Guide,

Bulletin 158, 2012, §7.3.
[2] Lombardi, G. and Deere, D. (1993). “Grouting Design and Using the GIN
Principle.” Water Power & Dam Construction, issue June 1993, pp. 15–22.
[3] ASCE (American Society of Civil Engineers). (2010) Compaction Grouting,

Consensus Guide, Standard ASCE/G-I53-10. ASCE, Reston, VA.
578
Q.105 - R.33
-------
-------
LESSONS LEARNT FROM THE FAILURE CASES ON

SEEPAGE CONTROL OF CFRD (∗)
XU ZEPING
Professor, CHINA INSTITUTE OF WATER RESOURCES AND
HYDROPOWER RESEARCH
CHINA
SUMMARY
Seepage control is an important issue for the normal operation and safety of
concrete faced rockfill dams (CFRD). By presenting the failure case of Baiyun CFRD
in China and several other failure cases on seepage control of CFRD projects, the
failure mechanism is analyzed and the lessons to be learnt are summarized. Based
on the analysis, principles of the seepage control system of modern CFRD is sug-
gested, which focuses on comprehensively exerting the function of each material
zone, and to form a comprehensive seepage control system with mutual protec-
tion through the cooperation of the permeability coefficient of each zone and the
gradation of key particle groups.
RÉSUMÉ
Le contrôle des infiltrations est une question importante pour le fonctionnement

normal et la sécurité des barrages en enrochement à masque amont en béton
(CFRD). En présentant le cas d’échec de Baiyun en Chine et plusieurs autres cas
∗ Leçons tires de cas de défaillance des systèmes de contrôle des infiltrations des barrages
en remblai à masque amont en béton
DOI: 10.1201/9781003211808-79 579

Q.105 - R.33
d’échec sur le contrôle des infiltrations des projets de ce type, le mécanisme de

défaillance est analysé et les leçons à tirer sont résumées. Sur la base de l’analyse,
les principes du contrôle des infiltrations des barrages à masque amont sont sug-
gérés. Le principe proposé se concentre sur l’exercice de la fonction de chaque
zone d’enrochement pour former un système complet de contrôle des infiltrations.
Il met l’accent sur l’action de protection mutuelle à travers le contrôle du coefficient
de perméabilité et la granulométrie de chaque zone d’enrochement.
1. INTRODUCTION
Concrete faced rockfill dam (CFRD) is a type of dam with rockfill for the main
body and concrete face slabs on the upstream surface as a watertight structure.
Today, CFRD can be regarded as one of the main types of rockfill dam. Their inherent
safety, adaptability and economy features make them very competitive and popular
in dam engineering.
With rapid progress on technologies of CFRD construction, more and more

high concrete faced rockfill dams have been constructed in the world. From 1990s to
the beginning of 21 century, numerous high CFRD with the height of more than 150 m
were constructed, include Bakun (205 m) in Malaysia, El Cajon (189 m) in Mex-
ico, Campos Novos (202 m) in Brazil, Karahnjukar (196 m) in Iceland, Nam Ngum
II (180 m) in Lao PDR, TSQ-1 (178 m), Hongjiadu (179.5 m), Sanbanxi (185.5 m),
Shuibuya (233 m) and Houziyan (223 m) in China.
The main issues in the design and construction of CFRD are stability of rockfill
dam body and safety of seepage control system. Due to the relatively high shear
strength of compacted rockfill, the stability of dam body usually is not an important
issue. Watertight system of CFRD is mainly composed with concrete face slabs and
joint waterstop system. Generally, the deformation of rockfill is the main cause for
the damage of concrete face slabs and joint waterstop system. Therefore, one of the
major tasks of CFRD design is to control the deformation of rockfill and to guarantee
the safety of face slabs and joint waterstop system. On the other hand, besides
the watertight system, rockfill in different zones are also have the functions in the
seepage control system of CFRD, with the semi-permeable cushion zone as the
second defense line behind concrete face slabs and the transition zone for drainage
and pressure relief.
For CFRD construction in recent years, most projects are successful in their
seepage control system. But there are also some failure cases, which caused large
amount water leakage. By presenting the failure case of Baiyun CFRD in China
and several other failure cases on seepage control of CFRD projects, the failure
mechanism is discussed and lessons to be learnt are summarized in this paper.
Based on the analysis, the general principles on seepage control system of modern
CFRD is suggested.
580
Q.105 - R.33
2. FAILURE CASES ON SEEPAGE CONTROL
In recent year, there are several CFRD projects have suffered large amount
leakage due to failure in seepage control system. Some of the cases are described
below.
2.1. BAIYUN CFRD, CHINA
Baiyun CFRD is located in Hunan Province, China. The main purpose of the
project is power generation and water supply. The total storage capacity of reservoir
is 3.6 × 108 m3 and the installed capacity of power station is 54MW. CFRD dam
height is 120 m and the crest length is 198.8 m. The rockfill for dam construction
is limestone. The project was completed in 2003. Fig. 1 shows the typical section
zoning of the dam.
Fig. 1
Typical section of Baiyun CFRD
Coupe typique de Baiyun CFRD.
After the first impounding of reservoir at the end of 1998, water leakage mea-
sured from downstream wire was basically normal in 10 years operation of the
project. From May of 2008, water leakage through the dam gradually increased.
On August of 2008, the measured leakage reached 200L/s. After that, the amount
reached 512L/s on October of 2019, 800L/s on October of 2010 and 1240L/s on
September of 2012.
Due to the large amount leakage of the dam, the reservoir water level was
limited around dead level (El. 505 m) and the leakage kept around 660L/s. From
underwater inspection, the main leakage inlet is located at lower part of the slabs
581
Q.105 - R.33
Fig. 2
Measured leakage from downstream wire
Fuite mesurée du fil en aval.
near left abutment. By emptying the reservoir and clearing upstream soil blanket,
the revealed damage of face slabs proved the results of underwater inspection.
In 2014, reservoir was emptied, and upstream soil cover was excavated. It
was found that there is a subsidence area in face slab L5 and L6 at EL. 490 m with
the area of 50 m2 and maximum depth of 76 cm (as shown in Figure 3). A larger
subsidence was found from face slab L4 to L7 below EL. 473 m with the area of
250 m2 and maximum depth of 245 cm (as shown in Figure 4). Concrete face slabs
were broken in the subsidence area, which formed the main inlet of water leakage.
Fig. 3 Fig. 4
Damage of slabs Damage of slabs
Dommages aux dalles. Dommages aux dalles.
2.2. ZHUSHUQIAO CFRD, CHINA
Zhushuqiao CFRD is located in Hunan Province, China. The main pur-

pose of the project is power generation. The total storage capacity of reservoir
582
Q.105 - R.33
is 2.78 × 108 m3 and the installed capacity of power station is 24MW. CFRD dam
height is 78 m with the elevation of dam crest of 171 m, and the crest length is
245 m. The rockfill for dam construction are limestone and siliceous slate. The dam
was completed in 1992. It is one of the earliest CFRD built in China.
Soon after being put into operation, the dam experienced serious water leak-
age, and the amount of leakage increased significantly year by year. By July 1999,
water leakage exceeded 2500L/s, which caused great concern of dam industry in
China. Table 1 shows the changes of dam leakage.
Table 1
Leakage of Zhushuqiao CFRD
Time (year) 1992.7 1994.8 1995.6 1996.9 1997.6 1998.7 1999.7
Reservoir level (m) 161.90 163.70 163.38 163.75 163.67 163.67 163.51
Leakage(L/s) 30 ∼ 40 970 1296 1344 1520 1684 2500
From the inspection after reservoir empty, it was found that some slabs were
collapsed or fractured. After excavation, further observation revealed that there was
a void between the damaged slabs and the cushion layer. All the places where
concrete face slabs were broken and collapsed were observed separation between
face slabs and cushion rockfill. Face slab L8 on the right side of river valley has the
most severe separation from the cushion rockfill at a height of 61 m from dam crest.
2.3. CAMPOS NOVOS CFRD, BRAZIL
Campos Novos CFRD is located in Santa Catarina, Brazil. The hydropower

station has installed capacity of 879 MW. The dam height is 202 m with the elevation
of dam crest of 665.8 m, and the crest length is 590 m. The valley factor (A/H2 ) is
2.59. Rockfill for dam construction is basalt. Both the upstream and downstream
slope of the dam is 1:1.3.
After the completion of dam construction, during the first reservoir impounding
in October 2005, when reservoir water level reached EL. 607 m, about 92% of normal
storage level, face slab rupture was discovered at the vertical joint between slab 16
and 17. The position of rupture located at the top part of face slab in the center of
riverbed section. The range of concrete rupture developed from EL. 660 m (bottom
of parapet wall) to EL. 535 m (underwater). Besides, At EL. 656 m (position of the
construction joint between second stage face slab and third stage face slab), a
horizontal crack was observed. After face slab rupture, water leakage through dam
rapidly increased to the value of 1400L/s. From site inspection, separation between
concreate face slabs and cushion rockfill was observed.
583
Q.105 - R.33
3. ANALYSIS OF FAILURE MECHANISM
For concrete faced rockfill dam, the performance of watertight system (con-
crete face slabs and joint waterstop) is mainly rely on deformation of rockfill
embankment. Large deformation of rockfill or differential deformation between rock-
fill and abutments will lead to damage of concrete face slabs and waterstop system.
When watertight system is damaged, reservoir water will enter into rockfill embank-
ment. If the cushion zone is not well protected, small particles of cushion zone will
be washed away. This process could cause void under face slabs or separation
between face slabs and rockfill, which may even worsen the operation status of
concrete face slabs.
For the case of Baiyun CFRD, the dam is located in a narrow valley (the
ratio of crest length to dam height of 1.67) with very steep abutment (especially
for the left abutment). Unfavorable conditions of river valley lead to large differential
displacement between abutments and dam body, which caused cracks of face slab
and damage of waterstop. From the results of site inspection, the main leakage
occurred on the location of left bank, indicating that the relative deformation between
dam and abutment were relatively large, which deteriorated the operation conditions
of face slabs and waterstop at this area.
According to the measured data on the construction materials of Baiyuan

CFRD, the content of particles smaller than 5mm of the material of cushion zone
is 22–52%, which does not meet the normal requirements for cushion material of
CFRD. In the checked samples for cushion material below EL. 475 m, more than
60% of the samples have the content for particle size smaller than 5mm less than
35%. In the filling construction of the dam, as the filling sequence was fine mate-
rials first and then coarse materials, the coarse particles at the interface between
transition zone and cushion zone cannot be removed. Therefore, the material in
transition zone cannot provide effective filter protection for the material of cushion
zone. Figure 5 shows the measure gradation of the material of cushion zone.
Fig. 5
Gradation of the of the cushion material of Baiyun CFRD
Gradation du matériau du coussin de Baiyun CFRD.
584
Q.105 - R.33
For the case of Zhushuqiao CFRD, the problems are also started from the
damage of face slabs and waterstop (watertight system), and then developed with
particles washed out of cushion layer. Finally, as the cushion zone cannot provide
a solid support for face slabs, concrete face slabs were broken, and it caused large
amount water leakage through the dam. Figure 6 is the gradation curves of the 4
samples taken from the cushion layer near face slabs L8 and L9 (left abutment) at
El. 110 m and El. 108 m (8–10 m above bedrock).
Fig. 6
Gradation curves of the cushion material of Zhushuqiao CFRD
Courbes de gradation du matériau du coussin de Zhushuqiao CFRD.
With the analysis of the gradation curves in Figure 6, the content of particles
with the size less than 5mm varies from 10% to 37%, the average value is only
25%, and the permeability coefficient of the material varies from 2.9 × 10−1 cm/s to
3.5 × 10−3 cm/s, which means that the rockfill of cushion zone has less fine particles.
Therefore, the cushion layer cannot take the function of the second defense line for
seepage control.
Figure 7 is the gradation curves of the rockfilll of cushion zone and transition
zone of Campos Novos CFRD. It is noticed that the content of particles with the
size less than 5 mm of the cushion material varies from 10% to 37%, D20 varies
from 0.75 mm to 10 mm. Obviously, the permeability of cushion zone is rather large
and the transition zone may not provide good protection of the fine particles of the
cushion zone. Therefore, the cushion zone is also impossible to play the role of the
second defense line for seepage control. This will inevitably lead to the large amount
of water leakage when watertight system was damaged.
585
Q.105 - R.33
Fig. 7
Gradation curves of the rockfill of cushion and transition zone
(Campos Novos CFRD)
Courbes de gradation de l’enrochement du coussin et de la zone de transition
(Campos Novos CFRD).
4. SUGGESTED PRINCIPLE OF SEEPAGE

CONTROL OF CFRD
As mentioned above, concrete slab and joint waterstop system are the main
body of the seepage control system of CFRD. But at the same time, the design of
rockfill zoning, especially the design of the cushion zone and transition zone, will
also play a very critical role to guarantee seepage safety of the dam.
Historically, the concept of seepage control of CFRD has gone through two
stages of development. In the first stage, it only emphasized the watertight function of
concrete face slabs. The major role of the cushion zone was designed as deformation
coordination and stress transition. Rockfill gradation of the cushion zone is relatively
coarse (with less fine grains) and the permeability coefficient of rockfill in cushion
zone is relatively large. Once face slabs or joint waterstop are damaged, serious
water leakage will be inevitable. In the 1980s, Sherard proposed the concept of
using cushion zone as the second defense line for seepage control of CFRD. The
permeability coefficient of cushion material should be controlled to the quantity of
10−3 cm/s to 10−4 cm/s. The proposal of Sherard is a major breakthrough in CFRD
design. It further improves the seepage control system of CFRD and the safety of
the dam. Therefore, it has been recognized by most engineers in the world and
widely applied in engineering practice. It marks the development of seepage control
technology for CFRD has entered the second stage.
In recent years, as the height of CFRD increases, the requirements for reli-
ability of seepage control system of CFRD and the safety for seepage stability of
586
Q.105 - R.33
rockfill have also increased. To addressing this end, the author further proposed a
concept for the seepage control system of CFRD, which is based on summary of
previous research results and analysis of related engineering practices. Key points
of the concept are: the design of seepage control system of CFRD should com-
prehensively consider the coordinated protection of each material zone, including
upstream soil covering (1A, 1B), concrete face slab, cushion layer (2A) and transition
zone (3A). The overall reliability of the whole system should be guaranteed by joint
operation and mutual protection of each rockfill zone, as well as the consideration of
the self-healing function of the seepage control system in case of watertight defense
damage.
In general, the seepage control system of CFRD mainly include the watertight
system (concrete slabs and joint waterstop) as the first defense line, the semi-
permeable cushion layer with the protection of transition zone as the second defense
line, and the main rockfill zone with strong water permeability and erosion resistance
as the drainage and pressure relief protection. From the perspective of seepage sta-
bility, the main rockfills zone (3B, 3C) have strong water permeability. Compared with
the permeability of cushion material, the difference between the two rockfill zone is
at least a hundred times. Once face slabs or joint waterstop loses its watertight
ability, cushion layer will become an inclined impermeable body, and most of the
seepage head will be borne by cushion layer. The sufficient drainage capacity of the
main rockfill zone will rapidly drop the seepage situation line, thereby to guarantee
the safety of the dam.
Engineering practices show that Sherard’s suggestion of using cushion layer

as the second line for seepage control is correct in concept. But there are still some
problems in implementation. The main issue is how should the permeability coeffi-
cient of cushion material be selected to ensure that it not only serves as the second
defense line for seepage control, but also can ensure that the selected material
has higher shear strength and lower compressibility. The smaller the permeability
coefficient, the better the seepage control performance, But the problem is that the
selected material contains too much particles with the size less than 0.1 mm, and
the shear strength and compression modulus are reduced accordingly, which is
not conducive to the stability of dam slope. In addition, according to the laboratory
test, the permeability coefficient of the rockfill with the gradation curve suggested
by Sherard cannot reach the requirement of 1 × 10−4 cm/s, especially the content
of the particles with the size less than 0.075 mm of the outer envelope curve is only
8%. This material can only reach the order of 10−2 cm/s even in the tightest state.
It is generally believed that permeability of rockfills are directly related to the

content of particles smaller than 5mm. But there are relatively few further studies
on the particle gradation of the part smaller than 5 mm. From laboratory test, it is
found that as the content of particles smaller than 5 mm increases, the range of fine
particles composition also increase, which causes great changes for the permeability
coefficient of each samples. When the content of particles smaller than 5 mm is
55%, the permeability coefficients for rockfill with different composition of particles
less than 5 mm may be different by 2 orders of magnitude. Therefore, only define
587
Q.105 - R.33
the content of particles with the size less than 5 mm cannot accurately reflect the
permeability coefficient of rockfill.
The permeability of cushion material mainly depends on the content of fine

materials, especially the particle size of d20 . When the content of particles with the
size less than 1 mm is below 20%, the permeability coefficient of the rockfill will
be greater than 10−3 cm/s. Therefore, for rockfill of cushion zone, the content of
particles with the size less than 1 mm should be greater than the recommended
value of Sherard (12%). The recommended gradation design for rockfill of cushion
zone shou be 35% to 55% of particles with the size less than 5 mm and 20% to
32% of particles with the size less than 1 mm. By optimizing the content of particles
smaller than 1 mm, the permeability coefficient of cushion material can reach the
requirement of 10−3 cm/s to 10−4 cm/s.
Regarding the self-healing ability of the cushion material, as the recommended

covering material above face slabs or the dumped soil used to repair concrete face
slabs damage is sandy soil, normally with d85 = 0.2 mm, d15 of rockfill in cush-
ion zone should be less than 0.8 mm, according to the Terzaghi filter criterion
(D15 ≤ 4d85 ).
5. CONCLUSIONS
Seepage control of concrete face rockfill dam is an important aspect related to

the overall safety of the dam, and it is also an important issue in dam design. As the
deformation of rockfill is the main cause for the damage of concrete face slabs and
joint waterstop, the major tasks of CFRD design is control the deformation of rockfill
to guarantee the safety of the watertight defense line. In addition, different rockfill
zones shall also be considered as the part of seepage control system of CFRD.
From the lessons learnt from the failure cases on seepage control of recent
CFRD projects, the new concept proposed on design of seepage control system of
CFRD will focuses on comprehensively exerting function of each material zone, and
to form a comprehensive system with mutual protection through the cooperation of
the permeability coefficient of each zone and the gradation of key particle groups.
The system includes the covering sandy soil for healing the possible damage of face
slabs and joint waterstop, the semi-permeable cushion zone as the second defence
behind concrete face slabs. In order to ensure the stable permeability of the cushion
zone, the transition material should have the function of filter, and the transition zone
should be designed according to the principle of filtering layer of the cushion zone.
REFERENCES
[1] Paulo T. Cruz, Manoel S. Freitas, “Cracks and Flows in Concrete Face Rockfill
Dams”, Proceedings of International Symposium on Dam Engineering, 2007,
Lisbon, Portugal.
588
Q.105 - R.33
[2] B. Materon, State of Art of Compacted Concrete Face Rockfill Dams,

Workshop on high dam know-how, 2005.8, Yichang, China.
[3] J.L. The upstream zone in concrete face rockfill dams, in “CFRD, Design,
Construction and Performance”, ASCE, 1985.
[4] Amaya, F. and Marulanda, A. Golillas dam-design, construction and perfor-

mance, in “CFRD, Design, Construction and Performance”, ASCE, 1985.
[5] J.M. Sierra, C.A. Ramirez, J,E.Hacelas. Design features of Salvajina dam, in
“CFRD, Design, Construction and Performance”, ASCE, 1985.
589
Q.105 - R.34
-------
-------
A STUDY ON THE COMMUNICATION HISTORY OF

THE THREE GORGES PROJECT (∗)
Yang JUN
Deputy Chief Economist Of CTG
Yan YI
Staff Of CTG,
CHINA
SUMMARY
The Three Gorges Project (“TGP”), the largest comprehensive water con-
servancy complex project in today’s world, shows different characteristics in its
communication in different stages from the proposal to the argumentation, design,
approval, construction and operation. From the perspective of project communica-
tion, the article uses historical retrospect to summarize the communication practice
of TGP in different stages and with different characteristics during its lifecycle, and
analyze the communication laws of TGP. According to the article, the project com-
munication allows information transfer, value interpretation and image building in
a complicated social context, and helps the public understand the project objec-
tively, deeply and vividly, thus boosting the decision making, leading the anticipation,
explaining the doubts, feeding back information, resolving disputes, and escorting
the project, and facilitating the smooth achievement of the project goals in different
stages.
TGP is the largest comprehensive water conservancy complex project in the

world today. From its construction which lasted more than a decade to its successful
completion and giving play to huge comprehensive benefits, TGP is of great signifi-
cance in the modernization process of China, and has produced an extensive social
∗ Une étude sur l’histoire de la communication relative au projet du barrage des Trois Gorges
590 DOI: 10.1201/9781003211808-80

Q.105 - R.34
effect across the world. A study on the communication history of TGP allows us to
analyze the communication laws of national major projects, enhance our awareness
of disseminating major projects, and provide reference for more major projects to
be better disseminated and play a better role.
RÉSUMÉ
Le projet des Trois Gorges (« TGP »), le plus grand projet complexe complet de
retenue dans le monde d’aujourd’hui, montre différentes caractéristiques de la com-
munication à différentes étapes de la proposition à l’argumentation, la conception,
l’approbation, la construction et l’exploitation. Du point de vue de la communication
de projet, le rapport présente cette communication à différentes étapes et avec des
caractéristiques différentes au cours de son cycle de vie, et analyse les lois de com-
munication autour de ce projet. La communication permet le transfert d’informations,
l’interprétation des valeurs et la construction d’images dans un contexte social com-
pliqué, et elle aide le public à comprendre le projet objectivement, profondément
et positivement, augmentant ainsi la prise de décision, menant l’anticipation, expli-
quant les doutes, alimentant l’information, résolvant les différends, et accompagnant
le projet, en facilitant la réalisation en douceur des objectifs du projet à différentes
étapes.
De sa construction qui a duré plus d’une décennie à son achèvement réussi et

en apportant d’énormes avantages complets, le projet des Trois Gorges est d’une
grande importance dans le processus de modernisation de la Chine. Il a produit
un effet social étendu à travers le monde. Une étude sur l’histoire de la com-
munication de ce projet permet d’analyser les lois de communication des grands
projets nationaux, de mieux nous sensibiliser à la diffusion des grands projets et
de référencer pour que d’autres grands projets soient mieux diffusés et jouent un
meilleur rôle.
1. THE COMMUNICATION OF MR. SUN YAT-SEN’S

“THREE GORGES DREAM”
Sun Yat-sen’s envisagement about the TGP was first recorded in The Interna-
tional Development of China written by Sun Yat-sen in English in 1918, in which The
Second Plan includes the development of hydropower resources in Three Gorges.†
From March 1919, The International Development of China was successively pub-
lished on the English newspaper of US in China, The Far Eastern Review. In August
1919, was translated into Chinese, and successively published on the Construction
†Zhang Yunchang and Luan Linbin (chief editors): A Compilation of Historical Literatures for
China Three Gorges Project (1948–1949), China Three Gorges Publishing House, Edition
2010, p19.
591
Q.105 - R.34
journal, an institutional publication of Kuo Min Tang (“KMT”), with an intention to pro-
mote consensus within the party. On October 10, 1921, the Chinese-version The
International Development of China by Mr. Sun Yat-sen was formally published by
Shanghai Minzhi Publishing House. In 1922, the English-version The International
Development of China by Mr. Sun Yat-sen was published in New York, US.‡ On
August 17, 1924, Sun Yat-sen delivered a speech titled the Principles of People’s
Livelihood at Guangzhou National Higher Normal School, in which he mentioned
again the development of hydropower resources in Three Gorges.
Sun Yat-sen fully employed multiple communication types including interper-

sonal communication, organizational communication, group communication and
mass communication, combined his project ideal with the rejuvenation of the nation,
gave the ideal of TGP a greater social influence gradually, and allowed the envisage-
ment about the construction of TGP to move difficultly from a dream toward a reality.
2. THE CONMMUNICATION OF TGP FROM DREAM TO FEASIBILITY
Mr. Sun Yat-sen’s political ideal about TGP gradually stepped into the techni-
cal research of water conservancy. The survey research reports and published by a
series of professional technical periodicals and sporadic media reports were highly
professional. For example, the “Powell Plan” proposed by Siaeny J. Powell, a British
engineer, was covered by the then Shun Pao. In 1929, Chen Zhan’en, the Chief
of Water Conservancy Section, serialized the Recent Situation of Yangtze River
and Remediation Opinions on the Yangtze River Watercourse Monthly § . In Octo-
ber 1932, the Survey Report on Hydroelectric Generation in the Upper Reaches
of Yangtze River was developed by Yun Zhen and other engineers, which was
serialized on Engineering, 1933 No.3 and No.4 proceedings of Chinese Institute of
Engineers.
In 1944, G.R. Paschal, a renowned economist in US, presented the well-known

Paschal Report, which was praised and strongly recommended by the Resource
Commission of the Nationalist Government, and the American government was
also interested in this report.
In the fall of 1944, Dr. J.L. Savage, the chief design engineer and a high dam
expert with the Bureau of Reclamation, US (“USBR”) presented the Initial Report
‡ Zhang Lixian (author): Memory about Three Gorges in the Period of Republic of China,
China Three Gorges Publishing House, Edition 2017, p18 and p19.
§ Zhang Lixian (author): Memory about Three Gorges in the Period of Republic of China,
China Three Gorges Publishing House, Edition 2017, p27.

Yun Zhen, Cao Ruizhi and Song Xishang: Survey Report on Hydroelectric Generation in
the Upper Reaches of Yangtze River by referring to Zhang Yunchang and Luan Linbin (chief
editors): A Compilation of Historical Literatures for China Three Gorges Project (1948–1949),
592
Q.105 - R.34
on the Yangtze River Gorges Plan¶ , or the stunning Savage Plan. The Nationalist
Government General sent the Savage Plan to US President Roosevelt via General
Nelson. The White House of US disclosed this world’s largest water conservancy
project to the press, and shocked the world. The then Commercial Counselor of UK
in Chongqing personally went to Changshou to visit Savage, and reported the TGP
to London for UK to make investment and supply equipment after the war. This was
the first time for TGP to arouse extensive interest of foreign countries in terms of
investment and equipment.∗∗
Central News Agency, UPI, Washington on December 26, 1945: “An engineer
from the Bureau of Reclamation, US will start to design a general plan for the devel-
opment of Yangtze River basin. Chinese Government intends to build the world’s
largest dam and hydropower plant in the Yangtze River basin, irrigate 10 million
acres of land, and protect this area from flooding again.”††
On February 28, 1946, an article on the Engineering News-Record (ENR) of

US reported that “China plans an unprecedented dam over the Yangtze River, which
integrates power generation, irrigation and shipping.” According to the agreement
by and between USBR and the Resource Commission of Nationalist Government,
the dam planning was under way.
In September 1947, the News Office of Executive Yuan, Nationalist Govern-

ment compiled and printed The Plan on Yangtze Three Gorges Water Conservancy
Project in order to disseminate the Three Gorges hydropower project.
This stage shows that, the earlier survey report and professional articles were
disseminated mostly by academic publications. Such industrial communication had
a less social effect, and promoted the government policies in a limited way. As the top
experts such as Paschal and Savage, who had both celebrity effect and professional
competence, pushed the project ideal onto a substantial step in a manner of official
document, the important models of “celebrity communication” and “official document
communication” were formed for TGP, and allowed the initial envisagement and
earlier survey of TGP to enter the stage of actual promotion by the government. The
communication of TGP also moved from the project “philosophy communication” to
the project “feasibility communication” stage.
¶ Zhang Yunchang and Luan Linbin (chief editors): A Compilation of Historical Literatures for
China Three Gorges Project (1948–1949), China Three Gorges Publishing House, Edition
2010, p48–72.
∗∗ Compilation Committee of China Three Gorges Construction Yearbook: China Three
Gorges Construction Yearbook 1994, Beijing, China Three Gorges Publishing House, 1995,
p197.
†† Zhang Lixian (author): Memory about Three Gorges in the Period of Republic of China,
593
Q.105 - R.34
3. THE COMMUNICATION OF TGP FROM CONTROVERSIAL

ARGUMENTATION TO DECISION MAKING
On October 1st , 1949, the New China was founded. The history of China
opened a new chapter, and the construction and communication of TGP also entered
a new phase, during which the communication of TGP had been an act of country.
In June 1956, Mao Zedong wrote the poem Swimming after swimming Yangtze
River in Wuhan, singing a new movement of the construction of TGP. The line “Till a
smooth lake rises in the narrow gorges” describes a nice picture of the harnessing
Yangtze river for national people. Through the poem, the project image was contin-
uously disseminated among the public by word of mouth. And the poem stimulated
the public’s yearning and anticipation for the construction of TGP.
In the first half of 1956, Lin Yishan, Director of Yangtze River Basin Planning
Office published an article titled “Discussion of Several Questions about Yangtze
River Basin Planning” on the China Water Resources journal No.5 and No.6.‡‡ In
September 1956, Li Rui, the then Director General of General Administration of
Hydropower Construction, Ministry of Electric Power Industry, wrote an article titled
Several Questions about Yangtze River Basin Planning,§§ which was demonstrating
the original differences of opinion during the construction of TGP. In December 1985,
an article titled “CPPCC Suggests to Suspend the Construction of Three Gorges
Dam due to Disadvantages Outweighing the Advantages” was published on the
Wenhui Daily of Hong Kong.
In June 1986, the Central Government decided that the Ministry of Hydropower
would organize experts in different fields to present a new feasibility report for the
TGP on the basis of wide solicitation of opinions and in-depth research and argumen-
tation. This time, the argumentation fell into 10 sessions, with 412 experts invited
from various industries across the country, who formed 14 expert groups. The TGP
argumentation players included the government officials at different levels, experts
and scholars from such circles of water conservancy, ecology, social sciences and
geology, journalists, networks and media, as well as relocated people of some
hydropower stations and the general public. The argumentation featured multiple
themes, and the construction of TGP became a mass communication event.
On May 22, 1991, relevant authorities of China convoked the press briefing
on TGP dissemination, clarifying the main direction of coverage and dissemination.
Later, Li Boning created, directed and personally produced a feature program Poor
Mountains Are Calling. On June 13, the office of the TGP Review Committee and
‡‡ Lin Yishan: Discussion of Several Questions about Yangtze River Basin Planning, China
Water Resources 1956 No. 5 and No.6.

§§ Li Rui: Several Questions about Yangtze River Basin Planning, Water Power, issue of
September 1956.
594
Q.105 - R.34
the Ministry of Water Resources organized heads of 23 press media in the capital
of China to have a field visit to the Three Gorges reservoir area and dam site. The
Dissemination Section of Yangtze River Office printed the Brief Report on Integrated
Utilization and Planning of Yangtze River Basin with a yellow cover and the Yangtze
Three Gorges Water Resources Argumentation with a green cover.
Before and after this, major media in the capital and local areas of China
opened special columns with their own unique features according to their own read-
ers base and focus of attention. Such targeted column coverage played a great role
in deeply disseminating the TGP.
To provide the people with visualized understanding of the scale and function
of TGP, 11 units including the Ministry of Water Resources, the Ministry of Energy
and the State Scientific and Technological Commission jointly held the Three Gorges
Project Exhibition at the Military Museum, which provided a powerful cooperation
for the dissemination of TGP, and produced a huge effect.
On April 3, 1992, the Resolution Concerning the Construction of Three Gorges

Project was adopted at the 5th Session of the 7th National People’s Congress with
1767 votes for and 177 against, and 664 abstentions.
More directly, the TGP communication in this stage featured controversy in

feasibility communication. Controversial argumentation could provide intellectual
support the decision makers, eliminate any potential dead zone of cognition, enable
the decision makers to obtain various project-related information in a wider range
and to a greater extent, promote the project decision making to be reasonable, and
contribute to a more democratic result of decision making for the project. Led by
the country, the TGP communication during decision making made preparations
in advance, fully created an environment of communication, and proactively dis-
played the necessity and feasibility arguments for the earlier argumentation, and
laid a social foundation for eliminating doubts, increasing cognition, and wooing the
acknowledgement of a greater majority of common people.
4. THE COMMUNICATION OF TGP FROM CONSRUCTION TO OPERATION
On September 27, 1993, China Three Gorges Project Corporation was

founded. During the construction of TGP, the communication of TGP was carried
out by the company rather than the country.
In October 1994, China Three Gorges Project Corporation established the

Press Release Center formally as a core organization guiding and disseminating the
TGP. On September 27, 1993, China Three Gorges Project Newspaper was formally

Cheng Shoutai: Before and After the Adoption of Resolution on Constructing the Three
Gorges Project [J]. Hundred Year Tide, 2010(5).
595
Q.105 - R.34
published, with a supplement of Three Gorges Weekly in 1998. On December 14,

1994, the first issue of China Three Gorges Construction journal was published
before the formal commencement of TGP construction. In 1994, preparations were
made for the publication of China Three Gorges Construction Yearbook. On May 18,
1995, the Three Gorges Exhibition Center opened formally. In 2002, the website of
China Three Gorges Project Corporation was revised to give full play to the Internet
dissemination.
To intensify the audience’s positive understanding of TGP construction, China
Three Gorges Corporation (CTG”) convoked press conferences and planned for
important topics at such milestones as the formal commencement of the project
construction, interception of river, power generation of generator sets, and success-
ful impoundment, spoke to the outside world through foreign and domestic media,
answered questions from the public and set up an open and transparent project
image while increasing the TGP’s social influence. At the same time, by relying on
Chinese National Committee on Large Dams, a national first-level committee, CTG
established the Committee on Public Cognition and Public Relations of Reservoirs
and Large Dams, held the Forum on Public Cognition of Reservoirs and Large Dams,
inviting experts and scholars to promote and boost the public’s cognition of TGP.
Since TGP’s successful construction and operation, CTG, as the main body
of communication, has been further improving the professional degree of news
planning, organization and dissemination in the lifecycle, eliminating doubts and
confusion, and leading the public opinion, with a sound and positive social opinion
field formed outside the Group day by day. The TGP communication in this stage
promoted acknowledgement with cognition, served the project effectively through
the project, and became part of the project construction.
5. EXPLORATION ON THE REGULARITY OF TGP COMMUNICATION
Looking through the TGP communication history, we can say that the com-
munication act was throughout the project lifecycle, and we had to solve different
problems in different communication stages. Today, it is necessary for us to grasp
relevant laws in the evolution of history.
Project communication is an important part of the project. Project commu-
nication is both a communication act and a project act, running through the project
lifecycle. It is an organic component moving forwarding synchronously with the phys-
ical project and equally important. Through project communication, it is possible to
pass the project philosophy and build the project image, reach a social consensus
around the project, and boost the construction of the physical project. Without a
better communication, without a better interaction with the public, there will be no
good project. The value of project communication is to better highlight the value of
project activities.
No dispute, no project. Any proposed significant project will inevitably
challenge the human’s existing cognition. Different cognition models and paths
are always likely to lead to the review of viewpoints, whereupon disputes will
596
Q.105 - R.34
occur. However, the “controversial nature” of a project is not always a negative

obstacle. Standing on a wider field of view, it is different opinions and different
“disputes” that further optimize and improve the project ideal, design and reality.
Sufficient opinion expression and effective dispute resolution would allow dissenting
opinions to become constructive ones, which enriches the connotation of project
communication.
Project image building is the core of project communication. The building
of a better project image may lead the project communication toward optimization
of project image. From the initial patriotism of “Constructing the Three Gorges for
China” to today’s “one mark and three examples”, TGP is the pillar of a big country.
The project image has kept deepening, enriching, and even destructing over time.
But the basic laws that the project image leads the project communication will not
change.
The aim of project communication is to boost the public’s cognition and
acknowledge of a project. Project communication is significantly public and inter-
active. On the one hand, it certainly has an aim of boosting the project construction.
On the other hand, it has to feed back the public’s cognition, emotion and demands
for the project. A mechanism of collecting the public opinions and feedback of pub-
lic emotion will enable continuous assessment, adjustment and optimization of the
construction of physical project, allow the project to be more consistent with the
social value and code of ethics, and finally help push the project activities to move
forward continuously.
In a new era, a series of national significant projects will also necessarily go
beyond the real entities and become a power force that shape the civilization, trans-
form the world and affect the future. In a project era, the harmonious coexistence
among the artificial nature, human society and natural ecology can only be achieved
through improving the project attainment, following the project laws, and conduct
proper project communication. This will drive the high-quality sustainable economic
and social development, help the Chinese nation realize a great rejuvenation, and
contribute more ideas, experience and solutions of China to the world’s development.
REFERENCES
[1] Li Bocong. Engineering Innovation: Breaking through barriers and eluding

traps [M]. Zhejiang University Press, Nov. 2010.
[2] Chen Xi. The Communist Party of China and the Three Gorges Project [M].
CPC History Press, Nov. 2014.
[3] Yin Ruiyu, Wang Yingluo and Li Bocong. Philosophy of Engineering [M]. Higher
Education Press, July 2013.
[4] Yin Ruiyu, Wang Yingluo and Li Bocong. Engineering Philosophy [M]. Higher
Education Press, Aug. 2017.
597
Q.105 - R.34
[5] Li Zhenfang. A Study on the PR Communication Effect of Three Gorges Project

[M]. Huazhong University of Science and Technology Press, Sep. 2017.
[6] Zhang Lixian. Memory about Three Gorges in the Period of Republic of China
[M]. China Three Gorges Publishing House, May 2017.
[7] Li Bocong, Yin Ruiyu and Zhai Guangming. Engineering Philosophy of Daqing
Oil Field: Practice and Theory [M]. Petroleum Industry Press, Jan. 2017.
[8] Lu Yuegang. Argumentation about Three Gorges in 50 Years. [M]. China Social
Sciences Press, Dec. 1993.
[9] Co-compiled by Lingnan Editorial Board of Collected Books and Guangdong
Chinese Culture Promotion Society. A Collection of Essential Articles by Sun
Yat-sen (Vol. 1) [M]. Guangdong People’s Publishing House, 1996.
[10] Zhang Yunchang and Luan Linbin (chief editors): A Compilation of Historical
Literatures for China Three Gorges Project (1918–1949) [M], China Three
Gorges Publishing House, 2010.
[11] Sun Yat-sen. The Three People’s Principles [M]. Yuelu Press, 2000.
[12] Compiled by Hubei Provincial Cultural and Historical Records Committee of
the Chinese People’s Political Consultative Conference and Yichang Munici-
pal Learning Culture and History Committee of the Chinese People’s Political
Consultative Conference. A Broad View of Three Gorges Culture and History
[M]. Chinese Culture and History Press, 1997.
[13] Compiled by the Literature Research Office of the CPC Central Committee.
A Chronicle of Mao Zedong (1946–1976) Volume 2 [M]. Central Literature
Press, 2013.
[14] Compiled by the Office of CPPCC Working Group. Selected Materials of
CPPCC National Committee Working Group [M]. Chinese Culture and History
Press, 1987.
[15] Lin Yishan: Discussion of Several Questions about Yangtze River Basin
Planning [J]. China Water Resources, May 1956.
[16] Li Rui: Several Questions about Yangtze River Basin Planning [J]. Water
Power: Sep. 1956.
[17] Guo Tao. Sun Yat-sen and the Three Gorges Project [J]. China Three Gorges,
Mar. 2016: 16–37.
[18] J.W.Morris and Zhang Jiliang. A letter from Morris, Retired Lieutenant General
of USACE to the Editor for the Three Gorges Project. Yangtze River, Feb. 1981.
[19] Cheng Shoutai: Before and After the Adoption of Resolution on Constructing
the Three Gorges Project [J]. Hundred Year Tide, May 2010.
[20] Wang Yihong, Jiang Xiaochuan, Zhang Xiaowei, Long Xinwei and Li Ke.
Project Communication As a New Target and New Content of Communication
Sciences Research [J]. Studies in Dialectics of Nature, Nov. 2011: 53–60.
598
Q.105 - R.35
-------
-------
CONSEQUENCES ON THE DESIGN PRACTICE OF WORLDWIDE DAM

ACCIDENTS (∗)
Roger BREMEN
Lombardi SA Ingegneri Consulenti
SWITZERLAND
SUMMARY
The available statistics on dam accidents and failures are used to identify some
approaches to be followed to reduce the risk of accidents of existing and new dams.
The available statistical information has to be combined with the evaluation of the
specific properties of each dam for a better risk evaluation. This approach is used
to reorient the priorities in the safety analyses of existing and new dams.
The concept of vulnerability of a dam is introduced which is felt of a better prac-

tical use in comparison to a failure probability. The concept of vulnerability does not
involve a quantitative evaluation of probabilities but allows to take into consideration
the consequences of events exceeding the considered limit. The identification of the
specific vulnerabilities and its consequences should thus guide future considerations
on dam safety.
RÉSUMÉ
Les données statistiques disponibles concernant la rupture des barrages

sont discutées pour identifier les orientations principales qui permettent de réduire
∗ Leçons tirées des accidents de barrage dans le monde
DOI: 10.1201/9781003211808-81 599

Q.105 - R.35
la probabilité de rupture des barrages existants et des nouvelles structures. Les

informations statistiques sont à combiner avec les caractéristiques spécifiques de
chaque ouvrage afin de mieux évaluer les risques associés. Cette approche est
proposée afin de réorienter les priorités actuelles dans le domaine de l’analyses de
risques pour les ouvrages de retenue.
L’article introduit le concept de vulnérabilité qui est considéré plus parlant

que la probabilité de rupture. Le concept de vulnérabilité ne comporte pas une
évaluation quantitative du risque mais permet de considérer les conséquences d’un
évènement dépassant la limite considérée au niveau du projet. La mise en évidence
des vulnérabilités et ses conséquences doit permettre de mieux cibler les analyses
de sécurité des barrages.
1. INTRODUCTION
Progress is often based on mistakes and accidents. It is thus of foremost impor-

tance to evaluate in detail the causes of dam accidents. Improvements in the dam
design have therefore to take into account and properly evaluate these accidents
to adopt the adequate measures. These measures range from the identification of
opportune design approaches and concepts but also the evaluation of the impact
of other aspects as monitoring on the risk of accidents. Design and supervision
practices are however only marginally driven by a consequent analysis and under-
standing of the statistics of dam accidents. The purpose of the present contribution
is to highlight the benefits of an appropriated use of failure statistics to reorient the
priorities in the safety analyses of existing and new dams.
2. USE OF STATISTICS
Having reached a fellow age, I recently decided to submit myself to a health

check. After having answered many questions on my health background, the young
doctor introduced all these data into a modern software and very quickly he came out
with his powerful diagnosis. The most relevant results of his report are the following:
• probability to die in the next year : 3.2 × 10−2

• probability to get a cancer in the next year : 7.3 × 10−2
• probability of heart attack in the next year : 4.5 × 10−2
• etc.
Then he explained me that I could reduce some of the previous probabilities by

taking some remedial actions including in particular more sport activities, no more
600
Q.105 - R.35
Fig. 1
Statistics on worldwide dam failures (except China)
Statistique mondiale des ruptures de Barrages (Chine exclue)
alcohol, less work, etc. At the end of the visit he handed me over a nice very colored
report with many graphs and probabilities and many suggestions how to reduce
some probabilities. Back at home I was asked by my wife about my health situation.
I was not sure what to answer and if I should start to speak about probabilities…
As we would never accept the sole use of statistics for our personnel health
check, (without any specific investigation) we should also never misuse statistics
and probabilities when we are dealing with safety aspects related to dams.
Exactly as for humans, statistics are correct and applicable for the entire pop-
ulation (of humans and dams) but are of a very limited use once we have to deal with
a specific sample. As shown in Fig. 1, the failure probability of dams is the result of
an average behavior of the entire population. To consider that any dam is behaving
as the average is equivalent to state that everybody is 1.85 m tall as the average
height of the world population.
As shown in Fig. 1 the annual failure probability of dams is continuously

decreasing having reached a maximum value of 7.6 10−4 in 1929. Applying in
our analyses an average probability to a specific dam is assuming that this dam
will behave as the average of its population. Any conclusion derived from such
an approach represents is giving us the illusion to quantify the uncertainty. We do
not know the specific failure risk of each dam, since otherwise we could intervene
were necessary, we only know the average behavior. Also subdividing the dams
according to their type does not help much. Statistics on dam failures are useless
601
Q.105 - R.35
to quantify the failure probability of a specific dam, they only give an idea on the
failure probability of the entire population without any indication on which dam will
be the next. Our general urge towards a quantification as done in the last decades in
the domain of dams (not only) is promoting the illusion that computers may quantify
the uncertainty. Decisions taken on a single object essentially based on the statis-
tical results of a population may result in a severe risk underestimate. We should
thus never forget that nobody (nor humans nor dams) behave as the average of the
population.
3. STATISTICS OF DAM FAILURES AND INCIDENTS
On the other hand, statistics are essential to address the main causes of
dam failures and orient any specific investigation. Learning from past accidents
is fundamental for progresses in dam safety in general and for the design of new
dams. The main information source about dam incidents and failures is represented
by ICOLD literature (mostly ICOLD Bulletin 99). Updated databases are registering
dam incidents and failures during the last 30 years. Fig. 2 shows as an example the
number of dam incidents subdivided according to the dam type and the main cause
based on dams in U.S.A.[2]
Fig. 2
Distribution of dam incidents according to dam type and cause.
Distribution des accidents de barrages selon type de barrage et cause.
602
Q.105 - R.35
Very few statistical data are available from dams in China as well as tailing
dams in order that the overall statistical information is not complete. Based on the
available statistics the following main aspects might be highlighted:
• The annual failure probability of dams is continuously decreasing being actu-

ally in the range of 15 10−5 /year and having reached a maximum value of
7.6 10−4 in 1929 (see Fig. 1). It is interesting to notice that ICOLD was created
in 1928.
• Overtopping is the most common cause of dam failure with approx. 30 fail-
ures recorded since 1930. All the failures are referred to embankment dams
resulting in an annual failure probability of 5 × 10−5 /year.
• The second cause of dam collapse is due to the dam body collapse (approx.
28 cases after 1930) with a failure probability of embankment dams of
4.4 × 10−5 /year. Foundation failure is the third cause (14 cases recorded after
1930).
• The annual failure probability due to earthquakes is 0.3 × 10−5 all concerning
embankment dams (4 recorded cases after 1930).
Overtopping as main cause of dam failure is often induced by an extraordinary

event as extreme weather conditions, failure of upstream dam or a large slope insta-
bility upstream of the dam. Embankment dams are significantly more vulnerable to
the consequences of an overtopping than most of the other dam types.
The collapse of the dam body and the foundation failure may have very different
causes and may affect any type of dam with rather similar annual probabilities.
Improving the dynamic dam safety as pushed during the last decades will thus
not affect significantly the failure probability of dams. Any contribution in reducing the
overtopping risk or its consequences on the dam might on the contrary be beneficial
to reduce the global failure probability. Or expressed differently it is statistically 15
times more probable that a dam fails due to overtopping than due to an earthquake.
The focus during the last decades to improve the seismic safety of dams does not
find a justification into the statistics of dam failures.
4. DYNAMIC SAFETY OF DAMS
As previously shown earthquakes are by far not the main cause of dam failures.
Statistically only 1 every 15/20 dam failure is due to an earthquake. It is thus ques-
tionable why the dam community is since various decades focused on the dynamic
safety of dams much more than for other domains causing by far more dam failures.
To explain this attitude, we have to consider the world statistics of global deaths
from natural disasters. If in the past drought and floods have been the main cause of
603
Q.105 - R.35
Fig. 3
World statistics on global deaths from natural desasters
Statique mondiale des décès en relation avec des évènements naturels
Table 1
Embankment dams collapsed after an earthquake
Ruptures de digues comme conséquence d’un séisme
FAILED DAM EARTHQUAKE RICHTER MAGNITUDE YEAR
Van Norman Lake San Fernando, California 6.6 1971
Lliu-Lliu
Algarrobo, Chile 8.0 1985
Lo Ovalle
Fujinuma Tohoku, Japan 9.0 2011
deaths, hurricanes and earthquakes are actually the most relevant causes of human
injury due to natural disasters as shown in Fig. 3.
The worldwide attention paid to improve the dynamic safety of structures and
buildings is thus a natural consequence of the previous figure. However, if seismic
loads are one of the main failure causes of buildings and other infrastructures, this
is by far not the case for dams as previously shown. An improved dynamic safety of
buildings is thus contributing to reduce the injuries due to earthquakes (as clearly
shown for example in Japan and Chile). On the contrary the improvement of the
seismic safety of dams will not substantially contribute to reduce the number of
injuries related to dams.
Table 1 includes the 4 embankment dams collapsed as consequence of an

earthquake.
604
Q.105 - R.35
However, many other structures have been submitted to very severe accelera-
tions due to earthquakes with some of them also being damaged. None of them have
seen their ability to retain water compromised and most of them are still in service.
The ability of dams in general to support dynamic loads is in general far better than
expected or resulting from our numerical analysis. On the other side and despite
the failure statistics we clearly underestimate the risks and potential consequences
related to failure mechanism not related to seismic loads.
In conclusion the attention paid to the dynamic safety of dams has been driven
by the dynamic vulnerability of infrastructures being the main cause of human injury.
However, extending this approach also to dams appears not appropriate since the
main causes of dam failures are not related to dynamic loads but are due to over-
topping, internal erosion, and foundation collapse. It is thus felt important that the
dam community will pay in future more attention on these aspects rather than focus
any safety consideration mostly on the dynamic behavior.
5. CONSEQUENCES ON THE DESIGN PRACTICE
To identify potential consequences on the design practice for new or existing

dams it is felt useful to introduce the concept of vulnerability. In the contest of dams,
the vulnerability might be defined as the openness of a dam to suffer severe con-
sequences due to unexpected conditions. The concept of vulnerability is implicitly
considered common design approaches but generally not systematically evaluated.
The vulnerability of an embankment dam to overtopping is definitively different from
a concrete dam on a rock foundation. Hoverer, for the spillway design criteria or
other risk scenarios resulting in a potential overtopping the significant difference in
terms of vulnerability of the two dam types is often not considered.
As previously mentioned significant attention has been paid during the last
decades in the reduction of the seismic vulnerability of dams, whereas relatively
little progress has been made in the evaluation and reduction of other vulnerabilities
as the consequences of overtopping or internal erosion, or increased pore pressures
in the dam body or its foundation.
The vulnerability of a given structure to conditions which differ from the design
assumptions is depending only from the specific properties of this structure and
its foundation. The vulnerability to internal erosion essentially depends from the
material properties of the earthfill.
The definition of a probability of occurrence of an extreme event is highly

difficult and this probability may vary over time. Most of the dam accidents are the
result of unexpected events which have not been considered in any previous risk
analysis.
605
Q.105 - R.35
Considerations based on the identification of the specific vulnerabilities and

the definition of measures allowing their reduction are considered more efficient to
improve in general terms the dam safety than an approach based on the estimate of
the probability of occurrence of a given event, which is rather subjective and depends
significantly from the available database. Thus rather than evaluate the probability
of an overtopping, which is rather difficult and questionable, it is felt more favorable
to identify the required measures to reduce the vulnerability of the dam due to such
an event in particular if the consequences of the failure would be significant.
Based on the previous approach, an embankment dam equipped with a gated

spillway in a remote mountainous area subject to potential landslides and located
upstream of inhabited areas should be considered with particular attention. A con-
crete dam under the same conditions would statistically show a significantly less
vulnerability due to overtopping. This significant difference has to be considered at
the design phase in order to properly address advantages and disadvantages of
different alternatives.
As regards structures under operation, a vulnerability analysis helps to identify

which conditions might result in a potential dam failure. Most of the failure modes
might be included in one of the following scenarios:
• Overtopping
• Internal erosion and seepage
• Instabilities of the dam body
• Foundation instabilities
Instabilities of the dam body and of the foundation might be due to different
causes as water pressures, external static or dynamic loads, material deterioration,
etc.
To reduce the failure risk, two approaches and their combination are basically
possible:
• Mitigation measures to reduce the vulnerability of the dam.

• Measures to reduce the uncertainties concerning the actions on the dam.
Mitigation measures are resulting into a strengthening of the structure and

its foundations. The measures to reduce the uncertainties related to the actions
on the dam are more diversified. Measures to reduce the uncertainties have to be
focused on the parameters and causes responsible for the specific vulnerability.
For the reduction of the uncertainty reference has only to be made to the specific
characteristics of the identified dam and no support might be given by statistics
based on a large population of dams.
606
Q.105 - R.35
6. CONCLUSIONS
In the present contribution, some statistical results on dam failures are dis-
cussed to identify a global approach to reduce the failure risk of dams. The approach
is based on the fact that no dam behaves as the average in order that it appears an
illusion to establish a failure probability of a single dam. In the design considerations
both for existing and new dams, the concept of vulnerability has to be introduced.
Any safety consideration has thus to be focused on a proper evaluation of the vul-
nerabilities in combination with the uncertainty on the actions on the dam. Both the
analysis of the vulnerability as well as the measures to reduce the uncertainty of the
causes have to be specifically focused on the characteristics of the single structure.
REFERENCES
[1] International Committe of Large Dams (ICOLD); Bullettin 99: Dam Failures –
Statistical Analysis; Mars 1995.
[2] Paul Scherrer Institut (PSI); Project GaBE: Comprehensive Assessment of

Energy Systems – Severe Accidents in the Energy Sector; First Edition; PSI
Bericht Nr. 98-16; November 1998.
[3] Bundesamt für Wasser und Geologie (BWG); Sicherheit der Stauanlagen –
Basisdokument zu dem Nachweis der Erdbebensicherheit; Version 1.2; März
2003.
[4] Dr. Eng. Giovanni Lombardi; Concrete dams and their foundation evalua-
tion for static loading; International Workshop on Dam Safety Evaluation;
Grindenwald, Switzerland, 26–28 April 1993.
[5] Pierre Duffaut; What modern rock mechanics owe to the Malpasset arch
dam failure; Harmonising Rock Engineering and the Environment; Taylor and
Francis Group, London, 2012.
[6] Ministère du développement durable; Rupture d’un barrage: Le 2 décembre

1959 – Malpasset, France; Avril 2009.
607
Q.105 - R.36
-------
-------
EMERGENCY PLANNING: LESSONS LEARNED REGARDING

REGLEMENTATION, ORGANISATION AND IMPLEMENTATION IN
SWITZERLAND(∗)
Jonathan FAURIEL
Surveillance and dam safety,
Alpiq SA
Alexandra BECKSTEIN
Supervision of Dams,
Swiss Federal Office of Energy SFOE
Olivier FOURNIER
Office cantonal de la protection de la population,
Canton Wallis
Nicolas ADAM
Civil engineer,
Alpiq SA
SWITZERLAND
SUMMARY
Ever since the construction of the first dams in Switzerland, the regulatory
framework has been evolving, but the Emergency Planning has always been a sub-
ject treated by the Swiss Authorities. If in the 40’s, most of the directives came
from army considerations, since 2010 several new legislations provide obligations
and indications to the stakeholders to improve the efficiency of the emergency plan.
∗ Planification en cas d’urgence : retour d’expérience sur la réglementation, l’organisation et
la mise en œuvre en Suisse
608 DOI: 10.1201/9781003211808-82

Q.105 - R.36
Cleuson dam was chosen as a pilot to establish its emergency regulations and carry
out an emergency exercise and hence test whether the regulations are sufficient in
the event of a safety-relevant event. The emergency regulations were tested right
after its first validation by the authorities. By involving the Operating Staff, the Experts
and the Owner, the internal and external communication channels were tested, as
well as the developed emergency strategies. It was shown that exercises are essen-
tial to detect and adapt weak points in the system. The exposure of stakeholders
into a realistic exercise scenario helps to identify several useful lessons learned to
improve the effectiveness of the emergency plan and allows an ongoing adaptation.
RÉSUMÉ
Depuis la construction des premiers barrages en Suisse, le cas d’urgence a

toujours été pris en compte par les Autorités mais le cadre réglementaire a con-
sidérablement évolué depuis. Si dans les années 40, la plupart des directives
étaient régies par l’armée, plusieurs textes de lois édités depuis 2010 redéfinis-
sent les obligations des parties prenantes dans le but d’améliorer l’efficacité des
plans d’urgence. Face à cette nouvelle législation, le barrage de Cleuson a été
choisi comme ouvrage pilote. Son règlement en cas d’urgence a été établi et un
exercice impliquant la sécurité de l’ouvrage a été mené pour pouvoir tester la per-
tinence du plan d’urgence directement après la première validation du règlement
par les autorités. En impliquant le personnel d’exploitation, les experts et le proprié-
taire, les canaux de communication internes et externes ont été testés, ainsi que les
stratégies d’urgence développées. Il a été démontré que les exercices sont essen-
tiels pour détecter et adapter les points faibles du système. L’exposition des parties
prenantes à un scénario d’exercice réaliste permet d’identifier un certain nombre
de leçons fort utiles pour améliorer l’efficacité du plan d’urgence et permettre une
adaptation continue des règlements.
1. INTRODUCTION
The history of dam construction and operation in Switzerland does not show
any major disorder involving a structural element. Even so, the emergency plan
remains one of the three fundamental pillars of dam safety. The Federal Act in
Water Retaining Facilities (2010), its associated Ordinance (2012) and the related
Directives (2015) provide very clear indications on the responsibilities, organization
and implementation of the emergency plan. Every operator has the duty to develop
and implement emergency regulations including the alarm procedures that have to
be implemented by civil protection responsibles. The aim is to plan the measures to
be taken by the owner/operator of a water retaining facility in the event that the dam
safety is no longer guaranteed. With the implementation of the Directive on the Safety
609
Q.105 - R.36
of Dams Part E: Emergency Planning, dam operators are required to execute an

emergency exercise every five years. The exercise to be conducted is intended to test
the robustness of the emergency concept and to identify weaknesses of the concept.
After a brief summary of the history of emergency planning in Switzerland

and its legal framework (chapter 2), the outlines of the Emergency Concept are
presented (chapter 3). The implementation of the emergency regulation is then
discussed based on the example of Cleuson dam and the exercise conducted in
August 2018 (Chapter 4). The final chapter is dedicated to an overall view given by
each of the three main stakeholders: the Operator, the Civil Protection Authority and
the Supervisory Authority.
2. REGULATORY FRAMEWORK
2.1. HISTORY OF EMERGENCY PLANNING AND DEVELOPMENT OF THE REGULATORY

FRAMEWORK
The history of Swiss emergency planning for dams is closely linked to the
context of the Second World War. Out of concern for an attack on dam structures, the
Swiss government had begun in 1945, based on a federal decree of 1943, to install
a nationwide water alarm siren network, which, in the event of an attack, would call
on the population to flee. This network is still a fundamental part of the emergency
concept today and the installation of the water alarm system is a precondition for
the further development of the security concept to its current extent [4]. A civil legal
basis for the implementation of the emergency concept was only created later, in
1954 and extended with the enforcement of the Dam Ordinance in 1957, which was
adapted in 1971 and again in 2010 to the legal basis that is still valid today. While the
emergency concepts at their time of origin in the 1950s were clearly designed for the
case of a military attack, with the creation of the new directive and its application, the
adaptation to more modern threats such as natural dangers or failure of IT systems
are integrated.
2.2. ACTUAL REGULATORY FRAMEWORK
The Federal Act on Water Retaining Facilities of 2010 [1] and the associated
Water Retaining Facilities Ordinance of 2012 [2] define the measures to be taken by
operators to prevent and manage an event. In 2015, the publication of the guideline
[3] additionally specified how the provisions of the law and the ordinance are to be
implemented. Specifically, the Federal Act on Water Retaining Facilities determines
that the operator must take precautions if the safe operation of a dam is no longer
guaranteed. It also stipulates that in the event of an emergency, the confederation,
cantons and communes must use the resources and structures of civil protection
610
Q.105 - R.36
to prepare instructions on how to behave and to evacuate the site if necessary.

Regarding alarming, dams with more than 2 million m3 storage capacity and all
dams that would endanger more than 1,000 people in the event of a dam failure,
must have a water alarm system installed. The ordinance specifies, that the operator
must draw up a dam-specific emergency regulation, and that the persons in charge
of cantonal civil protection must draw up evacuation plans [2].
The technical requirements for the water alarm systems as well as the nec-
essary structural facilities for the functioning of the system are defined in the Civil
Protection Act of 2002 and in the Alarm and Security Radio Regulations of 2010,
the implementation of which is the responsibility of the civil protection authorities.
3. EMERGENCY PLAN DESCRIPTION
3.1. EMERGENCY CONCEPT
As shown in Figure 1, Emergency Planning is the third of the three pillars

of the Swiss Dam Safety concept. If Structural Safety and Surveillance are mostly
dedicated to risk mitigation, Emergency Planning clearly has the objective to deal
with the residual risk. This approach is key to understanding the concept of the
Emergency Planning. In other words: “What to do if the first two pillars fail?”.
Fig. 1
Swiss Three-Pillar Dam Safety Concept
Les trois piliers du concept de sécurité en Suisse
611
Q.105 - R.36
The obligation for planning of the mastery of an emergency event lies with
the dam operator. He must involve the authorities responsible for civil protection in
the planning and must submit the concept to the supervisory authority for approval.
In addition, the emergency concept must be regularly exercised and, if necessary,
adapted. The planning for the event of an incident must be documented in an emer-
gency response regulation [1]. The content of these regulations is dependent on
the size of the water retaining facility and the potential damage caused by a dam
failure. The topics that must be included in every emergency regulation are : a flood
map that shows the flooded areas in the event of a sudden and complete failure, a
risk analysis to clarify which influences could impede emergency management, an
emergency strategy that defines the measures to be taken per danger level and the
organization of the responsible persons including their redundant availability. The
main results and all operational needed information are then recorded in the form
of an emergency response dossier which is distributed to all persons involved in an
event of an emergency.
In order to ensure that all those involved know which instructions to follow,
a danger level is determined based on the progress of the incident. Five danger
levels are defined for the event of a dam break, analogous to the danger levels of
natural hazards, whereby danger level 1 describes normal operation and danger
level 5 the imminent or uncontrolled outflow of water from the reservoir. Description
of the danger levels is given in Figure 2. For each danger level, clear instructions are
Fig. 2
Description of the danger levels (from 1 to 5)
Description des niveaux de dangers (de 1 à 5)
612
Q.105 - R.36
Fig. 3
Context of early warning, alarming and evacuation
Contexte d’alerte préalable, d’alarme et d’évacuation
defined for all those involved and included in the emergency regulations in the form of
a checklist. Triggering the sirens depends on the current danger level. In danger level
4 the population is warned by the general alarm, which means that the authorities of
civil protection still have time to advise the population via radio. As the event develops
into an uncontrollable event to danger level 5 where an uncontrollable outflow must
be considered, the water alarm signal signalizes the immediate obligation to leave
the flooding zone.
Emergency management follows the principle of the occurrence of the worst-

case scenario according to a deterministic approach without risk assessment. The
worst-case scenario is defined as the sudden water discharge of the entire vol-
ume, whereby the causes for the occurrence of the event are not differentiated in
the planning of the emergency management. The aim of emergency management,
which is pursued by all parties involved, is to prevent damage to any of its down-
stream inhabitants with a successful evacuation, which means that the time needed
to escape from hazard zone is higher than the time of the arrival of the wave front
(Figure 3). This can only be achieved if A) the threat is known, B) the time course
of the events can be estimated and C) the actors are organized in such a way that
everyone has left the danger area before the event occurs. Meeting these require-
ments requires detailed planning in advance with all those involved in the event of an
emergency. It also requires the installation of a technical infrastructure for detecting
events, communicating the event and alerting the population.
3.2. STRUCTURAL FACILITIES
An essential process for the protection of the population is the early warning
of the population after the operator has assessed the situation, so that the civil
613
Q.105 - R.36
Fig. 4
Siren systems and conditions for triggering
Signal d’alarme et conditions de déclenchement
protection has sufficient time to evacuate the population. This requirement is assured
by nationwide installed sirens, under responsibility of the civil protection authorities.
The siren systems’ principle is presented in Figure 4, their locations are determined
by acoustic modelling. If a region is affected by a flood wave from a very large dam
or a flood wave threatened more than 1,000 people, the siren sound is the sound
of the so-called water alarm. It is different from the sound related to small dams
or installations where fewer people are affected, i.e. the so-called general alarm.
The instructions for action in the case of a general alarm are to wait for instructions
from the civil protection services, which are broadcast on the radio or via an app
(AlertSwiss). In the event of a water alarm sounding, the instructions for action are
to flee the area. Evacuation- or escape plans for this purpose exist nationwide for
each municipality, which are made available to citizens in different ways.
In addition to the sirens installed and maintained by the civil protection author-
ities, the operator must provide other structural facilities that enable him to detect,
observe and communicate critical conditions at the dam. Each dam must be
equipped with a water level measurement device that can be read manually from
the reservoir side. Very large dams also have a copper wire under electrical tension
in their crown, which signals the control centre in the event of a power interruption,
thus indicating that the structure has broken. The water level and other relevant
measurements taken at the dam must be observable from a building next to the
dam, that has to meet strict security requirements.
614
Q.105 - R.36
4. EXAMPLE OF A SPECIFIC EMERGENCY EXERCICE AND

EMERGENCY PLAN IMPLEMENTATION
4.1. SCOPE AND GENERAL OBJECTIVES
The exercise was intended to test the robustness of the emergency regula-
tions and to identify the weaknesses of the general concept. The main elements
with a particular follow up were danger levels evaluations, i.e. making techni-
cal decisions based on the interpretation of measurements and observations,
triggering of water alarm (if water alarm device is present), internal communi-
cation, external communication/verification of communication interfaces, proper
functioning of internal processes, and good management and coordination of
ordered measures. The main goals of the exercise can be summarized as
follows:
– Have a feedback for exercises at other dams;

– Further develop the emergency management;
– Validate the emergency strategies;
– Optimize the coordination between the operating staff, all responsible levels
of dam surveillance, the owner and the authorities;
– Detect the emergency plan limitations and improve the emergency regulation;
– If necessary, adjust the directives.
The exercise was preceded by one day of training for the operating staff,
whereby the objective was to prepare them for emergencies and to understand
the processes , in particular regarding the legal framework, the understanding of
basic and operational documents, the link with the surveillance regulation, and the
general organization and scope of each participant.
4.2. DESCRIPTION
4.2.1. Dam site
The exercise was organized for Cleuson dam, which is a hollow gravity dam
built on the Printze river, located in the canton of Wallis. The dam was commissioned
in 1950 and its main characteristics are summarized in Table 1. In case of failure
with a full reservoir, the flooding area covers about 140’000 inhabitants.
615
Q.105 - R.36
Table 1
Main characteristics of Cleuson dam
Principales caractéristiques du barrage de Cleuson
Dam height [m] 87
Dam base thickness [m] 80
Maximum operating level [m a.s.l.] 2187
Volume [hm3 ] 20
4.2.2. Scenario
The choice of the scenario was the result of discussions during the elaboration
of the regulations. A high magnitude earthquake as a trigger was chosen for the
following reasons:
– In the available guidelines, the case of an “exceptional increase of the water

level” seems to be favored and this could mislead the operator. A new section
was included in the regulations to remove this bias and the implementation of
an event other than exceptional increase of the water level seemed to be an
interesting test of this section.
– Earthquakes are covered by the surveillance regulations which govern the
measures following such an event. The relationship and interaction between
the emergency regulations and the monitoring regulations was therefore
tested.
– Contrary to a flood which has a longer response time, there can be no
anticipation of this type of event.
Table 2 shows the scenario for the exercise and Table 3 explains the effects of
the triggers to the dam. Apart from those who prepared the exercise, no participants
or observers were notified in advance of the content of the emergency exercise and
the precise timing of the triggering events.
In order to focus on operating issues, the participants were limited to the first
three level of dam surveillance and the owner staff, but all actors involved in an
emergency event had at least a representative. The main stakeholder is clearly
Table 2
Main triggers for Cleuson 2018 exercice
Principaux déclencheurs de l’exercice 2018 au barrage de Cleuson
HOUR ACTIVITY
06:49 5.5-magnitude earthquake (mainshock) happens in Verbier (Canton of Valais, 7 km far
from the dam).
07:15 The 24/7 dispatching, is notified of the earthquake by the control room.
09:37 An aftershock happens with a magnitude of 5.
616
Q.105 - R.36
Table 3
Earthquake consequences and expected observations
from the operation staff
Conséquence du séisme et observations attendues
du personnel d’exploitation
TRIGGER CONSEQUENCES AWARENESS CRITERIA FOR
THE OPERATING STAFF
Mainshock Water leakage on the downstream The staff can visually see the leak-
face of the dam age when they arrive on site
Mainshock Infiltration discharge increasing Automatic measurements (seen
from power plan) or manual
measurements on site.
Mainshock Downstream displacement of 2 cm Automatic measurements (seen
for one pendulum from power plan) or manual
measurements on site.
Aftershock Unable to communicate between the non operational device
power plant and the dam
Aftershock Failure of the concrete block which Visual inspection
previously leaks
Table 4
List of people involved in the exercise and their respective roles
Personnes impliquées lors de l’exercice et rôles respectifs
STAKEHOLDER NUMBER COMPANY/ DAM SURV. ROLE
AUTORITIES LEVEL
Civil engineers 2 Alpiq Ltd Exercise staff
Operating staff 6 HYDRO Expl. Ltd Level 1 Participate
Competent civil engineer 1 Alpiq Ltd Level 2 Participate
Civil engineer expert 1 Stucky Ltd Level 3 Participate
Geological engineer expert 1 Norbert Ltd Level 3 Participate
Power generation dispatching 3 Alpiq Ltd Participate
Asset manager 1 Alpiq Ltd Participate
Owner’s representative 1 Alpiq Ltd Observe
Operating staff representative 2 HYDRO Expl. Ltd Observe
Federal authorities 2 SFOE Level 4 Observe
Cantonal autorities 1 State of Valais Observe
Cantonal police 2 State of Valais Observe
the operating staff, including the dam wardens and the Water Alarm Officer. The
exercise stakeholders and their respective roles are shown on Table 4.
4.3. CONDUCT OF THE EXERCISE
The exercise took place on Wednesday, August 29, 2018. A comprehensive

description of all events and actions is available in [6] and only the main phases are
617
Q.105 - R.36
Table 5
Conduct of the exercise: main phases
Phases principales lors de la conduite de l’exercice
HOUR DESCRIPTION DANGER
LEVEL
07h15 First trigger: in order to add a part of the unexpected, it had been announced that 1
the exercise should started at 8 am but the 24/7 Power Generation Dispatching was
informed of the earthquake at 7:15 am. All the persons involved in the dam safety
were then contacted by the dispatching.
8h05 At the Command Center, the operating staff notes the total infiltration discharge 2
increase and the 2 cm downstream displacement of a pendulum on the automatic
measurements. The dam wardens are ready and leave towards the dam. After dis-
cussions with the experts, the Water Alarm Officer announce the increase of the
danger level.
9h03 The dam wardens arrive to the dam, make a dam inspection and a measurement 2
campaign. They confirm the automatic measurements.
9h37 Second trigger: Aftershock! Without any possible contact with the Water Alarm Offi- 4
cer, the dam wardens themselves, on dam site, decide to activate the 4th danger level.
They can call the police center with the secured radio and announce the danger. They
open the bottom outlet.
9h45 At the Command Center, the Water Alarm Officer, without any news from the dam, 4
decide to send somebody to the police center. After discussion with the experts, he
decides to activate the 4th danger level.
TEMPORAL BREAK
10h35 When the lake level is low enough and the classic communication channels are 2
operational, the Water Alarm Officer decide, after discussion with the experts, to
lower the danger level.
10h38 End of the exercice
described in Table 5. The exercise was broadly conducted in accordance with the
forecast scenario. All participants were well engaged into the scenario and acted
with seriousness and professionalism.
During the exercise process, it was noted that as soon as the Power Generation
Dispatching alerted the Operating Staff, the Water Alarm Officer and the Operations
Staff went directly to the danger level 2, which was not necessary at this time. This
behaviour could be explained by the fact that everyone was aware of the existence
of the exercise and wanted to run to a higher level of danger.
It is also noted that if all participants had expected the first trigger, they were
surprised by the seismic aftershock. At that time, they had been pushed to their limit
and their behaviour was close to the one expected for real events.
4.4. OPTIMIZING THE EMERGENCY REGULATIONS
The exercise was analyzed by each participant and the feedbacks allowed
identifying several possible improvements. This paragraph summarizes the main
practical improvement identified for the existing regulations of Cleuson dam.
618
Q.105 - R.36
In case of an emergency event, the most important communication channel is

the one between the Operator and the Police Center for the announcement of a new
danger level. During the exercise, the exact content of the notification was difficult
to establish by the Operator and the proper understanding of the message could be
affected. It is therefore recommended that the announcement of the level of danger
at the Police Center must be accompanied by a predefined message.
For reasons of legal liability, the semantics of the content of the Decision
Support Sheet “Determining the danger level based on the state of the dam” could
be problematic. Indeed, the definition of the thresholds which determinate the danger
level could backfire on the dam wardens if they have not triggered the appropriate
danger level. The content of the Decision Support Sheet should therefore be stated
to be only indicative.
The consequences of the full opening of the bottom outlet during a long period
of time are unknown. Especially for high mountain dams, such an event has not
occured for several decades and the rivers are not used to have this discharge.
A flooding map including recent constructions and the discharge capacity of the
bottom outlet should be considered in the Emergency Planning.
Eventually, a detailed description (plans and explanations) of the communica-

tion system is necessary and should be subjected to a risk analysis.
5. LESSONS LEARNED FROM THE MAIN STAKEHOLDERS VIEW
This chapter provides a first assessment of the emergency regulations update.

The three main stakeholders, i.e. the Operator, the Civil Protection Authorities and
the Supervisory Authorities, are represented. Because each part has a different
scope, the three different points of view allow having different perspectives and
keeping a good overall view.
5.1. FROM THE DAM OWNER AND THE OPERATING STAFF
According to the Swiss Law, the owner is responsible for the surveillance, the
security of the dam and all the possible consequences on which the dam is involved.
First, he must guarantee that everything is done to ensure that the population is not
impacted negatively by the dam. In case of emergency, a small negligence can have
major effects and if any damage is notified, the Operator has to prove his good faith.
The objective of the exercise was to test the applicability of a new regulation
involving many actors and requiring a precise organization. It turned out that the
implementation of the Emergency Regulations as established by the Operation Staff
619
Q.105 - R.36
and the Owner was working efficiently and that the actors involved took the measures
of the regulations in a very professional and applied way.
The exercise, rather ambitious in its scenario, went perfectly and each step
could be analysed, which allowed for a rich feedback. The few recommendations
help to improve the following exercises and also to adjust the basic documents. The
best lesson learned would be the importance to carry out a well-designed emergency
exercise.
5.2. FROM THE CIVIL PROTECTION AUTHORITIES
The main protection measure for the population in the event of a major event
on a dam is the evacuation of the danger zone. This measure having a major impact
and implying a certain time period for its realization, it is essential that the interface
between the dam operating staff and the authorities takes place as far upstream as
possible in order to maximize the time remaining at the disposal of the authorities
to carry out the evacuation of the danger zone.
The regular awareness of the population to the alarm signals is part of the
preparatory measures. Every year, a full-scale test of the sirens is carried out, allow-
ing the population to be familiar with the general alarm and water alarm signals. For
the past two years, siren activation has been coupled with a national alarm app
(AlertSwiss), which allows the spreading of information and instructions also on the
cellular phones of people in the concerned zone.
Communal and cantonal authorities, at their own level, regularly carry out the
evacuation of the population as a transversal module of crisis management, spe-
cific to several problems, such as floods, torrential flows and avalanches. Crisis
management exercises with dams allow, in addition to their formative vocation, to
refine in even greater detail the critical points in the management of such an event.
It can be mentioned the reciprocal knowledge of the teams and their way of work-
ing (staff, operating rhythm…) but also the identification of small practical details,
apparently insignificant but which can quickly turn out to be problematic. The best
way to make these disruptive factors appear is obviously to make the situation as
true as possible to reality during exercises. The lessons learned from emergency
exercises largely relate to these factors that are the early interfacing of the right
people and functions, the various possible constraints to the right communication
flow as well as those imposed by the need to have a synchronized communication.
Equivocal information or simply information given at an inappropriate moment can
have an unfortunate influence on the smooth running of an evacuation. The emer-
gency management specific documentation must be as simple and comprehensible
as possible. The use of checklists seems to be efficient. In summary, the quicker
the interface between the dam and the public authorities is established in the event
of a rise in the level of danger, the greater the chances of reducing the impact on
the surrounding population and its livelihood. As such, future exercises should focus
620
Q.105 - R.36
on the means and pace of communication between the different teams involved in
crisis management, both internally and with the authorities.
5.3. FROM THE SUPERVISORY AUTHORITIES
The exercise has shown that only an exercise reveals the deficits of theoretical
emergency planning. Therefore, the supervisory authority considers the mandatory
execution of exercises to be justified. It has been shown that the organization of the
monitoring of a dam, in a system with 3-4 monitoring levels (depending on its size),
can lead to bottlenecks in case of an incident. Often one expert is responsible for
several dams in one area. In case of an event affecting several dams in an area,
the expert does not have enough time to evaluate the events for more than one
dam. One possible solution would be to appoint a second expert who can work
redundantly in case of an event.
The impact of an event is assessed based on a flood wave map, which is

prepared for each dam. This map assumes that the worst-case scenario occurs,
namely the sudden and total rupture of the barrier. In the exercise, a planning of the
measures to be taken in case of a partial failure was sought, but the flood wave of
such an event is not available. The supervisory authority has recognized this deficit
and intends to have partial failure calculated in addition to the total rupture in the
next update of the available flood wave maps.
Legislation and the associated guidelines require that the operator must main-
tain redundant means of communication. A functioning communication is necessary
for the correct information about the danger levels and thus the triggering of the right
measures to protect the population. The exercise has shown that as a further instru-
ment, the possibility of camera control would maybe be useful if redundant means
of communication also fail. The supervisory authority does not yet prescribe this to
operators but is considering a more detailed examination of this finding.
Furthermore, it has been shown that a very early inclusion of all potentially
involved parties makes sense, i.e. directly after the increase of the danger level from
normal operation to the next danger level. This is the only way to guarantee a good
preparation in case of a development of the event.
REFERENCES
[1] Federal Water Retaining Facilities Act of 1 October 2010 (SR 721,101).
[2] Federal Water Retaining Facilities Ordinance of 17 October 2012

(SR 721.101.1).
621
Q.105 - R.36
[3] Swiss Federal Office of Energy: Directive on the Safety of Water Retaining
Facilities Part E: Emergency plan (version 2.0), 2015.
[4] Geschichtliches zum Wasseralarm, Dr. R. Biedermann und Bundesamt für

Bevölkerungsschutz, Bern 2004.
[5] Swiss Federal Office of Energy: Recommendation for the design of emergency
exercises according to the directive on the safety of dams Part E Emergency
plan (2015), (version 2.0), 2017.
[6] Barrage de Cleuson, Formation et Exercice en cas d’urgence 2018, Rapport

complet, Alpiq, Lausanne, 8.10.2018, not published.
622
Q.105 - R.37
-------
-------
EMERGENCY PLANNING AND DAM FAILURE MANAGEMENT: STATE OF

PRACTICE IN SWITZERLAND, AUSTRIA AND GERMANY(∗)
Alexandra BECKSTEIN
Swiss Federal Office of Energy SFOE, Supervision of Dams,
Switzerland
Bettina GEISSELER
GEISSELER LAW, Freiburg i.Br., Germany
Burkhard RÜDISSER
Federal Ministry of Agriculture, Regions and Tourism,
Directorate General I –Water Management, Austria
SWITZERLAND
SUMMARY
The paper compares the practice of dealing with the residual risk of dam safety
in Austria, Germany and Switzerland. The residual risk, which remains despite com-
pliance with the required standards of safety and monitoring, is achieved through
the implementation of suitable emergency concepts, which are more or less dis-
tinctive in the respective countries. Further comparisons between the countries are
made for the handling of public safety and security. The measures taken by different
countries in the areas of Emergency Planning, Public Safety and Security are very
different and their implementation in legal regulations varies widely. The authors are
convinced that even in three countries with similar governmental structures and sim-
ilar dam safety philosophy and safety approaches many lessons can still be learned
∗ Planification d’urgence et gestion des ruptures de barrage : état de la pratique en Suisse,
Autriche et Allemagne
DOI: 10.1201/9781003211808-83 623

Q.105 - R.37
from each other. All players such as the dam owners, the state authorities/ agen-
cies or other experts should make an effort to further discuss all relevant questions
related to Emergency Planning by taking into account and balancing the interests
and needs of all parties concerned. The setting up of a working group within ICOLD
might be an appropriate way.
RÉSUMÉ
Le rapport compare la pratique en matière de traitement du risque résiduel de

la sécurité des barrages en Autriche, en Allemagne et en Suisse. Le risque résiduel,
qui subsiste malgré le respect des normes de sécurité et de surveillance requises,
est obtenu grâce à la mise en œuvre de concepts d’urgence appropriés, plus ou
moins spécifiques à chaque pays. D’autres comparaisons entre les pays sont effec-
tuées pour le traitement de la sécurité et de la sûreté publiques. Les mesures prises
par les différents pays dans les domaines de la planification d’urgence, de la sécurité
publique et de la sûreté sont très différentes et leur mise en œuvre dans les régle-
mentations légales varie considérablement. Les auteurs sont convaincus que même
dans trois pays ayant des structures gouvernementales similaires et une philoso-
phie et des approches similaires en matière de sécurité des barrages, de nombreux
enseignements peuvent encore être tirés des uns et des autres. Tous les acteurs,
tels que les propriétaires de barrages, les autorités publiques ou d’autres experts,
devraient s’efforcer de discuter plus avant de toutes les questions pertinentes liées
à la planification d’urgence en prenant en compte et en équilibrant les intérêts et les
besoins de toutes les parties concernées. La mise en place d’un groupe de travail
au sein de la CIGB pourrait être un moyen approprié.
1. INTRODUCTION
Dams have been built since the beginning of the civilization and contributed to
the social and economic development of the mankind. But – on the other hand – they
can cause, as every technical structure, harm to people and third parties’ property.
Whereas Dam Safety considerations always have been taken into account in some
way while constructing and erecting dams, the public awareness and a resulting
discussion of the remaining risks related to dams and thus the discussion on Dam
Safety and Emergency Planning have become increasingly important during the last
decades.
Recent incidents such as the Oroville dam’s main spillway failure (California,
USA) in 2017, the Xe Pian-Xe Namnoy (Laos) dam failure in 2018, or the incident at
624
Q.105 - R.37
the Toddbrook Reservoir, Derbyshire (Great Britain), in 2019 show the importance
of a well-functioning planning in case of a safety-related events at a dam.
This paper discusses how the risks are dealt with in Austria, Germany and
Switzerland. Following common practice, the issues of Dam Safety, Public Safety
and Security and their influence on third parties are dealt with. In this document,
“Dam Safety” is defined as the technical safety of a dam as it is regulated and
supervised by the authorities in all countries presented here. According to this defi-
nition, the objective of Dam Safety is to prevent a failure of a dam and in particular
the uncontrolled release of large water masses or to minimise their consequences.
“Emergency Planning” intends to handle the remaining risks and is one part of the
Dam Safety concept.
In this document, “Public Safety” refers to all threats to third parties arising
from the presence, including the operation, of the power plant complex. These can
be e.g. drop risks, but also surges and sunk, failure of pressure pipes, and others.
The term “Security” means the protection of the asset dam against any kind of
external threats and criminal acts potentially damaging the plant. This includes both
physical intrusion and intrusion into computer systems.
2. HISTORY OF DAM SAFETY PRACTICE
2.1. AUSTRIA
In 1918, after the catastrophic failure of the flood retention basin “Weisse
Desse” in Northern Czech Republic, former territory of the Austro-Hungarian Empire,
the Austrian Commission on Dams (“Staubeckenkommission”) was founded. Its pur-
pose was to support the authorities and plant owners in planning and approval
procedures. The Commission on Dams should consist of experienced experts of
all fields of engineering, which are relevant for the design, the construction and
the surveillance of large dams. From the beginning, the intention of the legisla-
tor was to create a committee in order to make collective and interdisciplinary
solutions. In 1934, the Commission on Dams was incorporated into the Austrian
Water Law. Therefore, projects involving the construction of a dam with a maximum
height exceeding 15 m or an Impoundment Volume larger than 500.000 m3 must be
approved by the Commission on Dams before construction.
After the Malpasset dam failure in 1959 and the Vajont catastrophe in 1963
the permanent Subcommittee Dam Supervision (“Unterausschuss für Talsper-
renüberwachung”) was established in 1966. This subcommittee is consisting of a
team of experts of each relevant field of dam engineering, in order to inspect large
dams in Austria periodically. In the Commission on Dams’ history, so far 106 meet-
ings were held, discussing projects of 103 dams for the production of hydroelectric
energy, 37 flood retention basins and 53 reservoirs for snowmaking. Additionally,
625
Q.105 - R.37
the Austrian Commission on Dams is responsible for the creation and actualiza-
tion of guidelines specifying the state of the art regarding dam construction and
surveillance.
2.2. GERMANY
For long-time a (remaining) risk-based approach regarding the evaluation

of the Dam Safety in Germany did not play an important role in the (public) dis-
cussion. [1] In Germany it always has been good engineering practise to assess
all risk aspects related to Dam Safety and take them into account during the
engineering, construction and erection phase and later during operation and main-
tenance. But even though of course Dam Safety rules always have been observed
in dam construction, operation and maintenance, the publicly communicated com-
mon understanding was, that dams are safe in the reasonable judgement of experts
(“nach menschlichem Ermessen”).
Only in 2004, with the revision of the technical norm DIN 19700 [2], which
reflects in Germany the generally accepted technical standard for dam construc-
tion, maintenance and operation, the subject of considering the consequences of
the remaining risk became part of the German legislation. DIN 19700 takes the
risk-based assessment e.g. into account by stipulating that despite their height
and/or storage volume dams can be classified into class 1 (large dams) or class
2 (smaller dams) according to their hazard potential. According to clause 11 of DIN
19700-10 the remaining risk resulting from the exceeding of the design flood BHQ2
must be assessed and to sufficiently be minimised by supplementary constructive,
operational and/or organisational measures.
Even though none of the dams classified as large dams according to the Ger-
man standard in clause 3 DIN 19700-11 has broken after the Second World War,
the potential damage would probably be enormous in case of the break of a dam in
Germany.
In the last decades it has as well in Germany become a generally accepted

principle that by evaluating the safety of new or existing dams the remaining risk has
to be taken into account and should be monitored in a transparent way. One of the
strategies to minimise the remaining risk is the Emergency Planning. Like in other
countries Dam Safety in Germany is nowadays based on the following three pillars:
• Construction, erection, and operation of the dam in accordance with the proven
(generally accepted) technical standards
• Continuous monitoring and surveillance of the dam as part of an adequate
safety management during the whole lifecycle of the dam
• Emergency concepts.
In Germany, there does not exist one law governing all safety aspects, includ-
ing emergency response aspects in connecting with potential risks resulting from a
dam. The principles of Emergency Planning are mainly governed by technical norms
626
Q.105 - R.37
reflecting the generally accepted technical standard and different laws of the Ger-
man federal states. In absence of a stringent detailed regulation and a centralised
supervising state authority each dam owner/ operator has a considerable discretion
on how to assess the risks and organise himself in order to assume his responsibility
regarding Emergency Planning.
2.3. SWITZERLAND
The history of Swiss emergency planning for dams is closely linked to the
history of the Second World War. In 1945, out of concern about an attack on the
dams, the Swiss government, on the basis of a federal decree of 1943, began to
install a nationwide water alarm siren network, which was intended to call on the
population to flee in the event of an attack and which is still a fundamental part of
the emergency plan today. Today, the water alarm system is regarded as a decisive
precondition for the further development of the security concept to its current extent.
[3] A civil legal basis for the implementation of the emergency concepts was only
created later, in 1954 (the dam ordinance was specified in more detail when it came
into force in 1957), and then only adapted in 2010 to the legal basis that is still valid
today. Even if the legal basis and the technical systems have developed further,
the emergency concept existing today, including the technical systems is clearly
designed for the case of war. Whether and how the system can be redesigned
to focus more on modern hazards such as natural hazards, which are becoming
increasingly common due to climate change, or attacks on computer systems.
3. REGULATORY FRAMEWORK
3.1. AUSTRIA
Water rights authority with notice of application of the Supreme authority for
large dams with the overall maximum height >30 m or an impoundment volume
>5 Mio m3 is the Federal Minister of Agriculture, Regions and Tourism. For smaller
dams the supreme authority is the governor of each province. Construction, oper-
ation and monitoring of dams are subject to the regulations of the Austrian Water
Law (WRG 1959, as amended). It stipulates, among other topics, that: (simplified)
• §12a (1) (…) facilities and operation must be State of the Art (…)
• §12a (2) (…) State of the Art can be set by the Federal Minister per decree
(…)
• §21a (1) (…) If facilities or operations are not corresponding to the State of the
Art, the Federal Minister can set constraints or prohibitions (…)
• §23a (…) For dams, reservoirs (…) with H > 15 m and/or V| > 500.000 m3 a
responsible and qualified person (Dam Safety Engineer) has to be appointed
627
Q.105 - R.37
and reported to the Federal Ministry and the local authorities. (…) This person
must be part of the senior technical management of the company who operates
the dam or exceptionally of a different consultant if this person is qualified
(Master Degree in Civil or Environmental Engineering) and familiar with the
dam (…)
• §104 (3) For the approval of dams and reservoirs higher than 15 m above
foundation level or an impoundment volume of more than 500.000 m3 an expert
opinion of the Austrian Commission of Dams must be obtained.
• §105 (2) Public interest: (…) Approval procedures must contain measures
concerning incident and failure management (…)
• §131 (1) (…) the Federal Minister …can monitor dams regarding the stability
and operational safety of dams higher than 15 m above foundation level or
impounding reservoirs with a capacity of more than 500.000 m3 due to the
Commission on Dams periodically, but the intervals must not exceed 5 years.
(…)
• §134 (7) (…) Insofar as this appears necessary in the interest of public safety,
the Water Rights Authority with notice of the §§23a and 131 (1) (…) the Federal
Minister can also monitor dams smaller than 15 m or with a capacity less than
500.000 m3 (…).
Mutual inspections of large dams (height >15 m, impoundment volume

>500.000 m3 ) are carried out by the Subcommittee Dam Supervision of the Federal
Ministry, the Province Supervisor and the plant owner in different intervals:
Fig. 1
Intervals of mutual inspections
628
Q.105 - R.37
Concerns about emergency preparedness are implemented in the Federal

Environmental Information Act (UIG), specified in a regulation (ordinance) on
Accident Information (both transposing EU directive 2012/18/EU).
Therefore, the owner of a dam which height exceeds 15 meters above the
foundation level or an impoundment volume exceeding 2 million m3 , has to provide
appropriate information about the dangers and effects of serious accidents and the
necessary action to be taken in the event of a major accident at regular intervals, not
exceeding five years. Flood maps are created by the dam owner, assuming a worst
case- scenario of a total dam failure. Thereupon the disaster management authority
of each Austrian province provides hazard maps and evacuation scenarios, which
are published.
Additionally, guidelines of the Austrian Commission on Dams stipulate, that

flood and hazard maps and evacuation scenarios should be provided also for smaller
dams, which have a high hazard potential (Fig. 2 [4]).
Fig. 2 defines the hazard potential for potentially effected people and
infrastructures according to the flood map.
Fig. 2
Hazard potential due to a flood wave caused by dam failure [4]
Concerning public safety and security there is currently no specific legal frame-
work existing in Austria. The matters are dealt with in internal guidelines of plant
owners or partly in training courses.
629
Q.105 - R.37
3.2. GERMANY
Until 2010 there was no unified water law in Germany, but each federal state
had its own Water Act. By completely revising and re-issuing the Water Resources
Act (“Wasserhaushaltsgesetz”/“WHG”) [5] in 2009, which came into force in 2010,
the federation of Germany made for the first time use of its competing legislative
competence in the field of water and inserted a regulation governing dams and
reservoirs. That means: the concurrent water laws of the different federal states
which existed long-time before the revision of the WHG remain in force, but might
have to be adapted to the new legislation of the federation. Only as far as the Water
Resources Act does not regulate a certain aspect, this aspect can be governed by
the water laws of the federal states.
According to §36(2) of this Act dams are to be designed, planned, built

and operated according to the generally accepted technical rules and standards
(the “state of the art” requirement), and if already erected dams do not fulfil this
specification, they have to be rehabilitated accordingly.
In Germany, DIN 19700 with its different parts (e.g. DIN 19700-10 or 11)
are considered to be the generally accepted technical rules and standards. Unlike
the former regulations of the federal states’ Water Acts, DIN 19700 applies to any
dam regardless its height/storage volume and/or classification. Only some specific
technical or operational requirements are related to the classification according to
§3 of DIN 19700–11.
The applicable standards and guidelines on dams and reservoirs include spec-
ifications of load combinations, safety coefficients, design principles (e.g. freeboard),
operational rules for important structures (for example spillway and outlet gates) and
follow the mentioned risk-based approach.
§8 of the Water Resources Acts requires a state permit for the erection or
rehabilitation of a dam, as this activity is considered to be a use of the resource
water. This single permit is granted in a concentrated administrative procedure in
which all legitimate aspects and stakeholders’ interests, which according to the law
have to be considered such as the Environmental Impact Assessment, are taken
into account.
3.3. SWITZERLAND
Several parties are involved in the emergency planning and its implementa-
tion. These are the operator and its commissioned experts, the federal supervisory
authority for the safety of the dams, and the federal and cantonal authorities
responsible for protecting the population. All parties involved in the planning and
implementation of emergency planning are based on the legal foundations laid down
630
Q.105 - R.37
Fig. 3
Overview of responsibilities in emergency planning and in the event of an incident
in Switzerland according to the regulatory framework
by the two federal agencies that are responsible, on the one hand, for the safety
supervision of the reservoirs (Swiss Federal Office of Energy) and, on the other
hand, for the protection of the population (Federal Office for Civil Protection). The
legislation on the safety of dams defines all measures to be taken by the operator in
order to prevent and manage an event, while the legislation on civil protection and
protection of the population defines all measures and technical equipment required
by civil protection authorities and operators in order to fulfil these tasks. The legisla-
tion on dam safety consists of the Dams Act, the Dams Ordinance and the Guidelines
on the Safety of Dams Part A to E, whereby Part E specifies the requirements for
emergency planning. Even though the guidelines are not legally binding, they rep-
resent the state of the art and the supervisory authority requires the implementation
of the requirements summarised in them. The implementation of safety supervision
is based on the Swiss Federal Act on Water Retaining Facilities, the Ordinance on
Water Retaining Facilities and the associated guidelines in Parts A-E. For the imple-
mentation of the requirements of the legal basis, the supervisory authorities, which
are either the federal government or the responsible cantons, depending on the size
of the facility, are both the federal and cantonal offices for civil protection matters.
4. DAM SAFETY
Dam Safety comprises the technical safety of the dam and its safety related
auxiliary installations. The aim of Dam Safety is to prevent uncontrolled water
631
Q.105 - R.37
leakage or to minimise its consequences. The practices of Dam Safety in Germany,

Austria and Switzerland are based on the internationally established safety concept,
which is based on three pillars. The first pillar is the obligation to design and build
the dam according to the state of the art in science and technology. The second
pillar describes the principle of regular monitoring of the dam in order to detect and
remedy possible deficiencies in technical safety at an early stage. The third pillar is
the control of the residual risk, i.e. it includes the planning of measures to be taken
if the first two pillars of the principle fail. In the following, the planning for dealing
with this residual risk, the Emergency Planning, in all three presenting countries is
described.
4.1. AUSTRIA
4.1.1. Philosophy and aim
In general, due to a high level of security awareness, high standards are imple-
mented regarding design, construction and monitoring of dams. Despite this high
level of security, every dam construction is a unique and complex system and- like
any technical structure- has a certain, extremely low probability of failure, which has
to be considered. According to [6], the overall annual risk of failure of dams older
than 5 years can be quantified as approximately 5 ∗ 10−5 . Therefore, the third col-
umn Emergency Preparedness and Failure Management is implemented in the dam
safety philosophy. The aim is to prepare for the worst possible (realistic) scenario.
The assumption is a total failure of the dam operating in normal conditions (Reservoir
at maximum operational water level assuming no additional flood scenarios, earth-
quakes, etc.). Risk assessment or the possible causes of failure are not considered
in the process. As described above, due to the Environmental Information Act and
guidelines of the Austrian Commission on Dams, flood mapping is obligatory for all
large dams (height >15 m, impoundment volume >500.000 m3 ). However, for each
dam with a significant hazard potential (Fig. 2) flood maps must be created by the
owner [4].
4.1.2. Technical concerns
In order to estimate a potential flood wave, the initial situation is a dam oper-
ating in “normal” conditions with its water level on the maximum operational height.
The hydrograph of arch dam failure is calculated by the sudden removal of the whole
dam cross- section. Failure of gravity dams is modelled by the sudden removal of
the three highest blocks. The failure of embankment dams is modelled with the
assumption of a breach at the highest cross section or the relevant cross section,
which causes the most potential harm to the downstream infrastructure. Empirical
632
Q.105 - R.37
models, such as e.g. Froehlich (1995) [7] are common. Also, more complex 2D and
3D Models are in use. However, using flood wave prediction models in engineering
practice in order to fulfil government requirements related to safety issues, it is nec-
essary to set equal standards. The application of process-based models is required
to define parameter values which impact on the high complex processes is currently
not fully understood. Within the possible range of feasible values, with some models
it is possible to “design” a flood wave, rather than calculate it. Empirical models are
strongly dependent on their often inhomogeneous and scarce database; the results
of commonly used equations display a large range. [8] With the dam breach hydro-
graph as initial condition, flood maps are created by using specific 1D- 2D hydraulic
models, which are suitable for the downstream terrain. Secondary processes, like
e.g. debris flow, woody debris, erosion, etc. must be considered. Preload in the
recipient is recommended with HQ1. [9].
4.1.3. Emergency planning and disaster management scenarios
The creation of emergency plans (“Sonderalarmpläne”) and disaster man-

agement scenarios is in the responsibility of the disaster management authority
(Katastrophenschutzbehörde) of each province. In the event of a catastrophe, the
regional province authority (Bezirkshauptmann) is in charge of disaster manage-
ment and therefore executing the relevant emergency plan for each scenario. The
emergency plan for a dam failure scenario is based on the flood map. It is created
under participation of all relevant emergency services as well as local and regional
authorities. Currently, there is no nationwide standard existing in Austria.
4.2. GERMANY
4.2.1. Erecting and operating a dam
As mentioned above, §36 WHG requires that dams must be erected and oper-
ated according to the actual generally accepted standard. The future owner of a dam
has to prove the “Zuverlässigkeit” (reliability) of the dam, which means, (clause 7 of
DIN 19700–11) that the dam
• can resist even extreme events, in particular flood events which might overtop
the dam or severe earthquakes,
• is designed and built for a long durability – a condition, which requires a regular
maintenance and
• is fit for its intended purpose (operation).
633
Q.105 - R.37
The DIN 19700–11 indicates in detail under which conditions the dam is con-
sidered to comply with the requirement of the reliability (e.g. suitable construction
materials, adapted calculation and dimensioning methods, cf. cl. 7.1.1.).
4.2.2. Monitoring and supervising
In view of their long life cycle the operator is obliged to install a safety man-
agement. The norms leave it at his discretion in which way he wants to supervise
his dam. According to common practise the safety management is threefold:
• the dam operator must continuously check the safety of his dam
◦ by testing the functioning of critical structural components such as the
spillways or the bottom outlet,
◦ by visual control and
◦ by checking the data transferred by the control instruments installed
inside or along the dam.
The operating team is responsible for supervising the dam.
• The engineer being in charge of the safety issues for the dam establishes
once a year a so-called safety report in which he assesses – based on the
data collected during the year – the safety of the dam. In particular, he has
to document and assess cases of occurred malfunctioning of critical compo-
nents and to evaluate the long- time behaviour of the dam. This report will be
submitted to the competent authorities.
The technical standard DIN 19700 does not prescribe in detail the structure
and content of this safety report. The DVWK Merkblatt 231 [10] indicates a
possible structure and content in order to unify these safety reports.
Depending on the classification and/or hazard potential the dam will be
inspected by the owner/operator together with the state authority once every
1 respectively 2–5 years. Special attention will be paid to components where
the safety report determines irregularities.
• Every 10–20 years an in-depth safety inspection takes place. This is initiated
by the owner of the dam, but usually performed by external experts engaged
by the owner. Construction materials such as concrete, the asphaltic core
or the hydraulic steelworks will be examined in detail by taking into account
hydrological data. In particular deviations from the original calculation for the
dimensioning of the dam will be assessed. It will be evaluated whether the
dam still complies with actual dimension and other requirements or has to be
adapted to new regulations in place.
4.2.3. Emergency planning
In Germany there is not one law governing all safety aspects, including emer-
gency response aspects in connecting with potential risks resulting from a dam.
634
Q.105 - R.37
The subject of Emergency Concepts and Planning today is mainly governed by the
mentioned DIN 19700, part 10 and 11, the laws of the different states regarding
emergency/disaster situations (Brand- und Katastrophenschutz-gesetze der Bun-
desländer), applicable ICOLD Bulletins and other technical guidelines. With the
revision of the WHG in 2009 the EU directive 2007/60/EC on the assessment and
management of flood risks has been integrated into national law; the water laws of
the different federal states stipulate the resulting obligations of the parties concerned.
According to clause 15.1 of DIN 19700-10 the owner and operator of a dam
is obliged to integrate instructions for an emergency case into his operation plan.
This operation plan has, among other documents, to be submitted to the compe-
tent authority when applying for the permit to erect a new dam or the continue to
operate an existing dam under the applicable water law WHG. In this context the
owner/operator of the dam has to evidence e.g. an adequate organisational struc-
ture (an organigram showing the responsibilities of the personnel of the emergency
team and their accessibility e.g. by mobile phone) and disclose the reporting and
alert chains.
Alarming the potentially affected population will nowadays be less done by

siren alarm, but rather by messages on the mobile phone or radio/TV. It depends
on the laws of the different federal states whether and to which extent the dam
owner/operator has to directly alert the downstream population potentially affected
by emergency events such as overtopping of the dam or mal-functioning/failure of a
gated spillway or whether this will be done by the competent state authorities (civil
protection agency, fire brigade etc.).
Though most of the emergency laws of the German federal states stipulate
the obligation that an owner of an installation with a considerable risk potential has
to support the state emergency response and civil protection authorities it seems
that so far there is not yet a common understanding to which extent this has to be
done – e.g. by providing of flood area maps only upon request or by an unsolicited
handing over to the state authorities.
In general, it can be said that Switzerland and other countries have much more
detailed and stringent legal regulations regarding Emergency Planning [11] in the
context of Dam Safety than Germany.
4.3. SWITZERLAND
4.3.1. Aims and actors of emergency management
Emergency management follows the principle of the occurrence of the worst-

case scenario according to a deterministic approach without risk assessment. The
worst-case scenario is the sudden water discharge of the entire volume, whereby
635
Q.105 - R.37
Fig. 4
Context of early warning, alarming and evacuation
the causes for the occurrence of the event are not differentiated in the planning of the
event management. The actors involved in the event management process are the
operator and the cantonal civil protection authorities. The operator and the federal
civil protection authorities are involved in planning the management of the event.
The aim of emergency management, which is pursued by all parties involved, is
to prevent damage to any of its downstream inhabitants or objects. In Switzerland,
emergency management is based on the principle of assessing the possible impact
of an event at a very early stage and assigning it a hazard level, the subsequent
measures of which are precisely defined. This is preceded by a detailed prepara-
tion plan, as well as the equipment with the appropriate infrastructure for detecting
events, communicating the event and alerting the population.
An essential process for the protection of the population is the early alerting
of the population after the operator has assessed the situation, so that the civil
protection has enough time to evacuate the population.
4.3.2. Sirens and alarm systems in Switzerland
The alarm is triggered by a system of sirens installed nationwide. The Civil

Protection is responsible for the installation and maintenance of these sirens. [12]
The siren locations are determined by acoustic modelling. If it is a flood wave from a
very large dam or a flood wave that would threaten more than 1,000 people, the siren
sound is the sound of the so-called water alarm, which is different from the sound of
small dams or installations where fewer people are affected. Here the sound of the
so-called general alarm is used. The instructions for action in the case of a general
alarm is to wait for instructions from the civil protection services, which are broadcast
on the radio or via an app. In the event of a water alarm sounding, the instructions
636
Q.105 - R.37
for action are to flee the area. Evacuation plans for this purpose exist nationwide for
each municipality, which are made available to citizens in different ways.
4.3.3. Danger levels
At what point in time the sirens are triggered depends on the current danger
level. In Switzerland, analogous to natural hazards, 5 danger levels are defined for
a safety related event of a dam, whereby danger level 1 describes normal operation
and danger level 5 the imminent or uncontrolled outflow of water from the reser-
voir. Each operator must determine the hazard levels depending on the expected
course of events and communicate them in the event to all those involved in emer-
gency management and to the supervisory authority. Close cooperation and mutual
coordination of measures is essential.
4.3.4. Emergency preparation: emergency regulations
The planning for an incident has to be prepared by the plant operators together
with the persons responsible for civil protection and has to be submitted to the super-
visory authority for review. In addition, the plans must be regularly exercised and, if
necessary, adapted. The planning in the event of an incident must be documented in
an emergency response regulation. The contents of these regulations are defined
and are more or less comprehensive depending on the size and purpose of the
water retaining facility. The contents that must be included in the emergency regula-
tions are a flood map that shows the flooded areas in the event of a sudden breach.
Furthermore, a risk analysis must be carried out to clarify which influences could
impede emergency management. In an emergency strategy it is defined which mea-
sures are to be taken per danger level. The organisation of the responsible persons
including their redundant availability in case of an incident must be clearly defined.
The summarised information is then recorded in the form of an emergency dossier
which is distributed to all persons involved. [13]
4.3.5. Components and structural facilities of the water alarm system
In addition to the civil protection sirens, the operator must also provide struc-
tural facilities that enable him to detect, observe and communicate critical conditions
at the dam. [14], [15] Each dam must be equipped with a water level measurement
device that can be read manually from the reservoir side. [13] Very large dams also
have a copper wire under electrical tension in their crown, which signals the control
centre in the event of a power interruption, thus indicating that the structure has bro-
ken. The water level of any large dam must be observable from a building next to the
dam. The requirements for the building vary according to the size of the dam. [13]
637
Q.105 - R.37
5. PUBLIC SAFETY
“Public safety” refers to all hazards to third parties arising from the presence of
the power plant complex. These are very different in nature and can be fall risks from
accessible parts of a dam, rapid changes in flow regime and currents downstream
of a dam (due to flushing), eddies and currents upstream of a dam (due to draining
and water extraction), rapid flow changes along waterways (due to flushing of sand
traps, failure of pressure pipes, occurrence of natural hazards such as rockfall along
public roads around the dam, hazards from locks and boat transfer points
5.1. AUSTRIA
Public Safety is covered by general laws, but it is not a special issue. Currently,
there are no specific acts or guidelines concerning Public Safety existing in Austria.
However, Public Safety is common practice and in the responsibility of the dam
owner. Technically, Public Safety is implemented by the dam owner e.g. by fencing,
signage, life belts as well as sirens and light signals. Training courses and exercises
are also performed as well as the distribution of public information. Currently, there
is no nation-wide standard for those measures. A standard requirement for projects
passing the appraisal of the Austrian Commission on Dams stipulates that for any
dangers that arise from the operation of the dam, appropriate precautions must be
taken to inform or warn the public.
5.2. GERMANY
Under German law any owner/operator of an installation with a hazard poten-

tial is responsible for the Public Safety and has to ensure that third parties will not be
damaged by the existence and/or operation of the installation. This is a general prin-
ciple resulting from the Civil Code and developed in detail over the decades by the
German jurisdiction. Preventive measures are e.g. the installation of warning signs
around the reservoirs respectively downstream the dam indicating that it might be
dangerous to swim in the reservoir or that the water level downstream may suddenly
rise even during normal operation of the dam and its installations.
5.3. SWITZERLAND
In Switzerland, the operators are responsible for the areas mentioned above.
There is no federal authority to supervise compliance with any regulations. The
applicable legal provisions are governed solely by the Swiss Code of Obligations
and plant owner liability.
638
Q.105 - R.37
6. SECURITY
The authors define “Security” as the protection against unauthorised intrusion

into a power plant complex with the aim of avoiding actions that could endanger third
parties and the environment. It concerns both physical intrusion and intrusion into
computer systems (cyber-attack).
6.1. AUSTRIA
Security is the responsibility of the dam owner. Every unauthorised access of

facilities, which could cause danger to the facility operation or to the public, has to
be prevented. Measures, like e.g. fencing, cameras, motion sensors are common.
There is currently no nation-wide standard in Austria. However, measures preventing
unauthorised access of relevant dam facilities like operation devices or control rooms
has to be included in the operating manual, which is part of the authority approval
procedure. Therefore, the Austrian Commission on Dams is providing a guideline
concerning the basic outline of the operating manual. [16]
6.2. GERMANY
The dam owner/operator is responsible for the security of the plant. There do
not exist specific legal requirements though it seems to be standard to establish
preventive measures such as fences or surveillance cameras. Not only from a mon-
etary point of view it is in the own interest of the owner to protect his asset against
destruction, but as well under Public Safety considerations. Any attack or malicious
action against the dam might lead to a failure and thus endanger third parties.
The European Directive 2008/114/EC of 8 December 2008 “on the identifica-

tion and designation of European critical infrastructures and the assessment of the
need to improve their protection” which applies among others to installations for the
power generation has been transformed into the German national law.
6.3. SWITZERLAND
There are neither general “security” regulations for the hydropower sector in
Switzerland, nor any official requirements. The latest requirements from [13] for
protection against physical intrusion into parts of a dam are an exception. These
requirements are imposed on installations, which, due to their size or the existence
of a high risk, have a water alarm system.
639
Q.105 - R.37
7. RESUME
As shown above the subject of Emergency Planning is differently regulated in

the countries concerned.
Some of the concerned countries prefer to regulate the safety issues and
Emergency Planning in detailed legal stipulations. In other countries only the general
principles are established by law leaving it to the owner and operator of a dam on how
to fulfil his obligations. Sometimes reference is made to proven technical standards.
Regarding organisation issues and the duties and competences of state

authorities the paper shows that for constitutional/historical reasons the safety super-
vision is in some countries primary responsibility of one centralised state agency,
whereas in other countries this is in the primary autonomous responsibility of the
dam owner. In practise, both systems seem to work well. Common to all systems
is the necessary cooperation between different actors in the field of dam safety and
civil protection. The authors are convinced that all players such as the dam own-
ers, the state authorities/agencies or other experts should make an effort to further
discuss all relevant questions related to Emergency Planning by taking into account
and balancing the interests and needs of all parties concerned. The setting up of a
working group within ICOLD might be an appropriate way.
REFERENCES
[1] DWA Themenheft T1/2017 – Stauanlagensicherheit und Folgen bei Über-

schreitung der Bemessungsannahmen nach DIN 19700, 2017.
[2] DIN (Deutsche Norm) 19700 (revision 2004), and its parts 10 (DIN 19700-10)
: “Dams plants – Part 10: General specifications” respectively part 11 (DIN
19700-11): “Dam plants – Dams”, July 2004.
[3] Geschichtliches zum Wasseralarm, Dr. R.Biedermann und Bundesamt für

Bevölkerungsschutz, Bern 2004.
[4] BMLFUW, Mindestanforderungen an den Stauanlagenverantwortlichen von

“Kleinen Stauanlagen”, FEDERAL MINISTRY OF AGRICULTURE, REGIONS
AND TOURISM (BMLRT), 2009 [online: https://www.bmlrt.gv.at/wasser/
nutzung-wasser/stauanlagen-talsperren/Richtlinien.html
[5] Gesetz zur Ordnung des Wasserhaushalts (Wasserhaushaltsgesetz – WHG),

31.7.2009.
[6] FOSTER, M.A., FELL, R. and SPANNAGLE, M. Analysis of Embankment Dam

Incidents. UNICIV Report No. R-374, The University of New South Wales,
Sydney, 1998.
640
Q.105 - R.37
[7] FROEHLICH, D.C. Peak Outflow from Breached Embankment Dam. Journal
of Water Resources Planning and Management, 1995, Nr. 121(1): 90–97.
[8] RUEDISSER, B. and TSCHERNUTTER, P. Failure of Embankment Dams due

to Overtopping- Experimental Study and Hydrograph Prediction, Proceedings
of the 26th ICOLD World Congress, Symposium Hydro Engineering, 2nd- 3rd
July 2018, Vienna: 2629–2639.
[9] BMLRT, Guideline for Flood Wave Prediction of Breaching Retaining Struc-
tures, Federal Ministry of Agriculture, Regions and Tourism (BMLRT), 2018
[not yet published].
[10] DVWK Merkblatt 231/1995 (as of 2015): Sicherheitsbericht Talsperren –

Leitfaden, version as of 2015.
[11] Sieber, H.-U. (2009): “Sicherheit von Talsperren”. In “Betrieb und Sanierung
von Talsperren, Hrsg.: Weiterbildendes Studium Wasser und Umwelt,
Bauhaus -Universität Weimar, Eigenverlag, April 2009.
[12] Ordinance on the Warning, Alarm and Security Radio Network of Switzerland
of 18 August 2010.
[13] Swiss Federal Office of Energy: Directive on the Safety of Emergency concept
(version 2.0), 2015.
[14] Federal Water Retaining Facilities Act of 1 October 2010 (SR 721,101).
[15] Federal Water Retaining Facilities Ordinance of 17 October 2012

(SR 721.101.1).
[16] BMLFUW, Leitfaden für Zentrale Warten beim Betrieb von Stauanlagen, FED-
ERAL MINISTRY OF AGRICULTURE, REGIONS AND TOURISM (BMLRT),
2007 [online: https://www.bmlrt.gv.at/wasser/nutzung-wasser/stauanlagen-
talsperren/Richtlinien.html]
641
Q.105 - R.38
-------
-------
EMERGENCY DAM SAFETY INSPECTIONS IN LAO PDR (∗)
Patrice DROZ
Dam Safety Expert, STUCKY LTD, on assignment from the
Swiss Agency for Development and Cooperation SDC,
Switzerland
Georges R. DARBRE
Dam Safety Consultant on assignment from the
Swiss Agency for Development and Cooperation SDC,
Switzerland
Bouathep MALAYKHAM
Director General, Department of Energy Management,
MINISTRY OF ENERGY AND MINES, Laos
SWITZERLAND
SUMMARY
In the wake of the Xe Pian-Xe Namnoy Hydropower Project saddle dam failure
incident on 23 July 2018 and the nationwide flooding during July-August 2018, the
Government of Laos initiated an Emergency Dam Safety Inspection (EDSI) program.
Dam Owners were notified by the Ministry of Energy and Mines (MEM) to provide
reports covering all the aspects of dam safety including design, construction, oper-
ation, surveillance and EPP. Then, International Advisors (IAs) teams have been
asked to assist the MEM in reviewing the Emergency Dam Safety Inspection Reports
and conducting site visits to confirm the safety conditions of the dams announced
by their Owners.
∗ Inspections d’urgence de la sécurité de barrages en RDP du Laos
642 DOI: 10.1201/9781003211808-84

Q.105 - R.38
The article presents the general context of inspection and the synthesis of the
results of this EDSI for all the dams analysed and inspected by the IAs. The synthesis
has been elaborated based on a pre-defined data base filled-up by all the IA Teams.
The structure of the data base covers the main points which have been assessed. A
simple approach has been adopted, aiming at showing up the most important and
meaningful issues which have been raised during the various reviews. The synthesis
consists in a statistical treatment of the information collected, pointing out the main
findings and dam safety aspects which should be improved, essentially in terms of:
• hydrology and flood evacuation

• seismic design and verification
• visual inspection
• monitoring data analysis and reporting
• maintenance of the monitoring system
• EPP
RÉSUMÉ
À la suite de la rupture de la digue de col de l’aménagement hydroélec-

trique de Xe Pian-Xe Namnoy le 23 juillet 2018 et aux crues qui se sont déroulées
sur l’ensemble du territoire national en juillet et août 2018, le Gouvernement du
Laos a initié un programme d’inspection d’urgence de la sécurité des barrages.
Le Ministère de l’Energie et des Mines (MEM) a invité les propriétaires de barrage
à fournir des rapports couvrant l’ensemble des aspects touchant la sécurité des
barrages: conception, construction, exploitation, surveillance et plans d’urgence.
L’appui d’équipes de conseillers internationaux a ensuite été demandé par le MEM
afin de l’assister à étudier les rapports d’inspection d’urgence et d’effectuer des vis-
ites des barrages afin de confirmer les conditions de sécurité annoncées par les
propriétaires.
Le rapport présente le contexte général de l’inspection et la synthèse des résul-

tats de ce programme de vérification de la sécurité des barrages pour l’ensemble
des rapports analysés et inspectés par les conseillers internationaux. La synthèse
a été élaborée à partir d’une base de données prédéfinie remplie par les équipes
de conseillers. La structure de la base de données couvre les points principaux
qui ont été évalués. Une approche simple a été adoptée afin de faire ressortir les
problématiques les plus importantes et les plus significatives. La synthèse consiste
en un traitement statistique des informations collectées mettant en évidence les
principales constatations et les aspects de sécurité des barrages qui devraient être
améliorés, essentiellement en termes:
• d’hydrologie et d’évacuation des crues,

• de résistance aux séismes,
• d’inspection visuelle,
• d’analyse des données d’auscultation et de documentation,
• de maintenance du système d’auscultation,
• de plans d’urgence.
643
Q.105 - R.38
1. INTRODUCTION
In the wake of the Xe Pian-Xe Namnoy Hydropower Project saddle dam failure
incident on 23 July 2018 and the nationwide flooding during July-August 2018, the
Government of Laos initiated an Emergency Dam Safety Inspection program (EDSI):
Owners of Hydropower Projects with Concession Agreements above 15 MW were
notified by the Ministry of Energy and Mines (MEM) to provide reports covering all
the aspects of dam safety including design, construction, operation, surveillance
and Emergency Preparedness Plan (EPP).
The Department of Energy Management (DEM) of MEM exercising its man-

date and responsibilities for dam safety monitoring, is the focal point for this EDSI
program and enforcement of the Lao Electric Power Technical Standards (LEPTS)
as provided in the Electricity Law.
Owners were notified to undertake the EDSI in strict adherence to the terms
set forth in the EDSI Terms of Reference and summarized here below:
• General information
• Project features (incl. construction drawings and EPP)
• Hydrology (incl. spillway and emergency drawdown characteristics)
• Geology (incl. cut-off, grouting and drainage)
• Construction history (incl. diversion, construction problems)
• Operation and regulation plan (incl. under normal and flood conditions)
• Operation experiences during past major floods
• Stability analysis of the dam, spillway and appurtenant structures
• Instrumentation and records of performance observations
• Any known deficiencies
• Inspection report (incl. main structures, monitoring system, reservoir, down-
stream channel, O&M)
International Advisors (IA) Teams have been asked to assist MEM in reviewing
the EDSI reports and conducting site visits with MEM to confirm the safety conditions
stated in the EDSI reports. In order to avoid conflicts of interest, MEM has decided
that any IA will not be permitted to review any EDSI report of a hydropower project
in which a corporate entity of the same nationality as the IA has been involved in
the project preparation, development, construction or operation or has a direct or
indirect interest in the hydropower project.
Among several other contributors and development partners, Switzerland

has been supporting several International Advisors through the Swiss Agency for
Development and Cooperation SDC.
The reports or information provided of 54 dams have been reviewed by the

IAs. Actually, since several hydropower schemes present more than one dam, it is
644
Q.105 - R.38
the safety of 79 dams which has been reviewed. Each of the 17 IAs teams were
composed of a Team Leader: a dam engineer, a geologist and a hydrologist.
2. METHODOLOGY FOR THE DAM SAFETY ASSESSMENT
It is commonly agreed that Dam Safety relies on three pillars as indicated

below:
• Structural Safety based on adequate design

• Operation, Maintenance and Surveillance based on visual inspection, moni-
toring, data analysis and reporting
• Early warning and Emergency Preparedness Plan, including flood mapping,
Early Warning System and Communication Strategy, Emergency Response
Plan, Preparedness activities and access as well as evacuation routes
identification
Therefore, the review of the documents provided by the Owners has been
performed by the IAs according to these 3 aspects of Dam Safety.
3. A DATA BASE FOR SYNTHESIS
The IA Team Leaders were requested to fill-in a pre-defined database. The

structure of the database covers the main points which have been assessed by the
members of the IA Teams:
• suitability of dam type

• bottom outlet
• seismic hazard considered for the design
• geology/foundation/geotechnical issues
• design and construction of the dam
• spillway
• surveillance (visual inspection and monitoring, maintenance, data analysis
and interpretation)
• floating debris and sediment issues
• access
• Emergency Preparedness Plan (EPP)
Actually, 46 aspects contributing to dam safety have been considered in

the data base. Hence, for instance, regarding dam Design and Construction, the
following aspects were considered:
• General aspect or design of the dam

• Cracks, slips and joints
645
Q.105 - R.38
• Excessive vegetation
• Appropriate freeboard
• Appropriate uplift pressure design hypothesis (concrete dams)
• Appropriate general design and calculation (stability, sliding, seepage, …)
• Appropriate drainage network
or regarding Monitoring aspects:
• Geodetic survey measurements (H and V)

• Appropriate internal settlement measurement (embankment)
• Appropriate uplift pressure or piezometric monitoring
• Appropriate leakage or seepage measurement
• Bathymetric survey of plunge pool (at least regular inspection)
It must be pointed out that a wide variety of approaches have been adopted
by the various IAs from different countries and backgrounds to review the EDSI
reports prepared by the Owners, the Developers of the projects or their Consultants.
Inevitably, although the structure of the EDSI reports had been defined in detail, this
results in focus and emphasis being put on different aspects of the design, the
construction and operation as it pertains to the safety of each dam. To avoid major
bias, a simple approach for elaborating a synthesis of the EDSI has been adopted,
aiming at showing up the most important and meaningful issues which have been
raised during the various reviews.
In addition, the review of the EDSI reports did not only aim to check the com-
patibility of the scheme with the LEPTS requirements, but to provide a general safety
assessment based on international practice in terms of dam engineering. Hence,
the IAs have been asked to qualify each detailed item using the following rating
instruction:
• 0: no identified problem, no issue, already protective measures exist

• 1: no information or possible problem, possible hazardous evolution, improve-
ments recommended,
• 2: identified problem, corrective measures are to be considered urgently
• 3: critical issue, partial or complete drawdown necessary
• blank: not relevant
Then, a statistical treatment of the information collected was possible.
4. MAIN RESULTS OF THE SYNTHESIS
The synthesis of the results of the 17 teams of IAs consists in a statistical

treatment of the information collected, pointing out the main findings, the frequent
646
Q.105 - R.38
identified issues such as the lack of bottom outlets, the poor maintenance of monitor-
ing systems or the absence of EPP as well as the main ways for improving dam safety
in Laos.A selection of the various domains under review, after aggregation of their
various components is presented hereafter. The figures represent the percentages
of the various ratings obtained for each aggregated domain.
4.1. HYDROLOGY AND FLOOD EVACUATION
Improvements are necessary in 62% of the cases (Fig. 1). This is mainly since
design or check flood values have been underestimated.
Considering the “n-1” rule is current practice in many countries: checking the
hydraulic capacity of the design flood with one gate failing to open contributes to
the 10% of identified problems, but other factors also influence this important figure:
severe underestimation of the design and check flood or weak emergency power
supply to the gates. Corrective measures are to be considered urgently.
Fig. 1
Hydrology and flood evacuation rating
Cotation de l’hydrologie et de l’évacuation des crues
4.2. SEISMIC HAZARD CONSIDERED FOR THE DESIGN
Inappropriate seismic hazard hypothesis considered for the design are one
of the most important weaknesses identified during the EDSI report review. About
70% of the dams under review present or may present such problem in regards of
current international practice and recent developments (Fig. 2). The verification of
the seismic safety of these dams is highly recommended.
647
Q.105 - R.38
Fig. 2
Seismic design rating
Cotation de la conception pour la résistance aux séismes
4.3. DESIGN AND CONSTRUCTION OF THE DAM
The main problems related to design and construction are linked to:
• inappropriate freeboard
• excessive vegetation (embankment dams)
• defective or weak drainage system
• but more important:
• inappropriate uplift pressure hypothesis (concrete dams)
• inappropriate stability analysis
Fig. 3
Design and construction rating
Cotation relative à la conception et à la construction
4.4. SURVEILLANCE
Regular visual inspections and monitoring (Fig. 4) is frequently neglected.

Surveillance activities including visual inspections, monitoring data acquisition,
including geodetic survey, data analysis and reporting are essential elements of
648
Q.105 - R.38
dam safety. These aspects must be improved to achieve a reasonable level of dam
safety soon.
Monitoring data analysis is neglected or absent in more than 80% of the cases.
Improvements or urgent actions are needed regarding the maintenance of

the monitoring systems. Only 30% of the dams under review present satisfactory
maintenance of their monitoring system.
Fig. 4
Monitoring rating
Cotation relative à l’auscultation
4.5. EMERGENCY PREPAREDNESS PLAN (EPP)
EPPs have been neglected or are highly insufficiently developed in more than
70% of the dams. EPP represents the third main element of dam safety since it
enables appropriate management of and response to residual risks.
Fig. 5
EPP rating
Cotation relative au Plan de Préparation et d’Urgence
649
Q.105 - R.38
A particular effort should be given for the preparation and implementation of

EPPs (Fig. 5) which should integrate:
• Flood mapping in case of dam failure

• Early Warning System and Communication Strategy
• Emergency Response Plan or Emergency Action Plan
• Preparedness activities
• Access and evacuation routes identification
5. GENERAL SYNTHETIC OVERVIEW
The following figure presents a synthetic overview of the EDSI report review
main findings. For each main categories of parameters as defined in Ch. 3, the per-
centage of 0, 1 and 2 are presented under a “radar” format. The various categories
of parameters are grouped according to their appurtenance to the main 3 pillars of
dam safety presented in Ch. 2.
One can see immediately on Fig. 6 that particular efforts should be provided to:

• seismic design and verification
• visual inspection
• monitoring data analysis and reporting
• maintenance of the monitoring system
• EPP
Improvement of the validation processes of hydrological studies, geological

investigations and dam design should also lead to a better safety of future projects.
It is also worth noting that:
• Structural measures may require important financial needs and can be long
to put in place
• Surveillance and Operation and Maintenance are less demanding in terms
of financial needs and can be put into operation in a shorter timeframe; but
surveillance may require re-organisation of dam safety resources due to the
importance of a systematic and appropriate data analysis and reporting
• Emergency Preparedness is also less costly than the other measures, except
when problems of access to the dam have been identified
650
Q.105 - R.38
Fig. 6
General synthetic overview (ratings 0, 1 and 2)
Vue synthétique (cotes 0, 1 et 2)
6. CONCLUSION
The IA reviews of the EDSI reports of the various dams did not reveal crucial
safety problems which would require immediate emergency actions such as partial
or complete drawdown. Nevertheless, it appears that several aspects of dam safety
have been neglected and should be improved to achieve a significant upgrade in
the dam safety management of the large HPP dams in Laos.
ACKNOWLEDGMENTS
The implementation of the EDSI program and the results obtained so far
would not have been possible without the indefectible commitment of Mr. Voradeth
Phonekeo who managed this project for the DEM as well as of Mr. Sychath
Boutsakitirah, former Director General of DE.
The Swiss Agency for Development and Cooperation (SDC) as well as the
other Development Partners of various countries who financed the IAs are warmly
thanked. The support of the World Bank has been highly appreciated.
651
Q.105 - R.39
-------
-------
LEARNING FROM INCIDENT DURING CONSTRUCTION: COFFERDAM

COLLAPSE AT KARALLOE DAM, INDONESIA (∗)
Anissa MAYANGSARI
Section Head of Program, Data, Evaluation and Information, Dam Safety Unit,
MINISTRY OF PUBLIC WORKS AND HOUSING
Yayuk WIJAYA
PPK of Dam Planning, BBWS Pompengan Jeneberang,
Duki MALINDO
Head of Dam Safety Unit,
INDONESIA
SUMMARY
Karraloe Dam was located at Gowa City, South Sulawesi Province. The pur-
pose of the dam is to provide water for irrigation 7000 ha, raw water 440 l/s and
hydropower 4.5 MW. This CFRD has 85 m high from the foundation with 396 m
length. On January 21–22, 2019 heavy rainfall occurred at Makassar and 6 (six)
cities nearby, including Gowa City for more than 24 (twenty-four) hours. This heavy
rainfall caused flood at the dam and the cofferdam collapsed after more than 3 (three)
hours overtopping at the crest of the cofferdam. This paper analyses the cause of
heavy rainfall and the collapse of the cofferdam. At the design phase, the diversion
tunnel was designed at 25 (twenty-five) of flood return period. When the flood came,
the rainfall was 140 mm for 24 hours, which is the same as the 100 years return
∗ Apprendre d’un incident pendant la construction: l’effondrement du batardeau au barrage
de Karalloe, Indonésie
652 DOI: 10.1201/9781003211808-85

Q.105 - R.39
period. The estimation of flood at the location was 524 m3 /s which is larger than
design criteria for the diversion and cofferdam 366 m3 /s, which is equal to 50 years
flood return period. The collapse of cofferdam was caused by an anomaly of rainfall
in the area.
RÉSUMÉ
Le barrage de Karraloe se situe dans la ville de Gowa, sur la province de

Sulawesi du Sud. Le but du barrage était d’irriguer 7000 ha, fournir l’eau de 440 l/s,
et l’hydroélectricité pour 4,5 MW. Ce barrage à masque amont en béton est de
85 mètres de haut de la fondation et 396 mètres de longueur. Du 21 au 22 janvier
2019, de fortes pluies sont tombées à Makassar et six villes à proximité, compris
Gowa City pendant plus de 24 (vingt-quatre) heures. Ces fortes pluies ont provo-
qué des inondations au barrage, et le batardeau s’est effondré après plus de trois
heures de dépassement à la crête du batardeau. Cet article analyse la cause des
fortes pluies et de l’effondrement du batardeau. À la phase de conception, le tunnel
de dérivation était conçu pour une période de retour des crues de 25 ans. Quand
l’inondation est arrivée, les précipitations ont été de 140 mm pendant 24 heures,
ce qui équivaut à une période de retour de 100 ans. L’estimation de l’inondation à
l’emplacement était de 524 m3 /s, ce qui est supérieur aux critères de conception
pour le détournement et le batardeau de 366 m3 /s, ce qui équivaut à une période
de retour des crues de 50 ans. L’effondrement du batardeau a été causé par une
anomalie des précipitations dans la région.
1. INTRODUCTION
Karraloe Dam was located at Karralloe River, ±2 (two) km at the upstream

of Karalloe Weir at Garing Village, Tompobulu District and Taring Village, Bringbulu
District at Gowa City, South Sulawesi Province. The coordinate was 5◦ 30 50 LS
and 119◦ 48 20 BT. The purpose of the dam was for watering irrigation 7000 ha,
raw water 440 l/s and hydropower 4.5 MW.
This Concrete Faced Rockfill Dam has 85 m (El. +253 m) high from the bottom
of foundation with 396 m length and 8 m width at the crest. The thickness of face
slab was 0.3–0.6 m, the minimum thickness of plinth was 0.6 m with 5.6 m width.
The slope at the upstream face was 1:1.4 and at the downstream 1:1.5. At the
upstream crest of the dam completed with the parapet at El. +254 m. The volume of
embankment was 1.4 MCM. The total capacity of the reservoir was 40.4 MCM. The
upstream cofferdam was rockfill with incline impervious core zone with 142 m length,
6 m width and 30 m from the bottom of foundation at the El. +206.5 m. The slope at
the upstream 1:2.4 and 1:2.1 at the downstream, with the volume of embankment
653
Q.105 - R.39
Fig. 1
Location of Karalloe Dam [1]
Lieu de Karalloe Dam [1]
Fig. 2
The Layout of Karalloe Dam [1]
La mise en page de Karalloe Dam [1]
654
Q.105 - R.39
±175.000 m3 . The diversion tunnel was designed at 25 years flood return period
366 m3 /s. The type of diversion tunnel was a shoe horse with 6 m diameter, and
539 m length.
2. BACKGROUND
Hydrology analysis in dam design was the most important thing to design
the capacity of diversion tunnels and the height of cofferdam which protected the
construction of the main dam. The design of diversion which is based on the proper
criteria design would minimize the incident during construction.
On January 21–22, 2019 heavy rainfall occurred at Makassar City and 6 (six)
cities nearby include Gowa City where Karalloe Dam is located. During the flood,
there were 3 people killed, 45 people injured, more than 2000 people evacuated
which spread at 13 locations. More than 500 houses flooded with 50–200 cm depth
at Gowa City.
Fig. 3
Cross Section of Main Dam [1]
Coupe transversale du barrage principal [1]
Floods also occurred at dam location. After the flow overtopped at the crest of
cofferdam for more than 3 (three) hours, it collapsed (Fig.3). The design criteria of
cofferdam and diversion tunnel was 25 years return period of flood. So, the cofferdam
and diversion tunnel could not accommodate the flood. This incident caused loss
for more than 7 (seven) million USD. The chronologies of flood at the dam site were
described at the table.
655
Q.105 - R.39
Table 1
The Chronologies of flood at the dam site
DATE–TIME EXPLANATION
JANUARY 21, 2019
3 p.m. to the next day Moderate rainfall at the Karalloe Dam
JANUARY 22, 2019
09.23 a.m. Flow from Karalloe River measured at the inlet diversion tunnel was El. +189 m. All
the flow still passes through the diversion tunnel. The water level raise slowly
10.39 a.m. The flood reached the crest of the upstream cofferdam which is still constructed at
El. +196 m. The diversion tunnel has already submerged. It was already overtopped
at the crest of cofferdam.
11.08 a.m. Overtopping occurred at the crest of upstream cofferdam
02.02 p.m. Overtopping continued for 3 hours and 20 minutes and eroded the crest cofferdam
causing the cofferdam to collapse.
03.26 p.m. The flow from Karralloe River gradually receded.
05.50 p.m. River water elevation at the inlet diversion tunnel reached El. +186 m.
JANUARY 23, 2019
04.00 a.m. All flows from Karalloe River flowed through the river and did not pass through the
diversion tunnel. River bed elevation was at El. +169 m and the inlet tunnel at El.
+184 m.
The purpose of this paper is to analyze the hydrology concern that caused the
big flood at Gowa City and nearby on January 21–22, 2019. And then compare the
capacity of the diversion tunnel and cofferdam from the criteria design with the flow
of flood.
Fig. 4
The collapse of Karalloe Dam in 3 hours [2]
L’effondrement du barrage Karalloe en 3 heures [2]
656
Q.105 - R.39
3. LITERATURE REVIEW
3.1. DESIGN FLOOD CALCULATION
Design flood calculation is the important thing at dam design. Based on

Indonesian Guideline SNI 2415–2016 [3] about flood design calculation for calcu-
lating flood design requiring data as follows:
– Catchment area characteristic, which consists of topography, geological

condition, and land use area
– Rainfall data, such as maximum rainfall per year, rainfall duration, rainfall
distribution, rainfall station
– Maximum flood flow from various event
– Flood hydrograph with single peak form
– Morphology of river which is related to flood calculation such as river geometry,
river infrastructure that usually use for flood monitoring
The method that can be used for the design flood calculation depends on
the availability of maximum flood discharge data for a moment to find the period of
observation time. There is no need the flood data in a series of time and no data
filling for the missing data flow.
In the catchment area of the Karalloe Dam, there is no flood data. So the design
flood calculation uses rainfall data. The empirical method can be used if the amount
of flood discharge is based on the rainfall variable and watershed characteristics. The
synthetic unit hydrograph method is used, which is the equation of the hydrograph
of the unit from watershed characteristics data such as river length (L), river length
to center of gravity (Lc), watershed slope, and others. This synthetic method has
been widely used such as the Snyder Method [4], the US-Soil Conservation Service
method.
The synthetic method for calculating the design flood discharge used is Snyder
[4,5] with the following equation:
Q = q_p (25.4A)/1000 (1)
For flood discharge calculated from rainfall data, it is necessary to calibrate

the model parameters by trial and error or automatic calibration to obtain a good
relationship between the calculated and the observed hydrograph. It is necessary
to verify the level of reasonableness for the amount of flood discharge calculated
without using calibration with observational data. The amount of flood discharge that
overtakes Cofferdam is calculated by the following equation [4,5]:
Q = CLHˆ(3/2) (2)
657
Q.105 - R.39
3.2. HYDRAULIC CALCULATION FOR DIVERSION TUNNEL
The capacity of diversion tunnel design is calculated based on open channel

method and submerged flow condition [4,5]:
3.2.1. Open Channel
Open channel calculation for diversion tunnels is calculated if the water is not
fully charged to the tunnel. The water flows through the tunnel when there is air
space from the water level to the top of the tunnel and at the outlet still did not sink.
(H/d δ 1,2–1,5). The formula that use is manning equation as follows:
v = 1/nRˆ(2/3)Sˆ(1/2) (3)
Q = A.V (4)
3.2.2. Submerged Calculation
This flow occurs when the tunnel is full and at the outlet already submerged.

Q = A (2g(h − D/2 + L sinα)/((1 + f )))
3.3. DIVERSION TUNNEL FLOOD ROUTING
The calculation of flood routing for diversion tunnel use this formula:
((I _1 + I _2))/2 + (S_1/δt − Q_1/2) = (S/2 /δt + Q_2/2) (5)
The output from this equation are as follow:
– Elevation of flood water level

– Flood hydrograph through diversion tunnel
4. ANALYSIS
Heavy rain with strong winds on January 22, 2019 poured in the city of Makas-
sar and 6 (six) districts in South Sulawesi causing big floods. Warm sea surface
658
Q.105 - R.39
temperature (>27◦ C) indicates extremely high evaporation which causes an excep-

tionally large potential for convective cloud formation and weather conditions tend
to be cloudy to rainy in the Makassar City area of South Sulawesi.
Fig. 5
SST (Sea Surface Temperature) – left and SST anomaly – right [6]
Anomalie SST (Sea Surface Temperature) – gauche et SST – droite [6]
MJO (Madden-Julian Oscillation) on January 22, 2019 is in quadrant 5 strong

(Maritime Continent) so it affects the formation of rain clouds in Indonesian territory
in general, including in the South Sulawesi region.
Wet air masses from Asia Continent and the Indian Ocean that flow the
Makassar Sea to the low-pressure center of 1005 hPa in the Flores Sea. From
the convergence of these two air masses, a convergent wind pattern is formed in
the Makassar Sea which causes the cumulation of air masses in the convergent
area, so that the growth of rain clouds is quite significant.
Fig. 6
Madden-Julian Oscillation on January 22, 2019
Oscillation Madden-Julian le 22 janvier 2019
659
Q.105 - R.39
From 00.00 UTC the temperature is −60◦ C, the cold temperature indicates
that the cumulonimbus clouds have formed in the mature phase, the peak is at
01.00 UTC with the cloud peak temperature reaching −80◦ C. This phase occurred
in a long duration of time until 11.00 UTC and became an extinction phase with a
significant temperature increase of −37◦ C. The cold clouds indicate the growth of
cumulonimbus clouds in Makassar’s sea and spread over a wide range with a long
duration, causing heavy rain and strong winds in the South Sulawesi region. Rainfall
around the Karalloe Dam can be seen in Table 2. The catchment area of the dam is
183 km2. The location of the rain station can be seen in Figure 5.
Table 2
The rainfall data from the rainfall station nearby the dam site
RAINFALL STATION RAINFALL DATA LOCATION
JANUARY 21, 2019 JANUARY 22, 2019
Kelara 55 50 Downstream of the dam
Malakaji 50 97 Upstream of the dam
Malino 63 186 Upstream of the dam
Limbungan 38 328 Upstream of the dam
From the table, the rainfall station that is in the catchment area of Karalloe
dam area Sta. Malino, Sta. Kelara and Sta. Malakaji. These rainfall stations are
also used for rainfall design at design phase. While the Sta. Limbungan, although
the rainfall is the highest among the other rainfall stations, the location is not in the
catchment area of Karalloe Dam. But still the rainfall also affects the flood at the
dam site. The regional rainfall calculated by using Thiesen analysis, it was found
that regional rainfall was 140 mm. Based on the design rainfall calculation (Table 3),
the amount of rainfall for 24 hours is almost equivalent to the 100 year return period.
At the design phase, the flood design was calculated from rainfall data using
the Snyder hydrograph unit method (Table 3). From the flood design calculation,
then flood routing is carried out to find out the diameter of the diversion tunnel and
the required height of the upstream cofferdam.
Table 3
Rainfall design and flood design return period
RETURN PERIOD RAINFALL DESIGN FLOOD DESIGN
MM M3 /S
2 48 137
5 75 249
10 93 330
25 116 436
50 134 515
100 151 595
1000 213 869
PMP/PMF 482 2,086
660
Q.105 - R.39
Fig. 7
The Hydrograph of Karalloe Dam
L’hydrogramme du barrage de Karalloe
At the design phase, the inlet diversion tunnel elevation was designed at El.
+182.5 m. Before the construction was started, the detail topography was measured
as MC 0. There were differences in topography between the design phase and con-
struction. So the bottom of the inlet elevation of the diversion tunnel became El.
+184 m.
When the flood occurred, the flood was measured from the water level in
the inlet of the tunnel and the height of the runoff that passed through the crest of
cofferdam. The amount of flood discharge that occurs at the dam site is estimated
at 524 m3/s. This value of flood is the same as the more than 50 years return period
of flood. It is higher than the design criteria.
This CFRD has 85 m high from the foundation with 396 m length. On January
21–22, 2019 heavy rainfall occurred at Makassar and 6 (six) cities nearby, including
Gowa City for more than 24 (twenty-four) hours. This heavy rainfall caused flood at
the dam and the cofferdam collapsed after more than 3 (three) hours overtopping at
the crest of the cofferdam. This paper analyses the cause of heavy rainfall and the
collapse of the cofferdam. At the design phase, the diversion tunnel was designed at
25 (twenty-five) of flood return period. When the flood came, the rainfall was 140 mm
for 24 hours, which is the same as the 100 years return period. The estimation of
flood at the location was 524 m3 /s which is larger than design criteria for the diversion
and cofferdam 366 m3 /s, which is equal to 50 years flood return period. The collapse
of cofferdam was caused by an anomaly of rainfall in the area.
661
Q.105 - R.39
5. CONCLUSION AND RECOMMENDATION
5.1. CONCLUSION
The conclusions of the analysis results are as follows:
1. On January 21–22 2019, the sea surface temperature was warm (>27◦ C),
the MJO (Madden Julian Oscillation) was in quadrant 5 (Maritime Continent)
which is strong, as well as abundant of wet air masses from Asia continent and
the Indian Ocean flow through Makassar Sea indicates very high evaporation
which causes the potential for the formation of very large convective clouds
caused heavy rainfall with long duration.
2. Regional rainfall in the catchment area of the Karalloe Dam which occurred
on 21–22 January 2019 was 140 mm, almost reaching the 100 years return
period of rainfall.
3. The flood at the dam site is estimated at 524 m3/s, which exceeds the capacity
of the diversion tunnel which can only accommodate a discharge of 366 m3/s.
This number is higher than criteria for the design of diversion tunnels and
cofferdam [1,4].
5.2. RECOMMENDATION
The Recommendations of the analysis results are as follows:
1. The Calibration of design flood calculation from rainfall data is the most impor-
tant thing to make sure the design flood calculation is suitable in the location
of study.
2. Due to climate change and land use, it causes extreme rains and longer
droughts. For this reason, reliable hydrological data are needed in hydrological
analysis for dam design.
3. The construction schedule should consider the seasonal conditions at the dam
site.
ACKNOWLEDGEMENTS
Special gratitude for Mr. Alexsander Nandar as the leader of the project, who
has great contribution for Karalloe Dam Project passing the difficult time when the
flood occured. Mr. Arif Paputungan as a team leader of consultant supervision who
always day to day supervise the project and provide the data for this paper. All the
team at the project who work hard day and night at Karalloe Dam project.
662
Q.105 - R.39
REFERENCES
[1] PT. METTANA. Final Report of Karalloe Dam Design, 2014.
[2] SUPERVISION CONSULTANT KARALLOE DAM. Flood Report at Karalloe

Dam, January 2019.
[3] SNI 2415–2016. Perhitungan debit banjir rencana. Badan Standardisasi

Indonesia. 2016.
[4] CHOW, V.T. Handbook of Applied Hydrology. McGraw-Hill, New York, 1964.
[5] DIREKTORAT JENDERAL SUMBER DAYA AIR, DEPARTEMEN PEKER-

JAAN UMUM. Panduan Perencanaan Bendungan Urugan, Volume II. Analisa
Hidrologi, 1999.
[6] AUSTRALIAN GOVERNMENT, BUREAU OF METEOROLOGY. www.bom.

gov.au. Accessed November 2020.
663
Q.105 - R.40
-------
-------
A QUANTITATIVE APPROACH TO THE RELIABILITY OF THE DAM EARLY

WARNING SYSTEM, AGAINST THE RISK OF LOSS OF LIFE IN PEOPLE AT
RISK (PAR) FACING DAM COLLAPSE DISASTER (∗)
Anto HENRIANTO
Senior Dam and Water Resources Engineer, PT. VIRAMA KARYA (PERSERO)
Lecturer of Civil Engineering Department, ESA UNGGUL UNIVERSITY
Esti WULANDARI
Senior Dam and Water Resources Engineer, PT. VIRAMA KARYA (PERSERO)
INDONESIA
SUMMARY
PENRIS response to disaster is described for sudden conditions (without early

warning), and ready conditions (with early warning), using the Graham formula to
test whether in Indonesia, the prediction of the number of LoL is appropriate using
this formula. This quantitative approach proves that an EWS can reduce the number
of LoLs by around 99% or even 100% of the total number of LoLs compared to dams
that have not used an early warning system. If based on this quantitative approach,
the relevant stakeholders are optimistic that they can guarantee LoL = 0.
RÉSUMÉ
La réponse de PENRIS à une catastrophe est décrite pour des conditions sans
alerte précoce et des conditions avec alerte, en utilisant la formule de Graham pour
tester si en Indonésie, la prédiction des pertes de vie est appropriée en utilisant cette
∗ Une approche quantitative de la fiabilité du système d’alerte précoce des barrages, contre
le risque de pertes de vie des personnes à risque lors de l’effondrement de barrages
664 DOI: 10.1201/9781003211808-86

Q.105 - R.40
formule. Cette approche quantitative prouve qu’une alerte précoce peut réduire les
pertes de vie d’environ 99% voire 100% par rapport aux barrages qui n’ont pas
utilisé de système d’alerte précoce. Si sur la base de cette approche quantitative,
les parties prenantes concernées sont optimistes quant à la possibilité de garantir
l’absence de perte de vie
1. INTRODUCTION
Based on references from many dam collapse disasters in the world, there
are always victims of the People at Risk (PAR or PENRIS in Indonesia), so that the
handling of risks to fatalities is a manifestation of the 2017 International Commit-
tee on Large Dams (ICOLD) Symposium agreement in Prague, Czech Republic,
and published by the Canadian Dam Association (CDA) in 2018, agreed that dams
around the world must be safe from the risk of loss of life (LoL). Prediction of LoL in
PENRIS depends on the characteristics of the flood, namely, velocity, flood arrival
time and flow depth, which threatens PENRIS, and their response to disaster.
The results of the PENRIS research from the Modified ICOLD Method with the
new equation for the LoL Prediction Formula (2019), the hazard classification level
from ICOLD and Risk Zoning Rule from Indonesian Dam Safety or Balai Keamanan
Bendungan [2], and the PENRIS response from Graham [3] concluded mathemati-
cally, that a dam with an early warning system (EWS) will reduce the amount of LoL
in the nearest village to almost 100% if implemented correctly.
2. METHODOLOGY
The methodology begins with the development of the framework of the 3 pillars
of the dam safety concept in Indonesia, it is proposed to add 1 more pillar, for
handling the risk of loss of life as shown in Fig.1 as follows:
Fig. 1
Development of the 3 Dams in Safety Pillar Concept
Développement du concept de pilier 3 barrages en sécurité
665
Q.105 - R.40
The concept of adding a fourth pillar to the concept of dam safety, shown in
Fig. 2 as follows:
Fig. 2
Development of the 4th Pillar of Dam Safety Concept
Développement du 4e pilier du concept de sécurité des barrages
The fourth pillar begins with a study of EAP products that are updated every
5 years, becoming a reference for the initial steps of applying the concept. PENRIS
readiness in the face of disasters depends on the design of the accuracy and relia-
bility of the early warning system by synchronizing the determination of risk levels
based on ICOLD, prediction of the number of LoL from the regression equation of the
event history of the event of a dam collapse in the world and the PENRIS response
index from Graham [3] for various level of understanding of disaster.
Next is a long-term effort to prepare a disaster-based spatial plan that will

reduce the risk of loss of PENRIS life to the maximum extent possible where each
dam has a different level of risk, expressed as an index of risk of LoL.
2.1. MODELING PREDICTION OF LOSS OF LIFE
The risk of LoL on PENRIS depends on the readiness of PENRIS to face

the catastrophic collapse of the dam that can occur at any time. Approach to the
prediction of loss of life is carried out by making a regression equation for 38 dam
666
Q.105 - R.40
events in the world, namely, Vega de Terra Dam in Spain, in 1959 until the last Situ
Gintung Dam in Indonesia, in 2009 and applied as a LoL prediction of 16 dams
in Indonesia that The EAP has been made, divided into conditions the number of
PENRIS that can be seen in Fig. 4 as follows:
Fig. 4
Regression Equation of LoL vs PENRIS (PENRIS > 10.000 people)
Équation de régression de LoL vs PENRIS (PENRIS > 10.000 personnes)
Prediction of LoL, with PENRIS > 10,000 people is far less than
PENRIS < 10,000 people. it can be assumed that the regression equation, here-
inafter referred to as LoL 2019 Equation as follows:
• PENRIS without Early Warning System:
LoL = 0, 7535(PENRIS)0,76 (1)
• PENRIS with Early Warning System:
LoL = 0, 0002(PENRIS) (2)
While PENRIS’s response to disaster, measured through an index from

Graham [3], can be seen in Table 1:
LoL 2019 equation is tested against the suggestion interval from Graham [3]
where the results will indicate whether for conditions in Indonesia, the recommended
interval is appropriate or needs to change because the downstream conditions of
667
Q.105 - R.40
Table 1
PENRIS Vulnerability Levels to Flood Category
dams in Indonesia are not the same as in the location where the Graham formula
[3] was made in Europe and the USA.
2.2. DETERMINING HAZARD CLASSIFICATION
Determination of the index refers to the classification of potential disasters

from ICOLD where the amount of flow rate, with a parameter H2 (V)1/2 indicating
water damage consists of 2 factors, namely inundation height (H) and magnitude of
flow velocity (V), can be seen in Table 2 as follows:
Table 2
Potential Hazard Classification
668
Q.105 - R.40
2.3. DETERMINING RISK ZONE CLASSIFICATION
Regarding the level of danger due to flood waves due to dam collapse, Balai
Keamanan Bendungan [2], referring to the provisions of ICOLD, has divided the
affected areas of disaster risk into 3 (three) zones, are:
1. High Danger Zone: PENRIS was directly affected by a major flood, where
H < 0,8 m and V < 2,0 m/s
2. Zone of Consideration: a lot of uncertainty. PENRIS was directly affected
by the flood, some of its status became loss of life, some survived, where
0,8 < H ≤ 1,4 and 0 ≤ V ≤ 2,50 m/s
3. Low Danger Zone: PENRIS is not affected by flooding, where H > 1,4 m and
V ≥ 0 m/s; H < 1,4 m and V > 2,0 m/s
Figure 4 represent the relationship between PENRIS and LoL which produces
a regression equation for estimating the number of LoL in dams in Indonesia that
still stands intact because no disaster has occurred. In the process, there will never
be an exact value for the number of LoL, but it can be predicted at intervals based
on research from the history of the dam collapse events that have occurred. The
nearest village to the dam will receive a higher risk than villages that are further
away. The higher the PENRIS response to EWS, the risk of LoL, is lower so that the
value of dam security risks can be lowered. The collaboration between ICOLD, LoL
equation 2019, Balai Keamanan Bendungan [2] and Graham (2010) for conditions
approaching, that is, without an EWS and with an early warning where the level of
PENRIS understanding of EWS is described as follows:
• Risk Level: ICOLD and Balai Keamanan Bendungan [2]

• Prediction of LoL: LoL Equation 2019 and Interval Response PENRIS using
Graham (2010) formula
• Level of Understanding of Floods: Not Apply (Without EWS) and >60 minutes –
Clearly (With EWS)
An overview of predicted LoL with and without early warning can be seen in
Table 3 and Table 4 below:
669
Table 3
Calculation of Prediction LoL (Without Early Warning System)
Condition: PENRIS are Not Ready for Disasters
Level of Understanding of Floods: Not Apply

Table 4
Calculation of Prediction LoL (Early Warning System)
Condition : PENRIS Ready for Disasters
Level of Understanding of Floods: >60 minutes (Clearly)

Q.105 - R.40
For conditions without EWS at 16 (sixteen) dams in Indonesia, where the

minimum interval limit is not constant one value as in Graham [3] but one equa-
tion function, namely the 2019 LoL Equation (Formula 1). For EWS, the minimum
interval limit is a function of the 2019 LoL Equation (Formula 2). This condition also
reflects the different characteristics of PENRIS at each dam location in Indonesia.
The amount of LoL can be seen in Table 5 as follows:
Table 5
Prediction LoL (With and Without Early Warning System)
From the comparison of Table 5, it can be seen that for each dam, without EWS
and with EWS there is a very large difference in the number of LoLs. Dams with an
EWS will provide more optimal protection for PENRIS compared to those without
an EWS. Cengklik Dam, Simo, Cipancuh, Greneng, Way Jepara, Gembong, Klego,
Gunung Rowo and Banyukuwung, for the nearest village, LoL = 0. Meanwhile, for
the remaining villages there is still a risk of LoL, but with a very small number so it
is hoped that with optimal efforts it can be zero. LoL reduction for the remaining and
nearby villages, can be seen the samp in Fig. 5 and Fig. 6, as follows:
672
Q.105 - R.40
Fig. 5
LoL Prediction at Remaining Village ( PENRIS > 10.000 people)
Prédiction LoL au village restant (PENRIS > 10.000 personnes)
Fig. 6
LoL Prediction at Nearest Village (PENRIS < 10.000 people)
Prédiction LoL au village le plus proche (PENRIS < 10.000 personnes)
673
Q.105 - R.40
4. CONCLUSION
The conclusions that can be drawn from the review of this paper are as
follows:
1. Between study sources, namely international regulations, literature [3] and

guidelines [2] and historical data (LoL 2019 equation) there is a close rela-
tionship to determine the prediction of the number of lives lost in PENRIS
in the face of a catastrophic dam collapse that can come at any time
unexpectedly
2. PENRIS response to disaster is described for sudden conditions (without early
warning), and ready conditions (with early warning), using the Graham formula
(2010), to test whether in Indonesia, the prediction of the number of LoL is
appropriate using this formula, is as follows:
• Sudden Dam Collapse Conditions
Nearest Village: The predicted LoL value is outside the PENRIS suggested
response interval. The minimum interval value for Indonesian conditions refers
to the 2019 LoL Equation Formula 1 so that it can be a prediction reference.
Remaining Village: LoL value of 16 dam EAPs, using Formula 1, all enter the
suggested response interval (Can be used as a reference).
• Condition with Early Warning > 60 minutes (Clearly).
An early warning system already exists, the level of risk of dam safety is con-
sidered to have dropped from high to low status, so the Nearest and Remaining
Villages, for LoL values, use Formula 2 into the interval suggested by Graham
(2010).
3. This quantitative approach proves that an EWS can reduce the number of LoLs
by around 99% or even 100% of the total number of LoLs compared to dams
that have not used an early warning system.
4. If based on this quantitative approach, the relevant stakeholders are optimistic
that they can guarantee LoL = 0, if a disaster occurs, they can convey to ICOLD
and then carry out the dam safety mandate at the 2017 ICOLD Symposium in
Prague like the previous 15 countries.
ACKNOWLEDGEMENT
The authors would say many thanks to KNIBB INACOLD and PT. Virama Karya
(Persero), for sponsorship to participate in ICOLD 2022, Esa Unggul University,
and many parties who cannot be mentioned one by one, who have supported the
publication of this paper.
674
Q.105 - R.40
REFERENCES
[2] ICOLD. World View: CDA International Strategy and Participation in ICOLD.
Canadian Dam Association. http://www.icold-cigb2019.ca/. 2018.
[3] BALAI KEAMANAN BENDUNGAN. Guidance of Determining Hazard Classi-

fication of Dams, Department of Public Works Indonesia. 1988.
[4] BROWN, G. Assessing Threat of Life from Dam Failure, USBR Published.
2010
[5] Mc. CLELLAND, D. Predicting Loss of Life in Case of Dam Failure and Flash
Flood, United States of Corps Engineer (USACE), Page 5–8. 2012.
[6] Mc. CLELLAND, D. Estimating Life Loss for Dam Safety Risk Assessment – A
Review and New Approach, United States of Corps Engineer (USACE), Page
10–18. 2002.
[7] FEMA Emergency Action Planning for State Regulated High – Hazard Potential
Dams, Federal Emergency Management Agency on Dam Safety USA, Page
6–2. 2006.
[8] HOMELAND SECURITY. Dam Sectors: Estimating Loss of Life for Dam Failure
Scenarios, U,S Department of Homeland Security, Page 42–45. 2011.
[9] MULYONO, J. Research on Situ Gintung Dam Failure Disaster, Indonesian

Committee on Large Dams (INACOLD). Page 1–72. 2011.
[10] PT. Wiratman-Lotti Associati. Dam Operation Improvement and Safety Project
(DOISP) Report, Department of Public Works. 2015
[11] PUSLITBANG SDA. Inventory of Dam in Indonesia, Department of Public

Works and Housing, Page 10–120. 1995.
[12] USBR. Reclamation Consequence Estimating Methodology (RCEM), U.S

Department of The Interior Bureau of Reclamation. 2015.
675
Q.105 - R.41
-------
-------
ANALYSIS OF SPILLWAY POSITION ON THE CIBEET DAM CONSTRUCTION

FROM SOIL GEOLOGICAL ASPECTS (∗)
Anang MUCHLIS
Head of Citarum River Basin Organization
Cecep Muhtaj MUNAJAT

Sub coordinator Planning of Integrated Infrastructure Development
Citarum River Basin Organization, Directorate General of Water Resource,
Najlawati Laitifah SYAZWANI

Staf of Data and Information Citarum River Basin Organization
Directorate General of Water Resource,
Sonny B. WICAKSONO
Engineering Geologist
INDONESIA
SUMMARY
Floods in the Karawang and Bekasi Regencies caused by the Cibeet River’s
overflowing occur almost every year. This flood needs special attention since its
effect is very detrimental for the community because it ruins the settlements and
irrigation areas in Karawang and Bekasi Regencies. One of the efforts made by
the Central Association of Citarum River is planning Cibeet Dam to control the flow
of Cibeet River, so it will not inflict the flood downstream. Cibeet River is one of
the Citarum tributaries, which has a watershed area of about 919,93 km2 . It covers
∗ Analyse de la position du déversoir du barrage Cibeet à partir des aspects géologiques
676 DOI: 10.1201/9781003211808-87

Q.105 - R.41
Cianjur Regency, Bogor Regency, Karawang Regency, and Bekasi Regency. The
upstream of Cibeet was started in Lemo Mountain, Cianjur Regency, which flows
along 101 km to the estuary of the Citarum River. According to the analysis result
in the dam’s planning, several problems were viewed from the aspect of soil type
condition in the ground related to spillway position determination and emergency
spillway. Existing retrieval structures were not in ideal situations, and prone to surface
erosion could disturb the slope stability. Therefore, it was necessary to analyze
determining alternative changes in the layout of several supporting water structures
in the Cibeet Dam design to function according to the dam’s design ages.
RÉSUMÉ
Les inondations à Karawang Regency et Bekasi Regency, causées par le

débordement de la rivière Cibeet, se produisent presque chaque année. Ces inon-
dations nécessitent une attention particulière car leur effet est très préjudiciable pour
la communauté car elle ruine les colonies et les zones d’irrigation des régences de
Karawang et de Bekasi. L’un des efforts déployés par l’Association centrale de la
rivière Citarum consiste à planifier le barrage Cibeet pour contrôler le débit de la riv-
ière, afin qu’il écrête les inondations en aval. La rivière Cibeet est l’un des affluents
Citarum qui a une surface de bassin versant d’environ 919,93 km2 , elle couvre Cian-
jur Regency, Bogor Regency, Karawang Regency et Bekasi Regency. Elle prend sa
source à Lemo Mountain, Cianjur Regency et coule sur 101 km jusqu’à l’estuaire de
la rivière Citarum. Selon le résultat de l’analyse, dans la planification du barrage, il
y avait plusieurs problèmes liés au sol de fondation pour la position du déversoir, de
déversoir d’urgence. Les structures de récupération existantes n’étaient pas dans
un état idéal et étaient sujettes à une érosion de surface pouvant mettre en cause
la stabilité des pentes. Il était nécessaire de mener une analyse pour déterminer les
modifications dans la conception du barrage Cibeet, afin qu’il puisse fonctionner à
toutes les époques de sa réalisation
1. INTRODUCTION
The Cibeet watershed has a 919.93 Km2 area formed from the Cibeet River
(617.93 Km2 ) and the Cipamingkis River (302.00 Km2 ). The Cibeet River consists
of two parts of the Sub-watershed, the Upper Cibeet Sub-watershed of the Cipam-
ingkis Confluence. The Cibeet River and Cigentis River have a river flow length
of about 18.1 Km2 , the Citalahab River with a river flow length of 18.8 Km2, and
the Cibeet Downstream Sub-watershed has 82.71 Km2 . The catchment area of
the Cibeet Watershed is 919.93 km2 ; it covers Cianjur Regency, Bogor Regency,
Karawang Regency, and Bekasi Regency. Cibeet River’s upstream starts from Lemo
Mountain, Cianjur Regency, which flows along 101 km to Citarum River’s estuary.
677
Q.105 - R.41
The Cibeet Dam Plan was conducted in 2017 by PT. Mettana Engineering
Consultant. It had a reservoir volume of 139.40 million m3 , where the development
plan was proposed to resolve the floods which often occurred in the Karawang and
Bekasi Area. However, from the analysis result of the Cibeet Dam design plan,
several problems were related to aspects of soil type conditions in the location. It is
categorized in position determination of Spillway, Emergency Spillway, and Retrieval
Construction, which had no ideal conditions and prone to experience erosion, also
it could disturb the slope stability
To overcome the mentioned problems, BBWS Citarum conducted further

investigations to provide alternative changes of the several supporting water
structures layout on Cibeet Dam Plan to function according to the dam’s design age.
2. RESEARCH STAGES AND METHODOLOGY
The research stages and methods were conducted in the following steps:
a. Analysis of geological condition, dam construction material, and layout of the

Cipunegara Dam Development Plan.
b. The Alternative Analysis Suggestion of Layout Changes in Diversion Tunnel
Buildings, Spillway, and Intake Conduit Position.
3. RESULTS AND DISCUSSIONS
3.1. ANALYSIS OF GEOLOGY CONDITIONS, CONSTRUCTION MATERIALS, AND DAM

LAYOUT IN THE CIBEET DAM DEVELOPMENT PLAN.
Cibeet Reservoir Plan is located in Cariu District, Bogor Regency, between

6◦ 18’0”–6◦ 47’10” South Latitude and 106◦ 23’45”–107◦ 13’30” East Longitude. Cariu
District is one of forty districts in Bogor Regency. Regional geology of the job site
based on the Regional Geology Map of Cianjur Java Page (Fig. 1).
On the plan area of the dam site, there are generally clay rock forms. It contains
hard marl limestone marl, dark-gray limestone layers, and nodules with a thickness
of about 2 or 3 meters. Sometimes, it also includes the inset of green glauconite
sandstones. The deposits of Alluvium surface (Qa) consisted of clay, silt, sand,
and gravel. Especially the present condition of river sedimentations. It includes the
erosion of the slope in the north and south of Cianjur, then, The Old Alluvium (Qoa)
consisted of conglomerates and river sand, which is composed of andesite and
basalt. Batuguling - Batuguling comes from excreted limestone, sandstone, silica,
and andesite compressions.
678
Q.105 - R.41
Fig. 1
Geological Map of Cianjur Java [1]
Carte géologique de Cianjur Java [1]
Table 1
Rock Classification and Rock Parameters (medium hard rocks)
CLASS MODULUS OF MODULUS OF SHEAR STRENGTH SEISMIC UCS
DEFORMATION ELASTICITY COHESION INT. FRICTION VELOCITY (KGF/CM2 )
(KGF/CM2 ) (KGF/CM2 ) (KGF/CM2 ) ANGLE (◦ ) (M/SEC)
CH 20,000 60,000 20 45 2,500 >500
CM 8,000 24,000 10 35 1,800 100–500
CL 4,000 12,000 8 30 1,500 40–100
D 2,000 6,0000 5 28 1,200 20–40
Based on the design stage’s drilling data, the foundation plan of Cibeet Dam
is generally limestone clay and sandstones (Fig. 2). Both of these rock types have
an excellent bearing capacity for the foundation of the embankment dam type. The
class of rocks is generally CL – CM or Moderately Weathered, with a compressive
strength value of about 40–200 kgf/cm2
The foundation rock’s permeability has a Lugeon value of about 5–15, fixed by
the curtain grouting. The reservoir area generally consists of dark gray marl, marl
claystone, and clay shale with the insets of quartz sandstone, quartzite, and marl
limestone. With the low geological conditions and undulating topography, the reser-
voir’s potential for leakage to adjacent valleys is minimal. On the dam construction
material, the form of a valley at Cibeet River in the dam site plan location is a low
679
Q.105 - R.41
680
Q.105 - R.41
Fig. 2
Map and Geological Cross Section of Dam Foundations [2]
Carte et coupe géologique des fondations des barrages [2]
undulating valley, gently sloping, and the condition of foundation rocks are claystone-
sandstone sedimentary rock. The type of dam that suits this area’s geological and
topographic requirements is the zonal fill type.
Based on the design details, the construction material required for the dam
body of zonal embankment type is around 3 million m3 . The landfill material is silt;
it can be obtained from the borrow area plan located in the inundation area with
a distance of 0.5 to 2.5 km. Clay soils will utilize the residual soil from the perfect
weathering result of claystone and sandstone [4]. The color of clay silt is brown; the
thickness is relatively thin by removing the entire top-soil as deep as 30 cm, then,
the under part has a thickness around 1.0–2.0 m, it can be used as soil pile material
(core material).
Sand and gravel materials are required for concrete and filter materials. For
concrete materials, the sand procurement will utilize the sediment materials of Cibeet
River in the reservoir area and crushing results from microdiorite igneous rocks.
681
Q.105 - R.41
The rock material for riprap will utilize the microdiorite igneous rock units found
in Antajaya Village, Tanjungsari District, with a distance of 5 ∼ 6 km from the dam
location. Based on the result of the design phase study. It can be concluded that
the construction material requirements of the dam body and concrete aggregate
materials can be met in the inundation area of the reservoir, with a maximum distance
of around 6 km from the dam location. The body of Cibeet Dam is a random landfill
dam with an upright core in the center; the dam’s top-length is 1,450 m, stretched
from the southeast to the northwest, the dam’s maximum height is 39 m from the
river bed and 44 m from the bottom of the excavation. The elevation of the dam top is
designed at El. +100.0 m with a keep height from average water level (El. +95.0 m)
of 5.0 m. The upstream slope is designed as 1V: 3.0H, and the downstream slope is
1V: 2.7H. Other facility buildings are the spillway, emergency spillway, and retrieval
construction; they are all placed on the left backrest position (Fig. 3).
The water elevation of Cibeet Dam in full reservoir condition is determined

based on the water availability (potential inflow) and Dam Storage Capacity. Based
on the graph of volume vs. elevation inundation area relationship and calculation of
inflow potential, so, it is determined as follows:
Total storage volume : 93,04 million m3

Effective storage volume: 63,55 million m3
Dead reservoir volume: 31,48 million m3
Normal reservoir elevation: +95,00 m (dpl)
Dead reservoir elevation: +83,95 m (dpl)
3.2. ANALYSIS OF ALTERNATIVE SUGGESTIONS FOR CHANGES IN LAYOUT AND

POSITION OF DIVERSION TUNNEL BUILDINGS, SPILLWAY, AND INTAKE CONDUIT
Other facility buildings are the spillway, emergency spillway, and retrieval
infrastructures; they are all placed on the left backrest position. The foundation rocks
are commonly claystone with very low strength rocks at its characteristics classified
in the left support position. However, it is adequate as a foundation structure (see
table 2). Yet, this claystone has quick weathering properties (rapid weathering).
Table 2
Deere and Miller’s Classification of Intact Rock Strength
DESCRIPTION UNIAXIAL COMPRESSIVE STRENGTH
KGF/CM2 MPA
Very low strength 10–250 1–25
Low strength 250–500 25–50
Medium strength 500–1000 50–100
High strength 1000–2000 100–200
Very high strength >2000 >200
682
Q.105 - R.41
Fig. 3
The Map of Preliminary Layout Design and Cibeet Dam Section [2]
La carte de la conception du schéma préliminaire et de la section du barrage
Cibeet [2]
The claystone and shale are categorized as soft rock, even if it is in fresh con-
dition. All clay stones in various situations (fresh or weathered) tend to form racks
in the surface area when it contacts air and water. A dry-wet-dry cycle over the time
scale of several hours to days, the claystone will break down quickly (disintegrated
soon). The mineral contents of claystone, such as kaolinite, montmorillonite, and
illite, generally have swelling and shrinking characteristics. Since the structure has
fragile bonds between the lattices, so the entry easily breaks it of water. When
excavated and exposed, these fast weathered rocks will be very vulnerable to
weather changes, so the excavation’s surface must be protected immediately. It
683
Q.105 - R.41
is to prevent surface erosion and the development of slope stability. With the condi-
tions described above, it is necessary to make several considerations regarding the
complementary building plan, there are:
a. Avoid digging many slopes in the claystone area because it will be prone to
erosion and landslides.
b. The spillway position, emergency spillway, requires digging a high enough
slope in the chute area (about 15–20 m) on the claystone.
c. It is necessary to protect a lot of excavated slopes.
d. The excavation of the building conduit for the foundation and the excavation
walls will also be in the claystone.
e. If the Diversion Tunnel conduit position is on the left side of the backrest, it will
also result in a temporary cofferdam requiring a length of about 400 m; this will
complicate the construction period.
See the result of the analysis above. When viewed from a geological aspect, it
is necessary to have an alternative change in the layout of several water structures
in the Cibeet Dam Development Plan.
A. Change Suggestion of Diversion Tunnel Position
The position of the Diversion Tunnel building is proposed on the right side
(Fig. 4) by utilizing the existing valley topographical formation, and the type of Diver-
sion Tunnel is open channel diversion. It will reduce the excavation of high cliffs. The
length of the temporary cofferdam is shorter.
B. Change Suggestion of Spillway Position and Intake Conduit
The position change of spillway building layout is in the center position, so the
Cibeet Dam is divided into Main Dam, Saddle Dam 1, and Saddle Dam 2 (Fig.5).
The spillway building in this position will rest on limestone sandstone and an intake
building or Intake Conduit on the backside right of Main Dam It will also rest on
limestone sandstones. The physical characteristics of limestone sandstones are
better than claystone.
C. The Suggestion of Position Change and Emergency Spillway Type
The Emergency Spillway building is proposed for the Fuse Plug type and is
placed between the Saddle Dam 1 and Saddle Dam 2; then, on the downstream
side, there is an open channel that connects to the former open channel of the river
defender.
684
Q.105 - R.41
Fig. 4
The Layout Map of Open Channel Diversion and Temporary Cofferdam [2]
La carte de disposition de la dérivation à ciel ouvert et du batardeau temporaire [2]
Fig. 5
The Layout Map of Spillway and Intake Conduit [3]
La carte de disposition du déversoir et du conduit d’admission [3]
685
Q.105 - R.41
4. CONCLUSION
Based on the result, it can be concluded the following things:
1. Cibeet Dam was built to solve the flood problem in Karawang and Bekasi area.
2. Cibeet Dam had a sufficient storage volume of 63,55 million m3
3. From technical geological analysis results, it was known that the plan to
determine the position of the spillway, emergency spillway, and retrieval infras-
tructures had a non-ideal condition and were prone to surface erosion. Then
also, it could disturb the slope stability. From the results of this analysis,
changes in the layout of the building were made. They were:
a. The Diversion Tunnel position was aimed at the right side by utilizing the
existing valley topography formation, and the Diversion Tunnel type was:
open channel diversion.
b. The spillway building layout’s proposed position was in the center, so
Cibeet Dam was divided into two parts: Main Dam, Saddle Dam 1, and
Saddle Dam 2.
REFERENCES
[1] SUDJATMIKO. Peta Geologi Lembar Cianjur, Skala 1 : 100.000 Dit Pusat
Penelitian dan Pengembangan Geologi, Bandung. 1972.
[2] PT. METTANA ENGINEERING CONSULTANT. Design Details of Cibeet Dam.

2017.
[3] PT. KWARSA HEXAGON. Feasibility Study of Cibeet Dam. 2016.
[4] SOSRODARSONO, S. Bendungan Type Urugan, Pradnya Paramita, Jakarta.

1977.
686
Q.105 - R.42
-------
-------
DAMS AND DIKES SAFETY MANAGEMENT IN ROMANIA; PAST, PRESENT

AND PERSPECTIVE (∗)
Altan ABDULAMIT
Assoc. Prof. of Civil Engineering,
TECHNICAL UNIVERSITY OF CIVIL ENGINEERING BUCHAREST,
PRES. OF ROCOLD
Dan STEMATIU
Romanian Academy of Technical Sciences,
ASTR
ROMANIA
SUMMARY
The present paper briefly presents some of the main aspects of the dam safety
management, including historical evolution, legislation, standards and norms that
regulate dam engineering field, dam monitoring and risk management. The con-
tribution of the Romanian National Committee on Large Dams (ROCOLD) in the
development of a modern dam safety concept is emphasized.
After 1990, the serious decline in the construction of new dams coincided
with some incidents and accidents occurring at these highly complex engineering
systems. Such difficult circumstances led to the development of a comprehen-
sive national legal package of laws, technical norms and guidelines in the field of
construction quality, water management and dam safety.
The law on dams’ safety was issued in 2001, followed by the law on safety of
small dams in 2006 and the law on safety of dikes and levees in 2010.
∗ Gestion de la sécurité des barrages et des digues en Roumanie; passé, présent et futur
DOI: 10.1201/9781003211808-88 687

Q.105 - R.42
The immediate consequence was the decrease in the number of dams that
suffered incidents, accidents or failures in the last two decades. The same conclu-
sions extend in the case of tailings dams and dikes. With a new system of safety
management, even in the case of an increased frequency of the extreme hydrolog-
ical events, most of the hydraulic infrastructure performed according to the design
requirements.
In case of existing dams, dikes and levees the reassessment of safety status
needs to be performed considering the new hydrological conditions as a result of
climate change.
New assessments of safety status of the most important structures, say those
of grades A and B, should be made performing comprehensive quantitative risk
analyses.
Risk-based decision analyses could become the new tools to validate pro-
grams of measures to assure the safety status of dams and dikes, to repair
or upgrade the existing hydraulic structures and equipment. Such tools can be
extremely important in conditions of funding constraints, a major issue for our
country.
RÉSUMÉ
Le présent rapport présente brièvement certains des principaux aspects de

la gestion de la sécurité des barrages, y compris l’évolution historique, la légis-
lation, les standards et normes qui réglementent le domaine de l’ingénierie des
barrages, la surveillance des barrages et la gestion des risques. La contribution
du Comité national roumain des grands barrages (ROCOLD) à l’élaboration d’un
concept moderne de sécurité des barrages est soulignée.
Après 1990, le déclin important de la construction de nouveaux barrages a

coïncidé avec certains incidents et accidents survenus au niveau de ces systèmes
d’ingénierie très complexes. Ces circonstances difficiles ont conduit à l’élaboration
d’un ensemble juridique national complet de lois, de normes techniques et de direc-
tives dans le domaine de la qualité de la construction, de la gestion de l’eau et de
la sécurité des barrages.
La loi sur la sécurité des barrages a été promulguée en 2001, suivie de la loi
sur la sécurité des petits barrages en 2006 et de la loi sur la sécurité des digues en
2010.
La conséquence immédiate a été la diminution du nombre de barrages qui ont

subi des incidents ou des accidents au cours des deux dernières décennies. Les
688
Q.105 - R.42
mêmes conclusions étendent dans le cas des barrages de résidus miniers et des
digues de protection contre les inondations.
Avec un nouveau système de gestion de la sécurité, même en cas

d’augmentation de la fréquence des événements hydrologiques extrêmes, la plupart
des infrastructures hydrauliques ont resté en fonction conformément aux exigences
de conception.
Dans le cas de barrages, digues et digues existants, la réévaluation de l’état de

sûreté doit être effectuée en tenant compte des nouvelles conditions hydrologiques
résultant du changement climatique.
De nouvelles évaluations de l’état de sûreté des structures les plus impor-

tantes, par exemple celles des catégories d’importance A et B, devraient être
effectuées par des analyses de risque quantitatives complètes.
Les analyses de décision sur la base des risques pourraient devenir les nou-
veaux outils pour valider les programmes de mesures pour assurer l’état de sécurité
des barrages et des digues, pour réparer ou améliorer les structures et équipements
hydrauliques existants. Ces outils peuvent être extrêmement importants dans des
conditions de contraintes de financement, un enjeu majeur pour notre pays.
1. INTRODUCTION
The present paper briefly presents some of the main aspects of the dam safety
management, including historical evolution, legislation, standards and norms that
regulate dam engineering field, dam monitoring and risk management. The con-
tribution of the Romanian National Committee on Large Dams (ROCOLD) in the
development of a modern dam safety concept is emphasized.
In the last half century, Romania successfully erected more than 240 large
dams and 2000+ small dams. The high rhythm of building such a large number
of dams stimulated the development of a national school of dam engineering, with
remarkable achievements especially in the years 1960–1980.
systems [1]. Such difficult circumstances led to the development of a comprehen-
small dams in 2006 and the law on safety of dikes and levees in 2010.
689
Q.105 - R.42
The application of such a complex system of safety laws in the field of hydraulic
structures was possible creating a dam safety management system with detailed
aspects regarding all phases of a dam or dike lifetime - conception and design,
construction, commissioning, operation, monitoring, post-use or decommissioning.
2. BRIEF HISTORY
The recent inventory of dams identified some 2500 dams on Romania’s ter-
ritory that correspond to the definitions specified by the Water Law and the Dam
Safety Law. Among them, more than 2300 are permanent water storage dams with
one or several purposes and circa 190 are tailings dams (closed or active).
The inventory of dikes and levees reached to a total length of more than
11000 km flood protection structures along country’s inland rivers and along Danube.
The systematic construction of large dams in Romania began in the early

1950s, and in the debut period their safety was a national priority. The lack of expe-
rience in the dam construction field was effectively replaced by a special care for
field tests, by a thorough study of the international technical reports and by research,
finally leading to the “adoption” and adaptation of the best international practice. The
Romanian school of dams, founded in the 1950–1960 decade, successfully built the
first large dams that are reference works even by nowadays standards. The need
for the in-situ testing of the adopted design assumptions and analysis methods pro-
moted also – at the same interest and exigency degree – the monitoring of the new
dams. For each dam the monitoring started during the construction phase and was
carried on during the commissioning and in operation.
The decades 1970–1980 represented a period of experience consolidation in

the field of dams when new standards and norms were created.
State was the sole owner of large dams. These guaranteed an organized frame
for the dam safety management under all its aspects.
Sadly, beginning with the 1980s the lack of resources simultaneous with an
excessive proliferation of dam construction, resulted in a decrease of exigencies
regarding the quality and, hence, the dam safety.
In the first years of the 1990 decade, the changing political conditions in the
Romanian society imposed a new approach in the dam safety management. In
the first stage the responsibilities moved to the primary owners but without hav-
ing the legal provisions and an organized system to govern the safety issue. This
deficiency associated also with the lack of funds for maintenance, rehabilitation
and safety assurance of dams generated a situation of real hazard for the society
and economic assets, as well as the environment. The Community of dam engi-
neers became aware of the risk presented by such an evolution in due time. The
690
Q.105 - R.42
Romanian National Committee on Large Dams (ROCOLD), as the Forum of dam

engineers began the development of the new national dam safety policy. The reg-
ulating stages that started with the issuing of the Dam Safety Law in 2001 were
encouraged by the dynamics of the changes in property and responsibility. In the
said period, the dam owners were diversified being now national authorities or com-
panies, companies with state capital and/or private capital, local authorities and
even private individuals [1].
Similar evolution marked the development of dikes and levees safety legal
package. Most of the national flood protection infrastructure was built in the 1960–
1980 decades, with little concern regarding safety standards. The safety status of
this system of hydraulic infrastructure is nowadays an issue of concern.
3. THOSE PAST EVENTS THAT TRIGGERED PROGRESS
3.1. BELCI DAM FAILURE, 1991
Belci Dam in northeast Romania was a 18 m high clay core earthfill dam that
dammed the Tazlau River valley. The 415 m long dam consisted of three sections,
a central concrete gated structure dam, a 234 m earthfill dam to the left abutment
and a 126 m earthfill dam to the right abutment (Fig. 1) [2].
Fig. 1
Plan view and typical cross section through Belci earthfill dam [2]
Vue en plan et coupe transversale type du barrage Belci
691
Q.105 - R.42
On the terrace of the right abutment the dam was extended by a 577 m long
earth dike. The spillway comprised four overflowing bays equipped with 2.5 × 11 m
flap gates. The two central openings were provided with bottom outlets, equipped
with 2.5 × 11 radial gates. Belci Dam’s main use was water supply for a thermal
power plant and for the local industry.
During the exceptional flood in the night of 28/29 July 1991 in a reservoir more
than 75 % silted, with flows that massively exceeded the spillway’s routing capacity,
the dam was overtopped and a 120 m wide breach was generated. As a result the
reservoir was completely taken out of work. The mechanism of the earthfill dam
failure was the erosion of the downstream prism face, followed by the sliding of
downstream face and collapse of a dam sector.
The consequences of the failure of the 18 m high earthen dam were dramatic –
17 casualties, 119 destroyed households, severe erosion and clogging of down-
stream river bed.
The report of the Commission of Experts who investigated the accident pointed
out that:
– the dam failed because of an exceptional flood which could not be adequately
routed through the dam discharge system;
– the lack of an adequate hydrological forecast system and of a communications
and remote transmission system prevented the anticipated operation of the
electro-hydromechanical equipment for flood control;
– the relative high number of casualties and the significant material damages
were, to a large extent, due to failure to comply with the construction discipline,
by allowing a large number of households to be built in the floodable area
downstream of the dam, even for small return period discharge values, as
well as to the lack of preparedness for emergency situations of the people
downstream of the dam.
The dam is still under the same condition as immediately after the rupture
(Fig. 2), while several alternative solutions were analyzed, either for the reoper-
ation of dam after major repair and rehab works or for total decommissioning of
the dam.
3.2. THE TECHNICAL ACCIDENT AT THE AURUL – BAIA MARE TMF, 2000
The Aurul TSF is located in Northern part of Romania, on the right bank
of the Sasar River. It covers an area of 89 ha, in the final stage the volume
of tailings deposited being planned to reach some 15 million tons at a height of
17–18 m.
692
Q.105 - R.42
Fig. 2
Downstream view of the collapsed dam
Vue en aval du barrage après la rupture
1-breach 1- brèche
Aurul TSF is a field-type pond, with contour dams made of the coarser fraction
of tailings into the pond, separated by hydro-cyclone.
On January 30, 2000, at 10:00 p.m. on the Southern side of the pond, about
50 m East of the access dike to the drainage pipe, a breach of about 20 m was
formed, through which more than 100,000 m3 of cyanide contaminated water were
spilled from the pond (fig. 3). The water leaked along the road from the southern side,
mainly to the drainage channel, entering the Lapus River and then the Somes-Tisa-
Danube hydrographic network. As a result of the accident, a significant area of land
was affected and cyanides and heavy metals were poured into the watercourses.
No casualties were reported, but in 24 localities the water supply was

temporarily affected [3].
The technical commission that investigated the accident came out with several
findings and conclusions regarding the event:
693
Q.105 - R.42
Fig. 3
Baia Mare, 2000 TSF failure [3]
Rupture du barrage de stérile Baia Mare, 2000
– The defective design concept. Primarily, the operation of the TSF with the
integral water recirculation in a closed system was practically impossible. On
site precipitation is superior in relation to evaporation. As a result, the water
accumulation was the dominant process within the pond. In the absence of any
possibility to discharge excess water, the water table ascended, the beach was
compromised, and the freeboard became insignificant. Secondly, in the first
phase of development of the pond, the contour dike could not be formed to the
prescribed parameters due to the lack of coarse material and the difficulty of
running the hydro-cyclones during winter.
– Excessive rain water supply. Over the large volume of water already accumu-
lated in the pond, in the days preceding the accident, an excess volume of
about 100,000 m3 of rain water and snow melt overlaid.
– Lack of special monitoring. The safety conditions specified in the project and in
the operating regulations (beaches, freeboard, exterior slope of contour dike,
maximum volume of free water in the pond) were not monitored and recorded.
3.3. SMALL DAMS COLLAPSE DURING 2005 FLOODS
Many small dams failures were determined by the inadequate maintenance or

operational mismanagement. Only in the Arges – Vedea catchment’s area, covering
roughly 10% of the country’s surface, approximately 14 % of the small dams (68 out
694
Q.105 - R.42
of 504) suffered severe incidents or accidents during the floods in the year 2005,
resulting in uncontrolled breach failures as result of overtopping due to inadequate
spillway capacity [4].
Some of the main findings of the technical inspections to the failed small dams
portfolio pointed upon probable causes of these events:
– most of the small dams were built by local authorities or companies during
the decades 1950–1960 to provide minor uses such as farming, irrigations,
fishing. In most of the cases, these dams were built using poor design projects
and standards;
– most of the affected dams did not have operation instructions;
– most of the dam owners are small communities and small private companies
without any qualified staff to operate the equipment and to manage the dam
safety;
– in most of the cases the existing discharge facilities were blocked by debris or
by the damaged mechanical equipment. The insufficient discharge capacity
of the spillways resulted in dangerous rise of the reservoir water level, crest
overtopping, erosion of the downstream face and, ultimately, in the failure of
the dams;
– high degree of sedimentation of the storage reservoirs which reduced the
storage capacity and therefore flood control capacity;
– the cascade development with a “domino” effect in the case of upstream
reservoir failures.
Fig. 4 shows some examples of small dam failures during 2005 floods in Arges-
Vedea river basin, mainly due to causes previously mentioned.
3.4. DIKE FAILURE DURING 2006 FLOODS
During the 2006 floods the maximum observed levels on Danube exceeded
by more than 60 cm the maximum levels recorded after the completion of the flood
protection works in the 1970s.
The prolonged pressures to which the dikes were subjected was partially
reduced by means of several controlled breaches, which resulted in a reduction
of the Danube levels by up to 28 cm.
A series of uncontrolled breaches were produced by the exceedance of the

design levels by up to 125 cm and due to the long – lasting action of the pressures
corresponding to these levels upon dikes and their foundations (some examples:
Rast, Bechet, Spantov, Oltina, Ostrov).
It is worth mentioning that a series of dike failures occurred in the zones of old,
“dead” river branches, where hydrodynamic erosion processes occurred to affect the
dikes’ foundations.
695
Q.105 - R.42
Fig. 4
Example of small dams affected by 2005 floods in Arges-Vedea river basin
Exemple de petits barrages affectés par les crues de 2005 dans le bassin
d’Arges-Vedea [5]
The 2006 flood, with a “historic” peak value of 15800 m3 /s and significantly
surpassing the maximum flood levels on the Gruia – Calarasi sector (i.e. the
upstream Romanian Danube), resulted in inundations and serious damage to the
flood protection dikes and levees.
2006 floods had disastrous consequences due to the high volume of mate-
rial damage, represented by flooded households and land, severely damaged or
completely destroyed infrastructure, evacuated population etc.
The main cause of the damage was the failure of the dike system in sev-
eral sectors: Isaccea, Rast – Bistreţ, Rasova, Obretin, Oltina, Spanţov, Bachet –
Dăbuleni – Potelu etc. Fig. 5 shows images of some dike failures during the 2006
floods [6].
Some recent studies acknowledged that, due to the poor exploitation and
maintenance funds allocated in the last 20…30 years, the existing dikes and their
appurtenant structures and systems show a number of deficiencies, the most
important of which are:
– the crest elevation lower than designed. This is the case of flood protection
works built after 1962 downstream of Calarasi (i.e. lower Danube), to withstand
696
Q.105 - R.42
Fig. 5
Failure of Danube flood protection dikes and levees during 2006 [6]
Ruptures des digues de protection contre les inondations du Danube au cours des
inondations de 2006
to only Q10%. Some other sectors, initially designed at Q1%, are currently
lower with up to 0.50 m, mainly due to an inadequate maintenance.
– local degradation of the cross section;
– voids in forest protection curtains of dikes;
– sectors with soil soaking phenomena affecting materials from the embank-
ments;
– suffosion in dikes foundations, with downstream boils.
4. WHAT PROGRESS?
4.1. DAM SAFETY
After the failure of Belci dam, in 1991, the premises of legislative regulations
in the field of dam safety were emerged together with the setting up of the National
Commission for the Safety of Dams and Hydraulic Works (CONSIB, 1994), as an
advisory body for the Ministry of Environment, Waters and Forests (MEWF).
697
Q.105 - R.42
At about one year from the first actions undertaken by CONSIB, Law No.
10/1995 regarding the quality in constructions is promulgated. The new law basi-
cally regulates the entire field of constructions, including dams, levees, dikes. The
Government’s Decision No. 766/1997 introduced a series of procedures concerning
the quality in constructions supplementing and properly detailing the Law No. 10.
State Inspectorate in Constructions (SIC) was created, to strengthen the application
of the law and its procedures.
Some of the new procedures have direct implications in the field of hydraulic
structures, namely:
– the rating into classes of the importance;

– monitoring, in-time interventions and post-use of structures as essential parts
of quality system associated with any construction.
By different internal or joint orders of the Minister of Public Works, Development

and Administration (MPWDA) and MEWF respectively, additional procedures were
set up, such as
– the creation of a corps of experts in the field of dams and other hydraulic
structures;
– Methodology for ranking into classes of importance of the above mentioned
structures.
Specific features regarding the hydraulic structures and hydraulic develop-

ments, including dams were furthermore detailed in the new Water Law No.
107/1996.
Simultaneously with the efforts for application of Laws No.10/1995 and

No.107/1996 CONSIB and ROCOLD (between them it existed and still exists a sym-
biosis by the nominal structure of members and a synergy of concerns) continued
the efforts for a distinct legislative bundle regarding dams. As a result of these efforts
and, unfortunately, of the technical accident occurred at Baia Mare tailings dams,
with some consequences on environment too, Romania’s Government issued The
Emergency Ordinance No. 244/2000 regarding the safety of dams. This ordinance
is subsequently approved by Law No. 466/2001 and becomes what today is usually
addressed as The Dams’ Safety Law.
The Dam Safety Law includes some of the already existing regulations (dam
ranking, the methodology of dam safety assessment, dam monitoring) but brings a
substantial progress in dam safety management.
The Law is organized in five chapters and 31 articles. Among the main
provisions one can mention:
– the dam safety requirements are mandatory in all stages of a dam life, starting
with the dam design and ending with its decommissioning.
698
Q.105 - R.42
– dams are rated into categories of importance on the basis of a quantitative

evaluation of their associated risk; the procedure uses a system of criteria and
indices that take into account the dam characteristics, the dam behavior and
the dam failure consequences;
– tailings dams are regarded as a special type of dams, requiring the fulfilling of
all exigencies included in the present law;
– the dam rating into categories of importance is used for defining the dam mon-
itoring level, for prioritization of dam safety assessment within a dam portfolio,
for establishing the level of dam safety evaluation and control and for drawing
the owner obligation concerning the risk reduction;
– the dam owners are responsible for observing the safety requirements
provided by current legislation and state standards.
– Chapter 2 – “Dam operating conditions” – is dedicated to issuing permit for
dam construction, licensing of the dam operation and dam monitoring.
– Chapter 3 – “Dam safety control“ – empowers MEWF to control the fulfilling of
the legal obligation concerning the dam safety.
Starting with the year 2001 the existing state norms were completed and
supplemented based on the Law provisions. These technical norms dedicated to
hydraulic structures (NTLH) that follow from the Dam Safety Law were approved by
orders issued by MEWF.
By quantifying the risk components and final risk associated with dams
and industrial waste landfills, Romanian dams are classified into 4 categories of
importance (grades):
Grade A – dams of exceptional importance;

Grade B – dams of special importance;
Grade C – dams of normal importance;
Grade D – dams of low importance.
The inventory of dams as a comprehensive database containing all major

information for technicians and stakeholders was created – REBAR.
Under the conditions stated by the relevant methodology, the safety assess-
ment is performed by technical experts certified by the MEWF. The operational safety
assessment reports for dams of grades A and B are analyzed and validated by a
central Commission within the MEWF, while the similar reports for dams of grades
C and D are analyzed and validated within 4 regional Commissions organized by
the National Administration “Apele Române” (ANAR), which is the administrator of
the water resources of Romania. The approval and validation of the safety reports
for dams are concluded by issuing of the safety operation license, for periods of
maximum 7 years. The license renewal is preceded by a new safety assessment of
the dam under scrutiny.
Presently, the Dam Safety Law creates a complete and up-to-date legislative
framework for dam safety management based on [1]:
699
Q.105 - R.42
– a more transparent approach dealing with risk related to dams;

– responsibility on the dam owner for the overall management of the dam during
all phases of the dam existence;
– entrusting of safety assessment to dam experts with appropriate qualifications
and experience in applying best engineering practice and guidelines in the
field of dams; the responsibility concerning the dam operation conditions is
placed on the certified expert;
– the state control on the safety requirements;
– financial and legal penalties for the dam owner or other involved persons in
case of non-fulfillment of law stipulation.
4.2. SMALL DAMS SAFETY
During the severe floods that faced Romania in 2005 and 2006 human behavior
proved to be a significant element of dam failure risk. Simple mistakes, operational
mismanagement, or unnecessary errors interacted with flood hazards to lead to
failures of dams rated into grades C and D. (the importance category C and D
are usually attributed to small dams creating small reservoirs for fish farming and
recreational use).
Technical inspections at the affected dams pointed out on some of the main
causes for such a high number of failures, briefly summarized in section 3.3.
Consequently, an additional law that concerns the safely operation of storages

for fisheries, recreation and local utilities created by small dams of grades C and D
was issued – The Small Dams’ Safety Law (Law No. 13/2006).
The Law is organized in four chapters and 15 articles. Among the main
provisions one can mention [4]:
– Every small dam should comply with the requirements of the Dam Safety Law;
– In the case of flood warning operation of small reservoirs is conducted accord-
ing to the dispositions of the River Basin Authority (RBA) of ANAR that can
impose operation reservoir levels, preventive drawdown of the reservoir and
mechanical equipment maneuvers;
– Operation of all small storages can only be performed by qualified personnel
certified by the RBA of ANAR;
– Dam owners with no operation license have to fulfill these requirement in a
time frame established by the Water Authority and in the time interval before
licensing the storage will be operated with imposed constrains;
– If the dam owner does not acquire the operation license in due time the storage
will be decommissioned;
– In the case of small dams having the spillway capacity compromised by
unauthorized constructive measures or by impediment of the equipment the
700
Q.105 - R.42
reservoir will be emptied, even by means of breaches performed in dam’s body.

All costs are to be supported by the owner.
– The Water Inspection Service of the RBA verifies the operation license and
the operation rules, the technical state of the structures and equipment and
the communication system between the operation personnel and the RBA.
4.3. DIKES AND LEVEES SAFETY
Damage to important dikes and levees, with significant consequences for

population, property, socio-economic activities or the environment, led to profound
changes in the regulatory and legislative system associated with the safety of flood
defense embankments.
The past flood associated events pointed out the need for a proper dikes’ safety
management. The components of such a dike safety management system include:
– the census/inventory of dikes to know all relevant information about the safety
status of these works;
– creation of a package of technical and legislative regulations and norms that
will introduce rigorous discipline in all phases of the life of these works, starting
from the design phases, to the decommissioning phase.
– creation of new, state-of-the-art design norms to highlight the experience
gained by specialists during the numerous flood events in the last 25…30
years;
– creating a normative on the monitoring of dikes, using monitoring devices or
based on inspections;
– creation of a corps of experts to periodically assess the safety of dikes and to
make operational recommendations in flood situations.
– assessment of the safety status of flood defense dikes and levees and licensing
the existing dikes or the implementation of new ones.
As a consequence, Law No. 259/2010 – Dikes’ Safety Law – was issued,

followed by the regulatory Norm for the assessment of the safety status of flood
defense dikes and levees, NP 131-2011.
In a quite similar approach as the one used in the dam safety law, i.e. quantify-
ing the risk components and final risk associated with dikes and levees, Romanian
dikes are classified into 4 categories of importance (grades), A to D, A including the
most important structures.
In recent years, technical norms for the implementation of the dikes’ safety law
were issued and the National Register of Dikes in Romania, REDIG, was created.
Licensing of dikes and levees is mandatory for the owners and the whole
procedure is organized within a Central Commission at the level of MEWF for grades
701
Q.105 - R.42
A and B dikes and within 3 Regional Commissions at the level of ANAR for grades
C and D dikes.
The approval and validation of the safety reports for dikes and levees are
concluded by issuing of the safety operation license, for periods of maximum 10
years. The license renewal is preceded by a new safety assessment of the dike
under scrutiny.
5. SAFETY MANAGEMENT SYSTEM
5.1. MONITORING OF DAMS
The role of dam monitoring within the safety management, well established
nowadays, is the result of an evolution that in Romania lasted for half a century. The
tracing of this evolution can be done by analyzing the changes recorded during time
both in dam instrumentation with measuring devices and in the data processing and
normal behavior assessment as well.
The dam monitoring is made by observations and measurements. The neces-

sity of a special accuracy in making these operations is obvious but is not enough.
The role of monitoring in safety management lies in detection as early as possible
of any phenomena not so abnormal or dangerous but first of all unusual, atypical. In
such a way there become possible additional measurements and analyses and even
measures for risk reducing. Hence, it results a special importance of processing and
interpretation of data collected within the dam behavior monitoring [8].
The Dam Safety Law specifies that dam monitoring is organized as a three-
level activity (Fig. 6):
– level I is performed at dam site and consists in visual inspections, measure-

ments and primary processing of the collected data as well as immediate
analysis of the results;
– level II is a periodical synthesis of observations and measurements as well as
of the annual technical inspections and their interpretation from standpoint of
dam safety; this synthesis is made by care of the owner and is performed by
qualified dam engineers who draw up safety assessment reports;
– level III is represented by the analysis and approval of the annual safety
assessment reports made by a dam monitoring commission whose structure
is agreed by CONSIB and approved by the minister in charge with the water
management [1].
The existing dams can be operated in normal state or in exceptional state.

The normal state is characterized by normal values of the external loads (levels in
reservoir, inflows or outflows, temperatures etc.), by an appropriate operation of all
702
Q.105 - R.42
Fig. 6
Organizational levels of the dam monitoring system [1]
Les niveaux d’organisation du système de surveillance du barrage
component parts and by a structural response corresponding to that predicted under

the acting loads. The non-compliance of any of these conditions brings about the
coming into exceptional state.
Within the exceptional state there are distinguished several degrees depending
on the seriousness of the deviation from the normal state and on the created risk:
state of attention, state of alert and state of alarm.
The warning criteria that settle the boundaries between the different dam oper-
ation states affect also directly the monitoring schedule. They are established in view
of an immediate application (without being expected the results of some additional
analyses) and are specified in the monitoring project that may be part of the operation
regulations [1].
The monitoring activity may be improved by automated data acquisition

systems, particularly for behavior parameters which are main safety indicators.
Automation leads implicitly to the enhancement of the volume and quality of the
database and also speeds up the diagnosing of the structure state if the flowsheet
is supplemented with monitoring forecast programs using behavior models.
The main owners of large dams in Romania, i.e. the National Administration
“Apele Române” (ANAR) and “Hidroelectrica” S.A., respectively implemented inte-
grated informational systems for monitoring and assessment of the behavior of the
hydrotechnical structures.
5.2. RISK MANAGEMENT
The reducing of risk associated to a certain dam is made both by increasing

of its safety and by reducing the potential consequences in case of its failure.
703
Q.105 - R.42
Fig. 7
Connection between the legislation, classification systems and risk management.
La connexion entre la législation, les systèmes de classification et de gestion des
risques [1]
The increase of safety is the objective of the safety management and, as it

was stated, comprises an aggregate of structural and non-structural measures. For
the dams being in operation the basis of the safety management lies in monitoring
of their behavior and in periodical safety assessments. The latter ones must confirm
that the dams are safe or, depending on the case, they should identify the safety
deficiencies and start efficient measures to prevent the evolution towards failure.
A synthesis of the risk reduction issue is presented in Fig 7 [1], [8].
The risk identification is explicitly required by the content of the safety assess-
ment report. The safety evaluation is made based on the safety criteria stipulated by
norms (STAS), which depend on the class of importance of the dam. The time inter-
vals between successive safety assessments and the level of their appraisal depend
on the category of importance. The risk control includes the safety evaluation, the
continuous dam monitoring, the state control and the emergency preparedness
plans.
Prior to issuing of the Dam Safety Law, the safety assessment was made
strictly based on the requirements provided by norms and standards.
At present, according to the provisions of law, the assessment of safety for

dams in operation is given to some experienced experts and their conclusions are
704
Q.105 - R.42
Fig. 8
Limits of tolerable risk. ALARP principle – risks as low as reasonably practicable
Limites du risque tolérable. Principe NBRF - niveau des risques le plus bas
raisonnablement faisable [1]
based not only on the norm provisions but also on the national or international
engineering practice. The safety assessment is made based on some analyses
of vulnerability and risk, considering all safety parameters, on expert own liability.
The decisions concerning structural corrective measures or operational constrains
are discernibly made considering the excessive costs generated sometimes by the
formal application of the norm provisions when the basic safety is not affected.
A new standard (NP 132-2011) was issued in order to decide whether the
existing risks associated to a certain dam are acceptable or tolerable and the present
risk control measures are adequate. The standard establishes the principles and the
methods that are to be used for risk analysis, risk quantification, risk evaluation and
risk control.
The standard promotes the acceptable/tolerable risk as a relationship between

the probability of failure and the number of fatalities due to a dam failure. For new
dams and in the case of upgrading of existing dams, the standard ensures that the
average risk of death to particular members of the public from dam failure does not
exceed 10−6 per annum. Do not subject any person to a risk greater than 10−5 per
annum. For existing dams individual risk can be up to 10 times larger than the ones
for new dams but it is subject of the principle “as low as rational possible” (ALARP).
The accepted F-N curve is presented in Fig. 8. The upper curve (Limit) refers
to existing dams. As originally conceived, the lower curve (Objective) refers mainly
705
Q.105 - R.42
to new dams. For an existing dam with risk levels higher than the Limit, compliance
would require that the risk be reduced to a value lower than the Limit, the acceptable
value being determined by application of the ALARP principle [1].
For the economic loss the standard accepts a risk rate of 104 Euro per annum
that corresponds to a probability of failure up to 10−4 if the total damage is less than
108 Euro and to a probability of failure up to 10−6 if the total damage is less than
1010 Euro.
6. PRESENT AND PERSPECTIVE
In the last two decades an intensive regulatory program was developed and
put in force in Romania. The Dam Safety Law (2001), The Small Dams’ Safety
Law (2006), The Dikes’ Safety Law (2010) and the subsequent technical norms and
guidelines created a sound framework for an efficient control in the field of safety of
dams and flood protection systems.
Almost the entire portfolio of grades A and B dams are regulated on the bases
of comprehensive safety assessments.
The portfolio of grades C and D dams is much consistent and diverse. Among
the 2200+ small dams only some 70% could be appropriately regulated according to
the existing safety laws and norms. There are still difficulties in regulating all the small
dams since most of these structures are privately owned and operated. However,
even in the case of the non-regulated dams the Water Inspection Service from ANAR
deployed a vast and permanent campaign of safety checks thus allowing the dam
safety commissions to be acknowledged on the safety status of every small dam.
suffered incidents, accidents or failures in the last two decades. As an example,
only for Arges-Vedea river basin the number of failed dams decreased from over
60 to maximum 3–5 affected dams during 2006–2020 floods. After the catastrophic
flood events of 2005, the repair and proper operation of spillways and bottom out-
lets, even of small storage dams, but also a modern system of safety regulation
and rules of coordinated operation of reservoirs significantly reduced flood related
consequences. Thus, the flooded area was moderate and the risk was maintained
at acceptable levels. Same evolution was noticed throughout the country.
Due to its natural resources, Romania has developed a strong sector of extrac-
tive industries, especially between 1960 and 1990. After 1990, but especially after
joining the EU, a large number of mining facilities, especially TSFs, had to follow
closure and ecological restoration or compliance programs.
Many of TMFs ceased to operate, are closed and, in some of the cases, even
having different ecological restoration works [3].
706
Q.105 - R.42
A small number of tailings ponds (circa 10% from total) are declared as active
having a permanent or a seasonal operation program.
For 33 TMFs the closure and ecological restoration works were completed.
Now these TMFs have post-closure monitoring programs.
By the legislation in force, it has been imposed an appropriate monitoring

of every closed TMF for a period of minimum 30 years. Periodical controls are
performed by water inspectors.
Some 21 TMFs are in a conservation state, with approved closure projects

awaiting the commencement of the closure works.
Closing and ecological restoration projects involve huge funds, which can not
be ensured only from the state budget. Over the years, closed projects financed by
the World Bank or the Environment Fund have been promoted, but there is still a
significant shortage of funds.
Lack of funding can affect such works in all stages of development: in

operation, by not providing the necessary resources for a good performance of
the activities; during the closure and ecological restoration works, due to delays
in progress; during the post-closure monitoring programs, with the possibility of
incidents and accidents, sometimes with cross-border effects.
Speaking of dikes and levees, huge efforts were performed to accomplish their
inventory, gaining all the required information to appropriately describe every flood
protection structure.
Due to the large costs associated with the survey, geotechnical investigations,
hydrologic, hydrogeologic and hydraulic analyses, the progress of dikes’ regulation
is rather slow.
Although a proper flood risk management has to be more and more based on
non-structural measures, the reality in Romania shows that structural measures still
have a primary position in controlling the flooding associated risks.
The large number of population in floodplain areas, private land ownership

and the insurance system for property and economic assets in floodplains make the
non-structural measures difficult to promote.
Under these circumstances, the structural measures, based on engineer-

ing solutions for developing water courses, but also on dikes, levees, dams for
permanent or non-permanent storage reservoirs, appear as the best alternatives
707
Q.105 - R.42
that ensure the protection of population, property, economic assets and the envi-
ronment in case of extreme hydro-meteorological phenomena, the frequency of
which has sensibly increased in the last 10-15 years.
It is important to notice that the greatest damage is still generally reported in

areas with non-engineered water courses or without storage reservoir dams and
flood protection dikes and levees [5].
It is the duty of water engineers to ensure the development of new and suitable
river development schemes and solutions, in which water resource management has
to be correlated with the need to guarantee all water uses but also with a modern
flood risk management [9].
In the same time, it is necessary to reconsider the structural and non-structural

solutions in terms of adaptation to climate change, which can lead to the devel-
opment of new multipurpose hydraulic works, adjustment of existing structures,
reconsideration of norms and design standards for new hydrotechnical construc-
tions, but also changes in operation rules or allocation of water volumes within the
existing storage reservoirs as possible non-structural measures.
In case of existing dams, dikes and levees the reassessment of safety sta-
tus needs to be performed considering the new hydrological conditions as a result
of climate change. For our country specialized studies concluded that in some
regions the hydrologic data are modified consequently challenging the design
projects for the majority of country’s dams, especially of the large dams, of grades
A and B.
Special attention must be allocated for a better adaptation of storage reservoirs

to climate change, even if the safety of the dams still proves to be fulfilled. Reopera-
tion of reservoirs is a possible countermeasure to mitigating adverse effects and for
a better adaptation to new conditions [10]. Depending on reservoir’s uses, one must
re-think how to operate the hydraulic facility, to maintain the initially designed uses
or to add new uses, not planned at initial stages of design, such as fish farming,
leisure, irrigation, navigation.
After 20 years of evolution of the dams’ and dikes’ safety legal and technical
provisions, the new assessments of safety status of the most important structures,
say those of grades A and B, should be made performing comprehensive quantitative
risk analyses, with the aim of evaluating the actual risks and to propose possible
measures to contain risks in an acceptable range.
grams of measures to assure the safety status of dams and dikes, to repair or
upgrade the existing hydraulic structures and equipment controlling inflow and/or
outflow.
708
Q.105 - R.42
7. CONCLUSIONS
In the last 50–60 years, Romania successfully built more than 240 large dams
and 2000+ small dams. Such an intense pace stimulated the development of a
national school of dam engineers, with remarkable achievements especially in the
years 1960–1980.
systems. Such difficult circumstances led to the development of a comprehen-
small dams in 2006 and the law of dikes’ and levees’ safety, in 2010.
suffered incidents, accidents or failures in the last two decades. The same conclu-
sions extend in the case of tailings dams and dikes. With a new system of safety
management, even in the case of an increased frequency of the extreme hydrolog-
ical events, most of the hydraulic infrastructure performed according to the design
requirements.
In case of existing dams, dikes and levees the reassessment of safety status
needs to be performed considering the new hydrologic conditions as a result of
climate change.
After 20 years of evolution of the dams’ and dikes’ safety legal and technical
provisions, the new assessments of safety status of the most important structures,
say those of grades A and B, should be made performing comprehensive quantitative
risk analyses, with the aim of evaluating the actual risks and to propose possible
measures to contain risks in an acceptable range.
grams of measures to assure the safety status of dams and dikes, to repair or
upgrade the existing hydraulic structures and equipment controlling inflow and/or
outflow. Such tools can be extremely important in conditions of funding constraints,
a major issue for our country.
REFERENCES
[1] STEMATIU, D. ABDULAMIT, A., CONSTANTINESCU, AL., “Recent aspects

on dam safety management in Romania”, Proc. 11th ICOLD European Club
Symposium, 2–4 Oct. 2019, Chania, Crete, Greece.
709
Q.105 - R.42
[2] ASMAN, I., BRATIANU, G., “Project for rehabilitation of Belci Dam after the
1991 failure”, Proc. 9th ICOLD European Club Symposium Sharing experience
for safe and sustainable water storage, Venice, Italy, 10–12 April 2013.
[3] ABDULAMIT, A., GROZEA, M., “Tailings dams in Romania”, Proc. Interna-
tional Symposium “Sustainable and safe dams around the world”, 87th Annual
Meeting of the International Commission on Large Dams, Ottawa, Canada,
9–14 June, 2019, editors Jean-Pierre Tournier, Tony Bennet, Johanne Bibeau,
CRC Press/Balkema, ISBN: 978-0-367-33422-2, pp 3437–3446.
[4] STEMATIU, D., ABDULAMIT, A., “Romanian Small Dams Safety Law”,
Proc. 8th ICOLD European Club Symposium Dam Safety-Sustainability in a
Changing Environment, Innsbruck, Austria, Sept. 2010.
[5] ABDULAMIT, A., “Role of dams and levees in the flood risk management in
Romania”, Proc. International Symposium “Sustainable and safe dams around
the world”, 87th Annual Meeting of the International Commission on Large
Dams, Ottawa, Canada, 9–14 June, 2019, editors Jean-Pierre Tournier, Tony
Bennet, Johanne Bibeau, CRC Press/Balkema, ISBN: 978-0-367-33422-2, pp
2603–2612.
[6] ABDULAMIT, A., STEMATIU, D., “Flood protection dikes founded on weak
soils. The case of lower Danube dikes and levees in Romania”, Q103 – R6,
26th International Congress on Large Dams, Vienna, Austria, 1–7 July, 2018.
[7] ABDULAMIT A., “Safety of Dams in Romania”, Q91 – R12, 23rd International
Congress on Large Dams, Brasilia, Brazil, May, 2009
[8] ABDULAMIT, A., STEMATIU, D., “New legal provisions concerning dams and
dikes safety in Romania”, Proc. 9th ICOLD European Club Symposium Shar-
ing experience for safe and sustainable water storage, Venice, Italy, 10–12
April 2013.
[9] ABDULAMIT, A., RÃDULESCU, D., “Aspects of water resources management

in Romania”, Proc. 9th ICOLD European Club Symposium Sharing experience
for safe and sustainable water storage, Venice, Italy, 10–12 April 2013.
[10] ABDULAMIT, A., STOICA, D., SÎRBU, N., “Reconsideration of Operation Rules
for an Existing Dam Reservoir Based on Advanced Flood Routing Analyses”,
Proc. International Symposium “Knowledge Based Dam Engineering”, 85th
Annual Meeting of the International Commission on Large Dams, Prague,
Czech Republic, 3–7 July, 2017.
710
Q.105 - R.43
-------
-------
CASES OF RUPTURE OF DAMS AND LESSONS (∗)
Geraldo PEREIRA MAGELA

Consultant Engineer,
Rio de Janeiro
BRAZIL
SUMMARY
This work presents the author’s opinion after studying some cases of dam’s
ruptures in Brazil in the last 40 years, namely the Euclides da Cunha and Limoeiro
dams in the State of São Paulo, the Camará dam in the State of Paraíba, and
the Algodões dam in the State of Piauí, as well as the lessons learned from these
accidents.
RÉSUMÉ
Ce rapport présente l’opinion de l’auteur après avoir étudié quelques cas de

ruptures de barrage au Brésil au cours des 40 dernières années, à savoir les bar-
rages Euclides da Cunha et Limoeiro dans l’État de São Paulo, le barrage de Camará
dans l’État de Paraíba et le barrage d’Algodões dans l’État de Piauí, ainsi que les
leçons tirées de ces accidents.
∗ Cas de ruptures de barrages et leçons
DOI: 10.1201/9781003211808-89 711

Q.105 - R.43
1. INTRODUCTION
It was only after the rupture of some famous dams in the 60s and 70s, Mal-
passet in France in December 1959 and Teton in the United States of America in
June 1976, that several countries began to study and plan systematic measures to
increase safety of its dams [1]. This study revealed that 59% of accidents are due to
geological-geotechnical causes, 23% to hydrological causes and 12% to deficient
construction, which totals 94%. The other 6% is due to acts of war (3%), improper
operation (2%) and earthquakes (1%).
On December 2, 1959, the Malpasset concrete dam broke on the Reyran River,
12 km from Fréjus, by displacing a large block of rock at the foundation of the dam
on the left bank. This rupture caused the destruction of the villages of Malpasset
and Bozon and more than 400 victims [2, 3].
On June 5, 1976, the Teton earth dam broke on the Teton River, Idaho, 100 m
high, by piping in the right abutment during the first filling. This famous rupture,
published in several references, killed 11 people and caused a huge financial loss [3].
On January 19, 1977, the earth dams of Euclides da Cunha and Armando
Salles de Oliveira broke on the Pardo River, west of the State of São Paulo, Brazil [4].
On June 19, 2004, the Camará dam on the Mamanguape River, in the State
of Paraíba – Brazil, broke due to a notable discontinuity on the left bank in the
foundation of the dam [5].
On May 19, 2009, the Algodões dam on the Piranji river in the State of Ceará
– Brazil broke through, due to erosion of the embankment downstream of the dam’s
back wall on the right abutment [5].
From these accidents, the process of raising awareness among proprietary

companies regarding the safety of dams in Brazil was long and persistent. In 1999
the Brazilian Committee on Dams – CBDB issued a Basic Dam Safety Guide [6].
2. EUCLIDES DA CUNHA AND LIMOEIRO DAMS
On January 19, 1977, on the Pardo River, near the cities of Caconde and
Mococa, west of the State of São Paulo, a water spout overcame the most intense
rain so far in the region, which was 230 mm in 24 hours. The center of the storm
reached the small reservoir of the Euclides da Cunha dam (1.0 km2 and 13 million
cubic meters), which is located 12 km upstream of the Limoeiro dam (officially named
after Armando Salles de Oliveira).
712
Q.105 - R.43
Fig. 1
Euclides da Cunha Dam Break
Fig. 2
Limoeiro Dam Break
The 56 m high dam was built between 1958 and 1960. The spillway, with two
12.00 m × 12.50 m gates, had a flow capacity of 2,100 m3 /sec with the reservoir at
maximum elevation. The free edge was 2.50 m.
The peak of the exceptional flood was 3,100 m3 /s, 50% higher than the spillway
design flow. The reservoir’s water level rose rapidly, overtopped and broke the dam
713
Q.105 - R.43
Fig. 3
Euclides da Cunha rebuilt including additional morning glory spillway upstream of
the original spillway
Fig. 4
Reconstructed Limoeiro dam including additional spillway in the right bank
714
Q.105 - R.43
Fig. 5
Top view of the dam already broken
in 4 hours. The water height over the dam was 1.2 m. The flood wave propagated
at a speed of 20 km/h to the downstream Limoeiro reservoir, overtopped the 35 m
high dam that broke after 20 minutes.
This accident leads us to conclude that the hydrological studies of extreme

flows were inaccurate as it did not predict the occurrence of a flood 50% greater
than the project discharge from the spillway. It should be noted that according to
Dams and Public Safety [3], this type of accident represents 23% of rupture cases.
Soon after the accident, the dams were promptly rebuilt, including complementary
spillways.
3. CAMARÁ DAM
The Camará dam, 50 m high and 300 m long on the crest, was built by the
Paraíba State Water and Sewer Company (CAGEPA) on the Mamanguape River,
between 2000 and 2002, to supply the city of Alagoa Grande.
The rupture occurred 1.5 years after construction, still during the filling of the
reservoir, due to the sliding of blocks along a discontinuity, left inexplicably after
excavations, in the foundation of the dam. The reservoir’s water level did not reach
the spillway crest – width = 39 m; Q = 270 m3 /sec for a head = 2.3 m; (see Figure 6).
The material of this zone after flooding for some time lost strength and under
water pressure it slid. The dam was without base for some time until it totally
collapsed.
The dam was rebuilt with the same characteristics between 2012 and 2016 –
see the Figures 10 to 12. The volume of the reservoir is in the order of 27 million
cubic meters.
715
Q.105 - R.43
Fig. 6
Upstream view of the dam already broken
Fig. 7
View of the dam before total rupture.
716
Q.105 - R.43
Fig. 8
Detail of the remnant of the foundation after sliding.
Fig. 9
Downstream view of the fully left part of the dam broken.
717
Q.105 - R.43
Fig. 10
Downstream view of the reconstruction.
Fig. 11
Downstream view of new dam.
4. ALGODÕES DAM
The Algodôes dam was built by the DNOCS (National Drought Works Depart-
ment) on the Piranji River, between 1996 and 2004, to supply the city of Cocal in
the Piauí State, north region of Brazil.
718
Q.105 - R.43
Fig. 12
Lateral view of new dam.
The dam broke in May 2009 (Sampaio, 2014). The site consists of sand-
stones from the Serra Grande formation, outcropping on the right abutment with
large amounts of talus. The layout is shown in the Figure 13.
The earth dam is 47 m high and 278 m long. The work has a 220 m long water
outlet in a gallery, diameter of 900 mm with a dispersing valve, with a flow capacity of
1.60 m3 /sec. The 120 m long side spillway has a capacity of 720 m3 /sec (TR = 100
years), with a head on the crest of 2.47 m.
After the start of the spillway operation, the embankment downstream of the
dam’s back wall was attacked by the flow. Erosion regressed and hit the concrete
plates of the wall. Access to the site was prevented because erosion disrupted the
road. The rains intensified and the increase in spills accentuated the process until
the dam burst on May, 2009 [5].
5. CONCLUSIONS
Based on the Brazilian experience acquired with the cases presented in this
work, as well as others, the following lessons can be extracted:
– that the projects must be prepared by teams with proven experience in works
of this nature, strictly observing what is specified in the Manual of Criteria for
Civil Project for Hydroelectric Plants (Eletrobras/CBDB, 2003);
719
Q.105 - R.43
Fig. 13
Algodões dam. Layout. Note the side spillway.
– One should not reduce or save on the budgets of hydrological and geological-
geotechnical investigations that support the basic studies of the projects of the
component structures of the projects;
– traditional basic studies of geology, geotechnics, hydrology and hydraulics
must be carried out in great detail, that is, in depth.
720
Q.105 - R.43
Fig. 14
Broken Algodões dam.
It is believed that by doing so, the likelihood of errors, accidents and disasters
will be further reduced.
Finally, the author notes that, in his opinion, dam owners should be legally
obliged to publish their failures, making all information about the accidents available
in full, which would enable others to learn the lessons and evolve the state of the art
of the issues involved.
REFERENCES
[1] Magela, G. P. Description of cases of dam ruptures from the 1950s world-
wide. Conclusion work of the Professional Master’s Course in Defense and
Civil Security at Universidade Federal Fluminense, Niterói-Rio de Janeiro
(2017).
[2] Bellier, J. The Malpasset Dam. The Evaluation of Dam Safety. ASCE, Ameri-
can Society of Civil Engineers. Asilomar Conference Grounds. Pacific Grove,
California (1976).
[3] USBR, United States Bureau of Reclamation. Dams and Public Safety (1983).
721
Q.105 - R.43
[4] BCOLD, Brazilian Committee on Dams. Design, Construction and Perfor-

mance (1982).
[5] Sampaio, M. V. N. Earth Dam Safety. An account of the experience in Piauí

State (Brazil). Masters dissertation. Federal University of Ceará (2014).
[6] BCOLD, Brazilian Committee on Dams. Basic Dam Safety Guide (1999).
722
Q.105 - R.44
-------
-------
THREE PRACTICAL EXAMPLES OF RIGID INCLUSIONS IN

CONCRETE DAM FOUNDATION (∗)
Ricardo ABRAHÃ
RICARDO ABRAHÃO GEOCIENCIAS, São Paulo
Alex CALCINA
INTERTECHNE, Curitiba
BRAZIL
SUMMARY
The influence of low strength and low stiffness geological discontinuities within
the rock mass, few meters below the dam foundation, plays an important role on
the concrete dam sliding stability assessment. If this scenario does not fulfil the
design criteria in terms of safety factor, one of the possibilities of dealing with such
a conditioning is introducing a rigid body in this discontinuity. This inclusion will lead
to a stress distribution in which great part of the shear stresses is taken by this
body. Since all inclusions are built in concrete they are also known as “concrete or
shear keys”. The stability evaluation has been historically assessed and certified by
means of rigid body strain and stress analysis and the corresponding forces evalu-
ated through equilibrium limit methods. Mathematical models have been employed
since the 70’s as a helping tool to support final decisions. Three special cases were
chosen, starting with the main dam of Itaipu power plant, built in the 70’s, almost
200 m high, followed by Teles Pires power plant and Camará water supply dam, both
built in the beginning of the 21st century.
∗ Trois exemples pratiques d’inclusion rigide dans la fondation de barrage en béton
DOI: 10.1201/9781003211808-90 723

Q.105 - R.44
RÉSUMÉ
L’influence des discontinuités géologiques de faible résistance et de faible

rigidité dans roche, à quelques mètres au-dessous de la fondation du barrage, joue
un rôle important sur l’évaluation de la stabilité au glissement des barrage en béton.
Si cette condition ne respecte pas les critères de calcul en termes de facteur de
sécurité, une des possibilités pour faire face à une telle condition est l’introduction
un corps rigide dans cette discontinuité. Cette inclusion conduira à une répartition
des contraintes, où une partie considérable des contraintes de cisaillement sera
absorbée par ce corps. Puisque toutes les inclusions sont construites en béton,
elles sont également appelées «clés en béton ou en cisaillement». L’évaluation de
la stabilité a été historiquement évaluée et certifiée au moyen d’une analyse de
déformation et de contrainte de corps rigide et les forces correspondantes évaluées
par des méthodes de d’équilibre limite. Les modèles mathématiques sont utilisés
depuis les années 70 comme un outil d’aide aux décisions finales. Trois cas parti-
culiers ont été choisis: le barrage principal de la centrale hydroélectrique d’Itaipu,
construit dans les années 70, avec près de 200 m de hauteur; la centrale hydroélec-
trique de Teles Pires et le barrage d’alimentation en eau de Camará, tous deux
construits au début du 21e siècle.
1. INTRODUCTION
Limit Equilibrium Methods (LEM) for sliding stability analysis have limited appli-
cation and are more reliable for the foundation at the concrete x rock contact where
stress distribution might be approximately assessed by means of rigid body theories.
Below this level, in the presence of geological features with complex behaviour,
the reliability can only be assessed by mathematical modelling.
For those used to employ LEM to certify stability analysis, it is possible to

transfer mathematical modelling data to LEM equations.
2. ITAIPU POWER PLANT
The Itaipu hydroelectric complex is located on a stretch of the Paraná river

forming the boundary between Brazil and Paraguay, as a program of cooperation
from both countries.
The project includes a hollow-gravity concrete structure with water intakes,

known as the main dam, just upstream of the powerhouse complex. The main dam
724
Q.105 - R.44
is flanked by concrete buttress structures as wing dams, with rock and earth fill dams
at their ends.
All structures overlie a trapp formation which consists of a series of upper

Mesozoic basaltic lava flows spread all over the Paraná sedimentary basin.
The whole rock mass is cut by a series of planar, nearly horizontal disconti-
nuities, which have regional expression. These discontinuities almost always follow
the contacts between flows. Otherwise they appear at the flow cores. On the val-
ley bottom, associated to the horizontal features, there are some levels of shearing
zones developed during the erosion of the valley. The unloading due to valley exca-
vation originated stress concentrations which induced buckling related phenomena
justifying the presence of such shearing features.
The highest blocks of the main dam of Itaipu are founded at the very bottom
of the Paraná river valley.
The engineering aspects of this part of the dam foundation includes some
levels of main discontinuities, being that of elevation 20 m the most important. It is
smooth, a few centimetres thick, presenting clay gouge in some parts and rock to
rock contact in others. This level also shows signals of slight differential movements.
Upper joints show characteristics of faults with milonite between walls and signs of
heavy shearing movements.
Due to the major difficulties of prospecting these features, located below water
level at the bottom of the river, the confirmation of their existence and the deter-
mination of their geomechanical characteristics could only be achieved after river
diversion, when construction was already underway. However, as soon as accesses
to these discontinuities were available, their strength and deformability parameters
have been estimated, based on direct shear tests and on the experience of experts
in the field of rock mechanics.
Figure 1 shows a simplified geological model of the foundation of the central

blocks, in which successive basaltic lava flows are represented. The location of the
main discontinuities is also indicated. For sliding stability analyses, the elevation
20.00 discontinuity was the most important geological feature due to its persis-
tence (covering the transversal length of the entire intake-powerhouse system),
smooth and low shearing strength (average friction angle φ = 25◦ and practically
no cohesion).
Although seemingly incompatible, the solutions for each aspect, i.e., deforma-
bility, shear strength and construction needs have converged into a final decision.
A fit solution would have been to remove the weakest part of the rock foundation
(down to the elevation 20.00 discontinuity) through open-pit excavation.
However, owing to the time aspects, such as construction schedule, this solu-
tion could not be applied, and consequently, underground work was adopted instead.
725
Q.105 - R.44
Fig. 1
General interpretation of the geology and structures,
with indication of the shear keys. Elevations in meters.
This procedure allowed dam construction and foundation treatment to be carried

out simultaneously. The final solution led to the construction of a system of concrete
shear keys, forming a grid pattern at elevation 20.00.
Due to their shear stiffness, much higher than that of the discontinuities, these
concrete keys would be able to reduce the tangential deformability of the foundation
to an acceptable level. Furthermore, the strength of both shear keys and remaining
untreated sections of the elevation 20.00 discontinuity lead to achieve the sliding
stability margin of safety required by the design criteria.
The design of the central stretch of the intake-powerhouse system of Itaipu

has been made in three successive phases, regarding its sliding stability.
During the first phase, the following loads were considered for limit equilibrium
and two-dimensional finite-element analyses: dead weight of the concrete dam and
powerhouse structure; hydrostatic pressure at the upstream face of the dam and at
the downstream face of the powerhouse structure, with water levels as indicated in
Figure 1; uplift pressure at elevation 20.00 level; equivalent static loads for taking into
account the seismic effects (horizontal acceleration equal to 0.05 g); water pressure
at a virtual vertical crack in the rock mass, from the heel of the dam down to the
discontinuity.
726
Q.105 - R.44
Geomechanical data for finite element models were as follows:
a. Rock masses idealized as no-tension, isotropic elastic media, with

E = 20 GPa for dense basalt, and E = 15 GPa for vesicular basalt and basaltic
breccia; Poisson’s ratio ν = 0.2 for both cases.
b. Elevation 20.00 discontinuity idealized as an elasto-plastic joint, with

ks = 0,05 GPa/mm and kn = 0.15 GPa/mm; strength parameters, φ = 25◦ and c = 0.
The elastic parameters adopted for the concrete were E = 30 GPa and ν = 0.2.
The results of the first numerical analyses showed that the safety margin
against sliding as required by the design criteria (which stipulated the partial safety
factor of 1.5 for the friction coefficient of the joint) could not be fully achieved for the
central blocks of the main dam unless some design modifications were introduced.
In order not to change the geometry of the dam, it was decided to improve the
drainage system, so as to reduce to a minimum the uplift forces. This decision was
the starting point of the second phase of the foundation design.
However, in order to remain on the safe side, it was not considered the lateral
collaboration of the adjoining rock masses when studying the sliding stability condi-
tions of the central blocks. The new drainage system included a pair of transversal
tunnels at the elevation 20.00 discontinuity, connected to two longitudinal ones, so
as to form a rectangular frame. The uplift pressure in the area was thus considerably
reduced. For the purpose of two-dimensional stability analyses, the assumption was
made of an uplift pressure corresponding to the difference in elevation between the
foundation level and the discontinuity level, with linear variations in hydrostatic head
towards the reservoir and tail water levels. With this premise, the safety factor of
1.5 was attained in the case of normal loading condition, but not for the exceptional
loading condition, which included the seismic effect. The design decision was thus
reached considering a construction of a net of concrete shear keys under the central
four highest blocks of the dam, in order to improve the strength characteristics of
the discontinuity and to increase the overall stiffness of the foundation.
In the third phase of the foundation design, the easy access to various sections
of the discontinuity, due to the excavation of a great extension of the tunnels to be
filled with concrete at a later date, allowed the confirmation, as desired, of the
geomechanical parameters used for the previous computations.
Limit equilibrium calculations, considering a 5 MPa ultimate shearing strength

considering the average vertical load applied to the concrete (with a partial safety fac-
tor of 1.5) revealed that the required safety margin against sliding along the elevation
20.00 discontinuity could be achieved by building few shear keys.
However, by means of mathematical models, it was proved that few shear keys
would take so much shear stresses, that their strength would not be enough.
727
Q.105 - R.44
Fig. 2
Illustration of the shear keys grid and detail.
(Souza Lima, 1982).
This is the first clue: within the foundation, stiffness contrast between disconti-
nuity and concrete completely modifies the stress distribution, so as the stress field
derived from LEM would no longer be applicable.
The mathematical models led to redesign the shear keys, leading to an

8.00 × 8.00 orthogonal grid formed by longitudinal and transversal concrete shear
keys, having a 3.50 × 2.50 m uniform cross section (schematically shown in
Figure 2).
In order to establish the distribution of the total shearing forces between the
concrete keys and the untreated sections of the discontinuity, finite element analyses
were performed, considering various alternatives for the grid, as regards the relative
positions of the concrete shear keys.
For the solution indicated in Figure 2, a distribution of shear stresses was

obtained, as illustrated in Figure 3, which represents the distribution in the transver-
sal concrete keys, as well as at the sections of the non-treated discontinuity. Due
to the much higher stiffness of the concrete grid, as compared to the discontinuity,
practically no shear stresses were found to be acting on it inside the grid’s cells. The
average values of the shearing stresses at the longitudinal keys are also shown in
Figure 3.
3. TELES PIRES POWER PLANT
The Teles Pires dam is owned by CHTP – Companhia Hidrelétrica Teles Pires.
The plant is located in Teles Pires river on the border of the states of Mato Grosso
and Pará (Midwest and North Brazilian regions, respectively).
Teles Pires dam is founded on a granitic rock mass known as Teles Pires
granite. This type of rock usually has high strength and high deformability modulus,
728
Q.105 - R.44
Fig. 3
Shear stress distribution along the line of the discontinuity.
Schematic illustration without scale. The integration of the dark and
light grey areas together is the tangential force T applied along
the discontinuity trace including the shear keys. (Souza Lima, 1982).
however, at the construction site, in depth, some low strength discontinuities

outcropped showing zones with weathering or hydrothermal alteration.
In Intake-Power House area the excavation works for the penstock exposed a
sound rock with joints which are persistent, rough and wavy sub-horizontally.
Throughout the rock mass, exfoliation joints, with gouge material whose thick-
ness varies from centimetric (5 cm) to metric (up to 2.0 m) with few rock to rock
contacts. One joint, in particular, proved to be continuous, weathered and very
thick, named “master joint” (Figure 6). Figure 7 shows the general cross section.
729
Q.105 - R.44
Fig. 4
Detail of the main discontinuity on the left side and preparation for blasting part of
the tunnel which later was filled with concrete (1980).
Fig. 5
Shear key tunnel, partially concreted.
As for the secondary joints, despite the significant number of rock drilling
carried out in the Intake region, there was no way to connect them in continuous
horizons that could have led to predict an unfavourable condition to the stability of
the structure.
To avoid the risks of instability resulting from this joint, the solution adopted
consisted on intercepting it with two longitudinal lines of the concrete shear keys, in
order to take into account an element of high shear resistance and low deformability.
In addition, it was included a new drainage system through the connection of the
drainage curtain between the gallery of the structure and a drainage tunnel in rock,
bypassing the entire structure.
730
Q.105 - R.44
Fig. 6
View of the excavated slope for the penstock. Note the spatial distribution of the
area between the master joint and the secondary ones indicated by yellow arrows.
Fig. 7
Typical section – Main geological and structural aspects.
The proposed solution also incorporated a HDPE impervious blanket upstream

the structure in order to reduce hydrostatic thrust and the uplift at the foundation of the
concrete structure. In the typical cross-section through the power circuit, showed on
Figure 7, it is possible to identify all the adopted solutions as well as the concept of the
existing weathered joints in the foundation. The adopted solutions efficiently solved
731
Q.105 - R.44
Fig. 8
Acting forces on intake structure.
the negative effects of lower strength and higher deformability of the master joint,
in addition to the fundamental reduction of the acting forces on concrete structures,
notably those related to hydrostatic pressures and uplift. The diagrams of acting
forces on the concrete structures were based on three-dimensional flow net studies,
besides the review of similar case studies, and later confirmed by instrumentation.
Figure 8 illustrates the acting forces on the Intake structure.
The question that remained was whether the effect of the persistent secondary
joints could become a problem regarding to the stability of the structures. Therefore,
aiming to better characterize the existing alteration materials in the relief joints, a
subdivision of them was performed into four groups classified mainly according to
the degree of alteration.
In order to evaluate the strength conditions of these materials, undisturbed

samples were collected and submitted to direct shear and triaxial compression tests.
Based on the results of these tests, it was possible to obtain the strength envelope
for each of the four groups of materials identified, as presented in Figure 9.
Since there is a considerable variation in the degree of alteration of the mate-

rials, stability studies were conducted using a finite element method considering the
existence of one joint continuity (1.0 m thick). For this case, it was adopted the lower
resistance parameter of the weathered filling material with a friction angle of 19◦ and
no cohesion. A summary of the geomechanical parameters adopted in numerical
models is presented in Table 1.
732
Q.105 - R.44
Fig. 9
Weathered materials – Strength envelopes.
Table 1
Geomechanical parameters.
These studies were performed using three-dimensional FEM modelling (soft-

ware MIDAS 3D), considering a perfect elastoplastic behaviour, in continuous joints
at elevations 167.00, 160.00, 155.00 and 150.00 masl.
Second clue: To emulate the Limit Equilibrium Method, the integration of τ × σ

curves along the analysed layer result in T and N forces that are used to calculate
the Safety Factor according to its very well-known equations.
Figure 10 illustrates the stresses distribution on the base of the Intake structure
for the critical plane at El. 167.00 masl for the Normal Loading Case (NLC).
733
Q.105 - R.44
Fig. 10
Stress distribution at critical surface.
Table 2
Stability Results – Critical case.
Table 2 shows the result of the analysis comparing to the design criteria. In
this way, even considering a secondary joint thickness greater than that found in the
field, and using the smallest parameter found for the weathered materials, it was
found that the secondary joints do not compromise the Intake Structure stability.
4. CAMARÁ DAM
The Camará dam is owned by the government of the state of Paraíba, North-
east region of Brazil on the Riachão River. It is a 50 m high rolled compacted concrete
structure designed to supply 2,4 m3 /s.
734
Q.105 - R.44
Fig. 11
Left: Camará Dam after cleaning the debris on left bank. Google Earth image,
June 15th, 2004. Right: after failure. (Abrahão, 2018).
Fig. 12
Vertical cross section through dam axis showing the geological model adopted.
(Abrahão, 2019).
The dam collapsed in June, 2004, due to erosion of heavily weathered exfolia-
tion joints occurring within the sound rock mass, a few meters below the foundation
(Figures 11 and 12).
The rebuilding process started in 2011 with a new program of investigation

and analyses and was completed by the end of 2016. The investigation program
encompassed surface mapping, coring, borehole recording, water pressure tests,
and lab tests of the joints gouge materials.
The final geomechanical model came up with a set of joints, parallel to the
surface and some faults with a high degree of weathering. Considering the aver-
age distribution, persistence, gouge and roughness of the joints, their strength was
globally taken as 35◦ of friction, with no cohesion.
735
Q.105 - R.44
Fig. 13
Camará spillway mathematical model (2D). Elevation in meter.
(Abrahão, 2016).
The geomechanical model based on this section considered the following

parameters:
• Foundation migmatite sound rock mass: deformability modulus = 20 GPa;

Poisson = 0.25.
• Disturbed surface rock mass: deformability modulus = 15 GPa; Poisson = 0.25.
• Exfoliation joints: average friction angle 35◦ , cohesion = 0.0 kN/m2 ; shear
stiffness ks = 0.07 GPa/m; normal stiffness kn = 2.87 GPa/m.
• Contact concrete × sound rock: friction angle 45◦ ; cohesion = 200 kN/m2 .
Hundreds of water loss tests registered a rock mass almost impervious with
its permeability varying from 5.10−5 cm/s to practically impervious. Exceptionally, in
a few spots where the weathered joints were sampled, some total water loss was
recorded.
By means of numerical models, the sliding stability was checked including the
downstream thrust of the rock, that is one of the remaining rock mass sliding on a
surface corresponding to the relief joint.
As a result, to a certain depth, concrete surface shear keys were needed

downstream of the existing and of the new blocks, which were a function of the
concrete block height.
In order to calculate the safety against the sliding through the discontinu-
ities in the foundation rock mass, a two-dimensional mathematical model (FEM)
was assembled to determine the stress distribution and magnitude to develop the
limit equilibrium analysis as stated in the design criteria. Figure 13 shows the
two-dimensional mathematical model and its main features.
736
Q.105 - R.44
Fig. 14
Tangential and normal stresses and φmob at potential sliding plane.
(Abrahão, 2016).
In order to better view the resulting stresses distribution along the joint within
the foundation, the tangential (τ ) and normal (σ ) stresses were plotted (Figure 14)
showing that the stiffer surface (contact between the concrete and rock in the shear
key) takes much more tangential stress, leaving the discontinuity with less loads.
Third clue: The idea of mobilized friction angle φ mob indicates the variation
of the friction angle that has to be available in order to accomplish the equilibrium.
The φ mob can be derived as: φ mob = arctan ((τ − c)/(σ − u)), where c is cohesion
and u the uplift.
The analyses for all representative blocks showed that the shear key would be
an adequate solution for the entire dam to achieve the design criteria, since 3/4 of
the dam were already built.
This text describes the spillway analysis; however, various calculation rounds
were performed to optimize the length and depth of each shear key for abutment
blocks, as well.
In order to improve the seepage control, as additional safety, the drainage

system was refurbished, and an additional grout curtain was built. The grout holes
were drilled from a slab built over the foundation rock surface at the upstream toe of
the dam face. This slab (Fig. 15) develops all along the dam and is 5.00 m wide, min-
imum, 1.00 m thick minimum and incorporated to the concrete structure by means
737
Q.105 - R.44
Fig. 15
Illustration for the as built section of the spillway.
(Abrahão, 2016).
of a watertight device (Fugenband). The slab was also considered to increase the
seepage path and increase the water head loss below the dam.
For the Spillway section the shear key width was defined with 12.00 m and on
the abutments where the concrete blocks are lower, the shear key width was defined
as 8.00 m.
To fulfil the design criteria, the existing drain holes were refurbished and the
grouting curtain built on a new position.
The dam construction started with the debris cleaning on the whole area. The
surface mapping started at this time.
After the design setting, the area of the shear key downstream of the spill-
way started to be excavated by means of cable cut to avoid major damage in the
remaining concrete.
As soon as the excavation reached the bottom, the concrete started to be

poured. In the remaining blocks the older concrete had to be partially cut to give
room to the shear key, while in the left bank, in the failure area, the shear keys were
incorporated to the new structure (Figure 16).
5. FINAL REMARKS AND CONCLUSIONS
Dealing with complex and marked geological units as low strength surfaces
below concrete structures has, since the 1970’s, being approached by various
design and construction procedures. For those cases the Limit Equilibrium Methods
does not reproduce reliable conditions to be modelled.
738
Q.105 - R.44
Fig. 16
(Left) Dam left abutment after cleaning the debris of the failed structure. The
dashed red line corresponds to the contact of the old rolled compacted concrete
and the rock foundation. Part of excavation of the left bank appears in the right side
of the picture. Some exfoliation joints can be seen in the rock walls. (Right)
Upstream wall of the excavated shear key in the left bank. A set of
exfoliation joints was intercepted by the excavation. The shear
key is partially concreted.
Going back to the clues in this text it is easy to understand that, within the
foundation, stiffness contrast between discontinuity and concrete completely mod-
ifies the stress distribution, so as the rigid body stress field derived for LEM would
no longer be applicable.
The stress distribution below the foundation has to be evaluated by means

of mathematical models supported by actual geomechanical parameters, meaning
reliable field and lab data acquisition to supply the model, besides competent geo-
logical in-depth mapping. If these models represent elastic-plastic behaviour, the
design has to evolve to some decision that eliminates or induces plastification to
converge.
In case the design criteria still require LEM definition of Safety Factor, one can,
based on τ and σ curves, calculate the curves area (integration) to obtain the forces
T (shear) and N (normal) to be applied in the respective equilibrium equations.
In all those 3 examples, the discontinuities, close to the shear keys, take almost
no shear load, as for example in Itaipu, where the internal areas of the discontinuity
within the concrete grid do not take any shear. In this case it is also recommended
to evaluate what should be the actual friction angle mobilized.
For the example, in the case of Camará the mean mobilized friction angle
φmob is less than 20◦ compared to the 35◦ available.
739
Q.105 - R.44
In all these 3 cases the LEM calculated safety factors complied with the require-
ments of the approved design criteria; however, the equilibrium forces were taken
from mathematical models.
Shear keys construction materials were engineered to deliver mechanical

properties to comply with the needs stablished in the design criteria and fulfil the
model’s requirements.
The design of the monitoring system was settled to check the accuracy of
structures behaviour, mainly with the introduction of the shear keys, and allow non-
expected intervention in a due time.
In all three cases the monitoring has shown that they all are behaving as
expected and accordingly to final design criteria.
ACKNOWLEDGEMENTS
The authors wish to thank OEC, Odebrecht Engenharia e Construção for kindly
allowing the inclusion of Teles Pires data in this paper.
REFERENCES
[1] R. Abrahão, Soft Rocks in Dam Foundation and Dam Sites, chapter 24 in “Soft
Rock Mechanics and Engineering”, Editors Milton Kanji, He Manchao and Luís
Ribeiro e Souza. Springer. ISBN 978-3-030-29476-2, 2019.
[2] R. Abrahão, F. Holanda, J. C. Degaspare. Camará Dam – Technical Aspects

of Reconstruction. Third International Dam World Conference, Foz do Iguazú,
2018.
[3] R. Abrahão, J. C. Degaspare. Rebuilding Camará water supply dam. II

International Specialized Conference of Soft Rocks, Cartagena, Colombia,
2016.
[4] V. M. Souza Lima et al. Rock Foundations with marked discontinuities, Criteria
and assumptions for stability analyses. 14th ICOLD, Rio de Janeiro, Brazil,
1982.
740
Q.105 - R.45
-------
-------
PLANNING AND DESIGN OF TEMPORARY COFFERDAMS – THE CASE OF

A COFFERDAM FAILURE IN BERGEN, NORWAY (∗)
Fjóla Guðrún SIGTRYGGSDÓTTIR

Department of Civil and Environmental Engineering,
NTNU, Trondheim
NORWAY
SUMMARY
In this paper a recent case of a cofferdam failure in Bergen, Norway, is outlined

and discussed in the framework of the Norwegian Dam Safety Regulation as well as
this in selected other countries. The case demonstrates that a cofferdam, originally
identified to be only of risk for the construction site, can quickly become a threat for a
larger area further downstream. For example, if an existing dam is partly weakened
or removed, thereby leaving the cofferdam the last barrier between the reservoir and
the downstream area.
The risk associated with the cofferdam in Bergen was misjudged by everyone
involved in the project. It was only a few days before the failure that it was recognized
that the cofferdam was a dam with potential consequences. In total 133 people were
evacuated, and a major road was closed in time before the cofferdam failure released
a flood which damaged a marina, cars, hiking trails and more.
The case study reveals the importance of careful planning and design of cof-
ferdams with due consideration of the associated risk. The discussion considers
general requirements for this in three different guidelines/regulations, as well as
∗ Planification et conception de batardeau temporaires. Le cas d’une rupture de batardeau
á Bergen, Norvége
DOI: 10.1201/9781003211808-91 741

Q.105 - R.45
return period of relevant design events from the Eurocode, probability of exceedance
and risk.
Lessons learnt from the case of the cofferdam in Bergen support the require-
ments of the guidelines and regulations reviewed in this paper. The main lessons
include, that a qualified dam engineer must plan dewatering and water diversion
during construction before the construction starts. Furthermore, a stricter require-
ment should be considered for a cofferdam that replaces the permanent dam and
for which downstream consequences are foreseen in case of a failure; compared
to a cofferdam for which the consequence of failure is limited to the construction
area. Additionally, risk assessments (such as in HSE plans) must be updated with
new information and similar for Emergency Preparedness Plans. The relevance of
the updates for safety must be realized. Evacuation, according to the emergency
preparedness plans updated on the day of the cofferdam failure in Bergen, was
successful and demonstrates the importance of such planning. Finally, it is vital to
ensure continuity between different phases of a project, e.g. with the involvement of
the designer (an approved consultant for dam safety), particularly on fundamental
safety issues in dam projects such as the control of water.
RÉSUMÉ
Ce rapport expose un cas récent de défaillance de baradeau à Bergen

(Norvège) dans le cadre du règlement norvégien sur la sécurité des barrages,
ainsi que dans certains autres pays. L’affaire démontre qu’un batardeau, identi-
fié à l’origine comme étant seulement à risque pour le site de construction, peut
rapidement devenir une menace pour une plus grande zone plus en aval. Ce peut
être le cas, par exemple, si un barrage existant est partiellement affaibli ou enlevé,
le batardeau devenant la dernière barrière entre le réservoir et la zone en aval.
Le risque associé au batardeau de Bergen a été mal évalué par toutes les per-
sonnes impliquées dans le projet. Ce n’est que quelques jours avant l’accident qu’il
a été reconnu que le batardeau était lui-même un barrage avec des conséquences
potentielles. Au total, 133 personnes ont été évacuées et une route importante a
été fermée à temps avant que la rupture du batardeau ne provoque une inondation
qui a endommagé une marina, des voitures, des sentiers de randonnée.
L’étude de cas révèle l’importance d’une planification et d’une conception

minutieuses des batardeaux en tenant compte précisément du risque associé. On
doit tenir compte des exigences générales dans trois guides/règlements différents,
ainsi que de la période de retour des événements de conception pertinents de
l’Eurocode et de la probabilité de dépassement et du risque.
Les enseignements tirés du cas de Bergen appuient les exigences des guides
et règlements examinés dans le présent rapport. Les principales leçons sont qu’un
742
Q.105 - R.45
ingénieur de barrage qualifié doit planifier l’assèchement et le détournement de

l’eau pendant la construction avant le début des travaux. En outre, une exigence
plus stricte devrait être envisagée pour un batardeau remplaçant un barrage perma-
nent et pour lequel des conséquences en aval sont prévues en cas de défaillance
par rapport à un batardeau pour lequel la conséquence de la défaillance se limite
à la zone de construction. De plus, les évaluations des risques doivent être mises
à jour avec de nouveaux renseignements et doivent être semblables à ceux des
plans d’urgence. La pertinence des mises à jour pour la sécurité doit être réalisée.
L’évacuation, selon les plans d’urgence prenant en compte la rupture du batardeau
de Bergen, a été un succès et démontre l’importance d’une telle planification. Enfin,
il est essentiel d’assurer la continuité entre les différentes phases d’un projet, par
exemple avec la participation du concepteur (consultant agréé pour la sécurité
des barrages), en particulier sur les questions fondamentales de sécurité dans les
projets de barrages.
1. INTRODUCTION
Cofferdams are required to dewater a construction site placed in rivers and

streams such as for dam construction and rehabilitation. While such cofferdams are
temporary structures, they need to be planned and designed considering health
and safety of the construction workers, safety of the downstream area and safety of
the permanent structure under construction. For any project, due consideration of
the hydrological conditions and diversion of potential floods during construction are
important.
Incidences related to cofferdams have been reported, such as leakage prob-

lems during the Folsom Dam Auxiliary Spillway Project in California in 2016, and
incidents of construction floods at the Jackson Lake Dam in 1986, Roosevelt dam in
1993 and Glendo Dam in 2010 and 2011, and the special case of the Auburn Cof-
ferdam breach in 1986 [1]. These are all cases from the USA, however, incidents at
cofferdams have without a doubt occurred worldwide on dam projects.
In this paper a recent case of a cofferdam failure in Bergen, Norway during

rehabilitation of an existing dam on the reservoir Munkebotnsvatn (see Fig. 1), is
outlined and discussed in the framework of the Norwegian Dam Safety Regulation [2]
as well as this in selected other countries. The case demonstrates that a cofferdam,
originally identified to be only of risk for the construction site, can quickly become a
threat for a larger area further downstream. For example, if an existing or permanent
dam is partly weakened or removed, thereby leaving the cofferdam the last barrier
between the reservoir and the downstream area.
The threat posed by the cofferdam in Bergen was identified some days before
the failure, thus in time for an evacuation plan to be activated. The police evacuated
133 people and closed a major road before the cofferdam failure released a flood
743
Q.105 - R.45
Fig. 1
The Munkebotnsvatn Reservoir in Bergen, Norway.
which damaged a marina, cars, hiking trails and more. Only the cleanup after the
event was estimated to 5 million NOK. Bergen Municipality instigated both an internal
and external investigation of the cofferdam failure [3], additionally, the case was
investigated by the police who proposed corporate penalties of 300 000 NOK to
be paid by both the Dam Owner and the Designer. However, in 2020 the public
prosecutor transferred the case to the Norwegian Water Resources and Energy
Directorate (NVE) for further evaluation and actions.
The external investigation [3] instigated by the Bergen Municipality concluded

that the risk associated with the cofferdam in Bergen was misjudged by everyone
involved in the project. It was only a few days before the failure that it was recognized
that the cofferdam was a dam of consequence.
2. BACKGROUND FOR THE REHABILITATION REQUIREMENT
Dam safety is emphasized in Norway with comprehensive Dam Safety Regu-

lation [2] and guidelines, and governmental supervision of dams carried out by the
Section of Dam Safety at the Norwegian Water Resources and Energy Directorate
(NVE). The entity responsible for the safety of a particular dam is the owner of that
dam (the Dam Owner).
The dam safety concept in the Norwegian Dam Safety Regulations [2] bases
first and foremost on dam structural safety, but also on supervision (monitoring) of the
dams and preparedness for unexpected events such as flooding or a dam break. The
concept of structural safety is to be reviewed regularly to ensure that an existing dam
744
Q.105 - R.45
fulfills the current requirements. Routine supervision is on the other hand to be com-
bined with regular maintenance and enable identification of necessary upgrading
and rehabilitation needs. Additionally, supervision and predefined danger limits and
alarm levels are to trigger emergency action according to predefined procedures,
defined in an emergency preparedness plan.
Dams in Norway are classified according to their consequence hazard poten-

tial into five classes (class 0, 1, 2, 3 and 4). Class 0 is for dams with no failure
consequences, while class 4 is the highest consequence class embracing dams of
very high consequence. Medium consequences are described with consequence
class 2, which is reached if 1–20 accommodation units and/or major infrastructure
are affected by a dam break. Conversely, consequence class 3 is reach if 21–150
accommodation units are affected and/or a major infrastructure. Requirements to
dam safety are stricter for the higher consequence classes.
Existing dams in Norway are to be reassessed regularly. The reassessment

also applies to design criteria such as floods arising from the hydrological conditions.
Dams in consequence class 1 need to be reassessed every 20 years, while dams
in consequence class 2 to 4 need to be reassessed every 15 years. Depending on
the outcome of the reassessment, an existing dam must be rehabilitated if needed
to fulfill the current requirements, e.g. for structural safety.
The need for added structural safety was identified during the required
reassessment of the dam Munkebotnsvatn North in Bergen, for which the cofferdam
considered herein was constructed.
3. THE CASE OF THE COFFERDAM IN BERGEN
The reservoir Munkebotnsvatn is owned by the Bergen Municipality, Norway.

The reservoir is retained by two dams (see Fig. 1), Munkebotnsvatn North and
West. The Dam Munkebotnsvatn North is (or was) a masonry dam (constructed ca
1830/1840) 42 m long and only 8.5 m high, with an integrated spillway. The catch-
ment area is relatively small or 1.12 km2 . Still, the dam was at the time defined to
be of medium consequences and thus in consequence class 2 according to the
Norwegian classification system. However, in 2020 the dam was placed in conse-
quence class 3† . Herein, consequence class 2 is referred for the dam since that was
the consequence class assigned to the dam during the events described. The area
downstream affected by potential dam break is highlighted in Fig. 2.
In 2013 a review of the ca 200-year old dam suggested rehabilitation of the

dam to fulfill requirements in the Dam Safety Regulation [2]. The process following
the review can be divided into two phases, marked by different dam solutions (new
† Information from NVE in an email from Goranka Grzanic 2020.09.30.
745
Q.105 - R.45
Fig. 2
Dam Munkebotnsvatn North. The area downstream affected by potential dam
break is highlighted. (Photo: Stavanger Aftenblad).
dam versus rehabilitation of the old dam), different consultants and different project
managers representing the dam owner. There were also different approaches for
handling of water in these two different phases. Phase 1 only extends to prepar-
ing design documents of a new dam, while Phase 2 extends into the construction
phase of the rehabilitation and ends with the cofferdam failure. A detailed account is
provided on the events leading up to the failure in a report prepared for the Bergen
Municipality (report [3]). The summary provided in this chapter 3 bases on this and
the information and documents reviewed in the report.
3.1. DEWATERING AND WATER DIVERSION (PHASE 1 AND 2)
In this section the two different phases mentioned above are discussed,
emphasizing solutions and plans for dewatering and water diversion during con-
struction. For both Phases 1 and 2, the spillway on the old dam, Munkebotnsvatn
North, would not be operational during construction/rehabilitation, hence, without
746
Q.105 - R.45
further measures, the only outlet from the reservoir was through a bottom outlet on
the other dam (Munkebotnvatn West) and an old waterpipe.
3.1.1. Phase 1 – New dam
Phase 1, comprised reassessment, feasibility study and design documents

for a new dam. In the design documents of Phase 1 the maximum outlet capacity
from the reservoir was assessed to be 1.6 m3 /s and this was compared to an inflow
of 4 m3 /s corresponding to a one in 10-year flood. In addition, the rather small
catchment area was assessed as large compared to the size of the reservoir, and
the flood damping in the reservoir assessed to be insignificant. Thus, the conclusion
in Phase 1, was that water diversion combined with a cofferdam was required during
the construction of the new dam. Furthermore, it was recommended to define the
crest elevation of a cofferdam and dimensions of a dewatering pipe beforehand.
In other words, it was not recommended to leave the decision to the contractor.
Phase 1 (new dam) did not continue since the preservation and rehabilitation of the
old dam was demanded by the relevant authorities.
3.1.2. Phase 2 – Rehabilitation of the old dam
In Phase 2, new consultants (the Designers) were assigned the task of reha-
bilitating the old dam. Phase 2 comprised design documents for the rehabilitation,
Health-Safety-Environment (HSE) plans, Emergency Preparedness plans for the
permanent dam, tender documents and the construction till the cofferdam failure.
The assessment carried out in Phase 1 on the outlet and inlet capacity of the
reservoir seem not to have been considered in the same manner in Phase 2. The
design documents of Phase 2 suggested, in a chapter on dewatering and water
diversion, that construction of a cofferdam should be avoided if possible, and that
the decision on whether to build this or not should lie with the contractor.
In the tender documents, dewatering was an all-inclusive term, and a coffer-

dam was optional. Thus, the contractor should decide on whether a cofferdam was
necessary, or if pumps would be sufficient. Furthermore, it was pointed out that a
new wall related to dam rehabilitation could be used as a temporary dam. Information
on flood sizes with different return periods was provided in the tender documents.
All plans for a potential cofferdam were to be accepted by the Dam Owner.
Even though the decision on important temporary structures was delayed until
after the construction started, the Designer in Phase 2 was not involved in the
contracting period and did not participate in a clarification meeting with the contrac-
tor. Furthermore, the Designer was hardly involved during the construction and did
neither participate in construction meetings (the first four) nor receive minutes of the
747
Q.105 - R.45
Fig. 3
Daily precipitation in mm and main events during the construction period.
(Precipitation data from www.eklima.no).
meetings. Additionally, the Designer was not, or little involved in following up on tem-
porary safety measures. This was unfortunate considering that decision relating to
dewatering and water diversion had been postponed until the construction period [3].
3.2. PLANNING, DESIGN AND CONSTRUCTION OF THE COFFERDAM (PHASE 2)
Fig. 3 shows a bar chart diagram of the daily precipitation measured at two sta-
tions (Florida and Skredderdalen) near the dam site during the construction period.
The daily precipitation is up to about 50 mm, but daily precipitation of up to 80 mm
is to be expected every other year at the site. Progress of the cofferdam construc-
tion is marked on the diagram, as well as critical water level and date of the dam
break. Additionally, completion dates of relevant project reports and assessments
(in framed text boxes) are given on the diagram.
Construction of the cofferdam was initiated in June 2018, however without

accepted plans and drawings. The Project Manager (Dam Owner) stopped the con-
struction and requested drawings. The Contractor prepared a simple sketch showing
the location of an earthfill embankment dam, and a cross-section with information
on slope inclination and the use of fiber-reinforced sheets on the upstream face.
No calculations were presented. The Project Manager forwarded the sketch to the
Designer for a quick review. The Designer replied with comments that included
a question related to the water level. The project manager rewrote the comments
before sending the contractor. Thus, all communication on the cofferdam plans went
through the Project Manager (the Dam Owner). In the last email from the Project
Manager to the Contractor, the Designer’s question on the water level is missing
and was not considered again until it was too late.
The sketch from the Contractor contained information that should have
alarmed those involved, namely that the crest elevation of the cofferdam was only
748
Q.105 - R.45
Fig. 4
The cofferdam under construction. The old dam has been removed and sheet piles
installed. (Photo: Bergens Tidende).
2 m below the planned spillway of the new dam. Thus, any weakening/removal of
the existing dam during construction, could result in that potential failure of the cof-
ferdam would have almost similar consequences as the existing dam. Furthermore,
no spillway or water diversion was provided on the cofferdam. Thus, the outflow from
the reservoir was limited to the capacity of the bottom outlet on other dam (Munke-
botnsvatn West) and the old water pipe, i.e. only at maximum 1.6 m3 /s, whereas the
10-year flood was 4 m3 /s.
At the construction site the same could have been observed. This can be
understood from Fig. 4, providing an overview of the small reservoir area, the cof-
ferdam under construction and the location of the permanent dam. In the Fig. 4, large
part of the permanent old dam has been temporarily removed and a reinforcement is
being installed as a part of the rehabilitation. Thus, the cofferdam had replaced the
permanent dam. Thereby the cofferdam became the last barrier between the reser-
voir and the population/property downstream. The crest of the cofferdam, which was
under construction when the photo was taken, is in the Fig. 4 close to the shoreline
of the reservoir, and thus the height of the old dam of consequence class 2 (later
evaluated to class 3).
The daily precipitation bar chart in Fig. 3 visualizes the fact that the majority
of July was very dry, and so was the construction site. But when it started raining
after this dry period, water quickly filled up behind the cofferdam. It is only the 19th of
August when the water level was 8 cm under the crest of the cofferdam that concerns
for the safety of the cofferdam arose, particularly since the weather forecast indicated
more precipitation the following days. Thus, to avoid or accommodate overtopping of
the dam some measures were considered, in addition to update of the Emergency
749
Q.105 - R.45
Fig. 5
Breaching of the cofferdam (Photo from: Bergen Municipality (left), TV2 (Right)).
Fig. 6
The cofferdam in Bergen after the failure. A view towards the reservoir
Munkebotnvatn. (Photo: Fjola G. Sigtryggsdóttir).
Preparedness plans. In haste, the Dam Owner decided to lower the dam crest in
the middle and incorporate a section into the dam that was supposed to withstand
overtopping and function as a spillway.
On August 22nd , 2018, the lowered dam section was overtopped and func-
tioned more like a pilot channel on a fuse plug than a spillway (Fig. 5). Thus, erosion
of the dam initiated at the lowered section and progressed into breaching of the dam
and consequent flooding of the downstream area. The breached section after the
flood can be seen in Fig. 6.
3.3. RISK IDENTIFICATION IN THE PROJECT
The above account of the events leading up to the cofferdam break indicates
lack of risk identification in the project. This section reviews the main documents
relating to risk assessment and preparedness in the project.
750
Q.105 - R.45
The Designer of Phase 2 prepared an HSE (health safety and environment)

plan for the construction work, issued in April 2018. Considerations of consequences
relating to a cofferdam were stated as limited to the construction site only. Thus,
potential consequences for the downstream area were not considered. Control of
water level and freeboard were defined as possible countermeasures. Neither of
these countermeasures was in practice considered for the constructed cofferdam,
other than control of water level by pumps.
The first emergency preparedness plan was issued in March 2018, revised in
July and again the day of the failure 22nd of August (see Fig. 3). In the first issue
from March, prepared by the Designer of Phase 2, it was stated that no special risks
were associated with a potential cofferdam, given that the height of the cofferdam
would be limited. Furthermore, no alarm limits were specified as required by the
Dam Safety Regulation [2]. It is only the revision made by the Project Manager (the
Owner) on 22nd of August, the day of the dam break, that included consideration of
flooded area from a dam break as well as alarm levels, including those triggering
evacuation. It was documented for the first time, that the elevation of the cofferdam
was only 2 m below the full supply level of the permanent dam. Furthermore, that
breaching of the cofferdam would have severe downstream consequences.
3.4. DECISIONS ON DEWATERING/WATER DIVERSION IN PHASE 2 EXPLAINED
The Designer in Phase 2 gave after the incident the following explanation on
the changes between Phase 1 and 2 on dewatering. First, that the cofferdam was
made optional in Phase 2 due to the new solution, that is a rehabilitation of an
old dam but not a construction of a new dam. The existing dam was expected to
protect the downstream area in case of water problems during construction. Second,
that the Dam Owner and the Designer agreed that a decision on whether to build
a cofferdam should be taken after the reservoir had been emptied. The contractor
should be allowed to find the most appropriate solution. Finally, the intention was that
a potential cofferdam would be so small (low) that the measure could be considered
trivial, i.e. of insignificant consequences, and thus that the cofferdam did not require
planning. [3]
4. DISCUSSION
The following discussion reviews requirements for planning and design of

cofferdams in Norway along with details regarding the requirement for actions
on temporary structures in the Eurocode. This is compared to requirements in
Sweden and regulation from the USA. The case of the cofferdam in Bergen is men-
tioned throughout the discussion as applies. Finally, some risk considerations are
discussed.
751
Q.105 - R.45
4.1. REQUIREMENTS FOR PLANNING AND DESIGN OF COFFERDAMS
In the following, the consideration of temporary structures such as cofferdams,

in dam safety regulations or guidelines in Norway, Sweden and the USA is discussed
and compared to the case of the cofferdam in Bergen.
4.1.1. Norway
The Norwegian Dam Safety Regulation [2] (referred to as the Norwegian Reg-
ulations in the following) presents dam safety requirements, including this for the
design and construction of permanent dams. Chapter 5 of the Norwegian Regula-
tion is on design documents (Technical plans and drawings), and Chapter 6 outlines
requirements for the construction works.
Here the requirements regarding the design documents and construction are
discussed and requirements for temporary structures is outlined.
• Design documents
An Approved Consultant for Dam Safety (referred to as Approved Dam Con-

sultant in the following) and a Chartered Engineer for Dam Safety (referred to in the
following as Chartered Dam Safety Engineer (CDSE) in the responsible entity (the
dam owner’s organization), have roles in the design process. The CDSE has, among
other things, the professional responsibility to follow up the safety of the dam and for
ensuring that the technical requirements of the Norwegian Regulation are fulfilled.
Thus, the CDSE assigns an Approved Dam Consultant to carry out the design of
new or reassessment/rehabilitation of existing structures. Design documents, pre-
pared by the Approved Dam Consultant, shall give information on water diversion
during construction and whether this poses a threat to safety and if so, how this is
to be handled. The design documents must be reviewed and accepted by the Dam
Safety Authorities (NVE).
The design documents for the rehabilitation of the cofferdam in Bergen

(Phase 2), did not provide the above required information, and thus did not fulfill
the Norwegian Regulation. Nevertheless, the design documents were accepted by
NVE without a comment on the issue of water diversion during construction.
• Construction
The Norwegian Dam Safety Regulation [2] defines the role of the Approved
Dam Consultant for the design process, while a role during construction is not
defined. It is the CDSE who shall ensure that everything related to planning and
construction fulfill relevant requirements in the Norwegian Regulation. For exam-
ple, the Norwegian Regulation (Chapter 6) requires the documentation of actions,
752
Q.105 - R.45
dimensions, materials and detailing of temporary dams (cofferdams) and other tem-
porary structures. However, it is not mandatory to have this documentation reviewed
and approved by the Dam Safety Authorities. In notes to the Norwegian Regulation,
it is further explained that temporary structures (cofferdams) need to fulfill necessary
safety requirements considering the above-mentioned required documentation. But
it is not specified clearly what these requirements for cofferdams consist of.
Considering the cofferdam in Bergen, the CDSE initially assigned the problem
of dewatering and water diversion to an Approved Dam Consultant (the Designer of
Phase 2). However, during the design process and preparation of the tender docu-
ments, it seems like the CDSE and the Designer agreed that the Contractor should
be responsible for solutions relating to this. Still, during construction, the Designer
became involved again when asked to review the Contractor’s plans for the coffer-
dam. The sketch of the cofferdam cannot be considered adequate documentation
that fulfills the requirements. Moreover, the necessary safety considerations were
not carried out.
• Requirements for temporary structures
In the Norwegian Dam Safety Regulation, the focus is on the permanent struc-
ture, and necessary safety requirements for a temporary dam are not clearly defined.
The question is: What are the formal requirements for temporary structures? This
question should be answered considering the required documentation of actions,
dimensions, materials and detailing.
A temporary dam of a certain material must fulfill minimum requirements to

counteract the failure mechanism of the dam type considered. For embankment
dams, it is crucial to prevent overtopping, and provide freeboard and an outlet for
the design flood (e.g. spillway). This was not provided on the cofferdam in Bergen.
Furthermore, the stability of slopes must have an acceptable level of safety against
sliding. Additionally, filters and seepage control should be provided to counteract
internal erosion, and surface protection to counteract external erosion. This was
not adequately considered for the cofferdam in Bergen, and the design was not
supported with calculations.
The requirements that apply for a permanent structure, for example regarding
actions, may be unreasonable when it comes to a temporary structure. In absent of
this information in the Norwegian Dam Safety Regulation, guidance on the minimum
return period for actions can be obtained from the Eurocode.
Table 1 is from the Eurocode and provides recommended return periods for
climatic actions. The cofferdam in Bergen was to be operated for less than a year.
Accordingly, a minimum of 10-year return period should have been taken as the
return period for floods into the reservoir. The Designer of Phase 1 (for a new dam
solution) considered this flood as described before, while in Phase 2 similar con-
sideration is not documented in the plans for the cofferdam. However, in the tender
documents for Phase 2, the Designer provided flood sizes for different return periods.
753
Q.105 - R.45
Table 1
Recommended return periods for the determination of the characteristic values of
climatic actions. (Adopted from [4])
Nominal duration of the execution phase Return period (years)
≤3 days 2a
≤3 months (but >3 days) 5b
≤1 year (but >3 months) 10
>1 year 50
a A nominal duration of three days, to be chosen for short execution phases, corresponds to the extent in
time of reliable meteorological predictions for the location of the site. This choice may be kept for a slightly
longer execution phase if appropriate organizational measures are taken. The concept of mean return
period is generally not appropriate for short term duration.
b For a nominal duration of up to three months actions may be determined taking into account appropriate
seasonal and shorter term meteorological climatic variations. For example, the flood magnitude of a river
depends on the period of the year under consideration.
4.1.2. Sweden
Recent Swedish dam safety guidelines, RIDAS, from the year 2019 [5], give
recommendations in line with the Eurocode. For example, the guidelines state that
actions should be assessed based on how long the temporary structure is to be
in use. Similarly, other requirements may be reduced considering limited lifetime.
Additionally, the guidelines [5] recommend strict requirements for a cofferdam that
replaces a permanent dam, as was the case for the cofferdam in Bergen.
The following two cases are identified, of which the cofferdam in Bergen falls
under the first case, but was planned as for the second case:
1. For a cofferdam that replaces the permanent dam and for which the same
consequences are foreseen in case of dam failure, > the requirements should
be equivalent to those defined for the permanent structure.
2. For a cofferdam that does not replace the permanent dam and is only of con-
sequence for the construction area, > the requirements should base on risks
for the working environment (HSE) and economic losses.
The Swedish guidelines also mention dewatering and diversion of water

through pipes and state that pipes should be located so that erosion of an
embankment or other parts of the dam does not occur in case the pipe breaks.
4.1.3. USA
USACE Regulation from 1994 [6] considers cofferdams in which collapse can
result in a potential risk to life or exceeds 10% of the project cost for the permanent
structure. The regulation states that for such cases, a cofferdam shall be planned,
754
Q.105 - R.45
designed, reported, approved, specified and inspected in the same manner as for
permanent project features. The USACE Regulation also requires that such a cof-
ferdam is designed and reviewed by experienced engineers. Furthermore, alternate
contractor-proposed designs must meet the approved design criteria and must be
reviewed by the authorities.
The cofferdam in Bergen was, if breached, a dam with consequences for life
and properties. Following the guidance of the USACE Regulation, the planning and
design of the cofferdam should have been carried out as for a permanent dam.
However, considering Eurocode, with appropriate selection of actions considering
lifetime, i.e. the duration of the risk.
4.2. RETURN PERIOD, LIFETIME AND RISK CONSIDERATIONS
The objectives of USACE Regulation from 1994 are to: provide criteria and
guidance for safe working conditions and life protection; optimize cost versus risk
of damage; and secure the integrity of permanent structures during construction.
The Swedish guidelines present similar safety and risk concept and differentiate
between two main cases from a consequence perspective. The Norwegian Dam
Safety Regulation is not that specific.
Here, failure mechanism for an embankment dam as in the case of the coffer-
dam in Bergen is considered, as well as, the return period of the design event and
the lifetime of the structure. With this information, in addition to cost estimates, a
risk analysis can be conducted.
4.2.1. Failure mechanism and probability of failure
Overtopping of an embankment dam is the cause of about 50% of failures

of this dam type [7]. Thus, one can roughly assume that if an embankment is
overtopped, this may ultimately result in breaching of the dam. In the case of an
embankment cofferdam, one can assume that when the design flood is exceeded,
and the full capacity of the spillway is reached, an overtopping of the cofferdam
is likely to occur. The overtopping is in turn likely to cause erosion developing into
breaching of the dam as in the case of the Bergen cofferdam. Thus, the probability
of exceedance of the design flood can, for the worst-case scenario, be taken as
almost equivalent to the probability of failure of an embankment dam or perhaps
one order of magnitude lower. This considers that the cofferdam failure probability
may, for example, be described verbally “as likely to happen as not”. In that case
the probability can roughly be considered in the range 0.33 to 0.66 [8]. However, the
failure probability must be evaluated for each individual cofferdam case.
755
Q.105 - R.45
Table 2
Probability of exceedance for different return periods and duration
4.2.2. Probability of exceedance during a predefined lifetime
Table 2 presents the return periods and probability of exceedance during a

lifetime of 3 days, 3 months, 1 up to 100 years. The permanent civil structure will
usually be designed considering a lifetime of 50 years. For dams, a lifetime of 100
years is reasonable. Reading from Table 2, for a construction project of one-year
duration and the consideration of a 1:10 year flood, then the probability is 10% that
the flood will occur during construction that one year. While there is a 2% chance for
occurrence of the 1 in 50-year flood. The permanent structure is designed for the
1000-year flood. In 100 years, the probability of the design flood is 10%. Thus, a
consideration of a 10-year flood for a construction project of one year duration, will
give the same probability of exceedance as consideration of the 1000-year flood for
a permanent structure with a defined lifetime of 100 years.
In Table 2, the orange color gives the minimum requirement of the Eurocode,
these should at least be followed for temporary structures like cofferdams built in
Norway or other countries where the Eurocodes are the national building standards.
However, in some cases one would additionally like to carry out risk analysis to
optimize cost versus risk of damage, as recommended in the US regulation.
4.2.3. Risk analysis
Risk analysis can be used to determine the appropriate return period of the
design event (the Eurocode must also be considered as a minimum requirement) in
756
Q.105 - R.45
the planning and design of cofferdams. Components of risk analysis should include;
evaluation of the probability of exceedance of the design event during the lifetime
of the cofferdam, evaluation of the failure mechanism for the cofferdam and the
probability that the exceedance of the design event will cause a cofferdam failure.
Estimation of the costs of a cofferdam failure, such as repairs, damage caused by
flooding of the downstream area and construction delays. Additionally, potential loss
of lives must be considered. The optimum return period of the design events may
additionally be estimated through evaluating the cost associated with strengthening
the structure when selecting a longer the return period, see e.g. such approach pre-
sented in [9]. However, the return period should never be lower than recommended
in a relevant standard for temporary structures.
5. FINAL REMARKS
The case of the cofferdam failure in Bergen has been reviewed in this paper and
considered through requirements in different dam safety guidelines and regulations.
The review indicates that plans for control of water during construction were not
realistic. The planning and design of the cofferdam was left to the contractor. The
cofferdam was optional in the tender documents, since the hope was that it would
not be needed, or that the contractor would find another solution. The intension
was that a potential cofferdam, should it be built, would be low and of no risk.
In view of the actual site conditions, the constructed cofferdam was not low but a
replacement of the permanent dam, and thus with consequences for the downstream
area in case of dam breach. Additionally, there was no spillway and no freeboard on
the cofferdam. Risk assessments and preparedness plans in the project were not
properly updated with respect to the actual situation. Continuity and oversight in the
project were missing. The cofferdam was not adequately planned, and it was not
properly designed.
The review demonstrates the importance of careful planning and design of cof-
ferdams with careful consideration of the associated risk. The same is emphasized
in a USACE Regulation as early as from 1994 and is also incorporated into recent
Swedish guidelines from 2019 through the identification of two main conditions to
consider. The Norwegian Dam Safety Regulation also requires that necessary safety
requirements are fulfilled.
Lessons learnt from the case of the cofferdam in Bergen support the require-
ments of the guidelines and regulations reviewed in this paper. The main lessons
include, that a qualified dam engineer must plan dewatering and water diversion
during construction before the construction starts. Furthermore, a stricter require-
ment should be considered for a cofferdam that replaces the permanent dam and
for which downstream consequences are foreseen in case of a failure; compared
to a cofferdam for which the consequence of failure is limited to the construction
area. Additionally, risk assessments (such as in HSE plans) must be updated with
757
Q.105 - R.45
new information and similar for Emergency Preparedness Plans. The relevance of
the updates for safety must be realized. Evacuation, according to the emergency
preparedness plans updated on the day of the cofferdam failure in Bergen, was
successful and demonstrates the importance of such planning. Finally, it is vital to
ensure continuity between different phases of a project, e.g. with the involvement of
the designer (an approved consultant for dam safety), particularly on fundamental
safety issues in dam projects such as the control of water.
ACKNOWLEDGEMENT
The cooperation with A. Instanes, J. Rognved and O. Bruland on reviewing the

cofferdam failure for Bergen Municipality as presented in [3] is acknowledged.
REFERENCES
[1] D.D. Mares; D.P. Keeney, V. Sandkovich Bahls. Construction Flood Case His-
tories. U.S. Department of the Interior. Bureau of Reclamation, DSO-14-01,
2014.
[2] Ministry of Petroleum and Energy. Forskrift om sikkerhet ved vassdragsan-

legg (Damsikkerhetsforskriften) [Dam Safety Regulation]. OED. Norway, 2009.
(FOR 2009-12-18-1600).
[3] A. Instanes; J. Rognved; F.G. Sigtryggsdottir; O. Bruland. Gjennomgang av

brudd på fangdam i Munkebotn. Report, 2018.
[4] Eurocode 1: Actions on structures – Part 1–6: General actions – Actions during
execution.
[5] RIDAS – Kraftföretagens riktlinjer för dammsäkerhet. Energi Företagen.

Sweden, 2019.
[6] U.S. Army Corps of Engineers (USACE); Planning and design of temporary
cofferdams and braced excavations. Regulation No. 1110-2-8152, 1994.
[7] ICOLD. Dam failures — Statistical analysis. Bulletin 99, 1995.
[8] IPCC. Managing the risk of extreme events and disasters to advance climate
change adaption. Special Report of the Intergovernmental Panel on Climate
Change. Cambridge University Press. 2012.
[9] M. Wieland. Design criteria. Int. Water Power and Dam Construction. 2004.
758
Q.105 - R.46
-------
-------
A WORLDWIDE DATABASE OF DAM FAILURE CASE STUDIES (∗)
Mayari BERNARD-GARCIA
Ph.D Candidate,
DÉPARTEMENT DES GÉNIES CIVIL, GÉOLOGIQUE ET DES MINES,
POLYTECHNIQUE MONTRÉAL.
Tew-Fik MAHDI
Professor,
DÉPARTEMENT DES GÉNIES CIVIL, GÉOLOGIQUE ET DES MINES,
POLYTECHNIQUE MONTRÉAL.
CANADA
SUMMARY
Numerical models have been developed over the years to simulate the variety
of dam failures, evaluate the flows and flood, estimate and quantify downstream
consequences. However, as a result of the current determinist available models
(especially for embankment dam failures, i.e. which only “mimic” the characteris-
tics of the breach development), standard frameworks have been formulated using
semi-empirical approaches. Indeed, most standard specifications thus remain inti-
mately dependent on the published number of failure parameters (e.g. final breach
dimensions, peak discharge, etc.). This technical paper expands the current avail-
able data collection of “detailed” dam failure cases. This data acquisition process
has led to record a total of 3,861 cases of historical dam failures around the world.
Technical description of the recorded data collection, and the lessons and learn-
ing concerning their compilation are also presented. This database and dataset
description are provided in open access, at: https://doi.org/10.5683/SP2/E7Z09B.
This Dataverse, hosted by Research Portal Dataverse and posted by Polytechnique
∗ Une base de données mondiale de cas de ruptures de barrage
DOI: 10.1201/9781003211808-92 759

Q.105 - R.46
Montreal, also facilitates future data acquisition and identification of dam incidents.
Overall, this cutting-edge and evolutive database coupled with this technical paper
allow to frame and to pave the way for data acquisition in the field of dam safety in
the actual era of “Big Data”.
RÉSUMÉ
Au fil des ans, des modèles numériques ont été développés afin de simuler la
variété des ruptures de barrages, évaluer les débits et les crues, estimer et quan-
tifier les conséquences en aval. Cependant, en raison des modèles déterministes
actuellement disponibles (en particulier pour les ruptures de barrages en remblai,
qui ne font qu’«imiter» les caractéristiques du développement de la brèche), des
cadres normatifs ont été formulés à l’aide d’approches semi-empiriques. De ce
fait, la plupart des spécifications standardisées restent donc intimement liées au
nombre de paramètres de ruptures disponibles (ex: dimensions finales de brèche,
débit de pointe, temps de formation de la brèche, etc.). Ce document technique
vise à élargir le jeu de données «détaillé» actuellement disponibles dans la littéra-
ture. Le processus d’investigation réalisé a ainsi mené à compiler un total de 3 861
cas de ruptures historiques de barrages dans le monde. Ce document technique
fournit une description technique de la collecte de données réalisée ainsi que des
leçons apprises au fil du processus. La nouvelle base de données obtenue ainsi
que la description du jeu de données sont fournies en libre accès, à l’adresse :
https://doi.org/10.5683/SP2/E7Z09B. Ce Dataverse, hébergé par Research Portal
Dataverse et publié par Polytechnique Montréal, vise à faciliter l’acquisition de futurs
«nouveaux» cas de rupture et l’identification des incidents de barrage. Globale-
ment, cette base de données évolutive, couplée à ce document technique, permet
d’encadrer et d’ouvrir la voie à l’acquisition de données dans le domaine de la
sécurité des barrages dans l’actuelle ère du «Big Data».
1. INTRODUCTION
Numerical models have been developed over the years to simulate the variety
of dam failures, to evaluate the flows and flood, to estimate and quantify downstream
consequences. However, as a result of the current determinist available models
(especially for embankment dam failures, i.e. which only “mimic” the characteristics
of the breach development), standard frameworks have been formulated using semi-
empirical approaches. Indeed, most standard specifications thus remain intimately
dependent on the published number of failure parameters (e.g. final breach dimen-
sions, peak discharge, etc.). This technical paper expands the current available data
collection of “detailed” dam failure cases. This data acquisition process has led to
record a total of 3,861 cases of historical dam failures around the world and has
been assembled from 196 references. It is this untapped potential that motivated
760
Q.105 - R.46
the writing of this article and the sharing of this database. It is also the amount and
variability of data available, which have progressively motivated the uniformization
and standardization of the set of data recorded (into one spreadsheet).
In this paper, a general description of the case studies’ compilation obtained,

the literature review, the tools used to record the database also as the challenges
encountered during the process are exposed. Thus, a general portrait of the data
record in this new compilation is presented. Finally, a discussion completes the
portrait proposed, which highlights the differences, particularities, strengths and
weakness of this new database compared to the previous database available as
presented in this paper.
2. CASE STUDIES’ COMPILATION
This new database (Bernard-Garcia and Mahdi, 2020) foresees the possi-
bility of improving safety studies and ensure better management of dams through
various fields of research. Therefore, the dataset elaborated aims to expand the
criteria for developing new and innovative relationships for predicting dam failures.
Knowing that failure studies, carried out in a predictive security mode, are mainly
based on “detailed” cases of historical dam failures (where the failure parameters
are recorded), the number and the quality of documented historical cases directly
influences the accuracy of these studies. To ensure the record of new “verified” his-
torical dam failures and (if possible) of new failure parameters, a special care has
been considered in order to record (when possible) the breach dimensions and the
failure hydrograph (peak outflow and time failure).
2.1. LITERATURE REVIEW
Overall, a total of 196 references have been assembled through this work, as
listed in the database file provided. The compilation of dam failure databases prior
to 1998 (e.g. MacDonald and Langridge-Monopolis, 1984; Costa, 1985; Froehlich,
1987, 1995a, 1995b; Singh and Scarlatos, 1988) are not explicitly listed in the
database file provided because the most recent dam failure databases have been
built on and added to these earlier databases (e.g. Wahl (1998)). The different types
of documentation (i.e. categories) that made this work possible are listed below.
• 49 databases compiling historical dam failures and/or event’s descriptions (e.g.

brief description of the failure, narrative description, …)
• 41 technical paper references (e.g. technical report, lessons learn, …)
• 44 general references both used for specific case study (e.g. news, …)
• 12 other types of references (e.g. presentations, analyses, …)
• 49 references for landslide dams adapted from Zhang et al. (2016b)
761
Q.105 - R.46
Excluding the 49 references adapted from the landslide dam failure database
from Zhang et al. (2016b), a total of 147 different references were obtained to record
and to corroborate this new database. Moreover, it is this particularity which has
allowed the authors to uniformize the dataset and standardized the record of new
dam failures.
2.2. CHALLENGES
In the course of the research, the documentation obtained has made it possible
to highlight the complexity of providing a representative and adaptive structure for
the various cases of breaks. The main challenges that were encountered and need
to be overcome are listed below.
• The presence of several dams with the same name but located in different
places in the same country.
• A nonuniform specification of the year of construction. Indeed, some
databases refer to the year of the beginning of construction of the structure,
and others refer to the year of completion of the dam.
• A classification of dam types that is difficult to categorize, in regard to the
heterogeneity of the descriptions available in the literature.
• The difficulty in tracking the historical breaks (for the same structure) due to
inconsistency, lack of available data and/or difficulty in locating the dam site
before the failure. In some cases, during the reconstruction of the structure,
an entirely different type of dam was sometimes built (hence the importance
of distinguishing events).
• A diversified compilation of the final breach dimensions in the structure
that does not always allow translation of an idealized geometry (trapezoidal,
rectangular or triangular).
• In some cases, more than one breach was formed during the break. Not only
is the total number of breaches not defined, but the position of the recorded
breach also is not specified.
• The heterogeneity of break times compiled in the various data sources (see
Wahl (1998, 2014) for more details).
• Descriptions of some dam failures such as ice dam failures, tailing dam failures
and temporary dam (cofferdam) failures are not listed/recorded.
• A diversification of values obtained by numerical simulations (such as the peak
outflow) becoming more extensive.
• A diversification of values attempting to estimate quantitative parameters (such
as the volume of the dam and/or eroded) and not specified.
Note that for the majority of the historical verified cases, the data from the
various sources are consistent with each other. For the few cases where an inconsis-
tency was identified, they involved either the type of dam, the height of the structure,
the year of construction, the failure time and sometimes the failure mode. Therefore,
those have been identified in bold in this database file.
762
Q.105 - R.46
3. DATABASE DESCRIPTION
In this database, various types of data are recorded (into one unique spread-
sheet) and reflect the diversity of data available in the literature. The five types of
data making up the 3,861 cases of failures obtained are as follows.
• 6 Cofferdam Data (COF-D)

• 1,049 Landslide Dam/Data (LS-D)
• 18 Laboratory Test Data (LT-D)
• 2,769 Man-Made Dam/Data (MM-D)
• 19 Tailing Dam/Data (TAI-D)
Regarding the variability of data available in the literature, it is the collection of

these failures into one same spreadsheet that greatly facilitated the collection and
the record of “new” dam failures.
3.1. USED TOOL
The compilation of the 3,861 historical cases of dam failures (Bernard-Garcia

and Mahdi, 2020) was carried out using a Microsoft Excel spreadsheet and is
available in the “DATABASE” tab of the XLSX file provide with this paper.
3.2. XLSX FILE STRUCTURE
To ensure a same comprehension, the global structure of the tabs available in

the XLSX file is presented below.
• “HOME” tab: Global description of the XLSX file;

• “DATASET” tab: Describing the global structure presented in “DATABASE
STRUCTURE” file available at the DOI provided;
• “DATABASE” tab: Recording this database compiling a total of 3,861 historical
dam failures around the world;
• “REFERENCES” tab: List of the 196 references assembled (and classified) in
this database presented in “APPENDIX 1” section of this paper;
• “(i)” tab (where “i” ª [1:13]): Referring to the different abbreviations employed
in this database and used in “RESULTS” section of this paper;
• “++”: Schemes for Landslide dams (LS-D) notably described in the sec-
tion/category “5. Dam properties”;
• “T.C.D.”: Pivot table of the database notably used in the “RESULTS” section;
763
Q.105 - R.46
It is suggested to open the XLSX file (provide at the DOI) for a better
understanding of the DATABASE STRUCTURE section and to fully understand the
investigation work encountered by the authors.
3.3. DATABASE STRUCTURE
Overall, the elaborate Excel spreadsheet, available in the “DATABASE” tab of

the XLSX file, is structured as follows.
• The lines represent the listed cases of ruptures.

• The 48 columns represent the dataset describing each historical dam break.
The 48 columns/variables proposed new dataset is classified into 12 cat-

egories of specifications. Each category/section contains one or more vari-
ables/columns constituting the dataset, as presented in Table 1.
Table 1
General structure of the dataset used for this database
SECTION/CATEGORY DESCRIPTION/VARIABLE/COLUMN
Dam Identification ID
Id2 From Zhang and al. (2016)
Dam Name∗
Dam Location State/Province (City/County/Region)
Country Name(1)
Dam Description Dam Trigger/Data Type(2)
Dam Type(3)
Dam Type Divided(4)
Dam Erodibility(5)
Formation Time(6) /Year of re-constuction
Failure Description Year of failure
Time before failure (Years)(7)
Failure type(8)
Failure cause divided(9)
Dam Properties Landslide volume (x106 m3 )∗∗
Dam volume (x106 m3 )
Dam height (m)
Dam crest length (m)
Crest dam width (m)
Base dam width (m)
Average width (m)
Upstream dam slope; (S_u : 1)
Downstream dam slope; (S_d : 1)
Reservoir/Lake Description Reservoir maximum Area (km2 )
Catchment area (km2 )
Lake length (m)
Reservoir/Lake maximum capacity (106 m3 )
Initial Conditions Water volume stored above breach invert (106 m3 )
Height of Water (Above de breach) (m)
Height of submersion; H_subm (m)
(Continued)
764
Q.105 - R.46
Table 1
(Continued)
SECTION/CATEGORY DESCRIPTION/VARIABLE/COLUMN
Breach Parameters Breach depth (m)
Breach top width (m)
Breach bottom width (m)
Breach Average Width (m)
Breach Average Side slope Z : 1 (h : v)
Breach shape (Idealized Geometry) (10)
ID_Ratio (11)
Ratio W_i/H_i
H_Breach VS H_Dam (12)
Peak discharge Peak outflow (m3 /s)
Failure Times Breach initiation end time (h)
Breach formation end time (h)
Emptying Reservoir end time (h)
Formation time tf (h)
Damages Loss of life
Notes References
Cotes(13)
Remarks
(i) : ID tab in the XLSX file, see DATASET DESCRIPTION file for more
details
The general structure of this new database (e.g., dataset, type of variables,
abbreviations, and nomenclature) has been worked meticulously to facilitate its use,
but, moreover, to facilitate the addition and/or tracking of the compiled data. There-
fore, its structure makes it possible to draw a portrait of the hydrographs of breaks
for different modes of breaks and types of dams. The “user guide” format of the
DATASET DESCRIPTION file is also provided and has deliberately been adopted,
notably in order to overcome the difficulties encountered by the author and clarify
it content. A significant concern has been given to the presentation of the general
structure of the dataset and the XLSX file, notably to facilitate and encourage its use.
4. RESULTS
In order to present a comprehensive and representative portrait of this

database in this paper (at the height of its contribution, but without distorting the scale
and/or impact of this paper), the results presentation has been adapted as follows.
– Description of the 3,861 worldwide historical dam failures compiled in this

database.
– Description of the 2,769 man-made dam (MM-D) failures recorded in this
database.
– Description of 355 historical embankment dam (EFL + RFL) failures compiling
at least one breach dimension (“Cote” = 3) and adapted from this database.
Notice that even if the database and dataset are considered as the main result
and contribution of this paper, it has not been possible to include it in the body of this
765
Q.105 - R.46
document (notably in regard to its dimensions : 48 columns and more than 3,861
lines) but it is freely available to any readers through this DOI 10.5683/SP2/E7Z09B
hosted by Research Portal Dataverse.
4.1. GENERAL DESCRIPTION OF THE 3,861 WORLDWIDE HISTORICAL DAM FAILURES

COMPILED IN THIS DATABASE
First, in regard to the geographical distributions of the 3,861 historical cases

recorded in this database, obtained for each “Dam Trigger” and illustrated in
Figure 1. It is possible to highlight the dominance of the 2,769 man-made dam
(MM-D) failures (i.e. 72% of this database) and 1,049 natural dams (LS-D) failures
(i.e. 27% of this database) recorded. It can be deduced that the 18 “Laboratory Test
Data (LT-D)”, the 19 “Tailing dams (TAI-D)” and the 6 “Cofferdams (COF-D)” only
represent 1% of the data record in this compilation.
Fig. 1
Geographical distribution of dam breaks recorded in this new database
Distribution géographique des cas de ruptures de barrages recensés dans
cette nouvelle base de données
766
Q.105 - R.46
Also, notice that:
• Data from 69 countries were compiled for this database.

• Data from 50 countries were compiled for man-made dam (MM-D) failures in
Figure 1-b;
• Also, a total of 2,542 man-made dams (MM-D) case study, located in the United
States, are also recorded (compared to the 65 cases located in China or the
14 cases located in Canada for example).
• Data from 42 countries were compiled for natural dam (LS-D) failures, as
illustrated in Figure 1-c;
• Data from 3 countries were compiled for data obtained by laboratory tests
(LT-D), as illustrated in Figure 1-d;
• Data from 11 countries were compiled for tailing dam (TAI-D) failures, as
illustrated in Figure 1-e;
• Data from 1 country were compiled for the cofferdam (COF-D) failures, as
illustrated in Figure 1-f;
Secondly, in regard to the distributions of “Failure Type” obtained for each

“Dam Trigger” and illustrated in Figure 2, it is possible to highlight the most recorded
“failure type” in this database is the overtopping failure (OT), except for “Landslide
Dams (LS-D)” (Figure 2-c).
Fig. 2
Distribution of the failure type, according to the “Dam Trigger”,
recorded in this database;
Distribution des types de rupture enregistrés en fonction du type
d’ouvrage compilés dans cette base de données.
767
Q.105 - R.46
In addition,
• As shown in Figure 2-b, 1,300 cases of man-made dam (MM-D) recorded have
failed by overtopping (OT) (i.e. 47% of the 2,769 MM-D cases).
• As shown in Figure 2-d, all the 18 “Laboratory Test Data (LT-D)” cases have
been classified as overtopping failure (OT).
4.2. DESCRIPTION OF THE 2,769 MAN-MADE DAM (MM-D) FAILURES

RECORDED IN THIS DATABASE
Regarding the distributions of the “Failure Types”, obtained for each “Dam
Types” and illustrated in Figure 3, it is possible to highlight the dominance of the
2,058 cases of embankment dam (EFL+RFL) failures (i.e. 74% of the man-made
dam (MM-D) failures’ compilation) recorded in this database.
• A total of 990 cases of overtopping failures (OT) of embankment dams

(EFL + RFL) are recorded (i.e. 945 “Earthfill dam (EFL)” dam failures shown
Fig. 3
Distribution of the number of breaks compiled for each failure type, according to
the type of dam recorded for man-made dams (MM-D);
Distribution des modes de rupture compilés en fonction des types de barrages
construits par l’Homme enregistrés (MM-D).
768
Q.105 - R.46
in Figure 3-a and 45 failures of “Rockfill dam (RFL)” dams shown in

Figure 3-b).
• A total of 280 rigid dam failures (CON+MAS) (i.e. 10% of man-made dam (MM-
D) cases) are also recorded as shown in Figure 3-c and Figure 3-d. Unlike the
other dam types recorded in this database, the most reported failure type of
concrete dams (CON) recorded is the internal erosion (IEE).
• Also, the other dam types (STL + TIM + OS), all grouped together in
Figure 3-e, only represent 2% of this compilation. Whereas for 13% of man-
made dams (MM-D), the dam type remains unknown (UNK), as shown in
Figure 3-f.
Knowing that each case study compiled in this database have different degrees
of details, Figure 4 then illustrates the distributions of the historical failures recording
the dimensions of the breach (“Cote” = 3), the peak outflow (“Cote” = 4) and/or the
failure times (“Cote” = 5), adapted from this database also as the 111 cases adapted
by Froehlich (2016a,b).
As illustrated in Figure 4:
• A total of 451 cases of man-made dam (MM-D) failures, where at least one
of the three failures parameters is available (“Cote” = 3 and/or 4 and/or 5) are
recorded in this database.
Fig. 4
Distributions of the number of detailed historical cases,
based on the column/variable “Cote”, present in this database and
compared to the distribution adapted from Froehlich (2016a,b);
Distribution des nombre de cas de rupture détaillé, basé sur la colonne
“Cote” disponible dans la base de données et comparés aux cas de
ruptures enregistrés par Froehlich (2016a,b).
769
Q.105 - R.46
• As highlighted in bold, a total of 355 historical cases of embankment dam

failures (EFL + RFL), where at least one breach dimension (“Cote” = 3), are
• Also, by comparing the embankment dam (EFL+RFL) cases failed by over-
topping (OT) and those failed by internal erosion failures (IEE) compiled in
this database to the 111 cases (OT + IEE) adapted from Froehlich (2016a,b),
it is possible to notice the among new “detailed” historical dam failure cases
4.3. DESCRIPTION OF 355 HISTORICAL EMBANKMENT DAM (EFL+RFL)

FAILURES COMPILING AT LEAST ONE BREACH DIMENSION (“COTE” = 3)
RECORDED IN THIS DATABASE
To enable a better comparison of the data described notably in this subsection,

the failure types distributions of the 355 “detailed” cases adapted from this database
(highlight in Figure 4), the 108 cases of historical ruptures adapted from Wahl (1998)
also as the 111 cases adapted from Froehlich (2016b) are resumed at Table 2.
Table 2
Failure types distributions of the 355 “detailed” embankment dam failures recorded
in this database, the 108 cases adapted from Wahl (1998) and the 111 cases
adapted from Froehlich (2016b).
As presented in Table 2, the most recorded failure types are the overtopping
failure (OT) and internal erosion failure (IEE) in those three databases. Indeed, these
two failure modes (OT+IEE) represent 87% of this database, 88% of Wahl (1998)
770
Q.105 - R.46
compilation and 99% of Froehlich (2016b) compilation. However, unlike the compi-
lations adapted from Wahl (1998) and the one adapted from Froehlich (2016b), this
database compiles more embankment dam (EFL+RFL) overtopping failures (OT)
than internal erosion (IEE) failures. Among the 41 embankment dam (EFL + RFL)
cases failed by overtopping (OT), adapted from Wahl (1998), 31 of the rupture cases
record the breach depth (HB ) and 19 cases also record the dam height (HD ). Also,
notice that Froehlich (2016b) compilation do not explicitly record the dam height.
Thus, the dam height distributions of the 355 embankment dam cases
(EFL + RFL) adapted from this database, the 108 adapted cases from Wahl (1998)
and the 593 cases of failure presented by Zhang et al. (2007) are illustrated in
Figure 5.
As shown in Figure 5:
• It is possible to highlight the dominance of small dams notably in this

compilation. Indeed,
◦ 222 cases of failure adapted from this database (i.e. 63% of this sample)
have a dam height (HD ) lower than 15 m,
◦ compared to the 43 cases from Wahl (1998) (i.e. 40% of the sample) and
the 301 cases reported by Zhang et al. (2007) (i.e. 51% of the sample).
Fig. 5
Distribution of the 355 embankments dam’s height recorded in this database and
the 108 embankments case study adapted from Wahl (1998) and compared to the
distribution from Zhang et al. (2007);
Distribution des hauteurs de barrages enregistrées pour les 355 barrages en
remblai enregistrés dans cette base de données, des 108 cas de ruptures de
barrages en remblai adaptés de la compilation de Wahl (1998) et des 593 cas
adaptés de Zhang et al. (2007).
771
Q.105 - R.46
Fig. 6
“Time before failure” distribution of the 355 embankments dam recorded in this
database;
Distribution des “Délai avant rupture” des 355 cas de ruptures de barrages en
remblai enregistré dans cette base de données.
• A total of 55 embankment dams (EFL + RFL) failed by overtopping (OT),

recorded in this database, have a dam height (HD ) higher than 20 m, compared
to the 18 cases recorded by Wahl (1998).
Furthermore, by filtering the column/variable “Time before failure” it is also

possible to illustrate, in Figure 6, that from the 355 overtopping embankment dam
failures (OT, EFL + RFL):
• 17 cases (i.e. 5% of the sample) do not compile the year of the rupture (UD,
UNK).
• 140 cases of failures do not compile the year of construction (UD) (i.e. 40% of
the sample).
• Almost 7% of this sample cases, the failure occurred during the dam
construction (UC) or during the reservoir filling phase (DRF).
In the same vein, the distributions of the “year of failure” recorded for the 355
historical embankment dams (EFL+RFL) failures adapted from this database, the
111 cases adapted from Froehlich (2016b) and the 108 historical failures adapted
from Wahl (1998) are illustrated in Figure 7.
Therefore, by comparing these distributions, it is thus possible to highlight

that most of the new cases studies recorded in this database have failed in 1970 to
2019.
772
Q.105 - R.46
Fig. 7
Distribution of the year of failure of the 355 “detailed” embankment dams
(where at less the breach dimensions are available) and compared to the 111
cases compiled by Froehlich (2016b) also as the 108 compiled by Wahl (1998);
Distribution des “années de rupture” des 355 cas de rupture de
barrage en remblai détaillé (enregistrant au moins une des dimensions de la
brèche), comparé aux 111 cas compilés par Froehlich (2016b) ainsi que des
108 cas compilés par Wahl(1998).
5. DISCUSSION
First, let’s highlight that it is the general dataset provided through this work that
has allowed to facilitate the generation of the previous results. Indeed, the results
illustrated in Figure 1 to Figure 7 aim to elaborate a general description of this com-
pilation. By comparing those distributions to the one recorded by Froelich (2016a,b)
and Wahl (1998), it also presents how this compilation provide (or not) different
picture on the previous distribution of dam types which have failed historically. Over-
all, even if these results only provide a “preliminary” portrait of the content of the
database (without presenting the database explicitly), this new database still repre-
sents the largest compilation of dam failures recorded to date and a considerable
attempt of data uniformization. Therefore, this compilation is also the first attempt to
combine and record various data adapted for the field of civil engineering concerned.
5.1. GENERAL DESCRIPTION OF THIS DATABASE
Following the validation and compilation work carried out, the need to develop
a database structure that quickly permits checking the credibility of the break param-
eters (or simply to put the dam failure into context) seems essential and have led
773
Q.105 - R.46
the diversification of the parameters, notably those used to translate dam properties
(−5. Dam Properties). Indeed, the proposed dataset makes it possible to quickly
identify duplicates and/or inconsistencies (even between the different “Dam Triggers”
recorded).
In addition, the major deficiencies identified in the CHALLENGES section have

been solved with the proposed new dataset composes by the 48 columns/variables.
Given the large amount of data available and still not processed in this database
(hence the magnitude of the task), an unfinished work was unavoidable. Due to the
main focus accorded to the man-made dams’ (MM-D) compilation, notice that more
data are still available in the literature, notably for “Cofferdams (COF-D)”, “Laboratory
test data (LT_D)” and “Tailing dams (TAI-D)” (e.g. For “Laboratory Test Data (LT-D)”:
Frank (2017), Coleman et al. (2002), Chinnarasri et al. (2002), Schmocker and
Hager (2009) and Walder et al. (2015)).
5.2. GENERAL DESCRIPTION OF THIS COMPILATION
This new database compiles historical dam failures regardless of the dam
height or any other parameters. The results obtained also corroborate some of the
“worldwide” dominance anticipated (e.g. the dominance of the embankment dam
failed by overtopping recorded in this database).
Thus, as illustrated in Figure 1-b, the over-representation of dam failures in

the United States (i.e. for the 2,543 cases, located in the United States) emphasizes
the accessibility of their data, which have facilitated their compilations (i.e. compared
to other countries). In fact, as presented by Zhong et al. (2017), according to the
statistical data of the “Dam Safety Management Center of the Ministry of Water
Resources, PR China” (MWR, 2015), 3,530 dam failures were recorded in China
between 1954 and 2014 (compared to the 65 recorded in this database), and more
than 50% of those resulted from an overtopping failure.
With regard to the level of detail in each case of failure (illustrated in Figure 4),
based on the 2,769 cases of man-made dam failures (MM-D) recorded in this
database, “only” 3% to 16% of the historical failures are sufficiently “detailed” and
therefore exploitable. However, those “detailed” cases studies are more numerous
than the ones recorded in previous database available in the literature (i.e. Froehlich,
2016b). Also, even if these cases (with very little data) hence have little impact on
more refined studies (e.g. peak outflow uncertainty analysis), for those specific
cases (with few details) a rigor was applied during investigation work to at least
identify the cause of failure, years of rupture, geographical location and type of
dam. Overall, these cases come mainly from Stanford University online database
(NPDP, 2016b) and that it is also the structure of this reference that has limited its
use. Indeed, the online database of Stanford University (NPDP, 2016b) requires
entering the name of the dam and clicking on different links to extract different data
774
Q.105 - R.46
(and thus a considerable time). The new version of this database is now provided
into a spreadsheet.
5.3. IMPROVEMENTS
Regarding the impressive compilation of worldwide historical dams’ failures

also as the variety of data recorded in this database, remember that this compilation
only reflects the “tip of the iceberg”. Despite the tremendous work encountered by
the author, as previously mentioned by Wahl (2014), this meticulous and fastidious
work is not recommended to anyone. As a guide, the improvements recommended
for an updated future version of this database are listed.
• Explicitly specified in the “Remarks” column the method employed for the esti-
mation of the dam volume (which is a derived quantity depending on the soil
type and the compaction level, on the method of calculation and on the dam
body dimensions).
• Refine/Validate data specifying “Dam Type Divided” for embankment dams
compiled in the database.
• Refine the historical cases compiled for which only the country is specified by
using the variable “State/Province (City/County/Region)”.
• For all cases where the failure type is unknown, search to identify the main
cause or to classify as “Not Reported (NR)”, which means that no data were
recorded.
• Add a column/variable “Failure Cause” and identify the main cause of the type
of “Failure Mode” specified (from the causes identified in the variable, “Failure
Cause Divided”).
• Verify the 594 cases of rupture from Zhang et al. (2016a), which have not yet
been corroborated by the literature.
• Adapt the few compiled historical cases for which several breaches were
formed within the same structure and during the same break.
6. CONCLUSIONS AND OVERVIEW
In conclusion, this article has provided a literature review of previous historical

dam failures around the world which compose the largest database of dam failures
recorded to date (Bernard-Garcia and Mahdi, 2020). In fact, it is this untapped
potential that motivated the writing of this article and the sharing of this database. The
exhaustive portrait of the variability and quantity of the data listed in this database
demonstrates its potential in various fields of dam safety study and civil engineering.
Indeed, the structure of the dataset and the XLSX file have been worked in order
to facilitate its use, and therefore, allow anyone to draw similar and/or diversified
775
Q.105 - R.46
portraits for other types of dams identified (e.g. concrete dam) and/or failure modes
(e.g. internal erosion) and/or failure parameters (e.g. formation time). Regarding the
variability and the diversity of data available, this vast compilation is a door opening to
a variety of statistical, descriptive and/or predictive studies, particularly in the field of
hydrology, hydraulic and dam safety. Therefore, the intent of sharing this database is
to provide useful data to researchers in various fields of civil engineering. Hence, by
providing the database through a XLSX file the author aims to facilitate the use of this
compilation, the extraction of samples but also to encourage others to perpetuate
the work that has been done by the authors.
Overall, the results were first elaborated to provide a portrait of the variability
and relevance of the data recorded in this database and composing the dataset.
Progressively, the samples illustrated in the distributions proposed allowed to put
back in context this compilation by adding samples adapted to the database recorded
by Wahl (1998), Froehlich (2016a,b) and Zhang et al. (2007). It has therefore made it
possible to identify new “detailed” cases of ruptures notably for embankment dams.
Leading to a sample composed of “detailed/exploitable” case studies, describing the
breach dimensions also as the dam height has aimed to provide technical findings
based on current standards’ specifications. Again, this database thus represents a
door opening for various studies in civil engineering.
Regard to the relevant data identified in the literature over the process, it finally
offers the possibility of converging the methods to a single database of dam breaks
(for all dam kinds), which as the most recent ones will allow a general consensus
on the reliability of the compiled data. The accessibility of historical dam failure data
will eventually make it possible to avoid having to observe breaks (or more breaks)
in its territory (or elsewhere) and to conduct more appropriate safety studies that
are accessible to all (empirical methods, probabilistic modeling and others). Please
note again that this new database (XLSX file) and the dataset description is freely
available through the DOI furnish in the “SUMMARY” of this paper.
DATA AVAILABILITY STATEMENT
The database, a XLSX file, and dataset description (PDF file) generated dur-
ing the study is available, is freely available to the public and can be downloaded
through Research Portal Dataverse, at this following DOI: 10.5683/SP2/E7Z09B.
This database has been released free of charge for open access data retrieval and
unrestricted use under a CC0 License, through a link hosted by Research Portal
Dataverse and posted by Polytechnique Montreal.
ACKNOWLEDGMENTS
This research was supported in part by a National Science and Engi-

neering Research Council (NSERC) Discovery Grant for the corresponding
author, application No: RGPIN-2016-06413, and a Hydro-Quebec and NSERC
776
Q.105 - R.46
Collaborative Research and Development Grant, application No: CRSNG RDC

491880 - 15. Special thanks is extended to Professor Pierre Léger, from Polytech-
nique Montreal, for providing some documents of concrete dams’ historical failures.
REFERENCES
[1] Bernard-Garcia, M. and Mahdi, T.-F. (2020). “A Worldwide Historical Dam

Failure’s Database”, https://doi.org/10.5683/SP2/E7Z09B, Scholars Portal
Dataverse, V1.
[2] Chinnarasri, C, Tingsanchali, T, Weesakul, S and Wongwises, S (2003). “Flow

Patterns and Damage of Dike Overtopping”. International Journal of Sediment
Research, Vol. 18, No. 4, p. 301–309. (Request PDF).
[3] Coleman, S E, Andrews, D P, and Webby, M G (2002). “Overtopping Breaching

of Non-cohesive Embankment Dams”. Journal of Hydraulic Engineering, Vol.
128, No. 9, p. 829–838.
[4] Costa (1985). “Floods from Dam Failures”. Open-File Report 85–560, US
Geological Survey, Denver, Colorado.
[5] Franck (2017), P-J R. “Hydraulics of Spatial Dike Breaches”. Dr. Sc. Thesis,
Diss. ETH No. 23938, ETH Zurich.
[6] Froehlich, D C (1987). “Embankment-Dam Breach Parameters”. Hydraulic

Engineering, Proc. 1987 ASCE National Conference on Hydraulic Engineer-
ing, Williamsburg, Virginia, 3-7 August, p 570–575.
[7] Froehlich, D. C. (1995). “Peak outflow from breached embankment dam”.

Journal of Water Resources Planning and Management, 121(1), 90–97.
[8] Froehlich (1995a). “Peak Outflow from Breached Embankment Dam”. Journal
of Water Resources Planning and Management, Vol. 121, No. 1, p. 90–97.
[9] Froehlich (1995b). “Embankment Dam Breach Parameters Revisited”. Water

Resources Engineering, Proc. 1995 ASCE Conference on Water Resources
Engineering, San Antonio, Texas, August 14–18th, 1995, p. 887–891.
[10] Froehlich D. C. (2016a). “Predicting Peak Discharge from Gradually

Breached Embankment Dam”, American Society of Civil Engineers, DOI:
10.1061/(ASCE)HE.1943-5584.0001424. (Table 1-p. 3–6)
[11] Froehlich D. C. (2016b). “Empirical model of embankment dam breaching”,

Conference Paper, June 2016, DOI: 10.1201/9781315644479-285. (Table 1;
p. 7–9)
[12] MacDonald, T C and Langridge-Monopolis, J (1984). “Breaching Character-

istics of Dam Failures”. Journal of Hydraulic Engineering, Vol. 110, No. 5,
p. 567–586.
777
Q.105 - R.46
[13] Ministry of Water Resources'CP. R. China, (MWR, 2015). “Dam Breach Reg-
ister Book of the National Reservoirs”. Dam Safety Management Center of the
Ministry of Water Resources, Nanjing, China. (IN CHINESE).
[14] Schmocker, L and Hager, W H (2009). “Modelling Dike Breaching due to

Overtopping”. Journal of Hydraulic Research, Vol. 52, No. 4, p. 585–597.
[15] Singh, V P and Scarlatos, P D (1988). “Analysis of Gradual Earth-Dam

Failures”. Journal of Hydraulic Engineering, Vol. 114, No. 1, p. 21–42.
[16] Stanford University (NPDP, 2016b). “NPDP Dam Incident Database”, Online
DATABASE, Last Updated 2016-04-07, National Performance of Dams Pro-
gram. <npdp.stanford.edu‘dam_incidents>
[17] Wahl, T. L. (USBR, 1998). “Prediction of embankment dam breach parame-

ters: a literature review and needs assessment”, Dam Safety Research Report
DSO-98-004, US Bureau of Reclamation.
[18] Wahl, T. L. (2014). “Evaluation of Erodibility-Based Embankment Dam Breach

Equations”, Hydraulic Laboratory Report HL-2014-02, Denver, Colorado: US
Bureau of Reclamation.
[19] Walder, J S, Iverson, R M, Godt, J W, Logan, M and Solovitz, S A (2015).

“Controls on the breach geometry and flood hydrograph during overtopping
of non-cohesive earthen dams”. Water Resources Research, Vol. 33, No. 10,
p. 2337–2348.
[20] Xu Y, Zhang LM (2009). “Breaching parameters for earth and rock-fill

dams”. Journal of Geotechnical and Geoenvironmental Engineering 135(12):
1957–1969. DOI: 10.1061/(ASCE) GT.1943-5606.0000162.
[21] Xu. Y. (2010). “Analysis of dam failures and diagnosis of distresses for
dam rehabilitation”, Doctoral thesis, Hong Kong University of Science and
Technology, January 2010, p. 317.
[22] Zhang, L M, Xu, Y and Jia, J S., (2007). “Analysis of Earth Dam Failure –
a Database Approach”. Proc. ISGSR2007 First International Symposium on
Geotechnical Safety and Risk, 18–19 October 2007, Shanghai, p. 293–302.
[23] Zhang, L., Peng, M., Chang, D., Xu, Y. (2016a). “Dam Failure Mechanisms
and Risk Assessment”, John Wiley & Sons, 2016. APPENDIX A.
[24] Zhang, L., Peng, M., Chang, D., Xu, Y. (2016b). “Dam Failure Mechanisms
and Risk Assessment”, John Wiley & Sons, 2016. APPENDIX B.
[25] Zhong, Q.-M., Chen, S.-S, Dend, Z. (2017). “Numerical model for homoge-
neous cohesive dam breaching due to overtopping failure”. Science Press
and Institute of Mountain Hazards and Environment, CAS and Springer-Verlag
Berlin Heidelberg 2017. DOI: 10.1007/s11629-016-3907-5.
778
Q.105 - R.47
-------
-------
FLOOD DESIGN EVALUATION AT BATUTEGI DAM, INDONESIA (∗)
Wanny K. ADIDARMA
Hydrologist,
Dam Safety Unit, MINISTRY OF PUBLIC WORKS AND HOUSING
Anissa MAYANGSARI
Section Head of Program, Data, Evaluation and Information, Dam Safety Unit,
Oky SUBRATA
Hydrologist,
Hydrology and Water Environment Unit,
INDONESIA
SUMMARY
Determination of a design flood should be as close as the condition in the

dam location, whether it is at the design phase and the operation of the dam. One
of the important factors in calculating the design flood plan is the calibration to
determine the input model parameters. Evaluation of the flood design at Batutegi
Dam used the rainfall-runoff relationship model by dividing it into 4 sub-watersheds
and consisting of 3 sub-models, namely the Losses Model, the Hydrograph Unit
Model and the Baseflow Model. The loss model parameter is based on the Soil
Conservation Services method in the form of Curve Number (CN), the hydrograph
unit parameter in the form of time lag is calculated based on the topography, and
the base flow parameter uses the Reversion Constant method. The calibration of
the flood model parameters uses the 2018–2019 reservoir flood water level data
∗ Évaluation de la crue de projet du barrage de Batutegi, Indonésie
DOI: 10.1201/9781003211808-93 779

Q.105 - R.47
which is read three times a day. Based on the parameters that have been calibrated
and the reservoir volume at an elevation of normal water reservoir El. + 274 m which
is 770 MCM and PMP 576 mm, the peak inflow of PMF is 3,361 m3 /s and spillway
capacity is 789 m3 /s. For the 3-day design rainfall, PMP 650 mm, the peak of the PMF
inflow is 3,687 m3 /s and the spillway capacity is 949 m3 /s the maximum reservoir
water level El. + 278 m. It can be concluded that the spillway capacity and height of
the dam in the Batutegi Dam are still sufficient and safe.
RÉSUMÉ
La détermination d’une crue de projet doit prendre en compte l’emplacement

du barrage et ses modes de fonctionnement, que ce soit lors de la phase de con-
ception et de l’exploitation du barrage. L’un des facteurs importants dans le calcul
des crues incidentes est l’étalonnage pour déterminer les paramètres d’entrée du
modèle. L’évaluation des crues au barrage de Batutegi a utilisé un modèle de rela-
tion pluie-ruissellement après avoir divisé le bassin versant en 4 sous-bassins ce
modèle comportant lui-même sous-modèles, à savoir le modèle des pertes, le mod-
èle de l’hydrogramme unitaire et le modèle de débit de base. Le paramètre de base
du modèle des pertes est basé sur la méthode des services de conservation du
sol sous la forme de numéro de courbe (CN). Celui de l’hydrogramme unitaire sous
forme de décalage temporel est calculé en fonction de la topographie. Le paramètre
d’écoulement de base utilise la méthode de la constante de réversion. L’étalonnage
des paramètres du modèle d’inondation utilise les données de niveau d’eau et de
crue du réservoir sur la période 2018-2019 et qui sont lues trois fois par jour. En
partant d’un réservoir à une cote du plan d’eau à 274 m (soit 770 hm3 ) et pour une
PMP de 576 mm, le pic d’entrée de la PMF est de 3361 m3 /s et le débit du déver-
soir est de 789 m3 /s. Pour la pluviométrie de conception de 3 jours et une PMP de
650 mm, le débit de pointe est de 3687 m3 /s et le débit du déversoir est de 949 m3 /s
avec un niveau maximal de la retenue à 278 m. On peut conclure que la capacité du
déversoir et la hauteur du barrage du barrage de Batutegi sont suffisant et assure
la sûreté.
1. INTRODUCTION
Batutegi Dam is located at Tanggamus City, Lampung Province. It is about

250 km from Jakarta. There are 3 rivers that flow to the reservoir, Way Sengharus,
Way Sekampung, and Way Rilau. The purposes of the dam are to provide water
irrigation 66,500 ha, raw water 2,75 m3 /s and hydropower 2 × 14 MW.
Batutegi Dam was constructed in 1995–2002. The type of the dam is rockfill
embankment with a central impervious core zone. The height of the dam from the
780
Q.105 - R.47
Fig. 1
The Layout of Batutegi Dam
Le plan du barrage de Batutegi
bottom of the foundation is 120 m (El. +283 m) with 701 m length and 12 m width at
the crest. The total capacity of the reservoir is 885 MCM at full water level (El. 278 m)
(Fig. 1 and Fig.2).
Fig. 2
Section of the Batutegi Dam (up) and Section of Spillway with Tunnel (bottom)
Section du barrage de Batutegi (en haut)
et section du déversoir avec tunnel (en bas)
The spillway was designed with ogee overflow at the inlet and connected with
the tunnel to the energy dissipator at the downstream. The diameter of the tunnel is
11.5 m. The crest of spillway at El. + 274 m (Fig.2).
781
Q.105 - R.47
Based on the Ministry of Public Works and Housing No. 27/2015 about Dam
[1], the reservoir operation rule should be evaluated every 5 years. It is necessary
to evaluate hydrology analysis to see if the design flood is still on the track as stated
in design. This paper will discuss the evaluation of flood design based on the new
hydrology data.
2. METHODOLOGY
The choice of a design flood analysis method is very dependent on the avail-
ability of hydrological data. There are many methods that can be used, but whatever
method is used, the main key in the design flood analysis method is calibration to
get the model parameter values. Based on SNI 2415-2016 about the procedure for
calculating design flood plans [2] are as follows:
a. A simple model (with less of model parameters) and a complex model (with
many parameters) can be used to estimate flood discharge from a water-
shed. Both types of models can be used if the parameter values can be
measured or calibrated (p.44). The SNI also describes the procedures for
utilizing mathematical models for determining design flood, as follows:
1) Select a mathematical model (Rainfall-Runoff model) to simulate the
flood hydrograph that occurs.
2) Select a hydrograph of the flow and rainfall that caused the flood.
3) Calibrate the model to get the parameter.
4) Verify the parameter by taking the hydrograph of the flood and rainfall. If
the results of the verification are good, proceed to the next stage (5) if it
is not good, go back to stage (3).
Fig. 3
Water level and inflow relationship
Relation entre le niveau d’eau et le débit
782
Q.105 - R.47
5) Select the maximum rainfall for each year using the verified model
parameters. The magnitude of the flood hydrograph due to the great-
est rainfall can be determined. Perform this process for the entire year
of available rain observations.
6) Determine the peaks from the flood hydrographs for each year.
7) Calculate the design flood for various return periods from peak flood
discharge data.
b. If the calibration cannot be done to get the model parameter, the design
flood discharge needs to be calculated using several approaches/models to
compare the results of several models (p.45).
At Batutegi Dam catchment area, the hydrological data is complete data.

Calibration has been carried out using reservoir water level data by applying the
HEC HMS model to obtain model parameter values, so there is no need for com-
parisons with other methods. There are two stages for design flood analysis for
Batutegi Dam [3]. First stage is calibration which results in a model parameter.
The second stage is the prediction stage which results in the design of flood
hydrograph [6]. Fig. 3 illustrates the performance of the calibration and prediction
models.
In the identification or calibration stage, the rainfall data and discharge hydro-
graph or Reservoir Water Level (MAW) are needed when the flood occurred. These
data are difficult to find in Dam Design. However, on a Dam Inspection, the data for
calibration can easily get at the local Dam Management Unit.
Rainfall runoff modeling is important to be used in flood analysis considering

several factors as following:
1) Limitation on ability to provide realistic modeling

a) Knowledge on modeling is continuously up-dating.
b) Necessity to always set a hypothesis
c) To find some missing information
d) To create simple system which perceived complex previously
2) Calculating gap or error and deviation or bias.
3) Precision (decimal range and iteration distribution) and accuracy (observation
and calculation gap vs true value) and projection of deviation result will be
similar with actual condition minus simulation result.
Rainfall-Runoff Model consist of three sub model as follows:
– Losses model
– Unit Hydrograph model
– Baseflow model
Since Batutegi watershed consists of 2 large rivers and one small river with
a combination of soil types between loam and clay as well as various land cover,
783
Q.105 - R.47
the Rainfall-Runoff Relationship or RRR Model chosen consists of at least 3 sub-

watersheds. Model parameters in each sub-watershed of the Losses Model are
determined based on the SCS (Soil Conservation Services, US) method in the
form of a Curve Number which is determined by measured parameters, namely the
Hydrologic Soil Group (HSG) derived from the soil type and land cover calculation
matrix [5].
The time lag of the Hydrograph Unit parameter from the SCS model is
calculated based on its topographic conditions. It is also determined for each sub-
watershed which are different from each other. The baseflow parameter uses the
Recession Constant Method which is determined by Trial and Error.
Based on SNI 2415-2016 [2] about the summary of the approach and method
of calculating the design flood used for dam design/reservoir (see Table 1). The
requirements must use the PMF (Probable Maximum Flood) design, therefore the
first step is necessary to calculate the PMP (Probable Maximum Precipitation) in
accordance with SNI No. 7746 of 2012 [3].
Table 1
Summary of flood design methodology
Based on SNI 7746 of 2012 [3] about Calculating Procedure for Maxi-
mum Probable Maximum Precipitation using Hersfield Method (p.2), it is stated
that the data required in calculating the maximum rainfall should be the maxi-
mum annual daily rainfall data. The probable maximum precipitation should be 24
hours.
If rainfall data are less than 24 hours, for example 12 hours, the data required is
the 12-hour maximum annual rainfall which only can be obtained from an automatic
rainfall station. Generally, in Indonesia the rain station measures the maximum daily
rain per year, so the maximum daily rainfall should be 24 hours. Consequently, all
other design rainfall must be 24 hours long as well.
784
Q.105 - R.47
3. COLLECTING HYDROLOGY DATA
Hydrology data generally can be collected from Batutegi Dam Monitoring Unit
as follows:
1. Three times a day of Water Level Reservoir (WLR) data which already sum-
marized as 2002–2020 daily average. Most of the original data of three times
a day of WLR accidentally were deleted, remaining only two years data with
three times recorded a day.
2. Observation outflow from reservoir intake from 2010–2020
3. 2010–2020 Reservoir Inflow –Indirect Observation Data, calculated from
Water Balance Equations with conditions “Outflow and Elevation-Storage Area
Relations” were known utilizing 2017 ESA data.
4. 1998–2019 Batutegi Daily rainfall data, with conditions 5.5 year data is not
available.
5. 1993–2020 ESA Table
6. Dam technical data
Data on point 1–3 graphically can be seen in on picture no 2, and 1993-2020

Elevation – Storage-Area (ESA) table can be seen on Table 2. Based on available
Table 2
1993–2020 Bathymetri measurement result [4]
785
Q.105 - R.47
data, Water Level Reservoir Recording Data is more consistent compared to Inflow
Data (Fig.3), which is dependent on Elevation-Storage-Area (ESA) data and it does
not consider reservoir evaporation, hence negative inflow result is inevitable.
4. RESULT ANALYSIS
Based on 2018-2019 Water Level Reservoir recording data (three times a day),
two times flood were identified as follow (table 3):
Table 3
Water Level Reservoir 2018 and 2018
VARIABLE FIRST FLOOD OCCURRENCE SECOND FLOOD OCCURRENCE
(FEBRUARY 16–18, 2019) (MARCH 4–5, 2018)
Reservoir water level raising (m) 1,4 m 0,4 m
Daily rainfall (mm) 117 mm Day 1 : 71 mm
Day 2 : 18 mm
Firstly in February 16th-18th 2019 and secondly in March 4th–5th 2018 (Fig.4).
On top of that, intake outflow date is also available. Water Level Reservoir position on
those two times flooded is under Normal Water Level at +274 elevation, water level
reservoir rose by 1.4 m in 1st accident and 0.4 m in 2nd accident. Daily rainfall in 1st
flood was 117 mm and in 2nd flood was 71 mm in first day and 18 mm in second day.
Recording of Water Level Reservoir is not dependent on ESA data, while for flood
calibration stage was using ESA, 2020 is the latest data. The result of calibration
and verification showed at Figure 3 and 4, normal water level describe as yellow line
for specific time and grey line describe reservoir water level simulation.
Fig. 4
Flooding March 2018 (left) and February 2019 (right)
Inondations mars 2018 (à gauche) et février 2019 (à droite)
Comparison between simulated water level reservoir and observations cannot

be determined visually through a graph. It should be checked mathematically by an
786
Q.105 - R.47
objective function. The objective functions that measure the performance of the
calibration model are [5,6]: 1) Observations Standard Deviations Ratio or RSR, 2)
Nash Sutcliffe Efficiency (NSE) and 3) Evaluation of the average deviation between
observations and simulations (PBIAS). In general, the three measuring instruments
are ranked as illustrated in Table 3.
The results of calibration model parameters used for the calculation of the
design flood for the Batutegi Reservoir are measured by the RSR, NSE and PBIAS
listed in Table 4. The calibration stage is ranked as Very Good and the verification
stage is in a satisfactory rating. Based on this assessment, it can be said that the
calibration results are acceptable and continue to the design flood prediction stage.
Table 4
Calibration and Verification Performance of Batutegi Dam
NO STAGE RSR NSE PBIAS
1 Calibration (Feb 2019) 0,19 (very good) 0,963 (very good) 1 (very good)
2 Verification (Maret 2018) 0,64 (satisfactory) 0,587 (satisfactory) 1,5 (satisfactory)
Furthermore, the parameters obtained from the calibration stage are used in
the prediction model with the input design rainfall and produce a design flood as
shown in Table 5.
Table 5
Summary of flood design calculation
PERIODE ULANG HUJAN RATA2 MAXIMUM
INFLOW OUTFLOW ELEVASI MA
TAHUN MM M3/S M3/S M
1 HARI
2 78.4 185.5 27.8 274.41
5 105.4 290.3 45.5 274.57
10 124.0 367.3 59.9 274.69
25 148.5 475.6 80.3 274.84
50 167.3 561.1 97.0 274.95
100 186.7 650.4 116.0 275.06
200 206.5 743.9 135.8 275.2
500 233.7 873.4 163.2 275.31
1000 255.0 975.3 185.0 275.43
PMF 576.2 3361.0 789.0 277.85
3 HARI
100 263.7 728.5 173.3 275.36
1000 345.9 1076.9 249.1 275.76
PMF 650.2 3686.7 949.4 278.36
Based on the explanation above, the design rainfall with 24 hours duration is
the input for the flood model. There is a connection between design flood at period
2–10 years return period with daily inflow as stated in Fig. 4. Because peak daily
discharge comes from the peak hourly discharge. The peak discharge ratio divided
787
Q.105 - R.47
Fig. 5
Discharge (m3 /s) relationship to catchment (km2 )
Relation débit (m3 /s) au bassin versant (km2 )
by the 24-hour average discharge for Batutegi Dam is 2. Although the observed
inflow of the Batutegi reservoir in Fig. 4 is not completely reliable, it can be used as
a rough reference for the following reasons:
1. The Floods data came from Batutegi Dam Monitoring Unit. In February 16–18,
2019 (used in the calibration as shown in Fig. 4, left) was 151 m3 /s and 4–5
March 2018 (used in the calibration as shown in Fig. 4, right) was 117.5 m3 /s.
If the peak discharges are multiplied by a ratio of two, it will produce a flood
peak of 302 m3 /s for February 2019 and 235 m3 /s for March 2018 respectively,
which is close to the peak flood calculated for February 2019 of 320 m3 /s and
March 2018 of 230 m3 /s.
2. Both floods occurred when the reservoir water level was below the normal
water level. It is shown that the data were accurate. If the reservoir water level
were above the crest of spillway, the difference in a few cm would result in a
large difference of discharge.
3. From the discharge data series for 21 years gives an estimated maximum
annual average value of 102 m3 /s. If the value is multiplied by the ratio of 2,
the peak discharge reaches 204 m3 /s almost equal to 2 (two) year return flood
period. In Table 5, the 2-year flood return period is 185 m3 /s almost equal to
204 m3 /s.
The largest design flood, which is called the probable maximum flood (PMF),
was derived from PMP. According to technical data (Nippon Koei, 2003), the reservoir
volume at normal water level (El. + 274 m) was 690 MCM with a peak inflow of PMF
788
Q.105 - R.47
5,350 m3 /s with PMP 629 mm. The spillway capacity was 1,930 m3 /s so that the
reservoir water level reaches El. + 281.5 m.
Based on analysis from this study, the reservoir volume at an elevation of

normal water level is 770 MCM with a peak inflow of PMF 3,361 m3 /s with a PMP of
576 mm and a spillway capacity of 789 m3 /s. For the 3-day rain design, the amount
of PMP is 650.2 mm and the peak of the PMF inflow is 3,688 m3 /s which makes the
overflow capacity to 949.4 m3/s, the maximum reservoir water level is El. + 278 m.
The difference between the design flood at the dam planning phase and the current
time can be caused by:
1. Updating the rainfall data. In 2003, when the impounding began, the Batutegi
rainfall station only had 5 (five) years data length. In recent study, the rainfall
data length is almost 16 (sixteen) years.
2. There was reservoir flood data for calibration that could be carried out. The
flood data used is a 2-10 years return period. The model parameters used in
this study are based on calibration and evaluation.
3. At the dam planning phase, the determination of model parameters through
an empirical approach without calibration. So it needs to be compared with
several approaches.
4. If the two PMF inflow magnitudes are plotted on the Creager Curve (Kasiro
et al, 1989) as shown in Fig.5, it can be seen that PMF-2003 is between
C = 100 and C = 150 and PMF-2020 is between C = 100 and C = 60.
5. CONCLUSION AND RECOMMENDATION
The conclusions of this paper are as follows:
1. Based on reservoir water elevation monitoring and daily outflow from 2010–
2020, there were 2 (two) floods which are predicted to have a 10 years return
period and a 2 years return period. These floods were used as a calibration
model for water availability.
2. There is a new benchmark of graphic elevation vs storage capacity based on
bathymetry in 2020.
3. It is suitable to use a calibration model based on 2 (two) floods occurred in
February 2019 and March 2018.
4. The result of daily simulation for 22 years period and flood design for 2–5
years return period, should be harmonized and could be used as reference for
calculation results.
ACKNOWLEDGEMENTS
Special gratitude to Dam Monitoring Unit BBWS Mesuji Sekampung for provid-
ing hydrology data for this paper. Dam Operational Improvement and Safety Project,
789
Q.105 - R.47
Phase II which can make the evaluation happen. And all the Operational and Moni-
toring Staff at Batutegi Dam who always give the best service, so that Batutegi Dam
can operate well.
REFERENCES
[1] Ministry of Public Works and Housing. Minister’s Regulation No. 27/2015 about
Dam. 2015.
[2] SNI 2415–2016. Tata cara perhitungan banjir rencana, Badan Standardisasi
Nasional Indonesia. 2016.
[3] SNI 7746:2012. Tatacara Perhitungan Maksimum Boleh Jadi Dengan Metode
Hersfield, Badan Standardisasi Nasional Indonesia. 2012.
[4] NIPPON KOEI CO. LTD. Buku Pedoman Operasi dan Pemeliharaan Batutegi,
Badan Pelaksana Proyek Induk Pengembangan Wilayah Sungai Way Seputih-
Way Sekampung, Direktorat Jenderal Sumber Daya Air. 2002.
[5] KASIRO, I., MAKMUR, M., MARTAWATI L. Metode Penentuan Banjir dan
Kapasitas Pelimpah Yang Lazim Digunakan Dalam Desain Bendungan di
Indonesia, JLP No.11-12 Th. 3-KW IV.1989
[6] PEREZ-SANCHEZ, M., SANCHEZ-ROMERO, F.J., RAMOS, H.M., LOPEZ-

JUMENEZ, P. A. Calibrating a flow model in an irrigation network: Case study
in Alicante, Spain, Spanish Journal of Agricultural Research 15(1), el202, 13
pages, 2017. https;//doi.org/10.5424/sjar/2017151_10144.
790
Q.105 - R.48
-------
-------
GAME ANALYSIS OF WATER AND SEDIMENT ALLOCATION BETWEEN

CASCADE RESERVOIRS AND LOWER CHANNEL (∗)
Xin WANG
Yellow River Institute of Hydraulic Research, Zhengzhou
Key Laboratory of Yellow River Sediment Research, MWR, Zhengzhou
Yuanjian WANG
Enhui JIANG
Xiang LI
China Institute of Water and Hydropower Research, Beijing
CHINA
SUMMARY
In the context of unified management of water resources, traditional allo-

cation models ignore the competition and cooperation over trans-regional water
resources and game theory has become an appropriate tool to solve this prob-
lem This paper proposes an integrated model framework to assess the water and
sediment benefits sharing schemes among the three stakeholders in the basin,
namely, Xiaolangdi, Xixiayuan, and lower channel. Two objectives are considered
∗ Simulation de l’allocation de l’eau et des sédiments entre des réservoirs en cascade et le
chenal inférieur
DOI: 10.1201/9781003211808-94 791

Q.105 - R.48
for this assessment—Hydropower (HP), and Sediment Release (SR). The optimiza-
tion model, and cooperative and non-cooperative game theory methods are jointly
used to analyze the water and sediment benefits of stakeholders under different
scenarios. This paper takes the measured hydrological data during the flood sea-
son from July to August 2019 as an example to calculate the results of the dynamic
allocation of water and sediment resources between cascade reservoirs and lower
channel under different game scenarios. The results show that, when the benefits
are no longer allocation, the maximum benefit can be obtained from the grand coali-
tion. From the individual optimal point of view, the Xiaolangdi reservoir can obtain
the maximum benefit when it chooses non-cooperative game, Xixiayuan reservoir
has the best benefit in the case of regional cooperative of Xiaolangdi-Xixiayuan
reservoir, and the benefit of the lower channel is the best under the condition of the
regional cooperation between Xixiayuan reservoir and the lower channel.
RÉSUMÉ
Dans le contexte d’une gestion unifiée des ressources en eau, les modèles
d’allocation traditionnels ignorent la concurrence et la coopération sur les ressources
en eau transrégionales, et la théorie des jeux est devenue un outil approprié pour
résoudre ce problème. Ce rapport propose un cadre de modèle intégré pour évaluer
les schémas de partage des bénéfices des sédiments entre les trois acteurs du
bassin, à savoir Xiaolangdi, Xixiayuan et le chenal inférieur. Deux objectifs sont
considérés pour cette évaluation: l’hydroélectricité (HP) et le rejet de sédiments
(SR). Le modèle d’optimisation et les méthodes de théorie des jeux coopératives
et non coopératives sont utilisées conjointement pour analyser les avantages de
l’eau et des sédiments des parties prenantes dans différents scénarios. Cet article
prend les données hydrologiques mesurées pendant la saison des crues de juillet
à août 2019 comme exemple pour calculer les résultats de l’allocation dynamique
des ressources en eau et en sédiments entre les réservoirs en cascade et le chenal
inférieur dans différents scénarios de jeu. Les résultats montrent que, lorsque les
avantages ne sont plus attribués, le bénéfice maximal peut être obtenu de la grande
coalition. Du point de vue individuel optimal, le réservoir Xiaolangdi peut obtenir le
maximum d’avantages lorsqu’il choisit un jeu non coopératif, le réservoir Xixiayuan a
le meilleur avantage dans le cas de la coopérative régionale du réservoir Xiaolangdi-
Xixiayuan, et l’avantage du canal inférieur est le meilleur sous la condition de la
coopération régionale entre le réservoir Xixiayuan et le chenal inférieur.
1. INTRODUCTION
Water is one of the most important resources in social development, however,

the construction of mega cities in China is facing the dual pressure of water shortage
and regional water competition, which presents a new challenge to the efficiency
792
Q.105 - R.48
and fairness of rational allocation of water resources(Su et al, 2019). Some achieve-
ments have been made in the research of water resource allocation at home and
abroad. The mathematical analysis model is usually used to maximize the objective
function for calculation and comprehensive evaluation to form the scheme results for
decision-making reference (Wu et al, 2004). Or use relevant software to configure,
such as MIKEBASIN (M K et al, 2003), RIVERWARE (Maimaitimin, 2003), IQQM
(Zhang, 2014), WEAP etc(Hu et al, 2009; Mutiga et al, 2010). But in the context of
unified water management, The traditional configuration model cannot fully consider
the inter-regional water competition and cooperation, and game theory becomes a
better tool to solve this problem. In recent years, scholars at home and abroad have
used game theory to solve the conflict of interest in cross-regional water use in terms
of project cost allocation (Fu et al, 2000), water resource quantity and quality alloca-
tion (Yin et al, 2004), initial water right allocation (Kong et al, 2005), compensation
mechanism (Xiao et al, 2005), and water fee regulation (Filho et al, 2008; Dinar
et al, 1992), and use the cooperative game theory to find the equilibrium stability
of cross-regional water use. Degefu (Degefu et al, 2016). based on the summary
of the cross-regional water resources game, the Nash bargaining model under the
cooperative game is used to establish the water resources allocation mechanism
under the condition of water shortage, which provides a new idea for solving the
problem of efficiency and equity of cross-regional water resources.
At present, in terms of the conflict of water resources allocation, most of the

relevant literatures focus on the conflict of water resources allocation among differ-
ent water users, and there are few researches on the dynamic analysis of water
resources -water quality between different players based on game theory. There
are few researches on the game model construction and solution analysis for the
dynamic regulation of water and sediment in the reservoir group and downstream
channel of multi-sandy river. In this paper, the empirical formula of the ratio of sand
discharge in Xiaolangdi reservoir (Zhang et al, 2018) and the empirical formula of
the ratio of sand discharge in downstream channels (Fei et al, 2008) is used to
quantify the relationship between water and sand, and the characteristics of water
and sand resources (namely, the relationship between water and sand in the sand-
heavy flow, the positive and negative characteristics of water and sand resource
benefits, etc.) are comprehensively considered. And combined with multi-sandy
river reservoir group and downstream river channel in the actual water-sand joint
operation scheduling principle and various constraints, the reservoir group in the
middle reaches of the Yellow River (Xiaolangdi – Xixiayuan) – downstream river
as an example, using the game theory to build the Xiaolangdi – Xixiayuan – the
downstream river water and sediment resources dynamic allocation model, And the
model is solved under different situations.
2. RESEARCH AREA AND DATA
Xiaolangdi reservoir is located between Mengjin County and Jiyuan City in

Luoyang City, Henan Province, China. It is 130 km downstream of Sanmenxia
793
Q.105 - R.48
Reservoir and 40km north of Luoyang City in Henan Province. The Xiaolangdi Reser-
voir controls a drainage area of 694,000 km2 , accounting for 92.3% of the area of
the Yellow River Basin. Xiaolangdi reservoir is a large-scale comprehensive water
conservancy project integrating sediment reduction, flood control, ice prevention,
water supply and irrigation, and power generation on the main stream of the Yel-
low River. Xiaolangdi reservoir has a crest elevation of 281m, a storage capacity of
12.65 billion m3 , a silt storage capacity of 7.55 billion m3 , and a long-term effective
storage capacity of 5.1 billion m3 The dead water level of Xiaolangdi reservoir is
230 m, the flood control limit water level in flood season is 254 m, and the limit water
level for ice prevention is 266 m. The maximum flood discharge is 17000 m3 /s.
Xixiayuan reservoir is a matching engineering of Xiaolangdi Reservoir on the

Yellow River. It is located on the main stream of the Yellow River 16 km downstream
of the dam site of Xiaolangdi reservoir, 25 km away from Luoyang City and 10 km
away from Mengjin County. The task of the reservoir is reverse regulation, combined
with power generation, irrigation and water supply. Xixiayuan reservoir has a total
storage capacity of 162 million m3 , a normal water level of 134 m, a flood control
level of 131 m, and an average annual power generation of 583 million KW ∗ H.
The lower channel of the Yellow River in this paper refer to the reach of the
Yellow River below Taohuayu in Zhengzhou, Henan Province. The length of the chan-
nel is 786 km, the drainage area is only 23000 km2 , accounting for 3% of the whole
basin area; the total head of the lower channel is 93.6 m, with an average gradient
of 0.12‰; the increased water volume in the interval accounts for 3.5% of the total
water volume of the Yellow River. Due to the large amount of sediment in the Yellow
River, the lower channel of the Yellow River have been silted up for a long time to
form a world-famous “The Aboveground River”.
This paper takes the measured hydrological data from July 1st to August 31st
during the flood season of 2019 as an example to design and calculate the game
model for the dynamic allocation of water and sediment resources in the Xiaolangdi-
Xixiayuan-lower channel under different scenarios. Among them, the daily flow
data and reservoir data used in this paper were obtained from the Yellow River
Conservancy Commission (http://10.4.0.21/yrccweb.nsf?open) and the Yellow River
hydrological information query and consultation system (http://10.4.1.9/sqweb/).
3. MODEL CONSTRUCTION
There are usually three ways to solve the joint operation scheme of multi reser-
voir group by using optimization algorithm (Deepti et al, 2010): (1) transforming part
of the objectives into constraints; (2) adding the weights of the original incommen-
surable multi-objective by mathematical method; (3) using Pareto optimal surface
to express the optimal solution group. Zhao et al [18] used the cooperative game
theory to analyze the allocation of transboundary water benefits in the Mekong River
794
Q.105 - R.48
Basin of Lancang River, and divided the construction of the game model into four
stages: (1) setting the cooperation scenarios; (2) calculating stakeholders’ direct
water benefits under different scenarios; (3) reallocating the incremental water ben-
efits from coalitions based on CGT, while ensuring that the water benefit sharing
schemes make all stakeholders better off; and (4) assessing the effects of damming
and model sensitivity. In this paper, the sediment resource allocation problem is inte-
grated into the dynamic allocation of water resources by using the empirical formula
of water and sediment, and the modeling process is consistent with that of Zhao
et al (Li et al, 2019).
3.1. GAME SCENARIOS SETTING
According to the actual situation of water and sediment joint operation of

Xiaolangdi reservoir, Xixiayuan reservoir and lower channel, the game scenarios
can be divided into 3 types: grand coalition (that is, all players follow the principle of
collective interests), independent (that is, all players follow the principle of individual
interests), and regional cooperation (that is, the coalition of two players follows the
collective interests principle), The third player follows the individual interests princi-
ple). The specific game scenarios are as follows: (1) Grand coalition with all three
stakeholders, ({Xiaolangdi, Xixiayuan, Lower channel}), where {} denotes a coalition;
(2) all Independent (Xiaolangdi, Xixiayuan, Lower channel); (3) Regional coopera-
tion, such as ({Xiaolangdi, Xixiayuan}, Lower channel), (Xiaolangdi, {Xixiayuan,
Lower channel}), ({Xiaolangdi, Lower channel}, Xixiayuan). For the convenience of
description, this paper uses the standard expression of n-person game to express
the above game scenario as follows:
(1) Grand coalition
Stakeholder set: N = {(R1 ,R2 ,C1 )}, Stakeholder strategy set: X = {XR1 ,XR2 },
Benefit function set: U = {(B1 + B2 + B3)};
(2) Independent
Stakeholder set: N = ({R1 },{R2 },{C1 }), Stakeholder strategy set: X = (XR1 ,XR2 ),
Benefit function set: U = ({B1},{B2},{B3});
(3) Regional cooperation
1) ({Xiaolangdi, Xixiayuan}, Lower channel)
Stakeholder set: N = ({R1 ,R2 },{C1 }), Stakeholder strategy set: X = ({XR1 ,XR2 }),
Benefit function set: U = ({B1 + B2},{B3});
2) (Xiaolangdi, {Xixiayuan, Lower channel})
795
Q.105 - R.48
Stakeholder set: N = ({R1 }, {R2 , C1 }), Stakeholder strategy set: X = (XR1 ,XR2 ),
Benefit function set: U = ({B1},{B2 + B3});
3) ({Xiaolangdi, Lower channel}, Xixiayuan)
Stakeholder set: N = ({R1 ,C1 },{R2 }), Stakeholder strategy set: X = (XR1 ,XR2 ),
Benefit function set: U = ({B1 + B3},{B2});
Among them, R1 and R2 are Xiaolangdi and Xixiayuan reservoirs respectively,

C1 is the lower channel, XR1 and XR2 are the strategic set of Xiaolangdi Reser-
voir and Xixiayuan reservoir respectively (i.e. daily outflow control variables of the
two reservoirs, such as daily average outflow, sediment discharge, daily average
reservoir water level, etc.), and the formula is as follows:

XR1 = Q1,i , Qs1 ,i |i = 1, . . . , T (1)

XR2 = Q2,i , Qs2 ,i |i = 1, . . . , T (2)
Among them, Q1,i is the i th average daily discharge of Xiaolangdi Reservoir,

m3 /s; Qs1 ,i is the i th average daily sediment concentration of Xiaolangdi reservoir,
kg/m3 ; Q2,i is the i th average daily discharge of Xixiayuan Reservoir, m3 /s; Qs2 ,i is
the i th average daily sediment concentration of Xixiayuan reservoir, kg/m3 ;. T = 62,
the total number of days is calculated in this model.
It is the basis of reservoir optimal operation to maximize the comprehensive

benefit of reservoir. Therefore, in this paper, taking the power generation benefit
reducing the sedimentation loss in the reservoir area as the comprehensive benefit
function consideration factor of the reservoir, the sedimentation amount of the down-
stream river channel as the benefit function consideration factor of the downstream
river channel, and the constraint conditions are the reservoir flood control (including
ice jam prevention), water supply, irrigation, eliminating harm and making profits.
The expressions of B1, B2 and B3 are as follows:
(1) Xiaolangdi reservoir

B1 Q1,1 , Q1,2 , . . . , Q1,62 = a1 E1 − b1 V1 (3)
a1 and b1 are weights respectively; E1 is the power generation benefit of

Xiaolangdi reservoir, kW ∗ H; V1 is the accumulated sedimentation capacity of
Xiaolangdi reservoir, billion m3
1) power generation benefit
62
E1 = K1 Q1,i H1,i − H0 t (4)
i=1
796
Q.105 - R.48
K1 is the output coefficient of Xiaolangdi Reservoir power station; Q1,i is the

average daily discharge flowing the turbine of Xiaolangdi Reservoir on the i th day,
m3 /s. H1,i is the average water level in front of Xiaolangdi Reservoir on the i th day,
m; H0 is the elevation of power tunnel of Xiaolangdi reservoir, m; t is the calculation
time of the period (1 hour).
2) sedimentation loss
62
V1 = WS − WS1,i (5)
i=1
WS is the total sediment inflow in the upstream within 62 days (from July 1 to
August 31), billion tons; WS1,i is the sediment discharge of Xiaolangdi Reservoir on
the i th day, billion tons.
3) Curve fitting of relationship between water level and reservoir capacity
H1,i = −0.0001054576V1,i
4
+ 0.0093516024V1,i
3
− 0.2930624492V1,i
2
+ 4.5318300134V1,i + 206.6162585568
R = 0.99183
2
(6)
V1,i is the i th day storage capacity of Xiaolangdi reservoir, billion m3
4) Relationship between storage capacity and discharge (water balance rule)
i−1
V1,i = V1,0 + wk i = 1, 2, . . . , 62 (7)
k=1

wk = Q0,k − Q1,k ∗ 0.000864 (8)
V1,0 is the initial storage capacity of Xiaolangdi Reservoir (June 31), billion m3 ;
wk is the capacity variable of Xiaolangdi Reservoir on the k th day, billion m3 ; Q0,k
is the average daily inflow of Xiaolangdi Reservoir on the k th day, m3 /S; Q1,k is the
average daily discharge of Xiaolangdi Reservoir on the k th day, m3 /s.
5) Empirical formula of the ratio of sediment discharge in Xiaolangdi

reservoir [16]

V1,i
WS1,i = −25.53ln − 81.219 ∗ WS0,i (9)
Q0,i + Q1,i /2
797
Q.105 - R.48
WS0,i is the inflow sediment storage capacity of Xiaolangdi Reservoir on the

i th day, billion tons.
(2) Xixiayuan Reservoir

B2 Q2,1 , Q2,2 , . . . , Q2,62 = a2 E2 (10)
a2 is the weight; E2 is the power generation benefit of Xixiayuan reservoir, kW ∗ H.
1) power generation benefit
62
E2 = K2 Q2,i H2,i − H0 t (11)
i=1
K2 is the output coefficient of Xixiayuan reservoir power station; Q2,i is the

average daily discharge flowing the turbine of Xiaxiayuan reservoir on the i th day,
m3 /s. H2,i is the average water level in front of Xixiayuan reservoir on the i th day, m;
H0 is the elevation of power tunnel of Xixiayuan reservoir, m; t is the calculation
time of the period (1 hour).
2) Curve fitting of relationship between water level and reservoir capacity
H2,i = −3.8186V2,i
2
+ 10.08V2,i + 127.4 (12)
V2,i is the i th day storage capacity of Xixiayuan reservoir, billion m3
3) Relationship between storage capacity and discharge (water balance rule)
i−1
V2,i = V2,0 + wk i = 1, 2, . . . , 62 (13)
k=1

wk = Q1,k − Q2,k 0.000864 (14)
V2,0 is the initial storage capacity of Xixiayuan reservoir (June 31), billion m3 ;
wk is the capacity variable of Xixiayuan reservoir on the k th day, billion m3 ; Q2,k
is the average daily discharge of Xixiayuan Reservoir on the k th day, m3 /s.
(3) Lower channel

B3 Q2,1 , Q2,2 , . . . , Q2,62 = −b2 V3 (15)
a2 is the weight; V3 is the total sediment deposition in lower channel, t.
798
Q.105 - R.48
1) sedimentation loss
62
V3 = WS3,i (16)
i=1
V1 is the daily sedimentation volume of the downstream channel, billion m3
2) Empirical formula of daily sedimentation in lower channel [16]
⎧ 0.47
⎪
⎨ 86.4Q2 QS2,k Q
− 0.108 QS2,k
QS2,k
≥ 0.015
2,k Q2,k Q2,k
WS3,i = 2,k
(17)
⎪
⎩ QS2,k
0 Q2,k < 0.015
QS2,k is the out-flow sediment concentration of Xixiayuan reservoir on the i th

day, kg/m3
3.2. OBJECTIVE FUNCTION UNDER DIFFERENT GAME SCENARIOS
(1) Grand coalition
U1 = Max (B1 + B2 + B3 ) (18)
U1 is the comprehensive objective function in Grand coalition.
(2) Independent
U21 = Max(B1 ) (19)
U22 = Max (B2 |B1 ) (20)
U23 = B3 (B1 , B2 ) (21)
Among them, U12 , U22 , U32 are the objective functions of Xiaolangdi reservoir,
Xixiayuan reservoir and downstream river channel in the Independent game coalition
separately.
(3) Regional cooperation
1) scenario ({R1 , R2 }, {C1 })
U31,2 = Max(B1 + B2 ) (22)
U33 = B3 (B1 , B2 ) (23)
799
Q.105 - R.48
Among them, U1,2 3 is the comprehensive objective function of Xiaolangdi-

Xixiayuan reservoir in scenario ({R1 , R2 }, {C1 }), U33 is the objective function of the
lower channel in this scenario.
2) scenario ({R1 , C1 }, {R2 })
U31,3 = Max(B1 + B3 ) (24)

U32 = B2 (B1 ) (25)
Among them, U1,2 3 is the comprehensive objective function of Xiaolangdi-Lower

channel in scenario ({R1 , C1 }, {R2 }), U2
3 is the objective function of the Xixiayuan
reservoir in this scenario.
3) scenario ({R1 }, {R2 , C1 })
U32,3 = Max(B2 + B3 ) (26)

U31 = B1 (27)
Among them, U2,3

3 is the comprehensive objective function of Xixiayuan-
Lower channel in scenario ({R1 }, {R2 , C1 }), U1
3 is the objective function of the
Xiaolangdi reservoir in this scenario.
3.3. CONSTRAINT CONDITION
The game schemes in different game situations should satisfy the following
constraints:
(1) Water level constraint
H1,i ∈ [205, 250] i = 1, 2, . . . , 62 (28)

H2,i ∈ [123, 134] i = 1, 2, . . . , 62 (29)
In order to meet the flood control requirements, the daily average water level
in front of Xiaolangdi and Xixiayuan dams should be within the range of [205,250]
and [123,134] according to the measured data in flood seasons over the years.
(2) Discharge flow constraint
Q1,i , Q2,i ∈ [200, 4500]i = 1, 2, . . . , 62 (30)
800
Q.105 - R.48
Table 1
Parameters of game model
MEANING OF PARAMETERS VALUE
output coefficient K1 8.5
K2 8.3
weight a1 0.3
a2 0.3
b1 5.66
b2 3.77
elevation of power tunnel H 206.61
H 137.4
initial storage capacity V1,0 16.26
V2,0 0.58
initial water level H1,0 236.38
H2,0 131.79
Targeted water level H1,62 242.67
H2,62 131.78
The lower limit of the reservoir discharge is determined according to the

requirements of downstream water supply and ecological water demand, and the
upper limit is determined by the flood control safety requirements in the down-
stream flood season. According to experience, the average daily discharge of the
two reservoirs should be within the range of [200,4500].
(3) Optimal scheduling constraints

H1,0 = H1,0 (31)

H1,62 = H1,62 (32)

H2,0 = H1,0 (33)

H2,62 = H2,62 (34)
The optimal operation scheme should meet the same initial water level, that
is, water balance. H1,0 , H1,62 are the starting and ending water levels of Xiaolangdi
Reservoir under real operation, M; H2,0 , H2,62 are the starting and ending water
levels of Xixiayuan reservoir under real operation, M.
(4) discharge flowing the turbine constraints
When Q1,i > 1800, take Q1,i = 1800; When Q2,i > 1800, take Q2,i = 1800.
The maximum discharge of Xiaolangdi and Xixiayuan reservoirs is 1800 m3 /s,

and various parameters are shown in the table below.
801
Q.105 - R.48
4. RESULTS
The design and calculation of the game scheme of the three parts: grand
coalition, independent, regional cooperation (such as scenario ({R1 , R2 }, {C1 }),
scenario ({R1 , C1 }, {R2 }), scenario ({R1 }, {R2 , C1 }), under different circumstances
are as follows:
Table 2
Calculation results of game model in different game scenarios
BENEFIT GRAND INDEPENDENT SCENARIO SCENARIO SCENARIO
(BILLION RMB) COALITION ({R1 , R2 }, {C1 }) ({R1 }, {R2 , C1 }) ({R1 , C1 }, {R2 })
Xiaolangdi 4.9502 5.0797 5.0717 1.4443 5.0023
Xixiayuan 0.5041 0.5232 0.5432 0.5949 0.1812
Lower channel −0.4090 −0.8967 −0.9056 0 −0.4090
Total 5.0453 4.7062 4.7093 2.0392 4.7745
It can be seen from the above table that the dynamic game results of reservoir
group-downstream channel water and sand distribution are different under differ-
ent game scenarios. From an overall perspective, the distribution effect of water
and sediment resources of reservoir group-downstream channel is the best under
the scenario of grand coalition (comprehensive benefit value is 504.5 million yuan).
The worst effect of water sediment distribution is scenario ({R1 }, {R2 , C1 }) (com-
prehensive benefit value: $203.9 million). According to the principle of individual
optimality of a single stakeholder, the schemes of optimal benefit and lowest benefit
are different for different stakeholder. Among them, Xiaolangdi Reservoir has the
best benefit (507.9 million yuan) without any cooperation, and the lowest benefit is
scenario ({R1 }, {R2 , C1 }) ($144.4 million), Xixiayuan Reservoir has the best benefit
in scenario ({R1 , R2 }, {C1 }) (54.3 million yuan), and the lowest benefit in scenario
({R1 , C1 }, {R2 }) (181.2 million yuan). The benefit of the downstream channel is the
best in scenario ({R1 }, {R2 , C1 }) (RMB 0 billion, i.e., the downstream channel is free
of sediment deposition), while the benefit is the lowest in scenario ({R1 , R2 }, {C1 })
(RMB −0.9056 billion). The distribution of sediment resources under different game
scenarios is shown in the figure below.
5. CONCLUSIONS
(1) Under the background of no redistribution of interests, it starts from the

perspective of individual optimality, in the dynamic game model of dynamic distri-
bution of water and sediment resources of xiaolangdi-xixiayuan-downstream river
channel, the maximum benefits can be obtained when xiaolangdi chooses Indepen-
dent. Xixiayuan reservoir has the best benefit in scenario ({R1 , R2 }, {C1 }), and the
downstream channel has the best benefit in scenario ({R1 }, {R2 , C1 }).
802
Q.105 - R.48
Fig. 1
Distribution of water and sediment resources in different game scenarios
(2) Under the background of no redistribution of interests, the distribution effect

of water and sediment resources in the reservoir group-downstream channel is the
best in the grand coalition, while the distribution effect of water and sand resources
in scenario ({R1 }, {R2 , C1 }) is the worst.
(3) The stability of cooperative game lies in the fact that the benefits of each
member participating in the grand coalition are higher than those under any other
scenarios. The cooperative game scenarios discussed in this paper (including grand
coalition and regional cooperation) do not consider the use of benefit redistribution
method (for example, Shapley value, Gately point, and Nash-Harsanyi solution,
etc.) to redistribute the benefits of each stakeholder. Therefore, the game scenario
calculation and analysis of dynamic distribution of water and sediment resources
in xiaolangdi-xixiayuan-downstream channel under different benefit redistribution
treatment schemes is the future research direction.
ACKNOWLEDGEMENTS
This research was funded by the National Key Research and Development
Program of China (2018YFC0407400), the National Natural Science Foundation of
China (51679104 and 51539004) and Central Public-interest Scientific Institution
Basal Research Fund (HKY–JBYW–2019–22)
803
Q.105 - R.48
REFERENCES
[1] Xinyue Su, Yang Yu, Jianshi Zhao, et al. Game Analysis of Trans-Regional
Water Resources Allocation in Beijing [J]. Journal of Basic Science and
Engineering, 2019, 27(02):239–251.
[2] Zening Wu Lisheng Suo. Advance about study of water resources optimal
distribution [J]. Journal of Irrigation and Drainage, 2004, (2):1–5.
[3] Jha M K, Gupta A D. Application of mike basin for water management strategies
in a watershed [J]. Water International, 2003, 28(1):27–35.
[4] Maimaitimin Kudereti. Application of RIVERWARE model in Urumqi River

Basin[D]. Urumqi:Xinjiang Agricultural University, 2015.
[5] Rentian Zhang. Study on sustainable utilization for regional water resources
[D]. Nanjing:Hohai University, 2014.
[6] Litang Hu, Zhongjing Wang, Robin Wardlaw, et al. Application of improved
WEAP model in water resources management[J]. Journal of Hydraulic Engi-
neering, 2009, (2):173–179.
[7] Mutiga J K, Mavengano S T, Su Z B, et al. Water allocation as a planning tool to

minimize water use conflicts in the Upper Ewaso Ng’iro North Basin, Kenya[J].
Water Resources Management, 2010, 24(14):3939–3959.
[8] Chun Fu, Zhenpeng Hu. A game analysis approach of cost allocation
for multipurpose water projects[J]. Journal of Hydraulic Engineering, 2000,
(4):57–63.
[9] Yunsong Yin, Fengping Meng, Zhongchun Mi. Game analysis on the double
conflicts between water resources quantity and quality distribution in River
Basins[J]. Quantitative & Technical Economics, 2004, (1):136–140.
[10] Ke Kong, Jiancang Xie, Xinli Yue, et al. Game analysis of water resources
market[J]. Journal of Hydraulic Engineering, 2005, (4):491–495.
[11] Zhijuan Xiao, Jiancang Xie, Ke Kong, et al. Game analysis for emergent water
dispatch compensation[J]. Advances in Water Science, 2005, (6):817–821
[12] Filho F A S, Lall U, Porto R L. Role of price and enforcement in water

allocation: insights from game theory[J]. Water Resources Research, 2008,
44(12):W12420
[13] Dinar A, Ratner A, Dan Y. Evaluating cooperative game theory in water

resources[J]. Theory and Decision, 1992, 32 (1):1–20.
[14] Degefu D M, He W, Yuan L, et al. Water allocation in transboundary river basins

under water scarcity:a cooperative bargaining approach[J]. Water Resources
Management, 2016, 30(12):4451–4466.
804
Q.105 - R.48
[15] Shuai Zhang, Junqiang Xia, Tao Li. Study on Flood-Season Sediment Delivery
Ratio of Xiaolangdi Reservoir [J]. Yellow River, 2018, 40(1): 7–11.
[16] Xiangjun Fei, Xvdong Fu, Ren Zhang. Study on Sediment Delivery Ratio, Rate
of Silting and Characteristic of Sediment Transport of the Lower Yellow River
Channel [J]. Yellow River, 20089, (11):6–8.
[17] Deepti Rani, Maria Madalena Moreira . Simulation-optimization modeling:

a survey and potential application in reservoir systems operation [J]. Water
Resources Management, 2010, 24: 1107–1138.
[18] Dongnan Li, Jianshi Zhao, Govindaraju Rao S. Water benefits sharing under
transboundary cooperation in the Lancang-Mekong River Basin[J]. Journal of
Hydrology, 2019, 577:123989.
805
Q.105 - R.49
-------
-------
SEDIMENTATION PROBLEMS AND MANAGEMENT STRATEGIES OF

LIUJIAXIA RESERVOIR, YELLOW RIVER, CHINA (∗)
Qiang WANG, Enhui JIANG, Yuanjian WANG & Like LI

Key Laboratory of Yellow River Sediment of the Ministry of Water Resources,
Henan Provincial Engineering Research Center of Reservoir-Lake Function
Restoring and Maintaining, Zhengzhou
CHINA
SUMMARY
Sediment deposition is the principal problem affecting the useful life of reser-
voirs, and the reservoirs sedimentation management is one of the important methods
to solve the sustainable development of reservoirs. Liujiaxia Dam, located on the
upper reach of Yellow River, in China, due to the desilting tunnel reserved at the
initial stage of the dam construction can no longer solve the sedimentation problem,
resulting in a significant decrease in the using capacity of the reservoir. In order to
solve this problem, creatively proposed a solution of coupling of “accessory path-
way sediment transport” and “density current sediment transport”, which effectively
solved the sedimentations problems of Liujiaxia reservoir. Starting from coordinat-
ing the water-sediment relationship, it is proposed to build an additional desilting
tunnel project at the bottom of the reservoir (the bottom of the reservoir bank) along
the path of the density flow of the reservoir, which is verified by a water-sediment
model with good results. This solution is to explore the use of engineering sediment
discharge and silt reduction measures, gradually improve the current situation of the
serious imbalance of the original water-sediment Technology gaps.
∗ Problèmes de sédimentation et stratégies de gestion du réservoir LIUJIAXIA sur le Fleuve
Jaune (Chine)
806 DOI: 10.1201/9781003211808-95

Q.105 - R.49
RÉSUMÉ
Le dépôt de sédiments est le principal problème affectant la durée de vie

utile des réservoirs, et la gestion de la sédimentation des réservoirs est l’une des
méthodes importantes pour résoudre le développement durable des réservoirs. Le
barrage de Liujiaxia, situé sur le cours supérieur du fleuve Jaune, en Chine, du fait
du tunnel de désensablage réservé au stade initial de la construction du barrage
ne peut plus résoudre le problème de sédimentation, entraînant une diminution
significative de la capacité d’utilisation du réservoir. Afin de résoudre ce problème,
a proposé de manière créative une solution de couplage du «transport de sédiments
par voie accessoire» et du «transport de sédiments par courant de densité», qui a
effectivement résolu les problèmes de sédimentations du réservoir de Liujiaxia. En
partant de la coordination de la relation eau-sédiment, il est proposé de construire
un projet supplémentaire de tunnel de dessalement au fond du réservoir (le fond
de la berge du réservoir) le long du trajet de l’écoulement de densité du réservoir,
qui est vérifié par une eau -sédiment modèle avec de bons résultats. Cette solution
consiste à explorer l’utilisation des mesures d’ingénierie de rejet de sédiments et
de réduction du limon, d’améliorer progressivement la situation actuelle du grave
déséquilibre des lacunes technologiques d’origine eau-sédiment.
1. INTRODUCTION
Reservoir sedimentation has always been a challenge for dam owners, which
impairing reservoir operation, decreasing storage and further affecting the channel
form and stability of the ecosystem (Fu et al., 2008; Kondolf et al., 2014; Song
et al., 2018). The average annual storage loss of large-scale reservoirs in the world
accounts for 0.5% to 1% of the remaining storage capacity, and China’s 2.3%, which
is much higher than the world average (Palmieri et al., 2003; Xie et al., 2013).
Sediment deposition upstream of dams reduces the water storage capacity and
hydroenergy generation potentials of the reservoirs and decreases in sediment load
downstream of dams often lead to accelerated erosion and loss of floodplains (Fu
et al. 2008; Wang et al. 2005). For example, the Sanmenxia Reservoir has attracted
the attention of the world due to its severe sedimentation problems (Wang et al.,
2006; Wu et al., 2007).
The historical conditions of dam construction, problems in reservoir design,

operation and management, underestimation of the role of water and soil conser-
vation and reservoir storage, etc. will all lead to serious reservoir capacity loss (Wu
et al., 2007). The sedimentation problem was considered when the dam was built,
but most reservoirs are still designed according to the “reservoir life” (for example,
50 or 100 years of use) (Morris and Fan, 1997; Sardar et al., 2014). Therefore, the
long-term sediment management strategy is an effective way to solve the current
reservoir sediment problem (Vorosmarty et al., 2003; Wu et al., 2004; Kondolf et al.,
807
Q.105 - R.49
2014; Morris, 2020). The use of advanced management concepts and the com-
prehensive use of engineering and non-engineering measures to intervene in the
operation of the reservoir is an effective way to realize the sustainable development
of the reservoir. Reservoir sedimentation prevention and control measures include
intercepting the upstream of the reservoir, using flood discharge and sediment dis-
charge facilities to discharge sediment, using mechanical equipment to excavate
the reservoir sediment, and the effective utilization of the discharged sediment (Wu
et al., 2007; Morris, 2020). However, these current methods are may still not be
possible to solve some heavily silted reservoir areas.
The Liujiaxia Reservoir in China offers an excellent example of the conse-

quences of reservoir sedimentation, sediment management strategies, and potential
solutions to critical sedimentation problems. After 1974, Liujiaxia Hydropower Sta-
tion mainly used the original left bank discharge channel, right bank sediment
discharge tunnel and other flood discharge and sediment discharge structures to
carry out more than 100 times of Tao River density current discharge and four times
of low water level discharge, which helped to relieve the sediment in front of the dam.
Siltation played a positive role. However, since the dead reservoir of the Tao River
became full, the reservoir area has become a natural river course, gradually losing
its ability to regulate sediment. The incoming sediment of the Tao River reaches
the dam in the form of density flow, although most of it can be discharged outside
the reservoir (sediment discharge ratio) 60%), but the siltation in front of the dam
is still obvious, and the effect on alleviating the siltation in front of the dam is not
significant, and the abrasion of the silt passing through the machine is more serious;
for the low water level dragging, due to the wide area involved and the large water
loss, it is difficult to coordinate the implementation It is also difficult to implement
for a long time. The existing sediment discharge facilities cannot completely solve
the serious harm caused by the sedimentation in front of the dam to the Liujiaxia
Reservoir. How to solve the sediment problem of Tao River is the key to improving
the sediment problem of Liujiaxia Reservoir.
In this paper, in response to the serious sedimentation problem of Liujiaxia

Reservoir and the serious inadequacy of the current sediment removal measures,
a creatively proposed “density flow desilting” and “bypass desilting” coupled desilt-
ing scheme and use a one-dimensional water and sediment model to explore the
scheme. The unique desilting scheme in this case study has certain guiding signifi-
cance for the sustainable development and management of the upper Yellow River
reservoir.
2. CONSTRUCTION OF LIUJIAXIA DAM
Liujiaxia Reservoir (Figure 1a) is a large reservoir on the main stream of the
Yellow River. It is a water conservancy and hydropower project that focuses on
power generation and has comprehensive utilization benefits such as flood control,
808
Q.105 - R.49
irrigation, snow prevention, and aquaculture. The project was built in September
1958, but stopped from 1961 to 1963, and then continued in 1964 until the water
was formally stored in October 1968. The normal storage level of the reservoir is
1,735 m, the flood control limit water level is 1,726 m, and the dead water level is
1,694 m. The reservoir area is composed of three parts: the main stream of the
Yellow River (Cross section: Huang 0-Huang 38), the tributary Daxia River (Cross
section: Da 1-Da 9) and the Tao River reservoir (Cross section: Tao 1-Tao 21) area
(Figure 1b). The Tao River and Daxia River merge at 1.5km and 26km upstream of
the dam site respectively. The original storage capacity is 5.74 × 109 m3 below the
normal storage level of 1,735 m, of which the mainstream reservoir area accounts
for 94%, the Tao River reservoir area accounts for 2%, and the Daxia River reservoir
area accounts for 4%. The storage capacity below the dead water level of 1,694 m
is 1.54 × 109 m3 , and the effective storage capacity is 4.2 × 109 m3 .
Fig. 1
Plan view of Liujiaxia Reservoir:
(a) Sketch of the upper Yellow River; (b) Sketch of the Liujiaxia Reservoir.
809
Q.105 - R.49
3. CURRENT MANGEMENT OF RESERVIOIR SEDIMENTATION
3.1. FLOW AND SEDIMENT CONDITIONS OF THE RESERVOIR
The Liujiaxia Reservoir’s inflow control hydrological stations include the main
stream Xunhua station, Hongqi station in the Tao River and Zheqiao station in the
Daxia River. Among them, the tributary Daxia River has less annual runoff and sed-
iment transport and will not be analyzed. The Xunhua Station on the mainstream of
the Yellow River controls a drainage area of 1.45 × 105 km2 . The measured runoff
from 1968 to 2010 was 2.05 × 1010 m3 , the annual measured sediment transport
was 2.589 × 107 t, and the average sediment content was 1.3 kg/m3 . Table 1 shows
the characteristics of flow and sediment in different periods of Xunhua station. With
the completion and operation of the upstream cascade reservoirs, the amount of
flow and sediment in different periods has changed greatly, and the amount of sed-
iment has decreased even more. From 1968 to 1986, the amount of sediment was
4.294 × 107 t, and the amount of sediment from 1987 to 1999 was 1.913 × 107 t. In
2010, the annual sediment volume was only 4.45 × 107 t and the sediment content
is also decreasing. The average annual sediment content from 1968 to 1986 was
1.9 kg/m3 , the average annual sediment content from 1987 to 1999 was 1.0 kg/m3 ,
and the sediment content from 2000 to 2010 was only 0.27 kg/m3 . Hongqi Station
is 17 km away from Liujiaxia, and the station controls an area of 24973 km2 . From
1968 to 2010, the average flow of Hongqi Station was 134 m3 /s, the annual runoff
was 4.2 × 109 m3 , the annual sediment transport was 2.141 × 107 t, and the aver-
age sediment concentration was 5.1 kg/m3 (Table 1). After 1986, the amount of flow
and sediment is reduced, and the reduction of the amount of sediment is greater
than the amount of water. From 1968 to 1986, the annual average water volume
was 5 × 109 m3 , the sediment volume was 2.727 × 107 t, and the average sediment
content was 5.1 kg/m3 . From 1987 to 1999, the water volume decreased more than
the sediment volume. The annual average water volume was 3.6 × 109 m3 and the
Table 1
Water and sediment conditions at Xunhua and Hongqi stations
GUAGING STATION XUNHUA STATION HONGQI STATION
PARAMETER YEAR 1968 ∼ 1987 ∼ 2000 ∼ 1968 ∼ 1968 ∼ 1987 ∼ 2000 ∼ 1968 ∼
1986 1999 2010 2010 1986 1999 2010 2010
Water quality Flood 140 71 66 100 30 18 20 24
(108 m3 ) period 231 184 184 205 50 36 36 42
yearly
Sediment Flood 3319 1475 366 2006 2192 1593 768 1668
quality period 4294 1913 445 2589 2727 2261 943 2141
(104 t) yearly
Sediment Flood 2.4 2.08 0.65 2 6.7 8.8 4.6 7
content period 1.9 1 0.27 1.3 5.1 6.3 3.2 5.1
(kg/m3 ) yearly
810
Q.105 - R.49
sediment volume was 2.261 × 107 t. The average sediment content is 6.3 kg/m3 .
From 2000 to 2010, the decrease in sediment volume is greater than that of water
volume. The annual average water volume is 3.6 × 109 m3 , the sediment volume
is 2.261 × 107 t, and the average sediment content increases to 6.3 kg/m3 . The
inter-annual water volume and sediment volume of Hongqi Station have a large vari-
ation. The maximum runoff is 9.509 × 109 m3 (1967), the minimum annual runoff is
2.3 × 109 m3 (2002); the maximum annual sediment transport volume is 6.59 × 107 t
(1979), the smallest annual sediment transport volume is 1.37 × 107 t (2009).
3.2. CURRENT STATUE OF DEPOSITION IN THE RESERVOIR AREA
Since the Liujiaxia Reservoir was impounded in 1968–2000, a total of

1.533 × 109 m3 of sediment has been deposited in the entire reservoir area, with
a storage capacity loss of 26.9%, an effective storage capacity of 7.22 × 108 m3 ,
and a remaining effective storage capacity of 3.428 × 109 m3 . The effective stor-
age capacity loss is 17.4%. The reservoir was in operation until 2015, with a total
of 1.686 × 109 m3 of siltation in the entire reservoir area, with an average annual
siltation volume of 3.59 × 107 m3 and a loss of 29.6% of the storage capacity. Accord-
ing to the compilation data of the siltation volume measured by the reservoir, the
cumulative siltation volume of different reservoir sections since the reservoir was
put into operation is shown in Figure 2. The sedimentation of Tao River reservoir
area is serious. The original storage capacity of the Tao River Reservoir area was
1.33 × 108 m3 , and the remaining storage capacity of the reservoir area was only
3.354 × 107 m3 (Cross section: Tao 0-Tao 13), and the storage capacity loss was
74.8%. In 2003, the Tao 0 section’s talweg elevation was deposited to 1,698.9 m,
which exceeded the design dead water level, and the dead storage capacity was
completely silted. In 2013, Tao 0 section’s talweg elevation was deposited to
1,711.3 m, which was 17.3 m higher than the designed dead water level.
Partial enlarged view of Figure 2 shows that the development speed of sed-
imentation in the reservoir area changes from fast to slow with the increase of the
operating life. Before 1987, the development speed of sedimentation in the whole
reservoir area was relatively fast, with an average annual sedimentation amount
of 5.991 × 107 m3 . From 1987 to 2000, the sedimentation rate slowed down, with
an average annual sedimentation volume of 3.091 × 107 m3 . After 2000, the depth
of sedimentation significantly slowed down, with an average annual sedimentation
volume of 9.73 × 106 m3 . Since 1994, the development speed of siltation in different
reservoir sections has obviously slowed down. The silt deposition in the reservoir
area above section Huang 21 of the main stream of the Yellow River has tended to
balance, and the average siltation volume for many years is only 2.7 × 105 m3 . The
Tao River and the front section of the main stream dam are still in a weak state. The
average annual siltation volume is 4.38 × 105 m3 and 4.46 × 105 m3 respectively.
The average siltation volume in the Tao River reservoir area only accounts for 3.6%
of the incoming sediment of the Tao River, and the sedimentation in the front section
of the main stream dam accounts for 3.7% of the incoming sediment of the Tao River.
811
Q.105 - R.49
Fig. 2
Cumulative sedimentation change curve of Liujiaxia Reservoir.
3.3. CURRENT STATUS OF SEDIMENT RELEASE OF DENSITY

CURRENT IN RESERVOIR
Whenever there is a flood in the river course in the river basin during the
flood season, a large amount of fine sediment enters the reservoir with the flood,
which creates favorable conditions for the formation of density currents. The mea-
sured data of the reservoir show that when the average sediment content of the
reservoir reaches about 20 kg/m3 , the density flow can be generated and run to
the front of the dam. The velocity of the density flow at the diving point is gen-
erally between 0.6 m/s ∼ 1.0 m/s, and the maximum velocity can reach 1.65 m/s.
When running to the front of the dam, the velocity of the density flow is generally
between 0.4 m/s ∼ 0.6 m/s. Figure 3a, b show the changes in runoff and density flow
of Liujiaxia Reservoir. According to the data of the Liujiaxia Reservoir from 1974 to
2008, a total of 3.542 × 108 t of sediment was discharged. The sediment discharge
of the density current accounted for 46.3% of the sediment in the Hongqi Station
(Figure 3c), and the sediment discharge accounted for the sediment in the Hongqi
Station. 80.6% of the amount. By adopting the method of discharging sediment with
density flow, more sediment from Tao River can be discharged into the reservoir,
and the speed of sedimentation in the front section of the dam is slowed down.
From 1974 to 1986, when the Liujiaxia Reservoir was used alone, the average
annual sediment discharge was 5 times (Figure 3d), and the density flow discharged
1.189 × 108 t of sediment. From 1987 to 2000, due to the joint operation of the
812
Q.105 - R.49
Fig. 3
Current situation of density flow sediment discharge in Liujiaxia Reservoir:
(a) Incoming flow and sediment content of the reservoir; (b) the amount of
sediment inflow from Tao River and the amount of sediment discharged by density
flow; (c) sediment discharge by density flow The proportion of total sediment
discharged; (d) the number of times of sediment discharge per year by density flow.
Longyangxia and Liujiaxia reservoirs, the abandoned water of Liujiaxia Reservoir

was reduced by 53%, and the operating water level of the reservoir was raised.
During this period, the density flow and sediment discharge were strengthened.
The average annual sediment discharge was 9 times. The amount of sediment
discharged was 1.659 × 108 t, and the amount of sediment discharged by the density
current accounted for 56.3% of the amount of sediment in the Hongqi Station. From
2001 to 2008, due to the significant reduction in the inflow of water and sediment
from the Tao River, the chance of opening the discharge channel during the density
current period was reduced. The average annual sediment discharge was 7 times,
and the density current discharge amount was 6.11 × 107 t. It accounts for 72.6% of
Hongqi Station’s sediment.
Due to the strengthening of the density flow monitoring and sediment dis-
charge of the reservoir, from 1987 to 2000, the average annual density flow was 9
times, and the density flow accounted for 58.3% of the total amount of sediment.
From 2001 to 2008, the annual average density flow was 7 times, and the average
density flow accounted for 81.5% of the sand out of the reservoir and 72.6% of the
sand in the Tao River, which is helpful to alleviate the reservoir sedimentation. After
2009, due to the significant decrease in the amount of sediment in the Tao River,
813
Q.105 - R.49
the number of times the density flow was formed decreased. The average annual
sediment discharge was only 2 times, and the amount of sediment discharged by
the density flow accounted for only 35.8% of the Tao River’s inflow. Due to the con-
tinuous development of sedimentation in the front section of the dam, the density
flow in front of the dam flows laterally into the discharge channel, the area of the
water passing below 1,680 m is seriously shrunk, the density flow channel is not
smooth, and the discharge channel density flow reduces the sediment discharge
effect.
4. IMPROVEMENT MANAGEMENT OF RESERVOIR SEDIMENT
4.1. THE ONE-DIMENSIONAL FLOW AND SEDIMENT MATHEMATICAL MODEL

VERIFICATION
In order to study the effect of adding desilting tunnels and the coupling of
density flow, this paper carried out a one-dimensional constant non-uniform unbal-
anced sediment transport model calculation. Mainly calculate the erosion and
siltation of each reservoir section of the reservoir, and use the measured sedi-
ment data to verify. This paper selects the model verification calculation period
from 1991 to 1996, and the calculation scope includes the reservoir area of the
main stream of the Yellow River (Cross section: Huang 0 ∼ Huang 26), Tao River
reservoir area (Cross section: Tao 0-Tao 13), Daxia River reservoir area (Cross
section: Da 1-Da 7). The calculated initial bed elevation is the data of the mea-
sured cross-section after the flood in 1990, and the boundary calculated conditions
are the discharge, sediment content and grain size distribution of the Xunhua Sta-
tion in Yellow River, Hongqi Station in Tao River, and Zheqiao Station in Daxia
River.
The reservoir area of Tao River, Daxia River and the front dam section of the
main stream verify that the calculated scour and deposition volume is consistent with
the actual measurement trend and the amount is similar. The main difference is the
middle reservoir section of the main stream of the Yellow River (Cross section: Huang
9-Huang 21). The value is 7.93 × 107 m3 smaller than the measured value. The main
reason is that the amount of sediment entering the reservoir in the Xunhua station to
main stream section of Liujiaxia Dam site in the verification calculation (analyzed by
the actual measurement data, the average annual amount of sediment entering the
reservoir in this section is about 1.69 × 107 t) is not included. After considering this
part of the amount of sediment, it is relatively close to the measured data of the reser-
voir. Judging from the longitudinal profile of the sedimentation of the main stream
of the Yellow River and the Tao River reservoir area (Figure 4a, b), the evolution
trend of erosion and deposition is also consistent. Therefore, this one-dimensional
814
Q.105 - R.49
Fig. 4
Schematic diagram for verification of the talweg of Liujiaxia reservoir area:
(a) Yellow River mainstream reservoir area; (b) Tao River reservoir area.
mathematical model can better simulate the changes in sediment erosion and depo-
sition of Liujiaxia Reservoir, and can be used for comparison and analysis of design
schemes.
4.2. BOUNDARY CONDITIONS FOR MODEL CALCULATION
Calculation series: using flow and sediment data from 1982 to 1996 as the
representative series of flow and sediment entering the Liujiaxia Reservoir, the cyclic
use calculation and analysis of the reservoir erosion and deposition changes within
60 years of operation of the reservoir. Flow and sediment data entering the reservoir:
The Xunhua Station, Hongqi Station, and Zheqiao Station use selected series of
actual measured daily average flow and sediment content, and the gradation of
suspended sediment particles in the reservoir adopts multi-year average (Figure 5).
Calculate the initial bed elevation: use the measured section data after the 2002 flood
as the initial bed elevation. Water level in front of the dam: Based on the monthly
average outflow and water level in front of the dam of the Longyangxia and Liujiaxia
reservoirs jointly dispatched at the end of the month, according to the principle of
daily inflow and water balance of Liujiaxia Reservoir and considering the need for
density flow and sediment discharge, and further calculate the daily average water
level in front of the dam and outflow from the reservoir month by month. The change
process of the water level in front of the dam and the outflow flow is shown in Figure 6.
4.3. MODEL PREDICTION AND ANALYSIS
This paper simulates the following two schemes: (1) the effect of sediment
removal by existing sediment removal facilities (2) the effect of sediment removal
after the addition of the desilting tunnel of Tao River and the joint operation of the
reservoir. Under the conditions of existing flow and sediment facility (without desilting
815
Q.105 - R.49
Fig. 5
Gradation of suspended load in Liujiaxia Reservoir
Fig. 6
The water level in front of the dam of Liujiaxia Reservoir (a–c) and the change
process of outflow flow (d–f)
tunnels), the calculation results of accumulated sediment erosion and sedimentation

in different periods (years) and different reservoir sections are shown in Figure 7.
The results show that the long-series average sediment discharge effect of Liujiaxia
Reservoir is similar regardless of whether the additional desilting tunnel is built or
816
Q.105 - R.49
Fig. 7
Comparison diagram of accumulated siltation volume of different schemes of
Liujiaxia Reservoir
not. Based on the statistical analysis of the amount of sediment discharged from the
reservoir, the annual average sediment discharge from the reservoir is 2.56 × 107 t.
Among them, 1.825 × 107 t of sediment are discharged through the desilting tunnel,
accounting for 71.3% of the total annual average sediment discharge. The existing
sediment discharge facilities have an average annual discharge of 2.527 × 107 t,
and the additional desilting tunnels increase the sediment discharge about 1.3%.
The addition desilting tunnels will not cause a significant increase in the amount of
sediment in the downstream river, but the path of sediment discharge is different.
According to the statistics of the measured density flow sediment discharge data
from 1987 to 2000, the density flow sediment discharge accounted for 56.3% of
the Tao River’s annual sediment discharge, and the addition desilting tunnel is the
71.37%. It can be concluded that the effect of sediment removal after the addition
desilting tunnels will be significantly improved compared with the original sediment
removal facilities.
After the addition of the Tao River estuary desilting tunnel, the reservoir has a
more obvious effect of reducing siltation, which is mainly manifested in the vicinity of
the Tao River mouth and in front of the main stream dam. The Tao River Reservoir
has reduced siltation by 9.5 × 105 m3 in 30 years and 1.99 × 106 m3 in 60 years.
The front section of the main stream dam (Huang 9 to the front of the dam) reduced
siltation by 1.616 × 107 m3 in 30 years and 2.304 × 107 m3 in 60 years. The sediment
reduction ratio of the Tao River and the front section of the main stream dam is 59%.
This is because the Tao River desilting tunnel directly intercepts and discharges the
density flow of the Tao River, which greatly reduces the sediment before reaching
the dam, and the siltation before the dam will also be significantly reduced.
817
Q.105 - R.49
Fig. 8
Comparison diagram of the variation process of the longitudinal profile
(deep stratum) of siltation in the main stream area of Liujiaxia Reservoir:
(a) Yellow River reservoir area; (b) Taohe reservoir area
The sand discharge of the Tao River estuary desilting tunnel accounted for
about 71.3% of the total sediment discharge of the reservoir, and the sediment
discharged from the discharge structure in front of the dam and the outlet of the
power station only accounted for the total amount of the reservoir discharge. 28.7%
after adding the desilting tunnel. This has greatly reduced the amount of sediment
passing by the turbine, and the particles of the sediment have become significantly
finer. The median size of the sediment reaching the front of the dam has been
reduced from 0.05mm to 0.002 mm. It can be shows that the addition of desilting
tunnels at the Tao River mouth can greatly alleviate the wear and tear on the turbine.
At the same time, it has a significant effect on reducing and controlling the elevation
of the sand sill. The sedimentation elevation (talweg) of the sand sill section with
the desilting tunnel does not change significantly, and it is basically maintained at
about 1698 m. However, the elevation of the sand sill of the no desilting tunnel has
increased significantly with the increase of the service life, and the siltation height is
4.78 m after 15 years and 7.1 m after 60 years. The comparison of siltation changes
in the longitudinal section of the siltation (talweg) section of the sand sill is shown in
Figure 8.
5. CONCLUSIONS
The current situation of siltation in Liujiaxia Reservoir and its harm were stud-
ied, and the “bypass sediment transport technology” was innovatively proposed. A
desilting tunnel is added to the path of sediment transport, and the sediment that
originally moved to the front of the dam is directly introduced into the desilting tunnel
and discharged from the reservoir, thereby greatly reducing the amount of siltation
in front of the dam. These measures have better adjusted the water-sediment rela-
tionship in front of the dam, so that the water-sediment relationship in the reservoir
has been gradually improved, and the problem of sedimentation in front of the dam
818
Q.105 - R.49
has been solved. Analyze the sediment discharge effects of existing reservoirs and
the main causes of sediment hazards, and firmly grasp the relationship between
water and sand, which is the key to the control of sediment problems, and study
ways and solutions to solve the sediment problem. Starting from coordinating the
water-sediment relationship, it was proposed to build an additional desilting tunnel
project at the bottom of the reservoir (the bottom of the reservoir bank) along the
path of the density flow of the reservoir, which was verified by a water-sediment
model with good results.
ACKNOWLEDGES
This study was supported by the National Key R&D Program of China
(2018YFC0407400) and the National Natural Science Foundation of China
(51539004, 51679104).
REFERENCES
[1] WANG G.Q., WU B.S., WANG Z.Y. Sedimentation problems and management
strategies of Sanmenxia Reservoir, Yellow River, China. Water Resources
Research, 2005, 41(9):477–487.
[2] KONDOLF G.M., GAO Y., ANNANDALE G.W., et al. Sustainable sedi-
ment management in reservoirs and regulated rivers: Experiences from five
continents, Earth’s Future, 2014, 2, 256–280.
[3] Fu K., He D., Lu X.X. Sedimentation in the Manwan reservoir in the Upper
Mekong and its downstream impacts. Quaternary International, 2008, 186(1),
91–99.
[4] SONG X., ZHUANG Y., WANG X., et al. Combined Effect of Danjiangkou
Reservoir and Cascade Reservoirs on Hydrologic Regime Downstream.
Journal of Hydrologic Engineering, 2018, 23(6).
[5] PALMERI A., SHAH F., ANNANDALE G.W., et al. Reservoir Conservation,
The RESCON Approach, World Bank, Washington, D. C. 2003, 1.
[6] XIE J.M., HOU B.S. Review of reservoir sedimentation management. Journal
of Sediment Research, 2013, 3, 71–80.
[7] WU B.S., WANG G.Q., XIA J.Q. Case study: Delayed sedimentation response
to inflow and operations at Sanmenxia Dam. Journal of Hydraulic Engineering,
2007, 133(5), 482–494.
819
Q.105 - R.49
[8] MORRIS G.L. Classification of Management Alternatives to Combat Reservoir

Sedimentation. Water, 2020, 12, 861.
[9] SARDAR B., SINGH A.K., RAGHUWANSHI N. S., et al. Hydrological modeling
to identify and manage critical erosion prone areas for improving reservoir life:
a case study of barakar basin. Journal of Hydrologic Engineering, 2014, 19(1),
196–204.
[10] VOROSMARTY C.J., MEYBECK, M., FEKETE, B., et al. Anthropogenic sed-
iment retention: major global-scale impact from the population of registered
impoun. Global and Planetary Change, 2003, 39, 169–190.
[11] MORRIS, G.L., Fan J.H. Reservoir Sedimentation Handbook, McGraw-Hill,

New York.1997.
[12] WU B.S., WANG Z.Y., LI C.Z. Yellow river basin management and current
issues. Journal of Geographical ences, 2004, 14(1), 29–37.
820
Q.105 - R.50
-------
-------
CONSÉQUENCES (ÉCONOMIQUES) DES RUPTURES DES PETITS

BARRAGES AU BURKINA FASO: ÉTUDES DE CAS (∗)
A. NOMBRÉ, F. MILLOGO & M. KABORÉ

Burkina Faso Committee on Dams (BUCOLD), Ouagadougou
BURKINA FASO
RÉSUMÉ
Le Burkina Faso Dispose d’un parc de barrage de l’ordre de 1001 Barrage dont
une trentaine de Grands Barrages selon les définitions de la CIGB. Pour des raisons
diverses chaque année connaissent quelques ruptures et/ou incidents sur les petits
barrages essentiellement. Durant les mauvaises années liées à la pluviométrie ces
ruptures peuvent s’approcher de la dizaine.
Il s’agit dans ce rapport au Congres de faire le point sur ces ruptures et inci-
dents aussi bien au plan technique pour en connaitre les causes et les mécanismes
mais surtout évaluer les conséquences de ces ruptures en vue d’en tirer les leçons
à partager avec la communauté des acteurs intéressés par ces ouvrages combien
critiques pour les populations et l’économie nationale des approches pour réduire
ces ruptures et incidents basées sur le retour d’expérience seront mis en exergue
dans le rapport.
∗ Economic consequences of small dam failures in Burkina Faso: case studies
DOI: 10.1201/9781003211808-96 821

Q.105 - R.50
SUMMARY
Burkina Faso has about 1001 dams, of which about thirty are large dams as
defined by the ICOLD. For various reasons, each year there are some failures and/or
incidents on the small dams mainly. During bad years linked to rainfall, these failures
can approach ten.
This report to the Congress aims at taking stock of these failures and incidents
both technically to know the causes and mechanisms but above all to evaluate the
consequences of these failures to draw lessons to be shared with the community
of actors interested in these structures, which are so critical for the populations and
the national economy. Approaches to reduce these failures and incidents based on
the feedback will be highlighted in the report.
1. INTRODUCTION
Des ruptures de petits barrages, notamment en terre, sont constatées chaque

année au Burkina Faso. Leur compilation permet de disposer d’une base de données
importante sur leurs causes et leurs conséquences. Dans le présent rapport porte
sur les conséquences de tous ordres des ruptures de petits barrages: pertes de
vies humaines, pertes économiques diverses, impact environnemental.
2. BRÈVE PRÉSENTATION DU BURKINA FASO
Le Burkina Faso est un pays en développement situé au cœur de l’Afrique de

l’Ouest, dans la boucle du fleuve Niger. Il s’agit d’un pays enclavé d’une superficie
de 274 000 km2 avec une population estimée à environ 20 millions d’habitants.
Le climat est caractérisé par une longue saison sèche et une saison plu-
vieuse qui dure de deux à trois mois. La pluviométrie annuelle varie de 500 mm à
900 mm avec un écoulement moyen de 30 mm à 40 mm. L’évaporation annuelle est
supérieure à la pluviométrie annuelle et peut atteindre 2000 mm. Malgré l’absence
de montagnes, le Burkina Faso dispose d’un large réseau hydrographique notam-
ment dans sa partie australe. Trois cours d’eau sont permanents (Mouhoun, Comoé
et Pendjari) et l’ensemble des cours d’eau sont regroupés en trois bassins nationaux
(Niger, Volta et Comoé). Le Burkina Faso compte plus de mille petits barrages avec
une capacité de stockage de 800 millions de m3 .
822
Q.105 - R.50
Fig. 1
Carte de situation du Burkina Faso
La construction des petits barrages a débuté au Burkina Faso dans les années
1920. Elle a été accélérée dans les années 1970–1980 suite aux sévères sècher-
esses dans la région du Sahel. Dans un contexte de pluviométrie insuffisante les
petits barrages ont aidé à mieux gérer les ressources en eau pour les communautés
locales. Depuis cette période, des centaines de petits barrages, dont la plupart sont
en service, ont été construits.
3. PETITS BARRAGES EN TERRE ET LEURS RÔLES
La construction d’infrastructures modernes pour l’habitat, le transport, l’eau,

l’électricité … est nécessaire pour le développement économique d’un pays et pour
un certain bien-être des populations. Parmi ces infrastructures, les barrages jouent
un rôle capital. A ce jour, le Burkina Faso compte une vingtaine de grands barrages
(selon les critères de la CIGB) et plus de mille petits barrages. Les retenues de ces
petits barrages sont utilisées pour l’alimentation en eau domestique de petites et
même de grandes localités. Les retenues de petits barrages sont aussi très utilisées
pour l’agriculture irriguée en zone rurale. Dans les pays sahéliens ou semi-arides
comme le Burkina Faso, les petits barrages permettent de retenir l’eau pour les
usages qui prolongent la courte saison pluvieuse.
823
Q.105 - R.50
Au Burkina Faso, les petits barrages sont aussi utilisés dans l’industrie minière.
Le pays connaît un boom minier et de petits barrages sont construits pour satisfaire
les besoins en eau des mines, aurifères surtout.
Les petits barrages permettent aussi l’alimentation en eau du bétail et de la

vie sauvage (flore et faune).
Enfin, l’eau retenue par les petits barrages participe à la préservation des
zones humides.
Une étude réalisée par le CNBB avec l’appui de l’administration en charge

des ressources en eau a conduit à une classification nationale des barrages. Cette
classification faite à partir de la taille des ouvrages répartit les barrages au Burkina
Faso en trois groupes comme suit:
– Petits barrages: barrages dont la hauteur est comprise entre 2 et 5 mètres

avec un volume inférieur à 5 millions de m3 ;
– Barrages moyens: barrages dont la hauteur est comprise entre 5 et 10 mètres
avec un volume inférieur à 5 millions de m3
– Grands barrages: barrages dont la hauteur est supérieure à 10 mètres ou avec
un volume supérieur à 5 millions de m3
– En relation avec la classification nationale, les barrages concernés par notre
étude sont les petits barrages et les barrages moyens.
4. ACCIDENTS AND INCIDENTS DE PETITS BARRAGES
Durant chaque saison humide, on constate quelques ruptures de petits bar-

rages en terre. Pendant les saisons très humides le nombre de ruptures de de ce
type de barrages peut atteindre dix.
Selon les statistiques de l’année 2015, les ruptures des petits barrages en
terre sont dues à deux causes principales: la submersion et l’érosion interne. Les
ruptures sont aussi liées aux études insuffisantes, aux défauts de construction ou à
une maintenance insuffisante ou quelque fois inexistante.
Plus spécifiquement, sur la base de l’analyse et de l’évaluation de nombreuses

défaillances de petits barrages en terre au Burkina Faso, la principale cause des
problèmes et des défauts à l’origine des accidents et incidents liés aux barrages
peut être résumée comme suit:
– Manque de données hydrologiques et climatiques et méthodes d’évaluation

des crues non actualisées
– Absence de critères de conception pour la conception de l’évacuateur de crues
des petits barrages
824
Q.105 - R.50
– Faiblesse des études géologiques et géotechniques et absence de spécifica-

tions appropriées
– Le mauvais compactage du remblais dû au fait que la teneur en eau n’est pas
toujours prise en compte comme critère pertinent
– Faiblesse de la supervision des travaux de construction et du contrôle de la
qualité
– Manque de filtres et de drains appropriés
– Critères de moindre coût conduisant à la sélection de contractants et de
bureaux d’études non qualifiés
– Manque d’entretien.
825
Q.105 - R.50
5. CONSÉQUENCES DES RUPTURES DES PETITS BARRAGES
Les principales pertes potentielles liées aux ruptures des petits et moyens
barrages en terre au Burkina Faso sont les pertes en vies humaines, les pertes
économiques et les impacts sur l’environnement.
Les risques de pertes en vies humaines lors des ruptures des barrages
sont faibles car les crues qu’ils provoquent sont de relative faible amplitude et il
y a rarement des habitations permanentes à l’aval des ouvrages; les statistiques
disponibles sur les ruptures des barrages ne signalent pas de pertes en vies
humaines.
Les pertes liées aux ruptures des barrages sont nombreuses et variées:
• Sur les populations: noyade, ensevelissement, blessure, isolées, déplace-

ment;
• Sur les biens:
– Pertes de l’ouvrage principal et des annexes: la rupture endommage
l’ouvrage et/ou les ouvrages annexes (EC ou PI) et dont les coûts de
réparations peuvent être importants;
– Pertes des aménagements agricoles: très souvent les retenues d’eau
des barrages alimentent les aménagements agricoles sis presque tou-
jours à l’aval des barrages de telle sorte que la rupture du barrage
entraine l’endommagement des aménagements;
– Les pertes de productions agricoles, animales et halieutiques;
• Sur l’environnement: destruction de la faune et de la flore, pollution diverses,
dépôt de déchets, boues, débris.
6. ETUDES DE CAS
6.1. BARRAGE DE BAMPELA
6.1.1. Présentation
Le barrage de Bampela est situé dans la Région du Centre-Est du Burkina,

dans la Commune de Garango. Il s’agit d’un barrage en terre homogène d’une
longueur de 762 m, une largeur en crête de 4 m et une hauteur maximale sur le
terrain de 5.80m. le barrage a un évacuateur de crue en forme de déversoir de
type poids de 120 m de long. Le barrage crée une retenue de 4.5 millions de m3 . Le
barrage est destiné à l’irrigation de centaine d’hectares soit une production annuelle
de milliers de tonnes de céréales, légumes et légumineuses.
826
Q.105 - R.50
Le barrage a coûté environs un (01) milliards de francs CFA, soit 1.9 millions
de Dollars US.
6.1.2. Rupture
La rupture du barrage de Bampela est intervenue le 16 juin 2007 par suite

d’une très forte pluie et d’une mise en eau très rapide de la retenue. La rupture
est due à un problème de mise en œuvre des matériaux et/ou de la qualité des
matériaux de la zone d’emprunt utilisée.
827
Q.105 - R.50
Le barrage est rompu en plusieurs trente-deux (32) points aussi bien dans
le corps de digue et au contact avec les bajoyers. La retenue du barrage a été
entièrement vidée.
6.1.3. Arbre logique de défaillance
L’analyse de toutes les causes et de tous les scénarios ayant conduit à la

rupture du barrage de Bampela est présentée sur la figure suivante qui constitue
l’arbre de défaillance:
6.1.4. Conséquences
Les conséquences de la rupture du barrage de Bampela sont:
– Pas de perte en vie humaine n’a été notifié consécutivement à la rupture du

barrage;
– L’aménagement hydroagricole prévu n’avait pas encore été réalisé au moment
de la rupture, mais les champs de cultures en aval ont été inondés. Des
animaux d’élevage ont également être emportés par les eaux;
– Le radier et la piste à l’aval du barrage est endommagé. Les dalles en béton
ont été emportées, les aciers de ferraillage sont mis à nu et les remblais
sont affouillés. L’accès d’une rive à l’autre est pratiquement impossible. Le
déplacement des populations de Bampela, vers la ville de Garango et ses
environnants, relève d’un parcours de combattant durant la saison des hautes
eaux. Cette situation a des impacts socio-économiques très importants;
828
Q.105 - R.50
– La rupture du barrage signifie d’abord la perte des ressources en eau si vitales

pour la population de la zone, pour la satisfaction de leurs besoins en eau,
l’agriculture et l’élevage
Plus de treize (13) ans après sa rupture, le barrage de Bampela n’a pas
encore été reconstruit. Il s’ensuit sur le plan économique une perte de revenus
correspondant aux produits d’exploitation. Avec un volume d’eau stocké de 4.5 Hm3 ,
on peut irriguer 150 ha avec une production de 600 T de riz et 3000T de produits
maraîchers. En admettant que la dotation moyenne de céréales de 200Kg/hbt/an
la production de riz assurerait la sécurité alimentaire d’environ 3000 personnes.
L’écoulement des produits maraîchers générerait 500 000 $US.
D’autres pertes, même si elles ne sont pas comptabilités en terme monétaire,

peuvent être signalées: la pêche, l’abreuvement des animaux domestiques.
En plus de la perte de l’ouvrage lui-même qui représente un investissement

élevé si on ajoute les difficultés de mobilisation des fonds et le temps que va prendre
une reconstruction de l’ouvrage les dommages économiques seront considérables.
Les coûts de reconstruction du barrage ou de construction d’un nouveau

barrage ont été estimés à 1.6 milliards de FCFA (3 millions US$).
6.2. BARRAGE DE GAZANDOURÉ
6.2.1. Présentation
Le barrage de Gazandouré est situé dans la Région du Centre-Est du Burkina,

dans la Commune rurale de Sangha. Il s’agit d’un barrage en terre homogène d’une
longueur de 750 m, une largeur en crête de 3,50 m et une hauteur maximale sur le
terrain de 6,75 m environ. Le barrage a un évacuateur de crue en forme de déversoir
de type poids de 125 m de long. Le barrage crée une retenue de 2 544 000 m3 . Le
barrage est destiné à l’irrigation de 39 ha de riz et 13 ha de maïs en saison humide
ainsi que 2 à 5 ha de culture maraichère en saison sèche. La production annuelle
est ainsi estimée:
– Riz: 156 tonnes

– Maïs: 39 tonnes
– Produits maraichers: 60 à 150 tonnes (tomates)
Le barrage a coûté 2 milliards de francs CFA, soit environ 4 millions de $US y

compris le coût du périmètre irrigué.
La réception provisoire eu barrage et du périmètre agricole est intervenue le

31 mai 2007.
829
Q.105 - R.50
La photo ci-dessous montre le déversoir du barrage en fonctionnement avant

l’accident.
6.2.2. Rupture
La rupture du barrage de Gazandouré est intervenue dans la nuit du 22 juillet

2007, environ un mois après la mise en eau de la retenue suite à une pluie qui
830
Q.105 - R.50
a engendré une crue supérieure à la crue décennale de projet. Les problèmes

constatés ont consisté en une rupture de la fondation par érosion interne et la rupture
du remblai de jonction entre la digue et le mur bajoyer du déversoir. On a aussi relevé
la rupture du plot N◦ 1 RG du déversoir, les désordres de la protection aval de la
digue.
Selon les investigations menées dans le cadre des études de réhabilitation du

barrage, les désordres constatées sont liés aux causes suivantes:
– L’insuffisance des investigations géologiques et géotechniques;

– Le sous dimensionnement de la fondation;
– L’insuffisance de la conception du barrage et les défauts de construction.
6.2.3. Arbre logique de défaillance
L’analyse de toutes les causes et de tous les scénarios ayant conduit à la

rupture du barrage de Gazandouré est présentée sur les deux figures suivantes qui
concernent le renard sous le seuil et le renard contre les bajoyers.
831
Q.105 - R.50
6.2.4. Conséquences
Les conséquences de la rupture du barrage de Gazandouré sont:
– La perte partielle du barrage;

– L’inondation du périmètre qui avait déjà été inondé par la crue naturelle la veille
de la rupture d’où sa destruction partielle;
– La perte de la réserve d’eau mobilisée par le barrage;
– La perte des productions prévues par le périmètre agricole jusqu’à la recon-
struction.
L’étude de réhabilitation du barrage de Gazandouré a été réalisée en 2008

tandis que les travaux se sont déroulés en 2010 le coût des études et des travaux
de réhabilitation est estimé à 1.85 milliards de FCFA ou 3.5 millions US$.
7. CONCLUSION
Dans un pays sahélien donc sec comme le Burkina Faso les petits barrages
jouent des rôles importants et sont très demandés par les communautés locales.
832
Q.105 - R.50
Sur la base des statistiques et renseignements disponibles, aucune perte en vie

humaine n’est notée comme conséquence d’une rupture de petits barrages.
Par contre les pertes économiques sont très importantes. Les coûts des
ouvrages initiaux et les travaux de reconstruction sont élevés dans le contexte
économique du pays (de nombreux ouvrages sont financés par la coopération
bilatérale ou autre). Lorsque la rupture intervient, le délai de réhabilitation de
l’ouvrage peut être excessivement et de façon inexpliquée long comme c’est le cas
du barrage de Bampela. Les pertes de revenus liées aux ruptures de barrages sont
très importantes dans les zones rurales où sont implantés les barrages. Cette situ-
ation a des conséquences sociales: scolarisation des enfants, immigration, famine,
perte de bétail, investissement.
REFERENCES
[1] Ifec & Gertec, Etudes diagnostiques et élaboration d’un projet de conforte-
ment. 2009.
[2] Ifec, Rapport de diagnostic du barrage BAMPELA, 2011.
833
Q.105 - R.51
-------
-------
CHANGE IN METHODOLOGY IN CONSTRUCTION OF PLASTIC CONCRETE

CUT-OFF WALL ADAPTED IN KISHANGANGA HE PROJECT (∗)
A.K. SINGH
Chairman & Managing Director, NHPC LIMITED, India
Sunil J. GANVIR
Senior Manager (Geology), NHPC LIMITED, India
INDIA
SUMMARY
In the Himalaya, rivers have carved out deeply incised valleys where construc-
tion of dams is a challenge as a number of sites have thick overburden in the river
bed. At Kishanganga HEP in J&K, a 37m high concrete face rock fill dam with 33m
deep cut-off wall was planned and efficaciously executed. Construction of the project
began in 2009 and all the three units (3 × 110 MW) were successfully commissioned
in March 2018.
One of the key aspects of the project was the construction of a 134m long, 1m
wide plastic concrete cut-off wall (diaphragm wall) below the dam upto the maximum
depth of 33m in overburden. Based upon the previous experience gained from dif-
ferent projects, significant amount of desk studies were taken up during the design
stage to develop a scope of work for the possibility of construction of cut-off wall
either through mechanical means like trench cutter or drill-blast method.
∗ Changement de méthode de construction d’un écran d’étanchéité en béton plastique
adapté au project hydroélectrique de Kishanganga
834 DOI: 10.1201/9781003211808-97

Q.105 - R.51
Unlike the prevailing mechanical method like trench cutters, the cut-off wall in
Kishanganga HEP was executed by drill and blast & hydraulic grab picking method
due to the presence of high strength boulders and other reasons. Highly specialized
methodologies as well as equipment deployed have also been discussed.
RÉSUMÉ
Dans l’Himalaya, les rivières ont creusé des vallées profondément encaissées
où la construction de barrages est un défi car, dans de nombreux sites, une grande
épaisseur de mort-terrain peut être trouvée dans le lit de la rivière. À la centrale
hydroélectrique de Kishanganga au Jammu-et-Cachemire, un barrage en enroche-
ment de 37 m de haut doté d’un écran étanche de 33 m de profondeur a été conçu
et réalisé de manière efficace. La construction du projet a commencé en 2009 et
les trois unités (3 x 110 MW) ont été mises en service avec succès en mars 2018.
L’un des aspects clés du projet était la construction sous le barrage d’un écran
étanche en béton plastique de 134 m de long, d’1 m de large (paroi moulée)
et jusqu’à la profondeur maximale de 33 m en mort-terrain. Tenant compte de
l’expérience antérieure acquise au cours de différents projets, de nombreuses
études documentaires ont été réalisées au cours de la phase de conception pour
définir un cahier des charges des travaux qui prendrait en compte la possibilité de
construire un écran étanche à l’aide de procédés mécaniques tels que la fraise de
havage ou le forage et le dynamitage.
Contrairement au procédé mécanique courant comme la fraise de havage,

l’écran étanche de la centrale hydroélectrique de Kishanganga a été réalisé par
forage et dynamitage et à l’aide de grappins hydrauliques pour déblayer en raison,
entre autres, de la présence de rochers très durs. Des méthodologies hautement
spécialisées ainsi que le matériel utilisé sont également discutés.
1. INTRODUCTION
The Kishanganga Hydroelectric project (KHEP) of 330 MW installed capacity

has been constructed on the river Kishanganga (a tributary of river Jhelum) and is
located in Bandipore District of Jammu & Kashmir. The power house site is acces-
sible by 75 km long all weather double lane road from Srinagar whilst the dam site
is further north and can be approached by 84 km long road which is single lane for
considerable length and crosses hilly terrain.
The project comprises of 37m high Concrete Face Rock fill Dam; 134m long,
33m deep, cut-off wall at upstream toe of dam; One number of 23km long, 6.24/5.2m
diameter, horse shoe/circular shaped HRT and 905m long, 5m dia, horse shoe
835
Q.105 - R.51
Fig. 1
Schematic layout plan of the project.
Schéma du projet.
shaped Tail Race Tunnel. The underground power house accommodated 3 units of
110MW capacity each designed to operate under the net rated head of 646m and
generating 1713 MU in a 90% dependable year. Construction on the project began
in 2009 and all the three units (3 × 110 MW) were commissioned and synchronized
with the electricity grid during March 2018.
The underground power house with installed capacity of 330 MW houses 3

units of 110MW capacity each designed to operate under the net rated head of 646m
and generating 1713 MU in a 90% dependable year.
Fig. 2
Schematic layout plan of Dam Area.
Plan des environs du barrage.
836
Q.105 - R.51
2. SITE GEOLOGY
Kishanganga project is located in Gurez valley in west central Kashmir

Himalaya and has a long history. Conceived in 1960s the project was investi-
gated by State as well as Central power utilities. However the active construction
was completed in 9 years beginning from 2009. This is a worthy achievement
considering the 84 km long mostly single lane access road from Bandipore to
Gurez going over Razdan pass (3500 M) which is closed in winter season and
other difficulties. Geologically the project is situated in Kashmir nappe within the
Basaltic/andesitic Panjal Volcanics of Permo-Triassic age and Madhumati Group
belonging to Cambro-Silurian age.
The dam is located in picturesque Gurez within a U-shaped valley having

steep abutments where Panjal Volcanics are manifesting. At this location river width
is about 90m with nearly 30m wide main channel being confined to the right bank
due to exposures of slope wash & low level fluviatile terrace deposit. The left bank
of river channel comprises of river borne material at river bed level. The overburden
materials consist of boulders, pebbles and cobbles of Andesite, granite, dolomatic
limestone and gneiss embedded in a sandy silty matrix. The depth of the overburden
along the dam axis varies from 1.9m to 33m with hole collars being at around El.
2364.00 m.
Both the abutments on either bank above the river bed are exposed with
bedrock of greenish in colour, fine grained and massive Andesite flows of Panjal
Volcanic rock. Laboratory tests on core samples as carried out indicate that the
uniaxial compressive strength (UCS) of the rock ranges from 95 to 265 MPa. Thus
the strength of rock was variable and quite hard. From the 15 number of drill holes
and other investigations carried out in dam area, the Andesite of Panjal Volcanics
rock were found to be deeply weathered (up to 7 to 9m depth) and permeability
of bedrock ranged from 12 to 50 Lugeons. The rock quality from the deformability
considerations however was considered from good to very good.
3. NEED FOR POSITIVE CUT-OFF WALL
The key part of the whole project was the construction of a 134m long, about
1m wide plastic concrete cut-off wall (or diaphragm wall) without reinforcement
underneath the main dam at upstream toe area to a maximum depth of 33m in
overburden consisting of bouldery zone along with 1m socket into bedrock through
drill-blast & grab picking method which was an innovative technology.
The objective of providing cut-off wall at upstream heel of the Concrete Faced
Rock fill Dam was to restrict seepage of water within the overburden and rock
837
Q.105 - R.51
formations beneath dam and to enhance long term stability/service life of the dam
against internal erosion and uplift pressure.
Unlike some of the other conventional cut-offs where the path of seepage is
diverted below the dam, in this case the placement of positive cut-off wall into socket
of bed rock sealed the water seepage below the dam body as illustrated below.
Fig. 3
Cross Section of CFRD.
Coupe transversale du barrage.
As already mentioned above, the thickness of the overburden along the cut-off
wall in the river bed section varies from 1.9 to 33.0m consisting of riverine materials
like boulders, pebbles and cobbles of Andesite, granite, dolomitic limestone and
gneiss embedded in a sandy silty matrix which are underlain by bedrock of massive
to moderately jointed, compact, hard Panjal volcanics.
During the construction of cut-off wall, riverine deposits have been excavated
and the resulting void backfilled with a barrier material of well defined properties.
Unlike the prevailing mechanical method like trench cutters, the constructions of cut-
off wall in Kishanganga HEP has been executed by drill and blast method (DBM) &
hydraulic grab picking method due to the presence of high strength boulders in an
relatively unknown quantity and size.
Based upon the previous experience gained from different projects, significant
amount of desk studies were taken up during the design stage to develop a scope
of work for the possibility of construction of cut-off wall either through mechanical
means like trench cutter or drill-blast method. With the deployment of Hydromill type
Trench cutter in geological conditions like hard boulder range more than 200MPa,
the excavation output may have dropped rapidly and wear & tear on the cutters may
838
Q.105 - R.51
have becomes high. As such, alternative excavation processes as combination of

drill-blast method, the mechanical grab system were adopted in Kishanganga HEP
and was found to be successful.
The concrete for the cut-off wall was required to be homogeneous, which can
be readily placed and uniformly workable. It was to have following main properties:
• Maximum grain size 30 mm

• Unconfined compressive strength (28 days) > 1.5 Mpa
• Confined compressive strength > 2.5 Mpa
• Strain at failure (unconfined test) > 5%
• Permeability < 10−8 m/s
In Kishanganga the plastic concrete cut-off wall was designed to be 1m

thick with maximum depth of 33m and average depth of 30m and surface area
of 3600 sq m.
4. GEOLOGY AROUND POSITIVE CUT-OFF WALL
The depth of the overburden along the cut-off wall in the river bed section varies
from 1.9m to 33.0m consisting of riverine materials underlain by bedrock of massive
to moderately jointed, compact, hard Panjal volcanics. The overburden materials
consists boulders, pebbles and cobbles of Andesite, granite, dolomitic limestone
and gneiss is embedded in a sandy silty matrix.
Fig. 4
Geological Section of Cut-Off Wall.
Coupe géologique de l’écran étanche.
839
Q.105 - R.51
5. SITE INVESTIGATION
In Dam area, total 10 no exploratory bore holes with cumulative length of

425.0m were drilled to assess the rock condition beneath the dam foundations.
Additional 04 nos of exploratory boreholes were drilled in which availability of fresh
rock along the cut-off wall at a depth 1.9 to 33m was observed. In Geophysical
survey, seismic refraction survey was also carried out by NHPC to assess the
overburden/bedrock interface.
5.1. KEY OUTCOMES FROM SITE INVESTIGATION INCLUDED
• Detailed geological section based on site investigation provided a high level of

confidence to decide the overburden, weathered rock and fresh rock levels.
• Permeability of the overburden ranges from 2.1 × 10−3 cm/sec to
98 × 103 cm/sec.
• Laboratory tests on core samples done indicate that the uniaxial compressive
strength (UCS) of the rock ranges from 95 to 265 MPa. The Andesite of Panjal
Volcanics rock deeply weathered (upto 7 to 9m depth) and Permeability of
bedrock ranges from 12 to 50 Lugeons.
6. DRILLING
Down the hole hammer (DTHH) drilling system of dia 115mm & 250 mm were
carried out as blast and relief holes respectively in riverine deposits in which one
meter in bedrock has been ensured.
The drilling pattern in a series of 5-metre stretch consisting of

12 No. x Ø 115 mm holes and 05 No. x Ø 250 mm holes were made for the drilling
and blasting preparatory operation to the cut-off wall excavation for entire length
consisting total of about 400 nos. of drill holes.
As the drillholes were carried out through DTHH, so logging of the individual
holes were done based on the penetration rate, hardness evidenced by the drilling,
the rate of advance, cuttings and the sound produced. This has facilitated determi-
nation of the location of the charges and their intensity to achieve the desired result
of rock shattering, without damaging the adjoining strata.
Subsequent to completion of drilling a hole, an inclinometer test was per-

formed on selected holes to accurately measure the verticality and the straightness.
Tolerance on hole final depth of ± 150mm has been ensured.
840
Q.105 - R.51
6.1. KEY FINDING DURING DRILLING:
The length was 2.5m for primary & secondary panels after the concreting.
According to the plan shown below, the depth was each module was adjusted as
per deepest hole of the panel that makes easy for grabbing the material.
7. BLASTING
For charging the hole, long delay Nonel has been used at multiple intervals
through Bottom Initiation System (BOS). A charge density of 0.7 to 1.5 kg/m was
used in the blast holes. As group of four numbers Ø115 mm holes with a Ø 250 mm
hole at the centre of this module was blasted all at once. This module of 4+1 holes
repeats itself every after 1000mm interval. This repetitive module was blasted one
after another with the suitable charges.
In the rock socket the first charge was placed in position with a nominal 300 mm
cushion below the charge to avoid any damage to the underlying bed rock. Depend-
ing on the depth of the cut-off, where 6 or more charges placed within the hole at
those locations, dictated by adverse soil conditions (e.g. large boulders). The blast
holes were cased with PVC pipes jointed with a socket joint (75mm O.D. & 70.4mm
I.D.). Each charge was pre-packed on the ground in a PVC charge-tube (63mm
O.D. with 57mm I.D.) which was lowered down hole to the designated depth defined
during drilling. Multiple charges was lowered down hole, one at a time, only after
adequate stemming has been performed for the previous charge.
The holes were initiated and blasting occurred according to the sequence
mentioned above. Prior to initiating the explosive charges, safety muffle pads made
Fig. 5
Module drilling pattern.
Schéma de perçage du module
841
Q.105 - R.51
Fig. 6
Primary & Secondary blast hole.
Trou de dynamitage primaire et secondaire.
of reinforced concrete (typically, 2500mm x 2500mm x 1500mm) in a truncated

square pyramid form, was placed above the charged holes to avoid fly rock and
surface damage. After completing the charging and blasting for one primary panel
length of 7 m, Ø 250 mm hole, only one in between the two earlier blasted 4-hole
patterns, were drilled with the necessary logging precautions.
8. CONSTRUCTION OF GUIDE WALL
Prior to construction of diaphragm wall it was important to construct guide

walls to maintain the alignment. The guide walls ensured the following:
• To maintain and confirm the correct alignment of the excavation Panel locations
and numbers were systematically set out on the guide walls
• Stability of the upper trench has been kept as it may have got disturbed by
cranes and heavy traffic close by
• It also gave protection against washing due to fluctuating levels of bentonite
slurry thus avoiding upper soil instability
• As stated above due various reasons the upper part of trench is protected
against loads due to traffic etc.
• The guide walls were mainly constructed as an in-situ cast reinforced unit and
offered support for vertical loads imposed by extraction of stop ends and ends
and whose extractor jacks in-turn were supported by guide wall.
842
Q.105 - R.51
Fig. 7
Cross section of guide wall.
Section transversale de la paroi de guidage.
9. GRABBING/EXCAVATION OF PANELS
• At Kishanganga, the excavation tool was a heavy wire rope-suspended

mechanical clamshell grab. That was operated by using a large lattice boom
hydraulic crawler crane (e.g., an 80-tonne Kobelco nominal bite dimension,
were approx. 4.5 metres tall with the jaws opened and weighed 7 tonnes or 9
tonnes).
• Prior to commencement of excavation of the primary panel, the mud reservoir
was created by excavating within the guide walls, for a depth of 800mm to
1.0m, a trench extending for approximately 3.0 metres beyond the ends of the
panel trench. This was filled with bentonite slurry and acted as a surge tank
to minimize the fluctuations of the bentonite level within the trench as the grab
was lowered into it and removed from it during grabbing operations.
• The excavation cut-off wall trench was performed by mechanical grabbing after
shattering of river borne materials by light controlled blasting was done.
• Walls of excavated trench were kept supported by bentonite slurry as usual.
• Chisel was used to break the boulders if by any chance some corners of
boulders still left unbroken and obstruct the grabbing.
10. PLACEMENT PLASTIC CONCRETE WITHIN PANEL
• Concrete was placed within the panel trench by means of tremie pipes sup-
ported by a crane, when necessary, subjected to jerk to promote the outflow
of the concrete from the submerged bottom of the tremie.
• With the steel tremie suspension device in place, the tremie pipe (minimum Ø
250 mm) was introduced into the trench down to 200 mm from the bottom. If
two tremie pipe assemblies were used, as in multi-bite primary panels, then
they placed at least 1.0 m from the panel ends and not closer than 3.0 m from
one another.
• Possible across the panel width. During concrete placement, the end of the
tremies was kept immersed below the surface of the fresh concrete at all times.
843
Q.105 - R.51
• As soon as placement was completed, scum which accumulated on the surface

of the concrete was removed by shovel or other means.
• All handling of concrete were such that discharge of each batch was completed
within 90 minutes of the introduction of the cement in the mix.
• Properties of plastic concrete is provided in Table 1
Table 1
Mix Design for Plastic Concrete.
Mélange pour béton plastique.
Permeability
Aggregate
Aggregate
Admixture
Bentonite
(kg/cum)
(kg/cum)
(kg/cum)
(kg/cum)
(kg/cum)
(kg/cum)
28 days
60 min
Slump
10mm
Water
(Mpa)
WRA
(mm)
(m/s)
OPC
UCS
Fine
At
150 80 551 1023 350 0.98 180 1.57 10−12
Fig. 8 Fig. 9
Crawler Crane. Mechanical Grab.
Grue sur chenilles. Grappin mécanique.
Fig. 10 Fig. 11
Heavy chisel. Placement of Bentonite Slurry.
Burin industriel. Déversage de la boue de bentonite.
844
Q.105 - R.51
Fig. 12 Fig. 13
View of Koden test Equipment. View of placement of plastic concrete
Matériel d’essai de Koden. inside cut-off wall panel.
Déversage du béton plastique.
dans l’écran étanche.
11. CONCULSIONS
Now-a-days deep diaphragm cut-off walls are more frequently in use and
provide optimum solution for an effective water barrier within the soil and weak
rock formations beneath dam. The successful construction of the cut-off wall has
also ensured foundation treatment work for dam which has resulted both in stability
and impermeability (water tightness) of the underlying foundation rock by avoiding
unacceptable under-dam water seepage.
In Kishanganga project, by selecting the most appropriate excavation method-

ology, cut-off wall with depth to the order of 30m in difficult ground conditions like
hard bouldery zone has been executed successfully through drill-blast method fol-
lowed by excavation through mechanically operated clamshell grab. Simultaneous
poring and bentonite slurry and later replacement by plastic concrete ensured the
success without collapse.
The Cut-off Wall was successfully constructed by advance planning to reduce

geological uncertainties by conducting seismic refraction survey and exploratory
drilling that helped in assessing the overburden-bedrock interface. Further, with the
use of drill-blast methodology instead of conventional trench cutter method resulted
in limited risk of cost variation and time overrun. Clearly the use of most modern
technology in the remotest corner of the country has indicated the highest degree
of logistic planning for transport and for beating the adverse winters also. The suc-
cessful execution of the cut-off wall in such a scenario in Kishanganga and in some
other projects such as Dhauliganga Stage-I, Parbati Stage-III, Teesta Stage-III and
MangdeChhu has opened new vistas in tackling deep bed rock profiles in river valley
development projects in the Himalayas. This solution can be used elsewhere also.
845
Q.105 - R.51
REFERENCES
[1] Feasibility Report (1989) – Geology Volume, Kishanganga H E Project-

Engineering Geology and Geotechnical Division, NHPC Limited.
[2] Dar, M.A., Dar, A.R., Bhat, S.S., Raju, J., Bhat, M.S.(2014) Construction of
Cut-off Wall for a 330 mw Hydropower Project at Kishanganga in (J&K). Journal
of Mechanical and Civil Engineering, Vol. 11, Issue 5. Sep-Oct 2014.
[3] Ojha, B.K. (2014) Teesta III Cut-off Wall. International Journal for Technological
Research in Engineering Vol 1 Issue 8 April 2014 ISSN (Online): 2347–4718.
[4] S J Ganvir, SS Bariha, A. Dutta, A. Mishra, S Sankhwar (2011-2017) Unpub-

lished Geological reports/data of construction of cut-off wall of Kishanganga
H E Project.
[5] Gajbhiye, P.K. (2016) Construction of deep plastic concrete cut-off wall in
upstream coffer dam at Punatsangchhu-I Hydropower project, Bhutan. Recent
Advances in Rock Engineering, NIRM/ISRM, Bangalore, November 2016.
846
GENERAL REPORT ON
QUESTION 105
RAPPORT GÉNÉRAL
SUR LA QUESTION 105
GR Q. 105
-------
-------
QUESTION 105

M. POUPART
Independent Consultant Dam Safety
FRANCE
General Reporter/Rapporteur Général
French text: original from the author Texte français: texte original de l’auteur
English text: translation by the author, Texte anglais: traduction par l’auteur,
proofreading by Russel GUNN relecture par Russel GUNN
(OFEN, Switzerland) (OFEN, Suisse)
DOI: 10.1201/9781003211808-98 849

GR Q. 105
TABLE OF CONTENTS
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 852
2. REPORTS SUBMITTED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 852
3. RECENT LESSONS LEARNED FROM DAM INCIDENTS
AND ACCIDENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.1 Available ICOLD Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853
3.2 Other available data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854
3.2.1. Books . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .854
3.2.2. Incident lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854
3.2.3. Annual reports by country . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
3.2.4. Incident reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855
3.2.5. Other sources of information . . . . . . . . . . . . . . . . . . . . . . . . . . 856
3.3 Categorisation of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856
3.4 Use of the data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
3.4.1. Incident list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .857
3.4.2. Incident reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857
3.5 Use of recent statistical data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 858
3.6 Examples of data analysis for some recent cases . . . . . . . . . . . . 861
3.7 How to improve? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866
3.7.1. Possible progress in data collection . . . . . . . . . . . . . . . . . . . 866
3.7.2. Possible progress in data analysis . . . . . . . . . . . . . . . . . . . . . 867
3.7.3. Possible progress in the dissemination of
information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 868
3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .869
4. FLOW AND FLOOD ASSESSMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .869
4.1 Breach hydrograph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 870
4.1.1. Simplified approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .871
4.1.2. Modelling coupling flow and erosion . . . . . . . . . . . . . . . . . . . 872
4.2 Assessment of flooded areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 872
4.2.1. Simplified methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 873
4.2.2. Calculation of the wave by numerical model . . . . . . . . . . . 873
4.2.3. Reporting of the results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874
4.3 Evaluation of the consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 875
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .877
5. EMERGENCY PLANING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 877
5.1 Why emergency plans? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 878
5.2 Regulatory aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 879
5.2.1. General regulations and external emergency plan . . . . . 879
5.2.2. Specific dam regulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 880
5.3 Owner’s actions and internal emergency plan . . . . . . . . . . . . . . . . 881
5.3.1. Internal emergency plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 882
5.3.2. Characteristics of emergency situations . . . . . . . . . . . . . . . 882
5.3.3. Preparatory operational activities . . . . . . . . . . . . . . . . . . . . . . 886
6. GOVERNANCE OF THE SAFETY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887
6.1 At the international level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 887
6.2 At the country level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .888
6.2.1. Regulations at the national level . . . . . . . . . . . . . . . . . . . . . . 889
6.2.2. Professional associations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 889
850
GR Q. 105
TABLE OF CONTENTS
6.3 At the individual owner level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 890

6.3.1. Governance of activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 890
6.3.2. Leadership and safety culture . . . . . . . . . . . . . . . . . . . . . . . . . 891
7. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 892
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 893
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 893
851
GR Q. 105
1. INTRODUCTION
In all hazardous industries, incident analysis is an important tool for improving

safety. Understanding the causes of the incident allows for improvement in what has
been identified as a weakness, whether in the design, construction or operation of
an industrial facility. Dams are no exception to this rule, which is why ICOLD has
always placed great importance on the collection and processing of dam incidents.
An emergency situation can never be ruled out and it is necessary to be pre-

pared for such an event. To do this, accident scenarios must be considered and
their consequences assessed by calculating dam-break waves. The emergency
plans integrate these results and define the roles and responsibilities of all the pro-
tagonists for the detection of abnormal situations, the means of alert and, in case of
extreme emergencies, evacuation of the downstream populations.
Dam safety is a major issue for several reasons: for the public safety of down-
stream populations, for the continuity of services provided by the reservoir (energy,
irrigation, etc.) and finally to preserve a very important asset for the owner. For all
of these reasons, dam safety must be central to the concerns of all stakeholders,
especially governments and owners. An efficient and rigorous governance is a tool
adapted to meet these challenges. It favours the development of a safety culture
that, beyond standards and rules, instils in organisations a “safety-oriented” state of
mind at both the individual and collective levels.
2. REPORTS SUBMITTED
51 reports were submitted for Question 105, 42 of which can be linked to one
of the 4 proposed themes.
Approximately half of the reports deal with theme a) “Recent lessons learned
from dam incidents and accidents”, which can be divided into several categories. 9
articles describe dam or cofferdam failures and explain the unfolding of the accident;
some of them comment on the lessons to be learned. 6 other articles deal with
accidents that did not lead to failure and it is interesting to understand what measures
were taken to avoid the failures. Finally, 4 articles describe in a more general way
the lessons that can be learned from dam incidents and the organisations that are
in favour of transferring these lessons to the younger generation of engineers.
Theme b) on the evaluation of the flows and flood is covered by 7 reports

addressing all the themes. Several very interesting reports present a global synthesis
of practices (R10) or the state-of-the-art of a specific problem such as the breach
hydrograph for embankment dams (R19). The networking of information during a
flood and the calculation capabilities now allow to share between all the stakeholders,
852
GR Q. 105
in real time, the extent of the consequences of a flood evaluated from the real flow
states and their forecasts (R11).
Theme c) on the regulation and organisation of emergency plans is covered

by 7 reports, several of which present the regulations in their respective countries
and the feedback from their implementation. One report deals with the issue of
cross-border emergency plans.
Theme d) on governance of the safety is the subject of 7 reports, only two of

which deal specifically with aspects pertaining to emergency situations. The other
reports describe organisations of the safety or risk review and analysis practices
and cover a broader scope.
3. RECENT LESSONS LEARNED FROM DAM INCIDENTS AND

ACCIDENTS
It is important to firstly recall the definition of the term incident used by ICOLD:
a dam incident can be a failure or an accident:
• A failure is defined as an uncontrolled release of water from the impoundment

or loss of integrity that no longer allows the dam to perform its function safely.
An example is the San Fernando Dam failure. Following a failure it is sometimes
possible to repair the dam, otherwise it is permanently abandoned.
• An accident is an incident that has been prevented from developing into a
failure by taking immediate corrective action to make the structure safe. An
example of a corrective action is reservoir emptying or lowering.
In this chapter, devoted to theme a) of question 105, paragraphs 3.1 and

3.2 describe the available data on dam incidents, and paragraph 3.3 proposes a
categorisation of data. Paragraphs 3.4 to 3.6 show how data analysis allows lessons
to be drawn and paragraph 3.7 proposes avenues for improvement.
3.1. AVAILABLE ICOLD PUBLICATIONS
ICOLD initiated three surveys of all its members to gather as much informa-
tion as possible on dam incidents. These surveys resulted in three publications:
“Lessons from Dams Accidents” in 1974, “ Deterioration of Dams and Reservoirs ”
in 1983, and Bulletin 99 “Dam Failures - Statistical Analysis” in 1995. Bulletin 188
updated Bulletin 99 in 2019 to incorporate failures and incidents that have occurred
since 1991 through 2019. In addition to these publications devoted exclusively to
dam incidents, several bulletins have mentioned dam incidents for specific loading
conditions or degradation processes, most notably Bulletin 82 (Selection of Design
853
GR Q. 105
Flood), Bulletin 109 (Dams less than 30 m high - Cost savings and safety improve-
ments), Bulletin 120 (Design Features of Dams to Resist Ground Motions), Bulletin
164 (Internal Erosion of Existing Dams, Levees and Dikes, and their Foundations),
and the recent Bulletin 180 (Dam Surveillance Lessons Learnt from Case Histories).
Several ICOLD congress’ questions have also dealt with the subject of dam dete-
rioration and incidents Q49 in 1973, Q52 in 1982, Q71 in 1994 and Q75 in 1997.
It should be noted that it has been 24 years since this topic was addressed at an
ICOLD Congress.
The material available in Bulletins, congress papers, and general reports on

the above issues provides a very rich base of “raw” information, that must however
be explored to fully take advantage of the material.
3.2. OTHER AVAILABLE DATA
Numerous documents or publications exist alongside those of ICOLD, which

can be grouped essentially into two categories. There are a few books published
mainly in the United States of America that summarize incidents that have occurred
either at the national or international level. The other documents, more numerous,
are on the one hand periodic reports on incidents, generally annual, published by
some countries, and on the other hand reports concerning the failure of a particular
dam. And for completeness, it is also important to mention papers published outside
of ICOLD conferences and information disseminated via the Internet. It is obvious
that the degree of reliability of these different sources of information is very variable.
The different categories are discussed below with some examples.
3.2.1. Books
The best-known books are those published by the ASCE on dam incidents
in the United States: “Lessons from dam incidents USA I” published in 1975,
and “Lessons from dam incidents USA II” published in 1988. The US Depart-
ment of the Interior published “Dams and Public Safety” in 1980, which includes
a detailed description of 41 dam failures worldwide and the lessons learned from
them. Another example is the UK Environment Agency’s 2011 publication “Deliv-
ering benefits through evidence - Lessons from historical dam incidents”. These
documents have a similar format to ICOLD’s “Lessons from dam incidents”, with a
synthetic presentation of dam characteristics and incidents.
3.2.2. Incident lists
In addition to these documents published as books accessible to all, many

countries publish summary lists of incidents that have occurred on dams in their
854
GR Q. 105
countries. The formats can be diverse but these lists are most often presented as a
table listing the dams and some parameters characterising the dam and the incident.
These documents are generally not directly accessible to the public but can often be
obtained through professional networks, including ICOLD. Examples of this category
of documents are the lists developed by South Africa, Germany, Australia, Finland,
Slovenia, etc.
3.2.3. Annual reports by country
Another source of information is the annual incident reports published by some

countries, some of which are publicly available and generally searchable on the
Internet. This practice of annual reporting is becoming more widespread, no doubt
in part because many countries’ regulations now require reporting of incidents to
the regulating authority.
For example, the ASDSO publishes an annual list of incidents in the United
States, and the Environment Agency in the United Kingdom publishes a similar list,
“Post incident reporting for reservoirs”. There are many other examples of these
annual reports that are more or less easy to find and consult. This information is
sometimes searchable on databases. It is interesting to note that it is almost always
the regulating authority that disseminates these reports, which is logical since it is
the authority that has all the information from the dam operators. In addition to these
annual reports, some supervisory authorities or other state or research bodies have
set up databases accessible via the Internet where all the incidents reported by
the operators can be found. This is the case, for example, of the NPDP (National
Performance of Dam Program) database published by Stanford University in the
USA or the ARIA database of the Ministry of the Environment in France.
The RESBA (Dam Resilience) project is an example of international coop-

eration in the inventory of incidents affecting small embankment dams on the
French-Italian border. It has resulted in a “bi-national” database, which is proba-
bly a first. The geometric and geomechanical characteristics, hazards, issues at
stakes and the incidents can be found. This database currently contains 26 dams
and 30 recorded incidents.
Accident and near-accident analysis reports are intended more for “expert”
stakeholders and address the context of the incident, the description and origins of
failure modes and mechanisms, the consequences, and the corrective actions.
3.2.4. Incident reports
An extremely important source of information is the analysis reports prepared

by experts after a significant incident, usually a failure. In most cases these analy-
ses are presented in papers at professional conferences: ICOLD, HydroPower and
855
GR Q. 105
Dams, SwedCold, DSIG, etc. The format of this type of paper of the order of 10 to
20 pages, and often much more, allows more information to be given than in lists
or summary descriptions, a very good example being the paper published at this
conference on the failure of the Saddle Dam D of the Xe-Pian Xe-Namnoy project in
Laos (R26). There are many articles describing dam failures that are more or less
easy to access, especially for older failures.
By far the richest sources of information are the analyses of incidents reported
by experts. Some countries provide access to these documents via the Internet,
which represents an invaluable opportunity for the international community to under-
stand the causes of incidents. The United States of America is exemplary in this
respect: among others, we can cite the reports on the failures of the upper reservoir
of Taum Sauk, the Delhi dam and more recently the serious incident of the Oroville
dam spillway. Other examples are the reports of the Camara Dam failure in Brazil (in
Portuguese) and the Toddbrook Dam failure in the United Kingdom. These reports
describe in detail the characteristics of the structure, the investigations carried out,
the analysis of the causes and consequences.
3.2.5. Other sources of information
Finally, the development of information technology has given access to a great

deal of information on dam incidents. These are mainly online newspaper articles
that hardly ever provide technical information. Their sole purpose is to warn of
the occurrence of an incident and its consequences. This at least has the merit
of drawing international attention to the incident and seeking further information.
3.3. CATEGORISATION OF DATA
As a preamble, it is important to note that the documents cited in the preceding

paragraphs are mostly from English-speaking countries. It is certain that publications
in other languages exist but are more difficult to identify. A search on the Internet in
Spanish or German, for example, yields interesting results, although it is necessary
to be able to read them. This is an important bias, and the dam community should
find ways to limit this gap as much as possible.
Finally, documents dealing with dam incidents can be categorised into two
types: incident lists and incident reports.
• The incident lists are usually quite comprehensive but do not provide much
detail about the dam and the incident. Their interest is to provide a basis
for statistical studies, such as the updating of the bulletin 99 for example.
These lists exist at the international level, mainly at ICOLD, and at the level
856
GR Q. 105
of each country. The “national” lists are generally more detailed than the lists
at the international level and include failures but also accidents. To compile
an international list from all this data is a difficult exercise because the same
dam may have several names and, worse, the data between these lists is
sometimes very different.
• The incident reports concern a single dam and provide information on the
dam, the incident, its causes and consequences, with a varying level of detail
depending on the report. It is from these detailed analyses that the main
lessons of the failures are obtained.
3.4. USE OF THE DATA
Having information on dam incidents is essential, but it will not be very useful
if it is not properly used. This usage is obviously different depending on whether it
is a question of incident lists or incident reports.
3.4.1. Incident list
For incident lists, statistical analysis is the most appropriate method of process.
The data must be available in numerical form and the parameters describing each
incident are numerical values or codes.
Statistical analyses are useful to understand the overall picture, but they do not
allow to change technical practices, they can give trends by classifying the failures
according to the types of dams, the evolution of failures in time, the most frequent
contexts in which the failures occur, the modes of failure, the causes. The update of
Bulletin 99 (see 3.5) is a good example of the benefits that can be derived from this
type of analysis.
This collection of data and the analyses testify to the effort of the profession
to have the best possible inventory of failures and accidents. However, it must be
kept in mind that one cannot want to have both an exhaustive inventory of failure
and detailed analyses.
The statistical analysis of failures should be done at the ICOLD level. That of
accidents at the level of countries and/or states and made public.
3.4.2. Incident reports
The analysis of incident reports first provides lessons for the dam that suf-
fered the incident. This type of report includes an analysis of the causes which is
857
GR Q. 105
the essential element to draw useful lessons. The quality of this cause analysis is
essential and must seek the root causes of the failure or accident. In the past, the
causes identified were essentially technical. They had the great merit of changing
design practices. These analyses have often been transmitted to the profession via
books, ICOLD technical bulletins or the publication of research works.
Nowadays, analyses of causes also include organisational and human causes.

This is a more difficult exercise, but it provides a lot of information on non-technical
factors: organisation of the operator, collective or individual human failure, relations
between the actors, transmission of information, supervision by the control authority,
etc. This practice is widespread in the industry and should be systematically applied
to all dam accidents. A well-conducted analysis of the causes often shows that
behind the technical factors, which are generally well identifiable, lie the deep-rooted
causes which are in the domain of the general safety organisation and human failure.
Accident reports thus lead to more detailed information on the technical causes
and can illustrate with concrete examples the trends indicated by the statistical anal-
yses. In general, the technical investigations are useful for a detailed understanding
of the failure of the analysed dam, but their extension to all dams is not directly pos-
sible: it is the role of the ICOLD technical committees to analyse how a particular
failure contains generic lessons. This is also true for the “human” causes, whose
lessons are sometimes more easily transposable to many cases in the world, despite
the differences in organisation and context. For these human factors are always the
root causes of a failure.
Finally, it should be noted that analyses of accidents are just as, or some-
times even more, useful than those of failures. An accident that has been properly
managed provides insights into the organisational management of the crisis, the
detection and evolution of the hazardous phenomena and the measures taken
to avoid the failure. The recent ICOLD Bulletin 180: “Lessons learned from case
histories” perfectly illustrates this accident analysis.
3.5. USE OF RECENT STATISTICAL DATA
The three ICOLD publications cited in 3.1 and various articles have succes-
sively evaluated failure statistics by looking for the influence of dam type, year of
construction, etc. The update of Bulletin 99 undertaken by a working group within
the Dam Safety Committee from 2017 to 2020 added failures occurring since 1991
through 2019 into the statistical analysis. This update also went back to previously
known cases to validate their data and add fields that were not included in previous
studies, such as context, failure mode, causes of the incident. The main findings are
as follows:
858
GR Q. 105
• There are significant differences in the accuracy and reliability of the failure
reports among the ICOLD countries, making it necessary to discard some data
so as not to distort the results of the statistical analysis.
• The ratio of the number of failures divided by the total number of existing large
dams is continuously decreasing from 1.42% during the years 1900-1925 to
0.12% since 2000.
• However, the ratio of failed dams built during a certain period provides a less
positive view. This ratio was 0.29% for the years 1975-1999 and is 0.38%
since 2000. This point is illustrated in Figure 1, which would tend to show a
slight increase in the failure rate over the last 20 years. Another explanation
could obviously be a better detection of dam failures since 2000 thanks to the
progress of information technologies (Internet, ...).
Fig. 1
Failure rate versus construction period
• As already indicated in Bulletin 99, the first ten years of a dam’s life are always
the period when 50% of the failures occur. This ratio is even 100% for arch
and buttress dams.
• For dams between 15 and 75 m in height, the failure rate relative to existing
dams of the same height is fairly constant. For higher dams, it decreases
significantly. To date, no dam over 100 m has failed.
• There is no significant influence of dam type and reservoir size on the failure
rate. The number of failures of multiple arch dams and buttress dams is too
small to be statistically significant.
• The three failure modes for embankment dams are overtopping (40%), internal
erosion (39%) and structural failure (21%).
• For rigid dams, foundation failure and internal foundation erosion are the
dominant failure modes. Structural failures occur primarily in masonry gravity
dams.
859
GR Q. 105
• Dam failures occur either during normal operation or during a flood, with these
two contexts accounting for 90% of failures. Since 2000, 70% of the failures
have occurred during a flood. In comparison, earthquakes have caused a very
low number of failures.
Fig. 2
Organisational causes for the overtopping failure mode
• Regarding the organisational causes, it can be stated that inadequate design

or construction are by far the most important causes identified for arch, but-
tress and multi-arch dams. For other types of dams (gravity, embankment),
inadequate operation during floods appears to play a role in about 20% of
failures. An analysis of the evolution of organisational causes of failure over
time for the crest overtopping mode of failure is presented in Fig. 2. From 1950
onwards, the organisational causes due to operational or maintenance defi-
ciencies become more and more important compared to design deficiencies.
This indicates that while design deficiencies were the most important cause
of dam failures prior to 1950, nowadays it is operation or maintenance defi-
ciencies that are the most important causes. Recent dam failures that have
made their mark since 2000, such as Taum Sauk, Delhi, Ivanovo or Oaky, bear
witness to this.
• The technical causes are different for different types of dams: Foundation
failures are the dominant cause for arch and buttress dams. For masonry
gravity dams, structural failures of the dam body are an important technical
cause. For other types of gravity dams, structural failures are as important
as foundation failures, while insufficient spillway discharge capacity is also a
cause. For earth dams, the two dominant causes are geotechnical problems
(66% of causes) and insufficient spillway discharge capacity (28%). For rockfill
dams, these two dominant causes are distributed differently, with geotechnical
860
GR Q. 105
failures accounting for only 32% and insufficient spillway discharge capacity
accounting for 64%.
3.6. EXAMPLES OF DATA ANALYSIS FOR SOME RECENT CASES
The last twenty-five years have been marked by dam failures, some of which
have unfortunately resulted in several casualties. The table below summarises the
dam accidents with a height of more than 15 m known to the reporter on which the
main known characteristics have been indicated. Other failures occurred during this
period (in total there are about 100) but the available data is insufficient to include
them in this table.
It should be noted, however, that some smaller dams have resulted in more
casualties, for example Shakidor in Pakistan in 2005 (70 casualties) and Koshi in
Nepal in 2008 (250 casualties). The Oroville spillway accident is included in this list
even though the dam was not damaged during the event, but it is considered to be
a loss of overall integrity of the facility and an uncontrolled release of water.
YEAR OF DAM FAILURE NUMBER OF

FAILURE NAME COUNTRY TYPE H (M) MODE CASUALTIES
1996 KENOGAMI Canada TE 21 Overtopping - 10
External erosion
1997 OPUHA New Zealand TE 50 Overtopping -
External erosion
1999 LYELL Australia ER 51 Important
DAM disorders
1999 SHIH Taiwan BM PG 25 Loss of support
KANG (from
foundation,
abutment
1999 EL Venezuela TE (Z) 60 Overtopping - 10
GUAPO External erosion
2000 FRY South Africa TE 21 Overtopping -
External erosion
2000 MAMBEDI South Africa TE 22 Important
LOWER disorders
2000 LEBEA South Africa TE/VA 18 Overtopping -
External erosion
2000 KOOS DE South Africa XX 15 Overtopping -
BEER External erosion
2001 CHANG India TE (Z) 15,5 Failure, cracks
or settlement
(foundation or
fill dam body)
861
GR Q. 105

2001 TAPPAR India TE (H) 15,5 Mass movement
(sliding, tilting,
settlement in the
dam body)
2002 ZEIZOUN Syria ER/TE (Z) 32 Overtopping - 22
External erosion
2003 BELLAIR South Africa TE 16 Overtopping -
External erosion
2004 CAMARA Brazil PG (RCC) 50 Foundation 7
failure (“rigid”
dams)
2004 BIG BAY USA TE 17,4 Interface erosion /
leakage
2005 NANDGAVHAN India PG (M) TE 19 Mass movement 14
(sliding, tilting,
settlement in
the dam body)
2005 TAUM USA TE ER 25 Overtopping -
SAUK External erosion
2006 KA LOKO USA TE/ER 15 Overtopping - 7
External erosion
2006 TUILIERES France PG (M) 31 Important disorders
2007 JASWANT India PG (M) 43 IE/Leakage
SAGAR in the foundation
2008 PALEM India TE 46 IE/Leakage
VAGU in the foundation
2008 BHIMLAT India CB (M) 17 Overtopping -
RESERVOIR External erosion
2008 BEDFORD Australia PG 16 Important disorders 1
WEIR
2009 ALGODOES Brazil TE 21,6 Overtopping - 9
External erosion
2009 SITU Indonesia TE/ER 16 Overtopping - 98
GINTUNG External erosion
2010 NIEDOW Poland TE PG 16,7 Overtopping - 1
(WITKA) External erosion
2010 GARARDA India TE 32 Internal Erosion /
leakage
(foundation
or fill dam body)
2010 DELHI USA TE PG 18 Overtopping -
(Hartwick dam) External erosion
2011 FUJINUMA-IKE Japan TE 18,5 Mass movement 8
(sliding, tilting,
settlement in the
dam body
2012 FORMIN Slovenia PG TE 49 IE/leakage
in the Dam
Body
2012 IVANOVO Bulgaria TE 19 Overtopping - 8
External erosion
2013 OAKY Australia ER/PG 18 Overtopping -
External erosion
862
GR Q. 105

2014 KREL_2 Vietnam TE (H) 27 Overtopping -
External erosion
2014 HA DONG Vietnam TE 27,5 Overtopping -
External erosion
2017 OROVILLE USA TE/ER 235 Overtopping -
External erosion
2018 SOLAI Kenya TE 25 Internal Erosion / 50
leakage
(foundation or
fill dam body)
2018 XE NAMNOY Laos TE 16 Internal Erosion / 65
saddle dam leakage
(foundation
or fill dam
body)
2019 TIWARE - India TE 28 Internal Erosion / 26
TIVARE leakage
(foundation or
fill dam body)
2019 SPENCER USA BM TE 14 Overtopping 1
2020 EDENVILLE USA TE PG 13 Internal Erosion /
leakage
(foundation or
fill dam body)
2020 SANFORD USA TE 11 Overtopping
2020 THIN Viet Nam TE 15 Internal Erosion /
leakage
(foundation
or fill dam
body)
2020 SPARMOS Greece TE ? ?
2020 SARDOBA Uzbekistan TE 32 Unknown or 6
Unclear
All of these failures are obviously dramatic when there are casualties, but also
because of the environmental damage and of course for the owner and operator of
the dam who suffer a considerable economic loss. The questions that arise after a
dam failure are always the same, as recalled in R12: Why did it happen, how did it
happen, and probably the most important question: how to prevent it from happening
again? To answer these questions, dam professionals have an essential tool at their
disposal: the in-depth analysis of the failure.
Several reports describe the process of post failure analysis. One example is
reference [19] published by the ASDSO in the United States, which provides very
useful recommendations and guides for conducting post-failure investigations. The
generic lessons that can be learned from the reports that address this topic are as
follows:
• The analysis of a dam failure must be undertaken systematically and rapidly

after the event, as some field evidence may disappear. The collection of
863
GR Q. 105
testimonies from those involved must also be rapid, both to avoid external inter-
ference and to gather the maximum amount of information while memories are
still “fresh”;
• It is very advisable to form a panel of independent and multidisciplinary experts
because the technical subjects are generally multiple and require various
skills;
• The analysis reports are very thorough on the technical causes of the failure.
These in-depth technical analyses are obviously essential to understand the
mechanisms of the failure of the dam concerned. Their transposition to other
structures is not always possible because each dam is a prototype and the
characteristics of the foundation and the materials of the dam that may have
played a role in the failure are never identical on other sites. On the other hand,
there is often much to be learned from a failure in terms of design, monitoring
and maintenance methods, which can then be transposed to other structures.
It is well known that the failure of the Malpasset dam has advanced the meth-
ods of evaluation of the foundations of arch dams and that the failure of the
Balderhead dam has allowed the discovery of hydraulic fracturing phenom-
ena, etc. ICOLD technical committees have a role to play in the consolidation
of these failure data and the consequences to be drawn from them to issue
generic recommendations in the form of a bulletin. It can be noted here that
many bulletins describe failures and the consequences to be drawn from them
for the profession.
• The authors of some of the reports submitted to question 105 make the
following recommendations, for example:
◦ Xanmoy (R26): The lessons learned focus a lot on the geotechnical
aspects of the design and on monitoring. But the authors also note that
there was no way to control the water level in the reservoir and that the
contingency plan did not address the saddle dam that failed. These are
two avenues of improvement that can be applied to many facilities.
◦ Malpaso (R31): This description of an incident on a CFRD and the asso-
ciated corrective measures can be a reference on incident management,
both technically and organisationally.
◦ Toddbrook (R16): The serious incident at Toddbrook Dam, following the
one at Ulley Dam, led to a review of safety management and the regu-
lation of dams in Great Britain. This review led to recommendations for
changes in regulations, particularly for inspection and incident reporting.
It can be noted here that in almost all countries serious incidents have
to occur for the rules imposed by the States to be enacted, and then
reinforced in case of new incidents. Article R12 also explains this for the
United States. Finally, the report states that the crisis management at
Toddbrook was successful because the owner had an emergency plan
and regularly performed practice drills.
◦ Oroville, Taum Sauk (R11): The authors note that safety assess-
ment using potential failure modes analysis (PFMA) does not assess
the probability and consequences of a failure. FERC has published
864
GR Q. 105
recommendations on “RIDM” (risk-informed decision making) methods,

but these are not widely used by consultants because they are not
mandatory. A review of two recent accidents in the USA (Oroville and
Taum Sauk) was undertaken to see if the use of RIDM analyses would
have prevented the incidents. For Oroville the authors present a detailed
description of the successive PFMAs in 2004, 2009, 2014 and the evo-
lution in quality of these reviews. They explain why the main spillway
slab uplift and emergency spillway regressive erosion scenarios that the
technical causes of failure were, had been ruled out.
• A fundamental point emphasised in several reports concerns the importance
of human factors. Following the analyses of the Taum Sauk and Oroville spill-
way failures, the authors of the R11 report write that “human factors are as
important as technical factors”. In absolute terms, one could even say that the
causes of dam failures are entirely due to organisational and human factors!
Indeed, since a dam is by nature an artificial structure designed and operated
by humans, its failure results either from a bad design, including a bad appre-
ciation of external stresses (floods, earthquakes, etc.), of the characteristics
of the foundation or of the materials used, or from an imperfect construction,
or from a faulty operation! But in general, the term “human factors” is used to
describe organizational failures and inappropriate actions by operators. The
feedback from failures over the last 25 years clearly shows that “technical”
causes are less important than in the past and that the cause of many failures
is primarily attributable to organisational and sometimes human errors.
• The corollary of this importance to be given to human factors, is to have analy-
sis methods dedicated to this objective. Indeed, the search for what are called
“root causes” requires analytical skills that engineers generally do not possess.
This awareness is emerging in some organisations such as ASDSO with con-
tributions from specialists in these fields, such as I.A. Alvi, but is far from
being widespread. Other industrial sectors such as aeronautics or the nuclear
industry have been practicing these “in-depth” analyses for a long time and
the analysis methods used are certainly transposable to the field of dams.
• Compared to the technical causes, the causes linked to the human factor are
very often generic because they do not concern a particular structure: they
can therefore be useful to a wide range of professionals. We can quote James
Reason: “We cannot change the human condition, but we can change the
conditions under which humans work”.
• Finally, several reports describe incidents that were managed correctly and
avoided dam failure. The analysis of these serious incidents is just as impor-
tant as that of the failures, if not more so. Indeed, everything related to crisis
management and the decision chain is often generic and can be applied to
other organisations. A fundamental aspect of these analyses is to show how
early detection of the incident allows the company to react with sufficient speed
to avoid the failure. The recent Oversight Committee Bulletin 180 “Lessons
learned from case histories” describes many incidents and how they were
handled. Management in a crisis is the subject of topic d) of this question 105
and is discussed in more detail in Chapter 6.
865
GR Q. 105
3.7. HOW TO IMPROVE?
The analysis of some recent dam incidents is rich in lessons learned. The
in-depth reports carried out for some of these incidents allow a good understanding
of the technical reasons for the incident and sometimes to trace the root causes
relating to the organisation or the human factor.
However, progress is still needed in several areas to make the best use of
these incidents for the benefit of the dam community. The difficulties in doing so are
several-fold:
• It is remarkable that the two most comprehensive documents published by

ICOLD date back to 1974 and 1983. The time spent by the members of the
“Committee on Failures and Accidents to Large Dams” to compile this col-
lection of incidents and analyse them must have been considerable. Their
respective organisations agreed at the time to contribute “in kind” to the work
of the ICOLD by giving them the necessary time. Unfortunately, it is certain
that today this would no longer be possible.
• Then there is the time and resources needed to analyse a failure. The
organisation and cost of a panel of experts are not within the reach of all
organisations.
• Perhaps the most significant difficulty is the lack of transparency about
incidents. There are two reasons for this:
◦ first of all, a necessary confidentiality linked to legal actions. The analysis
of the causes is then very delicate because the results can be used by
the judicial investigation and that can delay considerably the diffusion of
elements on the technical causes. In some cases, there is even a fear
that the conclusions of the analysis may be concealed.
◦ then a political will, on the part of certain countries or certain owners,
not to disseminate information deemed negative.
• Finally, language is sometimes an obstacle: English is the predominant lan-
guage of the international community, but publications available in other
languages should not be overlooked. Some failure reports present very useful
lessons but are not read even if their existence is known.
3.7.1. Possible progress in data collection
It is obvious that the availability of reliable and transparent data is a prerequi-

site for analysing incidents and drawing relevant lessons from them. Without data,
nothing can be done. Unfortunately, there are still many examples of dam failures
for which no information is available at the international level. The only information
is often newspaper articles on the Internet, which are not very reliable and do not
provide any technical information.
866
GR Q. 105
On the contrary, some countries are exemplary, such as the United States,
which posts databases of incidents at its dams on the Internet and then publishes
detailed analysis reports.
Many countries have enacted regulations requiring owners or operators to

report incidents that have occurred on dams under their control to the authorities.
These requirements are often accompanied by severity levels, which then determine
the speed with which incidents must be reported. These reports may be brief at first
but must be followed by an analysis report by the owner or operator. The number
of incidents can be quite large, and it is unlikely that this can be consolidated at
an international level, except for the most serious incidents. However, the analysis
of these near misses is extremely important to identify technical or organisational
deficiencies and should be conducted at the level of each member country, usually
on an annual basis. The most significant ones could be reported to the ICOLD.
For dam failures, it seems obvious that they should be declared at the interna-
tional level by the member countries, the simplest way being to go through ICOLD
and in particular through its Dam Safety Committee. The information transmitted to
this committee would then be entered in the database of dam failures available on
the ICOLD website.
3.7.2. Possible progress in data analysis
The use of incident data is also essential if lessons are to be learned. This anal-
ysis is generally to be done at two levels: that of the lists of incidents, the objective
of which is rather to have a fairly exhaustive but not necessarily very detailed col-
lection with the objective of carrying out statistical analyses. These lists can include
all the incidents declared by the owners and operators at the level of the member
countries and only the failures and serious incidents at the ICOLD level. Bulletin 188
is an example of what can be done at the ICOLD level. At the level of each country,
these analyses can be more detailed, on an annual basis, and used as indicators
of progress based on the number and severity of incidents reported, for example.
These incident lists lend themselves very well to the format of databases in
which many parameters characterising the incident can be entered. This is done at
ICOLD (Bulletin 188) and in many countries. A trap to be avoided when developing a
database is to provide many fields, sometimes a hundred (!) to characterise the dam
and the incident. Experience shows that most fields are not filled in, which makes
statistical analyses on these fields impossible. It is better to foresee fewer fields but
to know the values with a good degree of reliability and for most incidents.
As far as failures are concerned, the purpose of detailed analyses is to under-

stand the causes of the failure, whether they are technical or related to the human
factor. The analysis of the causes related to the human factor is a fairly new exercise
and it would be useful for the ICOLD to work on a guide to carry out these in-depth
867
GR Q. 105
analyses based on the existing guides in the industrial world. The presence in the
panel of experts of an analyst specialised in the human factor should be systematic
for failures and for serious incidents.
3.7.3. Possible progress in the dissemination of information
The R12 report is extremely explicit about the challenges of sharing lessons
learned from dam incidents: “Feedback on dam incidents is effective to the extent
that the industry makes an effort to investigate new incidents and research past
incidents. However, it is often difficult to share these lessons widely across individ-
ual organisations and international borders. Compounding this problem, the dam
safety industry is currently experiencing high rates of retirement among engineers
and those responsible for monitoring and regulation. Much of the knowledge and
experience gained from decades of building new dams and learning from incidents
will be lost unless these lessons are passed on to the next generation of dam safety
professionals.”
All this is summarized in the most important question and subject of this Q105:
“How can we prevent this from happening again?”
The various means of disseminating information on dam incidents can be

grouped into two broad categories: documents in the broad sense made available
to the profession, and conferences, training courses, etc. devoted to dam incidents.
• Documents in the broadest sense include expert reports, articles, national or

international lists, databases, etc. as presented in 3.3. Most of these docu-
ments should be available on the Internet. But having these documents is not
enough: first of all, there are many of them and, unless one does specific
research on a type of failure, or a type of dam, it is difficult for an engineer to
get acquainted with this set of documents, either because of lack of time, or
because of ignorance of the existence of these documents.
• Therefore, the second means of transmitting the acquired experience must be
through specific conferences or training sessions. Several organizations have
organized international conferences of this type, for example DSIG in 2009 and
SWEDCOLD in 2013. At the national level, a notable example comes from the
United States, where ASDSO includes a session dedicated to the presentation
of dam incidents at each annual meeting. The initiative called “decade series”
consists in presenting incidents that occurred in the years corresponding to the
decade of the year of the meeting (for example in 2019, dealing with incidents
from 2009, 1999, ...1909, etc.). This allows young professionals to be informed
by experienced peers about the lessons to be learned from past incidents. It
also represents a lot of work for those in charge of these presentations.
ICOLD could implement a similar initiative annually or at each congress under

the responsibility of the Dam Safety Committee.
868
GR Q. 105
3.8. CONCLUSION
The question in theme a) is “what lessons can be learned from recent

incidents?”
Summarizing the analyses developed in this chapter, the main lesson to be

learned from the incidents that have occurred over the last 25 years is the following:
these incidents are increasingly occurring in a flood context and, as a result, their
causes must be sought in operational or maintenance deficiencies. In these contexts,
the root causes linked to human factors are very present.
ICOLD Bulletins address many technical aspects of dam design and construc-
tion. Other bulletins provide recommendations for operation and maintenance. It can
be considered that the respect of the recommendations contained in these bulletins
would have allowed to avoid most of the recent failures.
But the root causes are the most important and are related to the organisation,
the human factor, etc. The two recent bulletins 154 and 175 on dam safety man-
agement for all phases of the dam’s life provide fundamental elements that all dam
owners and operators should implement. This would considerably limit the risks of
serious incidents and failures.
4. FLOW AND FLOOD ASSESSMENT
Topic (b) of this Question 105 deals with “Flow and flood assessment, esti-
mation and quantification of consequences in case of failure or incidents, including
social, economic and environmental aspects”.
ICOLD Bulletin 111: “Dam-Break Flood Analysis, Review and Recommenda-

tions, 1998” is the only ICOLD publication that deals exclusively with this subject.
Bulletin 59, “Dam Safety - Guidelines”, refers very briefly to the need for flood
mapping for emergency planning. This topic has also been discussed at ICOLD
congresses in connection with dam safety issues. For example, in Question 40
“Environmental Consequences of Dam Construction” or Question 68 “Assessment
and Improvement of Safety of Dams in Service”, there are articles that deal with the
assessment of the dam failure wave and its consequences.
The objectives of the flow and flood assessment were defined in Bulletin 111
and have not changed since its publication. These objectives are as follows:
1. Obtain basic data for the establishment of emergency plans (map of flooded
areas, dynamics of dam-break flood waves);
869
GR Q. 105
2. Collect data necessary to classify the dam by quantifying the effects of its
failure;
3. In some cases, assess the risks for large downstream facilities whose
destruction presents serious dangers (nuclear power plants, chemical
plants).
The authors of Bulletin 111 also noted that the technical requirements for
dam-break flood wave studies are not specified in laws or regulations. The dam
safety authorities define requirements on the principles of the calculations and the
assumptions to be adopted but leave the choice of methods to the dam operators.
These methods are then included in recommendations published by professional
associations. Reports R10 and R19 of this Question 105 fully illustrate this practice.
Compared to Bulletin 111, the reports issued for this Question 105 cite sce-
narios other than dam failure. One is the risk of a natural dam failure created by a
landslide damming a valley. The other scenario that is increasingly being studied is
the failure of a gate when its discharge can create a wave equivalent to a major flood
downstream. These scenarios are not always required by the dam safety authorities,
but it is the responsibility of the dam owner to study them.
The effectiveness of population evacuation and the safeguarding of economic

assets in the event of a dam failure relies in particular on the prediction of the
propagation of the flood wave. The characterisation of the flood following this prop-
agation of the flood wave is thus an essential element among the safety provisions
associated with any dam which presents significant downstream stakes (4.2). The
prediction of this flooding is determined by a hydraulic calculation, whose upstream
limit condition is the breach hydrograph (4.1).
The evaluation of flows and flooding can be broken down into three parts, which
are discussed below. First, there is 1) the evaluation of the breach hydrograph, 2)
then the calculation of the wave propagation and the extent of the flooded areas,
and finally 3) the evaluation of the consequences. All that concerns the operational
monitoring of the dam, the detection of anomalies that could lead to failure, the
organisation of the operator to face an emergency, the alerting of the population,
the distribution of responsibilities between the operator and the public authorities,
etc. is dealt with in chapters 5 and 6 below.
4.1. BREACH HYDROGRAPH
The knowledge of the breach hydrograph is the first essential step to evaluate
the characteristics of the flood wave and the flooded areas. This consists in evaluat-
ing the flow rate through the breach in the dam as a function of time. This evaluation
depends on the type of dam considered, as it is quite different depending on whether
it is a “rigid” dam (concrete or masonry) or an “embankment” dam (earthfill or rockfill
dam).
870
GR Q. 105
For rigid dams such as gravity dams, multi-arch dams or buttress dams, current
practice assumes a near-instantaneous failure of either the entire dam, or one or
more blocks (monoliths) or buttresses. The hypothesis of the total collapse of the
whole dam is certainly conservative because the history of failures of this type of
dam indicates that the general case is the failure of some blocks. The choice must
be made according to the specific characteristics of each dam and the shape of the
valley. For arch dams, the usual hypothesis is the total and instantaneous failure of
the whole dam.
The maximum flow rate and the shape of the hydrograph can be evaluated by
assimilating the breach to a hydraulic weir which allows to obtain the maximum flow
rate and then to construct a hydrograph from this maximum flow rate and the volume
of the impoundment. This approach is now replaced by the more accurate calculation
of the hydrograph of breach by the numerical simulations of the propagation of
the dam-break flood wave described in 4.2. In this case, the boundary condition
upstream of the numerical model is not a pre-established breach hydrograph but
results from the complete calculation of the flow through the breach taking into
account the emptying of the reservoir and the flow conditions downstream of the
dam.
The case of embankment dams is much more complex because the formation
of the breach depends on the event initiating the failure, which may be external ero-
sion by submersion of the crest or internal erosion by formation of a piping hole. In
both cases, the duration of the breach formation and the evolution of its dimensions
depend on the geotechnical characteristics of the material constituting the embank-
ment and on the height and velocity of the current. The physical processes result
from a complex interaction between the flow of water over or through the dam and
the dam material which may or may not have cohesive properties.
Current practice is well described in report R19, which discusses first simpli-
fied approaches based on empirical relationships and then modelling approaches
coupling flow and erosion.
4.1.1. Simplified approaches
These practices are very well covered in reports R10 and especially R19. They
consist of estimating a peak breach rate by simplified empirical relationships and
then constructing a breach hydrograph. It should be noted that these relationships
apply to two quite different breach processes, external erosion by crest flooding,
and internal erosion. Comparisons of the predictions obtained by these simplified
relationships with measured flows from dam failures are presented in Report R19.
These comparisons are also used to calibrate certain parameters of the relations
used. However, the results of these evaluations are sometimes quite different from
the measured values and lead to the recommendation to use several relationships
and to rely on the engineer’s judgement to select a major and a minor for this
871
GR Q. 105
breach flow. It is also recommended that a sensitivity study be performed to test

the consequences of these choices on further studies.
Two simplified approaches are then used to construct the breach hydrograph.
◦ When the maximum breach flow is estimated as described above, a triangular

shape is chosen for the breach hydrograph. The two important parameters
are the time to reach the peak flow, which is estimated by expert opinion
and usually varies from 30 minutes to a few hours. The second parameter is
the volume of the hydrograph, which must correspond to the volume of the
reservoir.
◦ The second approach is to estimate the peak breach flow using a thresh-
old law and an estimate of the maximum breach width, and then construct a
simplified breach hydrograph based on an average breach widening velocity.
This average breach widening velocity is an important parameter that is a
function of the erodibility characteristics of the materials and the time of con-
struction of the dam. According to R19 the range of these widening velocity
values is very wide, varying from 15 m/min to 0.01 m/min.
4.1.2. Modelling coupling flow and erosion
Physics-based deterministic modelling of breach initiation and development

mechanisms in embankment dams is currently not available for operational engi-
neering studies. The only exception is the case of overtopping erosion of a
homogeneous embankment, of simple geometry, made of cohesive materials. For
this last mechanism, a major research effort has been carried out since the end of
the 1990s on the physical description of the phenomena and their numerical and
experimental modelling on a large scale. This research effort has led to the vali-
dation of two calculation codes adapted to the needs of engineering, WinDAM and
EMBREA.
The report R20 presents ongoing research efforts to improve knowledge of

these internal and external erosion mechanisms.
4.2. ASSESSMENT OF FLOODED AREAS
The evaluation of the propagation of the breach hydrograph downstream of

the dam is undoubtedly the most mature stage. The practices described in Bulletin
111 remain for the most part current, the notable developments being mainly in the
refinement of numerical calculation tools made possible by the increase in computer
power and in numerical field modelling. However, it should not be forgotten that
simplified calculation models can be used in certain circumstances or for modest
structures.
872
GR Q. 105
4.2.1. Simplified methods
Two of these methods are presented in R19. They provide the flows at strategic
locations identified in the flood zone, which makes it possible to evaluate the level
reached when the height/discharge relationship is available at these points. These
simplified models are useful in case of emergency, for example to quickly evalu-
ate the characteristics of a breaking wave downstream of a natural dam formed
by a landslide. However, they present strong uncertainties, and it is usual to take
important safety margins on their results.
4.2.2. Calculation of the wave by numerical model
Numerical modelling by solving the St Venant equations is now the reference

method for establishing a break wave. In most cases, a 1D modelling is sufficient.
Indeed, many dams are located in rather deep valleys where flows follow the valley
bottom without recirculation or other complex flows. However, 2D modelling may be
necessary in large floodplains, in case of high river overflow. Most software packages
incorporate both approaches and a combination of the two. Among many others,
we can mention the HEC-RAS software developed by the USACE and TELEMAC
developed by EDF (Electricité de France), both freely available.
The input data required for these models are the geometry of the flood-
plain, the hydrograph entering the reservoir and/or the breach hydrograph, and
the assumptions about the scenarios.
◦ The field data necessary for the modelling are generally available and already
digitized. It is necessary to ensure that they have the necessary accuracy,
in the order of 5 to 30 m in spatial resolution depending on the topography
of the riverbed and submerged areas and 30 cm to 1 m in altimetry and
that their quality is consistent with the requirements of the study. Indeed,
the quality of the description of the geometry of the flooded area is funda-
mental to guarantee the quality of the results. If necessary, these data are
completed by spot surveys, for example with LIDAR technology, and by inte-
grating the detailed geometric characteristics of structures that may have an
influence on the flow, such as dykes, bridges, etc., and with bathymetries
(see R10).
◦ The calculations of the flooded areas are performed for several failure sce-
narios, at least for a failure during a flood event and a failure in normal
operation (“sunny day”). The flow conditions downstream of the dam and
the initial water level in the reservoir are initialized accordingly. Similarly, the
hydrograph entering the reservoir is generally chosen as the design flood
for the “flood failure” simulation, and as the mean inflow for the “dry weather
failure”. The R10 report presents a very complete overview of the different
scenarios that can be considered.
873
GR Q. 105
◦ For each scenario it is also usual to evaluate the flooded areas with and
without dam failure, at least for the flood failure case. This allows to
quantify the incremental consequences caused by the dam failure. This
type of incremental analysis can also be done in the context of spillway
rehabilitation studies (R19). The results allow quantifying the gain of the
rehabilitation works and provide important elements in the decision-making
process.
Finally, one should never forget that these simulations include an element of
uncertainty. A good practice is to increase the water level in the flooded areas and
to reduce the arrival time of the wave (R19).
The case of tailings dams is special (R28) and particularly difficult because of
the nature of the flows. The main difficulties specific to these dams are the avail-
ability of data and the non-Newtonian nature of the flow to be integrated into the
hydrodynamic models. The authors of the R28 report quote the English statistician
George Box: “all models are wrong, but some are useful” and conclude that this is
very applicable to tailings dam failure studies.
4.2.3. Reporting of the results
A fundamental objective of the submerged area calculations is to provide

the authorities with the necessary elements for emergency planning. It is there-
fore important to communicate the results of the calculations as clearly as possible.
The most common practice is to provide maps of the flooded areas, hydrographs at
various characteristic points and tables giving the total calculated values. The maps
should be provided at a scale that allows the habitats to be distinguished, typically
1:25,000. The extent of the flooded areas should be shown, with at least the time
of arrival of the wave. The maximum water height and the maximum current speed
can also be indicated.
Some organisations go much further and propose the real-time calculation

of flooded areas according to the flows actually recorded or forecasted and the
provision of this information in real time to the authorities in charge of the application
of emergency plans (R13).
The USACE has developed a web-based GIS tool to assist decision-makers

in flooding and emergency planning. This tool, called rapid flood inundation map-
ping, can be used to communicate the mapping of inundated areas following dam
failures or large floods. The mapping data can be shared with emergency planning
authorities within minutes. The tool improves coordination between stakeholders and
adapts to their specific needs, thus facilitating communication between all concerned
actors during a crisis.
874
GR Q. 105
4.3. EVALUATION OF THE CONSEQUENCES
Impact assessment is the rationale for determining flood zones with two key
objectives:
◦ Define the risk category of the dam when it is based on the consequences;
◦ Develop emergency plans for civil protection agencies and operators.
The consequences on the environment and cultural heritage can be assessed

as well, but few reports mention them.
The assessment of the consequences on the populations is the major concern

during these studies. It can be noted that this is done with very diverse methodologies
and extremely variable levels of detail.
The simplest methods attempt to assess the Population at Risk (PAR). This
is done by counting, when practicable, all the houses and buildings located in the
flooded area from georeferenced plans. When the flooded area is very large, and
the dwellings cannot be individualised (case of large densely populated areas) the
practice is to use the results of the population census and to apply an occupancy
ratio. The distinction between day and night, weekdays and weekends, winter and
summer can allow a better evaluation of the number of people at risk. In some
countries this assessment of the number of people at risk is considered sufficient
because the only objective is to develop emergency plans. The uncertainties on the
characteristics of the flooded area are taken into account by lumped increases of
the water heights and the arrival time of the wave.
But in many other countries, the objective of consequence assessment is

to evaluate the number of potential casualties (LOL). The methods used start, as
before, with the evaluation of the persons subjected to the risk and then use more
or less complex methods to go from these “PAR” to the number of casualties.
The empirical methods of Graham (1999), RESCDAM (2001), or DEFRA

(2006) can be used to estimate casualties under flood conditions. Other advanced
methodologies that fully account for spatially distributed physical processes, human
responses, and the performance of various systems are available. This is the case for
example of Life-SIM which models the entire flooded area and performs a dynamic
simulation of the wave arrival for the different selected flood scenarios. To evaluate
the number of casualties, this method integrates many parameters such as the delay
with which the alert is given, the reaction time of the people and their movements by
car, the possible flooding of the roads or their congestion, etc. For people moving
on foot, the classical curves of stability of a person according to the speed and the
height of water are used. Finally, the damage curves allow the evaluation of the
resistance of the buildings for the people who would have taken refuge on the roofs.
In order to take into account the uncertainties, sensitivity studies are performed, for
875
GR Q. 105
example with Monte Carlo methods. The orders of magnitude of potential casualties
are then associated with a probability (R10).
However, no single method is appropriate or applicable to a given situation.

The recommended practice is to assess the number of potential casualties using a
variety of methods.
One can legitimately wonder about the reasons for such a difference in
approach between simple and complex methods. There is certainly an explana-
tion in the fact that in the regulations of some countries the classification of dams is
not made according to the consequences, i.e. according to the number of potential
casualties. On the other hand, these regulations often require increases in the evalu-
ation of the people subject to the risk. When, on the contrary, the legislation provides
for the classification of dams according to the consequences in terms of potential
casualties, this encourages the operators to be as precise as possible. The use of
F-N acceptability criteria also provides an incentive to assess the number of potential
casualties with relative precision. Another explanation is probably that, especially in
Europe, the potentially flooded areas are densely populated and that approaches
such as Life-Sim become very cumbersome to implement. Finally, the legal system
also plays a role: in countries where the legal system is based on the Civil Code,
the authorities are reluctant to explicitly display a potential number of casualties,
whereas this is less of a problem for countries governed by Common Law.
For dam failures during floods, the assessment of the increment of conse-
quences due to the dam failure compared to those of the natural flood is often
performed. However, there is no well-established legal framework for this approach,
which is still controversial, especially for the assessment of potential casualties. It is
less problematic for the assessment of economic consequences where the human
element does not need to be taken into account: the houses and other structures
that are located in the “differential” fringe between the areas inundated by the failure
wave compared to the natural flood are clearly identifiable and cannot be moved.
A survey of consequence assessment practices and objectives is underway

at the Committee on Dam Safety. The interim results of this survey were used in
part in writing this paragraph. The survey indicates that the number of dams that
have undergone consequence assessment, and subsequently emergency planning,
is significant, for example several hundred in Italy and France. The usefulness of
risk assessment in dam management, the importance of emergency planning and
communication, and the need for accurate topography are important aspects that
emerge from the responses to this survey.
But the results of this survey also indicate the difficulties encountered in these
impact assessment studies:
◦ Model discrepancies in casualty estimates and the difficulty of determining

the likely number of casualties for a large population at risk;
876
GR Q. 105
◦ The difficulty of setting parameters for alert dissemination and population

mobilization;
◦ Estimating consequential damages;
◦ The fact that evacuation may result in more casualties in some cases (e.g.
people trapped in their cars).
4.4. CONCLUSION
The evaluation of the characteristics of the dam-break flood wave and its con-
sequences is a widespread practice among dam professionals. The quality of the
results of these studies depends a lot on the quality of the data used in the calcula-
tion models, and in particular on the topography data characterizing the potentially
flooded areas. The most delicate point at present is the characterization of the breach
hydrograph for embankment dams. Ongoing research should improve the prediction
of this hydrograph in the future. In the meantime, it is prudent to choose increased
ranges of values and to test the sensitivity of the results to these variations of the
breaching hydrograph parameters.
Consequence assessment is also an area that has reached a certain maturity.

The differences in approach between simplified methods and “dynamic” models are
more related to national legislative contexts. However, it should be kept in mind
that these assessments are subject to many uncertainties, especially when they
model the behaviour of populations. Regular exercises of emergency plans are
certainly the most effective way to ensure that the assumptions adopted are credible.
To conclude on this point, it is interesting to note that there is obviously a cost

to these studies, but the cost of avoided damage is far greater. For example, the
USACE estimates that between 2004 and 2013, its flood damage reduction projects
prevented about $8 in damage for every $1 invested.
5. EMERGENCY PLANING
Topic c) of Question 105 deals with emergency plans and covers everything
that concerns regulations, organisation, information for the population and examples
of implementation. It therefore covers all the means and organisations necessary to
manage crisis situations. These subjects have been addressed by ICOLD in several
bulletins and at congresses. We can mention bulletins 154 and 175 that deal with
dam safety management, bulletin 180 “Lessons learned from dam incidents”, bulletin
168 “Recommendations for operation, maintenance and rehabilitation” and bulletin
59, which is quite old but whose recommendations for emergency plans are still
current.
877
GR Q. 105
Many other documents developed by government agencies or financial

institutions are also available, for example FEMA guides or recent World Bank
recommendations that refer to numerous references.
5.1. WHY EMERGENCY PLANS?
Despite all the rigour brought to the design, construction and operation of
dams, we cannot exclude that a serious incident occurs, the worst scenario being
the failure of the dam. All high-risk industries are confronted with this problem where
the probability of a serious incident is very low, but its consequences are often very
important.
In many regulations or technical standards, the design assumptions of a dam

depend on the consequences of a possible failure: the more important these con-
sequences are, the more severe the stress and resistance assumptions will be. In
the “risk analysis” approach, “prevention barriers” can thus be explicitly defined,
which are structural or non-structural provisions implemented to reduce the proba-
bility of accidental scenarios. The control of the efficiency of these barriers during
the whole life of the dam allows to verify that the safety level of the dam remains
satisfactory.
This consideration of the consequences of a failure in the design and oper-

ation of a dam is essential. But as a serious incident, even of very low probability,
cannot be excluded, it is also essential to try to reduce the consequences of such an
incident.
This is the purpose of emergency plans that address the following four key
points:
◦ Detecting anomalies and warning signs of failure as early as possible;

anticipation of events is a key factor in any emergency plan;
◦ Define the relationship between the dam operator and the authorities, the
responsibilities of each, the alert procedures, etc.
◦ Define by whom and how the public is informed and the modalities of
preventive evacuation.
◦ Define possible interventions on the dam or reservoir.
The means and organisation necessary to achieve this objective are detailed
below.
878
GR Q. 105
5.2. REGULATORY ASPECTS
5.2.1. General regulations and external emergency plan
Regulations in many countries require the development of an emergency plan

for dams based on their size and downstream consequences. The recent World
Bank study of practices in 51 countries provides very interesting information on how
these emergency plans are prescribed to dam owners. Of the 51 countries surveyed,
only 27 have legislation making emergency plans mandatory, while 16 others have
emergency plans on a voluntary basis. The ICOLD European Club maintains a
summary of the regulations of 17 countries with a chapter devoted to the protection
of the population. Several reports submitted for this question describe the regulatory
context of their country, for example in Portugal, Switzerland and Italy, and one article
deals with the more delicate problem of emergency plans in a cross-border context.
In many countries there are legislative and regulatory provisions for “general”
emergency situations in the framework of civil protection, for example in the case of
serious incidents at chemical plants or natural floods, etc. These legislative provi-
sions also concern the field of dams, whose specific regulations must be compatible
with these more general texts. These general texts foresee the elaboration of an
emergency plan that is qualified as external because it applies to entities and ser-
vices that do not depend hierarchically on the owner of the dam. These external
emergency plans are developed by government services, based on studies and
data provided by the owner of the facility concerned, in our case the dam. These
plans identify the actors and their responsibilities and how in an emergency situation
these different actors and entities communicate, consult and act in a coordinated
manner. In addition to the owner and operator of the dam, the regulator representing
the State and all the emergency services, fire departments, civil protection and local
communities are involved to varying degrees. These external emergency plans are
drawn up by the public authorities on the basis of data provided by the dam owner
and in consultation with all the players. An external emergency plan should include
the following elements:
◦ List of all the actors involved: dam operator, regulator, civil protection, police,
hospitals, etc. and the means of communication between them all;
◦ Reminder of the characteristics of the dam, the downstream zoning, the
different alert levels and the actions to be taken at each alert level
◦ Responsibility of each actor for critical decisions including declaration of alert
levels, public information, public evacuation
◦ Modalities of communication to the public in emergency situations;
◦ Evacuation of areas at risk of flooding;
◦ Relief Operations and Other Emergency Arrangements;
879
GR Q. 105
◦ Emergency Transportation;
◦ Access to remote sites;
◦ Regular review and crisis exercises via room or functional exercises, involving
all stakeholders and parties concerned;
◦ Keeping emergency services on standby;
◦ Informing the public about the contents of the emergency plan, in particular
the modalities of alert and evacuation.
In transboundary basins, emergency operations are coordinated by ripar-

ian government agencies. The compatibility of requirements and precautionary
measures taken by riparian agencies should be checked periodically.
5.2.2. Specific dam regulations
The general civil protection regulations mentioned above concern industrial

establishments of various kinds. Specific texts define the field of application and the
specific requirements for each type of industrial establishment.
Depending on the country, this specific regulation for dams is “self-supporting”.

It must then define all the aspects described above (actors, responsibilities, etc.).
But whether it is self-supporting or included in a more general body of legislation,
the specific regulation for dams contains the following requirements:
◦ Definition of dams subject to regulation according to criteria based either

on the consequences of a serious incident, or on the size of the dam and
reservoir, or a mixture of the two types of criteria. It is possible at this stage
to define categories of dams that will be subject to graduated requirements;
◦ Establish an emergency plan for each dam within a watershed or as part of
an operating scheme (operating company, power system, etc.), but also for
the scheme as a whole. Creation of a liaison committee with representatives
from all operating companies when there are multiple companies operating
dams in the watershed.
◦ Selection of accidental scenarios whose consequences justify their consider-
ation in an emergency plan. The scenarios are essentially the total or partial
failure of the dam after the critical water levels in the reservoir have been
exceeded during an extreme flood or following a malfunction of the gates,
failures following an earthquake, landslide or rock fall in the reservoir. Gate
failure is considered when the released flows are damaging. It should be
noted here that some regulations require that the scenarios of gate opening,
whether voluntary or not, be included in the emergency plan. This is more a
matter of public safety than an emergency situation.
◦ Consideration of emergency situations that may result from riots, terrorism
or war;
◦ Analysis of the case of chain failures when several dams are built on the
river;
880
GR Q. 105
◦ For each accidental scenario retained in the risk analysis, a study of the
consequences with the production of flood maps and the estimation of the
impacted populations taking into account the kinetics of the events;
◦ The definition of different levels of alert;
◦ The definition of criteria for determining the zoning downstream of the dam:
immediately downstream, the operator is responsible for alerting the pop-
ulation. Beyond that, it is usually the responsibility of the public authorities
within the framework of the external emergency plan;
◦ The content of the internal emergency plan to be developed by the owner
(see a sample plan in 5.3.1 below);
◦ The need for monitoring devices to detect anomalies of the dam behaviour
that could lead to a serious incident;
◦ The need for communication between the operator and the authorities
◦ The need for population warning devices in the downstream area near the
dam.
Development along rivers increases the risk and damage caused by a possible
dam accident. Where necessary, external and internal emergency plans are revised
at regular intervals to adapt to changes in the natural and socio-economic environ-
ment. This revision is usually done every 5 years in rapidly developing regions, and
every 10 years in all other cases.
5.3. OWNER’S ACTIONS AND INTERNAL EMERGENCY PLAN
These regulatory requirements result in the owner producing the necessary

elements for the development of the external emergency plan and the development
of its own internal emergency plan.
881
GR Q. 105
For each incident scenario (partial or total failure of the dam, failure of a gate,
etc.) the owner provides the authorities with the results of hydraulic simulations
giving the extent of the flooded areas, the propagation time of the flood wave and the
estimated number of people present in the flooded area. These elements are used
by the public authorities to define several zones downstream of the dam. There are
generally two zones, the first one close to the dam where the alert to the population
will be given by the owner with his own means (sirens or others), the second one
where the public authorities will be responsible to give the alert and to proceed to
the evacuation of the population.
5.3.1. Internal emergency plan
Internal emergency plans are then developed by the owner or entity respon-
sible for the dam. These plans typically address the following points:
◦ Hydrographic Monitoring and Flood Warning Devices;

◦ Dam monitoring and specific anomaly detection devices;
◦ For the selected failure modes, identification of warning signs and thresholds;
◦ Correspondence between the observed phenomena and the alert levels;
◦ Map of the downstream zones of the dam with the location of sirens or other
warning devices for the population;
◦ Internal and external crisis communication to emergency services;
◦ Instructions for local operating crews including loss of communication with
dam officials and public authorities;
◦ Establishment and operation of a crisis cell;
◦ Flood maps for the project flood and for the flood wave resulting from the
selected incidental scenarios;
◦ Procedures for emergency emptying of reservoir(s);
◦ Equipment, materials and support services available for rescue actions;
◦ Emergency operation of structures associated with the dam such as plants,
water production or irrigation systems, locks, etc.;
◦ Emergency repair opportunities at the dam site;
◦ Regular review and crisis exercises via room or functional exercises;
◦ Periodic testing of communication and warning devices.
In some countries these internal emergency plans must be formally approved

by the authorities, but these specific provisions vary from country to country.
5.3.2. Characteristics of emergency situations
The fundamental importance of anticipation cannot be stressed enough.

When a crisis occurs, it is always too late to organise the relationships between
the actors involved and to think about the actions to be taken. It is the objective
of the external emergency plan to define the relationships between the actors, the
882
GR Q. 105
communication mode, the decision-making chain and therefore the responsibilities

of one another. As in any industrial process, interfaces, whether technical or between
organisations, are always a weak point. How can we ensure that the messages sent
out are understood in the same way by everyone? How can we ensure that everyone
knows exactly who is responsible for making which decisions and who is responsi-
ble for implementing them? For this purpose, clear and understandable emergency
instructions are established and distributed to all the actors and operating teams.
These emergency instructions are easily accessible and define the authorities autho-
rised or obliged to give orders in an emergency. All persons required to act in an
emergency have a copy of these instructions in their offices and homes. Emergency
plans therefore identify all the actors, the role of each and the actions to be taken
according to the levels of emergency.
This notion of emergency levels, from three to five varies from country to
country and from owner to owner. This seems normal because civil protection agen-
cies and warning systems vary from one country to another, and this influences the
choice of the number and the criteria for defining these emergency levels. The devel-
opment of double entry tables is a practice that is becoming more widespread. This
allows to have a quick correspondence between the failure and the urgency level.
An example of such a table is given in R36 for Switzerland. A very detailed example
of this type of table can be found in Appendix C of [24]. Depending on the type of
failure, it is not easy to establish criteria for the level of emergency. Failures during
a flood event are the easiest to deal with: the forecasted inflows, the availability of
gates, the presence of ice jams, etc. are objective criteria from which it is possible to
define emergency levels. This definition is more delicate for failure modes affecting
embankment dams or foundations: the occurrence of a sediment laden leakage,
the rise in pressure in a piezometer, etc. are phenomena that require expertise
before deciding on their degree of danger and the level of urgency to be associated
with them. This expertise must be identified in the emergency plan as well as its
availability in case of emergency. The problem is quite similar during an earthquake.
Emergency Structural
levels Floods Earthquake failure Other
Level 1 Reaching Z1 Intensity > I1 Leakage- Landslide, etc.
pressure
increase
Abnormally high
values/ history
and pressure Expert opinion
unfavourable increase Expert
developments opinion
and structural pressure Expert opinion
failure increase
Expert opinion
883
GR Q. 105
The delays available before declaring an emergency level is crucial. These

delays are very useful for public authorities and civil protection agencies to prepare
the actions to be undertaken before the next emergency level is reached. For flood
situations it is usually possible to estimate correctly the time available before the
next emergency level is reached. For example, the time available before reaching
a certain level in the reservoir can calculated from the gradient of the water rise, or
the time to operate a gate on normal or backup supply and its consequences on
the discharge is usually known. This evaluation of time is much more delicate for
failure modes affecting embankment dams or foundations and requires the expertise
mentioned above.
The setting up of several crisis cells is a common practice in emergency situa-

tions. Some crises can last several days or even longer. In such cases, emergency
plans provide for the rotation of leaders and participants.
Anticipation also concerns access to the dam site. When the weather fore-
cast calls for heavy precipitation accompanied by high winds, there is a risk of not
being able to access the dam site to monitor it and carry out equipment tests and
preliminary manoeuvres. A good practice is to send the operating teams before the
arrival of the disturbances to prepare for any eventuality. During the 1999 storms in
Europe, for example, access roads to some dams were impassable for more than
48 hours.
In case of severe bad weather, the impossibility of any communication between

the dam and the crisis cells cannot be excluded. In these cases, corrective actions
are taken immediately and independently of all current administrative procedures.
The technical operating manager has sufficient authority to order urgent preventive
measures or reinforcements without having to ask for special authorisation from
his managers. This can go as far as the decision to activate the public warning
systems. The communications between the different crisis cells can also be affected,
the external emergency plans foresee this eventuality and describe the degraded
processes for decision making.
Anticipation obviously includes the forecasting of phenomena that may lead

to failures:
◦ For floods, this implies that the operator has permanent access to infor-
mation from meteorological centres. Some large operators have their own
forecasting teams, but most operators rely on regional or national agencies
with which they have contracts. During a crisis, monitoring the level of the
reservoir is essential. It is often measured by sensors, but the presence of a
water level gauge or rod remains an essential device;
◦ For earthquakes, most operators rely on specialised agencies with whom
they contract to be alerted. The calculation of the intensity on the site is
generally part of the tasks delegated to these agencies. Some dams are
equipped with seismic sensors whose objective is more often to have mea-
surements for dynamic analysis a posteriori. The strong motion sensors are
884
GR Q. 105
probably more efficient in case of earthquake. In all cases, the actions to

be taken are immediate or deferred inspections according to the intensity,
then the declaration of emergency levels according to the findings of these
inspections;
◦ Finally, all phenomena that could lead to structural or geotechnical failure are
covered by the monitoring of the dam through inspections, monitoring and
periodic testing of safety equipment. Many ICOLD bulletins are devoted to
these essential activities. Not all sensors in a monitoring system are equally
important in detecting a failure. Some sensors are intended for understanding
the long-term behaviour of the dam. Others are primarily intended for the
early detection of emergency situations. The important parameter is then
the periodicity of the reading of these sensors, which must be compatible
with the kinetics of a potential failure.
Despite the analyses made a priori during the elaboration of the emergency
plans, it cannot be excluded to have a type of failure not foreseen. The failures of
the Oroville spillways are a good example. The reactivity of the actors and their
expertise are then essential. This further argues for a crisis organization that is both
efficient and able to adapt to various situations.
The decision to evacuate the population is a major responsibility that falls

to the public authorities, except in a zone close to the dam where the alert can
only be given effectively by the operators. Many national regulations define a zoning
downstream of the dam, with at least two zones:
◦ A zone close to the dam where the civil protection warning means are non-
existent or insufficient in number to alert the population. The owner designs
and installs warning means in this zone, generally sirens.
◦ Downstream of this first zone, a zone that will be covered by the existing
means of alert in the general framework of civil protection.
The development of communication means in the last two decades has led
to the emergence of new tools such as large-scale telephone alerts or the sending
of SMS to all smartphones located in a given area. Some countries have common
alert protocols (CAP) allowing the exchange of information between various alert
systems and civil protection agencies.
The importance of the time between evacuation and dam failure has been the
subject of recent studies, see for example [22] and R40 of this question 105. The
USBR study cited in [22] indicates that the mortality rate is divided on average by
100 in areas where the product of water height “D” and current velocity “V” is of the
order of 100 in feet2 /s (about 9 m*m/s, i.e., areas of high impact of the flood wave).
The evacuation decision must take this factor into account while evaluating the risks
of an unnecessary evacuation if the dam does not fail.
885
GR Q. 105
5.3.3. Preparatory operational activities
Many guides exist for the deployment of internal emergency plan actions. Ref-
erence [22] lists many guides developed for this purpose. We can also mention the
CDA guide [25] (ACB in French) published in 2019. The essential operational activ-
ities are derived from what is indicated in 5.3.2. Some guidance on specific aspects
is given below:
◦ Dedicated surveillance room: it is usual to provide a surveillance room,

outside the flood zone and positioned so as to have a clear view of the
downstream face of the dam and the spillway(s). This room is equipped with
the means of communicating with the internal and external crisis cells and
the control panel for triggering the warning systems for the populations in the
nearby downstream zone. Communication tests with the authorities and tests
of the warning systems are carried out periodically from this room. The emer-
gency plans and essential documentation of the dam are available there.
Unfortunately, the risk of malicious intrusion into these surveillance rooms
cannot be ruled out and some operators have removed sensitive documents,
for example, the operating manuals for the opening of bottom outlets.
◦ The weather conditions associated with very heavy rainfall are often very
severe and can lead to power and communication outages, for example due
to falling trees on power lines or widespread damage to the transmission
system. For this reason, emergency generators and batteries are provided
at the dam sites to ensure the continued operation of spillways and essential
equipment such as water level gauges, etc. These power supplies are tested
regularly to ensure that they are functioning properly. Particular attention is
paid to the autonomy of the generators’ operation: the replenishment time of
the diesel tanks can be important depending on the condition of the access
roads to the dam site.
◦ Some operators carry stocks of material (for embankment dams) or safety
equipment (sandbags, etc.) and spare parts for safety systems.
◦ Dedicated room for the crisis cells: this room is equipped with secure trans-
mission means and the necessary office and computer equipment. The
emergency plans and the essential technical documentation concerning the
dam(s) must be consultable from this room, which is now relatively easy when
this documentation exists in a digital format. Testing of emergency plans is
done from this room to ensure that all documents are up to date. Most large
dam operators and civil protection agencies have such a crisis room which
can also be used for other events.
◦ The information of the populations concerned on the warning signals and
the places of refuge in case of evacuation is ensured by the distribution of
brochures or flyers, of which an example is given in R6
◦ Operating crews are trained in emergency operations and periodically
updated. Training programs and procedures are updated in accordance with
changes in emergency plans, and to reflect technological advances.
886
GR Q. 105
◦ The organisation of crisis exercises is generally the responsibility of the

authorities, with a periodicity defined in the external emergency plan. Opera-
tors test their internal organisation more frequently by simulating deficiencies
scenarios and activating crisis cells.
6. GOVERNANCE OF THE SAFETY
Topic d) of this question 105 is entitled “Safety governance: definition of

responsibilities, periodic audits, implementation testing and organisation of feed-
back”. Some of these topics have already been addressed in previous chapters,
and others are covered by recent ICOLD Bulletins, in particular Bulletins 154 and
175, which largely cover dam safety governance and management. In what fol-
lows, the focus is on the role of safety governance in relation to “dam incidents and
accidents”, the subject of this Question 105. For governments and public author-
ities, safety governance has two aspects: the development of regulations setting
out requirements - and the ability to monitor their implementation - and the organ-
isation of the collection of data, analysis and reporting of incidents affecting dams
in the country and more broadly at the international level. The recommendations
and technical guides, even if they are not mandatory, constitute a substantial con-
tribution to safety governance for the profession. For entities owning or operating
dams, safety governance involves the establishment of a safety management sys-
tem (SMS), which implies compliance with regulatory requirements, the allocation of
qualified human resources, internal control, etc. All of this is described in the ICOLD
bulletins mentioned above. As far as incidents are concerned, these requirements
also include the detection of incidents and near misses, their analysis and reporting
to public authorities.
It seems relevant to address this issue at several levels of governance,

international, national and at the level of the entities that own dams.
6.1. AT THE INTERNATIONAL LEVEL
There are no international organisations with regulatory prerogatives on dam

safety. ICOLD is the only supra-national organisation that issues recommenda-
tions covering the whole spectrum of technical and organisational activities related
to dams. These recommendations concern all actors: States and governments,
investors, financial institutions, designers, construction companies, suppliers, oper-
ators etc. As such, ICOLD can be considered as the guarantor of the principles of
dam safety from the technical and organisational points of view. The recent publi-
cation of the World Declaration on Dam Safety is a perfect illustration of ICOLD’s
role.
887
GR Q. 105
ICOLD recommendations specific to the incident management are included

in the previously mentioned Bulletins 154 and 175. Other organisations publish
recommendations for the establishment of emergency plans. This is the case of the
World Bank [24] which publishes a collection of good practices to be implemented in
emergency situations. The application of these good practices for projects financed
by the Bank is a step towards better governance of the safety.
Practical guides for building an emergency plan and implementing the nec-
essary means have been developed by government agencies (e.g. FEMA [26]) or
professional associations (e.g. CDA [25]). These guides are available and often
serve as references beyond the countries where they were developed.
But in addition to these recommendations and guidelines, an important task

of ICOLD is the collection of data concerning dam incidents. Several bulletins or
publications have been produced on this subject, as already mentioned in 3.1. This
work of collecting incident data and making them available to the dam community
constitutes feedback that is used to improve design and operation practices, thus
contributing to the continuous improvement of safety. Moreover, the reference to
accidents is often explicitly present in ICOLD technical bulletins to explain the rea-
sons for the recommendations. The next step will be to put this information online
in the form of a database that can be consulted from the ICOLD website. Other
databases covering incidents beyond one country exist but are not generally in the
public domain and quite often thematic, thus not exhaustive.
To date, ICOLD has not made any recommendations on the analysis of inci-
dents. Analysis guides following a failure are available, for example that of FEMA [19]
and are very useful in these dramatic circumstances. To the best of the reporter’s
knowledge, there is no equivalent to this guide for the analysis of near misses, the
lessons of which would be very useful to the profession. On the other hand, it has
been noted that, in the event of a failure, analyses are often oriented towards the
technical causes, whereas it is recognised that the root causes, of an organisa-
tional and human nature, are predominant. This is what emerges from the excellent
report on the Oroville spillway accident in which the importance of these root causes
is emphasised. ICOLD could draw inspiration from the practices of other high-risk
industries such as the electro-nuclear energy production or chemical industries,
which have developed guidelines and have a long practice of in-depth analysis of
incidents.
6.2. AT THE COUNTRY LEVEL
At the national level, the governance of dam safety is carried out by two entities:
government departments or agencies responsible for regulation - and professional
associations and more specifically, ICOLD national committees.
888
GR Q. 105
6.2.1. Regulations at the national level
Paragraph 5.2.1 deals with this subject and the reader is encouraged to refer
to it. It can be recalled that governments are concerned with the management of
emergencies of all kinds, natural, industrial, malicious attacks, and establish regu-
lations to create and organize the civil protection organisation: fire department, civil
protection agencies, warning system, etc. These regulations specify the detection
methods, the processes for activating the civil protection entities and the means of
informing, alerting and evacuating the population with regard to these risks.
Specific regulations complete this general framework for hazardous industries.

They contain the requirement for external and internal emergency plans and specify
the different actors in crisis management and their responsibilities. Depending on
the industries concerned, they define the criteria for triggering alerts and the zones
for informing and evacuating the population. For dams, this concerns the definition
of the different downstream zones and the different alert levels. These regulations
may also specify particular means of communication between the actors and for
warning the population. Finally, they provide for the periodic testing of the system
and for prior information to the populations concerned.
Many countries require operators to report incidents on their dams. These

reports feed into databases and allow for global analyses of the level of safety, see
3.2.3. A good practice is to make this database available to the dam community and
to periodically disseminate a summary. The widest possible dissemination of this
information is beneficial for the acculturation of all actors.
Government oversight of compliance with regulatory requirements is an essen-

tial component of good safety governance. This rule obviously applies to the
regulatory corpus relating to emergency situations. Carrying out crisis exercises is
a requirement that concerns all actors, sometimes including the population. These
exercises are very important because they allow to verify that the recommendations
of the emergency plans “work”. They also have the advantage of making all the
players aware of the potential risks and training them to react appropriately. It is an
illusion to believe that emergency plans that have never been tested will be effective
in a real emergency situation. It is true that time is spent organising and conducting
these exercises, but this is disproportionate to the potential gains.
6.2.2. Professional associations
Many ICOLD national committees have task forces to develop guidelines and
recommendations. However, to the best of the reporter’s knowledge, the subject of
emergency situations is not often addressed. Each operator is often left to organise
his teams, choose an alert system, communication system, etc.
889
GR Q. 105
However, there are examples of national committees that have set up working
groups to produce incident summaries. These working groups bring together several
operators and representatives of the public authorities.
6.3. AT THE INDIVIDUAL OWNER LEVEL
The subject of safety governance is extensively covered in ICOLD bulletins.

For emergency-specific aspects, the activities that are the responsibility of the owner
or operator have been developed in the previous chapters and it is up to them to
implement them as part of their Safety Management System (SMS). Some addi-
tional information is provided in the following paragraph. Finally, the leadership of
managers and executives is a fundamental aspect of safety governance because
it contributes to the emergence of a safety culture in organisations. This point is
discussed in the second paragraph.
6.3.1. Governance of activities
The importance of detecting the early signs of incidents by careful monitor-

ing and surveillance can be recalled here. Bulletin 180 [10] presents several cases
where early detection prevented the development of an incident situation. An impor-
tant parameter is the time between each measurement or inspection. This delay
must be defined according to the kinematics of the failure modes. It is therefore up
to management to put in place the structure and skills necessary for this monitoring
to be effective.
The detection and analysis of incidents and near misses contributes to the con-
tinuous improvement of the process. A performance indicator based on the number
of incidents and near misses detected and analysed per year is very relevant to
verify the vigilance of the teams and the level of safety achieved. It is expected
that the total number of detected events, including all near misses, will remain
almost constant, but that the number of incidents with consequences will continue to
decrease. In-depth analysis of the most significant near misses provides elements
for improvement and helps prevent similar incidents from occurring elsewhere in the
organisation. This information is distributed to all teams, who, under the authority of
their managers, review and discuss it and check that the root and technical causes
of incidents are not present in their operations.
Large companies can also set up task forces to collect the most significant
incidents at the international level and to analyse the possibility of such incidents
occurring in their organization. For example, after the Taum Sauk dam failure, owners
of pump storage hydroelectric power generation facilities tested the resilience of their
facilities to the risk of over-pumping.
890
GR Q. 105
Finally, it is up to management to commission internal and external audits of

targeted activities. Anything that contributes to incident prevention is often a subject
for internal audits, and that is fine. On the other hand, the practice of using inde-
pendent panels of experts is widespread for dam design activities, but much less so
for operation activities. However, it is an aid to governance for managers and field
staff, providing them with recommendations on possible organisational or technical
weaknesses.
6.3.2. Leadership and safety culture
Leadership and safety culture are discussed in Chapter 2.5 of Bulletin 154.
The concept of a “safety culture” is implicit in the design and operation of dams but
needs to be further developed. It is particularly important in emergency situations.
Safety culture can be defined as follows:
“Safety Culture is the set of characteristics and behaviours that, at the collec-
tive and individual level, ensure that issues related to Dam Safety receive the priority
attention they deserve because of their importance...”
These notions of “characteristics” and “behaviours” are not recommendations

or guidelines but define a state of mind at all levels of organisations, from the
technician on the dam to the senior management and board of directors.
The practices of other high-risk industries can be a source of inspiration for

improving leadership and safety culture. The OECD guidance report [23] was devel-
oped for the chemical industry but can be applied directly to any industrial activity
with significant risks to the public and investors. This document is aimed at senior
managers to make them aware of their major role in the safety of their activities
and offers them a self-assessment questionnaire to evaluate their safety practices
in their companies. The importance of their duty to set an example is emphasised.
The report describes the characteristics of “high-reliability organisations”

that can provide relatively error-free production for long periods of time. “These
organisations have the following essential characteristics:
◦ They have a chronic sense of questioning, in that they are never compla-
cent. For example, they don’t dismiss the possibility of an imminent accident
because they haven’t had one in a very long time;
◦ They react strongly to the slightest warning signal: their intervention threshold
is very low. ... ”
Numerous executive testimonials illustrate this report, for example:
891
GR Q. 105
“Lessons learned from past incidents show that strong safety leadership is
critical to disaster prevention, and that these lessons must be learned and applied
across all sectors to prevent the same mistakes from reproducing future accidents.”
And again:
“To ensure process safety, it is absolutely essential to develop a culture in

which all employees expect to deal with the unexpected and make every effort to
work without making mistakes. This type of culture is only possible with leadership
at all levels of the organization (Bob Hansen, Dow CEO).”
On a practical level, the recommended good practices focus on five areas in

which CEOs and executives should consider [23]:
◦ Leadership and corporate culture: create an open and transparent environ-

ment for employees to ask questions, be present in their company and show
their unwavering interest in safety;
◦ Knowledge of risks: overall knowledge of vulnerabilities and risks involved,
the different levels of protection. Be responsible for the establishment of
emergency plans.
◦ Information: ensure that incident and near miss data is thoroughly anal-
ysed. Commission internal audits and be involved in the analysis and
implementation of their recommendations;
◦ Competencies: ensure that the organisation has the competencies at all hier-
archical levels, devote the necessary time and resources to the analysis of
hazards by specialists, trust the competencies of their staff without neglecting
to call upon external experts if necessary.
◦ Action: coordinate and drive safety management programs.
These five areas are the subject of a self-assessment questionnaire, the

objective of which is to detect any weaknesses and take action to improve them
if necessary.
7. CONCLUSION
The 2021 ICOLD Season Greetings can be used to conclude this report.
“There is no secret to success: it is the result of preparation, hard work and

learning from failures”.
This sentence summarises the themes addressed in this Question 105 and
their importance to dam safety: incident collection and analysis to “learn from fail-
ures,” “preparedness” through emergency planning and consequence assessment,
and “hard work” of which safety governance is one aspect.
892
GR Q. 105
8. ACKNOWLEDGEMENTS
The reporter warmly thanks all the members of the working group formed
within the Technical Committee on Dam Safety for their contribution to the analysis
of the incident data and their suggestions during the preparation of Bulletin 188.
Thanks also to the Chair of this committee, Andy Zielinski, who has supported and
encouraged this work for several years. The reporter would like to thank Russell
Gunn of the Swiss Federal Office of Energy for proofreading the English text.
Finally, I would like to thank my company “Electricité de France” that gave me

the time and the means to build an ICOLD database of dam incidents and to all my
colleagues from whom I learned so much.
REFERENCES
[1] ICOLD World Declaration on Dam Safety.
[2] CIGB-ICOLD Lessons from dam incidents, Leçons tirées des accidents de
barrages - 1974.
[3] CIGB-ICOLD Deterioration of Dams and Reservoirs – Examples and their

analysis – Détérioration de barrages et réservoirs – Recueil de cas et analyse,
1983.
[4] CIGB-ICOLD Bulletin 99: Statistical analysis of dam failure, Ruptures de

barrages, analyse statistique - 1995.
[5] CIGB-ICOLD Bulletin 111: Dam Break Flood Analysis, Review and recom-
mandations - Etude de l’onde de rupture d’un barrage, Synthèse et recom-
mandations, 1998.
[6] CIGB-ICOLD Bulletin 120: Design feature of dams to resist seismic ground
motion - Aspects de la conception parasismique des barrages, 2001.
[7] CIGB-ICOLD Bulletin 154: Dam safety management: Operational phase of the
dam life cycle, 2017.
[8] CIGB-ICOLD Bulletin 164: Internal erosion of existing dams - Erosion interne
dans les barrages existants, 2017.
[9] CIGB-ICOLD Bulletin 175: Dam Safety Management: Pre- operational Phases
of the Dam Life Cycle, 2021.
[10] CIGB-ICOLD Bulletin 180: Lessons learnt from case histories, preprint.
893
GR Q. 105
[11] CIGB-ICOLD Bulletin 188: Statistical analysis of dam failure, preprint.
[12] CIGB-ICOLD congress 13 New Delhi 1979: Q49 Deterioration or failure of

dams.
[13] CIGB-ICOLD Congress 14 Rio de Janeiro 1982: Q52 Safety of dams in

operation General Report.
[14] CIGB-ICOLD Congress 19 Firenze 1997: Q75 Incidents and failures of dams
General report.
[15] CIGB-ICOLD Congress 23 Brasilia 2009: Q91 Dam Safety Management,

General report.
[16] CIGB-ICOLD Congress 24 Kyoto 2012: Q93 Safety, General report.
[17] DEFRA Guide to risk assessment for reservoir safety management; Volume 2:
Methodology and supporting information; Report – SC090001/R2, 2013.
[18] Q101 C1: A data base for dam incidents listed by ICOLD, Une base de données
des incidents de barrages répertoriées par la CIGB – M. de Vivo, P. Le Delliou,
F. Lempérière, M. Poupart, 26th ICOLD Congress Vienna.
[19] Dam Failure Investigation Guideline (2017) completed by the ASDSO Dam
[21] Justin Larouzée: Théorie et pratique des modèles d’erreur humaine dans la
sécurité des barrages hydroélectriques: le cas des études de dangers de bar-
rages d’EDF. Thèse de l’Ecole Nationale Supérieure des Mines de Paris, 2015.
Français. NNT: 2015ENMP0067.
[22] Wishart, M.J., Ueda, S., Pisaniello, J.D., Tingey-Holyoak, J.L., Lyon, K.N.,
and Boj García, E. (2020) “Laying the Foundations: A Global Analysis of Reg-
ulatory Frameworks for the Safety of Dams and Downstream Communities.”
Washington, D.C.; World Bank Group.
[23] OECD (2017) Corporate Governance for Process Safety – OECD Guidance
for Senior Leaders in High Hazard Industries.
[24] World Bank. 2021. “Good Practice Note on Dam Safety – Appendix 4:
Emergency Preparedness Plan (Sample Framework).” World Bank, Wash-
ington, DC.
[25] Emergency Management for Dam Safety. CDA, Toronto, Ontario. Central
Water Commission. 2019. Guidelines for Assessing and Managing Risks
Associated with Dams.
[26] Federal Guidelines for Dam Safety: Emergency Action Planning for Dams
(FEMA 2013).
894
GR Q. 105
-------
-------
QUESTION 105

M. POUPART
Independent Consultant Dam Safety
FRANCE
General Reporter/Rapporteur Général
French text: original from the author Texte français: texte original de l’auteur
English text: translation by the author, Texte anglais: traduction par l’auteur,
proofreading by Russel GUNN relecture par Russel GUNN
(OFEN, Switzerland) (OFEN, Suisse)
DOI: 10.1201/9781003211808-98 895

GR Q. 105
Table des matières
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898
2. RAPPORTS SOUMIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .898
3. ENSEIGNEMENTS RÉCENTS TIRÉS DES INCIDENTS
ET ACCIDENTS DE BARRAGES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
3.1 Publications disponibles à la CIGB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
3.2 Autres données disponibles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 900
3.2.1. Livres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 900
3.2.2. Listes d’incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
3.2.3. Rapports annuels par pays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
3.2.4. Rapports d’incident . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 902
3.2.5. Autres sources d’information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 902
3.3 Typologie des données . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 903
3.4 Exploitation des données. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .903
3.4.1. Liste d’incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 903
3.4.2. Rapports d’incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 904
3.5 exploitation des données statistiques récentes . . . . . . . . . . . . . . . . . 905
3.6 exemples d’exploitation des données sur quelques
cas récents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 907
3.7 Comment s’améliorer ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 913
3.7.1. Progrès possibles dans la collecte des données . . . . . . . . . 914
3.7.2. Progrès possibles dans l’exploitation des données . . . . . . . 914
3.7.3. Progrès possibles dans la diffusion de l’information . . . . . . 915
3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916
4. EVALUATION DES DÉBITS ET DE L’INONDATION . . . . . . . . . . . . . . . . . 917
4.1 Hydrogramme de brèche . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .918
4.1.1. Approches simplifiées . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 919
4.1.2. Modélisation couplant écoulement et érosion . . . . . . . . . . . 920
4.2 Evaluation des zones inondées . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 920
4.2.1. Méthodes simplifiées . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 921
4.2.2. Calculs de l’onde par modèle numérique . . . . . . . . . . . . . . . . 921
4.2.3. Restitution des résultats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 922
4.3 Evaluation des conséquences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 923
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 925
5. PLANS D’URGENCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926
5.1 Pourquoi des plans d’urgence ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926
5.2 Aspects réglementaires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .927
5.2.1. Réglementation générale et plans d’urgence externe . . . . 927
5.2.2. Réglementation spécifique barrage . . . . . . . . . . . . . . . . . . . . . 928
5.3 Actions du propriétaire et plan d’urgence interne . . . . . . . . . . . . . . . 930
5.3.1. Plan d’urgence interne . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .931
5.3.2. Caractéristiques des situations d’urgence . . . . . . . . . . . . . . . 931
5.3.3. Activités opérationnelles préparatoires . . . . . . . . . . . . . . . . . . 935
6. GOUVERNANCE DE LA SÉCURITÉ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936
6.1 Au niveau international . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936
6.2 Au niveau des Pays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 938
6.2.1. Réglementations au niveau national . . . . . . . . . . . . . . . . . . . . 938
6.2.2. Les associations professionnelles . . . . . . . . . . . . . . . . . . . . . . .939
896
GR Q. 105
Table des matières
6.3 Au niveau de chaque propriétaire 939

6.3.1. Gouvernance des activités . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 939
6.3.2. Leadership et culture de sécurité . . . . . . . . . . . . . . . . . . . . . . . . 940
7. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 942
REMERCIEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 942
RÉFÉRENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 943
897
GR Q. 105
1. INTRODUCTION
Dans toutes les industries à risques, l’analyse des incidents est un outil
important pour améliorer la sécurité. Comprendre les causes de l’incident permet
d’améliorer ce qui a été identifié comme une faiblesse, que ce soit dans la concep-
tion, la construction ou l’exploitation d’une installation industrielle. Les barrages ne
font pas exception à cette règle, et c’est la raison pour laquelle la CIGB accorde
depuis toujours une grande importance à la collecte et l’exploitation des incidents
affectant les barrages.
Une situation d’urgence ne peut jamais être écartée et il convient de s’y

préparer. Il faut pour cela envisager les scénarios accidentels et évaluer leurs con-
séquences par un calcul des ondes de ruptures. Les plans d’urgence intègrent ces
résultats et définissent les rôles et responsabilités de tous les protagonistes pour
la détection des situations anormales, les moyens d’alerte et, en cas d’urgence
extrême, les évacuations des populations en aval.
La sécurité des barrages est un enjeu majeur à plusieurs titres: pour la sécu-
rité publique des populations en aval, pour la continuité des services assurés par le
réservoir (énergie, irrigation, etc.) et enfin pour préserver un patrimoine très impor-
tant pour le propriétaire. Pour toutes ces raisons, la sécurité des barrages doit être
centrale dans les préoccupations de tous les acteurs concernés, et en priorité les
gouvernements et les propriétaires. Une gouvernance efficace et rigoureuse est un
outil adapté pour répondre à ces enjeux. Elle favorise l’émergence d’une culture de
la sécurité qui, au-delà des normes et règles, insuffle dans les organisations un état
d’esprit orienté sécurité au niveau individuel et collectif.
2. RAPPORTS SOUMIS
51 rapports ont été soumis pour la Question 105 dont 42 peuvent être rattachés
à l’un des 4 thèmes proposés.
Environ la moitié des rapports concernent le thème a) Enseignements

récents tirés des incidents et accidents de barrages que l’on peut répartir suivant
plusieurs catégories. 9 articles décrivent des ruptures de barrages ou de batardeaux
et expliquent le déroulement de l’accident; certains d’entre eux commentent les
enseignements à en tirer. 6 autres articles traitent d’accidents n’ayant pas dégénéré
en rupture et il est intéressant de comprendre quelles mesures ont été prises pour
éviter la rupture. Enfin 4 articles décrivent de manière plus générale les enseigne-
ments que l’on peut tirer des incidents de barrages et les organisations favorables
à la transmission de ces enseignements aux jeunes générations d’ingénieurs.
Le thème b) sur l’évaluation des débits et de l’inondation est traité par 7

rapports abordant l’ensemble des thématiques. Plusieurs rapports extrêmement
898
GR Q. 105
intéressants présentent une synthèse globale des pratiques (R10) ou l’état des lieux
d’un problème spécifique comme l’hydrogramme de brèche pour les barrages en
remblai (R19). La mise en réseau des informations pendant une crue et les capac-
ités de calcul permettent maintenant de partager entre toutes les parties prenantes,
en temps réel, l’étendue des conséquences d’une crue évaluée à partir des états
réels d’écoulement et de leurs prévisions (R11).
Le thème c) sur la réglementation et l’organisation des plans d’urgence est

traité par 7 rapports dont plusieurs présentent les réglementations dans leurs pays
respectifs et le retour d’expérience de leur mise en œuvre. Un rapport traite la
problématique des plans d’urgence transfrontaliers.
Le thème d) sur la gouvernance de la sécurité fait l’objet de 7 rapports

dont deux seulement traitent spécifiquement les aspects propres aux situations
d’urgence. Les autres rapports décrivent des organisations de la sécurité ou des
pratiques de revues et d’analyse de risques et couvre un périmètre plus vaste.
3. ENSEIGNEMENTS RÉCENTS TIRÉS DES INCIDENTS ET ACCIDENTS

DE BARRAGES
Il est important de rappeler en préliminaire la définition du terme incident utilisé

à la CIGB: un incident de barrage peut être une rupture ou un accident:
• Une rupture est définie par un relâchement incontrôlé de l’eau de la retenue

ou par la perte d’intégrité ne permettant plus au barrage d’assurer sa fonction
en toute sécurité. Un exemple est la rupture du barrage de San Fernando. À
la suite d’une rupture il est parfois possible de réparer le barrage, sinon il est
définitivement abandonné.
• Un accident est un incident que l’on a empêché de dégénérer en rupture en
prenant des mesures correctives immédiates mettant l’ouvrage en sécurité.
Un exemple de mesure corrective est la baisse partielle ou totale du niveau
du réservoir.
Dans ce chapitre consacré au thème a) de la Question 105 les paragraphes 3.1

et 3.2 décrivent les données disponibles sur les incidents de barrages, le paragraphe
3.3 en propose une typologie. Les paragraphes 3.4 à 3.6 montrent en quoi les
analyses de ces données permettent de tirer des enseignements et le paragraphe
3.7 propose des pistes d’amélioration.
3.1. PUBLICATIONS DISPONIBLES À LA CIGB
La CIGB a lancé trois enquêtes auprès de tous ses membres pour recueil-
lir la plus grande quantité d’informations possibles sur les incidents affectant les
barrages. Ces enquêtes ont donné lieu à trois publications: Lessons from Dams
899
GR Q. 105
Accidents en 1974, Détérioration des barrages et réservoirs en 1983 et le

bulletin 99 Ruptures de barrages - Analyse statistique en 1995. Le bulletin 188
a mis à jour ce bulletin 99 en 2019 pour intégrer les ruptures et incidents qui se sont
produits depuis 1991 jusqu’à 2019. A côté de ces publications consacrées exclu-
sivement aux incidents de barrages, plusieurs bulletins ont mentionné des incidents
de barrages pour des conditions de chargement ou des processus de dégradation
spécifiques, en particulier le bulletin 82 (choix de la crue de projet), le bulletin 109
(barrages de moins de 30 m de hauteur), le bulletin 120 (aspects de la concep-
tion parasismique des barrages), le bulletin 164 (Erosion interne dans les barrages
existants) et le récent bulletin 180 (Lessons learnt from case histories). Plusieurs
questions de congrès CIGB ont également traité ce sujet de la détérioration des
barrages et des incidents Q49 en 1973, Q52 en 1982, Q71 en 1994 et Q75 en
1997. On peut noter que cela fait donc 24 ans que ce sujet n’a pas été traité lors
d’un congrès de la CIGB.
Les éléments disponibles dans les bulletins, les articles de congrès et les
rapports généraux des questions citées ci-dessus constituent une base très riche
d’informations brutes , qu’il faut cependant exploiter pour en tirer tout le bénéfice.
3.2. AUTRES DONNÉES DISPONIBLES
De nombreux documents ou publications existent à côté de ceux de la CIGB,

qui peuvent être regroupés essentiellement en deux catégories. On trouve ainsi
quelques livres publiés essentiellement aux Etats Unis d’Amérique qui dressent
une synthèse des incidents survenus soit au niveau national, soit au niveau inter-
national. Les autres documents, plus nombreux, sont d’une part des rapports
périodiques sur les incidents, en général annuel, publiés par certains pays, et
d’autre part des rapports concernant la rupture d’un barrage en particulier. Pour
être complet il faut mentionner des articles publiés lors de conférences hors CIGB
et les informations diffusées par Internet. Il est évident que le degré de fiabilité de
ces différentes sources d’information est très variable. On examine ci-dessous les
différentes catégories avec quelques exemples.
3.2.1. Livres
Les livres les plus connus sont les ouvrages édités par l’ASCE sur les inci-
dents survenus sur des barrages des Etats Unis: Lessons from dam incidents
USA I édités en 1975, et Lessons from dam incidents USA II édité en 1988.
Le US Department of the Interior a publié en 1980 Dams and Public Safety
qui comprend une description détaillée de 41 ruptures au niveau mondial et des
enseignements à en tirer. On peut citer également la publication faite par Envi-
ronment Agency du Royaume Uni en 2011 Delivering benefits through evidence
- Lessons from historical dam incidents . Ces documents ont un format similaire
900
GR Q. 105
à celui de Lessons from dam incidents de la CIGB, avec une présentation

synthétique des caractéristiques des barrages et des incidents.
3.2.2. Listes d’incidents
A côté de ces documents édités sous forme de livres accessibles à tous il y a

dans de très nombreux pays des publications de listes récapitulatives des incidents
survenus sur les barrages de leurs pays. Les formats peuvent être divers mais ces
listes se présentent le plus souvent comme une table énumérant les barrages et
quelques paramètres caractérisant le barrage et l’incident. Ces documents ne sont
pas en général directement accessibles au public mais peuvent souvent être obtenus
par les réseaux professionnels, dont la CIGB. On peut citer dans cette catégorie de
documents les listes développées par l’Afrique du Sud, l’Allemagne, l’Australie, la
Finlande, la Slovénie, etc.
3.2.3. Rapports annuels par pays
Une autre source d’information se trouve dans les rapports annuels d’incidents
publiés par certains pays et dont certains sont accessibles au public et en général
consultables sur Internet. Cette pratique de rapports annuels est de plus en plus
répandue, en partie sans doute parce que la réglementation de nombreux pays
impose désormais la déclaration des incidents vers l’entité de contrôle.
Ainsi, l’ASDSO publie annuellement la liste des incidents survenus aux Etats
Unis, Environment Agency au Royaume Uni édite chaque année une liste sim-
ilaire, Post incident reporting for reservoirs . Il y a beaucoup d’autres exemples
de ces rapports annuels plus ou moins faciles à trouver et à consulter. Ces informa-
tions sont parfois consultables sur des bases de données. Il est intéressant de noter
que c’est presque toujours l’autorité de contrôle qui diffuse ces rapports, ce qui est
logique puisque c’est elle qui détient l’ensemble des informations remontant des
gestionnaires des barrages. A côté de ces rapports annuels, certaines autorités de
contrôle ou d’autres organismes étatiques ou de recherche ont constitué des bases
de données accessibles par Internet où on peut retrouver l’ensemble des incidents
déclarés par les gestionnaires. C’est le cas par exemple de la base NPDP (National
Performance of Dam Program) publiée par l’Université de Stanford aux USA ou de
la base ARIA du ministère de l’Environnement en France.
Le projet RESBA est un exemple de coopération internationale d’inventaire

d’incidents ayant affecté des petits barrages en remblai à la frontière franco-italienne.
Il a débouché sur une base de données binationale ce qui est sans doute
une première. On y trouve les caractéristiques géométriques, géomécaniques, les
aléas, les enjeux et les incidents. Cette base contient actuellement 26 barrages et
30 incidents recensés.
901
GR Q. 105
Les rapports d’analyse des accidents et presque accidents s’adressent plutôt

à des intervenants experts et traitent le contexte de l’incident, la description et
les origines des modes et mécanismes de rupture, les conséquences et les mesures
correctives.
3.2.4. Rapports d’incident
Une source extrêmement importante d’information se trouve dans les rapports

d’analyse élaborés par des experts après un incident notable, en général une rup-
ture. Dans la plupart des cas, ces analyses font l’objet d’articles présentés lors de
conférences professionnelles: CIGB, HydroPower and Dams, SwedCold, DSIG, etc.
Le format de ce type de document, de l’ordre de 10 à 20 pages, et souvent beaucoup
plus, permet de donner beaucoup plus d’informations que dans des listes ou des
descriptions synthétiques, un très bon exemple étant l’article publié à ce congrès
sur la rupture du barrage de col D du projet Xe-Pian Xe-Namnoy au Laos (R26). Il
existe de très nombreux articles décrivant des ruptures de barrages plus ou moins
faciles d’accès surtout pour les ruptures anciennes.
Les analyses d’incidents faisant l’objet d’un rapport d’experts sont de loin les
sources les plus riches d’information. Certains pays donnent accès à ces documents
via Internet, ce qui représente une opportunité inestimable pour la communauté
internationale pour comprendre les causes des incidents. Les Etats Unis d’Amérique
sont de ce point de vue exemplaire: on peut citer parmi d’autres les rapports des
ruptures du réservoir supérieur de Taum Sauk, du barrage de Delhi et plus récem-
ment le grave incident du déversoir du barrage d’Oroville. D’autres exemples sont
les rapports de la rupture du barrage de Camara au Brésil (en portugais) et celui du
barrage de Toddbrook au Royaume Uni. Ces rapports décrivent de manière détail-
lée les caractéristiques de la structure, les investigations réalisées et l’analyse des
causes et des conséquences.
3.2.5. Autres sources d’information
Enfin l’essor des technologies de l’information donne accès à de très nom-

breuses informations sur les incidents de barrages. Il s’agit essentiellement d’articles
de journaux en ligne qui ne donnent pratiquement jamais d’informations techniques.
Leur seul intérêt est d’alerter sur la survenue d’un incident et sur ses conséquences.
Cela a au moins le mérite d’attirer l’attention de la communauté internationale sur
l’incident et de chercher à en savoir davantage.
902
GR Q. 105
3.3. TYPOLOGIE DES DONNÉES
En préambule il est important de noter que les documents cités dans les para-
graphes précédents proviennent pour une grande majorité de pays anglophones.
Il est certain que des publications dans d’autres langues existent mais sont plus
difficiles à identifier. Une recherche sur Internet en espagnol ou en allemand, par
exemple, donne pourtant des résultats intéressants, encore faut-il pouvoir les lire.
Ceci constitue un biais important et la communauté des barrages devrait trouver les
moyens pour limiter autant que possible cette lacune.
On peut finalement catégoriser les documents traitant des incidents de

barrages en deux types: les listes d’incidents et les rapports d’incidents.
• Les listes sont en général assez exhaustives mais ne donnent pas beaucoup
de détails sur le barrage et l’incident. Leur intérêt est de fournir une base à
des études statistiques, comme la mise à jour du bulletin 99 par exemple. Ces
listes existent au niveau international essentiellement à la CIGB, et au niveau
de chaque pays. Les listes nationales sont en général plus détaillées que
les listes au niveau international et comprennent les ruptures mais aussi les
accidents. Constituer une liste internationale à partir de toutes ces données
est un exercice difficile car le même barrage peut avoir plusieurs noms et, pire,
les données entre ces listes sont parfois très différentes.
• Les rapports d’analyse concernent un seul barrage et donnent des informations
sur le barrage, l’incident, ses causes et ses conséquences, avec un niveau de
détail variable suivant les rapports. C’est à partir de ces analyses détaillées
que les principaux enseignements des ruptures sont obtenus.
3.4. EXPLOITATION DES DONNÉES
Disposer d’information sur les incidents de barrages est indispensable, mais

ne sera pas très utile si une exploitation pertinente n’en est pas faite. Cette exploita-
tion est évidemment différente suivant qu’il s’agisse de listes d’incidents ou de
rapports d’analyse.
3.4.1. Liste d’incidents
Pour les listes d’incidents l’analyse statistique est la méthode d’exploitation la

plus appropriée. Les données doivent être disponibles sous forme numérique et les
paramètres décrivant chaque incident sont des valeurs numériques ou des codes.
Les analyses statistiques sont utiles pour comprendre le tableau d’ensemble

mais ne permettent pas de faire évoluer les pratiques techniques. Elles peuvent
903
GR Q. 105
donner des tendances en classant les ruptures suivant les types de barrages, les
évolutions des ruptures dans le temps, les contextes les plus fréquents dans lesquels
les ruptures se produisent, les modes de rupture ainsi que les causes. La mise à
jour du bulletin 99 (voir 3.5) est un bon exemple des bénéfices que l’on peut tirer de
ce type d’analyse.
Ce recueil de données et leurs analyses témoignent de l’effort de la profession

d’avoir le meilleur inventaire possible des ruptures et des accidents. Cependant il
faut avoir à l’esprit que l’on ne peut pas vouloir à la fois l’exhaustivité des ruptures
et des analyses détaillées.
L’analyse statistique des ruptures devrait être faite au niveau de la CIGB et

celle des accidents au niveau des Pays et/ou des Etats, ces analyses étant rendues
publiques.
3.4.2. Rapports d’incidents
L’exploitation des rapports d’incidents fournit d’abord des enseignements pour

le barrage ayant subi l’incident. Ce type de rapport inclut en effet une analyse des
causes qui est l’élément essentiel pour tirer des enseignements utiles. La qualité de
cette analyse des causes est essentielle et doit chercher les causes profondes de
la rupture ou de l’accident. Dans le passé les causes identifiées étaient essentielle-
ment techniques. Elles ont eu l’immense mérite de faire évoluer les pratiques de
conception. Ces analyses ont souvent été transmises à la profession via des livres,
des bulletins techniques de la CIGB ou la publication de travaux de recherches.
De nos jours, les analyses des causes incluent également les causes organi-
sationnelles et humaines. C’est un exercice plus difficile mais qui apporte beaucoup
d’enseignements sur des facteurs non techniques: organisation du gestionnaire,
défaillance humaine collective ou individuelle, relations entre les acteurs, transmis-
sion des informations, supervision par l’autorité de contrôle, etc. Cette pratique est
largement répandue dans le monde de l’industrie et devrait être désormais systé-
matique pour tous les accidents de barrages. Une analyse des causes bien conduite
montre d’ailleurs souvent que derrière les facteurs techniques, en général bien iden-
tifiables, se cachent des causes profondes qui relèvent du domaine de l’organisation
générale de la sûreté et de la défaillance humaine.
Les rapports d’accidents débouchent donc sur des enseignements plus fins
sur les causes techniques et peuvent illustrer par des exemples concrets les
tendances indiquées par les analyses statistiques. En général les investigations
techniques sont utiles pour la compréhension fine de la rupture du barrage analysé
mais leur extension à l’ensemble des barrages n’est pas directement possible: c’est
le rôle des comités techniques de la CIGB d’analyser en quoi une rupture parti-
culière contient des enseignements génériques. Ceci est également vrai pour les
904
GR Q. 105
causes humaines dont les enseignements sont parfois plus facilement trans-
posables à de nombreux cas dans le monde, malgré les différences d’organisation
et de contexte. Car ces facteurs humains sont toujours les causes profondes d’une
rupture.
Il faut noter enfin que les analyses des accidents sont tout aussi, voire parfois
plus, utiles que celles des ruptures. Un accident qui a été correctement géré apporte
des éléments de compréhension sur la gestion organisationnelle de la crise, la
détection, l’évolution des phénomènes dangereux et les mesures prises pour éviter
la rupture. Le récent bulletin 180 de la CIGB: Lessons learnt from case histories
illustre parfaitement cette analyse des accidents.
3.5. EXPLOITATION DES DONNÉES STATISTIQUES RÉCENTES
Les trois publications de la CIGB citées au 3.1 et divers articles ont succes-
sivement évalué des statistiques de ruptures en cherchant l’influence du type de
barrage, de l’année de construction, etc. La mise à jour du bulletin 99 entreprise
par un groupe de travail au sein du Comité sécurité des barrages de 2017 à 2020
a permis d’ajouter dans l’analyse statistique les ruptures survenues depuis 1991
jusqu’en 2019. Cette mise à jour a également repris les cas connus auparavant
pour valider leurs données et ajouter des champs qui n’étaient pas inclus dans
les études précédentes, tels que le contexte, le mode de rupture, les causes de
l’incident. Les enseignements principaux sont les suivants:
• Il y a des différences importantes dans la précision et la fiabilité des rapports

de défaillances entre les pays de la CIGB, ce qui oblige à écarter certaines
données afin de ne pas fausser les résultats de l’analyse statistique.
• Le ratio du nombre de défaillances et du nombre total de grands barrages
existants diminue continuellement de 1,42% pendant les années 1900-1925 à
0,12% depuis 2000.
• Cependant, le ratio des barrages défaillants construits pendant une certaine
période apporte une vision moins positive. Ce ratio était de 0,29% pour les
années 1975-1999 et il est de 0,38% depuis 2000. Ce point est illustré par la
figure 1. Cela tendrait à montrer une légère augmentation du taux de rupture
au cours des 20 dernières années. Une autre explication pourrait évidemment
être une meilleure détection des ruptures de barrages depuis 2000 grâce aux
progrès des technologies de l’information (Internet, ...).
• Comme cela était déjà indiqué dans le Bulletin 99, les dix premières années de
vie d’un barrage sont toujours la période où se produisent 50 % des ruptures. Ce
ratio est même de 100% pour les barrages-voûtes et les barrages à contreforts.
• Pour les barrages dont la hauteur est comprise entre 15 et 75 m, le taux de
rupture par rapport aux barrages existants de même hauteur est assez con-
stant. Pour les barrages plus hauts, il diminue de manière significative. A ce
jour, aucun barrage de plus de 100 m n’a cédé.
905
GR Q. 105
Fig. 1
Taux de rupture en fonction de la période de construction
• Il n’y a pas d’influence significative du type de barrage et de la taille du réservoir

sur le taux de rupture. Le nombre de ruptures de barrages à voûtes multiples
et de barrages à contreforts est trop faible pour être statistiquement significatif.
• Les trois modes de défaillance pour les barrages en remblai sont la surverse
(40%), l’érosion interne (39%) et la défaillance structurelle (21%).
• Pour les barrages rigides, la défaillance de la fondation et l’érosion interne de la
fondation sont les modes de rupture dominants. Les défaillances structurelles
se produisent principalement dans les barrages poids en maçonnerie.
• Les ruptures de barrages se produisent soit en fonctionnement normal, soit
pendant une crue, ces deux contextes représentant 90% des ruptures. Depuis
2000, 70% des ruptures se sont produites pendant une crue. On peut noter
qu’en comparaison, les tremblements de terre ont causé un nombre très faible
de ruptures.
• En ce qui concerne les causes organisationnelles, on peut affirmer qu’une
conception ou une construction inadéquate sont de loin les principales causes
identifiées pour les barrages voûte, à contreforts et multi-voûte. Pour les autres
types de barrages (poids, remblai), une exploitation inadéquate pendant les
crues semble avoir un rôle dans environ 20% des défaillances. La figure 2
montre une analyse de l’évolution des causes organisationnelles de ruptures
dans le temps pour le mode de rupture par déversement sur la crête. A partir
de 1950 on voit que les causes organisationnelles dues à des insuffisances
d’exploitation ou de maintenance deviennent de plus en plus importantes par
rapport aux insuffisances de la conception. Cela indique que si les insuffisances
de conception ont été la cause la plus importante des ruptures de barrages
avant 1950, de nos jours ce sont les insuffisances d’exploitation ou de main-
tenance qui sont les causes les plus importantes. Les ruptures récentes qui
906
GR Q. 105
Fig. 2
Causes organisationnelles pour différentes époques de rupture mode de rupture
par surverse de la crête
ont marqué les esprits depuis 2000 en témoignent, par exemple Taum Sauk,
Delhi, Ivanovo ou Oaky.
• Les causes techniques sont différentes selon les types de barrages: les défail-
lances des fondations sont la cause dominante pour les barrages-voûtes et
les barrages à contreforts. Pour les barrages-poids en maçonnerie, les défail-
lances structurelles du corps du barrage sont une cause technique importante.
Pour les autres types de barrages-poids, les défaillances structurelles sont
aussi importantes que celles des fondations, tandis que la débitance insuff-
isante du déversoir est également une des causes. Pour les barrages en
terre, les deux causes dominantes sont les problèmes géotechniques (66%
des causes) et la débitance insuffisante de l’évacuateur de crues (28%). Pour
les barrages en enrochement, ces deux causes dominantes sont réparties
différemment, les défaillances géotechniques ne représentant que 32% et la
débitance insuffisante de l’évacuateur de crues 64%.
3.6. EXEMPLES D’EXPLOITATION DES DONNÉES SUR QUELQUES CAS RÉCENTS
Les vingt-cinq dernières années ont été marquées par des ruptures de bar-
rages dont certaines ont malheureusement occasionné plusieurs victimes. Le
tableau ci-dessous récapitule les accidents de barrages connus du rapporteur dont
la hauteur est supérieure à 15 m, sur lequel on a indiqué les principales caractéris-
tiques connues. D’autres ruptures se sont produites pendant cette période (au total
907
GR Q. 105
il y en a environ une centaine) mais les données disponibles sont insuffisantes pour
les faire figurer dans cette table.
Il faut cependant noter que quelques barrages plus modestes ont cependant
occasionné davantage de victimes, par exemple Shakidor au Pakistan en 2005
(70 victimes) et Koshi au Népal en 2008 (250 victimes). L’accident de l’évacuateur
d’Oroville figure dans cette liste bien que le barrage n’ait subi aucun dommage
pendant l’événement, mais on considère qu’il y a eu perte d’intégrité globale de
l’aménagement et lâchures d’eau incontrôlées.
ANNÉE NOM MODE NOMBRE

DE LA DU DE DE
RUPTURE BARRAGE PAYS TYPE H (M) RUPTURE VICTIMES
1996 KENOGAMI Canada TE 21 Submersion - 10
Erosion externe
1997 OPUHA Nouvelle Zélande TE 50 Submersion -
Erosion externe
1999 LYELL Australie ER 51 Désordres impor-
DAM tants
1999 SHIH Taiwan BM PG 25 Défaillance des
KANG fondations ou
des appuis
1999 EL GUAPO Venezuela TE (Z) 60 Submersion - 10
Erosion externe
2000 FRY Afrique du Sud TE 21 Submersion -
Erosion externe
2000 MAMBEDI Afrique du Sud TE 22 Désordres
LOWER importants
2000 LEBEA Afrique du Sud TE/VA 18 Submersion -
Erosion externe
2000 KOOS DE Afrique du Sud XX 15 Submersion -
BEER Erosion externe
2001 CHANG Inde TE (Z) 15,5 Ruptures, fissures
ou tassement
(fondations et
barrages en
remblai)
2001 TAPPAR Inde TE (H) 15,5 Glissement,
déplacement,
tassement dans le
corps du barrage
2002 ZEIZOUN Syrie ER/TE (Z) 32 Submersion - 22
Erosion externe
2003 BELLAIR Afrique du Sud TE 16 Submersion -
Erosion externe
2004 CAMARA Brazil PG (RCC) 50 Rupture dans la 7
fondation
(barrages
rigides)
2004 BIG BAY USA TE 17,4 Erosion
d’interface /
fuites
908
GR Q. 105

DE LA DU DE DE
2005 NANDGAVHAN Inde PG (M) TE 19 Glissement, 14
déplacement,
tassement dans le
corps du barrage
2005 TAUM USA TE ER 25 Submersion -
SAUK Erosion externe
2006 KA USA TE/ER 15 Submersion - 7
LOKO Erosion externe
2006 TUILIERES France PG (M) 31 Désordres
importants
2007 JASWANT Inde PG (M) 43 Erosion interne /
SAGAR fuites dans la
fondation
2008 PALEM Inde TE 46 Erosion interne /
VAGU fuites dans la
fondation
2008 BHIMLAT Inde CB (M) 17 Submersion -
RESERVOIR Erosion externe
2008 BEDFORD Australie PG 16 Désordres 1
WEIR importants
2009 ALGODOES Brazil TE 21,6 Submersion - 9
Erosion externe
2009 SITU Indonésie TE/ER 16 Submersion - 98
GINTUNG Erosion externe
2010 NIEDOW Pologne TE PG 16,7 Submersion - 1
(WITKA) Erosion externe
2010 GARARDA Inde TE 32 Erosion interne /
fuites (fondation ou
corps barrage en
remblai)
2010 DELHI USA TE PG 18 Submersion -
(Hartwick dam) Erosion externe
2011 FUJINUMA-IKE Japon TE 18,5 Glissement, 8
déplacement,
tassement dans le
corps
du barrage
2012 FORMIN Slovénie PG TE 49 Erosion interne/
fuite dans le corps
du barrage
2012 IVANOVO Bulgarie TE 19 Submersion - 8
Erosion externe
2013 OAKY Australie ER/PG 18 Submersion -
Erosion externe
2014 KREL_2 Vietnam TE (H) 27 Submersion -
Erosion externe
2014 HA Vietnam TE 27,5 Submersion -
DONG Erosion externe
2017 OROVILLE USA TE/ER 235 Submersion -
Erosion externe
909
GR Q. 105

DE LA DU DE DE
2018 SOLAI Kenya TE 25 Erosion interne / 50
corps barrage en
remblai)
2018 XE NAMNOY Laos TE 16 Erosion interne / 65
saddle dam fuites (fondation ou
corps barrage en
remblai)
2019 TIWARE - Inde TE 28 Erosion interne / 26
TIVARE fuites (fondation ou
corps barrage en
remblai)
2019 SPENCER USA BM TE 14 Submersion de 1
la crête
2020 EDENVILLE USA TE PG 13 Erosion interne /
corps barrage en
remblai)
2020 SANFORD USA TE 11 Submersion de la
crête
2020 THIN Viet Nam TE 15 Erosion interne /
corps barrage en
remblai)
2020 SPARMOS Grèce TE ? ?
2020 SARDOBA Ouzbékistan TE 32 Inconnu ou 6
pas clair
Toutes ces ruptures sont graves en raison des dégâts environnementaux, pour
le propriétaire et l’exploitant du barrage qui subissent une perte économique con-
sidérable. Elles sont dramatiques lorsqu’il y a des victimes Les questions qui se
posent après une rupture de barrages sont toujours les mêmes, rappelées dans le
R12: Pourquoi cela est-il arrivé, comment cela est arrivé, et sans doute la question
la plus importante: comment éviter que cela se reproduise ? Pour répondre à ces
questions les professionnels des barrages ont à leur disposition un outil essentiel:
l’analyse approfondie de la rupture.
Plusieurs rapports décrivent le processus d’analyse post rupture. On peut citer

ici la référence [19] publiée par l’ASDSO aux Etats Unis qui fournit des recomman-
dations et des guides très utiles pour procéder aux investigations post ruptures. Les
enseignements génériques que l’on peut tirer des rapports qui évoquent ce sujet
sont les suivants:
• L’analyse d’une rupture de barrage doit être entreprise systématiquement et

rapidement après l’évènement car certains indices de terrain peuvent dis-
paraître. Le recueil des témoignages des acteurs doit être également rapide, à
910
GR Q. 105
la fois pour éviter des ingérences extérieures et pour rassembler le maximum

d’information pendant que la mémoire des intervenants est encore fraîche ;
• Il est très conseillé de former un panel d’experts indépendants et multi disci-
plinaires car les sujets techniques sont en général multiples et requièrent des
compétences diverses;
• Les rapports d’analyse sont très approfondis sur les causes techniques de la
rupture. Ces analyses techniques approfondies sont évidemment indispens-
ables pour comprendre les mécanismes de la rupture du barrage concerné.
Leur transposition à d’autres ouvrages n’est pas toujours possible car chaque
barrage est un prototype et les caractéristiques de la fondation et des matéri-
aux du barrage qui ont pu jouer un rôle dans la rupture ne se retrouvent jamais
à l’identique sur d’autres sites. Par contre, il y a souvent beaucoup à appren-
dre d’une rupture au niveau des méthodes de conception, de surveillance et
de maintenance, ce qui est alors transposables à d’autres ouvrages. Il est
bien connu que la rupture du barrage de Malpasset a fait progresser les méth-
odes d’évaluation des fondations de barrages voûtes et que celle du barrage de
Balderhead a permis de découvrir les phénomènes de fracturation hydraulique,
etc. Les comités techniques de la CIGB ont un rôle à jouer dans la consolida-
tion de ces données de ruptures et les conséquences à en tirer pour éditer des
recommandations génériques sous forme de bulletin. On peut faire remarquer
ici que de nombreux bulletins décrivent des ruptures et les conséquences à en
tirer pour la profession.
• Les auteurs de certains rapports soumis à cette question 105 font par exemple
les recommandations suivantes:
◦ Xanmoy (R26): Les enseignements tirés portent beaucoup sur les
aspects géotechniques de la conception et sur l’auscultation. Mais
les auteurs remarquent également qu’il n’y avait pas de possibilité de
maîtriser le niveau d’eau de la retenue et que le plan d’urgence ne
concernait pas le barrage de col qui s’est rompu. Ce sont deux pistes
d’amélioration transposables à de nombreux aménagements.
◦ Malpaso (R31): Cette description d’un incident sur un CFRD (barrage
en enrochement à masque amont en béton) et des mesures correc-
tives associées peut constituer une référence sur le management des
incidents, sur les plans techniques et organisationnel.
◦ Toddbrook (R16): Le grave incident survenu sur le barrage de Toddbrook,
faisant suite à celui du barrage de Ulley a donné lieu à une revue de la
gestion de la sécurité et de la réglementation des barrages en Grande
Bretagne. Cette revue débouche sur des recommandations d’évolution
de la réglementation, en particulier en ce qui concerne l’inspection et les
rapports d’incidents. On peut remarquer ici que dans presque tous les
pays il faut que des incidents graves se produisent pour que les règles
imposées par les Etats soient édictées, puis renforcées en cas de nou-
veaux incidents. L’article R12 explique cela également pour le cas des
Etats Unis. Enfin le rapport signale que la gestion de crise à Toddbrook
a été une réussite parce que le propriétaire avait un plan d’urgence et
pratiquait des exercices régulièrement.
911
GR Q. 105
◦ Oroville, Taum Sauk (R11): Les auteurs font le constat que l’évaluation
de la sécurité par l’analyse des modes potentiels de rupture (PFMA) ne
permet pas d’évaluer la probabilité et les conséquences d’une rupture.
La FERC (Federal Energy Regulatory Commission) a publié des recom-
mandations sur les méthodes de type RIDM (prise de décision
informé par le risque) qui sont cependant peu utilisées par les consul-
tants car non obligatoires. La revue de deux accidents récents aux USA
(Oroville et Taum Sauk) a été entreprise pour voir si l’utilisation de ces
analyses RIDM aurait permis d’éviter les incidents. Pour Oroville
les auteurs présentent une description détaillée des PFMA (Analyse des
modes potentiels de rupture) successives en 2004, 2009, 2014 et de
l’évolution en qualité de ces revues. Ils expliquent pourquoi les scénarios
de soulèvement des dalles de l’évacuateur principal et celui de l’érosion
régressive de l’évacuateur de secours, qui sont les causes techniques de
cet événement accidentel, avaient été écartés.
• Un point fondamental souligné dans plusieurs rapports concerne l’importance
des Facteurs Humains. Suite aux analyses des ruptures de Taum Sauk et de
l’évacuateur d’Oroville les auteurs du rapport R11 écrivent que les facteurs
humains sont aussi important que les facteurs techniques . Dans l’absolu
on pourrait même affirmer que les causes des ruptures des barrages sont
entièrement dues à des facteurs organisationnels et humains ! En effet un
barrage étant par nature une structure artificielle conçue et exploitée par des
humains, sa rupture résulte soit d’une mauvaise conception, y compris une
mauvaise appréciation des sollicitations externes (crues, séismes, etc.), des
caractéristiques de la fondation ou des matériaux utilisés, soit d’une construc-
tion imparfaite, soit d’une exploitation défaillante ! Mais en général on utilise le
terme facteurs humains pour décrire les défaillances organisationnelles
et les gestes inappropriés des opérateurs. Le retour d’expérience des ruptures
depuis 25 ans montre à l’évidence que les causes techniques sont en
proportion moins importante que par le passé et que la cause de beaucoup de
ruptures est attribuable d’abord à des défaillances organisationnelles et parfois
humaines.
• Le corollaire de cette importance à accorder aux facteurs humains est de dis-
poser de méthodes d’analyse consacrées à cet objectif. En effet la recherche
de ce qui est appelé les causes profondes demande des compétences
d’analystes que n’ont pas en général les ingénieurs. Cette prise de conscience
émerge dans certaines organisations comme l’ASDSO avec les contributions
de spécialistes de ces domaines (comme I.A. Alvi) mais est loin d’être très
répandue. D’autres secteurs industriels comme l’aéronautique ou l’industrie
nucléaire pratiquent ces analyses approfondies depuis longtemps et les
méthodes d’analyse utilisées sont certainement transposables au domaine des
barrages.
• Par rapport aux causes techniques les causes liées au facteur humain ont très
souvent un caractère générique car elles ne concernent pas un ouvrage en
particulier: elles peuvent donc être utiles à un large spectre de professionnels.
On peut citer ici James Reason: We cannot change the human condition, but
we can change the conditions under which humans work ( on ne peut pas
912
GR Q. 105
changer la nature humaine mais on peut changer les conditions dans lesquelles
les hommes travaillent ).
• Enfin plusieurs rapports décrivent des incidents qui ont pu être gérés correcte-
ment et éviter la rupture du barrage. L’analyse de ces incidents graves est tout
aussi importante que celle des ruptures, voire davantage. En effet tout ce qui a
trait à la gestion de crise et à la chaîne de décision est souvent générique et peut
s’appliquer à d’autres organisations. Un aspect fondamental de ces analyses
est de montrer comment la détection précoce de l’incident permet de réagir
avec la rapidité suffisante pour éviter la rupture. Le bulletin 180 Lessons
learnt from case histories publié récemment par le Comité de la surveillance
décrit ainsi de nombreux incidents et la façon dont ils ont été traités. Le man-
agement en période de crise est l’objet du thème d) de cette question 105 et il
est traité plus en détail au chapitre 6.
3.7. COMMENT S’AMÉLIORER ?
L’analyse de quelques incidents récents de barrages est riche d’enseignements.

Les rapports approfondis réalisés pour certains de ces incidents permettent de bien
comprendre les raisons techniques de l’incident et parfois de remonter aux causes
profondes touchant à l’organisation ou au facteur humain.
Cependant, des progrès sont encore nécessaires dans plusieurs domaines

pour que ces incidents soient utilisés au mieux pour le bénéfice de la communauté
des barrages. Les difficultés rencontrées pour ce faire sont de plusieurs ordres:
• Il est remarquable de constater que les deux documents les plus complets
édités par la CIGB datent de 1974 et 1983. Le temps passé par les membres
du Comité des ruptures et accidents de grands barrages ( Committee on
Failures and Accidents to Large Dams ) pour constituer ce recueil d’incidents
et les analyser a dû être considérable. Leurs organisations respectives accep-
taient à l’époque de contribuer en nature aux travaux de la CIGB en leur
donnant le temps nécessaire. Il est malheureusement certain qu’aujourd’hui
cela ne serait plus possible.
• Il y a ensuite le temps et les moyens nécessaires à analyser une rupture.
L’organisation et le coût d’un panel d’expert ne sont pas à la portée de toutes
les organisations.
• La difficulté sans doute la plus importante est l’absence de transparence sur
les incidents. Il y a deux raisons à cet état de fait:
◦ d’abord une nécessaire confidentialité liées aux actions en justice.
L’analyse des causes est alors très délicate car ses résultats peuvent être
utilisés par l’enquête judiciaire et cela peut retarder considérablement la
diffusion d’éléments sur les causes techniques. Dans certains cas on
peut même craindre que les conclusions de l’analyse soient dissimulées.
◦ ensuite une volonté politique, de la part de certains pays ou de certains
propriétaires, de ne pas diffuser d’informations jugées négatives.
913
GR Q. 105
• Enfin la langue est parfois un obstacle: l’anglais est la langue très majoritaire
de la communauté internationale mais on ne doit pas pour autant négliger des
publications disponibles dans d’autres langues. Certains rapports de rupture
présentent des enseignements très utiles mais ne sont pas lus même si leur
existence est connue.
3.7.1. Progrès possibles dans la collecte des données
Il est évident que la disponibilité de données fiables et transparentes est

une condition indispensable pour analyser les incidents et en tirer des enseigne-
ments pertinents. Sans données on ne peut rien faire. Encore aujourd’hui il y a
malheureusement de nombreux exemples de ruptures de barrages pour lesquelles
on ne dispose d’aucune information au niveau international. Les seules informations
sont trop souvent des articles de journaux sur Internet, peu fiables et ne donnant
aucune information technique.
Certains pays sont au contraire exemplaires comme les USA qui mettent en
ligne sur Internet des bases de données des incidents survenus à leurs barrages et
publient ensuite les rapports d’analyses détaillées.
De nombreux pays ont promulgué des règlements exigeant que les proprié-
taires ou exploitants déclarent aux autorités les incidents qui se sont produits sur
les barrages dont ils ont la responsabilité. Ces exigences sont souvent assorties de
niveaux de gravité dont dépendent alors la rapidité avec laquelle les incidents doivent
être déclarés. Ces déclarations peuvent être succinctes dans un premier temps mais
doivent ensuite donner lieu à un rapport d’analyse par le propriétaire ou l’exploitant.
Le nombre d’incidents peut être assez important et il est peu envisageable que
cela puisse être consolidé au niveau international, sauf pour les incidents les plus
graves. Pourtant l’analyse de ces presque accidents est extrêmement importante
pour identifier des lacunes techniques ou organisationnelles et doit être conduite au
niveau de chacun des pays membres, en général annuellement. Les plus significatifs
pourraient être remontés au niveau de la CIGB.
Pour les ruptures il paraît évident qu’elles doivent être déclarées au niveau
international par les pays membres, le plus simple étant de passer par la CIGB et en
particulier via son comité de la Sécurité des Barrages. Les informations transmises
à ce comité seraient alors entrées dans la base de données des ruptures de barrage
disponible sur le site Internet de la CIGB.
3.7.2. Progrès possibles dans l’exploitation des données
L’exploitation des données des incidents est également indispensable si on

veut tirer des enseignements. Cette exploitation est à faire en général à deux
914
GR Q. 105
niveaux: celui des listes d’incidents dont l’objectif est plutôt d’avoir un recueil assez
exhaustif mais pas forcément très détaillé, avec l’objectif de procéder à des analyses
statistiques. Ces listes peuvent intégrer tous les incidents déclarés par les proprié-
taires et exploitants au niveau des pays membres et uniquement les ruptures et les
incidents graves au niveau de la CIGB. Le bulletin 188 est un exemple de ce qui peut
être fait au niveau de la CIGB. Au niveau de chaque pays, ces analyses peuvent être
plus détaillées, au pas annuel, et servir d’indicateurs de progrès à partir de critères
de nombre et de gravité des incidents déclarés, par exemple.
Ces listes d’incidents se prêtent très bien au format des bases de données
dans lesquelles on peut renseigner de nombreux paramètres caractérisant l’incident.
Cela est réalisé à la CIGB (bulletin 188) et dans de nombreux pays. Un piège à
éviter lorsque l’on développe une base de données est de prévoir de très nom-
breux champs, parfois une centaine (!) pour caractériser le barrage et l’incident.
L’expérience montre que la grande majorité des champs ne sont pas renseignés, ce
qui rend impossible les analyses statistiques sur ces champs. Il vaut mieux prévoir
moins de champs mais dont on connait les valeurs avec un bon degré de fiabilité et
pour la majorité des incidents.
En ce qui concerne les ruptures, les analyses détaillées ont pour objet de
comprendre les causes de la rupture, qu’elles soient techniques ou liées au facteur
humain. L’analyse des causes liées au facteur humain est un exercice assez nou-
veau et il serait utile que la CIGB travaille à un guide pour réaliser ces analyses
approfondies en se basant sur les guides existants dans le monde industriel. La
présence dans le panel d’experts d’un analyste spécialisé dans le facteur humain
devrait être systématique pour les ruptures et pour les incidents graves.
3.7.3. Progrès possibles dans la diffusion de l’information
Le rapport R12 est extrêmement explicite sur les enjeux de la transmission

des enseignements tirés des incidents de barrages: Le retour d’information sur
les incidents de barrages est efficace dans la mesure où la profession fait l’effort
d’enquêter sur les nouveaux incidents et de rechercher les incidents du passé.
Cependant, il est souvent difficile de partager ces enseignements largement au-
delà de chaque organisation et des frontières internationales. Pour aggraver ce
problème, le secteur de la sécurité des barrages connaît actuellement des taux
élevés de retraite parmi les ingénieurs et les personnes en charge du contrôle et de
la réglementation. Une grande partie des connaissances et de l’expérience acquises
au cours de décennies de construction de nouveaux barrages et d’apprentissage
des incidents seront perdues à moins que ces enseignements ne soient transmis à
la génération montante de professionnels de la sécurité des barrages.
Tout ceci est résumé dans la question la plus importante et objet de cette
Q105: comment pouvons-nous éviter que cela se produise à nouveau ?
915
GR Q. 105
On peut regrouper les différents moyens de transmission de l’information sur

les incidents de barrages en deux grandes catégories: les documents au sens large
mis à disposition de la profession, et les conférences, stages de formation, etc.
consacrées aux incidents de barrages.
• Les documents au sens large comprennent les rapports d’expertise, les arti-
cles, les listes nationales ou internationales, les bases de données, etc. tels
qu’ils ont été présentés dans le 3.3. La plupart de ces documents devrait être
disponibles sur Internet. Mais le fait d’avoir ces documents n’est pas suffisant:
d’abord il y en a beaucoup et, à moins de faire des recherches spécifiques sur
un type de rupture, ou un type de barrage, il est difficile pour un ingénieur de
prendre connaissance de cet ensemble de documents, par manque de temps,
ou par méconnaissance de l’existence même de ces documents.
• C’est pourquoi le second moyen de transmission de l’expérience acquise doit se
faire par des conférences ou des sessions de formation spécifiques. Plusieurs
organisations ont organisé des conférences internationales de ce type, par
exemple DSIG en 2009 et SWEDCOLD en 2013. Au niveau national un exem-
ple remarquable nous vient des Etats Unis où l’ASDSO prévoit dans chaque
réunion annuelle une session consacrée à la présentation d’incidents de bar-
rages. L’initiative appelée decade series consiste à présenter les incidents
s’étant produits dans les années correspondants à la décade de l’année de la
réunion (par exemple en 2019, traiter les incidents de 2009, 1999, …1909, etc.).
Cela permet à de jeunes professionnels d’être informés par des pairs expéri-
mentés des enseignements à tirer des incidents passés. Cela représente aussi
un travail important pour ceux en charge de ces présentations.
• La CIGB pourrait mettre en place une initiative similaire, chaque année ou
lors de chaque congrès, sous la responsabilité du Comité de la sécurité des
Barrages.
3.8. CONCLUSION
La question du thème a) est quels enseignements peuvent être tirés des

incidents récents ?
En résumant les analyses développées dans ce chapitre, le principal

enseignement à tirer des incidents survenus ces 25 dernières années est le suiv-
ant: ces incidents se produisent de plus en plus dans un contexte de crues et, de
ce fait, il faut chercher leurs causes dans des défaillances de l’exploitation ou de la
maintenance. Dans ces contextes, les causes profondes liées aux facteurs humains
sont très présentes.
Les bulletins de la CIGB abordent de nombreux aspects techniques du dimen-

sionnement ou des dispositions constructives des barrages. D’autres bulletins
916
GR Q. 105
proposent des recommandations pour l’exploitation et la maintenance. On peut con-

sidérer que le respect des recommandations contenues dans ces bulletins aurait
permis d’éviter la plupart des ruptures récentes.
Mais les causes profondes sont les plus importantes et sont liées à
l’organisation, au facteur humain, etc. Les deux récents bulletins 154 et 175 relatifs
au management de la sécurité des barrages pour toutes leurs phases de vie four-
nissent alors des éléments fondamentaux que tous les propriétaires et exploitants
de barrages devraient mettre en œuvre. Cela limiterait considérablement les risques
d’incidents graves et de ruptures.
4. EVALUATION DES DÉBITS ET DE L’INONDATION
Le thème b) de cette Question 105 traite de l’ Evaluation des débits et de

l’inondation, estimation et quantification des conséquences en cas de rupture ou
d’incidents, y compris du point de vue social, économique et environnemental .
Le bulletin 111 de la CIGB: Dam Break Flood Analysis, Review and

recommandations - Etude de l’onde de rupture d’un barrage, Synthèse et recom-
mandations, 1998 est la seule publication de la CIGB qui traite exclusivement
de ce sujet. Le bulletin 59 Dam Safety – Guidelines; Sécurité des barrages –
recommandations) évoque quant à lui très brièvement la nécessité de disposer de
la cartographie des zones inondées pour élaborer les plans d’urgence. Ce sujet a été
traité également lors des congrès de la CIGB à l’occasion des questions relatives à la
sécurité des barrages. On trouve par exemple dans la Question 40 Conséquences
de la construction des barrages sur l’environnement ou la Question 68 Eval-
uation et renforcement de la sécurité des barrages en service des articles qui
traitent de l’évaluation de l’onde de rupture d’un barrage et de ses conséquences.
Les objectifs de l’évaluation des débits et de l’étendue de l’inondation étaient

définis dans le bulletin 111 et n’ont pas changé depuis sa parution. Ces objectifs
sont les suivants:
1. Obtenir des données de base pour l’établissement des plans d’urgence (carte
des zones inondées, dynamique de la propagation de l’onde de rupture);
2. Recueillir des données nécessaires à la classification du barrage en quantifi-
ant les effets de sa rupture;
3. Dans certains cas, évaluer les risques pour les grandes installations situées
à l’aval dont la destruction présente de graves dangers (centrales nucléaires,
usines chimiques).
Les rédacteurs du bulletin 111 remarquaient également que les prescriptions

techniques à appliquer pour les études d’ondes de rupture n’étaient pas précisées
par des lois ou des règlements. Les autorités de contrôle de la sécurité des barrages
917
GR Q. 105
définissent des exigences sur les principes des calculs et les hypothèses à adopter
mais laissent aux responsables des barrages le choix des méthodes. Ces méthodes
figurent alors dans des recommandations publiées par des associations profession-
nelles. Les rapports R10 et R19 de cette Question 105 illustrent tout à fait cette
pratique.
Par rapport au bulletin 111, les rapports publiés pour cette Question 105 citent
des scénarios autres que la rupture du barrage. Le premier concerne le risque de
rupture d’un barrage naturel créé par un glissement de terrain barrant une vallée.
L’autre scénario qui est de plus en plus étudié est la rupture d’une vanne lorsque
sa débitance peut créer une onde équivalente à une crue importante à l’aval. Ces
scénarios ne sont pas toujours exigés par les autorités chargées du contrôle de la
sécurité du barrage mais c’est la responsabilité du propriétaire du barrage de les
étudier.
L’efficacité des évacuations de population et la sauvegarde des biens

économiques en cas de rupture de barrage repose notamment sur la prédiction
de la propagation de l’onde de rupture. La caractérisation de l’inondation consécu-
tive à cette propagation de l’onde de rupture est donc un élément essentiel parmi
les dispositions de sûreté associées à tout barrage qui présente des enjeux aval
significatifs (4.2). La prédiction de cette inondation est déterminée par un calcul
hydraulique, dont la condition limite amont est l’hydrogramme de brèche (4.1).
L’évaluation des débits et des inondations peut être scindée en trois parties qui
sont traitées ci-dessous. Il y a d’abord 1) l’évaluation de l’hydrogramme de brèche, 2)
puis le calcul de la propagation de l’onde et l’étendue des zones inondées, et enfin 3)
l’évaluation des conséquences. Tout ce qui concerne la surveillance opérationnelle
du barrage, la détection des anomalies pouvant conduire à la rupture, l’organisation
de l’exploitant pour faire face à une situation d’urgence, l’alerte aux populations,
la répartition des responsabilités entre l’exploitant et les pouvoirs publics, etc. est
traité dans les chapitres 5 et 6 ci-après.
4.1. HYDROGRAMME DE BRÈCHE
La connaissance de l’hydrogramme de brèche est la première étape indispens-

able à l’évaluation des caractéristiques de l’onde de rupture et des zones inondées.
Cela consiste à évaluer le débit s’écoulant en fonction du temps à travers la brèche
qui se crée dans le barrage. Cette évaluation dépend du type du barrage consid-
éré car elle est assez différente selon qu’il s’agisse d’un barrage rigide (en
béton ou en maçonnerie) ou d’un barrage meuble (barrage en remblai ou en
enrochements).
Pour les barrages rigides de type barrage poids, multi voûte ou barrage à con-
treforts les pratiques actuelles font l’hypothèse d’une rupture quasi instantanée, soit
de la totalité du barrage, soit d’un ou de plusieurs plots (monolithes) ou contreforts.
918
GR Q. 105
L’hypothèse de la disparition totale de tout le barrage est certainement conservative

car l’historique des ruptures de ce type de barrages indique que le cas général est
la rupture de quelques plots. Le choix est à faire en fonction des caractéristiques
propres de chaque barrage et de la forme de la vallée. Pour les barrages voûtes
l’hypothèse habituelle est la rupture totale et instantanée de l’ensemble du barrage.
Le débit maximum et la forme de l’hydrogramme peuvent être évalués en

assimilant la brèche à un seuil hydraulique ce qui permet d’obtenir le débit maximum
puis de construire un hydrogramme à partir de ce débit maximum et du volume
de la retenue. Cette approche est désormais remplacée par le calcul plus précis
de l’hydrogramme de brèche par les simulations numériques de la propagation de
l’onde de rupture décrites au 4.2. Dans ce cas la condition aux limites en amont du
modèle numérique n’est pas un hydrogramme de brèche pré-établi mais résulte du
calcul complet de l’écoulement à travers la brèche en tenant compte de la vidange
du réservoir et des conditions d’écoulement en aval immédiat du barrage.
Le cas des barrages en remblai est beaucoup plus complexe car la formation
de la brèche dépend de l’événement initiateur de la rupture qui peut être une érosion
externe par submersion de la crête ou une érosion interne par formation d’un renard.
Dans les deux cas, la durée de la formation de la brèche et l’évolution de ses
dimensions dépendent des caractéristiques géotechniques du matériau constituant
le remblai et de la hauteur et de la vitesse du courant. Les processus physiques
résultent d’une interaction complexe entre l’écoulement de l’eau par-dessus ou à
travers le barrage et le matériau du barrage qui possède ou non des propriétés de
cohésion.
La pratique actuelle est bien décrite dans le rapport R19 qui traite d’abord
des approches simplifiées à base de relations empiriques puis des approches par
modélisation couplant écoulement et érosion.
4.1.1. Approches simplifiées
Ces pratiques sont très bien traitées dans les rapports R10 et surtout R19.
Elles consistent à évaluer un débit de pointe de brèche par des relations empiriques
simplifiées puis à construire un hydrogramme de brèche. Il faut noter que ces rela-
tions s’appliquent à deux processus de ruptures assez différents: l’érosion externe
par submersion de la crête, et l’érosion interne. Des comparaisons des prédictions
obtenues par ces relations simplifiées avec des débits mesurés lors de ruptures de
barrages sont présentées dans le rapport R19. Ces comparaisons sont également
utilisées pour caler certains paramètres des relations utilisées. Mais les résultats
de ces évaluations sont parfois bien différents des valeurs mesurées et conduisent
à recommander d’utiliser plusieurs relations et de s’appuyer sur le jugement de
l’ingénieur pour retenir un majorant et un minorant pour ce débit de brèche. Il
est également recommandé de réaliser une étude de sensibilité pour tester les
conséquences de ces choix sur la suite des études.
919
GR Q. 105
Deux approches simplifiées sont ensuite utilisées pour construire l’hydro-

gramme de brèche.
◦ Lorsque le débit maximum de brèche est évalué comme indiqué ci-dessus,

une forme triangulaire est choisie pour l’hydrogramme de brèche. Les deux
paramètres importants sont le temps d’atteinte du débit de pointe, évalué à
dire d’expert et qui varie habituellement entre 30 minutes à quelques heures.
Le second paramètre est le volume de l’hydrogramme qui doit correspondre
au volume du réservoir.
◦ La deuxième approche consiste à évaluer le débit de pointe de la brèche à
l’aide d’une loi de seuil et d’une estimation de la largeur maximale de brèche,
puis de construire un hydrogramme de brèche simplifié en se basant sur
une vitesse moyenne d’élargissement de la brèche. Cette vitesse moyenne
d’élargissement de la brèche est un paramètre important qui est fonction des
caractéristiques d’érodabilité des matériaux et de l’époque de construction du
barrage. D’après R19, la fourchette de ces valeurs de vitesse d’élargissement
est très large, variant de 15 m/min à 0,01 m/min.
4.1.2. Modélisation couplant écoulement et érosion
La modélisation déterministe basée sur la physique des mécanismes

d’initiation et de développement d’une brèche dans les barrages en remblai n’est
actuellement pas disponible pour des études opérationnelles d’ingénierie. La seule
exception concerne le cas de l’érosion par surverse d’un remblai homogène, de
géométrie simple, constitué de matériaux cohésifs. Pour ce dernier mécanisme,
un important effort de recherche a été mené depuis la fin des années 1990 sur la
description physique des phénomènes et leur modélisation numérique et expéri-
mentale à grande échelle. Cet effort de recherche a permis d’aboutir à la validation
de deux codes de calcul adaptés aux besoins de l’ingénierie, WinDAM et EMBREA.
Le rapport R20 présente des efforts de recherche en cours pour améliorer la

connaissance de ces mécanismes d’érosion interne et externe.
4.2. EVALUATION DES ZONES INONDÉES
L’évaluation de la propagation de l’hydrogramme de brèche en aval du barrage

est sans doute l’étape qui a atteint la plus grande maturité. Les pratiques décrites
dans le bulletin 111 restent pour la plupart d’actualité, les évolutions notables portant
essentiellement sur le raffinement des outils de calcul numérique rendu possible par
l’augmentation des puissances des ordinateurs et à la modélisation numérique de
terrain. Il ne faut pas oublier cependant que des modèles de calcul simplifiés peuvent
être utilisés dans certaines circonstances ou pour des ouvrages modestes.
920
GR Q. 105
4.2.1. Méthodes simplifiées
Deux de ces méthodes sont présentées dans R19. Elles fournissent les débits
en des endroits stratégiques identifiés dans la zone d’inondation ce qui permet
d’évaluer le niveau atteint quand on dispose en ces points de la relation hau-
teur/débit. Ces modèles simplifiés sont utiles en cas d’urgence, par exemple pour
évaluer rapidement les caractéristiques d’une onde de rupture en aval d’un bar-
rage naturel formé par un glissement de terrain. Ils présentent cependant de fortes
incertitudes et l’usage est de prendre des marges de sécurité importantes sur leurs
résultats.
4.2.2. Calculs de l’onde par modèle numérique
La modélisation numérique par résolution des équations de St Venant est

aujourd’hui la méthode de référence pour établir une onde de rupture. Dans la
majorité des cas, une modélisation 1D suffit. En effet, beaucoup de barrages sont
situés dans des vallées plutôt encaissées où les écoulements suivent le fond de
la vallée sans recirculation ou autre écoulement complexe. Toutefois, des modéli-
sations 2D peuvent s’avérer nécessaires dans les grandes plaines d’inondation, en
cas de fort débordement des cours d’eau. La plupart des logiciels intègrent ces deux
approches et une combinaison des deux. On peut citer parmi beaucoup d’autres
les logiciels HEC-RAS développé par l’USACE et TELEMAC développé par EDF
(Electricité de France) tous deux accessibles librement.
Les données d’entrée nécessaires à ces modélisations sont la géométrie de

la zone d’inondation, l’hydrogramme entrant dans le réservoir et/ou l’hydrogramme
de brèche et les hypothèses concernant les scénarios.
◦ Les données de terrain nécessaires aux modélisations sont en général

disponibles et déjà numérisées. Il faut s’assurer qu’elles ont la précision néces-
saire, de l’ordre de 5 à 30 m en résolution spatiale suivant la topographie du
lit de la rivière et des zones submergées, et de 30 cm à 1 m en altimétrie
et que leur qualité est conforme aux exigences de l’étude. En effet, la qual-
ité de la description de la géométrie de la zone inondée est fondamentale
pour garantir la qualité des résultats. Si besoin, ces données sont complétées
par des relevés ponctuels, par exemple avec la technologie LIDAR, et en
intégrant les caractéristiques géométriques détaillées d’ouvrages qui peuvent
avoir une influence sur l’écoulement comme les digues, les ponts, etc. et avec
des bathymétries (voir R10).
◦ Les calculs des zones inondées sont réalisés pour plusieurs scénarios de
rupture, a minima pour une rupture en crue et une rupture en exploitation
normale ( sunny day ). Les conditions d’écoulement en aval du barrage
et le niveau d’eau initial dans le réservoir sont initialisés en conséquence. De
même, l’hydrogramme entrant dans le réservoir est en général choisi comme
921
GR Q. 105
étant la crue de projet pour la simulation d’une rupture en crue, et comme le

débit moyen entrant pour la rupture en exploitation normale. Le rapport R10
présente un panorama très complet des différents scénarios envisageables.
◦ Pour chaque scénario, il est également habituel d’évaluer les zones inondées
avec et sans rupture du barrage, du moins pour le cas de rupture en crue.
Cela permet de quantifier l’incrément de conséquences causé par la rupture
du barrage. Ce type d‘analyse incrémentale peut également être fait dans le
contexte des études de réhabilitation des évacuateurs de crues (R19). Les
résultats permettent de quantifier le gain apporté par les travaux de réhabili-
tation et fournissent des éléments importants dans le processus de prise de
décision.
Enfin il ne faut jamais oublier que ces simulations comportent une part
d’incertitude. Une bonne pratique consiste à majorer les hauteurs d’eau dans les
zones inondées et de réduire les temps d’arrivée de l’onde (R19).
Le cas des barrages de stériles est particulier (R28) et particulièrement dif-

ficile en raison de la nature des écoulements. Les principales difficultés propres à
ces barrages sont la disponibilité des données et le caractère non newtonien de
l’écoulement à intégrer dans les modèles hydrodynamiques. Les auteurs du rap-
port R28 citent à ce sujet le statisticien anglais George Box: tous les modèles
sont faux, mais quelques-uns sont utiles (all models are wrong, but some are
useful) et concluent que cela s’applique tout à fait aux études des ruptures des
barrages de stériles.
4.2.3. Restitution des résultats
Un objectif fondamental des calculs des zones submergées est de donner

aux autorités les éléments nécessaires à l’établissement des plans d’urgence. Il est
donc important de communiquer le plus clairement possible les résultats des calculs.
La pratique la plus répandue est de fournir des cartes des zones inondées, des
hydrogramme en différents points caractéristiques ainsi que les tableaux donnant
la totalité des valeurs calculées. Les cartes doivent être fournies à une échelle telle
que l’on puisse distinguer les habitats, typiquement au 1/25000. On y fait figurer
l’étendue des zones submergées, en y indiquant a minima le temps d’arrivée de
l’onde. On peut aussi indiquer la hauteur d’eau maximale et la vitesse maximale du
courant.
Certaines organisations vont beaucoup plus loin et proposent le calcul en

temps réel des zones inondées en fonction des débits réellement enregistrés ou
prévus et la mise à disposition en temps réel de ces informations aux autorités
chargés de l’application des plans d’urgence (R13).
L’USACE a ainsi développé un outil GIS en ligne pour aider les décideurs
lors des inondations et de la planification des exercices d’urgence. Cet outil, appelé
922
GR Q. 105
cartographie rapide des inondations (rapid inundation mapping), peut être utilisé
pour la communication de la cartographie de zones inondées suite à une rupture
de barrage ou lors de crues importantes. Les données cartographiques peuvent
être partagées avec les autorités chargées des plans d’urgence en quelques min-
utes. L’outil améliore la coordination entre les parties prenantes et s’adapte à leurs
besoins spécifiques et permet ainsi de faciliter la communication entre tous les
acteurs concernés en période de crise.
4.3. EVALUATION DES CONSÉQUENCES
L’évaluation des conséquences est la raison d’être de la détermination des

zones inondées avec deux objectifs essentiels:
◦ Définir la catégorie de risque du barrage lorsque celle-ci est basée sur les
conséquences;
◦ Elaborer les plans d’urgence pour les autorités chargées de la sécurité civile
et pour les exploitants.
Les conséquences sur l’environnement et le patrimoine culturel peuvent être

évalués également mais peu de rapports en font état.
L’évaluation des conséquences sur les populations est la préoccupation

majeure lors de ces études. On peut noter que cela est effectué avec des
méthodologies très diverses et des niveaux de détail extrêmement variable.
Les méthodes les plus simples tentent d’évaluer la Population soumise Au

Risque (PAR). Cela est effectué en comptabilisant, lorsque cela est praticable,
toutes les habitations et édifices situés dans la zone inondée à partir de plans
géoréférencés. Lorsque la zone inondée est très importante et que les habitats
ne peuvent pas être individualisés (cas de grands agglomérations densément peu-
plées) la pratique est d’utiliser les résultats des recensements de la population et
d’appliquer un ratio d’occupation. La distinction entre le jour et la nuit, entre les jours
de semaine et ceux du weekend, entre l’hiver et l’été peut permettre une meilleure
évaluation de ce nombre de personnes soumises au risque. Dans certains pays
cette évaluation du nombre de personnes soumises au risque est jugée suffisante
car le seul objectif est celui d’élaborer les plans d’urgence. Les incertitudes sur les
caractéristiques de la zone inondées sont prises en compte par des majorations
forfaitaires des hauteurs d’eau et des temps d’arrivée de l’onde.
Mais dans beaucoup d’autres pays l’évaluation des conséquences a pour

objectif l’évaluation du nombre de victimes potentielles(LOL: Loss Of Life). Les
méthodes utilisées débutent comme précédemment avec l’évaluation des person-
nes soumises au risque et utilisent ensuite des méthodes plus ou moins complexes
pour passer de ces PAR au nombre de victimes.
923
GR Q. 105
Les méthodes empiriques de Graham (1999), RESCDAM (2001) ou DEFRA

(2006) peuvent être utilisées pour évaluer le nombre de victimes dans des condi-
tions de crue. D’autres méthodologies avancées qui tiennent pleinement compte
des processus physiques répartis dans l’espace, des réponses humaines et de la
performance de divers systèmes sont disponibles. C’est le cas par exemple de Life-
SIM qui modélise la totalité de la zone inondée et effectue une simulation dynamique
de l’arrivée de l’onde pour les différents scénarios d’inondation retenus. Pour éval-
uer le nombre de victimes cette méthode intègre de nombreux paramètres tel que le
délai avec lequel est donnée l’alerte, le temps de réaction des personnes puis leurs
déplacements en voiture, la submersion éventuelle des routes ou leur encombre-
ment, etc. Pour les personnes se déplaçant à pied les courbes classiques de stabilité
d’une personne en fonction de la vitesse et de la hauteur d’eau sont utilisées. Enfin
les courbes de dommage permettent l’évaluation de la tenue des bâtiments pour les
personnes qui se seraient réfugiées sur les toits. Pour tenir compte des incertitudes
des études de sensibilité sont effectuées par exemple avec des méthodes de Monte
Carlo. Les ordres de grandeur de victimes potentielles sont alors associés à une
probabilité (R10).
Cependant, aucune méthode unique n’est appropriée ou applicable à une sit-

uation donnée. La pratique recommandée consiste à évaluer le nombre de victimes
potentielles à l’aide de diverses méthodes.
On peut légitimement s’interroger sur les raisons d’une telle différence

d’approche entre méthodes simples et méthodes complexes. Il y a certainement
une explication dans le fait que dans la réglementation de certains pays la classi-
fication des barrages n’est pas faite en fonction des conséquences c’est à dire en
fonction du nombre de victimes potentielles. En contrepartie, ces règlements exi-
gent souvent des majorations de l’évaluation des personnes soumises au risque.
Lorsque, au contraire, la législation prévoit la classification des barrages en fonction
des conséquences en termes de victimes potentielles cela incite les exploitants à
être aussi précis que possible. L’utilisation de critères d’acceptabilité de type F-N
est également une incitation à évaluer le nombre de victimes potentielles avec une
relative précision. Une autre explication tient sans doute au fait que, notamment en
Europe, les zones potentiellement inondées sont densément peuplées et que des
approches comme Life-SIM deviennent très lourdes à mettre en œuvre. Enfin le
système légal joue également un rôle: dans les pays où le système légal est basé
sur le Code Civil les autorités hésitent à afficher explicitement un nombre poten-
tiel de victimes, alors que cela pose moins de problème pour les pays régis par le
système légal dit de droit commun (Common Law).
Pour les ruptures de barrages en crue l’évaluation de l’incrément de con-

séquences dû à la rupture du barrage par rapport à celles de la crue naturelle
est souvent effectuée. Il n’y a cependant pas de cadre légal bien établi pour
cette approche qui est encore controversée, surtout pour l’évaluation des victimes
potentielles. Elle pose moins de problème pour l’évaluation des conséquences
économiques où il n’est pas besoin de tenir compte de l’élément humain: les habi-
tations et autres structures qui se trouvent dans la frange différentielle entre les
924
GR Q. 105
zones inondées par l’onde de rupture par rapport à la crue naturelle sont clairement
identifiables et ne peuvent pas se déplacer.
Une enquête sur les pratiques et les objectifs de l’évaluation des con-
séquences est en cours au Comité sur la sécurité des barrages. Les résultats
intermédiaires de cette enquête ont été utilisés en partie pour la rédaction de ce
paragraphe. Cette enquête indique que le nombre de barrages ayant fait l’objet d’une
évaluation des conséquences, et par la suite de l’élaboration d’un plan d’urgence,
est important, par exemple plusieurs centaines en Italie et en France. L’utilité de
l’évaluation des risques dans la gestion des barrages, l’importance de la planifica-
tion et de la communication d’urgence et le besoin d’une topographie précise sont
des aspects importants qui se dégage des réponses à cette enquête
Mais les résultats de cette enquête indiquent aussi les difficultés rencontrées
lors de ces études d’évaluation des conséquences:
◦ Les écarts entre les modèles pour l’évaluation du nombre de victimes et

la difficulté de déterminer le nombre de victimes probable pour une grande
population à risque;
◦ La difficulté de fixer les paramètres relatifs à la diffusion des alertes et à la
mobilisation de la population;
◦ L’estimation des dommages indirects;
◦ Le fait que l’évacuation peut faire plus de victimes dans certains cas (par
exemple des personnes piégées dans leurs voitures).
4.4. CONCLUSION
L’évaluation des caractéristiques de l’onde de rupture et de ses conséquences

est une pratique largement répandue parmi les professionnels des barrages. La
qualité des résultats de ces études dépend beaucoup de la qualité des données
utilisées dans les modèles de calcul, et en particulier des données de topogra-
phie caractérisant les zones potentiellement inondées. Le point le plus délicat
actuellement est la caractérisation de l’hydrogramme de brèche pour les barrages
en remblai. Les recherches en cours devraient améliorer dans le futur les prédic-
tions de cet hydrogramme. En attendant il est prudent de choisir des fourchettes
de valeurs majorées et de tester la sensibilité des résultats à ces variations des
paramètres de l’hydrogramme de brèche.
L’évaluation des conséquences est un domaine qui a également atteint une

certaine maturité. Les différences d’approche entre méthodes simplifiées ou mod-
èles dynamiques tiennent davantage aux contextes législatifs nationaux. Il faut
cependant garder à l’esprit que ces évaluations sont soumises à de nombreuses
incertitudes notamment lorsqu’elles modélisent le comportement des populations.
Des exercices réguliers des plans d’urgence sont certainement le moyen le plus
efficace pour s’assurer que les hypothèses adoptées sont crédibles.
925
GR Q. 105
Pour conclure sur ce point il est intéressant de noter que ces études ont
évidemment un coût, mais que le coût des dommages évités est largement
supérieur. Ainsi l’USACE estime que, entre 2004 et 2013, ses projets de réduc-
tion des dommages causés par les inondations ont permis d’éviter environ 8 $ de
dommages pour chaque 1 $ investi.
5. PLANS D’URGENCE
Le thème c) de la Question 105 traite des plans d’urgence et recouvre tout

ce qui concerne la réglementation, l’organisation, l’information des populations et
des exemples de mise en œuvre. Il s’agit donc de l’ensemble des moyens et des
organisations nécessaires pour gérer des situations de crise. Ces sujets ont été
abordés par la CIGB dans plusieurs bulletins et lors de congrès. On peut citer en
particulier les bulletins 154 et 175 traitant de la gestion de la sécurité des barrages,
du bulletin 180 Enseignements tirés des incidents de barrages , du bulletin 168
recommandations pour l’exploitation la maintenance et la réhabilitation et le
bulletin 59 assez ancien mais dont les recommandations pour les plans d’urgence
sont toujours actuelles.
De nombreux autres documents élaborés par des agences gouvernementales

ou des institutions financières sont également disponibles, par exemple les guides
de la FEMA ou les recommandations récentes de la Banque Mondiale qui renvoient
à de nombreuses références.
5.1. POURQUOI DES PLANS D’URGENCE ?
Malgré toute la rigueur apportée à la conception, la construction et

l’exploitation des barrages, on ne peut pas exclure qu’un incident grave se pro-
duise, le pire scénario étant la rupture du barrage. Toutes les industries à risque
sont confrontées à cette problématique où la probabilité d’un incident grave est très
faible mais où ses conséquences sont souvent très importantes.
Dans de nombreuses réglementations ou standards techniques, les

hypothèses de conception d’un barrage dépendent des conséquences d’une
éventuelle défaillance: plus ces conséquences sont importantes, plus les
hypothèses de sollicitations et de résistance seront sévères. Dans l’approche
analyse des risques on peut ainsi définir explicitement les barrières de
prévention qui sont des dispositions structurelles ou non structurelles mises en
œuvre pour réduire la probabilité de scénarios accidentels. Le contrôle de l’efficacité
de ces barrières pendant toute la vie du barrage permet de vérifier que le niveau de
sécurité du barrage demeure satisfaisant.
926
GR Q. 105
Cette prise en compte des conséquences d’une défaillance au niveau de la

conception et de l’exploitation d’un barrage est indispensable. Mais comme un inci-
dent grave, même de très faible probabilité, ne peut pas être exclu, il est tout aussi
indispensable de chercher à réduire les conséquences d’un tel incident.
C’est l’objectif des plans d’urgence qui traitent les quatre points essentiels
suivants:
◦ Détecter les anomalies et signes précurseurs d’une défaillance le plus en

avance possible; l’anticipation des événements est un facteur clé de tout plan
d’urgence;
◦ Définir les relations entre l’exploitant du barrage et les autorités, les respons-
abilités de chacun, les modalités d’alerte, etc.
◦ Définir par qui et comment le public est informé et les modalités d’évacuation
préventive;
◦ Définir les interventions possibles sur le barrage ou le réservoir.
Les moyens et l’organisation nécessaires à l’atteinte de cet objectif sont

détaillés ci-dessous.
5.2. ASPECTS RÉGLEMENTAIRES
5.2.1. Réglementation générale et plans d’urgence externe
Les règlements de très nombreux pays imposent l’élaboration d’un plan

d’urgence pour les barrages en fonction de leur importance et des conséquences
en aval. L’étude récente de la Banque Mondiale sur les pratiques de 51 pays four-
nit des renseignements très intéressants sur la manière dont ces plans d’urgence
sont prescrits aux propriétaires de barrage. Sur les 51 pays objet de l’étude seuls
27 ont une législation rendant obligatoire un plan d’urgence, 16 autres ayant des
plans d’urgence sur une base de volontariat des propriétaires. De son côté, le club
européen de la CIGB tient à jour une synthèse des réglementations de 17 pays avec
un chapitre consacré à la protection des populations. Plusieurs rapports soumis pour
cette question 105 décrivent le contexte réglementaire de leur pays, par exemple
au Portugal, en Suisse, en Italie et un article traite de la problématique plus délicate
des plans d’urgence dans un contexte transfrontalier.
Dans de nombreux pays, il existe des dispositions législatives et réglemen-

taires pour les situations d’urgence générales dans le cadre de protection
civile, par exemple en cas d’incidents graves sur des usines chimiques ou des crues
naturelles, etc. Ces dispositions législatives concernent alors aussi le domaine des
barrages dont la réglementation spécifique doit être compatible avec ces textes plus
généraux. Ces textes généraux prévoient l’élaboration d’un plan d’urgence que l’on
qualifie d’externe car il s’applique à des entités et services qui ne dépendent pas
927
GR Q. 105
hiérarchiquement du propriétaire du barrage. Ces plans identifient les acteurs et

leurs responsabilités et comment en situation d’urgence ces différents acteurs et
entités communiquent, se concertent et agissent de manière coordonnée. Outre le
propriétaire et l’exploitant du barrage, le régulateur représentant de l’État et tous
les services de secours, pompiers, protection civile, communautés locales sont
concernés à des degrés divers. Les plans d’urgence externe sont élaborés par les
pouvoirs publics à partir des données fournies par le propriétaire du barrage et en
concertation avec tous les acteurs. Un plan d’urgence externe comprend en général
les éléments suivants:
◦ Liste de tous les acteurs concernés: exploitant du barrage, régulateur, pro-

tection civile, police, hôpitaux, etc. et les moyens de communication entre
tous;
◦ Rappel des caractéristiques du barrage, du zonage aval, des différents
niveaux d’alerte et des actions à entreprendre à chaque niveau d’alerte;
◦ Responsabilité de chaque acteur pour les décisions cruciales et notamment
la déclaration des niveaux d’alerte, l’information du public, l’évacuation du
public;
◦ Modalités de communication vers le public en situation d’urgence;
◦ Evacuation des zones menacées d’inondation;
◦ Opérations de secours et autres dispositions d’urgence;
◦ Transports en situation d’urgence;
◦ Accès aux sites isolés;
◦ Révision régulière et exercices de crise via des exercices en salle ou
fonctionnels, impliquant tous les intervenants et parties concernées;
◦ Maintien en veille de services d’urgence;
◦ Information du public sur le contenu du plan d’urgence en particulier les
modalités de l’alerte et d’évacuation.
Dans les bassins transfrontaliers, les opérations d’urgence sont coordonnées

par les agences gouvernementales riveraines. La compatibilité des exigences et
des mesures de précautions prises par les agences riveraines doit être vérifiée
périodiquement.
5.2.2. Réglementation spécifique barrage
La réglementation générale de protection civile mentionnée ci-dessus con-

cerne des établissements industriels de différentes natures. Des textes spécifiques
définissent alors, par nature d’établissement industriel, le domaine d’application et
les exigences spécifiques.
Suivant les pays cette réglementation spécifique aux barrages est autopor-
tante . Elle doit alors définir l’ensemble des aspects décrits ci-dessus (acteurs,
928
GR Q. 105
responsabilités, etc.). Mais qu’elle soit autoportante ou soit incluse dans un cor-
pus législatif plus général, la réglementation spécifique aux barrages contient les
exigences ci-après:
◦ Définition des barrages soumis à la réglementation en fonction de critères

basés soit sur les conséquences d’un incident grave, soit sur la dimension
du barrage et du réservoir, soit un mélange des deux types de critères. Il est
possible à ce stade de définir des catégories de barrages qui seront soumis
à des exigences graduées;
◦ Etablissement d’un plan d’urgence pour chaque barrage au sein d’un
bassin versant ou faisant partie d’un ensemble d’exploitation (entreprise
d’exploitation, système énergétique, etc.), mais aussi globalement pour
l’ensemble du système. Création d’un comité de liaison composé de représen-
tants de toutes les entreprises exploitantes lorsqu’il y a plusieurs entreprises
exploitant des barrages dans le bassin versant;
◦ Choix des scénarios accidentels dont les conséquences justifient leur prise en
compte dans un plan d’urgence. Les scénarios sont essentiellement la rup-
ture totale ou partielle du barrage après le dépassement des niveaux d’eau
critiques dans le réservoir lors d’une crue extrême ou suite à un dysfonction-
nement de vannes, les défaillances suite à un séisme, le glissement de terrain
ou l’éboulement rocheux dans le réservoir. La rupture de vannes est prise en
compte quand les débits relâchés sont dommageables. On peut signaler ici
que certaines réglementations demandent que les scénarios d’ouverture des
vannes, volontaires ou non, soient inclus dans le plan d’urgence. Il s’agit alors
de sécurité publique plus que de situation d’urgence;
◦ Prise en compte des situations d’urgence pouvant résulter d’émeutes, de
terrorisme ou de guerre;
◦ Analyse du cas de ruptures en chaîne lorsque plusieurs barrages sont
construits sur la rivière;
◦ Pour chaque scénario accidentel retenu dans l’analyse de risques étude des
conséquences avec production des cartes d’inondation et l’estimation des
populations impactées en tenant compte de la cinétique des événements;
◦ La définition de différents niveaux d’alerte;
◦ La définition des critères pour la détermination du zonage en aval du barrage:
à l’aval immédiat c’est l’exploitant qui a la responsabilité d’alerte aux popula-
tions. Au-delà, c’est généralement la responsabilité des pouvoirs publics dans
le cadre du plan d’urgence externe;
◦ Le contenu du plan d’urgence interne qui doit être élaboré par le proprié-
taire (voir un exemple de plan type au 5.3.1ci-dessous);
◦ La nécessité de dispositifs de surveillance de nature à détecter les anomalies
de comportement du barrage pouvant conduire à un incident grave;
◦ La nécessité de moyens de communication entre l’exploitant et les autorités;
◦ La nécessité de dispositifs d’alarme des populations dans la zone aval proche
du barrage.
Le développement le long des rivières augmente les risques et les dégâts

causés par un éventuel accident de barrages. Si nécessaire, les plans d’urgence
929
GR Q. 105
externes et internes sont révisés à intervalle régulier pour s’adapter aux évolutions
de l’environnement naturel et socio-économique. Cette révision est faite en général
tous les 5 ans dans les régions en développement rapide et tous les 10 ans dans
tous les autres cas.
5.3. ACTIONS DU PROPRIÉTAIRE ET PLAN D’URGENCE INTERNE
Ces exigences réglementaires aboutissent à la production par le proprié-

taire des éléments nécessaires à l’élaboration du plan d’urgence externe et au
développement de son propre plan d’urgence interne.
Pour chaque scénario d’incident (rupture partielle ou totale du barrage, rupture

d’une vanne, etc.), le propriétaire fournit aux autorités les résultats des simulations
hydrauliques donnant l’étendue des zones inondées, le temps de propagation de
l’onde de rupture et l’estimation du nombre des personnes présentes dans la zone
inondée. Ces éléments sont utilisés par les pouvoirs publics pour définir plusieurs
zones en aval du barrage. Il y a en général deux zones, la première au plus près du
barrage où l’alerte aux populations sera donnée par le propriétaire avec ses moyens
propres (sirènes ou autres), la seconde où les pouvoirs publics seront responsables
de donner l’alerte et de procéder à l’évacuation des populations.
930
GR Q. 105
5.3.1. Plan d’urgence interne
Les plans d’urgence internes sont ensuite développés par le propriétaire ou

l’entité responsable du barrage. Ces plans abordent en général les points suivants:
◦ Surveillance hydrographique et dispositifs d’alerte de crues;

◦ Surveillance du barrage et dispositifs spécifiques de détection d’anomalies;
◦ Pour les modes de rupture retenus, identification des signes annonciateurs et
des seuils d’alerte;
◦ Correspondance entre les phénomènes constatés et les niveaux d’alerte;
◦ Carte des zones aval du barrage avec l’implantation des sirènes ou autres
dispositifs d’alerte des populations;
◦ Communication de crise interne et externe vers les services d’urgence;
◦ Instructions pour les équipes d’exploitation locales y compris en cas de perte
de communication avec les responsables du barrage et les pouvoirs publics;
◦ Etablissement et fonctionnement d’une cellule de crise;
◦ Cartes d’inondation pour la crue de projet et pour l’onde de rupture résultant
des scénarios incidentels retenus;
◦ Modalités de vidange d’urgence du ou des réservoirs;
◦ Equipements, matériels et services d’appui disponibles pour les actions de
secours;
◦ Exploitation d’urgence des ouvrages associés au barrage comme les usines,
les systèmes de production d’eau ou d’irrigation, les écluses, etc.;
◦ Possibilités de réparations d’urgence sur le site du barrage;
◦ Révision régulière et exercices de crise via des exercices en salle ou
fonctionnels;
◦ Tests périodiques des dispositifs de communications et d’alerte.
Dans certains pays ces plans d’urgence internes doivent être formellement
approuvés par les autorités, mais ces exigences spécifiques varient d’un pays à
l’autre.
5.3.2. Caractéristiques des situations d’urgence
On ne soulignera jamais assez l’importance fondamentale de l’anticipation.

Lorsque la crise survient, il est toujours trop tard pour organiser les relations entre
les acteurs concernés et réfléchir aux actions à entreprendre. C’est le rôle du plan
d’urgence externe de définir les relations entre les acteurs, le mode de communi-
cation, la chaîne décisionnelle et donc les responsabilités de chacun. Comme dans
tout processus industriel les interfaces, qu’elles soient techniques ou entre organisa-
tions, sont toujours un point faible. Comment assurer que les messages émis soient
bien compris de la même façon par tous ? Comment assurer que tous sachent
parfaitement qui doit prendre tels types de décision et qui doit les appliquer ? Pour
cela des instructions en cas d’urgence, claires et compréhensibles, sont établies
931
GR Q. 105
et diffusées à toutes les acteurs et équipes d’exploitation. Ces instructions en cas

d’urgence sont faciles d’accès et définissent les instances autorisées ou obligées de
donner des ordres en situation d’urgence. Toutes les personnes devant agir en cas
d’urgence disposent d’une copie de ces instructions dans leurs bureaux et à leur
domicile. Les plans d’urgence identifient donc tous les acteurs, le rôle de chacun et
les actions à entreprendre en fonction des niveaux d’urgence.
Cette notion de niveaux d’urgence, de trois à cinq suivant les pays, est vari-
able suivant les pays et les propriétaires. Cela semble normal car les organisations
de sécurité civile et les systèmes d’alerte varient d’un pays à l’autre et cela influe
sur le choix du nombre et des critères de définition de ces niveaux d’urgence. Le
développement de tableau à double entrée est une pratique qui se généralise. Cela
permet d’avoir rapidement une correspondance entre la défaillance constatée et le
niveau d’urgence. Un exemple d’un tableau de ce type est donné dans R36 pour
la Suisse. On trouve dans l’annexe C de [24] un exemple très détaillé de ce type
de tableau. Selon le type de défaillance il n’est pas aisé de fixer des critères de
niveau d’urgence. Les défaillances lors d’un épisode de crues sont les plus faciles
à traiter: les prévisions de débits entrants, la disponibilité des vannes, la présence
d’embâcles, etc. sont des critères objectifs à partir desquels il est possible de définir
des niveaux d’urgence. Cette définition est plus délicate pour les modes de défail-
lance affectant les barrages en remblai ou les fondations: l’occurrence d’une fuite
chargée en sédiments, la hausse de pression dans un piézomètre, etc. sont des
phénomènes qui demandent une expertise avant de se prononcer sur leur degré de
dangerosité et le niveau d’urgence à leur associer. Cette expertise est à identifier
dans le plan d’urgence ainsi que sa disponibilité en cas d’urgence. La problématique
est tout à semblable lors d’un séisme.
Niveaux Défaillance
d’urgence Crues Séisme structurelle Autres
Niveau 1 Atteinte de la Intensité > I1 Fuites hausse Glissement
cote Z1 pression terrain, etc.
Mesures
d’auscultation
anormales
Niveau 2 Atteinte de la Intensité > I2 et Fuites Glissement
cote Z2 évolutions hausse pression terrain, etc.
défavorables Avis expert Avis expert
Niveau 3 Atteinte de la Intensité > I2 et Fuites hausse Glissement
cote Z3 défaillance pression Avis terrain, etc.
structurelle expert Avis expert
Les délais disponibles avant de décréter un niveau d’urgence sont une

donnée cruciale. Ces délais sont très utiles aux pouvoirs publics et à la sécu-
rité civile pour préparer les actions à entreprendre avant l’atteinte du prochain
niveau d’urgence. Pour les situations en crue, il est en général possible d’estimer
932
GR Q. 105
correctement le délai disponible avant l’atteinte du niveau d’urgence supérieur. On

calcule le temps disponible avant l’atteinte d’une cote dans la retenue à partir du
gradient de montée de l’eau par exemple, ou on connait le temps de manœuvre
d’une vanne sur alimentation normale ou secourue et ses conséquences sur le débit
évacué. Cette évaluation de délai est beaucoup plus délicate pour les modes de
défaillance affectant les barrages en remblai ou les fondations et requiert l’expertise
évoquée ci-dessus.
La mise en place de plusieurs cellules de crise est une pratique courante en

situation d’urgence. Certaines crises peuvent durer plusieurs jours voire davantage.
Les plans d’urgence prévoient alors la rotation des responsables et des participants.
L’anticipation concerne également les accès au site des barrages. Lorsque

les prévisions météorologiques annoncent des épisodes de fortes précipitations
accompagnés de vents violents, il y a un risque de ne pas pouvoir accéder au
site du barrage pour le surveiller et exécuter les manœuvres. Une bonne pratique
consiste à envoyer les équipes d’exploitants avant l’arrivée des perturbations pour
parer à toute éventualité. Lors des tempêtes de 1999 en Europe, les routes d’accès
à certains barrages ont été impraticables pendant plus de 48 h par exemple.
En cas d’intempéries majeures, l’impossibilité de toute communication entre

le barrage et les cellules de crise ne peut pas être exclue. Dans ces cas, des actions
correctives sont prises immédiatement et indépendamment de toutes procédures
administratives courantes. Le responsable technique d’exploitation est investi de
l’autorité suffisante pour ordonner d’urgence des mesures préventives ou des con-
fortements sans avoir à demander d’autorisation spéciale à ses responsables. Cela
peut aller jusqu’à la décision de déclencher les systèmes d’alerte aux populations.
Les communications entre les différentes cellules de crise peuvent également être
affectées, les plans d’urgence externe prévoient cette éventualité et décrivent les
processus dégradés pour la prise de décision.
L’anticipation inclut évidemment la prévision des phénomènes pouvant con-

duire à des défaillances:
◦ Pour les crues cela implique que l’exploitant dispose en permanence des infor-
mations de centres météorologiques. Certains grands opérateurs ont leurs
propres équipes de prévision mais la plupart des exploitants s’appuient sur
des agences régionales ou nationales avec lesquelles elles ont passé des con-
trats. Pendant la crise la surveillance du niveau de la retenue est primordiale.
Elle est souvent mesurée par des capteurs mais la présence d’une échelle de
niveau reste un dispositif indispensable;
◦ Pour les séismes, la plupart des exploitants s’appuient sur des agences spé-
cialisées avec lesquelles elles passent des contrats pour être alertés. Le calcul
de l’intensité sur le site fait en général partie des tâches déléguées à ces
agences. Certains barrages sont munis de capteurs sismiques dont l’objectif
est plus souvent de disposer de mesures pour des analyses dynamiques a
posteriori. Les capteurs de mouvement fort sont dans doute plus efficace
en cas de séisme. Dans tous les cas les actions à entreprendre sont des
933
GR Q. 105
inspections immédiates ou différées suivant l’intensité, puis la déclaration de

niveaux d’urgence en fonction des constats de ces inspections;
◦ Enfin tous les phénomènes pouvant amener une défaillance structurelle ou
géotechnique relèvent de la surveillance du barrage par inspections, auscul-
tation et tests périodiques des organes de sécurité. De nombreux bulletins
de la CIGB sont consacrés à ces activités essentielles. Tous les capteurs
d’un système d’auscultation n’ont pas la même importance pour la détec-
tion d’une défaillance. Certains capteurs ont un objectif de compréhension du
comportement à long terme du barrage. D’autres sont en priorité prévus pour
la détection précoce de situation anormale. Le paramètre important est alors
la périodicité de la lecture de ces capteurs, qui doit être compatible avec la
cinétique d’une défaillance potentielle.
Malgré les analyses faites a priori lors de l’élaboration des plans d’urgence il
ne peut pas être exclu d’avoir un type de défaillance non prévu. Les défaillances des
évacuateurs de crue d’Oroville en sont un bon exemple. La réactivité des acteurs
et leur expertise sont alors essentielles. Cela milite davantage encore pour une
organisation de crise à la fois efficace et pouvant s’adapter à des situations diverses.
La décision d’évacuation des populations est une lourde responsabilité qui

incombe aux pouvoirs publics, sauf dans une zone proche du barrage où l’alerte ne
peut être donnée efficacement que par les exploitants. De nombreuses réglemen-
tations nationales définissent alors un zonage en aval du barrage, avec a minima
deux zones:
◦ Une zone proche du barrage où les moyens d’alerte de la sécurité civile sont
inexistants ou en nombre insuffisants pour alerter les populations. Le proprié-
taire conçoit et installe dans cette zone des moyens d’alerte, en général des
sirènes;
◦ En aval de cette première zone une zone qui sera couverte par les moyens
d’alerte existant dans le cadre général de la sécurité civile.
Les développements des moyens de communication des deux dernières

décennies ont fait émerger des outils nouveaux comme les alertes téléphoniques
de masse ou l’envoi de SMS à tous les smartphones situés dans une zone donnée.
Certains pays disposent de protocoles d’alerte commun (CAP en anglais) permet-
tant l’échange d’informations entre divers systèmes d’alerte et diverses entités de
sécurité civiles.
L’importance du délai entre l’évacuation et la rupture d’un barrage a fait l’objet

d’études récentes, voir par exemple [22] et R40 de cette question 105. L’étude de
l’USBR citée dans [22] indique que le taux de mortalité est divisé en moyenne par
100 dans les zones ou le produit de la hauteur d’eau D et de la vitesse du
courant V est de l’ordre de 100 en feet2 /s (environ 9 m*m/s soient des zones
de fort impact de l’onde de rupture). La décision d’évacuation doit prendre en compte
ce facteur tout en évaluant les risques d’une évacuation inutile si le barrage ne se
rompt pas.
934
GR Q. 105
5.3.3. Activités opérationnelles préparatoires
De nombreux guides existent pour le déploiement des actions du plan

d’urgence interne. La référence [22] liste de nombreux guides développés à cet
effet. On peut citer aussi le guide de l’ACB [25] (CDA en anglais) paru en 2019.
Les activités opérationnelles essentielles dérivent de ce qui est indiqué au 5.3.2.
On donne ci-dessous quelques indications sur des aspects particuliers:
◦ Local de surveillance dédié: il est d’usage de prévoir un local de surveillance,

hors de la zone inondable et positionné de manière à avoir une vue dégagée
sur le parement aval du barrage et le ou les évacuateurs de crues. Ce local
est équipé des moyens de communication avec les cellules de crise interne et
externe et du pupitre de commande de déclenchement des systèmes d’alerte
aux populations dans la zone aval proche. Les tests de communication avec
les autorités et les essais des systèmes d’alerte sont réalisés périodiquement
depuis ce local. Les plans d’urgence et la documentation essentielle du bar-
rage y sont disponibles. Le risque d’intrusion malveillante dans ces locaux
de surveillance ne peut malheureusement pas être écarté et certains opéra-
teurs ont enlevé des documents sensibles, par exemple les modes opératoires
d’ouverture de vannes de vidange de fond;
◦ Les conditions météorologiques associées à de très fortes précipitations sont
souvent très sévères et peuvent donner lieu à des coupures d’alimentation
électrique et de communication, par chutes d’arbres sur les lignes par exem-
ple ou avaries généralisées sur le réseau de transport. C’est la raison pour
laquelle des générateurs de secours et des batteries sont prévues sur le site
des barrages pour assurer la continuité du fonctionnement des évacuateurs de
crues et celui des équipements essentiels, telles que les mesures de niveaux,
etc. Ces moyens d’alimentation en énergie sont testés régulièrement pour
vérifier leur bon fonctionnement. Une attention particulière est apportée à
l’autonomie du fonctionnement des générateurs: le délai de réapprovision-
nement des cuves de gasoil peut être important suivant l’état des routes
d’accès au site du barrage;
◦ Certains opérateurs prévoient des stocks de matériau (pour les barrages en
remblai) ou d’équipements de sécurité (sacs de sable, etc.) et pièces de
rechange pour les systèmes de sécurité;
◦ Salle dédiée pour les cellules de crise: cette salle est équipée des moyens
de transmission sécurisés et de matériels bureautiques et informatiques
nécessaires. Les plans d’urgence et la documentation technique essentielle
concernant le ou les barrages doit être consultable depuis cette salle, ce qui
est maintenant relativement facile lorsque cette documentation est numérisée.
Les tests des plans d’urgence ont lieu depuis cette salle pour vérifier la mise
à jour de tous les documents. La plupart des grands opérateurs de barrages
et des services publics de sécurité civile disposent d’une telle salle de crise
qui peut être également utilisée pour d’autres événements;
◦ L’information des populations concernées sur les signaux d’alerte et les lieux
de refuge en cas d’évacuation est assurée par la distribution de brochures ou
flyers, dont un exemple est donné dans R6;
935
GR Q. 105
◦ Les équipes d’exploitation sont formées aux opérations en cas d’urgence,

avec mise à jour périodique. Les programmes de formation et les procédures
sont mis à jour conformément aux modifications apportées dans les plans
d’urgence, et pour tenir compte des progrès technologiques;
◦ L’organisation des exercices de crise incombe en général aux autorités avec
une périodicité définie dans le plan d’urgence externe. Les exploitants testent
plus fréquemment leur organisation interne par des simulations de scénarios
de défaillance et activation des cellules de crise.
6. GOUVERNANCE DE LA SÉCURITÉ
Le thème d) de cette question 105 s’intitule Gouvernance de la sécurité:

définition des responsabilités, vérifications périodiques, essais de mise en œuvre,
organisation du retour d’expérience . Certains de ces sujets ont déjà été abor-
dés dans les chapitres précédents, et d’autres sont traités par des bulletins récents
de la CIGB, en particulier les bulletins 154 et 175 qui couvrent largement la gou-
vernance et le management de la sécurité des barrages. Dans ce qui suit, on se
focalise sur le rôle de la gouvernance de la sécurité vis-à-vis des incidents et
accidents concernant les barrages , sujet de cette question 105. Pour les gou-
vernements et les autorités publiques la gouvernance de la sécurité comprend deux
aspects: l’élaboration de règlements fixant les exigences et la capacité à contrôler
leur application – et l’organisation de la collecte, de l’analyse et de l’information
sur les incidents affectant les barrages du pays et plus largement les incidents au
niveau international. Les recommandations et les guides techniques, même si leur
prise en compte n’est pas obligatoire, constituent un apport substantiel à la gouver-
nance de la sécurité pour la profession. Pour les entités propriétaires ou exploitants
des barrages, la gouvernance de la sécurité passe par l’établissement d’un sys-
tème de gestion de la sécurité (SGS), lequel implique le respect des exigences
réglementaires, l’allocation de ressources humaines qualifiées, le contrôle interne,
etc. Tout ceci est décrit dans les bulletins CIGB mentionnés ci-dessus. Pour ce
qui concerne les incidents, ces exigences comprennent aussi la détection des inci-
dents et presque accidents, leur analyse et les rapports à adresser aux autorités
publiques.
Il semble pertinent de traiter ce sujet à plusieurs niveaux de gouvernance,

internationale, nationale et à celle des entités propriétaires de barrages.
6.1. AU NIVEAU INTERNATIONAL
Il n’y a pas d’organisations internationales ayant des prérogatives réglemen-

taires sur la sécurité des barrages. La CIGB est la seule organisation supra nationale
qui édite des recommandations couvrant tout le spectre des activités techniques
et organisationnelles relatives aux barrages. Ces recommandations concernent
936
GR Q. 105
tous les acteurs: les Etats et gouvernements, les investisseurs, les établissements
financiers, les concepteurs, les entreprises de construction, les fournisseurs, les
exploitants etc. A ce titre, la CIGB peut être considérée comme le garant des
principes de la sécurité des barrages des points de vue technique et organisation-
nel. La publication récente de la Déclaration mondiale sur la Sécurité des Barrages
illustre parfaitement ce rôle de la CIGB.
Des recommandations de la CIGB spécifiques au domaine des incidents sont

incluses dans les bulletins 154 et 175 déjà cités. D’autres organisations publient
des recommandations pour l’établissement des plans d’urgence. C’est le cas de la
Banque Mondiale [24] qui publie un recueil de bonnes pratiques à mettre en œuvre
pour les situations d’urgence. L’application de ces bonnes pratiques pour les projets
financés par la banque va dans le sens d’une meilleure gouvernance de la sécurité.
Des guides pratiques pour construire un plan d’urgence et mettre en œuvre

les moyens nécessaires ont été développés par des agences gouvernementales
(par exemple FEMA [26]) ou des associations professionnelles (par exemple CDA
[25]). Ces guides sont disponibles et servent souvent de références au-delà de leurs
pays d’élaboration.
Mais en complément de ces recommandations et de ces guides, une tâche

importante de la CIGB est le recueil des données concernant les incidents de bar-
rages. Plusieurs bulletins ou publications ont été produits sur ce sujet, comme déjà
indiqué au 3.1. Ce travail de collecte de données d’incidents et de leur mise à dis-
position de la communauté des barragistes constitue un retour d’expérience qui est
utilisé pour faire évoluer les pratiques de conception et d’exploitation contribuant
ainsi à l’amélioration continue de la sécurité. D’ailleurs, la référence à des acci-
dents est souvent présente explicitement dans les bulletins techniques de la CIGB
pour expliquer les raisons des recommandations. L’étape suivante sera la mise en
ligne de ces informations sous forme d’une base de données consultables depuis le
site internet de la CIGB. D’autres bases de données couvrant des incidents au-delà
d’un pays existent mais ne sont pas généralement dans le domaine public et assez
souvent thématiques, donc non exhaustives.
Il n’y a pas à ce jour au niveau de la CIGB des recommandations sur l’analyse

des incidents. Des guides d’analyse à la suite d’une rupture sont disponibles, par
exemple celui de la FEMA [19] et sont très utiles dans ces circonstances drama-
tiques. Il n’y a pas, à la connaissance du rapporteur, d’équivalent à ce guide pour
l’analyse des presque accidents dont les enseignements seraient très utiles à la
profession. D’autre part on constate que, lors d’une défaillance, les analyses sont
souvent orientées vers les causes techniques alors qu’il est reconnu que les causes
profondes, de nature organisationnelle et humaine, sont prépondérantes. C’est ce
qui ressort de l’excellent rapport sur l’accident de l’évacuateur d’Oroville dans lequel
l’importance de ces causes profondes est soulignée. La CIGB pourrait s’inspirer
des pratiques d’autres industries à risques comme la production d’énergie électro
nucléaire ou les industries chimiques qui ont développé des guides et ont une longue
pratique d’analyse approfondie des incidents.
937
GR Q. 105
6.2. AU NIVEAU DES PAYS
Au niveau national la gouvernance de la sécurité des barrages est portée

par deux entités: les services du gouvernement ou les agences gouvernemen-
tales chargées de la réglementation – et les associations professionnelles et plus
spécialement les comités nationaux de la CIGB.
6.2.1. Réglementations au niveau national
Le paragraphe 5.2.1 traite de ce sujet et on invite le lecteur à s’y référer.

On peut rappeler que les gouvernements se préoccupent de la gestion des situa-
tions d’urgences de toute nature, naturelle, industrielle, agressions malveillantes, et
établissent des règlements pour créer et organiser les services de sécurité civile:
pompiers, forces de sécurité, système d’alerte, etc. Ces règlements précisent les
modes de détection, les processus d’activation des services de sécurité et les
moyens d’information, d’alerte et d’évacuation de la population vis-à-vis de ces
risques.
Des règlements spécifiques viennent compléter ce cadre général pour les

industries à risques. Ils contiennent l’exigence des plans d’urgence, externe et
interne, et précisent les différents acteurs d’une gestion des situations de crise
et leurs responsabilités. En fonction des industries concernées ils définissent les
critères de déclenchement des alertes et les zones d’information et d’évacuation de
la population. Pour les barrages cela concerne la définition des différentes zones
aval et les différents niveaux d’alerte. Ces règlements peuvent aussi spécifier des
moyens particuliers de communication entre les acteurs et pour l’alerte à la popula-
tion. Ils prévoient enfin la réalisation périodique de tests du dispositif et l’information
préalable aux populations concernées.
De nombreux pays demandent aux exploitants de faire un rapport pour les

incidents sur leurs barrages. Ces rapports alimentent des bases de données et
permettent des analyses globales du niveau de sécurité, voir 3.2.3. Une bonne
pratique est de mettre cette base à la disposition de la communauté des barragistes
et de diffuser périodiquement une synthèse. La diffusion le plus large possible de
ces informations est bénéfique pour l’acculturation de tous les acteurs.
Le contrôle par des services du gouvernement du respect des exigences

réglementaires est une composante essentielle de la bonne gouvernance de la
sécurité. Cette règle s’applique évidemment au corpus réglementaire relatif aux
situations d’urgence. La réalisation d’exercices de crise est une exigence qui con-
cerne l’ensemble des acteurs allant parfois jusqu’aux populations. Ces exercices
sont très importants car ils permettent de vérifier que les préconisations des plans
d’urgence fonctionnent . Ils présentent aussi l’intérêt de sensibiliser tous les
acteurs aux risques potentiels et de les entraîner à avoir les bonnes réactions. Il est
938
GR Q. 105
illusoire de croire que des plans d’urgence qui n’ont jamais été testés seront effi-
caces lors d’une vraie situation d’urgence. Cela représente certes du temps passé
pour l’organisation et le déroulement de ces exercices mais c’est sans commune
mesure avec les gains potentiels.
6.2.2. Les associations professionnelles
Beaucoup de comités nationaux de la CIGB mandatent des groupes de travail

chargés de développer des guides et recommandations. Mais, à la connaissance
du rapporteur, le sujet des situations d’urgence n’est pas souvent traité. Chaque
exploitant est souvent livré à lui-même pour organiser ses équipes, choisir un
système d’alerte, de communication, etc.
Il existe cependant des exemples de comités nationaux qui ont constitué des
groupes de travail chargés de produire des synthèses d’incidents. Ces groupes de
travail réunissent plusieurs exploitants et de représentants des pouvoirs publics.
6.3. AU NIVEAU DE CHAQUE PROPRIÉTAIRE
Le sujet de la gouvernance de la sécurité est largement traité dans des bul-

letins de la CIGB. Pour les aspects spécifiques aux situations d’urgence les activités
qui relèvent de la responsabilité du propriétaire ou de l’exploitant ont été développés
dans les chapitres précédents et il leur appartient de les mettre en œuvre dans le
cadre de leur Système de Gestion de la Sécurité (SGS). Quelques compléments
sont développés dans le paragraphe suivant. Enfin le leadership des cadres et
dirigeants est un aspect fondamental pour la gouvernance de la sécurité car il con-
tribue à l’émergence d’une culture de sécurité dans les organisations. Ce point fait
l’objet du deuxième paragraphe.
6.3.1. Gouvernance des activités
On peut rappeler ici l’importance de la détection des signes avant-coureurs

d’incidents par une surveillance et une auscultation rigoureuse. Le bulletin 180 [10]
présente plusieurs cas où la détection précoce a évité le développement d’une
situation incidentelle. Un paramètre important est le délai entre chaque mesure ou
chaque inspection. Ce délai doit être défini en fonction de la cinématique des modes
de rupture. Il revient en conséquence aux dirigeants de mettre en place la structure
et les compétences nécessaires pour que cette surveillance soit efficace.
939
GR Q. 105
La détection des incidents et presque accidents et leur analyse contribue à

l’amélioration continue du process. Un indicateur de performance basé sur le nom-
bre d’incidents et presque accidents détectés puis analysés par an est très pertinent
pour vérifier la vigilance des équipes et le niveau de sécurité atteint. On s’attend
en effet à un chiffre quasi constant de la totalité des événements détectés qui com-
prennent tous les presque accidents, mais une baisse continue des incidents ayant
eu des conséquences. L’analyse approfondie des presque accidents les plus mar-
quants fournit des éléments d’amélioration et permet d’éviter que des incidents
analogues ne se produisent ailleurs dans l’organisation. Ces informations sont dif-
fusées à toutes les équipes qui, sous l’autorité de leurs managers, en prennent
connaissance, en discutent et vérifient que les causes profondes et techniques des
incidents ne sont pas présentes dans leur exploitation.
Les grandes entreprises peuvent aussi mettre en place des cellules de veille
en charge de collecter les incidents les plus marquants au niveau international et
d’analyser la possibilité que de tels incidents se produisent dans leur organisa-
tion. Par exemple, après la rupture du barrage de Taum Sauk, des propriétaires
d’installations de production d’électricité hydraulique par pompage ont vérifié la
résilience de leurs installations vis-à-vis de ce risque de sur-pompage.
Il revient enfin aux dirigeants de commanditer des audits internes et externes

sur des activités ciblées. Tout ce qui contribue à la prévention des incidents est
souvent un sujet pour des audits internes, et c’est très bien. Par contre la pratique
de faire appel à des panel d’experts indépendants est largement répandue pour
les activités de conception des barrages, mais beaucoup moins pour les activités
d’exploitation. Elle constitue pourtant pour les dirigeants et pour les équipes de
terrain une aide à la gouvernance en leur fournissant des recommandations sur les
éventuels points de faiblesse organisationnels ou techniques.
6.3.2. Leadership et culture de sécurité
Le leadership et la culture de la sécurité sont abordés au chapitre 2.5 du

bulletin 154. Cette notion de culture de sécurité est implicite dans le domaine
de la conception et de l’exploitation des barrages mais mériterait cependant d’être
mieux développée. Elle est en effet particulièrement importante pour tout ce qui
concerne les situations d’urgence.
La culture de sécurité peut se définir comme suit:
La Culture de Sûreté est l’ensemble des caractéristiques et des comporte-

ments qui, au niveau collectif et individuel, font que les questions relatives à la Sûreté
des barrages bénéficient, en priorité, de l’attention qu’elles méritent en raison de
leur importance …
940
GR Q. 105
Ces notions de caractéristiques et de comportements ne relèvent pas

de recommandations ou de directives, mais définissent un état d’esprit, à tous les
niveaux des organisations, du technicien sur le barrage jusqu’aux cadres dirigeants
et au conseil d’administration.
Les pratiques des autres industries à risque peuvent être une source
d’inspiration pour améliorer le leadership et la culture de la sécurité. Le rapport
d’orientation de l’OCDE [23] a été développé pour les industries chimiques mais
peut être appliqué directement à toute activité industrielle présentant des risques
importants pour le public et les investisseurs. Ce document s’adresse aux cadres
dirigeants pour leur faire prendre conscience de leur rôle majeur dans la sécurité de
leurs activités et leur propose un auto-questionnaire d’évaluation de leur pratique
de sécurité dans leurs entreprises. L’importance de leur devoir d’exemplarité est
soulignée.
Ce rapport décrit les caractéristiques d’organisations à haute fiabilité ,

capables d’assurer pendant de longues périodes une production relativement
exempte d’erreurs. Ces organisations présentent les caractéristiques essentielles
suivantes:
◦ Elles ont un sens chronique de la remise en question, en ce sens qu’elles

ne sont jamais portées à l’autosatisfaction. Par exemple elles n’écartent pas
la possibilité d’un accident imminent sous prétexte qu’elles n’en ont pas eu
depuis très longtemps;
◦ Elles réagissent fortement au moindre signal d’alerte: leur seuil d’intervention
est très bas. …
De nombreux témoignages de dirigeants illustrent ce rapport, par exemple:
Les enseignements tirés des incidents passés montrent qu’un fort leader-
ship en matière de sécurité est indispensable à la prévention des catastrophes, et
que ces enseignements doivent être acquis et appliqués dans tous les secteurs pour
éviter que les mêmes erreurs reproduisent à l’avenir de nouveaux accidents .
Et encore:
Pour assurer la sécurité des processus, il est absolument essentiel de

développer une culture dans laquelle tous les employés s’attendent à faire face
à l’imprévu, et mettent tout en œuvre pour travailler sans faire d’erreur. Ce type
de culture n’est possible qu’en faisant preuve de leadership à tous les niveaux de
l’organisation (Bob Hansen, P-DG de Dow)
Sur le plan pratique les bonnes pratiques recommandées portent sur cinq
axes pour lesquels les PDG et dirigeants devraient agir comme suit [23]:
◦ Leadership et culture d’entreprise: créer un environnement ouvert et trans-

parent pour favoriser l’émergence de questions posées par les collaborateurs,
941
GR Q. 105
être présents dans leur entreprise et y montrer leur intérêt sans faille pour la
sécurité;
◦ Connaissance des risques: connaître globalement les vulnérabilités et les
risques encourus, les différents niveaux de protection. Être responsables de
l’établissement des plans d’urgence;
◦ Information: s’assurer que les données des incidents et presque accidents
sont analysées en profondeur. Commanditer des audits internes et s’impliquer
dans l’analyse et la mise en œuvre de leurs recommandations;
◦ Compétences: s’assurer que l’organisation dispose des compétences à tous
les niveaux hiérarchiques, consacrer le temps et les moyens nécessaires à
l’analyse des dangers par des spécialistes, faire confiance aux compétences
de leur personnel sans négliger de faire appel à des experts externes si besoin;
◦ Action: coordonner et piloter les programmes de gestion de la sécurité.
Ces cinq axes font l’objet d’un questionnaire d’auto-évaluation dont l’objectif
est de détecter d’éventuelles faiblesses et engager des actions d’amélioration si
besoin.
7. CONCLUSION
Les vœux de la CIGB 2021 peuvent servir de conclusion à ce rapport.
Il n’y a pas de secret à la réussite: elle est le résultat de la préparation, du

travail acharné et de l’apprentissage à partir des échecs .
Cette phrase résume les thèmes traités dans cette Question 105 et
l’importance qu’ils revêtent pour la sécurité des barrages: la collecte des incidents
et leur analyse pour apprendre à partir des échecs , la préparation par les
plans d’urgence et les études de conséquences, et le travail acharné dont la
gouvernance de la sécurité est un des aspects.
8. REMERCIEMENTS
Le rapporteur remercie chaleureusement tous les membres du groupe de

travail formé au sein du Comité technique sur la sécurité des barrages pour leur
contribution à l’analyse des données d’incidents et leurs suggestions au cours de la
rédaction du bulletin 188. Merci également au président de ce comité, Andy Zielinski,
qui a soutenu et encouragé ce travail pendant plusieurs années. Le rapporteur tient
à remercier Russell Gunn de l’Office fédéral de l’énergie suisse pour sa relecture
du texte en anglais.
942
GR Q. 105
Merci enfin à mon entreprise Electricité de France qui m’a donné le temps
et les moyens pour constituer une base de données des incidents de barrages pour
la CIGB et à tous les collègues de travail au contact desquels j’ai beaucoup appris.
RÉFÉRENCES
[1] ICOLD World Declaration on Dam Safety.
[2] CIGB-ICOLD Lessons from dam incidents, Leçons tirées des accidents de
barrages - 1974.
[3] CIGB-ICOLD Deterioration of Dams and Reservoirs – Examples and their

analysis – Détérioration de barrages et réservoirs – Recueil de cas et analyse,
1983.
[4] CIGB-ICOLD Bulletin 99: Statistical analysis of dam failure, Ruptures de

barrages, analyse statistique - 1995.
[5] CIGB-ICOLD Bulletin 111: Dam Break Flood Analysis, Review and recom-
mandations - Etude de l’onde de rupture d’un barrage, Synthèse et recom-
mandations, 1998.
[6] CIGB-ICOLD Bulletin 120: Design feature of dams to resist seismic ground
motion - Aspects de la conception parasismique des barrages, 2001.
[7] CIGB-ICOLD Bulletin 154: Dam safety management: Operational phase of

the dam life cycle, 2017.
[8] CIGB-ICOLD Bulletin 164: Internal erosion of existing dams - Erosion interne
dans les barrages existants, 2017.
[9] CIGB-ICOLD Bulletin 175: Dam Safety Management: Pre- operational Phases
of the Dam Life Cycle, 2021.
[10] CIGB-ICOLD Bulletin 180: Lessons learnt from case histories, preprint.
[11] CIGB-ICOLD Bulletin 188: Statistical analysis of dam failure, preprint.
[12] CIGB-ICOLD congress 13 New Delhi 1979: Q49 Deterioration or failure of

dams.
[13] CIGB-ICOLD Congress 14 Rio de Janeiro 1982: Q52 Safety of dams in

operation General Report.
[14] CIGB-ICOLD Congress 19 Firenze1997: Q75 Incidents and failures of dams

General report.
943
GR Q. 105
[15] CIGB-ICOLD Congress 23 Brasilia 2009: Q91 Dam Safety Management,

General report.
[16] CIGB-ICOLD Congress 24 Kyoto 2012: Q93 Safety, General report.
[17] DEFRA Guide to risk assessment for reservoir safety management; Volume
2: Methodology and supporting information; Report – SC090001/R2, 2013.
[18] Q101 C1: A data base for dam incidents listed by ICOLD, Une base de données
des incidents de barrages répertoriées par la CIGB – M. de Vivo, P. Le Delliou,
F. Lempérière, M. Poupart, 26th ICOLD Congress Vienna.
[19] Dam Failure Investigation Guideline (2017) completed by the ASDSO Dam
[21] Justin Larouzée: Théorie et pratique des modèles d’erreur humaine dans la
sécurité des barrages hydroélectriques: le cas des études de dangers de
barrages d’EDF. Thèse de l’Ecole Nationale Supérieure des Mines de Paris,
2015. Français. NNT: 2015ENMP0067
[22] Wishart, M.J., Ueda, S., Pisaniello, J.D., Tingey-Holyoak, J.L., Lyon, K.N., and
Boj García, E. (2020) Laying the Foundations: A Global Analysis of Regu-
latory Frameworks for the Safety of Dams and Downstream Communities.
Washington, D.C.; World Bank Group.
[23] OECD (2017) Corporate Governance for Process Safety – OECD Guidance
for Senior Leaders in High Hazard Industries.
[24] World Bank. 2021. “Good Practice Note on Dam Safety – Appendix
4: Emergency Preparedness Plan (Sample Framework).” World Bank,
Washington, DC.
[25] Emergency Management for Dam Safety. CDA, Toronto, Ontario. Central
Water Commission. 2019. Guidelines for Assessing and Managing Risks
Associated with Dams.
[26] Federal Guidelines for Dam Safety: Emergency Action Planning for Dams
(FEMA 2013).
944
QUESTION 106
QUESTION 106
I
QUESTION
106
SURVEILLANCE, INSTRUMENTATION, MONITORING AND DATA
ACQUISITION
1. Long term performance of existing surveillance systems including reliability
and accuracy; importance of visual inspections.
2. New technologies in dam and foundation instrumentation and monitoring.
3. Data acquisition and processing to evaluate the behavior of dams, predict

and identify incidents.
4. Understanding and handling of large quantity of data, including artificial

intelligence approach.
SURVEILLANCE, INSTRUMENTATION, AUSCULTATION ET

ACQUISITION DES DONNÉES
1. Performances à long terme des systèmes de surveillance, y compris leur
fiabilité et leur précision ; importance des inspections visuelles.
2. Nouvelles technologies pour l’instrumentation et l’auscultation des barrages

et des fondations.
3. Acquisition et traitement de données pour évaluer le comportement des

barrages, prévoir et identifier les incidents.
4. Analyse et traitement de grandes quantités d’informations, y compris

l’approche par l’intelligence artificielle.
II
QUESTION
106
R. 1 MATTEO SBARIGIA, DAVIDE PAUSELLI, MATTIA PALMIERI,

PAOLO CHECCARELLI, ROSELLA CARUANA (Italy)
performance of large dams under the 2016-2017 seismic sequence
in central Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
R. 2 MASSIMILIANO CARCIONE, ROSELLA CARUANA,
MATTEO SBARIGIA, FABRIZIO FRASCA (Italy)
Enel experience of latest years in central Italy dam monitoring
through new technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
R. 3 MARCO BERSANO BEGEY, ROBERTO BERTERO (Italy)
Stability of the rocky slopes facing large reservoirs. innovative
verification and risk assessment approaches using 3d point cloud
analysis – the case of La Penna dam (Arezzo, Tuscany, Italy) . . . . . . . . . . . . . 38
R. 4 ARMANDO LANZI, MARIA CRISTINA BRAMATI,
ROSELLA CARUANA (Italy)
The Italian experience of post-earthquake inspection and safety
evaluation of large dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
R. 5 J.G. STENFERT, D. HONINGH, M. VAN HOEK, B. KOLEN,
I. VAN DE KERK, H. JANSSEN, K. MIDDELJANS, C.H. OOSTINGA,
E. BOERMA (Netherlands)
Data science pilot study: identifying risk of dam failure using
artificial intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
R. 6 FERNANDO SALAZAR, ANDRÉ CONDE,
CARLOS BARBERO (Spain)
An open-source software for dam monitoring data analysis: exploration,
curation and machine learning model fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
R. 7 STEFAN HOPPE, ÓSCAR PÉREZ ARROYO,
JÜRGEN FLEITZ (Spain)
Digital technologies to manage dam safety records . . . . . . . . . . . . . . . . . . . . . . . 87
R. 8 JUAN MATA, ANTÓNIO TAVARES DE CASTRO (Portugal)
Surveillance of large concrete dams aided by automated monitoring
systems and machine learning techniques. contribution from the
Portuguese experience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
R. 9 REPORT MOVED TO Q. 104
R. 10 REPORT MOVED TO COMMMUNICATION
III
R. 11 THIERRY GUILLOTEAU, MATTHIEU DOUAT,
OLIVIER RUSSO (France)
Retour d’expérience EDF de l’auscultation des barrages . . . . . . . . . . . . . . . . 111
R. 12 FRANÇOIS MARTINOT, GUILLAUME STOLTZ,
CYRIL GUIDOUX, MAXIME BOUCHER, SYLVIE NICAISE,
JEAN-ROBERT COURIVAUD, LAURENT PEYRAS
(France)
Méthodes de surveillance des fuites par fibre optique, mise en œuvre
au sein d’une organisation de surveillance des barrages et nouvelle
application aux ouvrages d’altitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
R. 13 XAVIEN MOLIN, PATRICE ANTHINIAC, PAULINE BOFFETY,
PHILIPPE KOLMAYER (France)
Mesures in-situ : un atout pour juger de la représentativité des
modélisations numériques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
R. 14 THÉO DEZERT, CHRISTOPHE PICAULT, BRUNO DAUMAS,
OLIVIER MAGNIN, CHRISTOPHE VERGNIAULT, JEAN-ROBERT
COURIVAUD (France)
Caractérisation des fuites et de la lithologie des digues de canaux par
méthodes géophysiques et fusion de données . . . . . . . . . . . . . . . . . . . . . . . . . . 190
R. 15 GUILLAUME TERRASSE, PAUL-HENRI FAURE, ERIC
VUILLERMET, YANN GAYET (France)
Utilisation des données haute résolution par drones pour la
surveillance des ouvrages hydrauliques: de l’acquisition au
traitement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
R. 16 ALEXANDRE SIMON, JEAN-PAUL FABRE,
MATHILDE DE GRANRUT (France)
Nouvelles analyses du comportement des barrages (piézométrie et
débits : comportements non linéaires) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
R. 17 GUILLAUME VEYLON, NATHALIE ROSIN-CORRE, YIFENG LIN,
CLAUDIO CARVAJAL, ANTOINE WAUTIER, FRANCK
TAILLANDIER, LI-HUA LUU, LAURENT PEYRAS (France)
Analyse des données d’auscultation de barrages par
des méthodes d’apprentissage automatique . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
R. 18 REPORT MOVED TO Q.105
R. 19 MATHIEU DESJARDINS, SEAN WHITAKER, ANTONY
TROLLOPE, CAIUS PRISCU (Canada)
Lessons learned from long term surveillance of two tailings storage
facilities in the Canadian arctic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
R. 20 BILL SHERWOOD, RYAN BUCHOON, ROBIN HOULIK (Canada)
“But where exactly are we?”: Positioning and navigation accuracy
in remote underwater surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
IV
R. 21 ERIC TIEDJE, C. RICHARD DONNELLY, KAI-SING HO,
PAUL TOTH, RICHARD A. DALE, RAMY SAADELDIN,
LUIGI PERRA (Canada)
Numerical analysis to support the rehabilitation and long-term
monitoring of the Waba dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
R. 22 SAM JOHANSSON, ARI DAVID, MICHAEL MONDANOS,
ANNA STORK, AURÉLIEN MORDRET (Sweden)
Distributed acoustic sensing for detection of changes in embankment
dams related to seepage and internal erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
R. 23 JASMINA TOROMANOVIC, JAN LAUE, HANS MATTSSON,
SVEN KNUTSSON, PETER VIKLANDER,
CHRISTIAN BERNSTONE
(Sweden)
Observations from initial impoundment of an experimental
embankment dam – field data and modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
R. 24 ROBERT TORNBERG AND CHRISTIAN BERNSTONE,
PETER VIKLANDER, HEDWIG HAAS (Sweden)
SAA measurement for positioning filter tips of standpipes in
embankment dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
R. 25 SAM JOHANSSON, CHRISTIAN BERNSTONE (Sweden)
Temperature modelling and distributed temperature sensing using
optical fibers in a test dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
R. 26 KUNIHIRO TOMITA, TETSUYA SUMI, AKIRA SUZUKI,
SHIGEHARU JIKAN, SHIGEYOSHI NOYORI, NOBUTERU SATO,
CHIKAKO, ARAYA (Japan)
Dam body behavior monitoring by GNSS in Hachisu dam and
applicability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
R. 27 MEGUMI NAKASHIMA, SATOSHI HARUNA, YUJI IWAMATSU,
SOTA UCHIDA, ATSUSHI GOTO, KEISUKE HATANO (Japan)
Benefits of introducing an ICT-based staff support system for
disaster response work along the lake Biwa shoreline and its
extended uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
R. 28 NARIO YASUDA, ZENGYAN CAO (Japan)
Seismic performance verification of a rockfill dam against large
doublet earthquakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
R. 29 MASAYUKI KASHIWAYANAGI, ZENGYAN CAO (Japan)
Investigation of damping characteristics of dams evaluated by
DE/TFM method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
R. 30 HIROFUMI OKUMURA, TETSUYA SUMI (Japan)
Characteristic evaluation and countermeasure planning of reservoir
sedimentation utilizing the long-term survey data in hydropower dam . . . . 489
V
R. 31 ABDELILAH BOUKAIDI LAGHZAOUI, SIHAM BELHACHMI,
MARIAM MAHDAOUI (Morocco)
Diagnostic des témoins d’auscultation des tirants précontraints du
barrage El Kansera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
R. 32 PEIWEI XIAO, BIAO LI, XINGGUO YANG, NUWEN XU, (China)
Characteristics of micro seismic b-value associated with large
deformation of rock mass in high geostress underground powerhouse
caverns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
R. 33 SHUAIDONG YANG, ZHIHUAI HUANG, XIAOLIANG WANG,
MI ZHOU (China)
Field text of soft soil preloading in the estuarine area of the
Pearl River . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
R. 34 FRANCK SCHMIDT, JONATHAN FAURIEL, JEAN-CLAUDE
KOLLY, REYNALD BERTHOD, VINCENT BARRAS
(Switzerland)
Lasergrammétrie et vidéo-tachéométrie au service de la surveillance
et de la sécurité des barrages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535
R. 35 DANIELE INAUDI, RICCARDO BELLI, RÉGIS BLIN
(Switzerland)
Fiber optic distributed sensing system for monitoring of tailings
storage facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 556
R. 36 MARIUS BÜHLMANN, DAVID F. VETSCH, GEORGES R.
DARBRE, RICCARDO RADOGNA, ROBERT M. BOES
(Switzerland)
Multi-objective calibration of dam behavior analysis model for
gravity dams: application to Robiei dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
R. 37 RUSSELL MICHAEL GUNN (Switzerland)
New concept for concrete swelling evaluations based on
surveillance and monitoring data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
R. 38 FRANCESCO AMBERG (Switzerland)
Advanced deterministic models to assess dam displacements
application example at 3 large arch dams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
R. 40 ANDRI P.W, LOLO W.R, RIZAL N.H (Indonesia)
Analysis and evaluation of short and long term Jatiluhur dam behavior
between instrument data and seep/w application . . . . . . . . . . . . . . . . . . . . . . . . 625
VI
R. 42 ARIS RINALDI, NALVIAN, JOKO MULYONO (Indonesia)
Hydrogeology study: sustainable groundwater monitoring of
Banyu Urip dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
R. 43 IRINEL DANIELA IACOB, CATALIN POPESCU (Romania)
Data acquisition and processing to evaluate the behavior of dams,
predict and identify incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
R. 44 JOAQUIM PIMENTA DE ÁVILA, GEAN LOPES TEIXEIRA
(Brazil)
Risk based monitoring planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
R. 45 FLORIAN LANDSTORFER, ERICH WAGNER (Austria)
Durlassboden – how old measurement data and new data processing
methods improve the understanding of a 50-year-old embankment
dam with underseepage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
R. 46 THÉO DEZERT, GANESH HIRIYANNA RAO RAVINDRA,
FJÓLA GUÐRÚN SIGTRYGGSDOTTIR (Norway)
Riprap and rockfill dam experimental models exposed to overtopping
events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676
Manuel G. MEMBRILLERA (Spain/Espagne) & Louis HATTINGH
(South Africa/Afrique du sud), General Reporters/Rapporteurs Généraux . . 695
VII
PAPERS ON
QUESTION 106
RAPPORTS SUR LA
QUESTION 106
VIII
Q.106 - R.1
-------
-------
PERFORMANCE OF LARGE DAMS UNDER THE 2016–2017 SEISMIC

SEQUENCE IN CENTRAL ITALY (∗)
Matteo SBARIGIA & Rosella CARUANA

ENEL GREEN POWER S.p.A. – Dams & Civil Infrastructures Safety
Davide PAUSELLI & Paolo CHECCARELLI

ITALIAN MINISTRY OF INFRASTRUCTURE AND TRANSPORT – General
Management for Dams and Water and Electrical Infrastructures
Mattia PALMIERI
CONSORZIO DI BONIFICA DELLE MARCHE
ITALY
SUMMARY
The Seismic Sequence occurred in Central Italy between August 2016 and
January 2017, with four major earthquakes with Moment Magnitude (Richter Scale)
Mw from 5.5 to 6.5, involved about 40 Large Dams.
The Owners for each dam, following each of the more than 50 earthquakes
with magnitude Mw > 4, carried out visual inspections and checks, in order to have
immediate information about evidences. Likewise, the Italian Dam Authority, in
collaboration with the Responsible Engineer of the dam, conducted extraordinary
surveillance audit in order to verify the safety conditions.
Generally, no significant effects either on the dam body or on the banks were
recorded. For all the dams, it was possible to deliver the extraordinary asseveration
of the safety conditions according to Italian law.
∗ Comportement des grands barrages pendant la séquence sismique 2016–2017 en Italie
centrale
DOI: 10.1201/9781003211808-99 1
Q.106 - R.1
During all the seismic sequence, the Owners normally raised the level of atten-
tion and the frequency of measurements for capturing the main effects and studying
the dam behavior. For the concrete dams closest to the epicenters millimetric con-
centrated displacements and slight settlements were observed. The deviation with
respect to the normal behavior was only in the first measures after the major earth-
quakes. Besides, several dams registered temporary increases of the uplift and
seepages, normally recovered in the following days.
For some dams, the Owners in agreement with the Italian Dam Authority
decided to take in action specific in-deep studies, even maintenance work and
improving projects. In addition, automatic and dynamic monitoring systems were
implemented in some dams, in order to provide better quality data on the dam
behavior, immediately available after an earthquake.
Summing up, the seismic sequence stressed the existing problems of the
dams, pushing for their resolution.
Concerning the appurtenant works, some damages were observed in the

buildings, mainly in terms of crack pattern of the masonry walls. The seismic
reassessment, already started, has been accelerated by the earthquakes, reveal-
ing several deficiencies with respect to the seismic actions, as severe as the ones
used for the dams. Consequently, upgrades of the appurtenant works have been
designed and some of that already carried out.
RÉSUMÉ
Une séquence sismique s’est produite dans le centre de l’Italie entre août
2016 et janvier 2017, avec quatre séismes d’une magnitude moment (échelle de
Richter) Mw de 5,5 à 6,5, et elle a concerné 40 grands barrages.
Les propriétaires de chaque barrage, après plus de 50 séismes de magni-

tude Mw > 4, ont effectué des inspections et des contrôles visuels, afin d’avoir des
informations immédiates sur les conséquences. De même, l’Autorité italienne des
barrages, en collaboration avec l’ingénieur responsable du barrage, a réalisé un
audit de surveillance extraordinaire afin de vérifier les conditions de sécurité.
En général, aucun effet significatif ni sur le corps du barrage ni sur les berges
n’a été enregistré. Pour tous les barrages, il a été possible de fournir une déclaration
extraordinaire des conditions de sécurité selon la loi italienne.
Pendant toute la séquence sismique, les propriétaires de chaque barrage ont

normalement élevé le niveau d’attention et la fréquence des mesures pour capter
2
Q.106 - R.1
les principaux effets et étudier le comportement du barrage. Pour les barrages en

béton les plus proches des épicentres, des déplacements concentrés millimétriques
et des légers tassements ont été observés. L’écart par rapport au comportement
normal n’était que dans les premières mesures après les séismes majeurs. En
outre, plusieurs barrages ont enregistré des augmentations temporaires des fuites,
généralement récupérés les jours suivants.
Pour certains barrages, les propriétaires en accord avec l’Autorité italienne

des barrages ont décidé de mener des études spécifiques en profondeur, voire des
travaux d’entretien et des projets d’amélioration. De plus, des systèmes de surveil-
lance automatiques et dynamiques ont été mis en place dans certains barrages,
afin de fournir des données de meilleure qualité sur le comportement du barrage,
immédiatement disponibles après un séisme. En résumé, la séquence sismique
a fait ressortir et accentué les problèmes existants des barrages, poussant à leur
résolution.
Concernant les ouvrages annexes, des dommages ont été observés dans les
bâtiments, principalement en termes de fissuration des murs de maçonnerie. La
réévaluation sismique, déjà entamée, a été accélérée par la survenue des séismes,
révélant plusieurs déficiences dont la sévérité est identique à celle des effets sur
les barrages. En conséquence, des mises à niveau des ouvrages ont été conçues
et certaines déjà réalisées.
1. INTRODUCTION
This paper deals with the performance of Large Dams involved in the severe
Seismic Sequence occurred in Central Italy on 2016–2017 with two earthquakes
with Moment Magnitude (Richter Scale) Mw more than 6.0 and four earthquakes
with Mw between 5.0 and 5.5, happened in just four hours.
A large number of dams was affected by the sequence. Both for each major
earthquake and dam, the Owner carried out visual inspections and several controls
in order to have immediate information about evidences. As well, the Italian Dam
Authority, in collaboration with the Responsible Engineer, conducted extraordinary
surveillance audit in order to verify the safety conditions.
Generally, no significant effects either on dam body or on the banks were

recorded. For all the dams, it was possible to deliver the extraordinary asseveration
of the safety conditions according to the Italian Law.
Some damages were observed only on the appurtenant structures and some
variations in respect to the normal behavior were measured by the monitoring system
installed.
3
Q.106 - R.1
Finally, in the paper some examples are given, in terms of evidences as well as
actions taken to deepen knowledge, analyze behavior, improve future performance
of the dams.
2. THE 2016-2017 SEISMIC SEQUENCE IN CENTRAL ITALY
Between August 2016 and January 2017, four major earthquakes occurred
in Central Italy. The first event, with Mw 6.0, took place on 24th August 2016, the
second one (Mw 5.9) on 26th October, the third one (Mw 6.5) on 30th October 2016
and the forth one (Mw 5.5) on 18th January 2017.
As shown in Figure 1, this earthquake sequence occurred in a space gap

between two earlier damaging events, the 1997 M6.1 “Colfiorito” earthquake in the
north-west and the 2009 6.1 Mw “L’Aquila” earthquake in the south-east.
Fig. 1
Seismic main events in the last 20 years in the interested area.
Principaux événements sismiques des 20 dernières années dans la zone
intéressée.
4
Q.106 - R.1
Fig. 2
Amatrice town before and after the 24th August Mw 6.0 earthquake.
Amatrice avant et après le séisme du 24 août Mw 6.0.
These events took place along the line of the Apennine Mountain range and
had rake angles ranging from −80 to −100 deg., which corresponds to normal fault-
ing. Each of these events produced substantial damage to local towns and villages.
2.1. EFFECTS ON TERRITORY
The 24th August event caused massive damages to the following towns:
Arquata del Tronto, Accumoli, Amatrice, and Pescara del Tronto. In total, there were
299 fatalities, generally from collapses of unreinforced masonry houses (fig. 2). The
October events caused significant new damages in the towns of Visso, Ussita, and
Norcia, although they did not produce fatalities, since the area had largely been
evacuated.
Furthermore, several important co-seismic effects were recorded (i.e. Monte

Vettore slip of about 1 m), as well as many rockfalls and landslides.
2.2. MAGNITUDE AND INVOLVED DAMS
The four main events of the 2016-2017 seismic sequence in Central Italy
(Amatrice-Norcia-Rieti) involved a large number of dams, subjected to controls if
included in the area of influence of earthquakes. The radius of these areas varied
between 60 and 100 km, starting from the epicenters. The 30th Oct. 6.5 Mw “Norcia”
event involved the largest number of dams, over 35, with different intensity.
The Italian Authority for the supervision of dam safety is the General Manage-
ment Office for dams, hydraulic and electrical infrastructures, inside of the Ministry
5
Q.106 - R.1
Fig. 3
24th August 2016 6.0 Mw “Amatrice” earthquake. Shakemap (PGA) and areas with
dams involved in safety checks.
24 août 2016 Séisme 6.0 Mw “Amatrice”. PGA et zones avec barrages impliqués
dans les contrôles de sécurité.
Fig. 4
30th October 2016 6.5 Mw “Norcia” earthquake. Shakemap (PGA) and areas with
30 Octobre 2016 Séisme 6.5 Mw “Norcia”. PGA et zones avec barrages impliqués
6
Q.106 - R.1
Fig. 5
18th January 2017 5.5 Mw “Rieti” earthquake. Shakemap (PGA) and areas with
18 Janvier 2017 Séisme 5.5 Mw “Rieti”. PGA et zones avec barrages impliqués
of Infrastructure and Transport. During the whole sequence, for each event with
Mw >4, the Directorate identified, in real time, dams to be checked according to its
own procedures, through an online tool database (Italian Large Dams Database).
So that, in a few moments, it allows to identify the dams involved and print reports
about their characteristics. Moreover, the tool provides immediately the contacts of
the professional technicians to be contacted for the safety supervision (Responsible
for the Owner, Dam Wardens, Responsible for the Authority). Shakemaps made
available in a short time by the National Institute of the Geophysics and Volcanol-
ogy, represented a further support for the organization of safety checks (fig. 3 to 5).
Shakemaps provide the range of variation of the maximum horizontal accelera-
tion on the ground (PGA), recorded by the accelerometric sensors of the national
accelerometric network.
2.3. PPROCEDURES AND EXTRAORDINARY INSPECTIONS
During the seismic sequence, the regulatory national instrument in force was
the PCM/SDI/3536 (01/07/2002) that provided the activation of extraordinary checks
for seismic events of magnitude greater than 4 and for areas gradually increasing
with Magnitude. As this protocol, for the events of Magnitude between 6 and 6.5, the
area with large dams to be immediately controlled by the Owners had radius between
7
Q.106 - R.1
Fig. 6
18th January 5.5 Mw “Rieti” earthquake. Poggio Cancelli embankment dam, only
6 km far from the epicenter.
18 Janvier 5.5 Mw Séisme “Rieti”. Barrage de Poggio Cancelli, à seulement 6 km
de l’épicentre.
80 and 100 km. In these conditions, the Owner, leaded by Responsible Engineers
of the dams, could have to carry out extraordinary inspections of dams and their
appurtenant works (spillways with mechanical tests, banks, buildings, etc.) and to
perform specific detailed checks based on the monitoring system for many dams at
the same time. The results of checks have to be reported as soon as possible to the
Dam Authority and summarized in an extraordinary asseveration of safety conditions
signed by Responsible Engineer, according to the Italian Law. After the analysis of
the results of extraordinary asseveration, the Authority organized a safety control
campaign of both engineering and geological nature.
The large dams affected by the seismic sequence were a total of 40, and more
than 150 extraordinary asseverations were acquired by the Authority, that conducted
at least one extraordinary inspection audit in situ for each dam. For dams closer
to the epicentral area or showing deviations from normal behavior, were particu-
larly focused on and inspected more than one time (Scandarello, Fiastrone, Poggio
Cancelli, Polverina, Comunanza and some others).
The experience of 2016/2017 about the safety checks allowed to test the acti-
vation protocol of 2002, showing its limits with a stressing long seismic sequence.
On November 2017, the Italian Dam Authority decided to update this protocol, bet-
ter based on seismic hazard and vulnerability of large dams. The updating faced
the issue of distinguishing among dams affected by strong motion events and dams
less involved in terms of intensity. Shakemaps guided and supported this kind of
development. The new and current protocol sets two control levels with increasing
severity, according to the PGA recorded in situ or estimated trough the latest GMPEs
Attenuation Relations (fig. 7).
8
Q.106 - R.1
Fig. 7
Current activation thresholds for safety controls of large dams after earthquake.
Seuils d’activation actuels pour les contrôles des grands barrages
après un séisme.
The two levels of safety controls are different from both the information flow
and the level of detail. Generalizing, both of them have an activation phase, a first
feedback in 3 hours based on immediate visual controls, reported into checklists and
a feedback in 12 hours based on simple instrumental response too. Level 2 controls
require in addition an extraordinary asseveration including the analysis of charts
of the main measurements and, where a dynamic monitoring system is installed,
an interpretation through numerical or analytic models of dam’s dynamic response.
Dynamic monitoring systems provide both developments in model updating issues
and structural heath monitoring.
3. PERFORMANCE OF THE DAMS
3.1. ACCELEROMETRIC RECORDINGS – DYNAMIC RESPONSE
In Italy, the main descriptive parameters of earthquakes are acquired, pro-

cessed and published by the Department of Civil Protection, by means of the national
accelerometric network (RAN).
9
Q.106 - R.1
Fig. 8
Recorded Peak Ground Accelerations and dam distance.
Peak Ground Accélérations enregistrées et distance du barrage.
During the seismic sequence, the accelerometric stations closest to dams

recorded important acceleration peaks (PGA) in the horizontal components, even
close to the value of 1 g (fig. 8).
For the four main seismic events of 2016/2017 sequence, three dams were
close both to the epicenter and to the accelerometric monitoring stations: Scan-
darello dam (gravity dam), Fiastrone dam (arch-gravity dam) and Poggio Cancelli
dam (embankment dam).
The triaxial accelerometer station of Amatrice (AMT), located less than 2 km

away from the Scandarello dam, recorded PGA of almost 1 g in the east/west com-
ponent and acceleration peaks in the vertical direction up to 0.4 g in correspondence
of the 24th Aug. 6.0 Mw “Amatrice” event. The response spectrum elaborated by the
Department of Civil Protection of the accelerometric signal in east/west direction
filtered with a band pass filter between 0.2 and 50 Hz shows the greatest spectral
amplifications (over 2 g) between 0.1 and 0.4 seconds (fig. 9). The frequencies of
the first three modes of the Scandarello dam, determined by both numerical models
and dynamic identification processes on environmental noise, vary between 6 and
9 Hz. Although Scandarello dam seems to have experienced phenomena of spec-
tral amplification, it must be highlight that the AMT station is not installed on a rigid
reference and the soil category is classified of type B, according to Eurocode 8.
The triaxial accelerometer station of Poggio Cancelli (PCB), located at the

base of the Poggio Cancelli dam, recorded PGA of about 0.5 g in the north/south
component and peaks of acceleration in the vertical direction up to 0.2 g in corre-
spondence of the 18th Jan 5.5 Mw “Rieti” event. The response spectrum elaborated
by the Department of Civil Protection of the accelerometric signal in east/west direc-
tion filtered with a band pass filter between 0.2 and 50 Hz shows the greatest spectral
amplifications (up to 2 g) between 0.1 and 0.2 seconds. As the frequencies of the
first three modes of the embankment Poggio Cancelli dam vary between 2 and 4 Hz,
the dam seems to have suffered amplification phenomena, even though the station
10
Q.106 - R.1
Fig. 9
24th Aug 6.0 Mw “Amatrice” earthquake. Horizontal acceleration time history
recorded by AMT Station and response spectrum (semi-logarithmic scale).
24 Août 6.0 Mw Séisme “Amatrice”. Historique du temps d’accélération horizontale
enregistré par la station AMT et spectre de réponse (échelle semi-logarithmique).
is installed on a soil classified of type E, according to Eurocode 8, as shown by a

specific recent survey.
3.2. EVIDENCES FROM INSPECTIONS
As mentioned in the par. 2.3, 40 dams were involved in the 2016–2017 seismic
sequence. Visual inspections were carried out for each dam, following each of the
11
Q.106 - R.1
Fig. 10
18th Jan 5.5 Mw “Rieti” earthquake. Horizontal acceleration time history recorded
by PCB station and response spectrum (semi-logarithmic scale).
18 Janvier 5.5 Mw Séisme “Rieti”. Historique du temps d’accélération horizontale
enregistré par la station PCB et spectre de réponse (échelle semi-logarithmique).
more than 50 earthquakes with magnitude Mw > 4. The first important conclusions of
these inspections were that no significant effects or damages were recorded either
on dam body or on the banks.
Some small evidences of damage were observed only for the few dams closer
to the epicenters, generally on the crest/parapet with an accentuation of the existent
cortical cracking pattern or the developing of slight sub-millimetric new cracks. This
kind of damages were observed after the 24th Aug. 6.0 Mw earthquake on Scan-
darello Dam (fig. 11), a 45.15 m high gravity concrete dam, built between 1921 and
1927, owned by Enel Company and located in the Municipality of Amatrice, only
7 km away from the epicenter.
Moreover, an evident effect was observed after the 30th Oct. 6.5 Mw earth-
quake on the Fiastrone Dam, a 81.50 m high arch-gravity concrete dam, built
12
Q.106 - R.1
Fig. 11
Scandarello Dam. Slight damages on the parapet and crack pattern on the crest.
Barrage de Scandarello. Légers dommages sur le parapet et fissures sur la crête.
between 1950 and 1955, owned by Enel and located in the Municipality of Fias-
tra (Macerata Province), 25 km away from the epicenter. The historical seepages,
observable on the downstream face when the reservoir is full, appeared increased
after the earthquake (fig. 12). The seepages are limited on the arch upper part of
the dam, where the drainage system is absent.
3.3. INSTRUMENTAL EVIDENCES
During all the seismic sequence, the Owners normally raised the level of
attention and the frequency of measurements in order to capture the main effects
and study the dam behavior, assuring periodically the citizenry about the safety
conditions.
Just the static monitoring system installed on the dams gave the most signifi-
cant evidences of the earthquakes. For examples, for the concrete dams closest to
the epicenters, millimetric concentrated displacements were observed, especially
13
Q.106 - R.1
Fig. 12
Fiastrone Dam. Increased seepages on the downstream face after 30th Oct 6.5 Mw
earthquake.
Barrage de Fiastrone. Augmentation des percolations sur la face aval après le
séisme du 30 octobre 6.5 Mw .
in the downstream-upstream direction. The deviation with respect to the normal

behavior was only in the first measures after the major earthquakes.
Besides, several dams registered temporary increases of the uplift and

seepages, normally recovered in the following days. One of the most important
case history is Scandarello Dam, specifically for the filtration process under the
foundation.
14
Q.106 - R.1
Fig. 13
Scandarello Dam. Correlation diagram between reservoir water level and internal
seepages before and after the major earthquakes, until nowadays.
Barrage de Scandarello. Diagramme de corrélation entre le niveau d’eau du
réservoir et les percolations internes avant et après les grands séismes, jusqu’à
nos jours.
Since it was built, Scandarello dam has always been interested by a signifi-
cant seepages phenomena, especially through the foundation and right abutment,
reduced during the life of the dam by means of grout curtains (main in 30s and
90s). When, at the end of 2002, a new drainage system into the foundation was
realized for effectively reducing the uplift pressure, seepages consistently increased
up to 10–12 l/s at maximum water level.
As shown in the diagram of figure 13, the days-after-24th August measure-

ments pointed out an important increase (about 8–10 l/s) of the internal seepages,
partially due to the efficiency of the drainage system. After the first earthquake, fol-
lowing the reservoir emptying, the seepages decreased, trying to recover a normal
behavior. The 30th October 6.5 Mw earthquake caused another slight increase of
seepages (about 3.5 l/s), that slowly reduced during the next months, keeping low
water levels. In 2019, maintenance works were carried out on the upstream face,
contributing to the progressive reduction of the seepages rate, especially for reser-
voir water level less than 860 meters above sea level. However, right now, about half
of the first seismic increase has not yet been recovered.
Furthermore, other dams (Fiastrone Dam; Colombara Dam, owned by Enel

and Comunanza Dam owned by the Consorzio di Bonifica Marche) were interested
15
Q.106 - R.1
Fig. 14
Scandarello Dam. Direct pendulum, downstream-upstream displacements. The
measurements are blue dots-lines; the models are the red lines.
Barrage de Scandarello. Pendule direct, déplacements aval-amont. Les mesures
sont les lignes de points bleus; les modèles sont les lignes rouges.
by seepages increase of up to 3 times the normal value, which was very modest
anyway. The increases were fully recovered in the following days or weeks.
Moreover, Scandarello dam is provided of two pendulums in order to measure

both the radial (downstream-upstream) and tangential (right-left) displacements of
the crown cantilever, in a movable point located at the bottom gallery level.
For both pendulums statistical models were developed, calibrated considering

as a reference the period between 1st October 2009 (last day of the 2009 L’Aquila
Seismic sequence) and 4th August 2016 (latest day of measurements before the
Amatrice 6.0 Mw earthquake).
As shown in the diagrams of figure 14, for direct pendulum, the days-after-
24th August measurements (blue line) deviate from the model (red line), showing
16
Q.106 - R.1
Fig. 15
Comunanza Dam. Plano-altimetric displacements (block 5 and 20) and foundation
rockmeters measurements (block 6). Historical and after major seisms trends.
Barrage de Comunanza. Déplacements plano-altimétrique (bloques 5 et 20) and
mesures des rockmeters dans la fondation (bloque 6). Tendances historique et
après les grands séismes.
millimetric downstream displacements. On the contrary, the days-after-30th October

measurements recorded comparable upstream displacement, bringing themselves
back almost overlap the model line. After that, the seasonal behavior seems regular,
in good agreement with the model.
Another example of monitoring evidences comes from Comunanza Dam, a

76 m high gravity dam, owned by the Consorzio Bonifica Marche, 23 km far from the
30th Oct 6.5 Mw earthquake. Plano-altimetric topographic survey and foundation
rockmeters measurements showed an irreversible global setting displacement of
about 1–2 mm (fig 15).
Finally, also the concrete dams of Campotosto Lake (Rio Fucino, 37 m high,
and Sella Pedicate, 21 m high), particularly interested by the 18th January 2017
17
Q.106 - R.1
Fig. 16
Fiastrone Dam warden’s house with the evidence of the damages due to 24th
August 6.0 Mw earthquake (important crack patterns on all the walls).
Maison du gardien du barrage de Fiastrone avec la preuve des dommages dus au
séisme du 24 août 6,0 Mw (motifs de fissures importants sur tous les murs).
5.2–5.5 Mw earthquakes (7–10 km far from the epicenters), recorded some concen-
trated millimetric displacements.
3.4. APPURTENANT WORKS
The 2016–2017 seismic sequence caused some damages to the appurtenant

buildings of the dams involved. Most of these were built before the 60s’ when many
of the Italian municipalities had not yet been classified as seismic (mostly classified
after the 80s’). Most of the dam warden’s houses are mixed masonry buildings
with reinforced concrete horizons. The damages normally were slight and localized,
especially if compared with the other civil constructions of the countryside and towns.
Probably the mostly rigid foundation soils and the best quality of the masonry texture
succeeded in showing a better seismic response. The main evidences were the
manifestation of crack pattern of the walls or not structural elements.
In any case, the warden’s houses affected by the 2016/2017 sequence have
fulfilled the task of the safeguarding human life.
Major damages were observed for the Scandarello, Fiastrone and Comunanza
warden’s houses, that need important rehabilitation works (fig. 16) and for the Scan-
darella hydroelectric power station (fig 17), that was already restored with local
interventions.
18
Q.106 - R.1
Fig. 17
Scandarella hydroelectric power plant – Damages (short columns failed in shear
and out of plane mechanisms).
Centrale hydroélectrique de Scandarella – Dommages (les colonnes courtes ont
échoué dans les mécanismes de cisaillement et hors plan).
Finally, no anomalies were recorded in the electro-mechanical equipment,

as well as the related civil structures. Each gate was successfully operate by the
Owners immediately after all the earthquakes, in according to the check protocols.
4. ACTIONS AFTER SEISMIC SEQUENCE
In general, the 2016–2017 seismic sequence stressed the little or big defects
and deficiencies that some dams have already shown in their history. Therefore,
some activities were carried on and accelerated, in order to improve knowledge and
safety conditions of the dams. In the following, some of them are described.
4.1. MONITORING SYSTEM
After the earthquakes, automatic monitoring systems were installed in several

dams (Campotosto, Scandarello, Fiastrone, Borgiano, etc.) in order to measure the
most important static magnitudes in real time and remotely (laser sensor to measure
downstream-upstream displacement of the pendulum weirs, magnetostrictive level
sensors for seepages and pressure sensors for uplift). This is an important target to
have information immediately available, especially after significant earthquakes.
19
Q.106 - R.1
The Owners have in mind further instrumental improvements or new monitor-

ing system to be installed on other dams.
Moreover, in February 2017, about six months after the August 24th main
shock, the Civil Protection Department (DPC), in collaboration with the National
Dam Authority, with the support of ENEL, equipped Scandarello and Fiastrone dam
with a dynamic monitoring system. In addition, Provvidenza dam (a 46 m high arch
gravity dam) was recently provided with dynamic sensors (MEMS).
4.2. IN-DEPTH ANALYSIS AND SEISMIC RE-ASSESSMENT
In the context of seismic re-assessment of the Italian dams and appurtenant

works, the 2016–2017 seismic sequence gave a boost to this activity, not only for the
involved dam. First of all, several survey campaigns have been carried out in order
to investigate the concrete and rock foundation strength parameters (after several
years by construction), dynamic parameters, ageing process, etc.
Moreover, in-depth seismic studies have been collected, re-valuating the basic
seismicity, reviewing the earthquake catalogues and so on, in order to define the
correct seismic action at the dam site.
Afterwards, these data have been elaborated and used in the complex process
of the re-assessment of the safety conditions of the dams, in respect to the new
Italian Technical Law, enacted in 2014.
For example, for Comunanza Dam site both new Probabilistic and Determin-
istic Spectra were calculated with Seismic Hazard Analysis (PSHA and DSHA) and
compared with the one provided by Law, finally choosing the heaviest (fig. 18).
Then, geophysical surveys were carried out for characterizing both the dam
body and the rock foundation, with seismic refraction and sonic tomography in
upstream-downstream sections (fig. 19), used to evaluate the homogeneity of the
materials and verify the presence of possible weakness zones or fissures
Finally, numerical models of the dam-soil-reservoir system were defined,

modal analyses to describe the dynamic behavior of the system and linear time
history FEM calculations were performed for estimating the potential level of dam-
age (Demand Capacity Ratio Cumulative Inelastic Duration Analysis) in case of
a serviceability limit state earthquake scenario. Positive results were obtained for
Serviceability States (Fig. 20).
In addition, setting the extreme earthquake scenario, potential kinematics of

sliding or overturning non-linear analysis are still in progress, in order to verify the
absence of uncontrolled release of water.
20
Q.106 - R.1
Fig. 18
Comunanza Dam. Comparison of different spettra (horizontal component),
obtained with PSHA, DSHA and 2014 Italian Technical Law.
Barrage de Comunanza. Comparaison de différents spectres (composante
horizontale), obtenus avec PSHA, DSHA et loi technique italienne 2014.
Fig. 19
Comunanza Dam. Results of the sonic tomography of the main section of the dam.
Barrage de Comunanza. Résultats de la tomographie sonique de la section
principale du barrage.
21
Q.106 - R.1
Fig. 20
Comunanza Dam. DCR Analysis of some nodes in serviceability limit state
earthquake scenario.
Barrage de Comunanza. Analyses DCR de certains nœuds en le scénario d’état
limite de service.
4.3. INVESTIGATION AND WORKS
Some brief examples of other activities carried out on the dams involved (for
further information please refer to other papers or publications) are:
1) An important research about the Campotosto Fault, considered Active by the

most part of the Scientific Community nowadays. In 2017, The University of
Camerino (Macerata Province) coordinated the excavation of a large paleo-
seismological trench, in order to investigate about evidences of the fault activity
and its exact position, with particular reference to the distance from Rio Fucino
dam. Finally, the dam resulted about 300 m away from the fault, so that it seems
not to be interfered with.
2) At Fiastrone Dam, diagnostic investigations aimed to find possible preferential
filtration routes and evaluate the state of conservation of the concrete with
particular regard to the upper part of the dam, subject to the increases in
permeation, observed on the downstream face after the 30th Oct. 6.5 Mw
earthquake.
3) Extraordinary maintenance works at Scandarello Dam, also to repair minor
earthquake damages, disconnecting the parapets, waterproofing the crest,
repairing buildings, etc.
4) At Scandarello Dam, in order to definitively solve the filtration problem that
affect the dam since it was built, Enel is now designing an important improve-
ment of the foundation sealing. It will consist in a new deep grout curtain in
the foundation rock, both in the right and central part, that are responsible of
over than 90% of the seepages.
22
Q.106 - R.1
Fig. 21
Scandarello hydroelectric power plant – Local works made using steel closed
frame and steel chains.
Centrale hydroélectrique de Scandarello – Travaux locaux réalisées à l’aide d’un
châssis fermé en acier et de chaînes en acier
Fig. 22
The Poggio Cancelli appurtenant work rebuilt project, using XLAM timber
technology.
Le projet de reconstruction des ouvrages annexes de Poggio Cancelli, utilisant la
technologie du bois XLAM.
4.4. SEISMIC REASSESSMENT OF THE APPURTENANT WORKS
The Italian technical law for the design and construction of dams was updated
in 2014 (NTD2014). The appurtenant works are considered directly related to
the safety and functionality of the dam including also “the interventions of water-
proofing and consolidation of the banks of the reservoir, the monitoring systems,
the alarm and lighting systems, the warden house, the service roads and the
collecting/supplying works”.
23
Q.106 - R.1
The seismic safety reassessment process of large dams started in Italy during
the 2000’s as a legal obligation for the Owners, due to the variation of seismic
classification of the Italian territory in 2003. The technical standards in force at that
time referred to a 1982 law. Now, the NTD2014, combined with the Italian technical
law for the design and construction of civil structures (NTC2018), comprehends all
the elements for a modern reassessment of the existing large dams and appurtenant
works, both considered “strategic” in terms of actions to be applied and safety factors
to be guarantee.
Among appurtenant works, many of them are similar to civil buildings (war-
den’s houses, gate control houses with electromechanical equipment), others are
particular structures for the operation of the dam (outlets, surface spillway, morning
glory spillway) or connected to the dam (bridges, galleries, channels, etc.) not so
easy to be modelled and analyzed.
The post-earthquake assessments and the seismic vulnerability studies high-

lighted in many cases vulnerabilities of the works similar to buildings, linked to local
mechanisms or, for reinforced concrete constructions, to shear breakings. In that
case, when the numerical analyzes show the not fulfillment of an ultimate limit state
(SLV/SLC), the NTD 2014 require at least improvement interventions.
The Owners, after seismic numerical analysis and technical-economic con-

siderations, in most case chose to reinforce the existing buildings by a set of local
interventions aimed to improve the global behavior of the appurtenant work (fig. 21).
In few cases, they chose to demolish and rebuild new simple structures, sometimes
through innovative materials and technologies (fig. 22).
REFERENCES
[1] Luzi L, Puglia R, Russo E & Orfeus WG5 (2016) Engineering Strong Motion
Database.
[2] Website Rete accelerometrica nazionale – ran.protezionecivile.it
[3] Website INGV shakemap.rm.ingv.it
[4] Paolo Zimmaro and Jonathan P. Stewart. Engineering Reconnaissance of the
24 August 2016 Central Italy Earthquake. Version 2. Geotechnical Extreme
Events Reconnaissance Association Report No. GEER-050B. 22 November
2016.
[5] Paolo Zimmaro and Jonathan P. Stewart. Engineering Reconnaissance fol-
lowing the October 2016 Central Italy Earthquakes. Version 2. Geotechnical
Extreme Events Reconnaissance Association Report No. GEER-050D. 08
May 2017.
[6] M. Dolce et Al. Italian structural seismic monitoring network. Bull Earthquake
Eng., vol.15, Issue 2, February 2015.
24
Q.106 - R.2
-------
-------
ENEL EXPERIENCE OF LATEST YEARS IN

CENTRAL ITALY DAM MONITORING THROUGH
NEW TECHNOLOGIES (∗)
Massimiliano CARCIONE, Rosella CARUANA,

Matteo SBARIGIA & Fabrizio FRASCA
ENEL GREEN POWER ITALIA srl –
Dams & Civil Infrastructures Safety
ITALY
SUMMARY
This paper deals with the experience in dam monitoring through new technolo-
gies, performed by Enel Green Power Company in many sites located in Central Italy
during latest years.
The traditional monitoring system, applied to the majority of Enel dams,

is based on analogic or digital measures directly obtained in situ by the oper-
ators, according to the Dam Operation and Maintenance Document. Through
high-precision mechanical and optical instruments, in some cases robotized, a solid
dataset on dam behavior has been achieved over the years.
However, especially during seismic sequences, both instant and remote avail-
able data are necessary in order to evaluate rapidly the dam behavior and the safety
conditions. Therefore many dams, involved in recent 2016–2017 Central Italy Seis-
mic sequence, are now equipped with automatic monitoring systems to control with
∗ Expérience de l’ENEL acquise dans les dernières années en Italie centrale en matière
d’auscultation de barrages utilisant de nouvelles technologies.
DOI: 10.1201/9781003211808-100 25
Q.106 - R.2
an higher frequency the main monitoring parameters, such as displacements and

seepages of the structure.
Some of the epicenter dams were also provided with dynamic monitoring sys-
tems consisting of cabled or Wi-Fi accelerometric sensors in both force-balanced
and MEMS technologies. In particular, the quakes registrations provided important
data of the dynamic structural behavior, site response effects and real waveforms
affecting the structure.
Finally, other specialist controls with new technologies, developed in the latest
years by Enel are represented by laser scanning, applied to hydraulic studies or
slope instability monitoring, and aerophotogrammetry applied to 3D modeling in
structural analysis and near-distance high quality inspections.
In conclusion Automatic, instant and detailed information data, checked by

manual, analogic or digital measurements, seems the best way to control the dam
behavior in both normal and extraordinary conditions.
The systems of data management and the professional skills required for their
analysis are transformed together with technological progress and allow to observe
in a better way the behavior of large works.
RÉSUMÉ
Le système de surveillance traditionnel, appliqué à la majorité des barrages

dépendant de Enel, est basé sur des mesures analogiques ou numériques directe-
ment obtenues in situ par les exploitants, selon le document d’exploitation et
d’entretien des barrages. Grâce à des instruments mécaniques et optiques de
haute précision, dans certains cas robotisés, un solide ensemble de données sur
le comportement des barrages a été obtenu au fil des ans.
Cependant, en particulier lors de séquences sismiques, des données

disponibles instantanées et distantes sont nécessaires pour évaluer rapidement
le comportement du barrage et les conditions de sécurité. Par conséquent, de nom-
breux barrages, impliqués dans la récente séquence sismique de l’Italie centrale
2016-2017, sont désormais équipés de systèmes de surveillance automatique pour
contrôler avec une fréquence plus élevée les principaux paramètres de surveillance,
tels que les déplacements et les fuites.
Certains des barrages proches de l’épicentre ont également été équipés de

systèmes de surveillance dynamique constitués de capteurs accélérométriques
câblés ou Wi-Fi dans les technologies à équilibrage de force et MEMS. En particulier,
les enregistrements de tremblements de terre ont fourni des données importantes
26
Q.106 - R.2
sur le comportement dynamique de la structure, les effets de la réponse du site et

les formes d’onde réelles affectant la structure.
Enfin, d’autres contrôles spécialisés avec de nouvelles technologies, dévelop-

pés ces dernières années par Enel sont représentés par le balayage laser,
appliqué aux études hydrauliques ou à la surveillance de l’instabilité des pentes,
et l’aérophotogrammétrie appliquée à la modélisation 3D dans l’analyse structurelle
et les inspections de haute qualité à distance.
En conclusion des données d’information automatiques, instantanées et

détaillées, vérifiées par des mesures manuelles analogiques ou digitales, semblent
être le meilleur moyen de contrôler le comportement des barrages dans des con-
ditions normales et extraordinaires. Les modalités de gestion des données et les
compétences professionnelles requises pour leur analyse se transforment avec le
progrès technologique et laissent entrevoir de nouveaux scénarios d’observation du
comportement des grands ouvrages.
1. INTRODUCTION
In recent years, modern technologies have been developing for dam monitor-
ing, introducing new and important opportunities.
The latest seismic sequences occurred in 2009 and 2016–2017 in Central Italy
accelerated the application of these technologies to Enel Green Power dams.
Therefore, the long history of monitoring applied to all of the Enel Green Power
(first Italian Company of electricity) dams and mainly based on the same measure-
ments and observations, obtained directly on site by surveyors, is now flanked by
measurements based on robotic or remotely controlled technologies.
Technological evolution allows to obtain a considerable quantity of high quality

information which require important analysis and interpretation effort
2. TRADITIONAL AND ROBOTIC TOPOGRAPHIC

MONITORING SYSTEM
Collimations, leveling and pendulums historically represented high precision

measurement systems of planimetric, altimetric and plano-altimetric deformations
of dams respectively.
27
Q.106 - R.2
Fig. 1
In recent decades, the collimation measurements show a slight
drift upstream of the dam shown by collimation measurements
in latest years
Au cours des dernières décennies, les mesures de collimation
montrent une légère dérive en amont du barrage illustrée par les
mesures de collimation des dernières années
In some cases the correlation between these measurement systems and

the analysis of the global deformation of the dam body could be difficult. This
problem can be attributed to the displacement of the hypothetical fixed reference
points caused by abutment strain or exceptional events such as earthquakes, thus
causing doubt on the real movement of the structure.
In spite of redundant measurement systems, the low frequency of manual

acquisition and systematic errors can also determine doubts about measurement
drift. In these cases, triangulation measurements obtained by means of robotic
total station has been implemented in order to avoid the existence of ongoing
deformations.
This is the case of the Provvidenza dam (a 46 m high arch gravity dam), located
in the territory of L’Aquila city, which was heavily damaged by the 2009 earthquake.
In recent decades, the collimation measurements show a slight upstream drift of the
dam. Specific laboratory tests on the structural materials excluded the presence of
alkali-aggregate concrete expansion phenomena.
Thus, a new triangulation system was installed to control the movements of

targets founded on deep piles and located on the crest and both abutments. The time
series recorded by a robotic total station is still too short; anyway it shows a relative
movement of the target in the left abutment. A hypothesis of a long-term geodetic
closing displacements of the abutments, which causes rigid upstream rotation of the
dam, is currently under discussion.
28
Q.106 - R.2
Fig. 2
Topographic monitoring system on Provvidenza Dam; blue lines represent
collimation measurements lines; green lines and points represent respectively
triangulation measurements lines and targets; red points are common targets for
triangulation and collimation systems.
Système de surveillance topographique du barrage de Provvidenza; les lignes
bleues représentent les lignes de mesures de collimation; les lignes et les points
verts représentent respectivement les lignes de mesures de triangulation et les
cibles; les points rouges sont des cibles communes pour les systèmes de
triangulation et de collimation
3. TRADITIONAL AND ROBOTIC TOPOGRAPHIC

MONITORING SYSTEM
The traditional monitoring system, that is applied to all the Enel Green
Power dams, is based on manual or analogic measures, according to the Dam
29
Q.106 - R.2
Fig. 3
Comparison between manual and automatic seepage
measurements
Comparaison entre la poignée et les mesures automatiques
d’infiltration
Operation and Maintenance Document approved by the Ministry of Infrastructure

and Transport.
Manual or analogic data acquisition, however appropriate for long term

evaluation of displacement trends, in the short term can be inaccurate.
Low frequency in acquisition and different sensitivity of the operators can only
partially reproduce the effects of earthquakes and the details in seasonal variations
and correlation with other parameters.
During the 2016-2017 Central Italy seismic sequence, new sensors were
installed and calibrated at the Scandarello dam (45.15 m high gravity concrete dam),
located in the territory of Amatrice town.
Although the best fit between manual and automatic measurements can be
satisfactory, it is evident that the automatic measurement with a higher acquisition
frequency better reproduces the real behavior of the parameter over time by captur-
ing the effects of impulsive phenomena such as a Mw = 4 aftershock (03 December
2017 quake), shown in the graph of figure 3.
Especially in emergency, the possibility of having an automatic and instant

information data gave pulse to the installation of automatic acquisition systems at
30
Q.106 - R.2
Fig. 4
Magnetostrictive sensors installed in 2017 at Scandarello
dam to measure in continuous the internal seepages
through dam body and foundation.
Capteurs magnétostrictifs installés en 2017 au barrage de
Scandarello pour mesurer en continu les infiltrations internes à
travers le corps et les fondations du barrage
the Enel Green Power dams located on Chienti (Polverina, Borgiano, Fiastrone
dams), Vomano (Rio Fucino, Poggio Cancelli, Sella Pedicate dams) and Tronto
(Scandarello dam) rivers in the Central Italian Regions.
In particular leakages, pendulums wire displacements, joints opening, tilting,

piezometric measures are now obtained by magnetostrictives, vibrating wire, mems
and other sensor types (fig. 4); they are conveyed to dataloggers remotely connected
to databases (fig. 5).
In parallel and at lower frequency, manual measurements continue to be car-

ried out to identify any instrumental drift of the sensors which require constant
maintenance.
31
Q.106 - R.2
Fig. 5
Datalogger installed in 2017 at Scandarello dam to collect data of the sensors
Datalogger installé en 2017 au barrage de Scandarello pour collecter les données
des senseurs
4. DYNAMIC MONITORING
In recent years, dynamic monitoring has gained increasing importance in the

field of dams.
The growing use of accelerometers is linked to the strong need to understand

and verify the dynamic response of large dams in high seismic areas.
In addition to automatically activating control procedures and providing initial

information on the acceleration undergone by the structure, this monitoring activity
allows the construction over time of a database of natural accelerograms to be used
in the context of seismic structural design upgrade of the dams.
In the sector of local seismic response studies, dynamic monitoring allows

to obtain valuable information on the possible differential behavior in foundation to
dynamic excitation.
In Central Italy, Enel Green Power currently owns four dams equipped with
accelerometers, all located in high seismicity areas. The oldest dynamic monitoring
is located on the Talvacchia Dam on Tronto River and is represented by a wired
system of high quality force-balanced accelerometers. This system allowed to record
the strong Mw 6.5 Norcia earthquake occurred on 30 October 2016.
32
Q.106 - R.2
Fig. 6
Force-balanced accelerometric system on Talvacchia Dam;
on the left the Mw 6.5 “Norcia Earthquake” recorded by the base dam
triaxial accelerometer.
Système accélérométrique à force équilibrée sur le barrage de Talvacchia;
à gauche le Mw 6,5 “Norcia Earthquake” enregistré par l’accéléromètre
triaxial du barrage de base
Fig. 7
Scandarello Wi-Fi accelerometric system: amplification caused by 03 Dec 2017
aftershock (Mw = 4) on abutment is higher than in the middle of the crown
(elaboration of MIT-Dam Department).
Système accélérométrique Scandarello Wi-Fi: l’amplification provoquée par la
réplique du 03 décembre 2017 (Mw = 4) sur l’épaule est plus haute qu’au milieu de
la couronne (élaboration du département MIT-Dam).
33
Q.106 - R.2
Fig. 8
Comparison between Mems and Force balanced accelerometer Power Spectral
Density on Campotosto Lake (Sella Santa Maria dam); a good fit is observed in
frequency range of structural interest.
Comparaison de Mems et Force l’accéléromètre équilibré sur le lac Campotosto
(barrage de Sella Santa Maria); un bon ajustement est observé dans la gamme de
fréquences d’intérêt structurel.
During the same seismic sequence, Scandarello and Fiastrone dams were
equipped with force-balanced accelerometers capable of transferring the data to
a datalogger in Wi-FI mode. They gave important information on local seismic
response. In particular on Scandarello dam a higher amplification can be observed
in the right abutment than on the central crown sector and at the structure base;
this anomaly is due to worse geomechanical characteristics on the right abutment
foundation.
An experiment consisting in a contemporary double acquisition of strong

aftershocks, conducted by Enel Green Power during the seismic sequence at the
Campotosto reservoir, made it possible to demonstrate that the modern MEMS tech-
nology allows to obtain seismic recordings and spectra of a quality similar to that
obtained through force-balanced models. This experimental test on real earthquake
recordings paved the way for a project to install a MEMS accelerometric network at
the Provvidenza dam, located near L’Aquila town.
5. LASER SCANNING, AEROPHOTOGRAMMETRIC SURVEYS AND

MONITORING
In latest years, the progress in laser, photogrammetry and sonar technology

has been remarkable. The laser scanning technique has been increasingly used on
34
Q.106 - R.2
Fig. 9
Pontefiume dam: Morpho-bathymetric surveys, obtained through aerial Lidar and
multibeam sonar, at the basis of hydraulic simulations.
Barrage de Pontefiume: levés morpho-bathymétriques, un lidar aérien et un sonar
multibeam, à la base de simulations hydrauliques.
Fig. 10
Piaganini dam: Rock sliding simulation applied to Laser scanning Digital Elevation
Model and Volume Rock Unit analysis.
Barrage de Piaganini: simulation de glissement de roche appliquée à l’analyse
laser du modèle d’élévationnumérique et de l’unité de volume de roche.
35
Q.106 - R.2
Fig. 11
Top Picture: High precision aerophotogrammetric combined with laser scanning
survey executed on Barrea dam-abutments system. Bottom Picture: San
Domenico Dam Wireframe 3D modelling for structural analysis obtained from UAV
survey.
Top : Aérophotogrammétrie de haute précision combinée à un levé par balayage
laser exécuté sur le barrage de Barrea- système d’épaules. Bottom: Modélisation
3D filaire du barrage de San Domenico pour l’analyse structurelle obtenue à partir
d’un levé UAV
36
Q.106 - R.2
Fig. 12
Fiastrone dam: high precision drone orthomosaic and high resolution details of
joints seepages.
Barrage de Fiastrone: orthomosaïque de drone de haute précision et détails haute
résolution des infiltrations de joints.
Central Italy Enel dams, starting from large-scale aerial surveys (LIDAR) aimed at
obtaining three-dimensional surfaces combined with bathymetric multibeam sonar
surveys for more precise hydraulic simulations (Pontefiume and Vulci dams).
However it is at an engineering scale (terrestrial survey) that the technique had

the greatest applications, especially in the field of studies of instability of rock slopes.
Interesting geomechanical studies, trough modern laser scanning applications, were
executed on Furlo and Piaganini dam-abutments system, to analyze dangerous
rock-mass volume, rolling or sliding paths and slope protection design.
Aerophotogrammetric technique trough UAVs (drones) has been more and

more used, also in coupling with terrestrial laser scanning, in the field of 3D mod-
elling. It is aimed to updating the consistency of the works and comparing with
historical drawings especially in structural design upgrade. Interesting results were
obtained on San Domenico and Barrea dam in Abruzzo Region.
Another important application recently experienced is the near-field defect

survey (Fiastrone and Talvacchia dam) through drone orthophotographic technique.
Georeferencing, through visible targets located on dam facing, made it possible to
obtain high resolution orthomosaic to support visual and time-repeatable inspections
especially on joints seepages.
37
Q.106 - R.3
-------
-------
STABILITY OF THE ROCKY SLOPES FACING LARGE RESERVOIRS.

INNOVATIVE VERIFICATION AND RISK ASSESSMENT APPROACHES
USING 3D POINT CLOUD ANALYSIS – THE CASE OF LA PENNA DAM
(AREZZO, TUSCANY, ITALY) (∗)
Marco Bersano BEGEY, Roberto BERTERO

HYDRODATA
Iacopo BROGI
ENEL
ITALY
SUMMARY
Rock slopes stability is one of the most relevant risk factors for the management
of dams and large reservoirs. The instabilities for collapse and topple represent
insidious phenomena, as they are generally sudden and rapidly developing, able to
mobilize large masses, so related to abnormal wave formation as well as possible
direct damage to the structures.
On the other hand, the security checks are complex and often not reliable, as
they require a detailed assessment of the discontinuities in the rock mass, which is
difficult to obtain if not at the level of punctual surveys and where compatible with
the conditions of safe access.
∗ Stabilité des pentes rocheuses des grands réservoir – approches innovantes de vérification
et d’évaluation des risques à l’aide de l’analyse des nuages de points 3D – le cas du barrage
de La Penna (Arezzo, Toscane, Italie)
38 DOI: 10.1201/9781003211808-101
Q.106 - R.3
The recent improvements of survey technologies by laser scan and Drone –

UAV photogrammetry allow to obtain extremely detailed 3D point clouds, to support
the analysis of discontinuity systems and joints predisposing instabilities.
This paper presents the application to the case of the ENEL Company dam of
La Penna, on the Arno River (Arezzo, Italy), of advanced technologies for the survey
by UAV, and the analysis, filtering and clustering of 3D point cloud data, aimed at
security checks for rock fall/topple, related to the possible direct exposure of plant
structures as well as the generation of hazardous waves.
RÉSUMÉ
La stabilité des pentes rocheuses est l’un des facteurs de risque les plus impor-
tants pour la gestion des barrages et des grands réservoirs. Les instabilités dues à
l’éboulement et au basculement représentent des phénomènes dangereux car ils
sont généralement soudains et à développement rapide, capables de mobiliser de
grandes masses, donc liés à la formation d’ondes anormales ainsi qu’à d’éventuels
dommages directs aux structures.
En revanche, les contrôles de sécurité sont complexes et souvent peu fiables,

car ils nécessitent d’une évaluation détaillée des discontinuités dans la masse
rocheuse, difficile à obtenir sauf au niveau des levés précis et lorsque cela est
compatible avec les conditions d’accès sécurisé.
Les améliorations récentes des technologies de levé à l’aide de scanners

laser et de photogrammétrie par drone permettent d’obtenir des nuages de points
3D très détaillés, pour aider à l’analyse des systèmes de discontinuité et des joints
qui prédisposent à l’instabilité.
Ce rapport présente l’application au cas du barrage de la société ENEL “La

Penna” sur le fleuve Arno (Arezzo, Italie), des technologies avancées de levé par
drone et l’analyse, le filtrage et le regroupement des données du nuage de points
3D, destiné aux contrôles de sécurité d’éboulement/basculement de roches, liés à
une éventuelle exposition directe des structures ainsi qu’à la génération de vagues
dangereuses.
1. INTRODUCTION
Landslides from rock slopes are characterized by a sudden and rapid

development with dynamics for sliding, fall and topple.
These instabilities therefore constitute a serious risk for the management

of the reservoirs, related to possible direct damage to the dam’s structures,
39
Q.106 - R.3
anomalous waves, up to the overflow scenario. Development times of the phenom-

ena are usually incompatible with emptying operations.
It is therefore essential to forecast and elaborate scenarios as precise as pos-

sible regarding the stability of the rock faces and the volumes involved, as well as
the overall monitoring of the physical state of the rock mass.
Rock instabilities are usually caused by movements along the fracture joints
present in the rock mass. Therefore, all the methods for the stability verification and
the quantification of unstable volumes are based on the field survey of the joints’
orientation.
This information is classically investigated in tedious ?eldwork campaigns

using a geological compass (Fig. 3).
In the real world, obtaining a description of the rock mass adequate to the
objectives of modeling and forecasting the instability with the “classic” field method-
ology is almost impossible and the resulting models are highly approximate and
substantially unreliable. This because the “manual” survey only allows to obtain data
on punctual measurement stations, with a limited number of representative joints,
and a reduced coverage of the entire mass, usually characterized by significant local
heterogeneities.
The survey must also be carried out by personnel on the rope (mountain
guides), and this is often not possible for safety reasons, or it can only be carried
out in “safe” sectors, therefore not representative of the real risk conditions.
Recent advances in Light Detection and Ranging (LiDAR) instrumentation,

and Drone – UAV photogrammetry currently allow quick and accurate 3D data
acquisition, yielding on the development of new methodologies for the automatic
characterization of rock mass discontinuities.
These methodologies support the analysis of discontinuity systems and joints

predisposing instabilities on the basis of a complete coverage of the rock mass,
at a level of precision and reliability not comparable with “classic” field survey
methodologies, allowing the development of objective and reliable risk models for
the reservoir.
The remote survey approach of the rocky front at La Penna dam (Arezzo) is
described below. For reasons of synthesis, the subsequent stability analysis refers
just to a sample sector of the rock wall (“A” sector in Fig. 1), considering that the
methodology was applied to the entire area by the same criteria and technology.
2. LA PENNA DAM CASE STUDY
The La Penna dam, built between 1955 and 1957, on the Arno river, in the
municipality of Laterina Pergine Valdarno (AR), is a gravity dam, overflowing with
40
Q.106 - R.3
interposed gates. It has an arched shape (R = 139.3 m) and is characterized by the

following data (taken from the Maintenance Operation Conditions Sheet).
– height of the dam (pursuant to Ministerial Decree 24.03.82) 36.30 m

– height of the dam (pursuant to Law 584/94) 30.00 m
– maximum retention height 25.50 m
– crowning altitude 206.50 m a.s.l.
– development of the crown 101.00 m
– volume of the dam 42,900 m3
– reservoir original volume 13 Mm3
– classification pursuant to D.M. 24.03.82: A.a.1 (ordinary gravity in concrete)
The instability concerns the sector of the rocky slope on the hydrographic right
located at a distance between about 100 ÷ 240 m upward from the La Penna dam
(Fig. 1).
Fig. 1
Cliff images and profiles, total height is around 40 m over reservoir average level
Images et profils de falaises, la hauteur totale est d’environ 40 m au-dessus du
niveau moyen du réservoir
41
Q.106 - R.3
The slope, starting from the upper surface of the fluvial-lacustrine terrace
(ref. Profiles in Fig. 1) at an altitude of about 240 m, has a development of about 36 m
compared to the current maximum reservoir level (203.5 m a.s.l.). The observed col-
lapses, endemic over time, concern arenaceous lithotypes, alternating with marly
levels of the so-called geological Unit of Falterona, with the involvement of weak
thicknesses of fluvial-lacustrine deposits from the top portion.
Anticipating the elements under investigation in the survey, study and stability
analysis phases, the failure is on the whole linked in both sectors to:
– alternation of arenaceous banks to more erodible marly levels, with endemic

undermining at the base of the arenaceous banks themselves;
– presence of sub-vertical open fracture systems, which allow the subdivision
into blocks and their mobilization, following the erosion of the marly levels at
the base.
2.1. DISCONTINUITY REMOTE SURVEY
The discontinuity survey was carried out in the overall frame of the landslide
risk assessment for the La Penna reservoir and dam. The specific goal was the con-
struction of a model of the fractured rock mass supporting the stability analysis and
the quantification/localization of the volumes possibly unstable for collapse/topple.
2.1.1. 3D point cloud survey through Drone – UAV survey
On 09/24/19, on the occasion of partial emptying operations (198.5 m above

sea level), an aerial photogrammetric survey was carried out in the specific sector,
by Drone – UAV, performing a flight in manual mode (Fig. 2).
Images were taken in frontal shooting, with a grid as regular as possible,

bidirectional (ref. diagram Fig. 2), with a minimum overlap between adjacent frames
of about 75% in the longitudinal direction (“overlap”, flight direction) and at least the
50–60% lateral overlap (“overside”, between strips).
For the correct calibration (georeferencing) of the frames and of the point cloud,
a support survey was carried out on known points (GCP: Ground Control Point).
Targets and natural control points were detected with a Leica TS16 total station
equipped with a prismless distance meter, enabling a unique topographic reference
network to be created.
Pix4D software (Simactive, 2019) was used for the reconstruction of the 3D
model. The software searches for the points in common between the photos and
42
Q.106 - R.3
Fig. 2
Layout of the flight plan and frame acquisition. b) positioning of the
UAM operator and of the total station for surveying the support network,
on the La Penna dam
Schéma du plan de vol et acquisition des photos.
b) positionnement de l’opérateur UAM et de la station totale de
levé pour le réseau de soutien, sur le barrage de La Penna
proceeds to align them on the basis of the analytical relationships existing between
the points in common.
At the end of the processing, the program shows on the screen the points in
common (tie points) between the photos that it used for the alignment and that it will
use for the construction of the point cloud.
Subsequently, again in the Pix4D environment, we proceeded to frame the

cloud on the support topographic survey (ground target net), and subsequent geo-
referencing of the point cloud itself.
On the processed and filtered (excluding some vegetated areas) point cloud
we proceeded to create polygonal “mesh” surfaces, generating 3D model.
2.1.2. Survey of discontinuities, extraction of joints from the

3D point cloud
As previously mentioned, the survey of the discontinuities is usually carried

out with a geological compass, detecting the direction and immersion of the plane,
placing the cover of the compass on the surface itself (Fig. 3-b). Using a customized
version of the well-known CloudCompare software (EDF), it is possible to perform
43
Q.106 - R.3
Fig. 3
3D point cloud compass Vs. field compass
Boussole par nuage de points 3D Vs. boussole de terrain
the same measurement on the single plane starting from the 3D model of the rock
wall (Fig. 3-a)† .
However, the complete schematization of the discontinuities in a rock mass

requires the continuous and complete definition of all the planes, not always
identifiable from the visual observation of the different systems.
For this purpose, the DSE calculation code was tested, capable of “extracting”
from the digital model the alignments of points corresponding to the single joints,
grouping them in homogeneous clusters from the point of view of orientation.
Discontinuity Set Extractor – DSE, developed by Alicante University (E),

applies a semi-automatically methodology to identify points members of an unorga-
nized 3D point cloud that are arranged in 3D space by planes.
This method identifies and defines the algebraic equations of the different
planes of the rock slope surface by applying an analysis based on [1]:
a) the user-supervised removal of noisy points through the creation of a copla-

narity test;
b) the semi-automatic identi?cation of discontinuity sets using a Kernel Density
Estimation (KDE) Analysis;
c) the automatic extraction of single discontinuities through a density-based
clustering algorithm.
† CloudCompare (developed by Électricité de France EDF) is a point cloud processing soft-

ware. It has been originally designed to perform comparison between two dense 3D points
clouds. Afterwards, it has been extended to a more generic point cloud processing soft-
ware, including many advanced algorithms (registration, resampling, statistics computation,
sensor management, etc.).
44
Q.106 - R.3
Fig. 4
Three-dimensional model developed in CloudCompare environment
Modèle tridimensionnel développé dans l’environnement logiciel CloudCompare
2.2. ROCK MASS MODELLING AND DEFINITION OF UNSTABLE VOLUMES FACING THE
RESERVOIR
Fig. 5 shows the result of the processing (extraction and clustering of discon-
tinuity plans) carried out in the DSE environment on the sample zone. Analysis is
performed over a 2.8E + 6 points cloud model (Fig. 4). The classification by joint-set
is represented with a continuous chromatic scale (on the right in the image).
Considering the finalization of the stability checks, we opted for an initial clas-
sification of the joints in n. 10 “joint sets”, meaning a significantly higher number
than discontinuity systems normally used for structural analysis.
The above refers to the need, in the practice of stability analysis with a geome-
chanical criterion, to avoid the calculation with reference to a few average planes
representative of systems with strong dispersion, i.e. where stability conditions could
result different if calculated assuming the average joint orientation rather than the
extremes within the orientation range.
Fig. 6 shows the projection of the same joint systems on a polar diagram
(strike/dip [◦ ] reported in the table on the left).
From a general characterization level, the system J1 ÷ J2, consistent with the
position of the slope, J8 ÷ J9 the sedimentation plane (anti-dip slope), J4–J7, and
remaining, subvertical fracture joints.
45
Q.106 - R.3
Fig. 5
Projection of discontinuities sets defined by DSE (chromatic scale on the right,
stereographic projection on following Fig. 6)
Projection d’ensembles de discontinuités définis par DSE (échelle chromatique à
droite, projection stéréographique sur la Fig.6 suivante)
Fig. 6
Stereographic projection on polar diagram of systems 1 ÷ 10 (screenshot software
DSE; N.B. the software represents by default North/0◦ downwards and South/180◦
upwards) Projection stéréographique sur diagramme polaire des systèmes 1 ÷ 10
(capture d’écran du logiciel DSE; N.B. le logiciel représente par défaut le Nord/0◦
vers le bas et le Sud/180◦ vers le haut)
46
Q.106 - R.3
2.3. SLOPE STABILITY CHECKS, MARKLAND TEST
Starting from the 3D point cloud, the stability analysis was conducted at two
levels:
– from the orientation of the joint systems extracted with DSE, a preliminary
analysis with Markland Test, leading to the evaluation of the discontinuity
sets possibly predisposing collapse (for planar sliding, translation of isolated
wedges, topple);
– as for the instabilities related to unfavorable joint geometries (as shown by
Markland test results), a detailed analysis in the CloudCompare software envi-
ronment has been performed, through the 3D modeling of the specific joint sys-
tems for the definition of unstable blocks, defined by significant intersections
2.3.1. Markland test
The Markland test allows to verify intersections between the plans extracted
from point cloud, predisposing phenomena of instability, based on the following
standard conditions (Fig. 7):
a: If the representative orientation vector of a system of discontinuity falls within

the area indicated in gray (resulting: Discontinuity inclination > joint friction
angle φ, and Discontinuity inclination < slope front inclination), the condition
of potential planar failure exists.
b: If the intersection of the two representative circles of two systems of dis-
continuity falls within the gray area (area between the friction angle φ and
the angle of inclination of the face), the condition of potential breakage and
collapse of rocky wedges exists.
c: The potential topple condition exists if the orientation vector representative of
a system of discontinuity (central point of the arc of a circle in blue) falls within
the highlighted triangular sector.
Fig. 7
Stereographic projection showing the conditions for block mobilization on:
a) a planar surface, b: wedge failure, c: topple (Hoek and Bray, 1981, Watts 2003).
Projection stéréographique montrant les conditions de mobilisation de bloc sur:
a) sur une surface plane, b: rupture de coin,
c: basculement (Hoek et Bray, 1981, Watts 2003).
47
Q.106 - R.3
2.4. RESULTS
Results, as already said, are reported for the sample sector. Fig. 8 shows the
stereographic projection, on an equiangular Wulff diagram, of the fracture systems
identified through the analysis in the DSE environment from the point cloud. Joint
systems J9, J10 have been excluded since, from the single projection in the 3D
model, they represent punctual and discontinuous alignments.
From the observation of the diagram, and with reference to the scheme of the
stability verification with Markland test referred to Fig. 8, the following is noted:
– the direction and average immersion of the front (Slope) is consistent with the
position of the main system of joints (system 1);
– there are no predisposing conditions for breakages on planar surfaces, as
none of the single orientation vectors representing a system of discontinuity
(central point of the arcs corresponding to the single planes) falls within the
“critical zone” indicated in color;
– there are no predisposing conditions for wedge-shaped breaks/unstable
wedges; the two systems J7, J4, with intersection in the “critical zone” actually
constitute a single sub-vertical system, with dispersion of the planes on the
two opposite verges, and do not give rise to unstable wedges;
– there are no conditions predisposing to kinematics for topple.
Fig. 8
Stereographic projection of discontinuity systems and stability verification with the
Markland test
Projection stéréographique des systèmes de discontinuité et vérification de la
stabilité avec le test de Markland
48
Q.106 - R.3
Under the current conditions, the rocky front therefore does not present
unfavorable joint conditions, predisposing phenomena of instability of rocky prisms.
Conditions of instability can instead derive from differential erosion, also

induced by reservoir wave motion.
Erosion, in particular of the marly banks, determines the undermining along

the system of joints J8 (stratification plane). This can change stability conditions,
causing sub-vertical sliding of blocks delimited by systems J1, J4–J7 (Fig. 9).
Fig. 9
Surfaces (system 8, dark orange color) from undercutting due to the action of wave
motion, resulting in the creation of blocks sliding on the surfaces intersections
between systems 1 (blue), 4–7 (green)
Surfaces (système 8, couleur orange foncé) de contre-dépouille due à l’action du
mouvement des vagues, entraînant la création de blocs glissants sur les surfaces
intersections entre les systèmes 1 (bleu), 4–7 (vert)
Therefore, even in the presence of favorable conditions in geometry of the

joints, wave erosion can cause the release and fall of single rocky prism.
This means that:
– on the whole the slope is stable, no significant landslides are expected;

– isolated prism fall is possible, as far as wave erosion proceed.
The extent of the volumes involved in isolated falls, can be analyzed starting
from the projection of the individual fracture planes selected in the DSE environ-
ment starting from the 3D model, i.e. delimiting volumes between joint sets number
J1 and J4–J7. The extracted plans, imported into the CloudCompare environment,
can in fact be selected individually to define the volumes underlying the relative
intersections. In CloudCompare environment this can be calculated automatically,
or “cutting” cloud adding new planes. As for the sample area a plan has been
added parallel to J1 system, to consider the location of current traction fractures
(Fig. 10–11).
49
Q.106 - R.3
Fig. 10
Delimitation of unstable blocks based on stress fractures, on point cloud and
georeferenced image
Délimitation de blocs instables basée sur des fractures de tension, sur nuage de
points et image géoréférencée
Fig. 11
Delimitation of unstable blocks on 3D point cloud for intersection between tension
fractures, sub-vertical joints (systems 4–7) and undercutting surfaces due to wave
motion (along joint system 8)
Délimitation de blocs instables sur un nuage de points 3D pour l’intersection entre
les fractures de tension, les joints sous-verticaux (systèmes 4-7) et les surfaces de
dégagement dues au mouvement des vagues (le long du système d’articulation 8)
50
Q.106 - R.3
2.5. MONITORING BY SEQUENCE OF 3D POINT CLOUDS COMPARISON
The monitoring of a rocky front is generally performed measuring local

displacements with high resolution instruments (ex. extensometers, crackmeters).
These technologies allow for accuracies down to <0.1 mm and real-time acqui-
sition and transmission where needed, and are therefore often irreplaceable in the
case of high risk related phenomena.
However, the local survey with fixed instrumentation does not allow to visualize
the overall deformations of the rock mass, through a complete coverage of the front.
Otherwise this information can be obtained by comparing sets of DTMs from

different acquisition dates.
The measurement accuracy at the level of individual DTM points is of the

millimeter order, i.e. significantly higher than that of local surveying with fixed
instrumentation.
However, the 3D point cloud is made up of an extremely high number of points,

for which a random positioning error is assumed, based on the same topographic
ground network for all the different surveys. We therefore assume a compensation of
the overall error and a consequent high overall reliability of the comparison between
point clouds acquired on different dates.
As an example, Fig. 12 shows the comparison (using CloudCompare by EDF)

referred to two sets of points, which highlights the overall rotation of the rock block
counterclockwise along an approximately sub-vertical central axis.
Fig. 12
Typical result of a distance computation process
Résultat typique d’un processus de calcul de distance
51
Q.106 - R.3
The comparison between point clouds referring to different survey dates there-
fore constitutes a fundamental element of interpretation and control of the monitoring
data from the local instrumentation.
3. CONCLUSIONS
3.1. GENERAL CONSIDERATIONS ON THE APPLIED TECHNOLOGY AND ITS

POTENTIAL
Evaluating the stability conditions of the rock slopes of a reservoir is frequently

a challenging issue, due to the difficulty in obtaining a reliable geomechanical model
through the direct survey of the discontinuity joints. Moreover, the direct survey of
discontinuities performed by the operators (mountain guides suspended on high-
line) is frequently not feasible in safety and refers to local measurement stations,
just allowing a very rough evaluation of the stability conditions. This is often a cru-
cial issue for reservoirs management, since reliable scenarios of possible collapse
cannot be evaluated (as well as related level rise, waves, morphological changes).
The proposed methodology allows to obtain a reliable risk model starting from
the remote survey by Drone-UAV, the construction of the 3D point cloud and the
subsequent extraction, with a semi-automatic methodology, of single digital models
corresponding to the sets of discontinuity joints.
3.2. CASE RELATED CONCLUSIONS – LA PENNA RESERVOIR
The survey and modeling of the discontinuity joints had the purpose of evalu-
ating the possible volumes of rock that could fall into the reservoir, possibly causing
waves and level rise. In this paper, for purposes of synthesis, the analysis has been
limited to a sample sector of the rock face, representative of the methodology
applied on a larger scale.
The model has shown that there are no joint geometries that can determine
instability due to the collapse and overturning of rock prisms, i.e. the Markland test
for stability is always verified, in all sectors of the rocky front.
The instabilities observed, consisting in the collapse of isolated prisms, occur

only as a result of undercutting due to erosion of the marly levels, with the collapse
of the overlying arenaceous banks, largely determined by the action of the wave
motion and favored by the level fluctuations in the reservoir.
52
Q.106 - R.3
Therefore, problems of instability of large volumes were excluded, resulting

compatible with the joints geometry, only the development of “endemic” collapses of
blocks and rock portions, as the erosion of the marly levels proceeds. It is therefore
the same phenomenology repeatedly observed in the past, precisely “endemic”,
without significant evolution.
REFERENCES
[1] ADRIÁN J. RIQUELME ET ALII, “A new approach for semi-automatic rock

mass joints recognition from 3D point clouds”, Computers & Geosciences, 68
(2014) 38–52, Elsevier.
[2] GIREARDEAU-MONTAUT D., “CloudCompare presentation”, Proceedings of

the 2nd Virtual Geoscience Conference, Bergen, Norway. 21–23 September
201.
53
Q.106 - R.4
-------
-------
THE ITALIAN EXPERIENCE OF POST-EARTHQUAKE

INSPECTION AND SAFETY EVALUATION OF LARGE DAMS (∗)
Armando LANZI
Ministero delle Infrastrutture e dei Trasporti – Direzione
Generale per le dighe e le infrastrutture idriche ed elettriche
Maria Cristina BRAMATI & Rosella CARUANA

Enel Green Power Italia srl – Dams & Civil Infrastructures Safety
ITALY
SUMMARY
This paper discusses state of the practice of post-earthquake inspection and

control of large dams in Italy.
To define national homogeneous criterion for evaluating dams safety condi-

tions after earthquakes, in 2002 the Italian Dam Authority issued the first procedure
for post-earthquake inspections, adopting the criterion given in ICOLD Bulletin
n.62 “Inspection of large dams following earthquakes” to identify the screening
area depending on earthquake magnitude and source-to-site distance. Dams
located inside the screening area required immediate inspection and controls. The
procedure also provided details on the type and extent of controls and on reporting.
Since 2002 this procedure has been applied many times, gaining significant
experience particularly after the seismic events in Central Italy in 2009 and 2016–
2017, when many large dams were involved. Lessons learned over more than fifteen
∗ L’expérience italienne de l’inspection post-séisme et de l’évaluation de la sécurité des
grands barrages
54 DOI: 10.1201/9781003211808-102
Q.106 - R.4
years of its use induced the Italian Dam Authority to update the post-earthquake
inspection criterion.
A new procedure was therefore issued in 2018 and it is currently in use. The
new procedure introduced three major changes: first, it differentiates two types of
inspections depending on the severity of the shaking; second, it defines the screen-
ing areas on the basis of explicitly defined acceleration thresholds; third, it allows
consideration of real-time earthquake data recorded by seismic monitoring sys-
tems; when not available, the peak ground acceleration at the dam site is estimated
from the earthquake magnitude and source-to-site distance, with the most recent
ground motion prediction equations. With respect to the 2002 criterion, the new
procedure leads to a narrower screening area in case of low-to-medium magnitude
earthquakes.
Post-earthquake inspections are divided into Level 1 and Level 2, respectively

activated when the shaking at the site reaches 0.04 g and 0.07 g. The procedure
provides a list of controls to be performed for both types of inspection.
ENEL Company operates many large dams, and provided each of them with
a specific document to implement the procedure for post-earthquake controls. This
document, called Annex 2 to the Dam Control Plan, is briefly described into this
paper.
RÉSUMÉ
Ce rapport examine l’état de la pratique de l’inspection et du contrôle post-

séisme des grands barrages en Italie.
Pour définir un critère national homogène d’évaluation des conditions de sécu-

rité des barrages après un séisme, l’Autorité Italienne des Barrages a publié en 2002
la première procédure d’inspection post-séisme, en adoptant le critère donné dans
le Bulletin CIGB n.62 “Inspection des barrages après séismes – Recommandations”
pour identifier les zones de contrôle en fonction de la magnitude du séisme et de la
distance source-site. Les barrages situés à l’intérieur de la zone de contrôle doivent
faire l’objet d’une inspection et des contrôles immédiats. La procédure a également
fourni des détails sur le type et l’étendue des contrôles et sur les rapports.
Depuis 2002, cette procédure a été appliquée à plusieurs reprises, acquérant

une expérience significative, en particulier après les événements sismiques du cen-
tre de l’Italie en 2009 et 2016–2017, lorsque de nombreux grands barrages ont
été impliqués. Cette vaste expérience a conduit l’Autorité Italienne des Barrages à
mettre à jour le critère d’inspection post-séisme.
Une nouvelle procédure a donc été publiée en 2018 et est actuellement util-
isée. Elle a introduit trois changements majeurs. Premièrement, elle différencie
deux types de contrôles en fonction de la magnitude du séisme. Deuxièmement,
55
Q.106 - R.4
elle définit les zones de contrôles sur la base de seuils d’accélération explicitement
définis. Troisièmement, elle permet de prendre en compte les données sismiques
en temps réel enregistrées par les systèmes de surveillance sismique ; lorsqu’elles
ne sont pas disponibles, l’accélération maximale du sol sur le site du barrage est
estimée à partir de la magnitude du séisme et de la distance source-site, avec les
équations de prédiction de mouvement sismique du sol les plus récentes. Par rap-
port au critère de 2002, la nouvelle procédure conduit à une zone de dépistage plus
étroite en cas de séisme de faible à moyenne magnitude.
Les inspections post-séisme sont divisées en niveau 1 et niveau 2, respective-

ment activées lorsque l’accélération maximale sur le site atteint 0,04 g et 0,07 g. La
procédure fournit une liste des contrôles à effectuer pour les deux types d’inspection.
La Société ENEL exploite de nombreux grands barrages et a fourni au per-

sonnel exploitant chacun d’eux un document spécifique pour mettre en œuvre la
procédure de contrôle post-séisme. Ce document, appelé annexe 2 du Plan de
Contrôle du Barrage, est brièvement décrit dans ce rapport.
1. INTRODUCTION
In order to ensure both functionality of the water infrastructure and safety

of downstream communities, large dams in Italy are continuously monitored by
operators and subject to close surveillance by government authorities, according
to the provisions stated in a dam-specific agreement named Dam Operation and
Maintenance Document (its Italian acronym is FCEM).
During normal working conditions, the FCEM typically requires dam operators
to record several response quantities, such as displacements, seepages and settle-
ments, with monitoring systems and frequencies defined on a case-by-case basis.
Special measures must be taken after any exceptional event such as a flood or an
earthquake.
To define a national uniform criterion for evaluating dam safety conditions after
earthquakes, in 2002 the Italian Dam Authority issued the first procedure for post-
earthquake inspections of large dams. The procedure introduced a simple rule to
identify dams requiring immediate inspection and defined the type and extent of
controls.
Since 2002 this procedure has been applied many times, gaining significant
experience particularly after the seismic events in Central Italy in 2009 and 2016–
2017, when many large dams were involved. Lessons learned over more than fifteen
years of its use induced the Italian Dam Authority to update the post-earthquake
inspection procedure.
A new procedure was therefore issued in 2018 and is currently in use. The
new procedure introduced three major changes: first, it differentiates two types of
56
Q.106 - R.4
inspections depending on the severity of the shaking; second, it defines the screen-
ing areas on the basis of explicitly defined acceleration thresholds; third, it allows
consideration of real-time earthquake data recorded by seismic monitoring systems.
With respect to the 2002 criterion, the new procedure leads to a narrower screening
area in case of low-to-medium magnitude earthquakes.
2. ITALIAN PROCEDURES FOR POST-EARTHQUAKE

INSPECTIONS ON LARGE DAMS
2.1. PROCEDURE ISSUED IN 2002
On the 1st of July 2002 the Italian Dam Authority issued the technical note
SDI/3536, addressed at dam operators in order to define homogeneous procedures
for inspection and control of large dams after seismic events.
The procedure adopted the criterion given in ICOLD Bulletin n.62 “Inspection of
large dams following earthquakes” to define the screening area (i.e. the geographic
area which included dams requiring immediate attention after the earthquake).
According to the ICOLD criterion, the screening area is a circle, centered on

the epicenter of the earthquake, with radius D defined as a function of the earthquake
magnitude M. The relation M-D which defined the screening area, represented
numerically in Table 1 and graphically in Figure 1, was obtained by application of
ground motion prediction equations in order to ensure that shaking at the dam site
was below a given acceleration threshold, although this acceleration threshold was
not explicitly stated.
The criterion drawn in Figure 1 applies for a given earthquake of magnitude M

and for a given site located at a distance D from the epicenter. If the point (M,D) is
above the curve, the shaking experienced by the dam is low and no post-earthquake
control is required, as the potential effects of the earthquake are expected to be
inconsequential for the dam. On the contrary, if the point (M,D) is below the curve,
special inspection and post-earthquake control is required for the dam and the
upstream reservoir.
Independently of the position of the point (M,D), post-earthquake inspection

and controls are always required if the shaking is felt by personnel on site (e.g. the
watchman).
Details of type and extent of inspections and controls of the dam and the basin
are defined by the Dam Authority in agreement with the dam operator for each site,
depending on peculiar characteristics of the structure and soil.
The dam operator communicates the results of post-earthquake inspection

and controls to the local Technical Office of the Dam Authority, as soon as possible
57
Q.106 - R.4
Table 1
Magnitude of earthquake and corresponding radius of the
circle centered in the epicenter for post-seismic controls
Magnitude M 4 5 6 7 8
Source-to-site distance (km) 25 50 80 125 200
Fig. 1
Curve of Magnitude and source-to-site distance under which special control is
necessary after earthquake
Courbe de magnitude et distance source-site sous lesquelles un contrôle spécial
est nécessaire après un séisme
after the event. Moreover, the Responsible Engineer of the dam must prepare a
report describing the earthquake effects on the dam and the basin, assessing the
safety condition of the structure and the actions taken to avoid or mitigate potential
risks related to the dam.
The 2002 procedure was effectively applied during many seismic events (of
magnitude M > 4) until 2018, when it was updated by the Dam Authority. Events of
particular interest occurred during this period are the seismic sequences of L’Aquila
2009, Po Valley 2012, and Central Italy 2016–2017.
2.2. PROCEDURE ISSUED IN 2018
After fifteen years of use of the 2002 procedure, it was realized that the cri-
terion defining the screening area was very conservative, resulting in the need of
58
Q.106 - R.4
performing the same types of immediate inspection and controls both for dams in
the epicentral area and for dams very far from it.
Therefore, in order to increase the efficiency of dam surveillance, in 2018 the

Dam Authority issued a new procedure, which is described in the technical note n.
25157 of November 8th 2018, which is currently in use.
The new procedure introduces three major changes with respect to the
previous one:
– it differentiates two levels of inspections (Level 1, less detailed and Level 2,

more accurate) depending on the severity of the shaking at the site;
– it explicitly defines values of acceleration thresholds which determine the acti-
vation of post-earthquake inspections, thus defining the radius of the screening
areas for both levels of inspections;
– it allows consideration of real-time earthquake data recorded by seismic moni-
toring systems at the dam site, when available. When such data is not available,
the peak ground acceleration at the dam site is estimated by use of the the most
recent ground motion prediction equations, once the earthquake magnitude
M and the source-to-site distance D are known.
In details, as introduced above, two levels of post-earthquake inspections are

defined: Level 1, less accurate, to be used for dams far from the epicenter and
subjected to a low intensity shaking, and Level 2, more accurate, to be used for
dams in the epicentral area subjected to a medium-to-high intensity shaking.
Assuming that the intensity of the shaking at the site is assumed to be

described by the Peak Ground Acceleration (PGA), Level 1 inspections are required
when the PGA is between 0.04 g and 0.07 g, while Level 2 inspections are required
when the PGA is above 0.07 g.
The horizontal PGA is -ideally- the acceleration measured in the free-field

next to the dam base, placed at the minimum distance from the foundation so that
dam-foundation interaction effects are negligible.
In case of dams equipped with such a reference seismic accelerometer pro-

viding real-time data, the measured peak acceleration can be compared with the
given thresholds (0.04 g and 007 g) in order to decide if the dam requires immediate
inspection.
In case of dams equipped with a seismic monitoring system but without a free-
field accelerometer, another reference sensor can be chosen as a measure of the
shaking intensity at the site, but then the activation thresholds must be re-evaluated
to account for the different installation conditions of the sensor.
When a seismic monitoring system is not available, an empirical criterion is

defined according to the magnitude-distance graph shown in Figure 2.
59
Q.106 - R.4
Fig. 2
Activation thresholds for post-earthquake controls
Seuils d’activation des contrôles post-séisme
For a given earthquake of magnitude M and for a given site located at a dis-
tance D from the epicenter, no post-earthquake controls are needed if the point
(M,D) lies in the white area, Level 1 controls are needed if the point (M,D) lies in the
blue area, Level 2 controls are needed if the point (M,D) lies in the red area,
Moreover, Level 1 controls are always required if on site personnel feels the
earthquake as quite strong, defined as level V of the Mercalli-Cancani-Sieberg
(MCS) intensity scale. Level 2 controls are always required if damages are observed
after Level 1 inspections.
The scope of Level 1 controls is to immediately verify the absence of damages

or failures in the dam or part of it, including potential instability of the abutments or
the basin slopes. The operator must communicate the inspection results to the Dam
Authority within 3 hours from the event.
The scope of Level 2 controls is to verify the safety conditions of the dam.
The operator must communicate the inspection results to the Dam Authority within 3
hours from the event, and the significant response quantities measured at the dam
within 12 hours from the event. The dam Engineer must also provide a technical
note to describe the structural effects of the earthquake and safety measures which
had been put in place.
60
Q.106 - R.4
Fig. 3
Comparison of the criteria to select screening areas
Comparaison des critères de sélection des zones de criblage
The 2018 procedure also suggests a list of controls to be performed for both
types of inspections and provides details for its application in the case of aftershakes.
Finally, it is worth to mention that for a given dam the 2018 procedure allows to
modify the proposed activation thresholds (0.04 g and 0.07 g), if a seismic evaluation
study of the dam is available that provides a reliable estimate of the fragility curve
of the dam.
The rationale for the proposed acceleration thresholds is presented in

Appendix I of the technical note n. 25157, which also includes a comparison of
post-earthquake inspection procedures adopted worldwide. Criteria are not homo-
geneous and depend on the type of information available for the dam and the
knowledge of its seismic behavior.
Figure 3, taken from Appendix I, presents a comparison of different empir-

ical criteria to define the radius of the screening area: the ICOLD 1988 criterion
(brown line), the 2002 procedure (light blue), the 2018 procedure (blue and red
lines, respectively for Level 1 and Level 2), the result of the study of Tani (2000) on
damages suffered by earthfill dams in Japan (green line).
Analysis of Figure 3 shows that the 2002 procedure envelopes the maxima of
the ICOLD criterion and that both the new and the old procedures are conservative
with respect to the results presented by Tani (2000). Moreover, it is clear that the
2018 criterion leads to a narrower screening area for low-to-medium magnitude
earthquakes, while it is more conservative for magnitudes higher than 6.5.
61
Q.106 - R.4
Fig. 4
Shakemap of the 5.5 M Rieti earthquake 2017, with screening circles
Shakemap du séisme de 5,5 M de Rieti 2017, avec cercles de dépistage
The technical note n. 25157 concludes (Appendix II) with a validation study of
the new criterion with reference to the major earthquakes occurred in Central Italy
during the last seismic sequences of 2009 and 2016–2017.
Figures 4 to 6 below show the shakemaps and the corresponding screening

areas computed according to the old and new procedures. The 2002 procedure
(black dashed circle) is compared with the 2018 procedure (blue and red circles).
The new procedure was found to provide a good estimate of the shaking intensity
at the sites. In the case of low magnitude earthquakes, the number of dams to be
inspected is significantly reduced, this being consistent with the observed behavior
of the dams.
3. IMPLEMENTATION OF THE POST-SEISMIC INSPECTION PROCEDURE

IN ENEL COMPANY
ENEL Company operates 181 large dams in Italy, of which 74 are located in
high seismic hazard areas, with estimated peak ground acceleration higher than
0,15 g.
62
Q.106 - R.4
Fig. 5
Shakemap of the 6.5 M Norcia earthquake 2016, with screening circles
Shakemap du séisme de 6,5 M de Norcia 2016, avec cercles de dépistage
According to provisions of the FCEM, each dam is provided with an internal

document, called Control Plan, which describes in detail all the controls that are per-
formed by the operator. The monitoring system of the dam is accurately described,
showing the position and type of measure points, together with the parameters that
have to be measured, the instruments and formulas to calculate these parameters.
Periodical discharge tests and visual inspections are also described into the Con-
trol Plan, focusing on the relevant structural part of the dam, the discharge control
mechanisms, the upstream basin and downstream riverbed.
According to the FCEM, the Control Plan describes the frequencies of mea-
sures, controls and inspections, which may change in case of ordinary operation
and in emergency conditions (i.e. flood and earthquake).
There is a specific Annex to the Control Plan, i.e. Annex n.2, which implements
the procedure for post-earthquake controls issued by the Dam Authority. This docu-
ment is printed and available in the dam guardhouse, and the guardian brings it with
him during the controls. It includes the inspection paths to be walked on the dam,
with photographs of the parts of the dam to control and a brief description of the type
of control to be implemented and potential failures to be excluded. The document
also describes the communications to be activated by the guardian to the reference
63
Q.106 - R.4
Fig. 6
Shakemap of the 6.1 M L’Aquila earthquake 2009, with screening circles
Shakemap du séisme de 6,1 M de L’Aquila 2009, avec cercles de dépistage
technicians of the dam, and the time schedule to be respected, which depends on
the type of controls performed.
The Annex n.2 also includes the list of necessary equipment, such as dam
schemes, notebook, graded container to measure leakages, binoculars, etc.
REFERENCES
[1] Circolare DG Dighe 08 novembre 2017, n.25157 “Controlli straordinari a

seguito di eventi sismici. Aggiornamento procedura art.6.2.3 del FCEM.
Revisione novembre 2017.” http://dgdighe.mit.gov.it/leg.html
[2] Circolare S.N.D. 1 luglio 2002, n.3536 “Controlli straordinari a seguito di eventi
sismici” http://dgdighe.mit.gov.it/leg.html
[3] Tani S. (2000), Behavior of large fill dams during earthquake and earthquake
damage, Soil Dynamics and Earthquake Engineering, vol. 20. https://doi.
org/10.1016/S0267-7261(00)00055-5
64
Q.106 - R.5
-------
-------
DATA SCIENCE PILOT STUDY: IDENTIFYING RISK OF DAM FAILURE

USING ARTIFICIAL INTELLIGENCE (∗)
J.G. STENFERT & D. HONINGH

Advisor,
HKV
M. van HOEK & B. KOLEN

Senior Advisor,
HKV
I. van de KERK
Advisor,
Rijkswaterstaat, Ministry of Infrastructure and Water Management
H. JANSSEN
Senior Advisor,
K. MIDDELJANS & C.H. OOSTINGA

Head of department,
E. BOERMA
Advisor,
NETHERLANDS
∗ Étude pilote sur la science des données : identifier le risque de rupture des barrages à
l’aide d’intelligence artificielle
DOI: 10.1201/9781003211808-103 65
Q.106 - R.5
SUMMARY
Predicting risk of dam failure is conventionally investigated in a physics-driven

way. However, a pure data-driven approach can provide new insights for new
questions about estimating risk of failure, comparing dams all over the world and
identifying vulnerabilities of dams. This study uses the ICOLD World Register of
Dams in combination with the ICOLD Incident database Bulletin 99 update to find
and clarify differences between failed and non-failed dams by using Self-Organizing
Maps. It is a pilot study in which this relatively new technique is used to do a purely
data-driven, big data research. Clusters were found as a result of extremely large
dams. No clear patterns were found which could have provided new insights into
the failure of dams. A limitation of this study was the selection of parameters. This
methodology needs a fully filled database. A trade-off was made between the amount
of parameters and the amount of remaining dams in the database. Although unavoid-
able, the choice of parameters led to significant shrinkage of the usable database.
This methodology should be used with a more complete database to gain new
insights. Additionally, external data sources should be added to enrich the analysis.
To achieve this, more information on the exact location of dams in the database
would be required. These kinds of studies will not directly lead to statistically sound
results. However, the use of self-organizing maps can provide an objective and unbi-
ased view of the dataset, which in turn can lead to new questions about the physics
of dam failure. In this way, this kind of method could be useful in combination with
expert knowledge on dam failure.
RÉSUMÉ
La prédiction du risque de rupture de barrage est classiquement étudiée d’une

manière basée sur la physique. Cependant, une approche purement basée sur les
données peut fournir de nouvelles perspectives pour des questions sur l’estimation
du risque de défaillance, la comparaison des barrages partout dans le monde et
l’identification des vulnérabilités des barrages. Cette étude utilise le Registre mondial
des barrages de la CIGB en combinaison avec la mise à jour Bulletin 99 de la base
de données des incidents de la CIGB pour trouver et clarifier les différences entre les
barrages défaillants et non défaillants en utilisant des cartes autoadaptatives. Il s’agit
d’une étude pilote dans laquelle cette technique relativement nouvelle est utilisée
pour effectuer une recherche de méga-données purement axée sur les données.
Des grappes ont été trouvées à la suite de barrages extrêmement grands. Aucun
modèle clair n’a été trouvé, ce qui a conduit à de nouvelles connaissances sur la
défaillance des barrages. Une limite de cette étude était la sélection des paramètres.
Cette méthodologie nécessite une base de données entièrement remplie. Un com-
promis a été fait entre la quantité de paramètres et la quantité de barrages restants
dans la base de données. Bien qu’inévitable, le choix des paramètres a conduit
à une réduction significative de la base de données utilisable. Cette méthodologie
66
Q.106 - R.5
doit être utilisée avec une base de données plus complète pour obtenir de nouvelles
informations. De plus, des sources de données externes doivent être ajoutées pour
enrichir l’analyse. Pour y parvenir, on a besoin d’information plus spécifique sur
la localisation des barrages qui sont inclus dans la base de données. Le présent
type d’études ne mènera pas directement à des résultats statistiquement solides.
Cependant, l’utilisation des cartes autoadaptatives peut fournir une vue objective et
impartiale de l’ensemble de données, qui à son tour peut conduire à de nouvelles
questions sur la physique de la rupture de barrage. De cette manière, ce type de
méthode pourrait être utile en combinaison avec des connaissances d’experts sur
la rupture des barrages.
1. INTRODUCTION
Predicting risk of dam failure is conventionally investigated in a physics-driven

way. A pure data-driven approach can provide new insights on estimating the risk
of failure, comparing dams all over the world and identifying the vulnerability of
dams. More and more techniques become available to investigate large data sets
with complex dependencies (Ardabili et al., 2019). Artificial Intelligence (AI) studies
are also carried out within ICOLD to investigate new analysis possibilities (Lacasse,
2019). In this study, we continue to explore these possibilities of AI for dams.
This study is an experiment for using Artificial Intelligence in combination with

the ICOLD World Register of Dams. This dam register contains information on
58.713 dams regarding design and construction, and is available for scientific work
and statistical evaluations. Using unsupervised learning, clusters of similar dams
could be formed based on a wide set of parameters. The objective of this exper-
iment is to investigate the usage of data techniques, specifically self-organizing
maps (SOM), in combination with the ICOLD database to identify similarities and
differences between dams all over the world concerning failure of these dams. This
analysis could lead to new insights for research into the physics of dams.
2. MATERIALS AND METHOD
2.1. DATA
The ICOLD World Register of Dams was used as a starting point for this study.
This World Register of Dams contains dams from member and non-member coun-
tries, after validation by the Committee of the Register. The database is filled with
fixed criteria including information about geometry, catchment area, construction,
location and purpose.
67
Q.106 - R.5
In the context of the ICOLD Incident database Bulletin 99 update (ICOLD,

2019), a database with comparable characteristics has been developed in which
only failed dams are present. This database has been added to the ICOLD database.
The failure of a dam depends on the dam itself, but also on the environ-
ment. Therefore, we searched for data that describe the environment of dams and
hydrological parameters. We have added global geological parameters and general
precipitation information.
2.2. SELF-ORGANIZING MAPS
The use of the Artificial Intelligence approach of self-organizing maps for such
an application is new in relation to the analysis of dam failure statistics. Therefore,
we introduce the concept of this analysis and present an example case in which
self-organizing maps are used for the analysis of an international database with a
wide set of parameters.
2.2.1. Concept
The method of self-organizing maps is a machine learning technique in which

complex patterns and relationships are investigated, while preserving topological
properties. It is an objective and unbiased method to investigate complex relations in
large databases with many variables. Self-organizing maps are artificial neural net-
works capable of clustering multidimensional data. It groups a collection of objects
into different groups based on multiple properties per object. The self-organizing
map uses competitive learning, unlike many other artificial neural networks that use
error-correction learning. Competitive learning is a form of unsupervised machine
learning in which output nodes in the network compete, based on the given input.
The node that is most activated by the given input during training is considered the
winner and moves more towards the given input, while the rest of the nodes remain
unchanged. This technique is well applicable to unbalanced data sets where the
patterns to be investigated are not known in advance. For an in-depth explanation
of the method we refer to Miljkovic (2017).
After training, the self-organizing map returns a two-dimensional representa-

tion of a multi-dimensional input space. The so called U-matrix (Unified Distance
Matrix) is a commonly used representation of the self-organizing map and visual-
izes the distance between surrounding data points (Ultsch, 2007). The distance
between neurons is the parameter of interest, giving information on the appearance
of clusters. In the U-matrix, the dark colored nodes depict closely spaced nodes and
blue-reddish colored nodes indicate more widely separated nodes. A cluster of dark
colored nodes surrounded by reddish nodes is therefore seen as a very unique
68
Q.106 - R.5
Fig. 1
Example of a U-matrix (left), a 2D-presentation of an input space in which
neurons represent underlying data. Distance between neurons is visualized
with color. High values (red) correspond to peaks. Large heights (values) mean
that there is a large difference in the underlying data for the specific neuron. Low
values for the U-matrix mean that neurons are close together.
Exemple de matrice U (à gauche), une présentation 2D d’un espace d’entrée
dans lequel les neurones représentent des données sous-jacentes. La distance
entre les neurones est visualisée par des couleurs. Les valeurs élevées (rouge)
correspondent aux pics. Ces pics signifient qu’il existe une grande différence
dans les données sous-jacentes pour le neurone spécifique. Des valeurs faibles
pour la matrice U signifient que les neurones sont proches les uns des autres.
cluster. In this cluster the combination of parameters is apparently different

compared to other data.
2.2.2. Example for European catchments
This example shows the use of a self-organizing map and how the U-matrix
can be interpreted. It is in detail available at https://ai.hkvservices.nl/european-
catchments.
In this example, characteristic data is used of all catchments in Europe, each

describing an aspect of the corresponding catchment area (discharge, information
on terrain, land use, etc.). Figure 2 shows the resulting U-matrix (left).
Analysis of the U-matrix patterns itself is very complicated. Linking each neu-
ron in the U-matrix to the corresponding location, not included in training, helps to
69
Q.106 - R.5
Fig. 2
Left, U-matrix derived from European catchments characteristics; middle,
geographical locations of input locations linked to U-matrix node-locations; right,
indicated geographical regions in the U-matrix; screenshots from Hoek (2019)
À gauche : matrice en U dérivée des caractéristiques des bassins versants
européens; Au centre: emplacements géographiques d’entrée liés aux
emplacements des n***uds de la matrice U; á droite: régions géographiques
indiquées dans la matrice U; captures d’écran de Hoek (2019)
understand the spatial perspective of the U-matrix. This way of linking the U-matrix
to data not already included in the training can be done for every type of data. The
right part of the figure shows geographical regions corresponding to the nodes in
the U-matrix.
The U-matrix is structured using all input data. It is also possible to view the
representation of each individual input variable. As an example, we show the per-
centage of forest in a catchment area in figure 3. In contrast to the U-matrix, the
nodes are now colored by the weight vector. Dark values represent low percentages
of forest and blue-reddish values correspond to high percentages of forest. The
selected box in the upper left corner clearly shows a cluster in South-Sweden and
Finland with a high presence of forest. This cluster becomes apparent by using all
input variables. It is obvious that these two areas have a high presence of forest and
therefore resemble each other. However, other forests do not light up in this cluster,
as they appear to be different based on other characteristics. For example, a forest
area may be at a different level of altitude, receive more or less annual rainfall or
have a different type of subsoil.
2.3. DATA PROCESSING AND APPLICATION OF SELF-ORGANIZING MAPS FOR

ICOLD DATABASE
To derive a self-organizing map for the ICOLD database the following data
processing steps were performed:
70
Q.106 - R.5
Fig. 3
Selection of ‘forest’ – parameter connected with (bottom-right)
spatial locations (screenshots from Hoek (2019))
Sélection du paramètre forêt lié aux emplacements spatiaux
(en bas á droite) (captures d’écran de Hoek (2019))
2.3.1. Parameter selection and dam selection
To demonstrate the validity of the approach, a selection of parameters was

made in such a way that it contained parameters that could be related to dam failure.
At same time it was important to include as many dams as possible in the analy-
sis, because this type of analysis demands a fully filled database. The selected
parameters relevant for failure were: area of reservoir, catchment area, dam type,
length of crest, height, length of reservoir, purposes, reservoir capacity, and year
of completion. The area of reservoir, catchment area and reservoir capacity are
related to the loading of the dam. The dam type and the crest length is relevant for
the sensitivity for failure and year of completion for the aging effect as well as the
design and construction practices of that period. Other parameters, like dam height
or spillway capacity, would have been interesting to include in the analysis, but these
parameters were only available for a small number of dams and were therefore not
included. This resulted in a reduction in the number of dams included in the analysis,
from 58,000 to 430.
2.3.2. Additional data
Based on the location of these remaining 430 dams, global geological and
hydrological information was gathered to enrich the dataset. Because the exact
location of the dams is not known in the database, it is not possible to include
specific data such as average deformation at the location of the dam in this analysis.
71
Q.106 - R.5
Table 1
Transformation of qualitative data of dam types to numbers of dam types for
use of self-organizing maps. Explanations of abbreviations are available at
www.icold-cigb.org
Transformation des données qualitatives des types de barrages en nombre de
types de barrages pour l’utilisation de cartes auto-organisées. Les explications
des abréviations sont disponibles sur www.icold-cigb.org
NR. DAM TYPE NR. DAM TYPE NR. DAM TYPE NR. DAM TYPE
0 TE 5 CB 10 TE/PG 15 TE/VA
1 ER 6 TE/PG/TE 11 TE/BM 16 TE/ER
2 PG 7 PG/TE/ER 12 PG/BM 17 ER/TE
3 VA 8 PG/ER 13 CB/PG 18 XX/VA
4 PG/TE 9 BM 14 CB/TE 19 PG/TE/XX
20 XX
Nevertheless, the soil type was included, based on a general geological map (USGS,
1995), and the average rainfall corresponding to the country of the dam (World
Bank, 2020), although the latter is a somewhat crude approximation of the rainfall
conditions at the dam site.
2.3.3. Preparing for training data
Self-organizing maps only work with quantitative data. However, the database
also contains a lot of qualitative data. In preparation, the qualitative data of purposes,
geology and dam type is therefore transformed into quantitative data for each vari-
able. Table 1 shows the transformation of qualitative to quantitative data of dam
types.
Information about the specific location in coordinates or city name, and failure
status is not included in the dataset for training. After training, data about failure
status is added to the dataset to interpret the U-matrix. This makes it possible to
investigate the amount of failed and non-failed dams within clusters. After the first
training of the self-organizing map, errors were found in the data on the year of
completion. Three dams with these data errors have been removed, followed by
new training of the self-organizing map. Thus, future users of self-organizing maps
should be aware that in practice, this often needs to be an iterative approach.
3. RESULTS
Figure 4 shows the trained self-organizing map, based on the combined

dataset, which contains 330 dams that did not fail and 97 failed dams. This
can be interpreted in the same way as described in the example case. The
72
Q.106 - R.5
Fig. 4
U-matrix of the ICOLD SOM, with numbers of selected dams in the right figure.
In light blue the total number of dams selected and in light red the number of
failed dams (dark blue and dark red are the total amount of dams).
Matrice U de la carte auto-organisée des données du CIGB, avec le nombre de
barrages sélectionnés dans la figure à droite. En bleu clair, le nombre total de
barrages sélectionnés et en rouge clair le nombre de barrages défectueux
(bleu foncé et rouge foncé sont la quantité totale de barrages).
complete resulting self-organizing map can be accessed on: http://ai.hkvservices.nl/

experiment-dam-risk
The U-matrix is shown in conjunction with a bar chart (right). The bar chart
represents the amount of failed and non-failed dams within a selection of the U-
matrix, to help investigate clusters. In this figure, the entire U-matrix is selected and
shown.
Two cluster regions were found, in the upper left corner and in the lower right
corner. Based on all input characteristics, these locations are apparently different
compared to their surroundings, because these nodes are enclosed and separated
by high (red) values. The center of the U-matrix contains mainly low values, which
means that there is relatively little distinction between the different dams based on
the input characteristics.
3.1. INTERPRETATION OF CLUSTERS
Figure 5 shows the selection of the clusters at the upper left and the lower
right corner of the U-matrix. Based on these selections, it can be seen that these
73
Q.106 - R.5
Fig. 5
Selection of formed clusters in U-matrix. Selection is shown left. The amount of
failed or non-failed dams within the selection is given in the right part of the figure.
Sélection des clusters formés dans la matrice U. La sélection est affichée á
gauche. Le nombre de barrages défaillants ou non défaillants au sein de la
sélection est indiqué dans la partie droite de la figure.
clusters are not formed as a result of failure of these dams, because all dams in the
cluster are represented in the bar chart of non-failed dams
With the interactive visualization it is possible to interpret each individual

parameter used in this experiment. When a selection is made in the U-matrix it
74
Q.106 - R.5
Fig. 6
Individual parameter visualization of catchment area (left) and length of reservoir
(right)
Visualisation individuelle des paramètres du bassin versant (à gauche) et de la
longueur du réservoir (à droite)
is possible to navigate through parameters to investigate the contribution of each

parameter to the result in the U-matrix. Based on this interactive visualization per
parameter, it appears that the cluster in the center below is the result of an extremely
large catchment area compared to other dams and the upper left cluster is the result
of a large reservoir length. This is shown in Figure 6 in which the data on catchment
area is presented at the left and data on the length of reservoir is presented at the
right.
3.2. INTERPRETATION OF U-MATRIX
Figure 7 shows a global interpretation of the location of failures in the U-matrix.

Approximately 70% of the failed dams are present in the right part of the figure, and
the other 30% failed dams are mainly present at the left part of the figure. Although
no clear clusters have formed, the failed dams are distributed in some way across
the U-matrix, as shown in the figure.
Figure 8 shows three individual parameter visualizations of the dam type,

amount of annual precipitation and the year of completion. The location of approxi-
mately 70% of the failed dams in Figure 7 (in the U-matrix), are in some way related
to a certain dam type (see Table 1), relatively high precipitation and older dams (‘low’
year of completion).
75
Q.106 - R.5
Fig. 7
Regions of U-matrix with failed dams
Régions de matrice en U avec des barrages défaillants
Fig. 8
Individual parameter visualization of dam type (left), precipitation (middle) and year
of completion (right).
Visualisation des paramètres individuels du type de barrage (à gauche), des
précipitations (au centre) et de l’année d’achévement (à droite).
4. CONCLUSIONS
This article elaborates on the results of a data-driven, big data experiment

with use of a combined dataset of failed and non-failed dams from the ICOLD World
76
Q.106 - R.5
Register of Dams and the Dam Failure Database of the ICOLD Dam Safety Commit-
tee. The objective of this experiment was to investigate the possibilities for using data
science techniques such as self-organizing maps in combination with the ICOLD
database to create insights in similarities and differences between dams all over the
world concerning failure of these dams for new research on the physics of dams.
Two clusters have been found as a result of data outliers. Dams within these
cluster appear to be extremely large compared to other dams. No specific clusters
have been found as a result of failed dams. However, the dam type, precipitation
and year of completion do show relevance for a large amount of failed dams.
The experiment did not point at a clear and dominant cause for the failure of
dams. However, since no specific clusters have been identified in relation to the
failing of dams, it can be concluded that in this experiment, the population of failed
dams does not seem to be correlated to a single parameter, nor a combination of
these.
A limitation of this study was the limited amount of parameters and dams
included in the analysis. A trade-off was made between the amount of parameters
and the amount of remaining dams. Although unavoidable, the selection of parame-
ters led to a significant decrease in the number of dams included in the analysis. Of
the approximately 58,000 dams, 427 dams were suitable for a data-driven analysis.
The main reason for this relatively small number of dams included in the analysis is
that not all relevant characteristics were available for the dams. It was attempted to
include relevant external data, such as soil characteristics and loads. This type of
data could only be included on a general level, because a more accurate location
of the dams was not available. In many cases it was not possible to determine the
exact location of the dam on the basis of the name of the dam and the nearest place.
Therefore the addition of this general information on soil characteristics and annual
precipitation did not lead to additional insight.
Apart from challenges related to the amount and quality of the data, there is
another limitation to this study. Smaller dams are often built and managed differently,
compared to larger dams. In the ICOLD World Register of Dams, there is relatively
little information on small dams, which is unfortunate for this type of analysis since
small dams are more likely to fail (ICOLD, 2018). Therefore, the question is whether
the present type of study will give a full picture of dam failure statistics when only
the dams in the ICOLD Dam Register are included in such a study.
5. RECOMMENDATIONS
This study was a first experiment for using self-organizing maps in combination
with the World Register of Dams to create insight in the failure of dams, using a
77
Q.106 - R.5
data-driven approach instead of a more common physics-driven type of analysis.

Of course, this analysis can be improved and expanded on the basis of current
knowledge and data. To get more insight, a more complete dataset is needed for
a follow-up experiment in order to check whether the present data analysis method
will be more successful then.
In addition, it can be very valuable to enrich the dataset with other data sources.
In order to do this, the exact location of a dam is important. At the moment, only the
dam name and nearest city or village is available. This can provide an indication of
the location, but it is not possible to systematically determine the exact location of
the dam. When the exact location becomes known, it might become possible to use
remote sensing data.
These kinds of studies will not directly lead to statistically sound results. How-
ever, the use of self-organizing maps can potentially provide an objective and
unbiased view of the dataset, which in turn can lead to new questions about the
physics of dam failure. In this way, data-driven analysis of dams could be useful in
combination with expert knowledge on dam failure.
REFERENCES
[1] BILJKOVIC, D., Brief Review of Self Organizing Maps, 2017.
[2] LACASSE, S., Reliability and risk approach for the design and safety evaluation
of dams. ICOLD-CIGB 2019 Symposium. Ottawa, Canada, 2019.
[3] ULTSCH, A., Emergence in Self Organizing Feature Maps, Proceedings of the
6th International Workshop on Self-Organizing Maps (WSOM), 2007.
[4] HOEK, M., Understanding Classification of European Catchments using

Artificial Intelligence. In Geophysical Research Abstracts (Vol. 21), 2019.
[5] FAIZOLLAHZADEH ARDABILI, S.; MOSAVI, A.; DEHGHANI, M.; R.

VÁRKONYI-KÓCZY, A., Deep Learning and Machine Learning in Hydrological
Processes, Climate Change and Earth Systems: A Systematic Review, 2019.
[6] ICOLD, available online: https://www.icold-cigb.org/GB/icold/organization__

mission.asp, 2020.
[7] ICOLD, Incident database Bulletin 99 update, Statistical analysis of dam fail-
ures. Committee on Dam Safety, International Commission on Large Dams
(ICOLD), 2019.
78
Q.106 - R.5
[8] WORLD BANK, Data Bank The World Bank. Average precipitation in depth
(mm per year) from 1962 till 2014, https://data.worldbank.org/indicator/AG.
LND.PRCP.MM, 2020.
[9] USGS, General geologic map of the world. https://mrdata.usgs.gov/

geology/world/, 2015.
[10] ICOLD, General Report 103 – Small dams and levees, 26th ICOLD Congress,
Vienna, 2018.
79
Q.106 - R.6
-------
-------
AN OPEN-SOURCE SOFTWARE FOR DAM MONITORING DATA ANALYSIS:

EXPLORATION, CURATION AND MACHINE LEARNING MODEL FITTING (∗)
Fernando SALAZAR
Project Development Director, CIMNE
André CONDE
Research Engineer, CIMNE
Carlos BARBERO
Head of Dam Safety Department, Catalan Water Agency
SPAIN
SUMMARY
Advances in the performance of measuring devices allow greater accuracy

and reading frequency in dam monitoring systems, but also require advanced tools
for the analysis of the large amount of data recorded. At the same time, many
dams worldwide feature simple, low-frequency, time series records, often including
missing data. An open-source software tool has been developed including func-
tionalities for data acquisition, curation and exploration, which can be applied in
both kinds of dams. Once the data is preprocessed, the tool can also be used for
creating predictive models using Boosted Regression Trees, a robust and easy-
to-use machine learning technique. In this contribution, the functionalities of the
application are showcased by analyzing monitoring data from La Baells arch dam,
a double-curvature arch dam with 100 m height above foundation. The results show
∗ Un logiciel open-source pour l’analyse des données de surveillance des barrages:
exploration, sélection et ajustement du modèle d’apprentissage automatique
80 DOI: 10.1201/9781003211808-104
Q.106 - R.6
the capability of the tool to fill missing data, generate new variables and create accu-
rate predictive models, which can also be analyzed to identify the effect of each load
on the response of the dam.
RÉSUMÉ
Les progrès dans les performances des appareils de mesure permettent une
plus grande précision et une plus grande fréquence de lecture dans les systèmes de
surveillance des barrages, mais ils ont besoin également des outils avancés pour
l’analyse des grandes quantités de données enregistrées. En même temps, de nom-
breux barrages dans le monde ont des enregistrements des séries chronologiques
simples à basse fréquence, souvent avec des données manquantes. Un logi-
ciel libre, comprenant des fonctions d’acquisition de données, de prétraitement
et d’exploration, a été développé et peut être appliqué aux deux types de bar-
rages. Une fois les données prétraitées, l’outil peut également être utilisé pour
créer des modèles prédictifs à l’aide d’arbres de régression renforcés, une tech-
nique d’apprentissage automatique robuste et facile à utiliser. Les fonctionnalités de
l’application sont présentées en analysant les données de surveillance du barrage-
voûte de La Baells, d’une hauteur de 100 m au-dessus des fondations. Les résultats
montrent la capacité de l’outil à remplir les données manquantes, à générer de nou-
velles variables et à créer des modèles prédictifs précis, qui peuvent également être
analysés pour identifier l’effet de chaque charge sur la réponse du barrage.
1. INTRODUCTION
Safety has always been a priority for dam engineers throughout history. Today,
it is becoming even more important as the average age of existing structures
worldwide increases. There is a growing number of dams with several decades
of operation, which are approaching the end of their service life. Many of these
structures, however, continue to provide a fundamental service without viable alter-
natives, which aroused interest in intensifying control and safety measures. The
monitoring system is a crucial element in the control of dam safety. Advances in the
performance of measuring devices allow greater accuracy and reading frequency in
the most recent dams, but at the same time require advanced tools for the analysis
of the large amount of data recorded.
In other technical and engineering fields, and especially in other sectors such
as social networks or communication, the volumes of data handled are much greater
and complex than those of the best instrumented dams, which has led to the
development of very powerful analysis tools and techniques. This technology has
been introduced in dam engineering in recent years at research level, and is close
81
Q.106 - R.6
to penetrating professional practice. In this regard, the Technical Committee on

Dam Surveillance (TCDS) of the ICOLD is preparing a bulletin provisionally enti-
tled “Acquisition and Interpretation of Dam Surveillance Data and Observations”.
The authors of this contribution are collaborating in the writing of the mentioned
document, on the basis of the knowledge acquired in the last years through the
development of several research projects related to the use of machine learning
techniques for the improvement of dam safety [1,2]. One of the results of this work is
a software tool that includes functionalities specifically developed to meet the needs
of dam safety managers in terms of data acquisition, curation and exploration and
its subsequent use for the creation of predictive models using machine learning
techniques. In this contribution, some of the functionalities of the mentioned soft-
ware are briefly described. The tool is open source and free to use. All features
have been developed using the programming language R [3]: the preprocess of the
data, the generation of predictive models and graphical utilities [4]. Monitoring data
from the La Baells dam in the period 2008–2019 have been used for illustrating the
functionalities. La Baells is a double-curvature arch dam with 100 m height above
foundation, located in Barcelona.
2. DESCIPTION OF THE SOFTWARE TOOL
2.1. DATA EXPLORATION
The traditional and simplest way of representing monitoring data involves

showing the time evolution of one or a group of variables. When several variables
of different nature are to be shown, a secondary vertical scale is added. However,
interesting information can be obtained by exploring other types of graphs, such as
a scatterplot of response variables of the dam as a function of the reservoir level
or temperature, or the relationship among response variables (e. g. a pendulum
against another). Thus, one of the functionalities of the software allows the user to
plot three-dimensional and interactive graphs where up to 4 variables of any type
can be represented, including the reading date. Figure 1 shows an example: one
of the radial displacements measured in pendulums is plotted on the vertical axis,
while the reservoir level and the 90-day moving average of the air temperature are
shown on the horizontal axes. Additionally, the points are colored according to the
date of reading. The usefulness of this representation can only be appreciated by
interacting with the graph, which can be rotated and zoomed. The figures below
show the same data from different points of view. It is observed (Fig. 1 left) how in
this case a view can be found that clearly reflects a high linear correlation between
a combination of the variables represented on the horizontal axes and the radial
displacement. This relationship can only be seen by rotating the graph (not seen in
the right view).
82
Q.106 - R.6
Fig. 1
Data exploration. A radial displacement is plotted in the vertical axis, while
reservoir level and air temperature (90-days moving average) are plotted in the 2
horizontal axes.
Exploration des données. Un déplacement radial est tracé sur l’axe vertical, tandis
que le niveau du réservoir et la température de l’air (moyenne mobile sur 90 jours)
sont tracés sur les 2 axes horizontaux.
2.2. DATA CURATION
Dam monitoring databases often include errors of different nature: missing

data, different reading frequencies between devices, and reading errors, among
others. Sometimes, changes in the systems cause the same series to present dif-
ferent reading frequencies in different periods. The formats in which the data is
stored also vary between dam owners. Different file formats can be used to load
data into the application, including some of the most common ones such as csv or
Excel. Exploratory charts allow the identification of those errors and periods without
data, which can be corrected interactively: missing data can be interpolated, read-
ing errors can be fixed. Further features, which include generating derived variables
(e.g. moving averages), among others, are described in [5].
2.3. MACHINE LEARNING BASED MODEL FITTING
The development of machine learning techniques and the improvement in

computing power allowed the use of sophisticated methods for the generation of
prediction models based on monitoring data. Although the HST model [6] and similar
linear regression approaches are still the most common in professional practice,
there is a growing interest in the use of machine learning techniques.
The software includes an interface (Fig. 2) for the generation of predictive

models based on a machine learning algorithm called Boosted Regression Trees
(BRTs). It was selected after a comparative study among several of the most popular
83
Q.106 - R.6
Fig. 2
BRT-based predictive model generation.
Génération d’un modèle prédictif basé sur la BRT.
techniques [7] which showed that BRTs offered a greater prediction capability for
different types of response variables, besides being robust and simple to apply: it
depends on a small set of parameters and these have a minor effect on the final
result. In the case of predicting radial displacements of arch dams, the thermal
inertia of the dam can be considered by using as input the moving average of the
ambient temperature in different periods, as shown in previous studies [8].
2.4. IDENTIFICATION OF BEHAVIOUR PATTERNS
One of the main obstacles for the extensive use of advanced models is the dif-
ficulty in their interpretation and therefore in drawing conclusions about the behavior
of the dam. Machine learning based models require the use of specific methods to
draw conclusions. These methods are commonly used in other fields, and, more
importantly, they offer reliable results because they generally better represent the
true behavior of the system. Numerous works have shown that these models allow
consideration of non-linear effects, as well as collinearity between input variables.
BRT-based models can be analyzed by studying two metrics: the relative influ-
ence of the variables measures the decrease in prediction accuracy when each one
of them is discarded: it is based on the assumption that if a variable not associated
to the dam response is removed, the accuracy of the model should not change, and
vice versa.
84
Q.106 - R.6
Fig. 3
Model interpretation. Relative influence (left) and partial dependence plot for the
two most relevant inputs (right).
Interprétation du modèle. L’influence relative (à gauche) et. le graphique de
dépendance partielle pour les deux entrées les plus pertinentes (à droite).
Figure 3 (left) shows the result in the example used, where it can be seen that
the radial displacement considered is influenced by the reservoir level and the 90-day
moving average temperature. It is interesting to see that the ambient temperature on
the day of the measurement features very low influence, which reflects the thermal
inertia of the dam.
As for the partial dependence plot (Fig. 3 right), the result obtained in the
example corresponds to engineering knowledge: the highest downstream displace-
ment occurs under high hydrostatic load and low air temperature, and vice versa.
What is relevant in the example is that the model has detected it automatically, with-
out the need for preliminary variable selection. In particular, it identifies the greater
relevance of the 90-day moving average of the temperature despite the introduction
of other moving averages, which are highly correlated.
3. CONCLUSIONS
A software tool has been developed that allows loading dam monitoring data for
exploration, error correction, completion of missing data and generation of derived
variables, among other functionalities. With these data, prediction models can also
be created based on BRTs, a robust, flexible and easy-to-apply machine learning
algorithm. It offers in general a good accuracy and has procedures for the analysis of
the model, also implemented in the tool. This allows to identify the actions that have
85
Q.106 - R.6
greater influence on the response of the dam, and analyze that effect. The software
is free to use, allowing it to be available to dam safety engineers worldwide, even
those in less developed countries.
REFERENCES
[1] SALAZAR, F., MORÁN, R., TOLEDO, M. Á., & OÑATE, E. (2017). Data-based
models for the prediction of dam behaviour: a review and some methodologi-
cal considerations. Archives of computational methods in engineering, 24(1),
1–21.
[2] SALAZAR, F., TOLEDO, M. Á., GONZÁLEZ, J. M., & OÑATE, E. (2017). Early
detection of anomalies in dam performance: A methodology based on boosted
regression trees. Struct. Contr. Health Monit, 24(11), e2012.
[3] R DEVELOPMENT CORE TEAM, R. 2018. R: A Language and Environment

for Statistical Computing. R Foundation for Statistical Computing.
[4] CHANG, W., CHENG, J., ALLAIRE, J., XIE, Y., MCPHERSON, J. 2018. Shiny:
Web Application Framework for R. R package version 1.1.0 https://CRAN.R-
project.org/package=shiny.
[5] SALAZAR, F., KOHLER, A., CONDE, A., LANDSTORFER, F. 2019b. Inter-
pretation of dam monitoring data combining visualisation tools and machine
learning. Eberlaste Dam Case Study. 15th International Benchmark Workshop
on Numerical Analysis of Dams. Milano, 9–11 Sep. 2019.
[6] WILLM AND BEAUJOINT: Les méthodes de surveillance des barrages au ser-
vice de la production hydraulique d’Electricité de France, problèmes anciens
et solutions nouvelles, in IXth International Congress on Large Dams. Istanbul,
pp. 529–550. (In French) (1967).
[7] SALAZAR, F., TOLEDO, M. A., OÑATE, E., & MORÁN, R. 2015. An empiri-
cal comparison of machine learning techniques for dam behaviour modelling.
Structural Safety, 56, 9–17.
[8] SALAZAR, F., TOLEDO, M. A., OÑATE, E., & SUÁREZ, B. 2016. Interpretation
of dam deformation and leakage with boosted regression trees. Engineering
Structures, 119, 230–251.
86
Q.106 - R.7
-------
-------
DIGITAL TECHNOLOGIES TO MANAGE

DAM SAFETY RECORDS (∗)
Stefan HOPPE
Dam Safety Engineer, OFITECO
Óscar PÉREZ ARROYO

Dam Safety Engineer, OFITECO
Jürgen FLEITZ
General Manager, OFITECO
SPAIN
SUMMARY
A complete dam documentation file is of vital importance to perform reliable

dam safety evaluation and risk assessment studies. It should combine updated
monitoring records with other dam safety documentation. To this end, easy access
to data and information for the authorized personnel must be provided. Moreover,
sorting and classification of documents in combination with powerful search engines
helps to quickly locate any information. In any case, permanent updating of the
Technical File is fundamental.
Finally, the importance and potential of today’s technology e.g. digitization,

asset management systems, web document managers, BIM, Virtual and Augmented
Reality based on 3D models must be highlighted and should be applied more and
more to facilitate modern dam management. Dam owners should invest in an ade-
quate collection and organization of all documents, data and records and assure to
∗ Technologies numériques pour la gestion des données de sécurité des barrages
DOI: 10.1201/9781003211808-105 87
Q.106 - R.7
update the data management system permanently. Especially for long-term infras-
tructure assets such as dams this is an aspect of vital importance and a means to
transform the living memory into an efficient institutional memory allowing knowledge
transfer to new generations.
RÉSUMÉ
Un dossier complet de documentation sur les barrages est d’une importance

vitale pour effectuer des études fiables sur la sécurité des barrages et l’évaluation
des risques. Il devrait combiner les dossiers de surveillance mis à jour avec d’autres
documents sur la sécurité des barrages. À cette fin, il faut faciliter l’accès aux don-
nées et à l’information pour le personnel autorisé. En outre, le tri et la classification
des documents en combinaison avec de puissants moteurs de recherche permet de
localiser rapidement toute information. Dans tous les cas, la mise à jour permanente
des archives techniques est fondamentale.
L’importance et le potentiel de la technologie actuelle, par exemple la numéri-

sation, les systèmes de gestion d’actifs, les gestionnaires de documents Web, le
BIM, la réalité virtuelle et augmentée basée sur des modèles 3D, doivent être mis
en évidence et devraient être appliqués de plus en plus pour faciliter la gestion
moderne des barrages. Les propriétaires de barrages devraient investir dans une
collecte et une organisation adéquate de tous les documents, données et dossiers
et mettre à jour le système de gestion des données de façon permanente. En par-
ticulier pour les infrastructures à long terme telles que les barrages, il s’agit d’un
aspect d’une importance vitale et d’un moyen de transformer la mémoire vivante en
une mémoire institutionnelle efficace permettant le transfert des connaissances aux
nouvelles générations.
1. INTRODUCTION
During Dam Operation, abundant safety relevant information is generated

every day: Monitoring data, photos, inspection forms, O&M records etc. The amount
of data is constantly growing and provides information on the current safety status of
a dam. In combination with dam safety reports and information about the historical
behaviour of the structure, it forms the knowledge base to allow for comprehensive
safety assessments and evaluations. This knowledge must be easily accessible and
available to the technical staff involved. Further it is very important to continuously
update the file. Nowadays, technology provides a variety of tools to facilitate this
challenge: digitization of all information, document manager in a web environment,
BIM methodologies (Building Information Modelling), etc.
88
Q.106 - R.7
2. MANAGEMENT OF DAM SAFETY DOCUMENTATION
2.1. TECHNICAL FILE OF A DAM
Every engineer involved in preparing dam behaviour reports or safety reviews

of dams may have experienced how tedious and time-consuming it can be to search
for documents and reliable data and validate them, especially if there is no central
archive or chronological gaps. Often more time has to be invested for data collection
and data preparation than for the actual behaviour and safety analysis itself. An
incomplete documentation and data set with lack of information is also a source
of uncertainty and may lead to poor or incorrect diagnoses. Therefore, it is the
responsibility of the dam owner to collect, organize and update all documents and
records within a complete Technical File and to allow to access and manage this
information in an efficient way.
The Technical File of a dam comprises all relevant documents and data related
to the design, construction, operation, maintenance, and ultimately the safety of the
dam. ICOLD Bulletin 158 insists on the importance of this information: “Dam records
are of enduring value and are a critical part of the owners and operator’s archives,
which constitute the institutional memory for these very long lasting facilities.” In case
of dams with a life span of many decades and even centuries this is especially impor-
tant to conserve and transfer the knowledge and information from one generation
of responsible engineers and operators to the next one. Knowledge management
over the years, is essential to keep “memory alive”. The Technical File is a key part
of dam safety data management.
2.2. IMPORTANCE AND CHARACTERISTICS OF SURVEILLANCE RECORDS
Management of surveillance records and monitoring data is more complex

than managing conventional files from reports and requires specialized tools.
An important aspect is the frequency of receiving new information, especially

in case of automated systems. Monitoring data is received on a daily or weekly
basis from many sensors, in some cases even more frequently. Depending on the
inspection program, several visual inspections are made by different experts and
have to be documented as well. Moreover, continuous O&M works and functional
tests generate additional information. In order to keep updated a Technical File which
correctly reflects this dynamic, a continuous actualization process is needed, giving
an essential added value to dam safety management.
Another peculiarity of surveillance records and monitoring data is the diversity

of data: manual, automatic, quantitative, qualitative, different formats of file formats,
tables and databases, raster and vector files, drawings and photographs. Digital
89
Q.106 - R.7
Fig. 1
Converting surveillance records into information
Conversion des enregistrements de surveillance en information
inspection forms can have additional formats. Especially a big amount of photos
has to be managed.
2.3. DIGITIZATION AND DOCUMENT MANAGEMENT
This process seeks to arrange the existing information in a digital and central
database. Migrating from having numerous unordered and scattered documents in
paper format requires digitization (applicable to the documents of the Technical File
but also to the observations during surveillance tasks such as visual inspections and
maintenance) and the use of modern document management systems, preferably
customized to the specific characteristics of dams and dam owners.
Most of the recent developments of software tools to manage dam data and
information are working in a web environment simplifying the user access throughout
an internet connection and through multiple devices. As an example some features
of such a document manager are shown in the following figure.
It combines updated monitoring data with dam safety documentation in one

platform, including the following functionalities:
– Allowing different levels of access depending on the user profile.

– Intuitive navigation tree to access the different folders and to access the
command menus for visualization and editing.
90
Q.106 - R.7
Fig. 2
Dam document and monitoring data manager
Gestionnaire de documents et de données de surveillance du barrage
– All types of data and information can be managed: numerical, text, images,
etc. (e.g. inventories and O&M manuals of equipment and installations, pho-
tographs, monitoring and operational data, templates and sheets for visual
inspections and maintenance tasks, etc.).
2.4. INFORMATION SYSTEMS USING BIM METHODOLOGY
One of the biggest challenges for modern societies in the beginning of the
21st Century is to develop, implement and apply digital technologies to practically all
type of industries and organizations. Dams are no exemption and some dam owners
have already started to use powerful tools such as BIM not only for the design and
construction of new dams, but also for improving operation and maintenance tasks of
existing dams. There are solutions that allow to retrieve any information interactively,
providing access to data and observations in real time.
An example is shown in the following figure. This tool allows the use of 2D and
3D models, facilitating to schedule and design O&M and inspection and monitoring
tasks, to perform cost studies and budget follow-ups as well as to link different pro-
cesses of a project. It is a web platform where different multidisciplinary specialists
have collaborated:
The main features of such type of tools are the following:
– Increases accessibility to information.

– Inspection and maintenance sheets uploaded to the central database are
available to all users, even on handhelds during field work.
91
Q.106 - R.7
Fig. 3
Visualizing monitoring data in 3D model in ODIS
Visualisation des données d’auscultation dans un modèle 3D en ODIS
– Observations made during visual inspections such as photographs, videos,

numerical data, comments, etc. can be uploaded easily during the inspections
and are available in quasi real-time.
– Flexible evaluation tools.
– Improves knowledge on the dam safety conditions.
The following figures illustrate how visual inspections, field observations and
monitoring data can be processed efficiently, interactively and in quasi real-time.
The fact that all data records are stored in a centralised database avoids duplicities
and uncertainties.
3. FUTURE TRENDS FOR DAM DOCUMENTATION MANAGEMENT
3.1. ASSET MANAGEMENT SYSTEMS
An increasing number of dam owners are using asset management systems to

maintain and operate efficiently their infrastructures. Asset management has already
a considerable tradition in applications to industrial facilities and other civil infras-
tructures such as bridges and roads. Among dam owners, especially hydropower
companies started to introduce asset management for dams. For that purpose, a
92
Q.106 - R.7
Fig. 4
Processing of visual inspections, observations and monitoring readings linked to a
3D model data base
Traitement d’inspections visuelles, d’observations et de relevés de surveillance liés
à une base de données de modèles 3D
digital technical file that faithfully reflects the inventory of all the works and installa-
tions of the dam-reservoir environment is a key element. The main purpose of asset
management is to maintain and preserve infrastructures to guarantee its function-
ality over time and to extend its service life within a sustainable context considering
safety, social and economic aspects and climate change adaptation.
The implementation and setup of an asset management system requires that

the dam owner clearly defines the objectives and collects all available data and
information. A complete and updated Technical File comprising the details of all
elements is a necessary source to set up a management system. Software tools
are employed to organize and implement multidisciplinary strategies for preventive,
corrective and predictive operation and maintenance tasks.
Dam safety aspects and the applicable regulations must be considered in asset
management systems for dams. The conjunction of a dam and its foundation is a
unique combination and the potential failure modes might be quite different, even for
apparently similar dams. It is important to take into account this fact, which is quite
specific to dams compared to other types of infrastructures or industrial facilities
characterized by a higher repeatability of its behaviour.
3.2. VIRTUAL REALITY
The existing software technology enables the display of 3D models in a realistic

environment as Virtual Reality (VR) or Augmented Reality (AR). The necessary
basis for such complex models is a complete and updated Technical File and data
management system as described above.
93
Q.106 - R.7
Fig. 5
Example of a VR model: General view and detail of monitoring system
Exemple de modèle RV : vue générale et détail du système de surveillance
The applications of VR and AR models in dam surveillance and safety activities

are numerous and comprise predefined and system-guided inspection routes, online
assistance to dam operators, access to previous records and observations, etc. A
very important use is also training of personnel involved in the management of the
dam, from engineers to dam attendants.
REFERENCES
[1] General approach to Dam Surveillance. Basic elements in a “dam safety”

process. ICOLD Bulletin 138, 2009.
[2] Dam surveillance guide. ICOLD Bulletin 158, 2016.
[3] HOPPE, S. Gestión de auscultación y seguridad de presas con BIM, EUBIM

2017, Universidad Politécnica de Valencia, 2017.
94
Q.106 - R.8
-------
-------
SURVEILLANCE OF LARGE CONCRETE DAMS AIDED BY AUTOMATED

MONITORING SYSTEMS AND MACHINE LEARNING TECHNIQUES.
CONTRIBUTION FROM THE PORTUGUESE EXPERIENCE (∗)
Juan MATA
Researcher of the Monitoring Division of the Concrete Dam Department,
LABORATÓRIO NACIONAL DE ENGENHARIA CIVIL
António TAVARES DE CASTRO

Researcher, head of the Monitoring Division of the Concrete Dam Department,
LABORATÓRIO NACIONAL DE ENGENHARIA CIVIL
PORTUGAL
SUMMARY
One concern of the dam surveillance activities is the assessment of the real
structural behaviour, to detect possible malfunctions early. For real-time decision be
effective, there must be confidence in the measured data, and it must be possible
to interpret the data (through adequate data-based methods) in order to properly
assess the structure’s behaviour and condition (based on reliable numerical mod-
els). In the actual context, the main threats are the inability to detect in a timely
manner some scenarios of abnormal structural behaviour that may originate an
accident. Therefore, the achievement of the objective of real time structural safety
control through the use of automated monitoring systems requires the improvement
of the management, validation, archiving and exploitation of information, and the
implementation of a quality control process for the measured data.
∗ Surveillance des grands barrages en béton utilisant un système de mesure automatique
et des méthodes d’apprentissage automatique. Contribution de l’expérience portugaise.
DOI: 10.1201/9781003211808-106 95
Q.106 - R.8
In Portugal, the Concrete Dam Department of the National Laboratory for Civil
Engineering (LNEC) has been involved on the safety control of the Portuguese large
dams presenting more significant risks since their design and construction stage.
One of the more important objectives of LNE’s activity is the development of a man-
agement information systems and several methodologies, aiming the improvement
of structural dam safety control in real time that, using automatic data acquisition
systems data, allows:
i) the evaluation of the quality of instrument readings, considering the redun-

dancy of measurement systems and using adequate outlier’s identification
tools
ii) the analysis and interpretation of the structural behaviour using machine learn-
ing techniques, like, among others, neural networks and short time Fourier
transform.
iii) the support to decisions resulting from the structural safety assessment,
namely the emission of early warning messages due to real-time classifica-
tion of measurements based on thresholds previously defined and the early
detection of patterns related to dam failure scenarios
It should be noted that the ongoing improvements in new automated monitoring

systems combined with new data analysis approaches, supported in a solid dam
behaviour knowledge, are important to achieve the main goal of safety assessment
of dams, preferably in real-time.
RÉSUMÉ
Une des préoccupations des activités de surveillance des barrages est

l’évaluation du comportement réel de la structure, afin de détecter rapidement
d’éventuels dysfonctionnements. Pour que la décision en temps réel soit effi-
cace, il doit y avoir confiance dans les données mesurées, et il doit être possible
d’interpréter les données (par des méthodes adéquates basées sur les données)
afin d’évaluer correctement le comportement et l’état de la structure (sur la base
de modèles numériques fiables). Dans le contexte réel, les principales menaces
sont l’impossibilité de détecter en temps opportun certains scénarios de comporte-
ment structurel anormal pouvant être à l’origine d’un accident. Par conséquent, la
réalisation de l’objectif de contrôle de la sécurité structurelle en temps réel grâce
à l’utilisation de systèmes de surveillance automatisés nécessite l’amélioration de
la gestion, de la validation, de l’archivage et de l’exploitation des informations, et la
mise en œuvre d’un processus de contrôle de la qualité des données mesurées
Au Portugal, le service des barrages en béton du Laboratoire National de

Génie Civil (LNEC) a été impliqué dans le contrôle de sécurité des grands bar-
rages portugais présentant des risques plus importants depuis leur conception et
leur construction. ’un des objectifs les plus importants de ’activité du LNEC est le
développement ’un système d’information de gestion et de plusieurs méthodologies,
96
Q.106 - R.8
visant l’amélioration du contrôle de la sécurité structurel des barrages en temps

réel qui, à l’aide des données des systèmes ’acquisition automatique de données,
permet:
i) l’évaluation de la qualité des lectures des instruments, en tenant compte de

la redondance des systèmes de mesure et en utilisant des outils appropriés
d’identification des valeurs déraisonnable
ii) l’analyse et l’interprétation du comportement structurel à l’aide de techniques
d’apprentissage automatique, comme, entre autres, les réseaux de neurones,
l’analyse par transformées de Fourier à court terme et les modèles récurrent
à mémoire court et long terme,
iii) ’aide aux décisions résultant de l’évaluation de la sécurité structurelle, comme
l’émission de messages d’alerte précoce dus à la classification en temps
réel des mesures en fonction de seuils préalablement définis et la détection
précoce des modèles liés aux scénarios de rupture du barrage
Il convient de noter que les améliorations en cours dans les nouveaux sys-
tèmes de surveillance automatisés combinés à de nouvelles approches d’analyse
des données, appuyées par une solide connaissance du comportement des bar-
rages, sont importantes pour atteindre l’objectif principal de l’évaluation de la
sécurité des barrages, de préférence en temps réel
1. INTRODUCTION
Europe is at the beginning of a new industrial revolution, considered to be the

fourth such leap forward and hence labelled Industry 4.0. The ubiquitous use of
sensors, the expansion of wireless communication and networks, the deployment
of increasingly intelligent robots and machines (as well as increased computing
power at lower cost and the development of ’big data’ analytics) has the potential to
generating new kinds of products and services [1–3]. The actual level of development
and dissemination of these techniques and methodologies allows their use in several
domains, where the access and processing in real time of remote data are of crucial
importance. It is the case of the safety control of large structures, like large dams,
which dam failure can cost a large number of human lives and properties, as well
as significant environmental damages.
In Portugal, the Concrete Dam Department of the National Laboratory for

Civil Engineering (LNEC) has been involved on the safety control of the Portuguese
large dams presenting more significant risk since its design and construction stage.
In line with the international best practices, it is fully committed on the development
of automated monitoring systems and it is the main responsible for the development
of the gestBarragens system, the support system used by LNEC, EDP-Energias de
Portugal and other dam owners, for monitoring, diagnosis and safety control of con-
crete dams in Portugal and abroad. LNEC is also responsible for the development
97
Q.106 - R.8
of several methodologies and procedures to support decision-making in the context

of real-time safety control of large concrete dams. These developments contribute
to enhance: i) the quality control of the different measurements, ii) the analysis and
interpretation of the structural behaviour, namely through long- and short-term analy-
sis and prediction, and iii) the structural safety assessment and the decision-making.
In order to improve the knowledge in these subjects, several developments

and achievements reached by LNEC in the structural safety control of concrete
dams aided by automated monitoring systems are presented in this work.
2. STRUCTURAL SAFETY CONTROL ACTIVITIES
2.1. CONTEXT AND MOTIVATION
The safety control of large concrete dams, involving monitoring data and
numerical modelling, is an important challenge being faced by dam engineers [4].
The main concern, for detecting possible malfunctions as early as possible, is the
real-time assessment of the structural behaviour under operating or extreme environ-
mental conditions. For an effective real-time decision, confidence in the measured
data is crucial. Additionally, it must be possible to interpret these data (through ade-
quate data based methods) in order to properly assess the structural’s behaviour and
condition (with the support of reliable numerical models). Nowadays, for controlling
the structural safety of large concrete dams under operating conditions, the collected
data is usually analysed using statistical models for effects separation (environment
loads and time effects) [5,6]. The results of these statistical models are compared
against deterministic models, like finite element models, and synthesis graphics are
automatically generated and periodically sent to the technicians in charge. In the
recent context, the main threat for real-time decision is the inability to timely detect
scenarios of abnormal structural behaviour that may originate an accident. There-
fore, the actual main challenge of real time structural safety control, through the use
of (automated) monitoring systems, is related to the continuous improvement of the
management, validation, archiving and exploration of the information, and to the
implementation of a quality control process for the measured data.
2.2. TECHNOLOGY’S IMPACT ON DATA COLLECTION FOR DAM SURVEILLANCE

APPROACHES
In Portuguese dams, automated monitoring systems have a conventional

structure, comprised of an automated data acquisition system (ADAS), data trans-
mission system, and data processing and management system. From 2005 until
2010, deluge of data can be seen with a linear growth estimation of total acquired
98
Q.106 - R.8
records of 0,9 × 106 /Year. This is mainly caused by the deployment of gestBarra-
gens∗ [7,8], which supports a larger number of instrumentation types and the archive
of records from many decades of dam monitoring. From 2011 until 2018, a linear
growth estimation of acquired records of 2,8 × 106 /Year is verified because new
schedule of measurements in weather station were defined and in remaining ADAS
devices (each hour instead). The gestBarragens supports more than 30 types of
manual data acquisition instruments and 25 types of automatic monitoring devices,
as well as a larger number of instrument types. The investment in automated data
acquisition additionally has lead to a higher frequency of data acquisitions. The cur-
rent growth rate is expected to be steady or increase over the next years, since new
large dams are currently under construction and the major dam owners are investing
in new automated monitoring systems to further increase the real time monitoring
of critical infrastructures.
2.3. THE USE OF MACHINE LEARNING TECHNIQUES IN SUPPORTING DATA

CLEANSING, ANALYSIS AND INTERPRETATION
“Machine Learning is the study of computer algorithms that improve automat-

ically through experience”. The major focus of ML research is to extract information
from data automatically, by computational and statistical methods. Applications
range from datamining programs that discover general rules in large data sets,
to information filtering systems that automatically learn users’ interests [9]. For
example, Hydrostatic-Seasonal-Time models (HST models), based multiple linear
regression (MLR) methods, are a type of machine learning model traditionally used
by dam engineers. There are two main learning approaches used to train the data
and get the output: supervised learning and unsupervised learning, Fig 1. In super-
vised learning, an algorithm learns from a dataset in order to develop a model. The
input as well as the corresponding output are known in the learning phase. The
algorithm learns from the given data and next time when a new data is present to
the model, it will give the output (predict next value) in accordance with the pattern
presented in the learning phase. In unsupervised learning, the target classes or
labels are not available. The model learns with the input data only being the output
or the result of the data is unknown. The algorithm groups the data or make clusters
of the data to make labels and when a new input is provided next time, the model
checks for the best cluster to which the input data should belong, to give the output.
The main difference between supervised and unsupervised learning is that

supervised learning uses labeled data, while unsupervised learning uses unlabeled
data. Several mathematical algorithms can be adopted, depending the goal and the
data available, Fig 2. An interesting description of the main algorithm used in the
context of the interpretation of the behaviour of dams can be found in [10, 11].
∗A data processing and management system for monitoring, diagnosis and dam safety
control developed in Portugal since early 2000.
99
Q.106 - R.8
Fig. 1
Supervised and unsupervised learning approaches
Approches d’apprentissage supervisé et non supervisé
Fig. 2
Techniques used in machine learning on the dam safety assessment context
Techniques utilisées en apprentissage automatique dans le contexte de
l’évaluation de la sécurité des barrages
2.4. IMPROVING STRUCTURAL SAFETY ASSESSMENT APPROACHES
The success of safety assessment approaches, whether these are based on

numerical modelling or data-driven strategies, relies on the quality of the data gath-
ered and the effectiveness of its analysis and interpretation. Errors can appear into
data analytics processes at any stage of the structure’s life and serious issues can
result when they do. Hence, one task of the surveillance activities is to develop
methodologies suitable to ensure that reliable monitoring data is acquired in real-
time. The following surveillance task aims at being capable of analysing, interpreting,
predicting and classifying the structural behaviour in the short and long terms, after
which the third task main objective consist of assessing the structural condition and
early-warning detection of abnormal scenarios. Based on the results and achieve-
ments of all these three tasks, decision support tools, as a part of the gestBarragens
system, capable of outputting the results of the structural assessment are impor-
tant to allow an effective near real time assessment of the dam behaviour and
100
Q.106 - R.8
Fig. 3
Main ideas regarding to the safety assessment and surveillance of large concrete
dams. From data to the decision making
Principales idées concernant l’évaluation de la sécurité et la surveillance des
grands barrages en béton. Des données à la prise de décision
condition [12]. The main ideas and the relation between the tasks are illustrated
in attached Fig. 3.
In order to improve the knowledge in these subjects, several developments

and achievements reached in the structural safety control of concrete dams aided
by automated monitoring systems are presented in the following sections.
3. QUALITY ASSESSMENT OF MONITORING DATA
In a recent past, when the first automated data acquisition systems in concrete
Portuguese dams were in an experimental phase, there were no procedures for
quality control of measurements. Then, three procedures were developed:
1. A procedure for the assessment of stored measurements (in the management

information systems, gestBarragens [13];
101
Q.106 - R.8
2. A procedure for quality control of measurements based on measurement

uncertainties in order to take advantage of the redundancy of measure-
ment systems (automated and manual) and the confidence in the manual
measurements [12]; and
3. Outliers identification through the use of DBSCAN, a clustering algorithm [14].
More details can be obtained through the referred documents that can made
available by the authors if required.
4. ANALYSIS AND INTERPRETATION OF THE CONCRETE DAM

BEHAVIOUR
This component is responsible to implement the analysis and interpretation

mechanisms that entail the “expert” component of the overall decision support sys-
tem. The analysis and the interpretation of the monitoring data is an important activity
of the structural safety control of concrete dams. The purpose of this activity is to
provide the necessary background about dam behaviour for a better definition of
the requirements (data selection, type of models, etc.), to enhance the conceptual
understanding and to represent the dam’s behaviour.
4.1. SEPARATION FOR THE HYDROSTATIC, THERMAL AND TIME EFFECTS USING
ARTIFICIAL NEURAL NETWORKS
Further than traditional HST models (based on MLR methods), artificial neural
network (NN) models can also contribute in characterizing the normal structural
behaviour for the actions to which the structure is subject using the past history of the
structural behaviour. Different NN models have been proposed since its conception
in the 1940s, but the Multilayer Perceptron (MLP) model is the most widely used.
MLP models learn by an iterative process, by adjusting the weights so as to be able
to correctly learn the training data and hence, after the testing phase, to predict
unknown data. The NN models themselves identify the influence that each variable
has on the structural response, by taking advantage of parallel processing in the
interaction of inputs. In 2011, a study showed that the effect that each variable has
on the output of the NN model can also be obtained from ceteris paribus analysis. [5].
Another interesting application can be found in [15].
The study was carried out on the analysis, ceteris paribus, for the effect of
the hydrostatic pressure changes and the effect of the temperature changes on the
horizontal crest displacement in a concrete arch dam. For example, the analysis
performed for horizontal displacements due to the effect of the temperature was
achieved for several reservoir water level heights previously established, covering
the entire field for the temperature variation. The process was repeated several
times for various reservoir water level heights (74 m, 79 m, 84 m and 89 m that
102
Q.106 - R.8
Fig. 4
Thermal effect on the horizontal crest displacement in a concrete arch dam
obtained through MLR and NN models
Effet thermique sur le déplacement horizontal de la crête dans un barrage-voûte
en béton obtenu grâce avec des modèles MLR et NN
correspond to a water level equal to 860 m, 865 m, 870 m and 875 m, respectively).
The results illustrated in Fig. 4 showed that the effect of the temperature changes
on the structural response is practically independent from the reservoir water level
variation.
The effects for both models shows to be similar. In the MLR model, the effect
of the temperature changes is given by a sinusoidal curve with an amplitude and a
well defined phase. However, it seems that in the NN model, the response due to
the thermal effect has a wave with smaller opening in the warm months than in the
cold months and it is better adapted to the observed values than the MLR model.
This flexibility is an advantage that the NN models have over MLR models. NN
models showed flexibility and proved to be more adequate for months with extreme
temperatures than the MLR models with the same variables.
4.2. THE IDENTIFICATION OF THE EFFECT OF THE DAILY VARIATION OF THE AIR
TEMPERATURE ON THE STRUCTURAL RESPONSE OF CONCRETE DAMS
The use of ADAS data makes it possible the identification of the effect of the
daily variation of the air temperature on the structural response of concrete dams
[17]. The main purpose is to obtain a better knowledge about structural behaviour of
concrete dams. In current day to day activities, quantitative interpretation models are
103
Q.106 - R.8
used for the interpretation and assessment of structural dam behaviour of concrete
dams. Most models ignore the temperature effect of a wave with a daily variation.
However, in dams with automated data acquisition systems this daily effect can be
used to anticipate the detection of abnormal behaviour.
A new methodology based on the use of frequency domain techniques was

developed in LNEC in order to provide information that is difficult or sometimes
impossible to realise in the time domain. The methodology consists in the use of
Short Time Fourier Transform analysis of the residuals, obtained by the traditional
HST models and measurement data, to identify the signature that the daily vari-
ation of the air temperature has on the structural behaviour of a concrete dam,
Fig. 5. A case study was presented in [16] based on the analysis of a horizontal
Fig. 5
Radial displacements in FP3-326.5 m in a large arch concrete dam: observed
values, predicted values and residuals over time. The effect of the daily variation of
the air temperature on the structural response is not observed in the cold time (a)
and it is observed in the warm time (b)
Déplacements radiaux dans FP3-326,5 m dans un barrage-poids:
valeurs observées, valeurs prédites et résidus dans le temps.
L’effet de la variation diurne de la température de l’air sur la réponse
structurelle n’est pas observé pendant le temps froid (a) et il est
observé pendant le temps chaud (b)
104
Q.106 - R.8
displacement measured on a pendulum, in a large arch concrete dam. As a result,

the relation between magnitudes of daily variations of the horizontal displacement
analysed and the air temperature was defined. The relation obtained can be used
to assess if there is alterations or not in the dam response to short period loads
corresponding to the daily variations analysed.
5. STRUCTURAL SAFETY ASSESSMENT AND

DECISION-MAKING
In terms of the structural dam behaviour, if the results obtained from dam
behaviour models (that represent the normal behaviour) and of the observed
behaviour (measured) are divergent, then some hypothesis/es which support the
model are no longer valid, and the reason for this change shall be identified to assess
the consequences. Over the years, new developments in monitoring systems have
increased the capability of measuring the structural response and the main loads
at dams, with the desired frequency. Also, new developments in information sys-
tems allow easy data access to interpret the information, and the emission of early
warnings about the dam’s structural behaviour. Two new approaches have been
developed since the use of automated monitoring systems: i) novelty identification
based on a local threshold definition, and ii) early detection of patterns related to
concrete dam failure scenarios.
5.1. NOVELTY IDENTIFICATION BASED ON A LOCAL THRESHOLD DEFINITION
This approach is related to the quantitative definition of local threshold

and its implementation based on a graphical representation, as suggested in
Fig. 6 [18].
The graphical representation of the information when an abnormal situation is

verified can be very important for the identification of the potential causes and conse-
quences of a hypothetical problem. The main idea is to classify the quantities based
on thresholds previously defined, based on recorded measurements or results from
mathematical models. Taking into account the signal convention of the quantities,
this representation may allow an overview of the dam condition and a timely iden-
tification of any potential accident scenario. If some threshold is achieved, an early
warning based on an automatic message should be send to the person responsible
for the dam safety.
105
Q.106 - R.8
Fig. 6
Graphical example of a dashboard for dam behavior assessment
Exemple graphique d’un tableau de bord pour l’évaluation du comportement des
barrages
5.2. EARLY DETECTION OF PATTERNS RELATED TO CONCRETE

DAM FAILURE SCENARIOS
This contribution is based on the development of a methodology for the early

detection of patterns related to concrete dam failure scenarios [19,20]. The main
objective is to assess whether ADAS measurements indicate the development of
scenarios related to abnormal dam behaviour. The expected magnitudes of the phys-
ical quantities to be measured within these scenarios are usually estimated using
existing mathematical models. To improve the effectiveness of the structural safety
control in real time, methods based on the construction of decision rules for the early
detection of developing abnormal dam behaviour scenarios were developed. These
106
Q.106 - R.8
decision rules are based on the use of supervised machine learning models devel-
oped with data obtained through mathematical models of the structural behaviour,
for scenarios related to abnormal dam behaviour, and real observed data (physical
quantities measured through the use of ADAS) that represent the normal structural
behaviour. The aim is to combine the physical measured quantities, appropriately
weighted, into a new single index allowing the classification of the observations into
one of two classes: normal or abnormal behaviour (with potential to develop an
incident or accident scenario), Fig. 7.
Fig. 7
Classification of the observed dam behaviour based on decision rules for early
detection of a developing concrete arch dam failure scenario
Classification du comportement des barrages observé en fonction des règles
de décision pour la détection précoce d’un scénario de rupture d’un
barrage-voûte en béton en développement
6. MAIN REMARKS
LNEC’s work has been focused on the development of management informa-

tion systems and several methodologies, aiming the improvement of structural dam
safety control in real time, to:
107
Q.106 - R.8
• improve the reliability of the measured data provided by automated data

acquisition system, through the implementation of a quality control routine
for the validation of the data measured, taking into account the measurement
uncertainty of the measurement systems;
• the analysis and interpretation of the structural behaviour using machine learn-
ing techniques, like, among others, neural networks and short time Fourier
transform;
• the decision rules definition for early warnings related to the identification of
abnormal behaviour and of a potential failure scenario.
So, it should be noted that the ongoing improvements in new automated

monitoring systems combined with new data analysis approaches, supported in
a solid dam behavior knowledge, are important to achieve the main goal of safety
assessment of dams, preferably in real-time.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the Portuguese National Labo-

ratory for Civil Engineering (LNEC), through its research program RESTATE
(0403/112/20970).
REFERENCES
[1] DITTRICH, P. Reskilling for the fourth industrial revolution. Formulating a

European strategy. Jacques Delors Institut, 2016.
[2] EUROPEAN COMMISSION. Industry in Europe. Facts & figures on com-

petitiveness & innovation. European commission, ISBN 978-92-79-65262-2,
Brussels, 2017.
[3] O’HALLORAN, D., KVOCHKO, E. Industrial Internet of Things: Unleashing

the Potential of Connected Products and Services. World Economic Forum,
Switzerland, 2015.
[4] ICOLD. Dam surveillance guide. Bulletin number 158. International Commis-
sion on Large Dams, Paris, 2013.
[5] MATA, J. Interpretation of concrete dam behaviour with artificial neural net-
work and multiple linear regression models. Engineering Structures. Elsevier.
48(3):903–910. doi:10.1016/j.engstruct.2010.12.011, 2011.
108
Q.106 - R.8
[6] MATA, J.; TAVARES DE CASTRO, A.; SÁ DA COSTA, J. Constructing input

for statistical models for arch dam deformation prediction. Structural Control
and Health Monitoring. Wiley Online Library. doi:10.1002/stc.1575, 2013.
[7] PORTELA, E.; PINA, C.; SILVA, A.; GALHARDAS, H.; BARATEIRO, J. The
modernisation of dam information systems: The gestBarragens system (in Por-
tuguese). In Dams – Technology, Safety and Interaction with Society. LNEC,
Lisbon, Portugal, 2005.
[8] TAVARES DE CASTRO, A.; MATA, J.; BARATEIRO, J.; MIRANDA, P. Informa-
tion management systems for dam safety control. The Portuguese experience.
Dam World. IBRACON, Brazil, 2012.
[9] CMU. Machine Learning textbook. www.cs.cmu.edu. Retrieved 2020-05-

28.2020, 2020.
[10] SALAZAR, F., MORÁN, R., TOLEDO, M.A., OÑATE, E. Data-Based Models
for the Prediction of Dam Behaviour: A Review and Some Methodological
Considerations. Archives of computational methods in Engineering. State of
the Art Reviews. Springer. ISSN 1134–3060, 2017.
[11] SALAZAR, F., TOLEDO, M.A., OÑATE, E., MORÁN, R. An empirical compar-
ison of machine learning techniques for dam behaviour modelling. Structural
Safety. Elsevier. doi: 10.1016/j.strusafe.2015.05.001, 2015.
[12] MATA, J. Safety control of control of concrete dams aided by automated

monitoring systems. PhD Thesis, Technical University of Lisbon, Portugal,
2013.
[13] MATA, J.; TAVARES DE CASTRO, A. Assessment of stored automated mea-

surements in concrete dams. Second International Dam World Conference.
LNEC, Lisbon, Portugal, 2015.
[14] LORVÃO, A. Adding value to sensor data of civil engineering structures:

Automatic outlier detection. 978-1-5386-8161-9/18, 2018 IEEE, 2018.
[15] GRANRUT, M., SIMON, A., DIAS, D. Artificial neural networks for the inter-
pretation of piezometric levels at the rock-concrete interface of arch dams.
Engineering Structures. Elsevier. doi: 10.1016/j.engstruct.2018.10.033, 2019.
[16] MATA, J.; MARTINS, L.; TAVARES DE CASTRO, A.; RIBEIRO A. S. Statistical
quality control method for automated water flow measurements in concrete
dam foundation drainage systems. FLOMEKO2019 – 18th International Flow
Measurement Conference. 26–28 June 2019. Lisbon, 2019.
[17] MATA, J.; TAVARES DE CASTRO, A.; SÁ DA COSTA, J. Time-frequency anal-

ysis for concrete dam safety control: Correlation between the daily variation
of structural response and air temperature. Engineering Structures. Elsevier.
48(3):658–665. doi:10.1016/j.engstruct.2012.12.013, 2013.
109
Q.106 - R.8
[18] MATA, J.; TAVARES DE CASTRO, A.; SÁ DA COSTA, J. “Threshold defini-

tion for Internal early warning systems for structural safety control of dams.
Application to a large concrete dam”. Dam World Conference 2012, Brazil,
2012.
[19] MATA, J.; SCHCLAR LEITÃO, N.; TAVARES DE CASTRO, A.; SÁ DA COSTA,
J. Construction of decision rules for early detection of a developing concrete
arch dam failure scenario – A discriminant approach. Computers and Struc-
tures. Elsevier. 142 (2014) 45–53. doi: 10.1016/j.compstruc.2014.07.002,
2014.
[20] LEMOS, J.; GOMES, J. Modeling seismic failure scenarios of concrete dam
foundations, chapter 30. Taylor & Francis, 2007.
110
Q.106 - R.11
-------
-------
RETOUR D’EXPÉRIENCE EDF DE

L’AUSCULTATION DES BARRAGES (∗)
Thierry GUILLOTEAU
Expert surveillance des barrages, EDF-DTG
Matthieu DOUAT
Chargé d’affaires, EDF-DTG
Olivier RUSSO
Ingénieur génie civil, EDF Hydro Alpes
FRANCE
RÉSUMÉ
En 1982, Électricité de France (EDF) publiait au congrès CIGB de Rio de

Janeiro un rapport faisant le bilan de 30 années de surveillance et d’auscultation
de ses barrages. Ce rapport vise à actualiser ce bilan sur la base de plus de 70
ans d’expérience dans la gestion de la sûreté d’aménagements hydrauliques. Deux
aspects essentiels de l’activité d’auscultation des barrages sont développés: la
maîtrise sur de longues périodes de temps de la qualité des mesures d’auscultation
réalisées sur les barrages et la gestion d’un processus de surveillance dynamique
permettant de traiter au plus tôt les précurseurs d’anomalies de comportement. La
conclusion synthétise et illustre ces propos par la formulation de recommandations
pouvant être reprises dans d’autres contextes et par d’autres organisations.
∗ Dam monitoring experience at Électricité de France (EDF)
DOI: 10.1201/9781003211808-107 111

Q.106 - R.11
SUMMARY
In 1982, Électricité de France (EDF) published at the ICOLD congress in Rio

de Janeiro a report summarizing 30 years of monitoring of its dams. The purpose
of this paper is to update this outlook, with now more than 70 years of experience
in the management of the safety of hydroelectric facilities. On a portfolio of several
hundred dams, this constitutes a unique sum of data and acquired knowledge and
confirms the need for good organization and continuous improvement of practices.
After a reminder of the framework and perimeter of the monitoring of EDF

dams, the report develops two essential aspects of this activity:
Controlling, over long periods of time, the quality of measurements carried out
on dams. This is based on a variety of activities, ranging from the qualification of
materials to the training of operators in charge of measurements,
Managing a dynamic monitoring process, exploiting all the collected data to

detect and treat precursors of behavioral anomalies as soon as possible. The results
obtained at the portfolio level, illustrated by a concrete case, confirm the effective
contribution of monitoring to the control of dam safety.
The conclusion summarizes the main lessons learned from this long and broad
experience and makes recommendations that can be taken up in other contexts and
other organizations.
1. INTRODUCTION
En 1982, Électricité de France (EDF) publiait au congrès CIGB de Rio de

Janeiro un rapport [1] faisant le bilan de 30 années de surveillance et d’auscultation
de ses barrages. Il nous a paru intéressant de lui donner une suite, avec désormais
une expérience de plus de 70 ans dans la gestion de la sûreté d’aménagements
hydroélectriques. Sur un parc de plusieurs centaines de barrages, cela constitue
un patrimoine unique de données et de connaissances acquises, et conforte la
nécessité d’une bonne organisation et de l’amélioration continue des pratiques.
Après un rappel du cadre et du périmètre de l’auscultation des barrages EDF,

le rapport développe deux aspects essentiels de l’activité d’auscultation:
• La maîtrise sur de longues périodes de temps de la qualité des mesures

d’auscultation réalisées sur les barrages. Celle-ci repose sur un ensemble
d’activités variées, qui va de la qualification de matériels à la formation des
exploitants en charge des mesures,
• La gestion d’un processus de surveillance dynamique, exploitant toute
l’information remontée au fil de l’eau afin de détecter et traiter au plus
112
Q.106 - R.11
tôt les précurseurs d’anomalies de comportement. Les résultats obtenus à

l’échelle du parc, illustrés par un cas concret, confirment l’apport effectif de
l’auscultation à la maîtrise de la sûreté des barrages.
La conclusion synthétise les principaux enseignements de cette longue et

large expérience, et formule des recommandations pouvant être reprises dans
d’autres contextes et par d’autres organisations.
2. ÉLÉMENTS GÉNÉRAUX SUR L’AUSCULTATION DES

BARRAGES EDF
2.1. DE 1982 À 2020, UNE FORTE ÉVOLUTION DU PÉRIMÈTRE
Certains chiffres du rapport de 1982 [1] étaient déjà marquants, mais leur
réactualisation (cf. Tableau 1) permet de mesurer tout le poids de la continuité des
activités de surveillance, qui se déroulent sans interruption.
Cet accroissement notable du périmètre s’explique par la mise en service de

nouveaux aménagements, mais surtout par la mise en place d’une auscultation sur
de plus petits barrages. En effet, le contexte réglementaire de 1982 ciblait surtout
les grands barrages (H > 20 m); la loi sur l’Eau de 2006 et les textes qui ont suivi ont
conduit à renforcer la surveillance des ouvrages de taille plus modeste. La présence
d’un dispositif d’auscultation est ainsi exigée pour tous les barrages classés A, B et
C (ouvrages présentant des enjeux de sécurité publique), avec un contrôle et des
inspections effectués par les services de l’État [2] et [3].
Les figures ci-dessous présentent les caractéristiques principales du parc de

barrages actuellement auscultés par EDF.
L’âge moyen des barrages auscultés s’établit actuellement à 67 ans, âge cor-
respondant à la décennie [1950, 1959] au cours de laquelle on observe le pic du
nombre de barrages mis en service.
La typologie des barrages est très diversifiée: la majorité est en béton, princi-
palement des poids (dont environ 20% en maçonnerie) et des voûtes, barrages
Tableau 1
Évolution de chiffres significatifs du parc EDF (1981–2019)
FIN 1981 FIN 2019
Nombre de barrages auscultés 122 329
Équivalent années-barrage d’auscultation en exploitation 2 500 12 200
Nombre de points de mesure (hors repères topographiques) 6 000 13 000
113
Q.106 - R.11
Fig. 1
Âge des barrages du parc EDF
Age of EDF dams
Fig. 2
Typologie des barrages du parc EDF
Typology of EDF dam
mobiles en rivière (BMR) et contreforts. Après la seconde guerre mondiale, le

nombre des barrages en remblais s’est accru et l’équipement en centrales hydro-
électriques s’est poursuivi, ce qui a entraîné la construction d’un linéaire important
de digues de canaux.
114
Q.106 - R.11
Fig. 3
Hauteur des barrages du parc EDF
Height of EDF dams
Fig. 4
Altitude des barrages du parc EDF
Altitude of EDF dams
Les caractéristiques géométriques sont très variables. La hauteur moyenne

est de 33 m, le plus haut (180 m) étant le barrage-voûte de Tignes. Un nombre
significatif de barrages sont construits en zone montagneuse (∼20% à une altitude
supérieure à 1000 m), avec des conditions météorologiques parfois sévères.
115
Q.106 - R.11
Les caractéristiques des dispositifs d’auscultation sont adaptées aux

ouvrages, à leur environnement et aux enjeux associés. Quelques points de repères
sont donnés ci-dessous:
• 40 points de mesure en moyenne par barrage (hors repères topographiques,

mais incluant les mesures individuelles de débits sur les drains), le plus instru-
menté ayant environ 300 points de contrôle. Cet indicateur a diminué par
rapport à 1982, ce qui s’explique par la réduction des dispositifs dans les
années qui suivent la mise en eau, et un recentrage sur les phénomènes les
plus pertinents pour la surveillance (en lien avec les modes de défaillance) et
le diagnostic de comportement (cf. §3.1),
• 136 barrages avec un dispositif de mesures topographiques (planimétrie et/ou
nivellement): il s’agit des barrages sur lesquels les mesures de déforma-
tions/déplacements sont difficilement envisageables par d’autres moyens (tels
que pendules habituellement),
• 33 barrages équipés d’un système de télémesure: ce sont les ouvrages forte-
ment instrumentés, avec une surveillance nécessitant des mesures à pas de
temps rapproché (périodicité horaire à journalière), ou d’accès difficile (en
altitude par exemple),
• 31 barrages équipés de systèmes d’alarmes, dont l’activation conduit à une
action immédiate de l’exploitant: concerne surtout les ouvrages en rem-
blais (débits et piézométrie) pour lesquels des cinétiques d’évolution rapide
pourraient se manifester (également des glissements de terrain).
2.2. L’ORGANISATION EDF, LES OUTILS ET LES MÉTHODES
Le processus de surveillance est défini au niveau national par EDF, et mis

en œuvre de manière homogène sur l’ensemble des barrages. Ce processus est
structuré autour de:
• Un référentiel commun d’exigences et de procédures d’exploitation et de

maintenance,
• Une standardisation des dispositifs d’auscultation (cf. §3),
• Un système d’information unique et partagé par tous les intervenants (logiciel
KOALA),
• Une offre de formation centralisée et un dispositif d’habilitation des personnels
intervenant sur site.
Sa mise en œuvre sur chaque barrage s’appuie sur 3 différents niveaux de

responsabilités portés par différentes entités EDF:
• Celle de maître d’ouvrage, au niveau régional: responsable de la sécu-

rité du barrage, de la relation avec les services de contrôle de l’État, de
116
Q.106 - R.11
l’établissement des consignes de surveillance et d’auscultation, de l’allocation

des moyens et ressources nécessaires,
• Celle de l’exploitant, au niveau local: responsable de la mise en œuvre des
consignes, de la réalisation des actions de surveillance, de l’analyse de pre-
mier niveau des mesures d’auscultation, de la maintenance courante des
dispositifs,
• Celle des services d’ingénierie: préconisation d’évolution du dispositif et
du contenu des consignes, réalisation d’activités techniques spécialisées
(mesures topographiques, interventions sur le dispositif), analyse de deux-
ième niveau des mesures d’auscultation, instruction et traitement des écarts
en relation avec le maître d’ouvrage et l’exploitant (cf. §4.1).
3. GARANTIR LA QUALITÉ DES MESURES RÉALISÉES

DANS LA DURÉE
3.1. CYCLE DE VIE DU DISPOSITIF D’AUSCULTATION SUR UN BARRAGE
Le dispositif d’auscultation et de surveillance d’un aménagement hydraulique

n’est pas immuable: il s’agit d’un composant vivant, qui peut être amené à évoluer au
fil de la vie de l’ouvrage afin d’en prendre en compte les contraintes propres (modifi-
cation de comportement, changement d’exploitation…), mais également extérieures
(contraintes réglementaires…).
La logique de conception d’un dispositif d’auscultation procède d’une anal-

yse de risques, dans la mesure où sa finalité première est de détecter des signes
précurseurs d’anomalie. Ceux-ci variant en fonction du type d’ouvrage ausculté, les
moyens mis en place pour les détecter sont adaptés au cas par cas. Cette mise
au point fait l’objet d’un processus de définition piloté par l’ingénierie en charge
de l’auscultation des ouvrages. La mise en exploitation du dispositif répond à une
logique similaire: en fonction des phénomènes précurseurs redoutés, les modalités
d’auscultation du dispositif, notamment les fréquences associées, sont élaborées
par l’ingénierie en charge de l’auscultation. C’est en règle générale l’exploitant qui
se charge de la réalisation des gestes liés à l’auscultation, même si certaines presta-
tions spécialisées peuvent être confiées à des prestataires dédiés; le dépouillement
et l’analyse de ces données est à la charge de l’ingénierie.
La maintenance des dispositifs est elle aussi déclinée selon plusieurs

principes: si la maintenance courante est du ressort de l’exploitant, des opérations
de maintenance spécialisée et de vérification de la conformité métrologique des
dispositifs sont réalisées par des services spécialisés, à des périodicités définies
pour l’ensemble du parc.
L’adéquation du dispositif d’auscultation à l’ouvrage surveillé est régulièrement

réinterrogée au fil du temps. Le dispositif peut en effet être appelé à évoluer pour
117
Q.106 - R.11
assurer son maintien en conditions opérationnelles (obsolescence de composants,

défaillances…), mais aussi du fait de la ré-interrogation de la pertinence de certains
appareils. Les grandes échéances réglementaires que sont les études de dan-
gers, que la réglementation française rend obligatoires pour les «grands barrages»
(ouvrages de classe A et B) à échéance de 10 ou 15 ans, constituent une opportunité
de réexamen de la pertinence du dispositif d’auscultation à la lumière d’une analyse
de risques exhaustive de l’ouvrage concerné. Le dispositif d’auscultation est en effet
examiné en tant que «barrière» de prévention dans la survenue d’événements non
désirés: cette analyse peut conduire l’étude de dangers à proposer une modification
ou un étoffement du dispositif.
Au fil de la vie du dispositif, certains composants peuvent être affectés de

défaillances, ou simplement parvenir en fin de vie. Si les capteurs directement
accessibles peuvent faire l’objet d’un remplacement, il en va différemment des
composants «noyés» (capteurs piézométriques scellés, cellules de pression inter-
stitielle, témoins sonores…). Dans ce dernier cas de figure, une analyse est réalisée
au cas par cas afin de statuer sur la marche à suivre pour maintenir un dispositif
d’auscultation adapté. La conception initiale du dispositif doit tenir compte, par des
mesures adaptées (accessibilité, redondance…), de ces limites et de la durée de
vie des différents composants.
Une fois défini, mis en place et en exploitation, un dispositif d’auscultation doit

offrir toutes les garanties de conformité technique tout au long de son cycle de vie.
Pour cela, il est nécessaire de définir une politique industrielle robuste qui garantisse
la satisfaction aux exigences de l’auscultation et intègre le retour d’expérience de
l’exploitation de tels dispositifs sur de nombreux aménagements.
3.2. POLITIQUE INDUSTRIELLE MISE EN ŒUVRE POUR GARANTIR DANS

LA DURÉE LA QUALITÉ DES MESURES D’AUSCULTATION
Pour les grandeurs habituellement mesurées dans la durée sur un barrage,

EDF a défini et mis en place une politique et une organisation s’appuyant sur son
système qualité (ISO 9001), qui permettent de garantir dans la durée des mesures
fiables, pérennes et contrôlables. Pour des besoins de mesures temporaires, ces
procédures peuvent être allégées.
Cette politique concerne l’ensemble des matériels qui constituent les disposi-
tifs d’auscultation (les capteurs et appareils de mesure, les équipements associés
aux dispositifs d’auscultation, les systèmes d’acquisition et de transmission des
données…) et implique tous les acteurs en lien avec l’auscultation [4]. Les axes
principaux sont exposés dans la suite de ce paragraphe, en prenant l’exemple
des mesures de déplacements par pendules (directs ou inversés) pour l’illustrer
concrètement.
118
Q.106 - R.11
Tableau 2
Exigences métrologiques pour les mesures de déplacement
AMPLITUDE INCERTITUDE MAXIMALE
GRANDEURS PHYSIQUES MAXIMALE DE MESURE TOLÉRÉE (À +/−2 ÉCARTS-TYPE)
Déplacements absolus horizontaux: 50 mm ±2 mm
Ouvrages en béton (tous types)
Déplacements absolus verticaux: 50 mm ±1 mm
Ouvrages en béton (tous types)
3.2.1. Politique industrielle d’EDF concernant les dispositifs

d’auscultation
Le domaine du génie civil se distingue des autres milieux industriels par des
durées de vie des dispositifs qui sont particulièrement longues, une grande disper-
sion géographique des ouvrages, des environnements disparates, des conditions
climatiques parfois sévères, une culture de la mesure et de la métrologie plutôt faible
et une offre de matériels limitée. Une politique industrielle d’Entreprise présente
l’intérêt de fixer des exigences et des règles internes sur le domaine d’activités
essentiel qu’est l’auscultation des ouvrages, pour lequel les références normatives
sont insuffisantes ou inadaptées.
Cette politique industrielle structure le processus de qualification, de

développement et d’exploitation des composants des dispositifs d’auscultation
d’EDF. La standardisation permet d’obtenir une forte connaissance des technolo-
gies, de mettre en place des procédures techniques pour l’installation des capteurs,
la réalisation des mesures, le contrôle et la maintenance des composants, de for-
mer et maintenir les compétences des personnels, de limiter les référentiels et
d’optimiser les relations avec les différents fournisseurs.
3.2.2. Définition des exigences métrologiques
En réponse aux besoins de l’auscultation, pour chaque phénomène (ou

grandeur) physique ausculté, des exigences en termes d’amplitude et d’incertitude
de mesure ont été définies. Elles formalisent les besoins de l’ausculteur pour assurer
le suivi du comportement d’un ouvrage avec un niveau de confiance satisfaisant.
Ces incertitudes sont indépendantes des méthodes et des appareils de mesure util-
isés. Elles ont été définies «à dire d’expert» et sont le fruit du retour d’expérience
important acquis sur la connaissance du comportement des grands ouvrages de
génie civil: un exemple en est donné dans le Tableau 2.
119
Q.106 - R.11
3.2.3. Qualification des matériels d’auscultation et de leurs fournisseurs
La qualification des matériels d’auscultation se fait à la suite d’un processus

permettant de s’assurer de l’adéquation des matériels aux exigences métrologiques
et fonctionnelles (fiabilité, résistance aux environnements difficiles, ergonomie…)
pour répondre aux besoins techniques, mais aussi valider la capacité à travailler
durablement avec le fournisseur (qualité de fourniture et de service, maîtrise des
coûts, service après-vente, organisation et compétences internes, relation de
confiance…). Cette validation des matériels se déroule en plusieurs étapes:
• Veille technologique via un retour d’expérience interne, externe avec nos

partenaires, rencontres/visites de fournisseurs…
• Validation métrologique et fonctionnelle en laboratoire (interne EDF et/ou
externe) pour s’assurer que les matériels répondent à nos exigences
métrologiques et fonctionnelles,
• Réalisation d’essais sur site pour tester les matériels en conditions réelles
d’utilisation,
• Établissement d’un dossier de synthèse de la qualification, incluant les référen-
tiels nécessaires à l’installation, l’exploitation et au maintien de la qualité des
dispositifs dans la durée.
3.2.4. Installation et réception des matériels d’auscultation
Pour chaque type de matériel, une procédure d’installation et de réception a

été rédigée. Ces procédures permettent de garantir l’installation de chaque matériel
suivant des exigences définies. L’installation d’un dispositif d’auscultation fait l’objet
de contrôles et d’une réception par un service d’ingénierie d’EDF (points d’arrêt,
réceptions partielles/finale) quelle que soit l’organisation de chantier.
EDF assure la traçabilité de ses interventions et le relevé de tous les

paramètres nécessaires à l’exploitation future des matériels dans le logiciel
d’auscultation KOALA. La maîtrise de ces paramètres, tout au long de la vie du
dispositif, est essentielle car ce sont eux qui garantissent la conversion correcte des
mesures en grandeurs physiques pour chacun des appareils installés.
3.2.5. Exploitation et maintenance courante des dispositifs d’auscultation
Les exploitants des ouvrages ont en charge la réalisation des mesures

d’auscultation ainsi que l’entretien courant des dispositifs. Les exploitants chargés
des mesures sont formés dans le cadre de stages spécifiques, de tutoriels, de
formations en ligne ou encore par compagnonnage.
120
Q.106 - R.11
Les mesures sont réalisées suivant un mode opératoire accompagné de pré-

conisations de maintenance de premier niveau pour chacun des appareils. Ces
gestes du quotidien sont indispensables au bon fonctionnement des matériels et à
la qualité des mesures réalisées.
Pour les ouvrages équipés de systèmes télémesurés, la réalisation des

mesures et le rapatriement des données sont effectués à distance par l’exploitant
avec le logiciel d’auscultation KOALA.
3.2.6. Contrôles métrologiques et maintenance corrective
En complément de l’entretien régulier effectué par les exploitants lors des

tournées d’auscultation, un contrôle périodique des dispositifs est réalisé par un
service d’ingénierie d’EDF. Ce contrôle comporte deux volets complémentaires: le
«contrôle fonctionnel» du dispositif pour vérifier que l’appareil d’auscultation fonc-
tionne de manière appropriée dans son environnement, ainsi qu’une «vérification
métrologique» pour contrôler que les capteurs accessibles sont conformes à un
critère métrologique (EMT: Erreur Maximale Tolérée) calculé à partir des exigences
métier métrologiques (cf. §3.2.2).
Ce contrôle préconise les actions d’entretien et éventuellement de remise

en conformité. La périodicité des contrôles n’est pas figée, et peut évoluer au
cours du temps. Elle est définie à partir des résultats des contrôles périodiques
et du retour d’expérience des problèmes constatés; elle est actuellement fixée à
5 ans sur la plupart des barrages. Par ailleurs, l’exploitant effectue la vérification
des appareils mobiles qui permettent de réaliser les mesures (sondes lumineuses,
pieds à coulisse…).
Pour les ouvrages disposant d’une télémesure, des mesures simultanées

entre les appareils manuels et automatiques sont faites périodiquement. Elles
sont comparées afin de vérifier leur cohérence. Elles permettent de détecter une
dérive éventuelle d’un capteur et s’assurer que les mesures sur l’appareil manuel
demeurent opérationnelles en cas de défaillance de la télémesure.
Pour finir, la défaillance ou le dysfonctionnement d’un matériel est instruit et

traité par l’exploitant et l’ingénierie pour établissement du diagnostic, mise en place
de mesures compensatoires temporaires et mise en œuvre des actions correctives
pour la résolution de l’écart.
3.3. EXEMPLE DE GESTION DU RETOUR D’EXPÉRIENCE

MATÉRIEL SUR UN CAPTEUR
À partir de toutes les données récupérées sur les matériels exposées

précédemment, EDF effectue un retour d’expérience sur les matériels qui
121
Q.106 - R.11
Fig. 5
Qualification du SAFTEL TOP 2 en laboratoire
SAFTEL TOP 2 laboratory qualification
composent les dispositifs d’auscultation en prenant en compte les dépan-

nages/défaillances, les éléments issus de la maintenance corrective, les obsoles-
cences de composants…
Cela permet d’orienter les besoins en qualification ainsi que les recherches
de nouveaux capteurs mais également de revoir le contenu et la périodicité de notre
politique de maintenance.
On prend ci-dessous pour exemple la qualification d’une nouvelle version du

télépendule SAFTEL TOP version 1 initialement développée en 1992. En effet, la
gestion de l’obsolescence de ce capteur avait identifié l’arrêt de commercialisation
de composants nécessaires à sa fabrication, tandis que les besoins à court/moyen
terme des maîtres d’ouvrages (principalement pour le parc nucléaire) étaient large-
ment supérieurs au stock disponible. Cela s’est vérifié en 2015, alors qu’il n’était
plus possible de fabriquer le capteur existant dénommé SAFTEL TOP 1.
En 2014, EDF a validé le développement d’un nouveau palier technique de

ce matériel alors que les contrôles montraient une très bonne fiabilité du capteur
avec un taux de défaillance qui restait inférieur à 2% annuellement sur environ 600
capteurs installés.
Après avoir déroulé tout le processus de qualification (cf. §3.2.3) avec la réali-
sation d’actions de développement (refonte du système optique, refonte de la carte
122
Q.106 - R.11
mère, développement d’un nouvel applicatif…), d’essais en laboratoire (essais fonc-

tionnels, essais de compatibilité électromagnétique, essais de sensibilité thermique,
essais métrologiques…), d’essais sur site (installation sur le barrage de Rizzanèse
pendant plus d’un an) et de rédaction du référentiel, EDF a validé le déploiement
de la nouvelle génération de télépendules SAFTEL TOP 2 en 2017.
4. LES APPORTS DE L’AUSCULTATION À LA SÛRETÉ DES BARRAGES
4.1. L’AUSCULTATION: UNE LIGNE DE DÉFENSE DYNAMIQUE
4.1.1. Mise en œuvre et pilotage du processus de surveillance
L’auscultation des barrages constitue une ligne de défense fondamentale de

la politique de sûreté hydraulique d’EDF. Elle s’appuie sur une organisation et des
exigences dont la robustesse et les résultats produits sont contrôlés et évalués
annuellement. L’objectif est de corriger et améliorer le processus en continu, selon
des principes d’assurance qualité bien connus.
Le déroulement du processus d’auscultation sur un barrage peut être résumé

de la manière suivante:
• L’exploitant local programme et réalise les tournées d’auscultation sur son

barrage selon les indications de la consigne de surveillance et d‘auscultation.
Après leur réalisation, il dispose de 24h pour leur saisie dans le système
d’information. Il contrôle (niveau 1) la complétude de la tournée, la cohérence
des lectures et l’absence d’écart majeur par rapport aux mesures précédentes,
• Une permanence d’un service d’ingénierie EDF valide ensuite sous 48 heures
l’absence d’anomalie de comportement du barrage (niveau 2).
Si le barrage n’est pas télémesuré, la collecte des mesures est assistée par un
terminal de saisie de terrain contenant: la liste des appareils de la tournée à réaliser
et les valeurs des lectures acquises précédemment, ce qui permet de réduire les
risques d’oubli de mesure de capteurs et de contrôler la lecture in situ. Les étapes
de transmission et de traitement des mesures (transformation de la lecture en valeur
de grandeur physique) sont complètement automatisées.
Les niveaux 1 et 2 de contrôle s’appuient sur des algorithmes de détec-

tion d’écarts, un écart pouvant être défini et configuré dans le logiciel KOALA de
différentes manières:
• Valeur en dehors de la plage de mesure normale du phénomène,

• Valeur dépassant un seuil défini de manière physique ou statistique,
123
Q.106 - R.11
• Valeur du phénomène différente de celle modélisée à partir de son historique,

en prenant en compte un niveau d’incertitude raisonnable (généralement 2σ),
• Valeur du phénomène en écart par rapport à un groupe de phénomènes
statistiquement comparables.
Des indicateurs annuels permettent de contrôler et de s’assurer de la

robustesse du processus, les principaux mesurant:
• Le respect du contenu des consignes de surveillances: décompte des

tournées d’auscultation et des mesures topographiques non réalisées, tout
écart significatif à la consigne devant faire l’objet d’une déclaration au service
de contrôle de l’État,
• Le respect des exigences internes de déroulement du processus de réalisa-
tion, traitement et validation des tournées d’auscultation (niveau 1 exploitant
et niveau 2 ingénierie) et des mesures topographiques: suivi des délais entre
ces étapes et de leur évolution au cours du temps,
• Le bon fonctionnement du système d’information: mesure et suivi des temps
d’indisponibilité, erreurs détectées et correctifs apportés.
Ces indicateurs sont produits pour chaque barrage et agrégés au niveau de

nos différentes strates d’organisation (régionales et nationale), ce qui permet de
détecter des écarts à caractère générique ou local. Par exemple, le non-respect
du délai de traitement des tournées sous 48 heures va conduire à des actions
de sensibilisation et éventuellement de formation des techniciens du groupement
d’usines concerné.
4.1.2. Résultats du processus de surveillance
L’analyse du fonctionnement et des résultats produits par le processus, dans

son ensemble, est probablement plus inhabituelle au sein de la profession, mais
pour autant essentielle car elle légitime et justifie les moyens mis en œuvre ainsi
que l’apport de l’auscultation à la sûreté des barrages. C’est aussi ce qui permet de
donner du sens à une activité qui est souvent vécue comme routinière, répétitive et
peu motivante.
À l’échelle du parc de barrages EDF, l’auscultation représente (chiffres pour

l’année 2019):
• 9 500 tournées d’auscultation,

• 340 mesures topographiques,
• et in fine, 560.000 mesures acquises, traitées et validées.
Les niveaux 1 et 2 de contrôle ont détecté environ 1 900 écarts (0,3% du

total des mesures) conduisant à une instruction dont la complexité et la durée sont
extrêmement variables en fonction de la nature de l’écart: de la simple confirmation
124
Q.106 - R.11
d’une mesure erronée à des actions beaucoup plus contraignantes en cas de

suspicion d’une anomalie de comportement ou de franchissement d’un seuil asso-
cié à un critère de sûreté (renforcement de la surveillance, études/investigations
complémentaires, restrictions d’exploitation).
La typologie des écarts qui ressort des instructions menées est la suivante:
• 3% sont des erreurs de configuration dans le logiciel d’auscultation (paramètre

de configuration capteur erroné, seuil mal calé…),
• 11% sont des problèmes matériels sur le dispositif d’auscultation (défaillance
capteur, problème fonctionnel…),
• 14% sont des erreurs de lecture ou de saisie,
• et les 72% restants sont des valeurs inhabituelles de phénomènes physiques.
Dans la très grande majorité des cas, l’écart s’explique par une situation
météorologique atypique (pluie, fonte de neige, situations thermique atyp-
iques…) ou des perturbations amenées par des travaux ou modifications sur
le barrage ou le dispositif. Le(s) phénomène(s) retrouve(nt) généralement une
valeur habituelle lors du retour aux conditions habituelles d’exploitation, ce qui
conduit alors à la fin de l’instruction et à la clôture de l’écart.
Présenté autrement, cela revient à dire que sur chacun de nos barrages,
nous détectons en moyenne 6 écarts par an: 1 lié à un problème d’acquisition de la
mesure, 1 lié à un problème matériel sur le dispositif et 4 traduisant des variations
de phénomènes nécessitant une instruction mobilisant l’exploitant et l’ingénierie. Il
s’agit d’une indication générale, bien entendu variable selon le barrage et son dis-
positif d’auscultation, et qui n’est pas forcément transposable dans un autre contexte
d’entreprise et de parc de barrages.
Pour finir, il est important de préciser que le nombre d’écarts constituant

d’éventuels signes de l’initiation d’un possible mécanisme de défaillance est très
limité: de l’ordre d’une dizaine par an. Sur ces épisodes, on mesure pleinement la
plus-value et la contribution de l’auscultation à la sûreté. Avec du coup, le constat (et
c’est heureux) que la probabilité qu’une tournée d’auscultation conduise à détecter
et prévenir une situation ayant un enjeu sûreté est très faible. Mais cela contribue
aussi à la difficulté, déjà mentionnée, de motiver dans la durée les techniciens en
charge de la surveillance. C’est pour cela que le partage de ces situations excep-
tionnelles a été identifié comme un moyen de donner du sens et de maintenir dans
la durée la rigueur et la vigilance indispensables à la réalisation des tournées sur
site. Pour illustrer le propos, c’est une situation de ce type vécue récemment qui est
mise en avant dans le paragraphe suivant.
4.2. ILLUSTRATION SUR UN CAS CONCRET: LE BARRAGE DE PONT-DES-CHÈVRES
On expose, dans le cas concret présenté ci-dessous, un exemple d’anomalie

de comportement détectée par l’auscultation courante ayant conduit à l’adaptation
125
Q.106 - R.11
Fig. 6
Vue globale de l’aménagement de Pont-des-Chèvres
Global view of Pont-des-Chèvres dam
de la surveillance aux phénomènes constatés et au déclenchement de recon-

naissances complémentaires afin de permettre le maintien de l’exploitation de
l’aménagement avec un niveau de sûreté acceptable.
Le barrage de Pont-des-Chèvres est situé sur l’Arc, dans la vallée de la Mau-

rienne. Il s’agit d’un barrage en enrochements de 18 mètres de haut, sur fondation
alluvionnaire. L’étanchéité est assurée par un masque béton pour le corps de rem-
blai et une paroi moulée en fondation. Cet aménagement, situé au cœur de la
chaîne hydroélectrique de l’Arc (2300 MW de puissance installée, deux stations de
transfert d’énergie par pompage), est constitué d’une digue «Arc», barrant le cours
d’eau principal, et d’une digue «usinière», les deux ouvrages étant séparés par un
terre-plein supportant les voies de circulation (routière, autoroutière et ferroviaire)
desservant la vallée.
Ce barrage présente la double particularité d’être affecté, depuis sa mise

en eau en 1969, de fuites élevées (de l’ordre de 1000 l/s) par la fondation et
l’étanchéité amont et d’être le siège de tassements non amortis. Ce comporte-
ment en écart par rapport aux données de conception initiales justifie l’existence
d’un dispositif d’auscultation particulier et adapté aux pathologies rencontrées: le
barrage est équipé d’un réseau de nivellement, relevé à fréquence mensuelle, et de
piézomètres, qui font l’objet de relevés hebdomadaires. La configuration du barrage
ne permettant pas le captage et la mesure directe des fuites, une mesure par explo-
ration des champs de vitesse à l’aval du barrage est réalisée annuellement selon
126
Q.106 - R.11
Fig. 7
Résultats de l’étude InSAR
Results of the InSAR survey
une procédure complexe: cette mesure permet de surveiller l’évolution des débits
transitant en fondation et de prévenir le risque d’érosion interne sous le barrage. La
surveillance est complétée par des examens visuels réalisés au pas hebdomadaire
par l’exploitant.
Suite à la réalisation de la mesure de nivellement de juin 2019, l’ingénierie en

charge de l’auscultation informe l’exploitant d’une accélération faible mais anormale
des tassements sur les repères situés en rive droite du barrage. La vérifica-
tion de la conformité intrinsèque de la mesure n’ayant montré aucune anomalie,
l’exploitant procède alors à des vérifications visuelles, complétées par un examen
visuel exhaustif à l’occasion d’une chasse réalisée quelques jours plus tard, afin
de détecter tout indice d’une dégradation de l’étanchéité amont ou d’une érosion
interne en fondation, sans qu’aucun phénomène anormal ne soit signalé.
Les mesures suivantes confirment l’existence d’une cinétique anormale, car-

actérisée par une nette accélération affectant principalement la rive droite du barrage
«Arc». Certains des repères de nivellement voient des cinétiques multipliées par un
facteur supérieur à 15, avec des vitesses jamais observées depuis la mise en eau.
La mesure de nivellement d’octobre 2019 ayant confirmé le changement de

comportement constaté depuis le mois d’avril, il a été décidé d’adapter la surveil-
lance de l’ouvrage aux phénomènes observés. Cet état de surveillance renforcée a
conduit aux actions suivantes:
• Modification de la surveillance opérationnelle sur le barrage: nivellement passé

du pas mensuel au pas 15 jours, examen visuel de l’exploitant et relevés
piézométriques passés du pas hebdomadaire au pas quotidien,
127
Q.106 - R.11
Fig. 8
Nivellement du repère 1 jusqu’en août 2020
Leveling graphs of mark 1 until August, 2020
• Réalisation d’investigations visant à détecter un éventuel changement de

comportement hydraulique de la fondation: mesure de fuites à conditions
d’exploitation identiques, mesures par traçage.
Ces contrôles ont permis d’écarter tout risque d’érosion interne en fondation;
les mesures de nivellement continuaient quant à elles à montrer le maintien de ciné-
tiques anormalement élevées sur l’ensemble du barrage «Arc», particulièrement en
rive droite.
En complément de la surveillance renforcée, il a été décidé de réaliser une

étude interférométrique (InSAR) afin de contrôler le comportement du versant rive
droite, celui-ci étant le siège d’un chantier de travaux souterrains. Cette étude a
montré une bonne cohérence avec les mesures de nivellement, et révélé une nette
subsidence du versant, concomitante avec le début des tassements observés sur
le barrage.
En janvier 2020, les mesures de nivellement ont mis en évidence un début

d’amortissement des tassements, confirmé par les mesures suivantes. L’absence
d’évolution des débits transitant en fondation ayant été confirmée par une mesure à
conditions identiques et compte tenu des résultats de l’étude InSAR, il a été décidé
fin mars 2020 de lever la surveillance renforcée.
Ce barrage continue toutefois de faire l’objet d’une attention soutenue, en

particulier par le biais de la surveillance opérationnelle réalisée. À ce jour, les raisons
de la subsidence généralisée du versant rive droite sont encore mal comprises;
des investigations complémentaires sont à l’étude afin d’affiner la connaissance du
comportement du versant.
128
Q.106 - R.11
En conclusion, la surveillance opérationnelle réalisée sur cet ouvrage a mon-

tré son efficacité en permettant de détecter immédiatement les signes précurseurs
d’une évolution anormale, pouvant potentiellement avoir des impacts sûreté.
L’adaptation de la surveillance aux phénomènes rencontrés et la mise en place
de mesures de surveillance renforcée adaptées ont permis de maintenir l’ouvrage
en exploitation dans des conditions de sûreté acceptables.
5. CONCLUSION
Il n’y a pas de conclusion définitive sur les meilleures pratiques à met-

tre en œuvre pour l’auscultation des barrages, chacun mettant en œuvre le
processus le plus adapté aux exigences réglementaires, à son organisation
et ses moyens et compétences internes. Forts de leur vécu au quotidien sur
un grand parc de barrages, les auteurs de ce rapport souhaitent néanmoins
conclure par quelques recommandations permettant d’assurer que l’auscultation
joue efficacement son rôle de ligne de défense de la sûreté des barrages, et
ce de manière pérenne. Elles sont proposées ci-dessous sous la forme de
«10 règles»:
• Réinterroger périodiquement la pertinence du dispositif vis-à-vis des modes

de défaillances possibles,
• Assurer la qualité et la performance du dispositif par un choix maîtrisé des
capteurs et des fournisseurs, en intégrant les progrès technologiques,
• Maintenir en conditions opérationnelles les dispositifs d’auscultation (fonction-
nel, métrologique), afin de garantir la qualité et la continuité de l’auscultation,
• Former et motiver les personnels dans la durée,
• Piloter le processus de réalisation des mesures et leur traitement en continu,
avec des exigences clairement définies et contrôlables, et des responsabilités
bien identifiées,
• Disposer de 2 niveaux de validation des mesures, afin de renforcer la
robustesse du processus et la capacité de détection des évènements anor-
maux,
• Sécuriser l’acquisition, le traitement et la conservation des données dans un
système d’information adapté,
• Instruire/investiguer tout signal inhabituel ou anormal, en sollicitant toutes
les compétences et responsabilités utiles (exploitants, services d’ingénierie,
maître d’ouvrages, services de contrôle),
• Être convaincu que chaque mesure d’auscultation a une valeur immédiate
pour la surveillance et le diagnostic, mais qu’elle sera aussi utile et valorisée
demain en réponse à d‘autres finalités,
• Donner au processus de surveillance les ressources et les moyens néces-
saires, avec une maîtrise des coûts proportionnée aux enjeux.
129
Q.106 - R.11
REFERENCES
[1] LE MAY Y., DOUILLET G. Trente ans d’expérience dans la surveillance

et l’auscultation des barrages d’Électricité de France. 14th ICOLD-CIGB
Congress, Rio de Janeiro. 1982. Q52 R13.
[2] LE DELLIOU P., GUILLOTEAU T., PEREZ M., LE VERGER C. New French
regulations concerning dams; improving practices to ensure their safety. 24th
ICOLD-CIGB Congress, Kyoto. 2012.
[3] LE DELLIOU P., HOONAKKER M. Évolution de la réglementation des barrages

en France. Colloque «Sûreté des barrages et Enjeux». Chambéry: CFBR,
23–24 novembre 2016.
[4] PAVAILLIER F., PIRON A. Auscultation des ouvrages de Génie Civil – Maîtrise
de la qualité de la mesure. Colloque «Auscultation des barrages et des
digues». Chambéry: CFBR, 27–28 novembre 2012.
130
Q.106 - R.12
-------
-------
MÉTHODES DE SURVEILLANCE DES FUITES PAR FIBRE OPTIQUE,

MISE EN ŒUVRE AU SEIN D’UNE ORGANISATION DE
SURVEILLANCE DES BARRAGES ET NOUVELLE
APPLICATION AUX OUVRAGES D’ALTITUDE (∗)
François MARTINOT
Spécialiste fibre optique, Division Technique Générale, EDF HYDRO
Guillaume STOLTZ
Ingénieur de Recherche, INRAE, Centre PACA
Cyril GUIDOUX
Expert fibre optique, geophyConsult
Maxime BOUCHER
Ingénieur fibre optique, geophyConsult
Sylvie NICAISE
Jean-Robert COURIVAUD
Expert barrages en remblai, Centre d’Ingénierie Hydraulique, EDF HYDRO
Laurent PEYRAS
FRANCE
∗ Fiber optic leak monitoring methods, industrialisation for dyke monitoring and innovative
applications to lining systems of altitude dams
DOI: 10.1201/9781003211808-108 131

Q.106 - R.12
RÉSUMÉ
La technologie de mesure répartie par fibre optique, qui permet la mesure

de températures sur des linéaires plurikilométriques avec une grande résolution
de mesure, est arrivée à maturité technique depuis quelques années. Le présent
article décrit les progrès méthodologiques réalisés dans le cadre de deux champs
d’applications: i) industrialisation EDF des méthodes de détection passives des
écoulements préférentiels potentiellement initiateurs d’érosion interne dans les
ouvrages en remblai et ii) expérimentation innovante INRAE de recherche des
défauts par méthode active dans les dispositifs d’étanchéité par géomembrane
(DEG) des barrages d’altitude.
SUMMARY
The distributed fiber optic measurement technology, which allows temperature

measurement on multi-kilometer linear reaches with high measurement resolu-
tion, has reached technical maturity in recent years. This article describes the
methodological progress made in the context of two fields of application: i) EDF
industrialization of passive detection methods for preferential flows potentially initi-
ating internal erosion in embankments and ii) INRAE innovative experimentation of
fault detection with active method in the lining systems (LS) of altitude dams.
1. INTRODUCTION
La technologie de mesure répartie de température par fibre optique a trouvé

depuis plusieurs décennies de nombreuses applications dans des champs aussi
divers que la sécurité incendie, le contrôle de l’intégrité de lignes à haute tension,
l’étude des échanges hydrauliques entre nappe et rivière, etc. Dans le domaine de
la surveillance et de l’auscultation des ouvrages hydrauliques, cette technologie est
employée avec succès, depuis une vingtaine d’années, afin de détecter et surtout
de localiser des fuites pouvant initier des phénomènes d’érosion interne.
L’intérêt des mesures réparties de température pour ces applications réside

dans la capacité des systèmes employant cette technologie à relever des tem-
pératures précises (0.1◦ C) sur de longues distances (30 km), et ce en tout point
du linéaire instrumenté, avec une résolution spatiale métrique et une fréquence
d’acquisition de l’ordre de 10 minutes. Or, il est un fait bien connu que – de manière
générale – les écoulements localisés agissent comme un perturbateur du champ
de température local, et que la présence d’une anomalie thermique permet d’inférer
la présence d’un écoulement (Fig. 1).
132
Q.106 - R.12
Fig. 1
Illustration de l’effet d’une fuite sur le champ des
températures au sein d’un ouvrage en remblai
Illustration of the effect of a leak on the temperature field
within an embankment structure
Fig. 2
Chaîne de mesure fibre optique simplifiée
Simplified fiber optic measuring chain
Composée d’un câble de mesure très similaire à ceux utilisés dans les télé-
communications et d’un appareil optique compact (Fig. 2), la technologie fibre
optique apparue dans les années 1990 s’est constamment améliorée pour atteindre
les spécifications annoncées plus haut.
Par rapport à la thermométrie classique utilisée pour mener des investigations

souvent ponctuelles et localisées, la couverture spatiale, la rapidité des mesures et
la capacité de localisation des capteurs à fibre optique ont rapidement ouvert la voie
à de nouvelles applications: lever plus rapidement les réserves sur les ouvrages
neufs, cibler les zones de fragilité sur les ouvrages anciens pour effectuer des
confortements ciblés, donner aux exploitants un moyen de supervision du com-
portement des ouvrages en crue et des moyens de diagnostic complémentaires à
l’auscultation conventionnelle en contexte post-évènements tels que les séismes.
Les mesures de température ne donnant qu’un accès indirect aux

phénomènes d’écoulement recherchés, les méthodes d’analyse des données fibre
optique jouent un rôle déterminant dans l’efficacité de ces systèmes de surveillance.
Une typologie de ces méthodes accompagnée de recommandations sur leur choix
et leur mise en œuvre est proposée au chapitre 2 du rapport. Aux chapitres 3 et 4,
133
Q.106 - R.12
deux exemples d’application fournis par EDF et INRAE avec la contribution de geo-
phy Consult sont ensuite proposés afin de souligner la nécessité d’adéquation des
méthodes aux configurations d’ouvrages et aux objectifs de surveillance.
2. TYPOLOGIE DES MÉTHODES D’ANALYSE DES FUITES
Dès les premières expérimentations de détection de fuites à l’aide de mesures

de température par fibre optique réalisées à partir de 1995, il est apparu que la
visualisation de ces mesures brutes, seule, ne permettait pas, dans de nombreux
cas, d’identifier correctement les fuites: de nombreux autres facteurs que les fuites
influent en effet sur la température d’un sol et notamment, la nature géologique
des grains de sol, la position de la fibre par rapport à l’air ou au réservoir amont,
ou encore la proximité d’ouvrages traversants. Le développement de méthodes
d’analyse des données de température par fibre optique s’est donc avéré très
rapidement un enjeu majeur pour rendre cette technologie opérationnelle.
2.1. MÉTHODES PASSIVES
Les méthodes passives désignent les traitements de signaux applicables à

des mesures de température de sol différentes de celles de la retenue d’eau, poten-
tiellement influencées par l’air et la nappe (cf. Fig. 3). Ces mesures sont obtenues
grâce à des câbles optiques simples; c’est-à-dire ne comportant pas de brins de
cuivre visant à ce qu’ils puissent être chauffés pour créer un contraste thermique
temporaire.
Plusieurs méthodes passives ont été successivement développées dans

une logique de réponse à des utilisations variées, de la détection automatisée
Fig. 3
Influences externes sur les mesures exploitées par méthode passive
External influences on the measurements used by passive methods
134
Q.106 - R.12
d’anomalies thermiques en surveillance immédiate avec peu d’historique de

mesures à l’estimation fine de vitesses de fuite grâce à des modèles élaborés et à
un historique de mesures plus consistant.
2.1.1. Application en milieu saturé, surveillance long terme
Les premières méthodes de détection de fuites par capteurs à fibres optiques

ont été développées par Sam JOHANSSON – HYDRORESEARCH AB, (1997, [1]).
Elles reposent sur l’analyse des variations annuelles des signaux de température
mesurés dans l’eau de la retenue et dans un sol saturé. L’approche du temps retard
consiste à calculer la vitesse de diffusion de l’onde thermique à partir du temps retard
et de l’épaisseur de l’ouvrage. La vitesse de fuite peut ensuite être déterminée grâce
à une loi de Darcy tenant compte des capacités calorifique de l’eau et du sol saturé.
L’approche des amplitudes est basée sur une modélisation simplifiée des échanges
thermo-hydrauliques à l’intérieur de la zone de fuite. Elle fournit une relation sous
forme d’abaque entre une variation d’amplitude de la température normalisée et un
débit de fuite adimensionnel. Relativement accessibles, ces méthodes nécessitent
une mesure de température externe de l’eau du canal et un historique de mesure
suffisant. Elles ne doivent par ailleurs pas être influencées par la température de
l’air ce qui peut être le cas lorsque la fibre optique est installée à proximité de la
surface et que la nappe est profonde.
2.1.2. Application en milieu saturé ou non, surveillance long terme
Les modèles IRFA (INRAE) [2] et MORITO (EDF HYDRO) [3] permettent de
prévoir les températures mesurées par les capteurs à fibre optique dans l’hypothèse
où les transferts de chaleur au sein de l’ouvrage ne sont gouvernés que par le
phénomène de conduction de la chaleur sans advection. Connaissant les condi-
tions aux limites de température de l’air et de l’eau, des paramètres d’atténuation
et de déphasage de fonctions de réponse impulsionnelle paramétriques sont cal-
culés pour minimiser l’écart entre les températures mesurées et modélisées. Pour
être correctement exploitées, ces méthodes nécessitent une connaissance a priori
des atténuations et déphasages possibles au regard des paramètres de sol, des
mesures externes de température d’eau et un historique de mesure suffisant.
La méthode DIGUOMATIC, fruit d’un partenariat entre EDF HYDRO [4] et

geophyConsult [5,6] est basée sur une modélisation couplée thermo-hydraulique de
l’ouvrage en remblai et de sa fondation. Les influences des températures de l’eau et
de l’air sont modélisées séparément, suivant deux directions distinctes et en tenant
compte des propriétés physiques des matériaux. Le transport de chaleur par advec-
tion est modélisé pour l’écoulement. D’autres grandeurs telles que la hauteur de
nappe, sa température ou le rayonnement [5] en surface peuvent être intégrées. La
135
Q.106 - R.12
mise en œuvre de cette méthode nécessite une connaissance avancée de l’ouvrage,

de son environnement ainsi qu’un regard critique sur les choix de modélisation.
Cette méthode est particulièrement performante pour critiquer des détections issues
d’autres modèles plus simples d’utilisation ou pour préciser des vitesses de fuite.
2.1.3. Application en milieu saturé ou non, surveillance immédiate
Le modèle AJOUT développé par EDF HYDRO [8] est principalement utilisé à
des fins de détection sur un historique de données court, typiquement d’une journée
(Analyse JOUrnalière des Températures). Il repose sur une méthode de traitement
du signal réalisée en deux étapes: filtrage des influences indésirables à partir d’une
décomposition en valeurs singulières puis construction d’un indicateur journalier
d’anomalie thermique à partir des sommes de dissimilarités entre les résidus issus
de chaque position au sein des zones auscultées. La détection des anomalies ther-
miques s’effectue en deux temps: en l’absence d’historique de mesures, on cherche
d’abord les positions pour lesquelles l’indice de dissimilarité AJOUT présente des
valeurs élevées. On observe ensuite l’évolution des séries chronologiques en les
comparant à des seuils historiques.
Le calcul du modèle AJOUT suppose une acquisition des mesures à pas

horaire mais ne nécessite pas de bénéficier d’un historique de mesures conséquent:
quelques jours peuvent suffire. Il ne fait pas non plus appel à des données d’entrée
complémentaire à la température du sol comme la température de l’air ou celle de
l’eau du canal. Aucun paramètre de modélisation thermo hydraulique du sol n’est
requis.
2.2. MÉTHODE ACTIVE
La méthode active ou Heat Pulse Method (HPM) [10] est principalement util-
isée dans les cas où la fibre optique doit être implantée dans un milieu dont la
température est directement influencée par l’eau de la retenue en l’absence de
fuite. Elle fait référence à la réalisation et à l’exploitation de mesures de tempéra-
ture au sein d’un câble optique installé au voisinage de la retenue et préalablement
chauffé par effet Joule grâce à des brins de cuivre électriquement alimentés (cf.
Fig. 4).
La puissance linéique retenue – typiquement de 5 à 10 W/m – dépend de

la conductivité du sol, du temps de chauffe et de l’augmentation de température
souhaitée. En présence de fuite au voisinage du câble, la convection générée par
celle-ci va entraîner une moindre élévation de la température du câble (cf. Fig. 5).
L’interprétation des mesures repose sur la comparaison des valeurs de dT

mesurées avec une valeur de référence dTREF . Moyennant des essais d’étalonnage
136
Q.106 - R.12
Fig. 4
Influence de la température de l’eau et câble optique utilisé dans ce context
Influence of water temperature and optical cable used for the active method
Fig. 5
Principe de fonctionnement de la méthode HPM
HPM principle
en laboratoire ou sur site avec des fuites contrôlées [11], il est possible de quantifier
la vitesse de l’écoulement pour une gamme de vitesses de Darcy comprises entre
10−5 et 10−3 m/s. Le rayon de détection des fuites autour du câble optique, à l’aide
de cette méthode, est souvent moindre que pour les approches passives, de l’ordre
de 15 à 20 cm. Les moyens nécessaires à la chauffe rendent cette méthode plus
difficilement opérante sur des linéaires d’ouvrage d’une dizaine de km ou plus.
Certains paramètres techniques pouvant varier d’une chauffe à l’autre (la

puissance électrique injectée, le temps de chauffe, l’évolution des températures
extérieures), les profils de dT bruts calculés sont systématiquement recalés, en
ajustant leur moyenne et leur écart-type à ceux de la mesure de référence consid-
érée. Les valeurs de dT d’une chauffe donnée sont ensuite retranchées aux valeurs
dTREF relevées aux mêmes positions lors de la chauffe de référence. On obtient ainsi
en tout point du câble une mesure qui constitue à proprement parler le paramètre de
détection de la méthode active, à laquelle il suffit d’appliquer un seuil de détection
dépendant des conditions de terrain (variabilité spatiale caractéristiques du câble
utilisé). Cette méthodologie est nommée SACRE (Surveillance Active par Centrage
et Réduction avec Etat-zéro).
137
Q.106 - R.12
3. MISE EN ŒUVRE DE MÉTHODES PASSIVES AU SEIN DU

PROCESSUS DE SURVEILLANCE EDF
En 2019, EDF a industrialisé un processus de surveillance opérationnelle

suivant trois étapes: vérification de conformité des données d’entrée, exploitation
reproductible d’un modèle de détection polyvalent puis diagnostic de comporte-
ment. Ce processus a été appliqué sur trois remblais de configuration similaire à
celle présentée sur la Figure 3 et pour lesquels un niveau de maîtrise suffisant en
conception et en réalisation était acquis, soit un linéaire total instrumenté avoisinant
30 km. Pour le mettre en œuvre, EDF s’est appuyé sur l’expérience pratique de la
société geophyConsult.
3.1. QUALITÉ DES DONNÉES D’ENTRÉE EN AMONT DES TRAITEMENTS
La qualité des analyses étant conditionnée à celle des données d’entrée, EDF
s’assure du fonctionnement optimal de ses chaines de mesure, du bon achem-
inement des données jusqu’à l’application et de leur préparation en amont des
traitements.
3.1.1. Vérification des performances métrologiques et de

la fiabilité des matériels
Un référentiel de conception mis à jour annuellement intègre l’ensemble des

exigences métier nécessaires à la sélection des matériels et des recommandations
de mise en œuvre. Des contrôles métrologiques sont effectués par le Laboratoire
National d’Essais [12], en particulier sur deux critères importants pour la surveillance
des fuites: dispersion temporelle et résolution vraie à longue portée. Des certificats
ou preuves de durabilité sont demandés aux constructeurs de matériels afin de
limiter les opérations de maintenance ultérieures et donc les interruptions de séries
chronologiques préjudiciables aux traitements.
3.1.2. Contrôle des dispositifs sur les sites
Les principaux contrôles de conformité réalisés en chantier, à la réception et

au cours des contrôles de dispositifs périodiques (2 ans) sont les suivants:
– vérification métrologique de la chaîne de mesure. Il s’agit de s’assurer que

la fibre optique ne vieillit pas anormalement et que la qualité des éventuelles
soudures est préservée. Il s’agit ensuite de vérifier principalement, la justesse
138
Q.106 - R.12
Fig. 6
Surveillance opérationnelle des imports dans Fox’Hy
Operational monitoring of imports in Fox’Hy (software in french)
des mesures de températures dans les bobines d’étalonnage à l’aide d’une

sonde Pt100 préalablement étalonnée, la répétabilité et l’absence de dérive.
– vérification fonctionnelle. Il s‘agit de vérifier le bon état de l’appareil et de son
environnement par des contrôles visuels et des tests élémentaires.
3.1.3. Sécurisation des transferts de données
Pour contrôler la disponibilité, l’intégrité, la confidentialité, et la traçabilité des

données (critères de sécurité des systèmes d’information D.I.C.T.), EDF s’appuie sur
des infrastructures informatiques à haut niveau de service également utilisées pour
ses applications de surveillance à forts enjeux de sûreté. Des outils de supervision,
de prise en main et de maintenance élémentaire à distance permettent d’assurer une
détection rapide des interruptions de transmission, de les signaler et dans certains
cas de les résoudre rapidement à distance.
3.1.4. Vérification de la qualité des données importées
Il s’agit ensuite de vérifier que les fichiers de mesure ne présentent aucune

anomalie informatique et que les mesures sont réalisées dans le domaine nominal
d’utilisation du capteur, indépendamment du sens physique.
Une application nommée Fox’hy (Fig. 6) a été développée pour permettre à

l’équipe EDF DTG en charge du contrôle des mesures d’auscultation de niveau 2 – le
niveau 1 étant assuré par l’exploitant – de superviser la capitalisation et l’exploitation
des données. Compte tenu de la quantité de données à contrôler – de l’ordre de
deux millions par jour avant traitement – en contexte opérationnel contraint, il a été
prévu dans le logiciel un “ tableau de bord de surveillance ” qui permet de visualiser
facilement, entre autres, les défauts d’imports des données et les valeurs aberrantes
filtrées.
139
Q.106 - R.12
Fig. 7
Surveillance opérationnelle de la ventilation des données
Operational monitoring of data proofing in Fox’Hy
3.1.5. Vérification du récolement spatial et de la synchronisation temporelle
Le logiciel Fox’Hy (Fig. 7) permet également aux services de DTG de con-

trôler la préparation des données en amont des traitements. Après avoir été
associées et synchronisées, les données issues de plusieurs fibres optiques sont
récolées; c’est-à-dire qu’elles sont associées à un repère kilométrique unique, et
ce même si plusieurs fibres sont implantées à différentes hauteurs ou positions
transversales pour une même localisation longitudinale (PK). S’ensuit une opération
d’interpolation des données à pas métrique de telle sorte que même si le repère
fibre optique évolue suite à une réparation ou à un changement d’appareil, les
séries chronologiques continuent à être alimentées. Enfin, les données sont ven-
tilées dans des “zones d’auscultation” ou zones où le comportement thermique de
la digue est présumé homogène, hypothèse importante d’application du modèle
d’analyse AJOUT dont l’utilisation sera détaillée dans la suite. Un suivi informatique
du récolement garantit la traçabilité des modifications.
3.1.6. Vérification du sens physique des données
Fox’Hy comporte un module de visualisation instantanée des données. Il per-

met de s’assurer du sens physique des données reçues dans l’application: absence
de dérive ou de sauts imputables à une perte de calibration par exemple. Deux
représentations graphiques en interaction sont disponibles: un profil spatial des tem-
pératures d’une zone d’auscultation à un instant donné (graphique du haut, Fig. 8) et
une série chronologique des températures à une position (graphique du bas, Fig. 8).
La sélection d’une position sur le profil permet d’afficher la série chronologique asso-
ciée et inversement, la sélection d’une date sur la série chronologique génère son
profil spatial. Les données de températures brutes sont affichées en bleu.
3.2. MAÎTRISE DE L’EXPLOITATION DES DONNÉES
En surveillance opérationnelle, l’objectif prioritaire d’EDF est de détecter au

plus tôt les fuites importantes ou rapides, c’est-à-dire susceptibles de menacer
140
Q.106 - R.12
Fig. 8
Visualisation des données brutes de température dans Fox’Hy
Viewing raw temperature data in Fox’Hy
Fig. 9
Submersion d’un pied de digue lors des crues du Rhône de 2018
Submersion of a dike foot during the Rhône floods of 2018
l’ouvrage. Pour y parvenir, EDF DTG a défini une méthodologie d’analyse des
données simple et structurée.
3.2.1. Détection et localisation des anomalies thermiques avec les mesures

brutes
L’évolution des profils de température est examinée périodiquement, en par-

ticulier aux positions ayant fait l’objet de détections antérieures grâce au modèle
AJOUT (cf. §3.2.3). Sachant que l’incertitude de mesure globale d’un appareil atteint
un maximum de 1◦ C dans les conditions les plus défavorables; si la position con-
sidérée est peu sensible aux influences externes, une anomalie thermique majeure
peut directement être mise en évidence sans avoir recours aux méthodes d’analyse.
141
Q.106 - R.12
Fig. 10
Analyse AJOUT sur une digue du Rhône suite au séisme du Teil (M = 5.2)
AJOUT analysis on a Rhône embankment following the Teil earthquake (M = 5.2).
Cas d’usage n◦ 1: lors de la crue du Rhône en janvier 2018, un pied de digue

a été submergé par l’inondation naturelle en aval, ce qui a généré une augmenta-
tion rapide et spatialement étendue (350 m) des températures au sein de la digue.
Des mesures de températures de qualité peuvent présenter un intérêt en cas de
changement rapide de comportement thermique.
3.2.2. Détection et localisation des fuites avec le modèle AJOUT
EDF a fait le choix d’industrialiser en priorité la méthode d’analyse immédi-

ate AJOUT pour faciliter la détection et la localisation au plus tôt des anomalies
thermiques. Le logiciel Fox’Hy permet de représenter des indicateurs de fuite
représentés en distance et en temps comme pour les mesures de température.
Les points sont affichés en violet (Fig. 10). Un paramétrage par défaut pour cha-
cun des sites permet de restreindre la configuration des traitements à la période
d’étude. La définition d’un seuil s’effectue en sélectionnant un traitement AJOUT de
référence et en précisant un niveau de vigilance en écarts-type. Ce seuil s’applique
uniquement sur le profil spatial des dissimilarités et s’affiche en noir.
L’exploitation des résultats du modèle AJOUT se fait en deux temps. Les

dépassements de seuils sur le profil spatial de la zone d’auscultation sont d’abord
consignés. On analyse ensuite l’évolution temporelle de l’indicateur aux points
considérés. De manière simplifiée, les dépassements issus d’évolutions à allure
saisonnière sont rapidement écartés car souvent reliés à l’hétérogénéité du sol. À
contrario, les évolutions rapides isolées dans le temps, les tendances à la hausse,
les apparitions récurrentes en lien avec la saison ou la cote sont à investiguer,
surtout si la position considérée n’est ordinairement pas sensible à la pluie. Deux
142
Q.106 - R.12
Fig. 11
Validation d’une détection sur le Rhône par le modèle DIGUOMATIC
Validation of a detection on the Rhône river with the DIGUOMATIC model
cas d’usage récents illustrent la valeur ajoutée de cette méthodologie dans un con-
texte d’utilisation prometteur: lorsqu’il s’agit d’attester de l’intégrité des ouvrages en
situation post-séisme ou post-crue.
Cas d’usage n◦ 2 et 3: sollicitation d’un Centre National de Production

d’Électricité EDF suite au séisme du Teil. Le 11 novembre 2019, le Centre Régional
d’Auscultation de DTG (EDF HYDRO) et son partenaire geophyConsult ont été con-
tactés pour porter un avis au plus tôt sur l’éventuelle présence de fuites sur une digue
du Rhône (2 km). Après s’être assuré que les seuils de surveillance n’étaient pas
significativement dépassés (cf. Fig 10), seules les évolutions temporelles AJOUT
proches du seuil spatial ont été analysées. Les séries ne montraient aucune évolu-
tion. Un avis a été rendu dans la journée du 12 novembre. Au regard des enjeux de
sûreté, cet avis a tout de même été consolidé dans les jours qui ont suivi par une
analyse de niveau 3 (cf. §2.2.6). Une démarche similaire de vérification d’absence
de fuite a été mise en œuvre pour EDF HYDRO Est, unité de production hydraulique,
le 19 décembre 2019 suite à une variation de débit sur le Rhin.
3.2.3. Approfondissement des analyses avec des modèles dédiés
En cas de fuite détectée avec le modèle AJOUT, des modèles plus spécifiques
sont mis en œuvre dans le cadre de la surveillance de niveau 3 pour confirmer les
détections de fuites (cf. cas d’usage n◦ 4, Fig. 11) et parfois estimer leurs vitesses. Ce
travail est systématiquement réalisé dans le cadre d’études annuelles présentées
dans une forme comparable aux rapports d’auscultation conventionnelle, le champ
des anomalies thermiques analysées étant élargi aux fuites lentes et distantes.
Cas d’usage n◦ 4: sur une digue du Rhône, EDF a fait appel à geophy Consult
pour confirmer une détection AJOUT au PK 30.450 (valeur modifiée pour raisons
de confidentialité, Fig. 11). Le modèle DIGUOMATIC révèle une dégradation du
modèle air au profit du modèle eau. La vitesse de fuite estimée est de 1.0e-3 m/s.
143
Q.106 - R.12
3.3. DIAGNOSTIC DU COMPORTEMENT
Toujours dans le cadre de la surveillance de niveau 3, il s’agit ensuite de réc-

oncilier les résultats d’analyses issues des capteurs à fibre optique avec l’ensemble
des informations de la surveillance conventionnelle.
3.3.1. Croisement de l’information et diagnostic au moyen d’un schéma

itinéraire
Les éléments de site, les résultats d’examens visuels, les résultats

d’auscultation conventionnelle et de surveillance par capteurs à fibre optique sont
regroupés au sein d’un “schéma itinéraire”, ou tableau synthétique avec des codes
couleur. La dernière ligne de ce tableau est intitulée “diagnostic”. Si une fuite est
avérée, une vitesse de fuite estimée est notée. Si davantage d’historique est requis
ou qu’une source d’information donne une information discordante, on précise que
cette position doit faire l’objet d’une vigilance particulière.
3.3.2. Valorisation des résultats issus de cette surveillance
Les systèmes de mesure fibre optique permettent d’orienter les actions de

surveillance et d’affiner le diagnostic du comportement réalisé avec les moyens
conventionnels. Une réflexion reste à mener sur le meilleur formalisme de présen-
tation des résultats. Si le détail des analyses présentées montre un intérêt pour
la traçabilité et le suivi long terme, les retours d’expérience récents indiquent que
les utilisateurs finaux, majoritairement non spécialistes, sont plutôt en attente de
formats plus synthétiques: quelques graphiques ciblés supportés par les éléments
de méthode clé semblent favoriser l’acceptation de cette technologie.
4. APPLICATION DES MÉTHODES ACTIVES À LA SURVEILLANCE DE

BARRAGES D’ALTITUDE ETANCHÉS PAR GÉOMEMBRANES
4.1. CONTEXTE ET ENJEUX
Les retenues collinaires ou d’altitude sont des ouvrages principalement dédiés

à la production de neige de culture (objet principal des retenues d’altitude) et/ou
destinés à l’irrigation (objet principal des retenues collinaires). D’autres applications
sont possibles comme la production d’énergie hydroélectrique, l’alimentation en eau
144
Q.106 - R.12
potable, ainsi que l’accès au plan d’eau pour la pratique de la pêche de loisir. La
capacité de stockage des retenues d’altitude est variable, pouvant aller de capacités
modestes (inférieures à 20 000 m3 ) à des capacités significatives pour des ouvrages
d’altitude (supérieures à 150 000 m3 ). Le projet RISBA (2013–2015), “risques des
barrages d’altitude”, a mis en évidence que la majorité des retenues récentes ou
en projet ont une capacité moyenne d’environ 100 000 m3 [13]. En France, la hau-
teur des remblais (soit la différence entre le couronnement et le terrain naturel)
des retenues d’altitude se situe entre quelques mètres jusqu’à un maximum de
19,95 m, ce qui, en regard de leur capacité de stockage, correspond théoriquement
aux classes administratives C et D. En Italie, en vallée d’Aoste, ces même types
d’ouvrages présentent quasi systématiquement des hauteurs de remblai comprises
entre 10 m et 15 m. Même si la taille des retenues d’altitude est modeste au regard
des grands barrages, compte tenu des conditions météorologiques sévères en péri-
ode hivernale à une altitude supérieure à 1800 m [14] et de leur impact potentiel
en cas de rupture, ces ouvrages sont fréquemment surclassés administrativement
(passage de la classe C théorique à la classe B).
Concernant la conception des retenues d’altitude, à l’exception des plus

anciennes, la majorité de ces ouvrages disposent d’un dispositif d’étanchéité par
géomembrane (DEG). Outre la géomembrane qui assure la fonction d’étanchéité,
ce type de dispositif comporte plusieurs couches assurant chacune une ou des
fonctions comme la structure support (sur laquelle repose la géomembrane et
ayant des fonctions de protection et de drainage de la géomembrane) et la struc-
ture de protection, sus-jacente à la géomembrane, qui a un rôle de protection de
la géomembrane vis-à-vis de sollicitations mécaniques extérieures. Un descriptif
complet du DEG peut être trouvé dans [14]. La géomembrane qui assure seule la
fonction d’étanchéité, sauf dans les cas exceptionnels de double étanchéité, peut
présenter des défauts d’étanchéité, dont l’origine peut se trouver lors de la mise
en œuvre (endommagement mécaniques accidentels, défauts de soudures, etc.)
ou en exploitation (endommagements mécaniques par poinçonnement des granu-
lats sous ou sus-jacents à la géomembrane, efforts mécaniques de la glace, etc.).
De façon générale, la présence d’une fuite dans les ouvrages décrits ci-dessus
est préjudiciable et, dans des cas extrêmes, peut nuire gravement à la stabilité de
l’ouvrage. Ainsi, pour garantir la stabilité et assurer la durabilité de l’ouvrage sur le
long terme, toute fuite doit être réparée.
Aujourd’hui, la détection d’une fuite au travers de la géomembrane peut être

visuellement observée à l’exutoire du système de drainage. En revanche, la localisa-
tion de la (ou des) fuite(s) en vue d’une réparation est une opération difficile et parfois
longue, délicate et couteuse. En effet, dans le cas d’une géomembrane exposée, il
est nécessaire que la retenue soit complètement vide, ce qui exclut toute opération
d’exploitation, pour qu’une inspection visuelle puisse être menée sur l’ensemble de
la surface de la retenue. Cette inspection visuelle reste cependant une méthode
relativement peu efficace notamment pour localiser des défauts de soudures. Dans
le cas d’une géomembrane recouverte, généralement par une couche granulaire
pluridécimétrique, la présence d’une fuite, pouvant se produire lors du premier rem-
plissage, est dramatique car il est alors nécessaire de retirer toute la couche de
145
Q.106 - R.12
granulats pour accéder à la géomembrane et procéder à une inspection visuelle.

Il est alors évident que cette opération est longue et potentiellement préjudiciable
pour la géomembrane car l’engin qui va retirer la couche de recouvrement granulaire
peut lui-même endommager à nouveau la géomembrane.
Dans ce contexte, une action spécifique sur le développement d’une méthode

simple et robuste de localisation de fuite a été menée dans le cadre de l’action
W.P.3.4.1 du projet de recherche Européen Interreg Alcotra dénommé RESBA,
“ résilience des barrages d’altitude”, démarré en 2017 et regroupant des parte-
naires français (INRAE et l’Université de Savoie Mont Blanc) et italiens (Région
Piémont et Vallée d’Aoste, Ecole Polytechnique de Turin). Cette action a consisté
à tester la méthode active par fibre optique, dans une configuration optimale pour
ce type d’ouvrage. Bien que la méthode soit déjà opérationnelle, il n’existe actuelle-
ment pas de retenues d’altitude équipées d’un tel dispositif. Ce constat peut être
expliqué par plusieurs raisons possibles: méconnaissance de la technique par les
acteurs des retenues d’altitude, absence de démonstrateur à l’échelle du site, apri-
ori négatif concernant la difficulté de transport et le coût lié au générateur de chauffe
à emporter en contexte montagneux, etc.
Pour démontrer la performance de la méthode dans des conditions proches

d’un ouvrage réel, la méthode a été mise en œuvre par l’intermédiaire de geo-
phyConsult sur le bassin expérimental du site INRAE d’Aix-en-Provence. Trois
principales questions ont notamment été investiguées:
– Quelle est la puissance (de chauffe) minimale à fournir pour une détection
optimale de fuite?
– Quel est le débit de fuite minimal détecté pour une énergie de chauffe mini-
male ? Cette question est notamment liée d’une part au critère de détection,
c’est-à-dire la différence de température qui ne peut être expliquée que par la
présence d’une anomalie (soit une fuite), et, d’autre part, à la distance de la
fuite par rapport au positionnement de la fibre optique;
– Le système est-il aussi performant en période chaude (été) qu’en période
froide (hiver)?
4.2. DESCRIPTION DU BASSIN EXPÉRIMENTAL D’INRAE
Situé sur le site d’INRAE à proximité d’Aix-en-Provence, le bassin expérimen-

tal (Fig. 12) présente une surface de 200 m2 , pour un volume d’eau retenu de 210 m3 .
Le bassin s’élève en remblai, d’une hauteur de 2,5 m, avec des pentes de talus de
2H/1V sur les parements internes et 3H/2V sur les parements externes. Le remblai
a été réalisé avec les matériaux limono-argileux du site, présentant un Indice de
Plasticité Ip = 17 et une perméabilité voisine de 10−9 m.s−1 dans les conditions de
mise en place (mesure en laboratoire).
146
Q.106 - R.12
Fig. 12
Vues du bassin expérimental d’INRAE (crédit photo INRAE)
Picture of INRAE’s experimental (photo credit INRAE)
Le dispositif d’étanchéité est composé d’un géocomposite de drainage à mini-

drains (Draintube® , 20 mm de diamètre de mini-drains, espacés de 2 m sur le
produit), placé directement sur le remblai (Fig. 13). Les mini-drains sont placés dans
le sens de la pente et connectés à un collecteur central (drain granulaire) qui évacue
les eaux recueillies par un conduit traversant le remblai nord. Sur le géocomposite,
une géomembrane en polypropylène d’épaisseur 1 mm assure l’étanchéité.
Quatorze fuites artificiellement créées traversent la géomembrane, avec un

exutoire sur le géocomposite drainant (Fig. 13). Ces quatorze fuites sont celles qui
ont été activées dans l’expérimentation décrite. D’autres fuites traversent par ailleurs
le remblai, l’exutoire étant alors en aval du remblai – elles ne font pas l’objet de la
communication.
Les dix fuites modélisant un endommagement de la géomembrane ont été

réparties de façon à étudier l’influence de l’exposition du talus (effet de la tempéra-
ture) et de leur position respective par rapport aux mini-drains collectant les eaux
de fuite. Chaque fuite est équipée d’un chapeau d’injection relié à l’alimentation en
eau par des tuyaux d’arrosage.
Une partie de l’instrumentation consiste en une fibre optique associée à un

câble en cuivre, l’association étant appelée fibre active, placée au contact du remblai
en pied de talus interne. Des capteurs complémentaires permettent de valider les
acquisitions des fibres optiques: détecteurs d’humidité et sondes de teneur en eau
placées en aval des points d’injection. Des sondes de température sont réparties
147
Q.106 - R.12
Fig. 13
Géocomposite de drainage à mini-drains (Fig. à gauche) et fuites artificielles
traversant la géomembrane, réparties sur le parement intérieur du bassin: vue
pendant le remplissage du bassin (Fig. à droite)
Drainage geocomposite with mini-drains (Fig. on the left); artificial leaks passing
through geomembrane, on the inner part of the basin: picture during filling the
basin (Fig. on the right)
de façon à mesurer la température de l’eau retenue, et la température de l’eau au

niveau des points d’injection.
L’eau du bassin alimente directement les fuites, par un moyen de pompage,

de façon à assurer une homogénéité de température entre l’eau injectée et l’eau
retenue dans le bassin. Un dispositif de régulation et de mesure de débit et pression
d’injection est placé en amont des fuites.
4.3. MATÉRIELS, MÉTHODES ET PROGRAMME EXPÉRIMENTAL
Lors des campagnes de mesures, l’interrogateur Sensornet Oryx de mesure

de température le long de la fibre optique a été utilisé. Il a été fixé une résolution
spatiale de 1 m ainsi qu’une fréquence d’acquisition de 5 minutes. Lors de la seconde
campagne, cet interrogateur a été gracieusement prêté par EDF par l’intermédiaire
de geophyConsult. La puissance électrique injectée au cours des opérations de
chauffe (tension et intensité) a été enregistrée au moyen d’un wattmètre enregistreur
(Voltraft Energy Logger 4000F), placé en amont du coffret de chauffe.
La détection d’une fuite par la fibre optique correspond à un calcul de la dif-

férence de température mesurée au voisinage de la fibre optique en l’absence de
fuite avec celle mesurée en présence de la fuite. Ce paramètre est noté dT. Le
critère de détection d’une fuite, correspondant à une anomalie thermique mesurée,
a été fixé à 0,3◦ C. Dans un objectif de quantification, il est calculé l’intensité de
détection (◦ C.m) qui correspond à l’aire sous la courbe dT le long du linéaire de
fibre optique.
Le bassin expérimental a fait l’objet de deux campagnes expérimentales de

5 jours de mesures: la première menée du 11 au 21 octobre 2019 et la seconde
148
Q.106 - R.12
du 10 au 14 février 2020. L’objectif de ces deux campagnes était d’une part de

tester la performance du dispositif pendant deux périodes où les températures
extérieures sont différentes et, d’autre part, tester la robustesse du dispositif dans
le temps, à l’échelle d’un projet de recherche. Lors des campagnes d’injection, il
a été testé cinq débits d’injection (0,005–0,01–0,02–0,05–0,1 l/s), 5 puissances de
chauffe (2–4–6–8–10 W/ml) et trois distances point d’injection – câble de chauffe
(0–1,6–3,0 m). Chacune des puissances de chauffe a fait l’objet d’une mesure de
référence c’est-à-dire une mesure d’élévation de température au voisinage de la
fibre optique et du câble de chauffe, en l’absence de fuite.
4.4. PRINCIPAUX RÉSULTATS ET RECOMMANDATIONS
Il a été montré que l’ensemble des fuites présentant un débit supérieur ou égal
à 0,05 l/s et placées en amont de la fibre optique en base du remblai ont toutes été
détectées. Avec le critère de détection de 0,3◦ C, et avec une puissance de chauffe
supérieure ou égale à 4 W/m, aucun faux positif n’a été relevé.
Il n’a pas été observé d’impact significatif de la puissance de chauffe sur la

détectabilité des fuites. En d’autres termes, une fuite détectée avec une puissance
de 8 W/m l’a également été avec une puissance de 2 W/m. Cependant, l’utilisation
d’une faible puissance a induit des faux positifs provoqués par le trop faible contraste
de température induit par la chauffe. En conséquence, une puissance minimale de
4 W/m est recommandée.
Il a été observé des intensités de détections croissantes avec les débits injec-
tés. A débit équivalent, les intensités de détection diminuent si la distance entre la
fuite et le câble de mesure augmente, en cohérence avec le fait que la vitesse de
fuite diminue le long de la propagation de celle-ci (diffusion dans le géocomposite
de drainage). Ainsi, sur site réel, il est recommandé d’installer des fibres optiques
à plusieurs côtes du bassin, avec un espacement qui dépendra du niveau souhaité
de détectabilité de fuite.
5. CONCLUSION
De manière générale, pour les ouvrages contenant de l’eau, la détection et la

location d’une fuite est une donnée essentielle pour les gestionnaires afin de garan-
tir la stabilité des ouvrages et d’assurer leur durabilité. En adressant ce besoin
de surveillance de façon performante, la technologie de mesure fibre optique se
démocratise à présent et s’étend à de nouveaux usages avec une maturité indus-
trielle des matériels, des méthodes et des coûts d’installation plus accessibles, de
l’ordre de 30 k/km. Pour l’exploiter pleinement, il est nécessaire de définir en amont
149
Q.106 - R.12
la méthode d’exploitation la plus appropriée à la configuration d’ouvrage et à ses

enjeux.
Dans l’objectif de détecter et localiser au plus tôt les fuites les plus dan-
gereuses pour ses remblais, les services d’auscultation d’EDF ont intégré un
nouveau processus de surveillance opérationnelle reposant sur trois notions clé:
utilisation de données de qualité, industrialisation basée sur Fox’Hy, application
industrielle permettant l’exploitation d’AJOUT, méthode passive performante en
détection et ne nécessitant ni entrées externes, ni historique long, et enfin croise-
ment des résultats avec d’autres modèles pour validation ou quantification et
réconciliation avec les données de surveillance conventionnelle.
Pour les retenues d’altitude où les fuites sont à la fois préjudiciables pour
l’exploitation de l’ouvrage mais également dangereuses vis-à-vis des risques
d’instabilités, la méthode active a été appliquée de façon concluante dans un
démonstrateur INRAE représentatif de la configuration des retenues d’altitude. En
particulier, il a été montré que la technique était capable de localiser sans faux posi-
tifs ni négatifs une fuite d’un débit supérieur ou égal à 0,05 l/s et placées en amont
de la fibre optique. Cependant, le niveau de détectabilité de fuite dépendant de la
distance entre la fuite et la fibre optique, le dimensionnement du réseau de fibres
optiques à positionner dans l’ouvrage sera une étape clé vis-à-vis de la performance
du dispositif.
REMERCIEMENTS
L’expérimentation réalisée sur le bassin d’INRAE a été soutenue par le

projet RESBA financé par l’Europe dans le cadre du programme INTERREG
ALCOTRA.
REFERENCES
[1] KAPPELMEYER O., The use of near surface temperature measurements for
discovering anomalies due to causes at depths, Geophysical Prospecting,
5(3):239–258, 1957.
[2] JOHANSSON S., Seepage monitoring in embankment dams. Thèse de

doctorat, KTH, Stockholm, 1997.
[3] RADZICKI K., BONELLI S., Impulse response function analysis of pore
pressure in earthdams. 18ème Congrès Français de Mécanique, Grenoble,
France. 6 p., Aug 2007.
150
Q.106 - R.12
[4] CUNAT P., Détection et évaluation des fuites à travers les ouvrages
hydrauliques en remblai, par analyse des températures réparties, mesurées
par fibre optique, Thèse de doctorat, LTHE, Grenoble, 2012.
[5] KERZALE S., Estimation des débits de fuites dans les digues en terre par
thermométrie. Thèse de doctorat, université de Lorraine. 2013.
[6] GUIDOUX C., Développement et validation d’un système de détection et de

localisation par fibre optique de zones de fuite dans les digues en terre. Thèse
de doctorat, université Joseph Fourier de Grenoble. 2006.
[7] COURIVAUD J.-R., LORRAIN N., MARTINOT F. BEGUIN R., Quantifying

seepage flow velocities in embankment dams from optical fibre distributed tem-
perature measurements. Long-Term Behaviour and Environmentally Friendly
Rehabilitation Technologies of Dams. 2017.
[8] GARANDET A., BROTHIER C. MARTINOT F., Techniques Nouvelles

d’Auscultation et de Surveillance des Digues de Grand Linéaire, ICOLD
Vienne, 2018.
[9] KHAN A.A., VRABIE V., MARS J., GIRARD A., D’URSO G., A source sepa-
ration technique for processing of thermometric data from Fiber-Optic DTS
measurements for water leakage identification in dikes. Sensors Journal,
IEEE, 8(7):1118–1129. 2008.
[10] DORNSTADTER J., HEINEMANN B., In situ Detection of Internal Erosion, 8th
ICOLD European Club Dam Symposium, Innsbruck, 481–485, 2010.
[11] PERZLMAIER S., Verteilte Filtergeschwindigkeitsmessung in Staudämmen.

Thèse de doctorat, TU München, Munich, 2007.
[12] FAILLEAU G. et al., Development of facilities and methods for the metrological
characterization of distributed temperature sensing systems based on optical
fibres, Meas. Sci. Technol. 28 (2017) 015009 (12pp).
[13] BOUTRY M. RISBA, RAPPORT FINAL, Action 2.1, 65 p, 2014.
[14] PEYRAS L. ET MERIAUX P., Retenues d’altitude. Quae. 352 p, 2009.
151
Q.106 - R.12
1. INTRODUCTION
The technology of distributed temperature measurements by optical fiber has

found numerous applications for several decades in fields as diverse as fire safety,
high voltage lines integrity controls, study of hydraulic exchanges between ground-
water and river, etc. In the field of monitoring of hydraulic structures, this technology
has been successfully used for twenty years to detect and locate leaks that could
initiate internal erosion phenomena.
The advantage of distributed temperature measurements for these applica-

tions lies in the ability of systems using this technology to record – at any point of
the instrumented line – precise temperatures (0.1◦ C) over long distances (30 km),
with a metric spatial resolution and a 10-min acquisition time. Now, it is a well-
known fact that localized flows generally disturb the local temperature field, so that
the presence of a thermal anomaly makes it possible to infer the presence of a flow
(Fig. 1).
Composed with a measurement cable very similar to those used in telecom-

munications and a compact optical device (Fig. 2), the fiber optic distributed
Fig. 1
Illustration of the effect of a leak on the temperature field within an embankment
structure
Fig. 2
Simplified fiber optic measuring chain
152
Q.106 - R.12
measurements which appeared in the 1990s have constantly improved to reach

the specifications announced above.
Compared to the classic thermometry used to carry out often punctual and
localized investigations, the spatial coverage, the speed of the measurements and
the localization capacity of the optical fiber sensors have quickly opened the way to
new applications: remove more quickly reservations on new structures, target areas
of weakness on old structures to carry out localized reinforcements, give operators
means of supervising the behavior of structures during floods and diagnostic means
complementary to conventional auscultation in a post-event context such as the
earthquakes.
Since temperature measurements only provide indirect access to the desired

flow phenomena, fiber optic data analysis methods play a decisive role in the effi-
ciency of these monitoring systems. A typology of these methods – accompanied by
recommendations on their choice and their implementation – is proposed in chapter
2 of the paper. In chapters 3 and 4, two application examples provided by EDF and
INRAE with the contribution of geophyConsult are then proposed in order to under-
line the need for the methods to be suited to the configurations of structures and the
monitoring goals.
2. TYPOLOGY OF LEAKAGES DETECTION METHODS
From the first leak detection experiments using fiber optic temperature mea-
surements carried out from 1995, it appeared that the visualization of raw measure-
ments only usually did not allow to identify leaks correctly. Indeed, many factors
other than leaks influence the temperature of a soil: the geological nature of the soil,
the relative position of the fiber with respect to the air or the upstream reservoir, or
even the proximity of crossing structures. The development of methods for analyzing
temperature data by optical fiber therefore quickly turned out to be a major challenge
in making this technology operational.
2.1. PASSIVE METHODS
Passive methods refer to signal processing applicable to soil temperature mea-

surements different from those of the water retention, potentially influenced by air
and groundwater (see Fig. 3). These measurements are obtained using simple opti-
cal cables; that is, not having copper wires so that they can be heated to create
a temporary thermal contrast. Several passive methods have been successively
developed in order to match various uses, from the automated detection of thermal
anomalies in real-time monitoring with short chronicle of measurements to the fine
153
Q.106 - R.12
Fig. 3
External influences on the measurements used by passive methods
estimation of leakage velocity thanks to elaborate models and longer measurement

chronicle.
2.1.1. Long-term monitoring, saturated soil
The first leakage detection methods by optical fibers sensors were developed
by Sam JOHANSSON – HYDRORESEARCH AB (1997, [1]). They are based on the
analysis of annual variations in temperature signals measured in the reservoir water
and in saturated soil. The delay time approach consists in computing the diffusion
speed of the thermal wave from the delay time and the thickness of the structure.
The leakage speed can then be determined using Darcy’s law taking into account
the heat capacities of water and saturated soil. The amplitude approach is based on
a simplified modeling of the thermo-hydraulic exchanges inside the leakage zone.
It provides a relationship between a normalized temperature amplitude variation
and a dimensionless leak rate. Relatively accessible, these methods require an
external temperature measurement of the canal water and a sufficient measurement
chronicle. They must also not be influenced by the air temperature, which may be the
case when the optical fiber is installed near the surface and the water table is deep.
2.1.2. Long-term monitoring, saturated or unsaturated soil
The IRFA (INRAE) [2] and MORITO (EDF HYDRO) [3] models make it possible
to predict the temperatures measured by the optical fiber sensors in the hypothesis
where the heat transfers within the structure are only governed by the phenomenon
of heat conduction without advection. Knowing the boundary conditions of air and
water temperature, attenuation and phase shift parameters of parametric impulse
response functions are calculated to minimize the difference between measured
and modeled temperatures. To be properly used, these methods require a priori
154
Q.106 - R.12
knowledge of possible attenuations and phase shifts with regard to soil parameters,
external water temperature measurements and a sufficient measurement chronicle.
The DIGUOMATIC method (partnership between EDF HYDRO [4] and geo-
phyConsult [5, 6]) is based on a coupled thermo-hydraulic modeling of the backfill
structure and its foundation. The influences of water and air temperatures are mod-
eled separately, following two distinct directions and taking into account the physical
properties of the materials. The heat transport by advection is modeled for the flow.
Other quantities such as the height of the water table, its temperature or the solar
radiation [5] at the surface can be integrated. The implementation of this method
requires advanced knowledge of the structure and its environment, as well as a crit-
ical look at the modeling choices. This method is particularly effective for criticizing
detections from other models that are easier to use or for specifying leak velocities.
2.1.3. Real-time monitoring, saturated or unsaturated soil
The AJOUT model developed by EDF HYDRO [8] is mainly used for detection
purposes on a short data history, typically one day (Analyse JOUrnalière de Tem-
pérature in French, Daily Temperature Analysis). It is based on a signal processing
method carried out in two stages: filtering of undesirable influences from a decom-
position into singular values then construction of a daily thermal anomaly indicator
from the sums of dissimilarities between the residues resulting from each position
within the auscultated areas. The detection of thermal anomalies is carried out in
two steps: in the absence of a measurement history, high values of dissimilarity
index AJOUT are pointed out. Then the evolution of the time series are observed by
comparing them to historical thresholds.
The calculation of the AJOUT model assumes an acquisition of hourly mea-

surements but does not require a significant measurement history: a few days are
enough. It also does not use additional input data such as air temperature or water
temperature in the canal. No thermohydraulic soil modeling parameters are required.
2.2. ACTIVE METHOD
The active method or Heat Pulse Method (HPM) [10] is mainly used in cases
where the optical fiber must be installed in an environment whose temperature is
directly influenced by the water in the reservoir (even with no leakage). It refers to the
realization and use of temperature measurements within an optical cable installed
near the reservoir and previously heated by the Joule effect using electrically
powered copper wires (see Fig. 4).
The needed electrical power – typically 5 to 10 W/m – depends on the con-

ductivity of the soil, the heating time and the aimed temperature increase. In the
155
Q.106 - R.12
Fig. 4
Influence of water temperature and optical cable used for the active method
Fig. 5
HPM principle
presence of a leak in the vicinity of the cable, the convection generated by the flow
will cause a lower rise in the temperature of the cable (see Fig. 5).
The interpretation of the measurements is based on the comparison of the

measured values of dT with a reference value of dTREF . Through calibration tests
in the laboratory or on site with controlled leaks [11], it is possible to quantify the
flow velocity for a range of Darcy’s velocities between 10−5 and 10−3 m/s. The leak
detection radius around the optical cable using this method is often less than for
passive approaches, in the range of 15 to 20 cm. The means required for heating
make this method more difficult to operate on measurement lengths of ten kilometers
or more.
As some technical parameters may vary from one heating to another (the
injected electrical power, the heating time, the evolution of the outside temperatures),
the raw dT profiles are systematically readjusted, by adjusting their average and their
standard deviation to those of a reference measurement. The dT values of a given
heating are then subtracted from the dTREF values recorded at the same positions
during the reference heating. A measurement is thus obtained at any point of the
cable which is strictly speaking the detection parameter of the active method, to
which a detection threshold depending on the field conditions (spatial variability
156
Q.106 - R.12
characteristic of the cable used) is applied. This methodology is called SACRE (as
for Surveillance Active par Centrage et Réduction avec Etat-zéro in French, Active
Monitoring by Centering and Reducing with Reference-State).
3. IMPLEMENTATION OF PASSIVE METHODS WITHIN THE EDF

SURVEILLANCE PROCESS
In 2019, EDF industrialized an operational monitoring process according to

three stages: verification of compliance of input data, reproducible use of a ver-
satile detection model, then behavior diagnosis. This process was applied to three
embankments with a configuration similar to that presented in Figure 3 and for which
a sufficient level of control in design and construction was acquired, i.e. a total instru-
mented line of around 30 km. To implement it, EDF relied on the practical experience
of geophyConsult.
3.1. QUALITY OF INPUT DATA BEFORE PROCESSING
Since the quality of the analyzes is conditioned on that of the input data, EDF
ensures the optimal functioning of its measurement chains, the proper routing of
data to the application and their preparation before processing.
3.1.1. Verification of metrological performance and equipment reliability
Annually updated design guidelines bring all of the business requirements

necessary for the selection of materials and recommendations for implementation.
Metrological controls are carried out by the National Testing Laboratory [12], in
particular on two important criteria for the monitoring of leaks: temporal dispersion
and true resolution at long range. Certificates or proofs of durability are requested
from equipment manufacturers in order to limit subsequent maintenance operations
and therefore interruptions in time series that are detrimental to processing.
3.1.2. Control of devices on sites
The main conformity checks carried out on site, upon receipt and during
periodic device checks (2 years) are as follows:
– metrological verification of the measuring chain. This is to ensure that the opti-
cal fiber does not age abnormally and that the quality of any welds is preserved.
157
Q.106 - R.12
Fig. 6
Operational monitoring of imports in Fox’Hy (software in french)
The main task then is to verify the accuracy of the temperature measure-
ments in the calibration coils using a previously calibrated Pt100 probe, the
repeatability and the absence of drift.
– functional verification. This involves checking the condition of the device and
its environment by visual checks and elementary tests.
3.1.3. Secured data transfers
To control the availability, integrity, confidentiality and traceability of data (secu-

rity criteria for DICT information systems), EDF relies on high-level IT infrastructure
that is also used for its monitoring applications with strong safety issues. Remote
supervision, handling and basic maintenance tools make it possible to quickly detect
transmission interruptions, report them and in some cases quickly resolve them
remotely.
3.1.4. Quality check of imported data
The next step is to verify that the measurement files do not present any com-
puter anomalies and that the measurements are carried out in the nominal range of
the sensor.
An application named Fox’hy (Fig. 6) has been developed to allow the EDF
DTG team in charge of controlling level 2 auscultation measurements – level 1 being
provided by the operator – to supervise capitalization and data exploitation. Taking
into account the quantity of data to be checked – of the order of two million per day
before processing – in a constrained operational context, a "monitoring dashboard"
has been provided in the software which makes it possible to easily visualize, among
other things, data import faults and filtered outliers.
158
Q.106 - R.12
Fig. 7
Operational monitoring of data proofing in Fox’Hy
3.1.5. Verification of spatial registration and temporal synchronization
The Fox’Hy software (Fig. 7) also allows DTG services to control data prepa-
ration before processing. After having been associated and synchronized, the data
coming from several optical fibers are collected; that is to say, they are associated
with a single mileage mark, even if several fibers are implanted at different heights
or transverse positions for the same longitudinal location (PK). This is followed by a
metric-pitch data interpolation operation so that even if the fiber optic optical lenght
changes following a repair or device change, the time series continues to be fed in
a right way. Finally, the data are divided into "auscultation zones" or zones where
the thermal behavior of the dike is presumed to be homogeneous, an important
assumption for the application of the AJOUT analysis model, the use of which will be
detailed below. Computer monitoring of these operations guarantees the traceability
of modifications.
3.1.6. Verification of the physical meaning of the data
Fox’Hy includes an instant data visualization module. It allows to assess the

physical meaning of the data received in the application: no drift or jumps attributable
to a loss of calibration, for example. Two interacting graphical representations are
available: a spatial profile of the temperatures of an auscultation zone at a given
time (top graph, Fig. 8) and a time series of temperatures at one position (bottom
graph, Fig. 8). Selecting a position on the profile displays the associated time series
and conversely, selecting a date on the time series generates its spatial profile. Raw
temperature data is displayed in blue.
3.2. MANAGEMENT OF DATA EXPLOITATION
In operational monitoring, EDF’s priority is to detect large or rapid leaks as

early as possible, that is to say leaks likely to threaten the structure. To achieve this,
EDF DTG has defined a simple and structured data analysis methodology.
159
Q.106 - R.12
Fig. 8
Viewing raw temperature data in Fox’Hy
Fig. 9
Submersion of a dike foot during the Rhône floods of 2018
3.2.1. Detection and localization of thermal anomalies with raw measurements
The evolution of the temperature profiles is examined periodically, in particular

at the positions that have been the subject of previous detections using the AJOUT
model (see §3.2.3). Knowing that the overall measurement uncertainty of a device
reaches a maximum of 1◦ C under the most unfavorable conditions; if the position
considered is not very sensitive to external influences, a major thermal anomaly can
be directly highlighted without needing analytical methods.
Use case 1: during the Rhône flood in January 2018, a foot of the dike was
submerged by natural flooding downstream, which generated a rapid and spatially
extensive increase (350 m) in temperatures at within the dike. Quality temperature
measurements may be of interest when there is a rapid change in thermal behavior.
160
Q.106 - R.12
3.2.2. Detection and localization of leaks with the AJOUT model
EDF has chosen to industrialize as a priority the AJOUT immediate analysis

method to facilitate the early detection and localization of thermal anomalies. The
Fox’Hy software makes it possible to represent leak indicators in distance and time
as for temperature measurements. The points are displayed in purple (Fig. 10). A
default setting for each of the sites makes it possible to restrict the configuration
of treatments to the study period. The definition of a threshold is carried out by
selecting a reference AJOUT treatment and specifying a level of vigilance in standard
deviations. This threshold only applies to the spatial profile of dissimilarities and is
displayed in black.
The results of the AJOUT model are exploited in two stages. The exceedances
of thresholds on the spatial profile of the auscultation zone are first recorded. Oper-
ators then analyze the temporal evolution of the indicator at the points considered.
In a simplified manner, the overruns resulting from changes at a seasonal pace are
quickly ruled out because they are often linked to the heterogeneity of the soil. Con-
versely, rapid isolated changes over time, upward trends, recurring appearances in
connection with the season or the hydraulic load should be investigated, especially
if the position considered is usually not sensitive to rain. Two recent use cases illus-
trate the added value of this methodology in a promising context of use: when it
comes to attesting to the integrity of structures in a post-earthquake or post-flood
situation.
Use case 2 and 3: request of an EDF nuclear plant safety team following the Teil
earthquake. On November 11, 2019, the DTG Regional Auscultation Center (EDF
HYDRO) and its partner geophyConsult were called to advise as soon as possible
on the possible presence of leaks on a Rhône dike (2 km). After making sure that
the monitoring thresholds were not significantly exceeded (see Fig 10), only the
AJOUT temporal changes close to the spatial threshold were analyzed. The series
showed no evolution. An opinion was issued on November 12. With regard to safety
issues, this opinion was nevertheless consolidated in the days that followed by a
level 3 analysis (see §2.2.6). A similar procedure to verify the absence of leaks was
implemented for EDF HYDRO Est, a hydraulic production unit, on December 19,
2019 following a variation in flow on the Rhine river.
3.2.3. Deepening analyzes with dedicated models
In case of a leak detected with the AJOUT model, more specific models are
implemented as part of level 3 monitoring to confirm the leak detections (see use
case 4, Fig. 11) and sometimes estimate their speeds. This work is systematically
carried out within the framework of annual studies presented in a form comparable
to conventional auscultation reports, the field of thermal anomalies analyzed being
extended to slow and distant leaks.
161
Q.106 - R.12
Fig. 10
AJOUT analysis on a Rhône embankment following the Teil earthquake (M = 5.2)
Fig. 11
Validation of a detection on the Rhône river with the DIGUOMATIC model
Use case 4: on a Rhône embankment, EDF called on geophyConsult to con-

firm an AJOUT detection at PK 30.450 (value modified for confidentiality reasons,
Fig. 11). The DIGUOMATIC model reveals a degradation of the air model in favor of
the water model. The estimated leakage velocity is 10−3 m/s.
3.3. BEHAVIOR DIAGNOSIS
Still within the framework of level 3 surveillance, it is then a matter of reconciling

the results of analyzes from the fiber optic sensors with all the information from
conventional surveillance.
162
Q.106 - R.12
3.3.1. Cross-referencing of information and diagnosis using an itinerary

diagram
The site elements, the results of visual examinations, the results of conven-
tional auscultation and surveillance by fiber optic sensors are gathered in a "route
diagram", or synthetic table with color codes. The last row of this table is titled “diag-
nosis”. If a leak is found, an estimated leak rate is noted. If more history is required
or a source of information gives conflicting information, the position is marked and
has to be investigated.
3.3.2. Promotion of the monitoring value-added
Fiber optic measurement systems make it possible to orient monitoring actions

and refine the diagnosis of behavior carried out with conventional means. It is still
necessary to find out the best formalism of presentation for results. While the detail
of the analyzes presented shows an interest in traceability and long-term monitor-
ing, recent feedback indicates that end users, mostly non-specialists, expect more
synthetic formats. A few targeted graphics supported by key method elements seem
to accelerate the acceptance of this technology.
4. USE OF ACTIVE METHOD FOR SURVEILLANCE OF ALTITUDE

RESERVOIRS USING GEOMEMBRANE LINING SYSTEMS
4.1. CONTEXT AND ISSUE
Hill or altitude reservoirs are hydraulic works mainly dedicated to the pro-
duction of artificial snow (main purpose of altitude reservoirs) and/or intended for
irrigation (main purpose of hill reservoirs). Other applications are possible such as
the production of hydroelectric power, the supply of drinking water, as well as for
the practice of leisure fishing. The storage capacity of altitude reservoirs is variable,
ranging from modest capacities (less than 20,000 m3 ) to significant capacities for alti-
tude structures (greater than 150,000 m3 ). The RISBA project (2013–2015), “Risks
of high altitude dams”, has shown that the majority of recent or planned reservoirs
have an average capacity of around 100,000 m3 [13]. In France, the height of the
embankments (i.e. the difference between the crest and the natural terrain) of the
altitude reservoirs is between a few meters up to a maximum of 19.95 m, which, com-
pared to their storage capacity, theoretically corresponds to administrative classes
C and D. In Italy, in the Aosta Valley, almost all these types of structures have backfill
heights comprised between 10 m and 15 m. Even if the size of the altitude reservoirs
is modest compared to large dams, given the severe meteorological conditions in
163
Q.106 - R.12
winter at an altitude above 1800 m [14] and their potential impact in the event of
failure, these structures are frequently administratively upgraded (transition from
theoretical class C to class B).
Today, detection of a leak through the geomembrane can be visually observed

at the outlet of the drainage system. On the other hand, locating the leak for repair
is difficult and sometimes long, delicate and expensive operation. Indeed, in the
case of an exposed geomembrane, it is necessary for the reservoir to be completely
empty, which excludes any use of the reservoir, so that a visual inspection can be
carried out on the entire surface of the reservoir. However, this visual inspection
remains a relatively inefficient method, in particular for locating weld defects. In the
case of a covered geomembrane, generally by a multi-decimetric granular layer, the
presence of a leak, which may occur during the first filling, is dramatic because it is
necessary to remove the entire layer of aggregates to access the geomembrane and
perform the visual inspection and reparation. It is then obvious that this operation is
long and potentially detrimental for the geomembrane because the machine which
will remove the granular cover layer can itself damage the geomembrane again.
Regarding the design of the altitude reservoirs, with the exception of the oldest
ones, the majority of these structures have a geomembrane lining system (GLS). In
addition to the geomembrane which ensures the waterproofing function, this type of
device comprises several layers each providing one or more functions such as the
support structure (on which the geomembrane is set and having functions of protec-
tion and drainage of the geomembrane) and the structure protection, overlying the
geomembrane, which has a role of protecting the geomembrane against external
mechanical stresses. A full description of GLS can be found in [14]. The geomem-
brane, which performs alone the waterproofing function, except in exceptional cases
of double waterproofing, may exhibit waterproofing defects, of which the origin may
be found during installation (accidental mechanical damage, weld defects, etc.) or in
operation (mechanical damage by puncturing of the aggregates under or overlying
the geomembrane, mechanical forces of the ice, etc.). In general, the presence of
a leak in the structures described above is detrimental and, in extreme cases, can
seriously affect the stability of the structure. Thus, to ensure the stability and ensure
the durability of the structure over the long term, any leak must be repaired.
In this context, a specific action on the development of a simple and robust

method for locating leaks was carried out within the framework of action WP3.4.1 of
the European research project Interreg Alcotra called RESBA, “resilience of dams d
‘altitude’, started in 2017 and bringing together French partners (INRAE and the Uni-
versity of Savoie Mont Blanc) and Italian partners (Piedmont and Aosta Valley region,
Polytechnic School of Turin). This action consisted of testing the active method by
optical fiber, in an optimal configuration for this type of structure. Although the method
is already operational, there are currently no altitude reservoirs equipped with such
a device. This observation can be explained by several possible reasons: lack of
knowledge of the technique by the administrator of altitude reservoirs, absence of a
demonstrator at the site level, negative prejudice concerning the difficulty of transport
and the cost linked to the heating generator to take away in mountainous context, etc.
164
Q.106 - R.12
Fig. 12
Picture of INRAE’s experimental (photo credit INRAE)
To demonstrate the performance of the method under conditions close to a real

structure, the method was implemented through geophyConsult on the experimental
basin of the INRAE site in Aix-en-Provence. Three main questions were particularly
investigated:
– What is the minimum (heating) power to be supplied for optimum leak

detection?
– What is the minimum leakage rate detected for minimum heating energy?
This question is particularly linked on the one hand to the detection criterion,
i.e. the temperature difference which can only be explained by the presence
of an anomaly (i.e. a leak), and, on the other hand, at the distance of the leak
relatively to the positioning of the optical fiber;
– Is the system as efficient in hot period (summer) as in cold period (winter)?
4.2. INRAE’S EXPERIMENTAL BASIN DESCRIPTION
INRAE’s experimental basin is located in the vicinity of Aix-en-Provence. The

surface area of the basin is 200 m2 , while the volume of water retained is 210 m3
(Fig. 12). The basin is built as a 2.5 m high embankment; slopes are 2 horizontal/1
vertical for the inside part, and 3 horizontal/2 vertical on the outside of the basin.
The embankment is made of a local silt and clays material with a Plasticity Index of
17 and a hydraulic conductivity of 10−9 m.s−1 (laboratory measurement).
A device of geomembrane waterproofness is placed on the inner part of the

basin. One part of the device is a Draintube®drainage geocomposite made up of
165
Q.106 - R.12
Fig. 13
Drainage geocomposite with mini-drains (Fig. on the left); artificial leaks passing
through geomembrane, on the inner part of the basin: picture during filling the
basin (Fig. on the right)
mini-drains with a diameter of 20 mm spaced 2 m apart (Fig. 13). The drainage

geocomposite is installed directly on the soil. Mini-drains follow the slope and are
connected to a central drain on the basin, filled with gravels. This drain allows water
evacuation thanks to a pipe crossing the embankment. The second part of the
waterproofness device is a 1mm thick polypropylene geomembrane placed above
the geocomposite.
Fourteen artificial leaks cross the geomembrane, outlet is the drainage geo-
composite (Fig. 13). These leaks were used during the experiment described. Other
leaks pass through the embankment; they are not the topic of this paper.
Leaks reproducing geomembrane damage are localized in order to study the

influence of exposure to the sun (temperature effect) and the influence of their posi-
tion compared to that of mini-drains. Each leak is connected to the water supply with
a garden hose.
Part of the monitoring consists of an optical fibre associated with a copper

cable, together called active fibre, placed in contact with soil at the embankment
toe. More sensors are used to confirm fibre optic measurements: humidity sensors
and water content located downstream of the leaks. Temperature sensors measure
the temperature of the water, both in the basin and at the leaks.
Water from the basin supplies directly the leaks, using pumps, so that the
temperatures are close to each other. Flow rate and water pressure are controlled
through a valve upstream of the leaks, and recorded during the experiments.
4.3. MATERIALS, METHODS AND EXPERIMENTAL PROGRAM
During the measurement campaigns, the Sensornet Oryx interrogator for mea-
suring temperature along the optical fiber was used. A spatial resolution of 1 m
has been set as well as an acquisition frequency of 5 minutes. During the second
campaign, this interrogator was kindly loaned by EDF through geophyConsult. The
166
Q.106 - R.12
electrical power injected during the heating operations (voltage and intensity) was
recorded by means of a recording power meter (Voltraft Energy Logger 4000F),
placed upstream of the heating device.
The detection of a leak by the optical fiber corresponds to a calculation of the

temperature difference measured in the vicinity of the optical fiber in the absence of a
leak with that measured in the presence of the leak. This parameter is denoted dT.
The criterion for detecting a leak, corresponding to a measured thermal anomaly,
was set at 0.3◦ C. For the purpose of quantification, the detection intensity (◦ C.m)
is calculated, which corresponds to the area under the curve dT along the linear
optic fiber.
The experimental basin was subjected to two experimental campaigns of 5

days of measurements: the first was carried out from October 11 to 21, 2019 and
the second from February 10 to 14, 2020. The objective of these two campaigns was
on the one hand to test the performance of the device during two periods when the
outside temperatures are different and, on the other hand, test the robustness of the
device over time, on the scale of a research project. During the injection campaigns,
five injection flow rates (0.005–0.01–0.02–0.05–0.1 l/s), 5 heating powers (2–4–6–
8–10 W/ml) and three distances injection point – heating cable (0–1.6–3.0 m) were
tested. Each of the heating powers has been the subject of a reference measure-
ment, that is to say a measurement of temperature rise in the vicinity of the optical
fiber and the heating cable, in the absence of leaks.
4.4. MAIN RESULTS AND RECOMMENDATION
It has been shown that all the leaks with a flow rate greater than or equal
to 0.05 l/s and placed upstream of the optical fiber at the base of the backfill have
all been detected. With the detection criterion of 0.3◦ C, and with a heating power
greater than or equal to 4 W/m, no false positives were detected.
No significant impact of heating power on leak detectability was observed. In

other words, a leak detected with a power of 8 W/m was also detected with a power
of 2 W/m. However, the use of a low power induced false positives caused by the too
low temperature contrast induced by the heating. Consequently, a minimum power
of 4 W/m is recommended
Increasing detection intensities were observed with the injected flow rates. At
equivalent flow rate, the detection intensities decrease if the distance between the
leak and the measurement cable increases, in line with the fact that the leakage
speed decreases along the propagation of the leak (diffusion in the drainage geo-
composite). Thus, on a real site, it is recommended to install optical fibers at several
sides of the basin, with a spacing which will depend on the desired level of leak
detectability.
167
Q.106 - R.12
5. CONCLUSION
In general, for structures retaining water, the detection and localization of a

leak is essential for managers in order to guarantee the stability of the structures
and ensure their durability. By addressing this need for monitoring in an efficient
manner, fiber optic measurement technology is now democratizing and extending
to new uses with an industrial maturity of materials, methods and more accessible
installation costs, in the order of 30 k/km. To fully exploit it, it is necessary to first
define the most appropriate operating method for the structure and its challenges.
With the aim of detecting and localizing the most dangerous leaks for its
embankments as early as possible, EDF’s monitoring services have integrated a new
operational monitoring process based on three key concepts: use of quality data,
industrialization based on Fox’Hy, industrial application allowing the exploitation of
AJOUT, a passive method performing in detection and requiring neither external
inputs, nor long history, and finally crossing of the results with other models for
validation or quantification and reconciliation with the conventional surveillance data.
For the altitude reservoirs where the leaks are both detrimental to the opera-
tion of the structure but also dangerous with regard to the risks of instabilities, the
active method has been successfully applied in a representative INRAE demonstra-
tor. In particular, it has been shown that the technique is capable of locating without
false positives or negatives a leak with a flow rate greater than or equal to 0.05 l/s
and placed upstream of the optical fiber. However, as the level of leak detectabil-
ity depends on the distance between the leak and the optical fiber, the sizing of
the optical fiber network to be positioned in the structure will be a key step in the
performance of the device.
ACKNOWLEDGEMENTS
The experiment performed in the INRAE site was supported by the

RESBA project funded by European Union under the INTERREG ALCOTRA
program.
REFERENCES
[1] KAPPELMEYER O., The use of near surface temperature measurements for
discovering anomalies due to causes at depths, Geophysical Prospecting,
5(3):239–258, 1957.
168
Q.106 - R.12
[2] JOHANSSON S., Seepage monitoring in embankment dams. Thèse de

doctorat, KTH, Stockholm, 1997.
[3] RADZICKI K., BONELLI S., Impulse response function analysis of pore
pressure in earthdams. 18ème Congrès Français de Mécanique, Grenoble,
France. 6 p., Aug 2007.
[4] CUNAT P., Détection et évaluation des fuites à travers les ouvrages
hydrauliques en remblai, par analyse des températures réparties, mesurées
par fibre optique, Thèse de doctorat, LTHE, Grenoble, 2012.
[5] KERZALE S., Estimation des débits de fuites dans les digues en terre par
thermométrie. Thèse de doctorat, université de Lorraine. 2013.
[6] GUIDOUX C., Développement et validation d’un système de détection et de

localisation par fibre optique de zones de fuite dans les digues en terre. Thèse
de doctorat, université Joseph Fourier de Grenoble. 2006.
[7] COURIVAUD J.-R., LORRAIN N., MARTINOT F. BEGUIN R., Quantifying

seepage flow velocities in embankment dams from optical fibre distributed tem-
perature measurements. Long-Term Behaviour and Environmentally Friendly
Rehabilitation Technologies of Dams. 2017.
[8] GARANDET A., BROTHIER C. MARTINOT F., Techniques Nouvelles

d’Auscultation et de Surveillance des Digues de Grand Linéaire, ICOLD
Vienne, 2018.
[9] KHAN A.A., VRABIE V., MARS J., GIRARD A., D’URSO G., A source sepa-
ration technique for processing of thermometric data from Fiber-Optic DTS
measurements for water leakage identification in dikes. Sensors Journal,
IEEE, 8(7):1118–1129., 2008.
[10] DORNSTADTER J., HEINEMANN B., In situ Detection of Internal Erosion, 8th
ICOLD European Club Dam Symposium, Innsbruck, 481–485, 2010.
[11] PERZLMAIER S., Verteilte Filtergeschwindigkeitsmessung in Staudämmen.

Thèse de doctorat, TU München, Munich, 2007.
[12] FAILLEAU G. et al., Development of facilities and methods for the metrological
characterization of distributed temperature sensing systems based on optical
fibres, Meas. Sci. Technol. 28 (2017) 015009 (12pp).
[13] BOUTRY M. RISBA, RAPPORT FINAL, Action 2.1, 65 p, 2014.
[14] PEYRAS L. ET MERIAUX P., Retenues d’altitude. Quae. 352 p, 2009.
169
Q.106 - R.13
-------
-------
MESURES IN-SITU: UN ATOUT POUR JUGER DE LA REPRÉSENTATIVITÉ

DES MODELISATIONS NUMÉRIQUES (∗)
Xavier MOLIN
Ingénieur, TRACTEBEL ENGINEERING
Patrice ANTHINIAC
Expert, TRACTEBEL ENGINEERING
Pauline BOFFETY
Ingénieur, EDF CIH
Philippe KOLMAYER
Expert, EDF CIH
FRANCE
RÉSUMÉ
Le gonflement du béton et les séismes sont des sujets de mieux en mieux

appréhendés. Il y a eu des évolutions importantes dans la modélisation de ces
phénomènes. Afin de valider les modèles numériques parfois complexes, il est
important de les confronter à des mesures in-situ spécifiques.
Les mesures de bruits ambiants sont utilisées pour caler les rigidités
dynamiques de l’ouvrage dans le modèle numérique. Les résultats sur deux
ouvrages sont présentés. L’erreur entre la fréquence mesurée et celle calculée à
partir des rigidités classiques peut être importante. Dans un des cas, cela aurait
conduit à sous-estimer l’amplification du séisme par le barrage.
∗ In-situ measurements: an asset for assessing the representativity of numerical modeling
170 DOI: 10.1201/9781003211808-109

Q.106 - R.13
Les contraintes évaluées par le modèle numérique sont directement dépen-

dantes du module de béton considéré: un calage du modèle numérique sur
uniquement des déplacements issus de l’auscultation n’est pas suffisant et peut
conduire à une importante surestimation des contraintes. Les mesures de con-
traintes in-situ permettront de vérifier si le niveau de contraintes dans le barrage est
acceptable à l’instant de la mesure, et de valider la représentativité du modèle. Le
REX d’EDF montre qu’une modélisation par analogie thermique peut suffire pour
les barrages à faible taux de gonflement tandis qu’un modèle d’endommagement
sera nécessaire afin de reproduire le niveau de contrainte réel mesuré d’un barrage
à fort taux de gonflement. Les mesures de contraintes in-situ constituent donc une
bonne pratique de validation des modèles numériques.
SUMMARY
The understanding of behavior of concrete swelling dams and behavior during

seismic loading is improving. Some important evolution occurred recently, thanks to
numerical modeling. To validate those complex models, it is important to compare
them to in-situ measurements.
Ambient vibration recordings are used to adjust the dynamic stiffnesses of the
structure in the numerical model. The results for two dams are presented. The error
between the first frequency measured and the one estimated in the model can be
important. For one dam, the error would have led to underestimate the dynamic
amplification of dam displacements.
The stresses estimated by numerical modeling are linked to the deformation

modulus of concrete considered: the validation of a numerical model only on dis-
placements of dam from monitoring can lead to an important over-estimation of the
stresses. The in-situ stress measurements will allow to assess the actual stress
state of the dam and validate the numerical model. EDF experience is that, as a
simple numerical model can be representative of a dam behavior with slow swelling
rate, it is not for dams with high swelling rate. The measurement of in-situ stresses
is therefore recommended to validate a dam behavior analysis.
1. INTRODUCTION
L’analyse du comportement des barrages en béton est en constante évolu-

tion. Elle tire profit des capacités de modélisations qui deviennent de plus en plus
élaborées. Néanmoins, le recours à ces modèles sophistiqués nécessite une atten-
tion particulière, du fait de leur complexité et de leur sensibilité. L’application de ces
modèles à des ouvrages existants requiert ainsi quasiment systématiquement de
les confronter à de données d’auscultation ou des mesures in situ spécifiques.
171
Q.106 - R.13
Deux grands sujets sont abordés dans le présent article.
Le premier concerne le comportement dynamique des barrages en béton. A

travers le retour d’expérience de Tractebel sur deux ouvrages, l’apport de mesures
de bruits ambiants dans l’analyse dynamique des modèles est exposé.
Le second sujet montre les apports des mesures de contraintes in situ dans la
validation et le recalage des modélisations du comportement des barrages atteints
de gonflements. Le retour d’expérience d’EDF sur les différents essais réalisés sur
quatre de ses ouvrages est ainsi présenté.
À l’issue de cette analyse, des recommandations sont proposées pour prendre

en compte ces spécificités dans l’analyse du comportement des barrages en béton.
2. APPORT DES MESURES DE BRUITS AMBIANTS À L’ANALYSE DU

COMPORTEMENT SISMIQUE DES BARRAGES BÉTON
2.1. SURVEILLANCE DES OUVRAGES AU SÉISME: ENTRE PRÉVENTION ET

PRÉDICTION
Les sollicitations sismiques sont par nature brutales, instantanées et spéci-

fiques. Ces caractéristiques font qu’il n’est pas possible d’agir sur l’ouvrage pendant
un séisme. Il n’est pas non plus possible d’extrapoler le comportement habituel
de l’ouvrage à partir des observations réalisées quotidiennement: les oscillations
amont et aval lors d’un séisme sollicitent le barrage de manière très différente des
chargements habituels.
La sûreté et la sécurité des ouvrages pendant un séisme sont donc assurées

en amont dans son dimensionnement et dans la définition de l’organisation à met-
tre en place après une secousse d’une part, et d’autre part par la réalisation
après l’évènement d’un diagnostic post-séisme afin de vérifier que l’ouvrage s’est
comporté conformément à ce qui avait été anticipé.
La sûreté repose donc:
• sur la prédiction du comportement sismique, en particulier sur la modélisation

du barrage et son dimensionnement, pour évaluer l’amplification du barrage
pendant le séisme, et les champs de contraintes, accélérations, déplacements
associés, ainsi que sur le retour d’expérience des ouvrages ayant subi des
séismes majeurs
• sur la prévention, en s’assurant qu’après un séisme il sera possible de réaliser
les visites d’inspections, de contrôler les équipements hydromécaniques
(organes sensibles) et de préserver la fonctionnalité d’un évacuateur de crues
172
Q.106 - R.13
ou de dimensionner la vidange de fond afin d’avoir une capacité suffisante

pour réduire rapidement la pression de l’eau qui s’applique sur le barrage en
cas d’alerte.
Dans les paragraphes suivants, il sera discuté des moyens et méthodes

pour prédire le comportement de l’ouvrage, à partir du retour d’expérience,
mais également des modélisations numériques, qui peuvent donner des résul-
tats assez différents en fonction des méthodes mises en œuvre et des paramètres
(caractéristiques) qui ont été retenus.
2.2. RETOUR D’EXPÉRIENCE: DES OUVRAGES PLUTÔT RÉSISTANTS
Peu d’ouvrages sont équipés d’une auscultation spécifique permettant de

mesurer les sollicitations du barrage au cours d’un séisme (accéléromètres). Cela
ne diminue en rien la sûreté ni la sécurité des ouvrages existants, car le diag-
nostic post-sismique reposera principalement sur les grands principes usuels de
l’auscultation (inspection, et auscultation des sous-pressions, déplacements, etc.).
Il s’agit de s’assurer que le barrage peut être maintenu en service en conditions
normales d’exploitation. En revanche, lorsque la mesure des accélérations du bar-
rage pendant le séisme est disponible, cela est riche d’information pour analyser a
posteriori le comportement du barrage lors du séisme. Il s’agit d’un outil primordial
pour valider les méthodes de calculs comme on le verra dans le paragraphe suivant.
Le retour d’expérience ne montre pas de ruptures de barrage en béton au

cours d’un séisme, à l’exclusion du cas particulier du barrage de Shih-Kang à
Taïwan: traversé par une faille active qui n’avait pas été détectée lors de la construc-
tion de l’ouvrage, le séisme Chi-Chi a généré en 1999 un déplacement différentiel
des deux rives (provoquant un rejet vertical de 7,7 m) [1].
Différents ouvrages ont subi des séismes majeurs, et cela concerne tous les
types de barrage en béton.
Les dégradations sur les barrages-voûtes ont été limitées. Elles concernent
plutôt les ouvrages annexes, en particulier les piles des évacuateurs de crues inté-
grées dans l’ouvrage. Les barrages-voûtes de Rapel (109 m de hauteur, Chili)
Pacoima (113 m de hauteur, Etats Unis) ou Shapei (130 m, Chine) ont subi des
séismes majeurs. Seuls quelques désordres mineurs sont apparus, à proximité des
évacuateurs de crues (Rappel), ou des rives (Pacoima)
La grande majorité des barrages-poids ayant subi des séismes importants, ont
résisté sans dommage bien que les calculs prédisaient des contraintes de traction
élevées. Quelques fissures sont apparues toutefois dans certains cas. Les cas
suivants peuvent être mentionnés en particulier: le barrage Baozushi (132 m, Chine)
concerné par le séisme du Wechuan 2008 dont les fuites ont augmenté après le
séisme, le barrage de Lower Crystlal Sprinces (39 m, Etats Unis) qui n’a pas subi de
173
Q.106 - R.13
dommage malgré des contraintes de tractions supérieures à 5 MPa selon les rétro-
analyses réalisées ultérieurement, et le barrage Kyona (102 m, Inde) pour lequel
quelques fissures sont apparues.
Les barrages multi-voûtes et les barrages à contreforts ont un comportement

différent, et sont très sensibles aux sollicitations transversales. Le barrage à contre-
forts de Sefid Rud (106 m, Iran) a subi des dommages importants lors d’un séisme
de magnitude 7,3 à 7,7 en 1990. Une fissuration importante a été observée sur
la plupart des contreforts, en partie supérieure de l’ouvrage. Le barrage a pu être
conforté par la suite.
Finalement, si des désordres mineurs tels qu’une fissuration locale ou de

faibles déplacements différentiels ont pu être quelques fois observés à la suite à
de séismes majeurs, les ouvrages en béton présentent globalement une bonne
résistance au séisme, les barrages voûtes étant même particulièrement résistants.
En revanche, certaines vulnérabilités sont apparues:
• Les ouvrages hydromécaniques, en particulier les vannes. En effet,

l’oscillation en opposition des phases des piles des évacuateurs de crues
a généré certains désordres.
• L’augmentation des sous-pressions, en fondation, peut apparaître pour des
séismes mineurs. Un drainage confortable est ainsi une barrière de sécurité
appréciable, voire nécessaire pour les séismes importants; une augmenta-
tion démesurée des sous-pressions après un séisme pouvant conduire à une
instabilité générale.
2.3. MODÉLISATION: DES TECHNIQUES EN PLEINE ÉVOLUTION
La modélisation sismique des ouvrages a connu un véritable essor ces

dernières années, avec d’une part l’augmentation importante de la capacité de
calcul, et d’autre part la démocratisation de lois de comportement de plus en plus
élaborées. Cette complexité a été poussée par le constat d’une surestimation des
effets du séisme, parfois importante, des méthodes usuelles: lors des rétro-analyses
qui ont été effectuées, des tractions supérieures à 5 MPa ont été par exemple cal-
culées sur le barrage de Lower Crystal alors qu’aucune fissure n’a été observée
sur l’ouvrage. De la même manière, des amortissements proches de 20% ont été
nécessaires dans la modélisation du barrage de Kurobe, alors que le séisme initial
était très faible [2].
Les principales évolutions dans la modélisation concernent l’amélioration de la

prise en compte de l’interaction fluide structure, avec la modélisation de la retenue
sous forme de fluide compressible ou incompressible, et l’utilisation de frontières
absorbantes pour mieux modéliser les interactions sols structure.
174
Q.106 - R.13
Ces méthodes ont fait l’objet de nombreuses publications et peuvent être main-
tenant considérées comme matures et être appliquées sur projet. On citera le travail
d’A.K. Chopra, en particulier sur les barrages-voûtes de Mauvoisin et de Pacoima
[3][2], la coopération entre la JCOLD et le CFBR [4] qui tire profit d’une base de
données importante de mesures du comportement d’ouvrage au Japon, ou encore
les benchmarks de calcul organisés par le comité A de la CIGB dont les plus récents
concernent Gratz en 2015 (interaction fluide-structure), Stockholm en 2017 (inter-
action fluide structure et les frontières absorbantes) et Milan en 2019 (frontières
absorbantes).
On notera que ces dernières méthodes, parfois longues à mettre en œuvre et

difficiles à paramétrer, sont à réserver à l’analyse de cas spécifiques. Les méthodes
«historiques» doivent ainsi continuer à être utilisées si elles permettent de justifier
le comportement des ouvrages.
La modélisation fine des interactions sol structure ou fluide structure se des-

tine principalement aux ouvrages soumis à des aléas sismiques importants. En
diminuant les marges de calcul, en particulier sur les amortissements, il con-
vient de maîtriser l’ensemble des autres paramètres significatifs de l’ouvrage; les
approximations devenant de moins en moins acceptables.
In fine le comportement d’un ouvrage au séisme dépend de deux facteurs

principaux:
• La sollicitation (intensité, direction). L’analyse de l’aléa sismique, et notam-

ment la prise en compte des éventuels effets de site, n’est pas le sujet du
présent article. Un minimum de trois jeux d’accélérogrammes est requis pour
les calculs linéaires selon les bonnes pratiques;
• La résonance de l’ouvrage.
Cette résonance est gouvernée d’une part par la géométrie de l’ouvrage et la

rigidité des matériaux, qui vont déterminer la fréquence de résonnance (l’interaction
fluide structure tend également à diminuer cette fréquence), et d’autre part par
l’amortissement apparent, qui détermine le facteur d’amplification de la résonance.
La compressibilité de l’eau, mais surtout les frontières absorbantes apportent un
amortissement significatif, pouvant atteindre 20% pour les barrages-poids [4].
Pour des ouvrages neufs, la rigidité dynamique des matériaux est souvent
déterminée en multipliant par un facteur 1,25 à 1,5 la rigidité statique. Dans ce
cas, il est recommandé de vérifier sur le spectre la fréquence des premiers modes,
afin de s’assurer qu’une petite modification de la fréquence de résonnance ne peut
générer une augmentation significative des efforts, vitesses ou déplacements de la
structure.
Pour des ouvrages existants, il est possible de réaliser une mesure de ces
fréquences de résonances, fiabilisant ainsi les modèles de calcul et les résultats.
175
Q.106 - R.13
2.4. MESURES IN SITU: UN OUTIL QUI PERMET DE FIABILISER LA PRÉDICTION
Deux types de méthode sont utilisés pour mesurer la fréquence de réso-

nance des ouvrages. Bien utilisées, elles permettent chacune d’avoir des résultats
satisfaisants:
• les mesures par vibration forcée

• les mesures de bruits ambiants
Les mesures par vibration forcée consistent à faire vibrer une masse sur
l’ouvrage, et de mesurer les accélérations en plusieurs points de mesures sur
l’ouvrage. En choisissant correctement l’emplacement de ces mesures, il est ainsi
possible de déterminer la fréquence de résonnance, la déformée, et l’amortissement
de la sollicitation. L’énergie d’excitation est faible (5kW) comparée à l’énergie d’un
séisme. Elle permet de réduire le ratio entre le signal à mesurer et le bruit. De telles
mesures ont par exemple été réalisées sur le barrage multi-voûte Daniel Johnson
(Canada) dynamique, et ont permis de caler le modèle avec un module du béton et
du rocher respectivement de 44 et 40 GPa [5]. Le signal d’excitation étant connu,
les capteurs sont installés en crête de l’ouvrage.
Les mesures de bruits ambiants sont une autre alternative. L’ouvrage est
naturellement soumis à diverses sollicitations, de faibles voir très faibles inten-
sités, mais couvrant l’intégralité du spectre. Ces mesures consistent à «écouter»
le bruit ambiant sur l’ouvrage. Le signal initial n’étant pas connu, cette méthode
nécessite d’installer des capteurs à la fois en crête de l’ouvrage, mais également
en pied. Un traitement statistique des signaux entre la crête et le pied permet de
déduire l’amplification des sollicitations, et donc de calculer les fréquences de réson-
nance, ainsi que la déformée des modes propres lorsque le pas de mesure est
suffisant (plusieurs capteurs). Les mesures peuvent être réalisées sur une journée,
et l’appareillage de mesure est facilement transportable.
Cette méthode analyse les vibrations du barrage pour une faible énergie de
sollicitation. Par conséquent, même si ces vibrations permettent de déterminer la
rigidité des matériaux, elles ne permettent pas de mesurer les effets non linéaires
qui apportent de la souplesse à la structure et augmentent la dissipation de l’énergie.
Une revue un peu plus détaillée des avantages de cette auscultation a été
réalisée par EDF, en particulier dans les publications [7] et [8]. Une campagne d’un
an d’enregistrement a été menée sur la voûte de Saint Guérin en 2016 et permet de
s’interroger sur l’impact de la cote de la retenue et des conditions thermiques sur le
comportement vibratoire [9].
A noter qu’un groupe de travail de l’Association Française de Parasismique a

été constitué afin de rédiger un «Guide relatif à la mise en œuvre et à l’exploitation
des méthodes de caractérisation dynamique in situ des bâtiments et ouvrages».
176
Q.106 - R.13
Fig. 1
Mesure des bruits ambiants – Déformée du premier mode à environ 5 Hz.
Ambient noise measurements – Deformed shape of first mode at about 5 Hz
2.5. EXEMPLE DE MESURE SUR UN BARRAGE-VOÛTE
Les mesures de bruit ambiant ont été réalisées sur deux barrages-voûtes
de géométrie similaires. Il s’agit de deux voûtes fines, d’environ 50 m de hauteur
chacune.
Une dizaine de capteurs ont été installés sur la crête, ainsi qu’en pied du
barrage. Les mesures se sont déroulées sur une journée.
Les dix premiers modes du barrage ainsi que la déformée ont été déterminés.
La figure ci-dessus présente le premier mode, dont la fréquence de résonance est
de l’ordre de 5 Hz. La déformée n’est pas symétrique, les deux rives oscillent en
opposition de phase.
Une modélisation sismique avancée de l’ouvrage a été réalisée. L’interaction

fluide structure a été modélisée en considérant un fluide incompressible, et des
frontières absorbantes (PML, Perfectly Matched Layers) ont été introduites pour
prendre en compte l’interaction sol structure. La déformée du mode propre est bien
reproduite par le modèle aux éléments finis. En revanche pour des modules de
rigidité usuels, la fréquence de résonnance est plus faible (3 Hz) que celle qui a
été mesurée. Il faut augmenter les rigidités au-delà de 60 GPa pour la fondation, et
50 GPa pour le barrage pour atteindre une fréquence de 5 Hz. Ces paramètres ont
été jugés déraisonnables (valeurs trop élevées).
Une autre explication a été recherchée, et les blocages entre le barrage et la

fondation ont été modélisée: lors des excavations du barrage, la tranche superficielle
177
Q.106 - R.13
Fig. 2
Déformée du premier mode propre calculée par le modèle aux éléments finis.
Deformed shape of the first eigen mode calculated by the finite element model.
du rocher, jugée de qualité insuffisante a été excavée pour atteindre le rocher sain
sur lequel a été fondée la voûte. À l’issue de la construction, l’espace entre la voûte
et les talus des excavations a été comblé par du béton. C’est ce qui est appelé
blocage du barrage.
En rigidifiant le pied du barrage, ces derniers augmentent la rigidité globale de

l’ouvrage. La tranche superficielle du rocher étant de moins bonne qualité, celle-ci
a été distinguée dans le modèle, et trois rigidités différentes ont été considérées: le
barrage, le massif de fondation et les blocages.
En considérant ces blocages, une fréquence du premier mode de 5 Hz est

obtenue pour une rigidité du béton de 40 GPa, de 30 GPa pour le rocher, et 2,5 GPa
pour les blocages.
Au final, le ratio entre les modules statiques (court terme, calé à partir des
modèles HST réalisés sur les mesures d’auscultation des pendules) et dynamiques
est respectivement de 1,15 pour le béton, 2,0 pour la fondation, et 1,6 pour les
blocages.
Un calcul identique a été réalisé sur un second barrage-voûte similaire de

50 m de hauteur. La mesure de bruit ambiant a été réalisée de la même manière
que la précédente. Le barrage étant plus épais et les excavations moins profondes,
les blocages amont et aval du barrage n’ont pas d’influence significative dans la
réponse du barrage au séisme. Les calages statiques et dynamiques ont conduit
à considérer un ratio entre le module statique et dynamique de 1,0 pour le béton,
et 1,6 pour la fondation. Le module statique est respectivement de 30 GPa pour le
béton, et 10 GPa pour la fondation. La fréquence du mode propre est de 6,8 Hz.
178
Q.106 - R.13
Fig. 3
Comparaison entre la fréquence du premier mode mesuré (trait noir continu), celle
calculée en multipliant les modules statiques par 1,5 (traits pointillés noirs),
superposés au spectre (bleu)
Comparison between the frequency of the first mode measured (solid black line)
and that calculated by multiplying the static moduli by 1.5 (black dotted line),
superimposed on the spectrum (blue)
Pour ces deux projets, il a été observé que la rigidité de la fondation avait un
impact plus important sur la fréquence des modes propres que la rigidité du béton.
Si une attention particulière est portée sur la fréquence de résonance, c’est

que celle-ci joue un rôle fondamental dans la réponse du barrage. Lorsque ces
fréquences correspondent aux fréquences où l’énergie du séisme est importante,
l’amplification du barrage le sera également. Pour analyser cette sensibilité, il est
possible de superposer les fréquences des modes propres sur le spectre du séisme.
Les champs de contraintes, d’accélérations, et de vitesses calculés sont

d’autant plus importants que l’accélération spectrale est élevée. Dans le cas du
premier barrage, sans mesures, l’erreur qui aurait été réalisée est importante. Dans
le second cas, les 2 fréquences sont situées au maximum de l’accélération spectrale
et les résultats sont équivalents.
179
Q.106 - R.13
2.6. CONCLUSION
Les sollicitations sismiques constituent un chargement particulier. La sûreté

et la sécurité des ouvrages reposent essentiellement sur l’anticipation et le bon
dimensionnement des ouvrages.
Le retour d’expérience international montre que les ouvrages en béton ont

dans l’ensemble bien résisté aux secousses sismiques, même fortes. Une attention
particulière doit toutefois être portée à la vulnérabilité des équipements hydromé-
caniques, qui assurent parfois des fonctions essentielles comme l’évacuation de
crues. Il convient également de prévoir un drainage généreux pour les barrages dont
l’aléa sismique est important, la plupart des séismes, même de faibles intensités,
ayant été suivis par une augmentation des fuites ou des sous-pressions.
Le comportement des barrages béton sous sollicitations sismiques peut

être aujourd’hui bien approché par des modèles de calcul sophistiqués qui ont
connu une évolution importante ces dernières années. Les interactions fluide-
structure et sol-structure permettent maintenant de reproduire les amplifications
constatées sur les barrages. Ces méthodes permettent de tenir compte d’une dis-
sipation de l’énergie bien plus importante que celle considérée dans les modèles
traditionnels. En revanche, en diminuant les marges de calculs, les autres approx-
imations deviennent moins acceptables, en particulier sur le choix des fréquences
de résonance (qui dépendent principalement de la géométrie et la rigidité des
matériaux).
Pour des ouvrages en service, il est conseillé de réaliser des mesures de

bruits ambiants (ou de vibration forcée), qui permettent de déterminer la fréquence
de résonance et la déformée du barrage. Deux cas ont été traités. À l’issue du calage
des modèles numériques sur les mesures, les ratios entre les rigidités statiques et
dynamiques varient d’un facteur 1 et 1,15 pour le béton, et 1,5 et 2 pour le rocher.
Ces ratios qui peuvent être comparés à la valeur généralement adoptée de 1,25.
Si cette hypothèse classique conduit à des résultats équivalents pour le second

cas, l’erreur commise dans le premier cas aurait été importante (sous-estimation de
la fréquence du mode fondamentale d’un facteur 2): cette différence est attribuée à
la présence des blocages en béton entre le barrage et les excavations du barrage.
Pour des ouvrages neufs, il n’est pas possible de se référer à de telles

mesures. Les rigidités dynamiques sont généralement déterminées en multipliant
les modules statiques par un facteur 1,25 ou 1,5 en fonction des recommanda-
tions appliquées. Au regard des différences observées entre cette méthode et
les mesures de bruits ambiants sur les ouvrages en services, les auteurs recom-
mandent de réaliser une étude de sensibilité sur les modules afin de s’assurer
qu’une faible modification de la fréquence de résonance ne conduise pas à une
augmentation significative de la réponse de l’ouvrage.
180
Q.106 - R.13
3. GONFLEMENT DES OUVRAGES
Certains barrages du parc EDF souffrent de pathologies de gonflement du

béton. Les modèles numériques, utilisés dans les études de comportement des
barrages, sont aussi utilisés dans le cas des barrages gonflants pour affiner la con-
naissance du comportement et prévoir le comportement long-terme des ouvrages.
Dans ces modèles, l’évaluation des contraintes internes est une donnée primordiale
voire dimensionnante et il est donc nécessaire de comparer leur estimation avec la
mesure in situ.
Si les procédures de calage des modules réversibles sur les déplacements
hydrostatiques et saisonniers sont bien établies, Celle du calage d’un module pour
la prise en compte du gonflement par analogie thermique est plus sujette à discus-
sion. En effet sur la durée du gonflement pouvant atteindre plusieurs décennies,
le phénomène de fluage ne peut être négligé. Le choix forfaitaire de ce module de
fluage (cf. le Bulletin 145 de la CIGB) conditionne alors directement les contraintes
fournies par le calcul qui peuvent se révéler très importantes. La connaissance de
la contrainte in situ permet alors de valider le modèle numérique.
La comparaison de résultats modèles numériques/contraintes in situ est
présentée dans ce paragraphe sur 4 grands barrages du parc EDF: le barrage
C, le barrage J, le barrage M et le barrage S, atteints soit par la réaction sulfatique
interne (RSI) soit par une réaction alcali granulat (RAG).
3.1. DESCRIPTIONS SUCCINCTES DES BARRAGES
Les barrages présentés souffrent tous d’une pathologie de gonflement avérée,

avec un taux de gonflement plus ou moins important. Les caractéristiques princi-
pales des barrages sont données ci-dessous, ainsi que la pathologie observée et
les déplacements irréversibles mesurés par l’auscultation:
Tableau 1
Caractéristiques principales des barrages considérés
BARRAGE C J M S
Type Barrage-poids Barrage-voûte Barrage-voûte Barrage-poids
rectiligne mince (+culée poids) curviligne
Hauteur sur fondation 60 m 57 m 32 m 32 m
Hauteur sur TN 52 m 22 m – 27 m
Construction 1941–1949 1948–1949 1940–1947 1926–1930
Patho-logie RSI (plus RSI (partie RSI (partie RAG (plus
importante en supérieure rive supérieure rive importante en
RD, depuis 1955) gauche de gauche de partie supérieure
l’ouvrage) l’ouvrage) de l’ouvrage)
Déplacements Déplacement En RG, vitesse Déplacement Déplacement
irréversibles vers l’amont de de déplacement vers l’amont de vers l’amont de
0,4 mm/an au vers l’amont de 1,5 mm/an au 0,6 mm/an (entre
max. 0,6 mm/an. maximum. 1974 et 1994).
Exhaussement En RD, vitesse Exhaussement Exhaussement
de 1,4 mm/an. de déplacement de 0,9 mm/an. de 0,2 à 0,4 mm/an
de 0,4 mm/an. (entre 1981 et 1994)
181
Q.106 - R.13
Tableau 2
Modèles numériques des barrages et taux de déformation imposée pour
reproduire les déplacements irréversibles
BARRAGE (A) ETUDES NUMÉRIQUES RÉALISÉES (B) GONFLEMENT
C Modèle linéaire avec prise en Déformation maximale imposée
compte du gonflement par analogie thermique par analogie thermique, variant de
(gonflement imposé), puis 18 à 80 µm/m/an entre le haut
modèle non-linéaire avec loi de et le bas.
comportement RGI_BETON
J Modèle linéaire avec prise en compte Déformation imposée de 2 à
du gonflement par analogie thermique 10 µm/m/an pour les parties
(gonflement imposé) supérieures des plots.
M Modèle non-linéaire avec éléments Variable entre 45 et 100 µm/m/an
joints au contact béton-rocher ainsi en crête et 0 µm/m/an en
qu’au niveau d’une fissure RG. Prise partie inférieure.
en compte du gonflement par analogie
thermique.
S Modèle linéaire. Variable entre 0 en pied et
Prise en compte du gonflement par 40 µm/m/an en crête.
analogie thermique.
3.2. DESCRIPTION DES MODÈLES NUMÉRIQUES
Dans les exemples proposés, la prise en compte du gonflement dans les

modèles numériques peut être réalisée de deux manières:
• Avec une loi de comportement du béton élastique linéaire, le gonflement peut

être modélisé par «analogie thermique» qui correspond à l’application d’une
dilatation imposée au barrage,
• Avec un calcul plus sophistiqué, en utilisant la loi de comportement nommée
RGI_BETON, développée par EDF et le LMDC dans Code_Aster. Cette loi
couple deux modèles de comportements: un premier qui permet de décrire
la fissuration du béton grâce à un modèle non linéaire couplant des critères
plastiques et des endommagements anisotropes, et un second qui permet de
décrire le fluage du béton.
La méthode par analogie thermique est à utiliser dans un premier temps. Si

elle ne suffit pas à décrire le comportement du barrage, ou si les contraintes et la
fissuration de l’ouvrage sont trop importantes, l’approche non linéaire basée sur la
loi RGI_BETON est alors mise en œuvre. Les modèles numériques utilisés pour
les 4 barrages du parc EDF sont présentés dans le tableau suivant. La colonne
(B) précise la déformation imposée déterminée par calage sur les déplacements
irréversibles pour les calculs par analogie thermique.
182
Q.106 - R.13
3.3. MESURE DE CONTRAINTES IN SITU ET COMPARAISON AUX RÉSULTATS DES

MODÈLES NUMÉRIQUES
Ces 15 dernières années, EDF a commandé plusieurs campagnes de

mesures de contraintes in situ par surcarottage sur différents barrages. Le retour
d’expérience sur ces mesures met en évidence qu’elles permettent d’estimer l’ordre
de grandeur des contraintes, mais qu’elles présentent une variabilité locale assez
importante. Cette variabilité est due au caractère local de la mesure; les cellules
des essais de surcarottage ne mesurant des déplacements que sur quelques cen-
timètres, la présence d’un granulat au niveau de la mesure peut avoir une influence
non négligeable. La variabilité peut être lissée en réalisant un nombre d’essais
suffisants afin de constituer une valeur moyenne représentative.
Dans ce chapitre seront présentés des résultats des mesures de con-

traintes réalisées avec la même technologie: l’essai de surcarottage avec cellule
Doorstopper (réalisé par l’entreprise SIAL.TEC).
Le principe de l’essai est le suivant: un forage est réalisé dans le corps du bar-
rage et en fond de forage, la cellule Doorstopper est collée (remarque: la profondeur
maximale du forage est limitée à 4,5 m pour un problème de temps de prise de la
colle). Puis, un sur-carottage est réalisé en fond de forage: la cellule Doorstopper,
équipée de 4 jauges de déplacements, va mesurer le relâchement radial du béton
décomprimé par le surcarottage. La carotte issue du surcarottage est ensuite cassée
du fond de trou, et elle est testée en essai de compression biaxiale afin de déduire
le module d’Young du béton. En connaissant ainsi la déformation (relâchement des
contraintes) et le module, les contraintes principales in situ peuvent être calculées
ainsi que leur orientation. Le coefficient de Poisson est considéré par défaut à 0,2.
3.3.1. Barrage J
5 essais de mesure de contraintes par surcarottage ont été effectués sur le

barrage J, dans un seul forage, régulièrement espacés dans l’épaisseur de la voûte
(entre 0,40 et 2,6 m). Ils ont été réalisés depuis le parement aval au niveau du
plot supposé présenter le plus fort taux de gonflement. L’essai réalisé à 2,6 m de
profondeur est positionné à 40 cm du parement amont (voûte d’épaisseur 3 m).
Les essais ont été réalisés à cote de retenue basse et en automne. D’après le
modèle numérique, au point d’implantation des forages, le chargement thermique
saisonnier fait varier les contraintes principales d’environ plus ou moins 0.5 MPa, ce
qui est très faible. Il a donc été postulé que les contraintes mesurées à cote basse
sont essentiellement dues phénomène de gonflement.
Les contraintes mesurées pour chaque essai sont présentées à la Figure 1.

Elles présentent une certaine variabilité, ce qui est naturel pour un matériau
183
Q.106 - R.13
Fig. 4
Mesures in situ sur le barrage J et comparaison avec les contraintes du modèle
numérique à chargement équivalent (gauche) – vue des contraintes de
compressions en parement aval d’après le modèle numérique (droite)
In-situ measurement on dam J and comparison with stresses of numerical model
(left) – downstream face view of compression stresses from numerical model (right)
hétérogène comme le béton. Toutefois, une tendance apparait: le champ de con-

trainte présente une compression uni-axiale, plongeant vers la rive avec des valeurs
de 2 MPa en partie aval de la voûte et inférieures à 1 MPa en partie amont.
Ces valeurs sont cohérentes avec les valeurs de contraintes obtenues dans cette
zone par le modèle aux éléments finis dans le cas «retenue vide + phénomènes
irréversibles» (soit 1,5 MPa à 2 MPa de compression en parement aval et ∼0 MPa
de compression en parement amont).
Les hypothèses de modélisation du gonflement par déformation imposée

(analogie thermique) sont donc confortées par ces essais.
Cependant, il peut être noté que la direction des contraintes du modèle n’est
pas identique à celle identifiée par les essais Doorstopper (∼45◦ d’écart entre les
directions des contraintes principales).
3.3.2. Barrage S
18 essais Doorstopper ont été réalisés depuis le parement aval du barrage

de S, répartis dans 5 forages. La Figure 5 présente l’orientation et les valeurs des
contraintes de compressions moyennes mesurées au droit de chaque sondage (DR1
à DR5), ainsi que les résultats des contraintes de compressions en parement aval
du modèle numérique.
La comparaison des orientations de la contrainte principale majeure entre

mesures et modèle numérique montre une bonne corrélation, sauf au droit de l’essai
DR4. Dans ce sondage, les photos des éprouvettes montrent toutes la présence
de gros granulats dont la taille peut atteindre le diamètre des échantillons. De plus,
184
Q.106 - R.13
Fig. 5
Contraintes principales de compression moyennes mesurées par forage et issues
de la modélisation numérique au barrage S
Compression stresses measured for each drill and compression stresses from
numerical modeling for dam S
il peut être noté la présence de nombreuses fissures autour du carottage DR4 qui
peuvent conduire à une réorientation des contraintes.
Pour tous les autres sondages, les résultats des essais semblent confirmer
les hypothèses de gonflement considérées: absence de gonflement de la partie
inférieure de l’ouvrage car l’orientation de la contrainte principale mise en évi-
dence correspond à un état de contrainte dû à un effet réversible hydrostatique, et
gonflement en partie supérieure mis en évidence par les contraintes horizontales.
Les amplitudes des contraintes mesurées par essais Doorstopper ont été
utilisées pour caler le modèle numérique.
Elles ont alors permis de définir le scénario de gonflement le plus réaliste

avec un temps de latence de 30 ans et un coefficient de fluage du module de béton
de 4, au lieu de 3 considéré habituellement par défaut. Cette modélisation surestime
toutefois encore les valeurs de compression en partie supérieure du barrage, jusqu’à
un facteur 2.
185
Q.106 - R.13
3.3.3. Barrage C
16 essais Doorstopper ont été réalisés sur le barrage C à partir de 4 forages;

2 horizontaux réalisés depuis le parement aval, 2 verticaux vers le haut depuis les
galeries. Dans chacun des forages les essais ont été réalisés à des profondeurs de
1.7, 2.3, 2.9 et 3.5 m de profondeur.
Le comparatif entre le modèle numérique avec la loi de comportement

RGI_BETON et les essais in situ est présenté à la Figure 6.
L’ordre de grandeur des contraintes du modèle est validé par les mesures in
situ. Les contraintes du modèle numérique par analogie thermique étaient environ
deux fois plus élevées que celles du modèle avec la loi RGI_BETON.
Fig. 6
Mesures des contraintes avec essais Doorstopper et contraintes de compression
du modèle numérique pour le barrage S
In situ measurement of stresses from Doorstopper overcoring tests and
compression stresses from numerical model of dam S
186
Q.106 - R.13
Fig. 7
Contraintes de compressions en parement aval du modèle numérique, position
des forages et résultats des mesures de contraintes au barrage M
Compression stresses on downstream face from numerical model, positions of
drills for tests and tests results of in situ stresses measurement for dam M
3.3.4. Barrage M
30 essais Doorstopper ont été réalisés sur le barrage M dans 10 forages

réalisés depuis le parement aval, à des profondeurs de 0.5, 1, et 1.5 m. Les résultats
des essais sont présentés à la Figure 4.
Les essais ont été réalisés pour une cote de plan d’eau de 5 m inférieure à la
RN, au mois de septembre. Les résultats du modèle considérés pour la compara-
ison sont pour deux cas de charge: «RN + été» et «RN + thermique neutre» et ne
correspondent donc pas strictement au chargement du barrage lors des essais.
En tenant compte de ces limitations il apparait tout de même qu’en par-

tie haute, l’orientation des contraintes mesurées est plutôt cohérente avec les
prédictions de la simulation que ce soit en surface ou en profondeur.
187
Q.106 - R.13
Cependant, le modèle est en écart sur les sondages réalisés le long des
appuis: les compressions mesurées sont de l’ordre de 1 à 2 MPa dans la direction
parallèle au contact, qui ne sont reproduites que très partiellement par le modèle
en été et pas du tout en l’absence de chargement thermique.
Quantitativement (en termes d’ordre de grandeur mesurée), le modèle est

également en écart d’un facteur 2 environ avec les essais. Les contraintes mesurées
sont approximativement 2 fois plus faibles que celles calculées. Ainsi, pour le bar-
rage M, la pertinence de la modélisation par analogie thermique est questionnée.
Une étude avec un modèle RGI_BETON est en cours.
3.4. CONCLUSION
Les mesures de contrainte in situ mettent en évidence que la modélisation du

gonflement par analogie thermique surestime les contraintes.
Pour les barrages à faible taux de gonflement (barrage J), cette surestimation
est acceptable et la réalisation des mesures in situ permet généralement de valider
le modèle.
Pour les barrages à fort taux de gonflement (barrage M, barrage C), la

surestimation des contraintes ne devient plus acceptable: le module de fluage
habituellement considéré dans les modèles (module statique divisé par un facteur
3) est certainement trop élevé par rapport au module réel. La réalisation d’un mod-
èle avec endommagement du béton permet de s’affranchir de cette hypothèse et
permet donc de mieux reproduire l’état de contrainte réel de l’ouvrage.
En perspectives, il semble intéressant de continuer à réaliser des mesures de

contraintes in situ dans les barrages. Il est important d’avoir un nombre suffisant de
mesures pour obtenir un ordre de grandeur des contraintes et avoir une moyenne
représentative. Il semble opportun d’utiliser différentes méthodes de mesures de
contraintes afin de pouvoir confronter les résultats de différentes techniques (cellules
de surcarottage spéciales pour mesurer la contrainte en un point d’un forage à une
profondeur à 4,5 m pour un barrage épais, vérins plat pour les barrages minces…).
Enfin si ces mesures sont répétées dans le temps, elles peuvent servir à mieux
caractériser la cinétique du phénomène de gonflement (pour les barrages à fort
taux de gonflement).
REFERENCES
[1] K. Hansen & L. Nuss, Lessons learned from the earthquake performance of
concrete dams, Waterpower and dam construction, 2011.
188
Q.106 - R.13
[2] Long-term behavior and dynamic analysis based on observation records of

seismic motion of the kurobe dam, H. Akira, A. Mistuzumi, 24ième congrès
des grands barrage, Q95 R37, Kyoto 2012.
[3] Earthquake Analysis of Arch Dams: Factors to Be Considered, AK. Chopra,

ASCE Journal of structural Engineering, February 2012.
[4] Validation of Dynamic Analyses of Dams and Their Equipment: Edited Con-
tributions to the International Symposium on the Qualification of Dynamic
Analyses, J.J Fry, N. Masumoto & Al.
[5] Forced-vibration tests of Daniel-Johnson multiple arch dam, O Gauron & Al.,
Experimental vibration analysis for civil engineering structures, San Diego
2017.
[6] Pertinence du bruit de fond sismique pour la caractérisation dynamique et

l’aide au diagnostic sismique des structures de génie civil, F. Dunand, Doctorat
Sciences de la Terre et de l’Univers, Université Joseph Fourier, Grenoble I,
378 p.
[7] Tour d’horizon des mesures de réponse de barrages sous bruit ambiant, N.
Humbert, A. Diallo, T. Guilloteau, Colloque CFBR 2018.
[8] Mesures sur ouvrages hydrauliques, N. Humbert, AFPS 2018.
[9] A high-resolution seismological experiment to evaluate and monitor the seis-

mic response of the Saint-Guérin arch dam (French Alps), Koufoudi et Al.,
Seismological Research Letters (2018) 89 (4): 1576–1582.
189
Q.106 - R.14
-------
-------
CARACTÉRISATION DES FUITES ET DE LA LITHOLOGIE DES DIGUES DE

CANAUX PAR MÉTHODES GÉOPHYSIQUES ET FUSION DE DONNÉES (∗)
Théo DEZERT
Post doctorant, UNIVERSITÉ GUSTAVE EIFFEL
Christophe PICAULT
Expert Auscultation, CNR
Bruno DAUMAS
Chef de Projet, EDF
Olivier MAGNIN
Expert Géophysique, EGIS
Christophe VERGNIAULT
Référent Géophysique, EDF
Jean-Robert COURIVAUD
Expert, EDF
FRANCE
RÉSUMÉ
Depuis 2007 les canaux de la Durance et du Rhône doivent faire l’objet péri-
odiquement d’un Examen Exhaustif règlementaire, donnée d’entrée des Etudes
De Dangers. Dans cet article, une nouvelle méthodologie est développée et pro-
posée par la Compagnie Nationale du Rhône et Electricité De France. Elle est mise
∗ Characterization of leakages and lithology of channel dykes by geophysical methods and
data fusion
190 DOI: 10.1201/9781003211808-110

Q.106 - R.14
en œuvre directement sur les canaux en eau afin d’éviter les vidanges couteuses
et potentiellement préjudiciables pour la sûreté de l’ouvrage. Cette méthodologie
repose, en complément de la surveillance classique, sur l’utilisation de méthodes
géophysiques grand rendement dont la mise en œuvre ainsi que les avantages
et inconvénients propres à leur utilisation et interprétation sont détaillés dans
cet article. Les méthodes employées sont les suivantes: panneaux électriques
en roll-along, trainé avec électrodes à contact capacitif, méthode électromagné-
tique basse fréquence en champ proche, méthode radar, MASW, polarisation
spontanée en eau et à terre et méthode acoustique. Cet article présente un
retour d’expériences appliqué aux grands canaux du Rhône et de la Durance sur
l’identification d’éventuelles zones de fuite ainsi que d’hétérogénéités au sein des
remblais et/ou de la fondation mais également leur caractérisation lithologique.
L’utilisation de reconnaissances géophysiques apparaît comme une solution d’avenir
dans la mesure où elle permet i) une meilleure connaissance des zones de faib-
lesse et des zones de fuite, ii) d’éviter une vidange et iii) de minimiser les pertes
d’exploitation. Bien que les grands ensembles structurels aient pu être identi-
fiés sur les ouvrages du Rhône et que des écoulements aient été détectés sur
ceux de la Durance, donnant en partie satisfaction, de nombreuses améliorations,
notamment en termes d’interprétation sont encore à fournir. C’est pourquoi une
méthodologie de fusion d’informations géophysiques et géotechniques reposant
sur l’utilisation de la théorie des masses de croyance est proposée en ouverture.
Cette nouvelle approche propose une amélioration de la caractérisation des ensem-
bles lithologiques en tenant compte de la complémentarité et des spécificités des
données acquises en géophysique et par sondages carottés.
SUMMARY
Since 2007, the Durance and Rhône dikes are subject to a periodic compre-
hensive review, essential for Hazard Studies. In this article, a new methodology is
developed and proposed by the «Compagnie Nationale du Rhône» and «Électricité
De France». It is implemented directly on the channels to avoid draining process that
are expensive and potentially damaging to the safety of the structure. This method-
ology is based, in addition to conventional monitoring, on the use of high yield
geophysical methods. Their implementation as well as the pros and cons specific
to their use and interpretation are detailed in this article. The methods used are as
follows: ERT with roll-along technique, dragged capacitive contact electrodes, low
frequency near field electromagnetic method, radar method, MASW, spontaneous
polarization in water and on land, acoustic method. This article presents feedback on
the channels of the Rhone and Durance with the aim of identifying possible leakage
zones as well as heterogeneities in the structure or in the foundation, but also the
lithological characterization of these dikes. The use of geophysical reconnaissance
appears to be a solution for the future since it allows i) better knowledge of weak
areas including leakage areas, ii) to avoid draining and iii) to minimize operating
losses. Although the main lithological structures have been detected on the works
191
Q.106 - R.14
of the Rhône and leakages have been detected on those of the Durance, many
improvements, especially in terms of interpretation are still to be provided. There-
fore, a methodology for merging geophysical and geotechnical information based
on the use of the theory of belief masses is proposed. This new approach offers an
improvement in the characterization of the lithological assemblies, considering the
complementarity and specificities.
1. LES EXAMENS EXHAUSTIFS PAR VIDANGE: LIMITES ET

NOUVELLES APPROCHES
Jusqu’en 2007, une faible proportion des canaux du Rhône et de la Durance

était soumise à des exigences réglementaires. Depuis, l’ensemble des 400 km de
canaux est soumis à la nouvelle réglementation sur les barrages et la majorité
doit faire l’objet d’un Examen Exhaustif, donnée d’entrée des Etudes De Dan-
gers. Jusqu’à très récemment, ces examens (anciennement dénommés Examens
Techniques Complets (ETC)) étaient réalisés par un arrêt d’exploitation et un
abaissement très important du niveau d’eau (jusqu’à la vidange totale parfois)
donnant lieu à un examen visuel des parements intérieurs. A cause d’un envase-
ment important du radier et des bajoyers, ces examens visuels se sont avérés peu
pertinents, n’apportant pas d’information nouvelle à la connaissance des ouvrages.
De plus, l’abaissement du niveau d’eau pour un canal est toujours préjudicia-

ble pour son étanchéité, particulièrement lorsque celle-ci est obtenue par l’apport
des limons transportés par le cours d’eau venant colmater tout ou partie des lignes
d’écoulement. En effet, une vidange a toujours pour conséquences i) un lessi-
vage des limons lors de l’abaissement du plan d’eau, ii) une inversion des sens
d’écoulement et le décolmatage des limons, iii) une dessiccation des limons et
leur entraînement éolien lors de l’assec, iv) une augmentation, parfois irréversible,
souvent importante, des débits de fuite lors de la remise en eau.
Ainsi, cette méthode d’examen par abaissement important du plan d’eau ou

par vidange totale ne satisfait ni le service du contrôle (DREAL), en raison de la mau-
vaise visibilité de l’ouvrage mettant en cause l’exhaustivité attendue de l’examen,
ni les Maîtres d‘Ouvrage quant au coût de l’arrêt des aménagements, très impor-
tant au regard de la plus-value des informations apportées. Suite à ce constat de
manque de pertinence, une nouvelle méthodologie est développée à la fois par la
Compagnie Nationale du Rhône et Electricité De France. Elle est mise en œuvre
sur le canal en eau et se base sur la recherche d’écoulements ou de zones poten-
tiellement «faibles» au sein des remblais de grand linéaire et de leur fondation à
partir de mesures géophysiques, en complément de la surveillance et des mesures
d’auscultation plus courantes.
Complémentairement à l’utilisation de méthodes géophysiques, des sondages

géotechniques sont généralement menés afin de déterminer précisément les
paramètres géotechniques des matériaux. Une approche de fusion d’informations
192
Q.106 - R.14
Fig. 1
Vue d’une digue du Rhône (gauche) et d’un canal de la Durance vidangé (droite)
View of a Rhône dike (left) and of a drained channel Durance dike (right)
géophysiques (informations incertaines mais spatialisées) et géotechniques (infor-

mations fiables mais ponctuelles spatialement) permettant d’améliorer la caractéri-
sation des grands ensembles lithologiques est proposée sur une section de digue
du canal de la Durance en fin de rapport.
2. LES GRANDS CANAUX DU RHÔNE ET DE LA DURANCE
2.1. OUVRAGES DU RHÔNE
L’aménagement du Rhône, réalisé par la CNR entre 1930 et 1986, a permis

l’installation de 19 centrales hydroélectriques de la frontière suisse à la mer, et a
ouvert la navigation à grand gabarit de Lyon à la mer au moyen de 12 écluses.
L’aménagement a nécessité la réalisation de 400 km de digues fluviales (Fig. 1),
dont la charge permanente est de l’ordre d’une dizaine de mètres.
Ces digues sont construites avec les matériaux issus du terrassement du canal
(Fig. 2). De ce fait, en fonction des matériaux extraits, le profil et la constitution des
digues du Rhône vont varier. On identifie deux grands types: les digues en limon et
les digues en graviers, selon la constitution du noyau. Ces digues sont constituées
de matériaux perméables, et la limitation des écoulements au travers de l’ouvrage
est obtenue par le colmatage naturel des talus amont par les sédiments transportés
par le Rhône.
2.2. OUVRAGES DE LA DURANCE
La Durance est un puissant affluent rive gauche du Rhône qui a fait l’objet,
au début des années 1960, d’un vaste programme d’aménagement hydroélec-
trique. Ainsi, de 1963 à 1975, sont mis en service i) le plus grand réservoir
193
Q.106 - R.14
Fig. 2
Coupe type d’une digue fluviale du Rhône
Typical section of a fluvial Rhône dike
Fig. 3
Coupe type d’une section du canal de Sisteron.
Typical cross section of the Sisteron Channel.
artificiel d’Europe (retenue de Serre-Ponçon d’une capacité de 1,2 milliard de m3 )

en tête d’aménagement et ii) près de 200 km de canaux (Fig. 1), entrecoupés
d’une quinzaine d’usines exploitées par EDF, qui représentent une puissance totale
d’équipement de 2 000 MW et qui sont équipés de très nombreuses prises d’eau
agricoles et d’eau potable pour un débit de prélèvement avoisinant près de 100 m3 /s.
Ces canaux sont constitués d’un profil intérieur trapézoïdal (radier de 9 m de

large environ, pentes de 2/1 pour un tirant d’eau de 6 à 8 m au maximum, Fig. 3)
entièrement revêtu de béton de ciment ou de béton bitumineux. Ils sont également
«adossés» à un flanc de colline, ce qui engendre la présence de talus de très
grandes hauteurs, pouvant dépasser les 30 m au passage des talwegs les plus
profonds.
194
Q.106 - R.14
3. CAMPAGNES DE RECONNAISSANCES GÉOPHYSIQUES
3.1. OBJECTIFS
Dans le cas des canaux du Rhône, s’agissant de digues parcourues en

permanence par des écoulements internes (infiltration des eaux du canal), les
objectifs principaux des campagnes géophysiques sont d’abord d’identifier des vari-
ations et/ou anomalies de caractéristiques physiques pouvant correspondre à des
singularités géologiques ou géotechniques dans l’ouvrage ou sa fondation (zone
décomprimée, ou plus perméable, ou mécaniquement faible), et éventuellement
des écoulements internes excessifs. Dans le cas des canaux de la Durance, comme
l’étanchéité est assurée par un revêtement mince en parement amont, tout écoule-
ment à travers les remblais ou la fondation est considéré comme une défaillance
qui peut être préjudiciable à la pérennité de la digue. Les objectifs principaux des
campagnes géophysiques sont ainsi d’identifier d’éventuelles zones de fuite, puis
de les caractériser finement pour en définir le potentiel de danger.
3.2. MÉTHODES POUR LA CARACTÉRISATION GÉOLOGIQUE ET

GÉOTECHNIQUE
A ce stade des études, cinq méthodes géophysiques ont été testées sur un
secteur de 2.5 km de digues du Rhône pour réaliser un comparatif (rendement,
profondeur d’investigation, qualité et précision des informations obtenues). Il s’agit
de la tomographie de résistivité électrique (TRE), du traîné électrique à électrodes
capacitives (TEC), de la méthode électromagnétique basse fréquence en champ
proche (EM), de la méthode radar et la méthode sismique MASW.
3.2.1. Méthodes électriques et méthode électromagnétique

basse fréquence
La TRE a consisté ici à déployer un panneau électrique de 237.5 m de long

(96 électrodes espacées de 2.5 m) qui est prolongé par «roll-along» par moitié de
dispositif (soit 48 électrodes). Cette méthode permet d’obtenir un panneau élec-
trique continu. La profondeur d’investigation est de 20 m environ. Avec 1 km/j pour
10 h de travail à 3 personnes, cette méthode ne présente pas un grand rendement
mais elle a servi de référence pour évaluer les méthodes présentées ci-après.
Le TEC est une méthode qui consiste à tracter le dispositif électrique derrière
un engin motorisé. Les électrodes capacitives employées ici pour permettre cette
mesure en continue sont les OhmMapper, de Geometrics, USA. Le dispositif mis en
195
Q.106 - R.14
Fig. 4
Dispositif en phase de test déployé sur les digues du Rhône
Test phase device deployed on the Rhône dikes
œuvre (Fig. 4) permet une profondeur d’investigation de 8 à 9 m. Les mesures en

continu sont géoréférencées lors du tractage du dispositif à une vitesse de 3 km/h,
ce qui permet un rendement d’environ 15 km/jour avec un point de mesure tous les
mètres.
La méthode EM a été déployée avec le dispositif DualEM 642, développé

par le constructeur DualEM, Canada. Ce système de mesure se présente sous la
forme d’une perche de 6 m de long équipée de 3 récepteurs fixés respectivement à
2, 4 et 6 m de l’émetteur. La mesure est réalisée en horizontal et en vertical à une
fréquence de 9 kHz.
La profondeur maximale d’investigation atteinte est ici de 8 m. Le rendement

de cette méthode est de l’ordre de 15 km/j à 1 opérateur (la perche est fixée sur un
chariot et tirée derrière un véhicule). Les mesures sont réalisées avec un échan-
tillonnage d’une mesure par seconde, soit 1 mesure par 0.5 m, avec un véhicule
progressant à 2 km/h environ.
La figure 5 ci-après présente les profils de résistivité obtenus au moyen des

3 méthodes. En partie gauche du profil (Fig. 5), on observe une digue en graviers
de forte résistivité (>1500 .m) de 8 m de hauteur reposant, à partir de la cote 138,
sur des alluvions limono-sablo-graveleuses en place nettement moins résistantes
(couleur bleue, environ 100 .m). L’anomalie visible en TRE-WSR (Fig. 5.a) située
en fondation, entre les distances 400 et 500 m et entre les cotes 130 et 137 NGF
se situe en face d’une résurgence connue en parement aval de digue. Au-dessus
de cette anomalie on note que la digue en graviers présente corrélativement une
légère baisse de résistivité (Fig. 5a). En partie droite du profil, on identifie en surface
une couche graveleuse d’épaisseur métrique (>500 .m), puis un corps de digue
en limons (couleur bleue, moins de 200 .m) de 5 à 8 m d’épaisseur. Le corps de
digue limoneux repose sur les alluvions graveleuses en place (>500 .m).
196
Q.106 - R.14
Fig. 5
Vue comparée des graphiques d’inversion obtenus sur les digues du
Rhône avec les méthodes a) TRE (config. WSR), b) OhmMapper et c) DualEM.
Report schématique des singularités identifiées en TRE sur chaque profil.
Comparative view of the inversion graphs obtained on the Rhône dikes with the
a) ERT (WSR configuration), b) OhmMapper and c) DualEM methods.
Schematic report of the singularities identified with the ERT on each profile.
Par rapport à la méthode TRE, les valeurs de résistivités absolues du

OhmMapper (Fig. 5.b) et du DualEM (Fig. 5.c) sont globalement cohérentes.
Cependant, la profondeur d’investigation est plus faible (car la longueur du dis-
positif est plus faible) et atteint 9 m pour le OhmMapper et 8 m pour le DualEM. On
constate que les données OhmMapper et DualEM ne permettent pas d’imager la
couche résistive superficielle (couche de roulement identifiée en TRE en partie droite
du profil). En effet, les dispositifs présentent des longueurs de dipôles d’émission
et de réception trop importantes pour disposer d’une résolution inférieure ou égale
au mètre.
Comme on peut le constater, l’interface digue en graviers/alluvions en place

vers la cote 138 en partie gauche du profil n’est pas identifiée par le DualEM car
elle se trouve en limite d’investigation. On notera que les valeurs de résistivités du
DualEM sont moins contrastées que celles du OhmMapper, en particulier en partie
gauche du profil où la forte résistivité des graviers secs du corps de digue n’est pas
propice à la création du champ secondaire, ce qui donne une mesure plus bruitée
ayant une résolution moindre.
197
Q.106 - R.14
Fig. 6
Dispositif radar tracté par un véhicule sur les digues du Rhône
Radar device towed by a vehicle implemented on the Rhône dikes
3.2.2. Méthode radar
Deux campagnes ont été réalisées, avec des antennes de respectivement

100 MHz et 200 MHz. La profondeur d’investigation s’est avérée très limitée sur les
digues à noyau limoneux (entre 2 et 4 m) et assez limitée sur les digues en graviers
(6 à 7 m au plus). Les antennes radar émettrice et réceptrice étaient tractées à
l’arrière d’un véhicule roulant (Fig. 6) à une vitesse constante comprise entre 3 et
7 km/h en crête de digue. Passé le calage et l’étalonnage de la méthode (1 jour), le
rendement a été d’environ 15 km/j.
La figure 7 ci-après illustre les principaux résultats obtenus en 200 MHz: i)

un épaississement de la couche superficielle graveleuse au niveau d’un aqueduc
traversant le corps de digue est bien mis en évidence sur le radargramme (Fig. 7),
tandis que l’ouvrage en lui-même, à 4 m de profondeur, n’est quasiment pas révélé,
ii) la transition d’une digue en graviers à une digue en limon (Fig. 7) se traduit par
le passage d’une image radar hétérogène (partie gauche du profil) à une image
radar «blanche» (partie droite du profil) sans aucun écho au-delà de la couverture
graveleuse. Aucune interface remblai/sol d’assise n’est mise en évidence compte
tenu de la faible pénétration du signal.
3.2.3. Méthode sismique MASW (Multi-Channel Analysis of Surface Waves)
La méthode analyse la dispersion des ondes de surface (section pseudo-2D)

pour caractériser la vitesse sismique et l’interpréter ensuite en termes de propriétés
198
Q.106 - R.14
Fig. 7
a) Vue d’un approfondissement de l’interface couche graveleuse/noyau en limon
au droit d’un aqueduc traversant. b) Vue d’une zone de transition entre une digue
en graviers (à gauche) et une digue en limons (à droite).
View of the deepening of the gravel-on-silts interface above an aqueduct crossing
the dike. b) View of a transition zone between a gravel dike (left)
and a silt dike (right)
mécaniques du milieu (compacité notamment). Le dispositif est ici constitué de

24 géophones verticaux de type Gimbal tractables, espacés de 2 m. La profondeur
d’investigation est de l’ordre de 12 m. L’équipement est tracté au sol par un véhicule.
La source sismique, de type chute de poids accélérée, est activée tous les 10 mètres.
Le véhicule stoppe pour chaque tir. Le rendement de la méthode sismique est de
l’ordre de 2 km/j.
On observe bien en partie droite de la Fig. 8, la séparation entre le corps de

digue limoneux (Vs < 250 m/s, teinte bleue) et les sables et graviers en place sous-
jacents (Vs > 300 m/s), interface localisée entre les cotes 135 et 140. En partie
gauche du profil, la séparation entre le corps de digue en graviers et les alluvions
en place ne présente pas de contraste net en termes de Vs (interface localisée vers
les côtes 137/138 d’après la TRE). La dispersion des Vs dans les graviers (entre
260 et 360 m/s) est probablement liée à la présence de zones de graviers plus crus
et/ou plus compactés.
3.3. PRINCIPES ET MÉTHODES POUR LA RECHERCHE DE FUITES
Le principe de la recherche de fuites se déroule en deux phases: une pre-

mière mettant en œuvre des investigations grand rendement de pré-identification sur
l’ensemble des zones en remblais du canal, et une deuxième mettant en œuvre des
199
Q.106 - R.14
Fig. 8
Comparaison des profils TRE (config. dipôle-dipôle) et de la MASW
Comparison of the ERT (dipole-dipole configuration) and of the MASW sections
investigations locales pour en confirmer l’importance et les caractériser (localisation,

intensité, vitesse d’écoulement, cheminement).
3.3.1. Méthode à grand rendement de pré-identification
Les méthodes géophysiques grand rendement testées ont été la TRE et la

polarisation spontanée (PS) en eau et à terre. Le résultat de ces tests a mon-
tré que pour identifier des zones de fuites potentielles, ces méthodes possèdent
encore actuellement de nombreuses limites techniques. En effet, elles n’ont pas
permis de mettre en évidence avec un taux de réussite de 100% les zones de fuite
par ailleurs connues pour certaines, par exemple autour des ovoïdes constitués
de béton armé. En particulier, pour que la détection d’une fuite soit possible par
la méthode TRE, il faut disposer de suffisamment de contraste entre la résistivité
du sol et celle de l’eau: cela n’est donc pertinent que dans le cas de matériaux
suffisamment résistifs (non applicable pour des matériaux fins ou cohérents comme
les limons, l’argile ou les marnes). Enfin, la résolution spatiale de cette méthode
(i.e. la plus petite anomalie détectable) dépend de la distance entre électrodes
et de la profondeur d’investigation. Seules des anomalies, et a fortiori des fuites,
de grande taille (plusieurs mètres dans un plan vertical et une dizaine de mètres
dans le plan horizontal) sont identifiables avec cette méthode. Il serait donc illu-
soire de vouloir détecter des petites fuites saturant des zones de quelques dizaines
de centimètres à quelques mètres de diamètre. Cette méthode n’a donc pas été
200
Q.106 - R.14
retenue pour cette phase de pré-identification. Seule la méthode par PS (en eau ou
à terre) a été retenue. Elle repose sur la mesure du potentiel électrique généré par
le phénomène d’électrofiltration. Ce phénomène génère des variations de potentiel
de l’ordre de 5 à 50 mV (baisse de potentiel pour un entonnement de fuite, augmen-
tation de potentiel pour un exutoire). La principale difficulté dans la mise en œuvre
provient du masquage possible de ce potentiel d’électrofiltration par d’autres poten-
tiels électriques d’intensité plus importante (oxydo-réduction, présence d’ouvrages
métalliques à proximité, courants souterrains vagabonds…). En tout état de cause,
malgré ses limites actuelles, cette méthode grand rendement demeure plus perti-
nente que la détection de zones d’entonnement par examen visuel du parement
amont après vidange.
Ainsi, cette méthode met en évidence des anomalies qui peuvent être des
indices de fuite. En revanche, elle n’est pour l’heure ni exhaustive ni parfaitement
certaine (nombreuses anomalies causées par des singularités comme par exemple
une carcasse de voiture). Il est donc indispensable de croiser géographiquement
les anomalies détectées avec les données d’examens visuels et d’auscultation et
de constitution structurelle de la digue. In fine, l’analyse fine de toutes ces informa-
tions permet de retenir des zones dites «suspectes» nécessitant des investigations
locales.
3.3.2. Méthodes locales de confirmation et caractérisation
Sur les zones retenues à l’issue de la phase de pré-identification, des investi-

gations locales de confirmation et caractérisation sont réalisées par la méthodologie
suivante:
– Recherche de zones d’entonnement de fuite par méthode acoustique.

– Pour chaque zone confirmée, caractérisation (vitesse, cheminement) de
l’écoulement par traçage à la saumure (réalisé avec la méthode TRE).
Cependant, pour réaliser ces méthodes, le canal doit être à sa cote

d’exploitation normale et le débit transitant doit être limité très fortement, engendrant
des arrêts de production importants. Le retour d’expérience de ces diverses inter-
ventions réalisées sur les canaux de la Durance est très complet [1]. En particulier,
la détection des entonnements par recherche acoustique s’est avérée très satis-
faisante. C’est une méthode éprouvée et fiable dans la mesure où les écoulements
recherchés sont turbulents (à travers une fissure ayant un contraste de perméabilité
important mais pas pour des écoulements dans un matériau homogène). Sa finesse
de détection est suffisante pour les défauts recherchés (Fig. 9). Cette méthode est
systématiquement employée sur les zones préalablement identifiées, dans la limite
d’un rendement de mise en œuvre qui ne dépasse pas 100 m linéaire de canal
par jour.
201
Q.106 - R.14
Fig. 9
Cartographie des bruits permettant de localiser un point potentiel d’entonnement
de fuite sur le canal de Saint-Estève.
Noise mapping to locate potential inflow leakage on the Saint-Estève Channel.
4. INTÉRÊTS ET LIMITES DES MÉTHODES
Le tableau 1 a été établi sur la base des tests réalisés sur les digues du
Rhône. Il présente les aptitudes de chaque méthode à détecter une caractéristique
donnée. La clarté de la réponse de chaque méthode géophysique étant très liée au
contexte géologique rencontré, ce tableau n’a pas pour objet une hiérarchisation
des méthodes mais plutôt une identification des aptitudes de chacune d’entre elles
dans le contexte particulier des digues du Rhône. Ici la méthode TRE sert d’étalon,
mais n’entre pas dans le champ des méthodes à grand rendement.
Sur la base des informations connues sur le secteur d’essai (lithologie, pro-
fondeur des interfaces, transitions latérales de nature de remblai, singularités), les
différentes méthodes montrent des aptitudes très différentes à détecter la carac-
téristique recherchée, et aucune ne se révèle exhaustive. Néanmoins, les grands
ensembles géologiques sont identifiés et quelques singularités connues présentent
une signature nette.
Ainsi, du point de vue de la structure de l’ouvrage, les caractéristiques visi-

bles sur les plans et coupes d’exécution sont bien retrouvées. Dans ce contexte d’un
202
Q.106 - R.14
Tableau 1
Comparaison des aptitudes des différentes méthodes géophysiques à grand
rendement testées sur les digues du Rhône
Comparison of the aptitudes of different geophysical methods tested on the Rhône
dikes
TRE OHMMAPPER DUALEM RADAR MASW
Coût hors 1350 =
C/km 900 =
C/km 900 =
C/km 700 =
C/km 1600 =
C/km
interprétation
Rendement 1 km/j 15 km/j 15 km/j 15 km/j 2 km/j
constaté
Dispositif 237.5 ml/96 30 ml/6 6 ml/1 Antennes 100 46 ml/24
électrodes électrodes émetteur et et 200 MHz géophones
capacitives 3 récepteurs
Profondeur 20 m 9m 8m 2 à 4 m sur 12 m
d’investigation digue limon et
constatée 6 à 7 m sur
digue graviers
Identification Partielle (le Partielle (le Partielle (le Partielle Oui
de la nature contraste de contraste de contraste de (limon identifié
lithologique résistivité est résistivité est résistivité est par défaut par
(limons ou atténué par atténué par atténué par l’absence de
sables et la présence la présence la présence signal)
graviers) d’écoulements) d’écoulements) d’écoulements)
Identification Oui Oui Partielle Partielle Partielle
des interfaces (masqués en (uniquement (uniquement
horizontales cas de couche les interfaces interfaces
superficielle superficielles) superficielles)
très résistante)
Identification Oui Oui Oui Oui Partielle
des interfaces (interface
verticales «fondue»)
Détection des Oui Partielle Partielle Oui Partielle
singularités (anomalies (anomalies (ouvrage
déjà connues plurimétriques plurimétriques encastré et
(aqueducs, de fort de très fort aqueduc
résurgences) contraste contraste non vus)
uniquement) uniquement)
Détection de Oui Oui Oui Oui Oui
singularités
nouvelles
ouvrage bien documenté et localisé dans une plaine alluviale où la géologie présente
peu de variabilité, les méthodes géophysiques à grand rendement mises en œuvre
n’ont pas apporté d’informations supplémentaires déterminantes. En outre, les sin-
gularités significatives déjà connues n’étant pas systématiquement identifiées par
la géophysique, le test montre que la granularité de ces méthodes reste adaptée à
l’identification d’anomalies présentant une extension plurimétrique et un contraste
fort (qu’il s’agisse d’un contraste de résistivité ou de vitesse sismique). Néanmoins,
des singularités locales nouvelles ont été détectées, par une ou plusieurs méth-
odes simultanément selon le cas. La réalisation de sondages géotechniques ciblés
doit permettre de statuer sur leur caractère significatif ou non, et ainsi préciser si
la granularité des différentes méthodes cadre avec les attendus de notre étude
(variations de nature, de densité, de degré de saturation).
203
Q.106 - R.14
Tableau 2
Avantages comparés des méthodes de détection et de caractérisation de fuite
Comparative advantages of leakage detection and characterization methods
COÛT GRAND CONTRAINTES EFFICACITÉ DE
RENDEMENT D’EXPLOITATION DÉTECTION
PS en eau 5000 =
C/km ++ +++ ++si maillage adapté et peu
multitrace de perturbation antrhropique
TRE 1500 =
C/km + +++ 0 si résistivité du sol proche
de celle de l’eau
0 si fuite de petite dimension
spatiale
Acoustique 2000 = C/zone 0 0 (vitesse eau faible) +++ si écoulement turbulent
(4h00)
Trac˛age à la 1000 =
C/zone 0 0 (vitesse eau faible) +++
saumure (avec
ou sans TRE)
Le tableau 2, d’avantages comparés, a été établi sur la base des tests et opéra-
tions réalisées sur les canaux de la Durance pour la détection de fuites. Ainsi, les
deux méthodes de détection de fuite (PS et acoustique), retenues pour l’investigation
des canaux de la Durance car ayant des résultats encourageants, ont pour l’heure
encore quelques limites d’utilisation. Elles résident dans les possibles difficultés
d’interprétation des résultats pour la première et le faible rendement nécessitant en
plus un arrêt d’exploitation pour la deuxième. Ainsi, des développements sont en
cours pour lever spécifiquement ces limites. Les premiers résultats sont très encour-
ageants: i) la modélisation numérique du phénomène d’électrofiltration pour la PS
permet d’améliorer grandement l’interprétation des résultats, ii) l’automatisation de
l’acquisition et du traitement du signal acoustique permet de s’affranchir d’un grand
nombre de contraintes.
5. MÉTHODOLOGIE DE FUSION D’INFORMATIONS GÉOPHYSIQUES ET

GÉOTECHNIQUES
5.1. BESOIN DE COMBINAISON ET THÉORIE DES MASSES DE CROYANCE
Comme souligné dans les REX ci-dessus, les méthodes géophysiques à

grand rendement seules ne permettent ni d’identifier avec certitude les zones de
fuites potentielles ni de caractériser la répartition des ensembles lithologiques et la
position des interfaces avec la précision qu’apporterait un sondage géotechnique.
Plusieurs méthodologies reconnues pour les évaluations des digues suggèrent
l’utilisation conjointe de méthodes de reconnaissances géophysiques et géotech-
niques [2–3]. Les méthodes géotechniques sont intrusives et fournissent une
information ponctuelle spatialement, elles sont précises et adaptées à l’analyse des
modes de rupture d’ouvrage. Complémentairement, les méthodes géophysiques
sont non-intrusives et fournissent une information physique sur de larges volumes
204
Q.106 - R.14
de sols, avec d’importants rendements, mais avec des niveaux d’incertitudes poten-
tiellement significatifs. Ces incertitudes sont notamment à imputer aux aspects
intégrants des méthodes ainsi qu’à la résolution de problèmes inverses intrinsèque-
ment mal posés. Un enjeu majeur pour l’évaluation des ouvrages hydrauliques est de
pouvoir combiner les informations acquises par ces deux ensembles de méthodes,
en prenant en considération leurs niveaux d’imperfections (incertitude, imprécision,
inconsistance) et leurs distributions spatiales respectives [3], afin d’aller au-delà
d’une superposition graphique des résultats.
Nous proposons ici une méthodologie de fusion d’informations géophysiques

et géotechniques [4] respectant le cadre fixé par la théorie des masses de croyance
[5–6], reposant sur l’utilisation de deux règles de combinaison. Nous considérons
dans cet article des données acquises sur une digue du canal de la Durance (tronçon
EDF de St-Estève), à partir d’une source d’information géotechnique (sondages
carottés avec essais de granulométrie) et d’une source d’information géophysique
(TRE). Il s’agit de mettre en avant la capacité de la méthodologie de fusion à discrim-
iner les ensembles lithologiques constitutifs de l’ouvrage et la position d’interfaces,
où sont susceptibles de se développer des phénomènes d’érosion interne, tout en
associant un niveau de confiance à la caractérisation proposée.
Le cadre établi par la théorie des masses de croyance requiert: (1) la définition
d’un cadre de discernement (Frame of Discernement – FoD) du problème consid-
éré, (2) la détermination des distributions de masses de croyance pour chaque
source d’information (géotechnique et géophysique) et (3) l’utilisation d’une règle
de combinaison pour fusionner les masses de croyance.
5.2. CADRE DE DISCERNEMENT
Pour cette étude de digue de canal, nous considérons deux hypothèses de

matériaux distincts θ1 et θ2 . Alors que θ1 correspond à des matériaux à grains
fins (résistivité et Vs faibles), θ2 correspond à l’hypothèse de matériaux plus
grossiers (résistivité et Vs plus importantes). Le cadre de discernement, noté
, doit être composé d’hypothèses exhaustives et exclusives, correspondant à
l’ensemble des solutions possibles au problème posé. Nous considérons donc
une troisième hypothèse θ3 . Elle est associée aux valeurs physiques non incluses
dans les gammes de valeurs associées aux deux autres hypothèses. Ainsi, le FoD
s’écrit: Theta = {θ1 , θ2 , θ3 } L’espace des fonctions de masse de croyance, noté 2 ,
est fixé par toutes les disjonctions et par le conflit (noté ∅) existant entre les sources
d’informations [6] tel qu’ici 2 = {Ø, θ1 , θ2 , θ1 ∪ θ2 , θ3 , θ1 ∪ θ2 , θ1 ∪ θ2 , θ1 ∪ θ2 , }. Tout
comme dans un cadre de réflexion probabiliste, la fonction masse de croyance
m attribuée à l’hypothèse A (définie sur 2 ) est définie dans [0, 1]:

m(A) = 1 (1)
A2
205
Q.106 - R.14
Fig. 10
a) Profil de TRE et sondages réalisés sur l’ouvrage, b) coupe longitudinale de
résistivités après inversion.
a) TRE profile and boreholes b) longitudinal section of resistivities after inversion.
Plus m(A) tend vers 1 et plus la confiance en A est importante. La grande

différence avec la théorie probabiliste réside dans le fait que A peut ici représenter
l’union de deux hypothèses. Il est donc possible de représenter l’incertitude et le
manque de connaissances. L’attribution de masses de croyance doit être réalisée
pour chaque source d’information et pour toutes les hypothèses de .
5.3. CAMPAGNE TRE ET MASSES DE CROYANCE ASSOCIÉES
Une campagne TRE a été menée en 2014 sur le tronçon d’étude. A l’issue
d’une première série de mesures (48 électrodes espacées de 5 m), les 24 électrodes
du début du dispositif sont placées à la fin et la mesure est relancée (procédé de
«roll-along») en conservant la même profondeur de pénétration (Fig. 10.a) en config-
uration dipôle-dipôle. Les résultats de l’acquisition géoélectrique, obtenus après un
processus d’inversion (norme L1) à partir d’un modèle initial sans contrainte partic-
ulière, à l’aide du logiciel Res2Dinv, sont représentés Fig. 10.b. Une part importante
de la section est associée à des valeurs de résistivité plutôt faibles (<100 .m).
Cependant, quelques ensembles résistifs (>150 .m) apparaissent localement à
partir de 8 m de profondeur sur les 700 premiers mètres et au-delà de x = 1 600 m.
Une zone plus résistive est à relever en proche surface entre x = 250 et 400 m.
Les positions exactes des interfaces entre ces deux grands ensembles ne peuvent
cependant être déterminées avec précision.
Les intervalles de résistivités sont associés aux hypothèses de à dires

d’experts, en utilisant une représentation des valeurs sous forme de distribution
en classes modales. Les valeurs des classes, en .m, sont définies tel que: [2; 80]
est associé à θ1 , [150; 6, 5.103 ] est associé à θ2 et [5.10−2 ; 2[ ∪ ]6, 5.103 ; 2, 6.105 ] est
associé à θ3 . Afin de modéliser l’incertitude sur le choix des classes et puisqu’il est
difficile d’associer les valeurs de résistivités intermédiaires à θ1 et θ2 à l’une de ces
206
Q.106 - R.14
Fig. 11
a) Hypothèses ayant la plus grande masse de croyance et b) valeurs de masses
associées pour le traitement de l’information géophysique.
a) Hypotheses with the greatest belief mass and b) their associated mass values
for the processing of geophysical information.
deux classes, ]80; 150[ est associé à θ1 ∪ θ2 , suggérant que ces valeurs de résis-
tivités peuvent être rattachées à l’hypothèse θ1 ou θ2 . La méthodologie développée
nécessite de découper la section en mailles avec certaines contraintes liées à la
nature des informations à fusionner ainsi qu’au temps de calcul. Les distributions
de masses de croyance m1 (·) sont fixées pour chaque élément de la section inver-
sée à partir du calcul des distances entre les valeurs de résistivités inversées ±
imprécisions estimées après calcul d’inversion, et les intervalles associés à chaque
hypothèse présentée. Chaque maille est finalement associée à une distribution de
masses de croyance normalisées en respectant Eq. [1].
La Figure 11.a présente les hypothèses de ayant la plus grande masse de

croyance associée. Les valeurs de masses associées à ces hypothèses sont présen-
tées Fig. 11.b et nous renseignent sur le niveau de confiance. Cette représentation
met en évidence la répartition des deux ensembles de matériaux précédemment
décrits ainsi que les zones où la distinction est plus difficile, correspondant à des
valeurs de résistivités intermédiaires (en vert Fig. 11.a).
5.4. ESSAIS DE GRANULOMÉTRIE ET MASSES DE CROYANCE ASSOCIÉES
Une campagne de 7 sondages carottés a été menée en 2016 sur le tronçon

d’étude (Fig. 10.a). Après extraction des carottes, des échantillons sont prélevés
pour mener des essais de granulométrie au tamis (NF P94-056). Une distinction est
faite (NF P11300) entre les matériaux fins (θ1 ) avec présence ou non de gros élé-
ments («matériaux A, C1 A ou C2 A» au sens du Guide des Terrassements Routiers)
et les matériaux sableux à graveleux avec présence de fines (θ2 ) et présence ou
non de gros éléments («matériaux B, C1 B ou C2 B»). Les hypothèses associées
aux caractérisations par les essais de granulométrie sont présentées Fig. 12. Si
aucune analyse granulométrique n’a été effectuée, nous considérons l’attribution
207
Q.106 - R.14
Fig. 12
Position des sondages carottés et des hypothèses du FoD associées aux résultats
d’essais de granulométrie.
Position of the boreholes and FoD hypotheses associated with the particle size
test results.
de la masse de croyance sur l’incertitude totale (θ1 ∪ θ2 ∪ θ3 ) (Fig. 12, en noir). Les
résultats mettent en évidence la présence de matériaux fins, entre B3 et B7 de la
surface jusqu’à z ≈ 8 m, et de matériaux plus grossiers au-delà, dans les sondages
plus profonds (B3, B6 et B7) à l’exception de B6 entre z = 15 et 20 m. B1 et B2
laissent apparaître une granulométrie plus grossière en proche surface.
Afin de caractériser l’ensemble de la section du modèle, une décroissance

latérale exponentielle de la masse de croyance intrasondage (proche de 1), est
imposée à partir du sondage considéré, jusqu’au sondage le plus proche. Le taux
de décroissance est fonction des valeurs proposées par le sondage à proximité.
Ainsi, pour chaque maille de la section (de dimension 0,1 × 5,32 m, de même largeur
que les mailles du traitement géophysique), une nouvelle distribution de masses
de croyance m2 (·) est générée, entrant dans le processus de fusion. Étant donné
que B1 et B7 sont situés à l’extérieur de la section de tomographie (Fig. 12), ils
n’apparaîtront pas sur les résultats de la fusion. Cependant, les informations qu’ils
fournissent permettent de quantifier le taux de décroissance de la confiance en
l’information fournie par B2 et B6.
Pour le traitement géotechnique, la représentation (Fig. 13) met en évidence

le haut niveau de confiance placé au niveau des sondages et son extension en
fonction de l’information contenue dans les sondages voisins
5.5. RÉSULTATS DE FUSION ET DISCUSSION
La procédure de fusion consiste à combiner les distributions de masses de

croyance m1 (·) et m2 (·) affectées à chaque maille de la grille. Nous utilisons ici deux
208
Q.106 - R.14
Fig. 13
a) Hypothèses ayant la plus grande masse de croyance et b) valeurs de masses
associées pour le traitement de l’information géotechnique.
a) Hypotheses with the greatest belief mass and b) their associated values for the
processing of geotechnical information.
règles de combinaison: i) la règle dite de Smets [7], ii) la règle PCR6 (Proportional
Conflict Redistribution rule n◦ 6) [8]. L’hypothèse faite par Smets permet l’attribution
d’une masse de croyance au conflit, en-dehors de , tel que m12 (Ø) > 0. Avec m12
la masse de croyance fusionnée à partir des informations des sources 1 et 2. Il est
donc possible de quantifier le niveau d’ignorance de résolution du problème en plus
des masses allouées aux hypothèses de :

m12 (A) = m1 (X )m2 (Y ) (2)
X ,Y ⊆|X ∩Y =A
Avec ici m1 (X) et m2 (Y) les masses de croyance respectivement attribuées

aux hypothèses X et Y par les sources 1 et 2. La combinaison de PCR6 permet
en revanche la redistribution de tous les conflits partiels, proportionnellement aux
masses concernées par ces conflits tel que et m12 PCR6 (∅) = 0 et ∀ A ∈ 2 \{∅}:

m1 (A)2 m2 (Y ) m2 (A)2 m1 (Y )
12 (A) = m12 (A) +
mPCR6 + (3)
m1 (A) + m2 (Y ) m2 (A) + m1 (Y )
Y ∈ 2Theta
A ∩ Y =∅
Les résultats issus de la fusion d’informations de deux sources sont visi-

bles Fig. 14. L’intérêt d’une telle méthodologie apparaît lorsque les sources sont
en désaccord (présence de conflit, Fig. 14.a) mais également lorsque l’information
concorde (présence de masses de croyance plus grandes, Fig. 14.b et 14.d).
209
Q.106 - R.14
Fig. 14
(a, c) Hypothèses ayant la plus grande masse de croyance et (b, d) leurs valeurs
de masses associées après fusion. Fig a et b sont obtenues par combinaison de
Smets, Fig c et d sont obtenues par combinaison PCR6.
(a, c) Hypotheses with the greatest belief mass and (b, d) their associated values
after fusion. Fig a and b are obtained by combination of Smets, Fig c and d are
obtained by PCR6 combination.
En raison des informations localement contradictoires fournies par la carac-

térisation géoélectrique et les analyses granulométriques, le conflit prend une place
importante dans les résultats obtenus par règle de Smets, entre 6 et 10 m de pro-
fondeur entre B3 et B4 et vers z = 10 m au niveau de B6 (Fig. 13.a). La règle PCR6
(Fig. 14.c et 14.d) autorise cependant la répartition de cette masse conflictuelle sur
les autres hypothèses de . Les masses de croyances représentées (Fig. 14.b et
14.d) mettent en évidence le fait que la caractérisation des matériaux proposés est
moins fiable à proximité des interfaces entre les deux ensembles lithologiques. Si
l’investigation sur site devait être renforcée dans le cadre de futures campagnes
de reconnaissance, de tels résultats suggèrent un besoin d’informations supplé-
mentaires où les masses de croyance sont les plus faibles mais également là où la
caractérisation de θ1 ∪ θ2 est majoritaire.
Cette méthodologie permet la prise en compte des spécificités de chaque

méthode d’investigation en termes de niveau d’imperfection et de répartition spa-
tiale de l’information, pour représenter une section de digue de canal. Les résultats,
obtenus à partir de données de TRE et d’analyses granulométriques, mettent en
évidence la distribution de deux grands ensembles lithologiques et les positions
d’interfaces de façon plus lisible que les résultats d’inversion seuls. Ce travail per-
met aussi de mettre en avant les zones de conflit et celles de moindre confiance,
suggérant où les investigations doivent être renforcées.
210
Q.106 - R.14
La méthode de fusion est intéressante pour étendre spatialement de

l’information locale précise d’une propriété d’intérêt à partir d’information spatialisée
d’une autre propriété non directement liée. Toutefois, l’extension de cette méth-
ode implique deux axes d’amélioration en amont du traitement géophysique: i)
la caractérisation de l’incertitude et ii) la définition des classes physiques asso-
ciées aux hypothèses. Dans de futures recherches, il serait pertinent d’avoir à
disposition des données piézométriques concomitantes aux campagnes de recon-
naissance afin de prendre en considération les effets de hauteurs de nappe. Dans
le cadre d’un monitoring géoélectrique sur du long terme, il pourrait être envisagé
d’automatiser la procédure de fusion afin d’obtenir des résultats réguliers permettant
de suivre l’évolution de l’ouvrage. Il serait enfin intéressant de prendre en consid-
ération un plus grand nombre de méthodes géophysiques et géotechniques et de
plus nombreuses classes de matériaux puisque le cadre théorique des fonctions de
masses de croyance ne limite ni le nombre de sources d’information ni le nombre
d’hypothèses constitutives de .
6. CONCLUSION
Une vidange totale ou partielle d’un canal est préjudiciable pour son étanchéité
et ne se justifie pas pour la recherche de points de faiblesse dans ses remblais. En
effet, la plupart d’entre eux peuvent être recherchés et identifiés, canal en eau,
par mise en œuvre d’une surveillance et d’investigations adaptées. Le déploiement
de reconnaissances géophysiques sur les remblais constitue une solution d’avenir
dans la mesure où elles permettent i) une meilleure connaissance des zones de
faiblesse dont celles de fuite, ii) la garantie de l’intégrité de l’ouvrage en évitant une
vidange et iii) une minimisation des pertes d’exploitation (plus d’arrêt des usines
lors des vidanges). Si ces investigations géophysiques sont prometteuses, il existe
encore des limites d’utilisation que des développements pourront repousser. Enfin,
dans l’objectif de quantifier le niveau de sûreté des digues, des reconnaissances
géotechniques sont indispensables pour objectiver les paramètres géotechniques
des matériaux. Celles-ci sont ponctuelles pour des digues de plusieurs de kilo-
mètres. Ainsi, les méthodes de fusion de données permettent une amélioration de
la caractérisation de ces ensembles lithologiques.
REMERCIEMENTS
Nous tenons à remercier Sérgio Palma-Lopes (UGE), Yannick Fargier (UGE),

Vincent Guihard (EDF), Isabelle Penot (EDF), Jean-Christophe Gourry (BRGM),
Stéphane Pligersdorffer (GEOSCAN), Rachid Fellag (CNR) et Alexandre Tourette
(EGIS)
211
Q.106 - R.14
RÉFÉRENCES
[1] DAUMAS B., COURIVAUD JR., PENOT I., BILGER C., AIGOUY S. Diagnos-
tic Exhaustif (Ex. Examen Technique Complet) d’ouvrages de grand linéaire
(canaux) sans vidange. Colloque CFBR, 2018.
[2] DEZERT, T., FARGIER, Y., LOPES, S. P., & COTE, P. Geophysical and
geotechnical methods for fluvial levee investigation: A review. Engineering
Geology, 105206, 2019.
[3] ROYET, P., PALMA LOPES, S., FAUCHARD, C., MERIAUX, P., and AURIAU,
L. Rapid and cost-effective dike condition assessment methods: geophysics
and remote sensing. FloodProBE Project, 2013.
[4] DEZERT, T. Combinaison d’informations ponctuelles et volumiques pour le

diagnostic d’ouvrages en terre soumis à des risques hydrauliques (Doctoral
dissertation, Université de Nantes-Faculté des Sciences et Techniques), 2019.
[5] DEMPSTER A. P. Upper and lower probabilities induced by a multivalued

mapping. The annals of mathematical statistics, 325–339, 1967.
[6] SHAFER G. A mathematical theory of evidence. Princeton University Press,

1976.
[7] SMETS, P. The combination of evidence in the transferable belief model. IEEE
Transactions on pattern analysis and machine intelligence, 12(5):447–458,
1990.
[8] MARTIN, A. and OSSWALD, C. A new generalization of the proportional con-

flict redistribution rule stable in terms of decision. Advances and applications
of DSmT for information fusion: collected works Volume 2, 2:69–88, 2006.
212
Q.106 - R.15
-------
-------
UTILISATION DES DONNÉES HAUTE RESOLUTION PAR DRONES

POUR LA SURVEILLANCE DES OUVRAGES HYDRAULIQUES:
DE L’ACQUISITION AU TRAITEMENT (∗)
Guillaume TERRASSE
Ingénieur-chercheur, EDF – Recherche et Développement
Paul-Henri FAURE
Référent Topographie & Géopositionnement, Compagnie Nationale du Rhône
Eric VUILLERMET
Expert en Sûreté des barrages, BRL Ingénierie
Yann GAYET
Gérant et Directeur technique, GEOKALI
FRANCE
RÉSUMÉ
Dans ce rapport, nous décrivons des moyens rapides d’auscultation des sur-
faces d’ouvrages hydrauliques à partir de méthodes de numérisation 3D haute
résolution intégrant des capteurs montés sur drone. Ces technologies permettent
de suivre plus exhaustivement l’évolution de l’ouvrage que les mesures ponctuelles
classiques. De plus, elles permettent de compléter les visites terrain avec un contrôle
quantitatif des mouvements. Nous présentons deux méthodologies de numérisation
3D à partir de drone par mesures LiDAR et photogrammétriques. En particulier,
nous décrivons, au travers trois cas, l’utilisation respective de ces deux technolo-
gies pour la surveillance des ouvrages ainsi qu’une technologie hybride combinant
∗ Use of high-resolution data acquired by UAV for the monitoring of hydraulic structures:
From acquisition to processing.
DOI: 10.1201/9781003211808-111 213

Q.106 - R.15
LiDAR et photogrammétrie. Ces trois cas nous permettent de décrire les méth-
odes d’acquisition mise en œuvre où notamment le géoréférencement joue un
rôle prépondérant. Enfin, nous détaillons des exemples d’exploitation des données.
Nous montrons alors que les acquisitions mises en œuvre permettent de suivre
des mouvements centimétriques sur les ouvrages. Néanmoins, ces méthodes ont
également des limites que nous présentons sur la base du retour d’expérience des
numérisations. Si les progrès technologiques sur les capteurs et les algorithmes
permettent de suivre les ouvrages avec une plus grande facilité de mise en œuvre
et une meilleure précision, les acquisitions drones sont sujettes aux conditions
météorologiques et à la réglementation pour les autorisations de vols, ce qui limite,
pour l’instant, leur utilisation opérationnelle.
SUMMARY
In this report, we describe fast surface inspecting technologies of hydraulic

structures using high-resolution 3D scanning methods mounted on UAVs. These
technologies allow a more precisely monitoring of the evolution of the structure
than classical point measurements. In addition, they complete the field visits with a
quantitative control of the movements. We present two methodologies for 3D digiti-
zation from UAVs using LiDAR and photogrammetric measurements. In particular,
we describe, through three cases, the respective use of these two technologies for
the monitoring of structures as well as a hybrid technology combining LiDAR and
photogrammetry. These three cases allow us to describe the acquisition methods
used, in which georeferencing plays a predominant role. Finally, we detail examples
of data exploitation. We then show that the implemented acquisitions offer a centi-
metric movements monitoring on the structures. Nevertheless, these methods also
have limitations, which we present from the digitalization feedbacks. While techno-
logical improvement in sensors and algorithms make it possible to track structures
with greater ease of use and better accuracy, UAV acquisitions are subject to weather
conditions and flight authorization regulations, which limit their operational use.
1. INTRODUCTION
Dès leur construction, les ouvrages hydrauliques font l’objet d’une surveillance
régulière de la part des concessionnaires. Cette surveillance se décline selon trois
piliers: l’inspection visuelle, l’auscultation et les contrôles et essais. L’inspection
visuelle est essentiellement réalisée par des visites terrain à pied. L’auscultation est
assurée par des dispositifs d’auscultations topographiques, piézométriques ou de
mesure de débits de fuite. Deux types de surveillance peuvent être distingués: la
surveillance standard réalisée selon les consignes de surveillance et la surveillance
renforcée déployée sur des zones bien spécifiques nécessitant des fréquences ou
méthodes adaptées.
214
Q.106 - R.15
Sur l’ensemble du parc des ouvrages suivis par les 3 entreprises (CNR, EDF et
BRLi), les ouvrages hydrauliques en remblai peuvent se regrouper en 3 catégories
principales: les barrages, les digues fluviales et les digues de canaux.
La surveillance a pour objectif de déceler la défaillance de composants con-

stitutifs de l’ouvrage soit en détectant des signes précurseurs, soit en qualifiant
l’évolution d’une dégradation. A ce titre, l’érosion interne des ouvrages en remblai
est une des principales causes de rupture des digues et barrages en remblai. Lors
du processus d’érosion, un affaissement apparait en surface de la digue. Celui-ci
peut prendre 2 formes: une zone de tassement précurseur ou directement un fontis.
La détection de ces désordres est une alerte pour les exploitants.
La surveillance concerne aussi les protections contre l’érosion externe des

parements des ouvrages soit pour s’assurer de la bonne efficacité des protections
en place, soit en évaluant l’évolution des désordres en effectuant des mesures de
déformations.
Depuis une dizaine d’années, les technologies de mesures topographiques

sans contact à haute résolution se développent et sont utilisées pour compléter
les méthodes traditionnelles [1]. Les dernières années ont vu le développement de
nouveaux capteurs et nouvelles méthodologies d’acquisition permettant dorénavant
d’envisager l’utilisation à grande échelle de ce type de systèmes.
A travers 3 cas, EDF, CNR et BRLi présentent les dernières avancées

méthodologiques et opérationnelles permettant l’utilisation des technologies LiDAR
(Light Detection And Ranging) et photogrammétriques notamment portées par
drone. Un focus sera fait sur les derniers travaux d’automatisation des traitements
pour la détection de désordres. Pour terminer, une dernière partie fera le bilan des
acquisitions et de l’exploitation des données par les trois entreprises.
2. ÉTAT DE L’ART
2.1. MÉTHODES CONVENTIONNELLES D’AUSCULTATION
Historiquement, l’auscultation des ouvrages en remblais est réalisée à l’aide

de mesures ponctuelles: topographie classique (nivellement le plus souvent), pié-
zométrie et plus rarement de mesures aux tassomètres et inclinomètres, associées
à des mesures intégratrices telle que les mesures de débit de fuite ([5] et [6]). Il con-
vient de noter le déploiement depuis quelques années des systèmes d’auscultation
et surveillance par fibre optique qui permettent, dans des conditions ad-hoc et sur
des endiguements spécifiques, de détecter des signes précurseurs d’écoulements
anormaux dans le corps de digue. Les principales limites des méthodes actuelles
sont la densité des points de mesure qui ne permettent pas une vision et une
auscultation complète de l’ouvrage.
215
Q.106 - R.15
Tableau 1
Liste des méthodes conventionnelles
d’auscultation et leur composante respective mesurée
MÉTHODE COMPOSANTE MESURÉE
Piézométrie Charge hydraulique
Nivellement/tassomètre Tassement absolu/différentiel
Inclinomètre Déformation différentielle du corps de digue
Mesure de débit Débits de percolation et de drainage de l’ouvrage
2.2. L’INSPECTION VISUELLE
Plusieurs niveaux d’inspections sont mis en œuvre lors de ces visites: les
tournées hebdomadaires de l’exploitant, les Visites Annuelles de l’Exploitant (VAE)
et les Visites Techniques Approfondies (VTA). Respectivement tous les 10 ans et
15 ans (en France), les ouvrages de classe A et de classe B sont aussi soumis aux
Diagnostics Exhaustifs de l’Ouvrage (DEO) qui intègrent une forte composante liée
à l’inspection de l’ensemble de l’ouvrage (parties immergées et émergées). En règle
générale les tournées hebdomadaires sont réalisées par moyen motorisé et les VTA
et VAE à pied de manière à voir l’ensemble de l’ouvrage (du pied à la crête). Ces
inspections permettent d’avoir une information qualitative sur l’état de l’ouvrage.
Actuellement, ces visites consistent en la prise de photographies associées à des
commentaires. Elles permettent ainsi d’avoir une vision de l’état des parements,
des protections ou de la végétation. Ces inspections ne concernent que la surface
et sont souvent sujettes à l’expérience ou l’acuité de la personne. Cependant les
déformations structurelles ne sont souvent visibles que lorsque l’opérateur constate
un désordre (présence d’un fontis). En effet, visuellement, sans moyens de mesure
et de comparaison, il peut être extrêmement difficile de distinguer un affaissement
sur un ouvrage.
2.3. DE L’INTÉRÊT DES MÉTHODES SURFACIQUES
Actuellement, dans le cas d’une surveillance standard et selon le type

d’ouvrage, la densité et les dispositifs conventionnels sont suffisants. Cependant,
dans les cas de surveillance renforcée ou sur certains ouvrages sensibles, des
moyens complémentaires sont souvent nécessaires afin de pallier aux limites des
méthodes classiques.
Parmi les méthodes surfaciques, nous nous intéressons aux acquisitions

Haute Résolution de données topographiques. Ces méthodes couvrent à la fois
l’acquisition de nuages de points 3D et les photographies Hautes Définition. Ces
technologies permettent d’avoir une vision globale et exhaustive de l’ensemble de la
zone auscultée. Elles présentent l’avantage de fournir un relevé intégral de l’ouvrage
avec une plus grande densité d’information, particulièrement intéressante dans le
cas de structure non monolithique. Elles permettent maintenant de détecter des
mouvements ou des changements de la structure de l’ouvrage non nécessairement
216
Q.106 - R.15
Fig. 1
Illustration d’un nuage de points acquis sur une digue en
remblai avec une haute résolution par drone LiDAR
Illustration of a high resolution point cloud of an
embankment dike acquired from LiDAR UAV.
visibles à l’œil nu. Ainsi, avec une résolution et une précision adaptée aux défauts
recherchés, il sera possible de détecter des zones de tassements (inférieurs à 5
cm) peu importants mais étendus ou des désordres ponctuels. De plus, il est rare
que ces défauts apparaissent au droit des dispositifs d’auscultation ponctuels. De la
même maniére, il sera plus aisé de suivre l’évolution d’une protection en enroche-
ments via une comparaison d’un Modèle Numérique de Terrain1 (MNT) ou nuages
de points2 (Fig. 1) ou d’orthophotographies.
De la même manière, il sera plus aisé de suivre l’évolution d’une protection

en enrochements via une comparaison d’un Modèle Numérique de Terrain3 (MNT)
ou nuages de points4 (Fig. 1) ou d’orthophotographies.
2.4. RAPPELS UTILES
La numérisation de la surface d’un ouvrage nécessite un système à 3

composants principaux:
• Un porteur qui embarque les systèmes de mesures (par exemple: drone,

hélicoptère, voiture, bateau, humain)
• Un système de numérisation de la surface de l’ouvrage: caméra numérique
ou capteur LiDAR
• Un système de positionnement: pour associer à chaque image ou point du
nuage une coordonnée dans un référentiel global partagé entre les différentes
1 Maillage 3D de la surface du sol soustrait des éléments en surface (par exemple : la

végétation)
2 Un nuage de point est constitué d’un ensemble de positions (x,y,z) auxquelles sont associés
des attributs (intensité de la réflexion enregistrée, une information colorimétrique RGB, …)

3 Maillage 3D de la surface du sol soustrait des éléments en surface (par exemple: la
végétation)
4 Un nuage de point est constitué d’un ensemble de positions (x,y,z) auxquelles sont associés
des attributs (intensité de la réflexion enregistrée, une information colorimétrique RGB, …)
217
Q.106 - R.15
acquisitions. Le positionnement est nécessaire pour la comparaison entre

deux acquisitions. Ce dernier est généralement assuré par la combinaison
d’un système GNSS et d’une centrale inertielle.
Depuis quelques années, les développements techniques et algorithmiques

ont permis la miniaturisation des capteurs et l’optimisation des chaines de
traitements. Ces avancées permettent maintenant de déployer des systèmes
performants sur des porteurs légers (type drone).
3. MÉTHODOLOGIES D’ACQUISITION
Cette partie se propose de présenter à travers 3 cas différents plusieurs

méthodologies d’acquisition de données topographiques haute résolution. Les
processus de traitement seront abordés dans la 3ème partie de l’article.
3.1. DÉTECTION DE FONTIS PAR ACQUISITION LIDAR SUR DRONE

MISE AU POINT PAR EDF
Cette partie a pour but d’exposer la méthode d’acquisition par drone LiDAR
choisie pour la détection de fontis à la surface de digue en remblai avec des affaisse-
ments centimétriques. Nous présentons également la méthodologie expérimentale
mise en place pour évaluer les performances des traitements algorithmiques (§4.1).
Le choix du mode de numérisation s’est porté sur un système d’acquisition

LiDAR monté sur drone. En effet, les drones offrent la possibilité d’acquérir
d’importants linéaires de digue à moindre coût. La comparaison d’acquisitions
espacées dans le temps permet alors de localiser l’apparition de déformations. Afin
de se placer dans un contexte industrialisable, nous avons décidé de privilégier
l’utilisation d’un drone à voilure fixe associé à un capteur LiDAR qui assure, pour le
cas des digues, une meilleure répétabilité des mesures ainsi qu’une meilleure pré-
cision altimétrique. Le drone à voilure fixe choisi est le DT26X de la société Delair
équipé du LIDAR RIEGL VUX-1 UAV permettant de couvrir de plus longs linéaires
qu’un drone à voilure tournante. Le drone est également équipé d’un système de
positionnement GNNS et d’une centrale inertielle Applanix APX15. De plus, un GPS
différentiel est placé au sol.
Afin d’étudier la faisabilité de la méthode, deux dispositifs de trappe ont été

réalisés pour la simulation de 10 paliers de déformation allant de 0 à 23 cm et mis en
place lors d’une expérimentation en 2017 sur la digue du canal de Sisteron (EDF).
Les données collectées sont utilisées pour évaluer les traitements de détection de
fontis proposés (§4.1).
L’acquisition a été réalisée sur un linéaire de digue de 1.5 km. Pour chaque
niveau de la trappe, un nuage de points est acquis simulant ainsi des numérisations
218
Q.106 - R.15
Fig. 2
Illustration des acquisitions LiDAR drone sur la digue en remblai du canal de
Sisteron. A gauche: le lancement du drone DT26X à l’aide d’une catapulte.
A droite: l’installation du dispositif de trappe pour la simulation physique de fontis.
Illustration of LiDAR UAV acquisitions on the embankment dike of the Sisteron
canal. Left: launching of the DT26X UAV with a catapult. Right: the installation of
the trapdoor device for the physical simulation of distortion.
espacées dans le temps où se forme un affaissement progressif d’un fontis. Pour

chaque niveau de trappe, le drone a réalisé les acquisitions à une altitude de vol
de 80 m et en effectuant deux allers-retours afin de garantir une densité de points
suffisante (200 pts/m2 ). Sa vitesse moyenne de déplacement était d’environ 10 m/s
dans la phase aller et 25 m/s dans la phase retour. Un nuage contient entre 23 et
30 millions de points5 par acquisition.
Les données sont ensuite post-traitées en corrigeant la trajectoire du drone à

partir de l’exploitation des données du couple centrale inertielle/GNSS, des mesures
GPS au sol ainsi que des points de calage. L’association des coordonnées de
points de calage au nuage de points permet de corriger la trajectoire du drone et
in fine l’exactitude du nuage de points général. Nous avons également réalisé des
recalages relatifs entre les différentes bandes d’acquisition. Les nuages de points
présentent un bruit dans les données de +/−3 cm (1 σ). Ce bruit résulte des incer-
titudes de mesure liées au système d’acquisition, de la topographie sur terrain ainsi
que de la végétation basse.
En plus des simulations physiques de fontis par deux dispositifs de trappes,

nous avons simulé numériquement des fontis afin d’avoir une base d’exemples
plus importante pour l’évaluation. Ainsi 22 simulations numériques de fontis ont
été ajoutées en faisant varier:
• les diamètres de 20 à 60 cm
• les affaissements de 3 cm, 5 cm et 7 cm
• les 3 zones différentes de la digue, ce qui nous permettra d’évaluer la
robustesse de la méthode en fonction de la position de la déformation de la
5La variation du nombre de points est directement liée la variation de vitesse de vol causée
par la force du vent
219
Q.106 - R.15
digue dont le nuage de points est susceptible d’avoir une densité et un niveau
de bruit différent.
La Figure 2 compare qualitativement un fontis simulé numériquement et

physiquement par une trappe. On constate visuellement que les deux simulations
sont similaires.
Fig. 3
Simulation numérique (à gauche) et physique (à droite) de fontis
d’un rayon de 40 cm avec un affaissement de 7 cm dans un nuage de points.
Les points de fontis apparaissent ici en rouge. En haut: une illustration des
simulations en vue de dessus et en bas en vue de coupe verticale.
Numerical (left) and physical (right) simulation of distorsion in a point cloud.
The distorsion points are shown in red. Above: illustration of the simulations in
top view. Below: illustrations in vertical section view.
3.2. PHOTOGRAMMÉTRIE NUMÉRIQUE HAUTE PRÉCISION SUR OUVRAGES

LINÉAIRES DÉVELOPPÉE PAR LA CNR ET IGN
3.2.1. Contexte
Le contexte des mesures photogrammétriques de précision réalisées par la

CNR et l’IGN (Institut National de l’Information Géographique et Forestière) est
rappelé par [2]. Dans le cadre de sa surveillance, la CNR a défini deux besoins:
des acquisitions récurrentes et des acquisitions exceptionnelles.
220
Q.106 - R.15
Les acquisitions récurrentes concernent l’auscultation régulière de zones

ciblées et localisées (de l’ordre du km) dans le cadre de suivis spécifiques. Pour
ce type d’acquisition la précision escomptée est de ±1 cm afin de pouvoir déceler
des déformations “centimétriques” (±2 cm).
Les acquisitions exceptionnelles concernent des situations de crise où la CNR

souhaite obtenir un état des lieux de linéaires importants (plusieurs dizaines de km).
Dans ce cadre, la CNR souhaite avoir à disposition un outil rapide dans la mise en
place et dans le traitement des données (séisme, accès perturbés etc.). Sur ce type
d’acquisition les précisions escomptées sont de l’ordre de ±3 cm.
Le principe retenu est ensuite d’effectuer des comparaisons de MNT (Modèle

Numérique de Terrain) entre 2 campagnes successives.
3.2.2. Les travaux menés entre 2012 et 2019
La caractéristique linéaire des ouvrages CNR engendre des configurations ter-

rain spécifiques peu propices à la réalisation d’acquisitions de qualité. En effet, pour
des linéaires dépassant les 200 mètres, les géométries des vols engendrent des
problématiques d’accumulation d’erreurs nécessitant un grand nombre de points de
référence au sol pour caler les modèles photogrammétriques.
Une première thèse réalisée entre 2012 et 2015 [3], a permis d’améliorer la
compréhension et la correction des dérives observées en permettant d’atteindre
les précisions souhaitées avec un point de contrôle par portion de 100 m linéaire.
Entre 2016 et 2019, une deuxième thèse a été réalisée [4] afin d’améliorer encore
le processus. A l’issue de cette étude, il est désormais possible de réaliser des
acquisitions centimétriques avec un point de calage au sol, en utilisant des caméras
photogrammétriques légères synchronisées avec un récepteur GNSS embarqué.
3.2.3. Les stratégies d’acquisition
• Utilisation d’une caméra métrique Global-Shutter

La qualité du capteur photographique est le prérequis indispensable
pour réaliser des acquisitions de qualité. Pour une caméra Global-Shutter,
l’ensemble de la matrice du capteur réalise l’acquisition de manière simul-
tanée, contrairement aux capteurs Rolling-Shutter où l’acquisition se fait ligne
par ligne ce qui génère une déformation de l’image lorsque la caméra est
mobile.
• GNSS Low-cost (géocube de l’IGN) sur le porteur
L’utilisation d’un récepteur GNSS Low-Cost (mono fréquence), parfaitement
synchronisé avec la caméra et couplé à une station GPS au sol connue, est
suffisante pour réaliser un géoréférencement absolu (WGS84) à ±4 cm. A
221
Q.106 - R.15
Fig. 4
Illustration des bras de levier entre la caméra et l’antenne GNSS – crédit Yilin Zhou
Illustration of the lever arms between the camera and the GNSS antenna.
l’aide d’un point de référence au sol, il sera possible de descendre cette

précision à 1 cm.
• Des axes de vols spécifiques
Les axes de vols photogrammétriques doivent permettre de réaliser des
acquisitions nadirale6 et obliques à une hauteur de l’ordre de 50 m et avec
des recouvrements longitudinaux de 80% et transversaux de 70%.
• Connaissance parfaite de la calibration du couple Caméra-GNSS
Cette calibration consiste en la connaissance des paramètres intrinsèques
de la caméra (Calibration Camera) et des valeurs des bras de levier entre
le centre de phase de l’antenne GNSS et du point focal de la caméra. Si la
calibration est connue et calculée avant le vol, un seul vol nadir à 50 m de
hauteur permettra d’atteindre une précision de 4 cm en altitude sans point au
sol. En l’absence d’une pré-calibration il sera nécessaire de réaliser une série
de vols nadir et obliques à 2 hauteurs différentes afin de calculer, in situ, ces
paramètres.
A noter que pour relever un linéaire de 1 600 m avec une précision de ±1 cm,
une journée est suffisante pour réaliser les opérations de terrain en prenant en
compte les aléas. Les matériels utilisés lors de ces deux thèses sont des prototypes
développés par le laboratoire LOEMI de l’IGN. Il n’existait pas, au début des travaux,
de solutions commerciales répondant au cahier des charges métrologique (caméra
6 Nadir: direction opposée au zénith
222
Q.106 - R.15
légère global shutter et récepteur GNSS parfaitement synchronisé embarquables

sur drone).
3.2.4. Les calculs
Le calcul photogrammétrique validé lors des 2 thèses est réalisé par la

suite avec le logiciel Open-Source MicMac développée par le Laboratoire Matis
de l’IGN. L’ensemble des résultats des études a été intégré dans MicMac et est
donc disponible pour l’ensemble de la communauté.
3.3. SUIVI D’UN PAREMENT EN ENROCHEMENTS DE BARRAGE PAR MODÉLISATION

3D HYBRIDE MIS AU POINT PAR BRLI/GEOKALI
Les protections en enrochements amont des barrages en remblai sont un

composant important dans la protection des talus amont sollicités par le batillage et
le déferlement des vagues. Au cours du temps, ces protections peuvent présenter
des désordres et conduire à des défaillances. Les origines peuvent être multiples:
sous dimensionnement des blocs, sous-estimation des sollicitations, évolution de la
blocométrie par altération (mécanique, gélifraction,..). Les désordres observés sont
de natures diverses: ségrégation, enlèvement, et déplacement de blocs, dégrada-
tion des blocs et glissement de talus…Si l’inspection visuelle des parties émergées
(zone de marnage) permet de qualifier l’apparition de ces désordres, il est apparu
que le suivi quantitatif de leur évolution largement supérieure au centimètre restait
un exercice difficile. L’avènement de porteur “léger” a mis à disposition des moyens
de collecte de données faciles à mettre en œuvre, mais il restait à définir un pro-
tocole d’acquisition et à choisir une méthode à déployer intégrant une garantie de
répétabilité permettant un recalage à coût optimal.
La technologie ayant évolué, un levé hybride laser-photogrammétrique du

parement amont (laser statique + acquisition photo par un drone) a été identifié
comme pouvant être le vecteur de collecte des données. Restait à développer et
tester le protocole d’acquisition.
• Le matériel déployé:
Le choix du capteur a déterminé le choix du porteur. Un capteur photo plein format

de 43 Mpx, équipé d’une focale fixe de 25 mm a été installé sur un drone hexacop-
tère conçu et fabriqué par la société Escadrone. Ce drone lourd (11 kg, batteries
comprises) a été sélectionné pour porter une charge utile de plus d’1 kg tout en
assurant une bonne stabilité au vent, et une autonomie de 20 minutes.
Le scanner laser terrestre utilisé est de type statique à mesure de phase, à

très haute densité (1 million de points/s) et équipé d’une caméra pour la colorisation
223
Q.106 - R.15
Fig. 5
Partie émergée du parement en enrochements objet du levé
Riprap slope, object of the survey.
RGB des points. Le nuage de points fournit une contrainte géométrique forte lors
du calcul photogrammétrique (plusieurs millions de points d’appui), en substitution
de points de calage ponctuels, dès la phase de calcul de la géométrie interne et
externe des images: position, orientation, distorsion, focale, échelle. La corrélation
entre le nuage de points lasergrammétrique et les images est automatisée.
La position des images n’est pas mesurée par des capteurs en temps réel, elle
n’est déterminée qu’a posteriori lors du traitement des données. Le géoréférence-
ment est donc établi à partir du nuage de points laser uniquement. Celui-ci est recalé
dans un système de coordonnées local.
L’expérience a été menée sur le barrage de Laprade (France). Le protocole

testé en mode expérimental a été le suivant:
• exécution d’un premier levé du parement;

• déplacements artificiels de blocs en deux groupes:
◦ une série définie par le bureau d’étude sans informer le prestataire de
la localisation et de la typologie des déplacements;
◦ une série définie par le prestataire;
• exécution d’un second levé du parement une semaine après le premier levé.
Le protocole expérimental avait pour objectif multiple de:
• réaliser un levé photogrammétrique initial du parement amont en enroche-

ments,
• produire une modélisation 3D du parement pour permettre un examen
exhaustif du parement,
224
Q.106 - R.15
Fig. 6
Vue zénithale du nuage de points lasergrammétrique.
Les points rouges indiquent la position des stations de scans.
Overhead view of the 3D laser point cloud.
Red dots represent the scan station position.
• valider le protocole de suivi et de détection de l’évolution des déplacements

de blocs permettant d’en analyser la gravité.
• Acquisition:
Un relevé terrestre au laser statique de l’ouvrage est d’abord réalisé. Plusieurs

stations successives sur le couronnement sont nécessaires pour fournir une descrip-
tion géométrique extensive de l’ouvrage. Il est à préciser néanmoins qu’il n’est pas
nécessaire que le levé laser soit exhaustif. Les parties invisibles au scanner laser
seront complétées par la photogrammétrie.
Le levé laser a été réalisé avec 7 stations chacune étant géoréférencée dans
un système local par implantations de cibles, relevées à la station totale depuis une
station unique (au centre du couronnement).
Le levé par drone est réalisé avec un objectif 25 mm à 15 m de distance du

parement pour assurer une résolution de 3 mm/px. Cette résolution a une incidence
déterminante pour la précision des mesures. Les paramètres de prise de vue, de
recouvrement, de contrôle des données doivent être maîtrisés:
• L’appareil photo est paramétré afin de fournir le meilleur compromis piqué de

l’optique/profondeur de champ/sensibilité/minimisation du bruit numérique.
• Les recouvrements entre axes de vols sont, a minima, de 80% (recouvrement
longitudinal) et de 65% (recouvrement latéral).
Le relevé par drone est effectué en vol automatique, le plan du capteur

alternativement parallèle au parement et au nadir, sur 5 axes, pour un total
de 10 survols du parement. La vitesse de vol est déterminée pour générer un
minimum de flou de filé (inférieur à 1/3 de pixel). La totalité de l’intervention
(scanner + topo + photogrammétrie) sur un linéaire de 500 m aura duré 1 jour, pour
deux opérateurs.
225
Q.106 - R.15
4. TRAITEMENT ET EXPLOITATION DES DONNÉES
4.1. DÉTECTION AUTOMATIQUE DE FONTIS
Plusieurs facteurs caractéristiques de la qualité du nuage de points sont sus-

ceptibles de limiter la détection de fontis. Lors du choix d’une méthode de détection
de fontis, il est nécessaire de la confronter à ces facteurs d’influence afin d’en estimer
sa robustesse. Ces facteurs sont:
• La densité du nuage de points: celle-ci est dépendante du système

d’acquisition LiDAR (fréquence d’acquisition de points), de l’altitude de vol,
du nombre d’allers-retours effectué et de la vitesse de déplacement du drone.
Cette dernière peut notamment grandement varier en fonction de la force du
vent.
• Le bruit d’acquisition: cette incertitude est directement liée au système
d’acquisition.
• La surface acquise: la topographie ainsi que la végétation ont une influence
sur l’écho enregistré par le système. Ces dernières introduisent un bruit dans
les données. Une surface plane avec une végétation basse présente un bruit
moins important.
• Les incertitudes de géoréférencement: les incertitudes sur le positionnement
du système couplé aux incertitudes du recalage des acquisitions entre elles
introduisent un biais lors de la comparaison.
Pour détecter les fontis, nous proposons une méthodologie schématisée en

Figure 6. Celle-ci présente les différentes étapes de détection d’un fontis formé entre
deux acquisitions à t0 (nuage de points de référence) et à t1 (nuage de points de
comparaison):
1. Découpage des deux nuages de points selon une grille commune. La grille est
composée de cellules qui se superposent selon un facteur de recouvrement.
Le recouvrement permet de limiter les effets de bords lors de la détection
de fontis mais également qu’un point du nuage soit évalué plusieurs fois par
l’algorithme de détection et ainsi renforcer la confiance dans la détection.
2. Pour une même acquisition, on extrait les points à l’intérieur d’une cellule de
la grille et on recale les allers et retours entre eux
3. Pour une cellule commune aux deux acquisitions, on recale les deux nuages
de points entre eux
4. On calcule la différence d’altitude pour chaque point du nuage de comparaison
avec le nuage de référence.
5. On applique la méthode statistique de détection de fontis sur le nuage de points
extrait de la cellule associé aux écarts altimétriques précédemment calculés.
226
Q.106 - R.15
Fig. 7
Schéma de la méthode proposée pour la détection de fontis
Diagram of the proposed method for the distorsion detection.
Le recalage local des allers-retours d’une même acquisition permet de réduire

le bruit dans les données. Le recalage local entre deux acquisitions permet de
réduire le biais et ainsi d’avoir une distribution des écarts entre deux acquisitions
centrées en zéro.
La méthode de détection consiste à appliquer une méthode de clustering

localement dont les vecteurs à discriminer sont les écarts altimétriques de petits
groupes de points voisins (par exemple 30 points). L’idée est que dans une zone
inchangée, chaque vecteur constitué des écarts altimétriques dans un petit voisi-
nage se ressemble (au bruit prêt). Toutefois dans une zone ou une déformation
est apparue (i.e. un fontis), nous allons retrouver localement des vecteurs d’écarts
dissemblant entre ceux extraits dans la zone inchangée du nuage de points et ceux
extraits au niveau de la déformation (ou sur sa frontière). La méthode de clustering
permet alors de classifier (de manière non supervisée) ces vecteurs d’écarts. La
méthode de clustering la plus performante que nous ayons identifiée modélise ces
vecteurs par un mélange de Gaussiennes.
Nous avons évalué quantitativement les performances de la méthode pro-

posée à partir des acquisitions réalisées sur le canal de Sisteron (EDF) comprenant
des fontis simulés physiquement et numériquement sur la partie haute de la digue
227
Q.106 - R.15
Fig. 8
Schéma de la méthode de détection de fontis par clustering
Diagram of the detection method with clustering.
(§3.1). Nous montrons qu’en moyenne la méthode permet de détecter entre 81 %

et 87 % des points appartenant à un fontis (selon leur affaissement) pour un peu
moins 6 % de faux positifs.
4.2. COMPARAISONS DE MODÈLES NUMÉRIQUES DE TERRAIN (OU DE SURFACE)
Les différentes stratégies de comparaisons d’acquisitions surfaciques

(nuage/nuage, nuage/MNT, MNT/MNT) sont détaillées dans [1]. Dans le cas de
l’auscultation des endiguements CNR, le choix s’est porté sur la comparaison entre
2 modèles maillés.
4.2.1. Spécificités à prendre en compte
• Caractéristiques des nuages générés

La résolution finale des nuages obtenus par le système déployé par CNR
et IGN est de l’ordre de 1 point tous les 5mm. De par son mode de calcul,
les nuages de points issus de la photogrammétrie sont distribués sous forme
de grille (chaque point correspondant à un pixel d’une des photos). De fait, le
nuage est très dense et légèrement bruité.
• Caractéristiques de la surface auscultée
Les endiguements CNR sont essentiellement enherbés et composés de
graviers. Chaque pixel et donc chaque point calculé peut donc représenter un
gravier ou un brin d’herbe.
228
Q.106 - R.15
Fig. 9
Illustration d’un endiguement CNR
Illustration of CNR’s dike.
4.2.2. Modélisation des nuages de points
Du fait des deux caractéristiques exposées ci-dessus, nous nous retrouvons

confrontés à des nuages de points extrêmement denses décrivant un ouvrage dont
la surface peut évoluer ponctuellement: croissement de la végétation, mouvement
des graviers. Une comparaison point à point aura peu de sens, et amènerait à des
écarts non significatifs et non pertinents.
Afin d’éviter que les comparaisons entre campagnes successives soient

bruitées, il a été fait le choix de modéliser l’ouvrage. En effet, cette méthode permet
de réduire le bruit de mesure et l’incertitude finale. De plus, celle-ci peut être aussi
utilisée pour réduire le nombre de points et simplifier l’exploitation du nuage.
Les modélisations sont réalisées à l’aide du logiciel Leica Cyclone 3DR qui
permet la gestion de gros nuages de points ainsi qu’une analyse fine du résultat
de la modélisation via des comparaisons au nuage d’origine. Il conviendra de bien
choisir les mêmes méthodes de modélisations pour chacune des campagnes afin
de s’affranchir de ce potentiel biais.
4.2.3. Comparaisons
Après les phases de modélisation, les comparaisons se font de manière

classique par différence de modèles et affichage de cartes d’écarts. Ces cartes per-
mettant de visualiser de potentielles zones de tassement. Sur la figure ci-dessous,
on remarque que les écarts sont visibles sur le talus amont, la zone avec présence
de végétation, l’axe de la digue ne présentant pas d’écarts.
229
Q.106 - R.15
Fig. 10
Vue perspective entre deux comparaisons de modèles maillés.
Perspective view between two comparisons of mesh models.
4.3. MODÉLISATION 3D HYBRIDE
4.3.1. Traitement hybride des levés
Geokali a co-développé un procédé de modélisation 3D hybride par intégra-

tion de données laser et photogrammétriques: le calcul de corrélation dense entre
les images s’appuie sur une forte densité de points du scan laser. Cela permet
de produire des modèles 3D de haute précision: la qualité géométrique du scan
laser est conservée, et les zones invisibles au scanner sont complétées par le levé
photogrammétrique.
La qualité de l’acquisition laser est déterminante: consolidation géométrique

des différentes stations de scan d’une part et qualité de la colorisation du nuage de
points laser d’autre part.
230
Q.106 - R.15
Fig. 11
Levé 1/Traitement et localisation des écarts/levé 2 Code couleur: bleu: écarts
inférieurs à 2 cm, vert: écart entre 2 et 4 cm, rouge: supérieur à 4 cm.
Survey 1/Processing and locating gaps/Survey 2 Colour code: blue: gap less than
2 cm, green: gap between 2 and 4 cm, red: more than 4 cm.
4.3.2. Traitement pour le suivi du parement
Le traitement a permis de produire, pour chacun des levés, un nuage de points

colorisé d’une densité de l’ordre d’ 1 cm entre chaque point. La comparaison des
nuages de points issus des levés 1 et 2, pour la détection des mouvements de blocs,
est effectuée avec le logiciel 3DReshaper.
93,3% des écarts entre les nuages de points hybrides étaient de moins de 2 cm
en 3D, 4% des écarts se situaient entre 2 et 4 cm, le reste (2,7%) étant supérieur
à 4 cm. Cela révèle une cohérence géométrique satisfaisante entre deux relevés
séparés d’une semaine. Les sources identifiées des écarts sont: les déplacements
de blocs, le bruit naturel dû à la végétation et à l’eau, et le bruit numérique plus
spécifiquement lié aux effets d’ombres entre les pierres et au contre-jour (parement
amont orienté nord, levé en hiver, soleil rasant). Cette approche a permis de détecter
les blocs ou petits groupes de blocs ayant bougé d’au moins 2 cm, seuil largement
supérieur au seuil de détection nécessaire au suivi.
4.3.3. Validation du protocole
De par leur diversité, la détection de l’ensemble des blocs et groupes de blocs

a permis de confirmer que le protocole de suivi répondait aux attentes:
• Reproductibilité du levé, critère indispensable pour permettre un suivi,

• Seuil de détection supérieur aux limites de déplacements intéressant le suivi
des parements,
• Typologie des déplacements: petits blocs isolés, tailles des blocs, couleur des
blocs.
231
Q.106 - R.15
5. RETOUR D’EXPÉRIENCE ET PERSPECTIVES
Les 3 méthodes présentées dans cet article permettent maintenant aux

maitres d’ouvrages de déployer des nouvelles méthodologies et capteurs pour
la surveillance des ouvrages hydrauliques. Les réelles avancées qui ont eu lieu
ces dernières années permettent maintenant d’utiliser des porteurs légers tout en
garantissant des précisions de plus en plus fines.
Fort de leurs expériences, EDF, CNR et BRLi connaissent aussi les limites
de ces systèmes. L’un des points critiques est la capacité à déployer, à chacune
des campagnes, des systèmes de précision identiques et à maintenir un référentiel
topographique sur site suffisamment précis et stable au cours du temps pour perme-
ttre une comparaison robuste des campagnes. Les modes opératoires se doivent
donc d’être parfaitement maitrisés par des équipes multi-compétences afin de livrer
des résultats de précision connue et compatibles avec les attendus.
Il convient aussi de rappeler que la mise en œuvre opérationnelle de ces

solutions nécessite des conditions météorologiques et aérologiques adéquates. A
noter aussi qu’en France, les drones aériens peuvent être compliqués à déployer
de manière réactive sur le terrain du fait des exigences des scénarii réglemen-
taires, particulièrement pour des linéaires supérieurs à 2 km ou avec des porteurs
mobilisant des charges utiles importantes.
Les développements en cours sur le géoréférencement absolu avec un seul

récepteur par les méthodes PPP et PPK7 laissent entrevoir de nouvelles possibilités
pour simplifier les opérations de terrain en limitant voire supprimant le nombre de
point de calage au sol où en supprimant la nécessité de mettre en place un récepteur
GNSS de référence. La qualité des capteurs est aussi en constante évolution, ce
qui devrait permettre une meilleure maitrise des précisons finales.
La détection de fontis pourrait bénéficier des progrès algorithmiques, notam-

ment en intelligence artificielle, avec l’émergence depuis 2017 de réseaux de
neurones profonds appliqués aux nuages de points.
Le recours aux acquisitions (photogrammétriques et lasergrammétriques) est

amené à se démocratiser. Notamment avec les travaux du FUI FRELON, dont EDF
fait partie, visant à développer un capteur LiDAR français embarquable sur drone.
De plus, les travaux IGN-CNR pour le développement de la photogrammétrie haute
précision pour l’auscultation d’ouvrages linéaires sont maintenant implémentés dans
la suite logicielle open-source Micmac. Cela ouvre de nouvelles perspectives aux
gestionnaires pour assurer une surveillance toujours plus optimale
7 PPP: Precise Point Positioning/PPK Post Processed Kinematic
232
Q.106 - R.15
REMERCIEMENTS
Nous souhaitons remercier Rémy Boudon, Philippe Rebut et Gaël Verdun,

d’EDF DTG, qui ont réalisé les acquisitions drone LiDAR sur le canal de Sisteron
et mis en place le dispositif de simulation physique de fontis. Nous remercions
aussi Marc Pierrot-Deseilligny et Yilin Zhou pour les derniers travaux sur la thèse
IGN-CNR.
REFERENCES
[1] BOUDON, Rémy, et al., Les nouvelles techniques topographiques 3d sur-

faciques pour optimiser le suivi des ouvrages hydrauliques. CIGB, 25ieme
Congrès Stavanger 2015 – Q.99–R.53
[2] FAURE, Paul-Henri, et al. Rpas to monitor dikes: an optimised intervention.

International symposium on dams in global environmental challenges. BALI,
Indonesia: s.n., 1-6 june 2014.
[3] TOURNADRE, Vincent. Metrology with light aerial photogrammetry, applica-

tion to dyke monitoring. PhD Thesis of UNIVERSITÉ PARIS EST, 21 december
2015.
[4] ZHOU, Yilin. Métrologie 100% automatique par photogrammétrie aéroportée

légère et GPS de précision permanent et embarquée, application au suivi de
digues. PhD Thesis of UNIVERSITÉ PARIS EST, 6 december 2019.
[5] Bulletin ICOLD-CIGB n˚ 158: Dams Surveillance Guide/Guide de la Surveil-

lance des Barrages – 2018 – ISBN: 978-1-138-49020-8 (Pbk)/ 978-1-351-
03578-1 — 222 pages.
[6] Bulletin ICOLD-CIGB n˚ 138: Surveillance: Basic elements in a “DamSafety”

process/La surveillance: élément de base d’un processus “Sécurité des
barrages” – 2009 – 52 pages.
[7] Olivier DEGUELDRE. Les missions photogrammétriques réalisées par drone

au centimètre sans points de calage au sol. Revue XYZ N◦ 162 – Mars 2020.
233
Q.106 - R.16
-------
-------
NOUVELLES ANALYSES DU COMPORTEMENT DES BARRAGES

PIÉZOMÉTRIE ET DÉBITS: COMPORTEMENTS NON LINÉAIRES (∗)
Alexandre SIMON
Expert Auscultation des barrages, EDF
Jean-Paul FABRE
Expert Génie Civil, AB Fabre Consulting
Mathilde DE GRANRUT
Ingénieure d’études, ARTELIA
FRANCE
RÉSUMÉ
Le comportement des barrages dépend de plusieurs phénomènes explicatifs

réversibles (charge hydrostatique, effets thermiques) et irréversibles (consolida-
tion, fluage, extension de fissuration, pathologies, etc.). L’analyse classique HST
(Hydrostatique, Saisonnier, Temporel) développée par EDF additionne simplement
de manière indépendante les différents phénomènes. Les effets différés (diffusion
de la température, diffusion de l’eau dans les remblais, effet viscoélastique du béton)
sont pris en compte par un effet retard des variables explicatives correspondantes
(température et cote de retenue). En réalité, de nombreux phénomènes tels que la
piézométrie et les débits de drainage et de fuite sont des phénomènes non-linéaires
causés par l’action de plusieurs effets, qui interagissent au-delà de certains seuils
(passage du domaine linéaire au domaine non-linéaire). Par exemple, les pressions
d’eau au contact béton-rocher dépendent de l’influence simultanée (et non simple-
ment additive) de la charge hydrostatique et de la fissuration, qui elle-même dépend
∗ New analyses of dam behaviour - Piezometric levels and flowrates: Non Linear behaviour
234 DOI: 10.1201/9781003211808-112

Q.106 - R.16
du niveau de sollicitation imposé par les effets saisonniers, hydrostatiques et tem-

porels. Le modèle HST a donc évolué vers des versions linéaires par partie HyLP,
ou Non Linéaires NL HST, afin de mieux représenter les phénomènes physiques.
Finalement, à la lumière de ces analyses, le comportement réel des barrages et
leur évolution dans le temps peuvent être mieux mis en évidence, et mieux compris.
SUMMARY
The measurements in concrete dams foundations (crack opening, uplift-

pressure, flowrates, …) are influenced by several reversible phenomena (such as
hydrostatic load, thermal effects) and irreversible (such as setting up of founda-
tions, creep with modification of the stress in the foundations, etc.). The classic HST
(Hydrostatic, Seasonal, Temporal) analysis developed by EDF is often too simplistic
because it considers a mere addition of responses to external loads. The delayed
effects of external loads such as temperature, hydrostatic pressure or viscoelasticity
of the concrete are taken into account by delaying the corresponding explanatory
variables, i.e. temperature and water level. Many phenomena such as piezometric
levels in concrete-rock contact, opening of the dam contact, and flows, are non-
linear phenomena caused by the action of several effects, which interact beyond
certain thresholds. This threshold corresponds to the transition from the linear to
the non-linear domain. For example, the water pressures at the concrete-rock con-
tact depend on the combined (and not simply additive) influence of the hydrostatic
charge and the cracking, which itself is limited by the aperture of the crack and
depends on the level of stress imposed by the seasonal, hydrostatic and temporal
effects. The HST model has therefore evolved into Non-Linear versions, such as
NL-HST and Hylp, to better approach physical phenomena. Finally, thanks to these
analyzes, the actual behavior of dams can be better understood, and their evolution
over time better predicted.
1. INTRODUCTION
Les méthodes de justification des barrages (poids, voûtes, remblai...) ont

évolué vers un panel de méthodes plus ou moins sophistiquées, qui sont employées
selon les cas (calculs simplifiés de type Résistance des Matériaux, Éléments Finis
2D, 3D…). Les méthodes simples restent souvent préférables: elles sont mieux
comprises, bien connues, leurs hypothèses sont claires et leurs résultats éprou-
vés permettent l’analyse de la grande majorité des comportements des ouvrages.
Cependant, si les méthodes simples ne suffisent pas à préciser certains comporte-
ments, alors des méthodes plus fines peuvent être employées, pour permettre la
prise en compte de non-linéarités, de relaxation des tractions, d’endommagement,
235
Q.106 - R.16
ou de couplage hydromécanique. La méthode choisie permet de répondre aux

questions relatives à la sûreté des barrages.
Cette possibilité de complexifier les modèles pour affiner les analyses existe
également pour l’analyse des données d’auscultation à partir de la méthode HST
(Hydrostatique, Saisonnier et Temporel) [1] dont le modèle historique fonctionne
bien dans la majorité des cas, mais dont certaines améliorations sont utiles dans
les cas suivants:
• EFR: Effets retardés, pour prendre en compte la diffusion de la charge hydro-

statique en milieu peu perméable ou bien l’effet de la pluie sur les piézomètres
(1992),
• HSTT: Effets retardés de la température de l’air pour expliquer les déplace-
ments d’origine thermique des voûtes (2003) [2], [3].
Ces modèles sont intégrés au logiciel d’auscultation d’EDF (KOALA); ils

restent linéaires, avec de nouvelles variables explicatives: la cote retardée et la
température retardée avec un paramètre caractéristique du temps de retard. De
plus, le comportement viscoélastique de béton et de certains rochers se traduit
par des déformations différées mesurées notamment sur les voûtes, que l’on peut
modéliser en utilisant un modèle équivalent à HST+EFR [4].
De nouvelles méthodes d’analyse sont également employées lorsque les

méthodes «classiques» ci-dessus s’avèrent insuffisamment efficaces [4]. Leur mise
au point a été facilitée par l’utilisation de modèles à base de réseaux neuronaux,
qui seront présentés au §2.1.
En particulier, l’analyse de la piézométrie et des débits en fondation a connu

des développements récents très intéressants grâce au développement de modèles
basés sur une modélisation robuste des phénomènes physiques en jeu (§2.3 et 2.4).
Enfin, une meilleure modélisation et compréhension des composantes irréversibles
des modèles (§2.5) permet un diagnostic plus fin du comportement de l’ouvrage.
2. MÉTHODES D’ANALYSES SPÉCIALES APPLIQUÉES À DES

COMPORTEMENTS PARTICULIERS
2.1. LES RÉSEAUX DE NEURONES ARTIFICIELS (RNA)
Les réseaux de neurones artificiels sont une technique de data mining et

d’apprentissage automatique, autrement dit une méthode statistique avancée, per-
mettant l’exploration de données. L’application de cette méthode a pour objectif
d’identifier la manière dont les charges (hydrostatique, thermique, temporelle) influ-
encent le phénomène mesuré sur lequel porte l’analyse. Ce type de technique
236
Q.106 - R.16
Fig. 1
Structure du RNA utilisé pour l’analyse des mesures d’auscultation
Structure of Neural Network used for dam monitoring analysis
permet d’établir des liens entre la variable cible, qui est la variable que l’on cherche
à expliquer, et un ensemble d’autres variables, appelées les variables explicatives.
L’avantage de cette technique est qu’elle est capable de détecter des liens qui peu-
vent être très complexes, notamment non-linéaires, entre ces variables. Pour cette
raison, cette méthode a été choisie pour expliquer les liens entres les charges
réversibles et irréversibles pour certains comportements particuliers, de manière à
pallier les limites classiques des modèles de type HST, limites liées à leur caractère
linéaire.
La modélisation de ces charges passe par l’utilisation des mêmes variables

explicatives que le modèle HST et qui sont les suivantes [5]: le creux relatif Z pour
modéliser l’effet de la charge hydrostatique, le sinus et le cosinus de la saison S
pour modéliser les effets thermiques saisonniers, et le temps t pour modéliser les
effets temporels (Fig. 1). À partir de ces variables explicatives, le RNA construit une
estimation de la grandeur auscultée. L’écart entre la grandeur mesurée et modélisée
est quantifié par le terme d’erreur ε.
Ce modèle peut donc être vu simplement comme une fonction fRNA (Z,S,t) qui
relie les variables explicatives à la grandeur auscultée, modulo un terme d’erreur
ε. Comme pour les modèles de régression linéaire multiple, cette fonction fRNA
s’exprime par des paramètres (appelés poids synaptiques), qui sont ajustés aux
données grâce à un processus d’optimisation. L’avantage de cette technique est
son extrême souplesse, qui lui permet d’identifier des liens très complexes entre
les variables, mais cette caractéristique rend également les RNA sensibles au
sur-apprentissage, qui correspond au cas où le modèle s’est trop ajusté aux don-
nées, allant jusqu’à intégrer au modèle des éléments non physiques (comme le
bruit de mesure). Afin d’éviter cet écueil, l’application des RNA doit être mise en
œuvre en prenant un certain nombre de précautions, à la fois lors du choix de
237
Q.106 - R.16
l’échantillon d’apprentissage et de l’optimisation du modèle mais également lors de

l’utilisation du modèle en prévision, notamment afin de minimiser l’extrapolation,
c’est-à-dire l’utilisation du modèle avec une combinaison de variables explicatives
significativement différentes des observations. La sensibilité aux hyper-paramètres
de l’algorithme d’apprentissage (typiquement nombre d’itérations, répartition de la
base d’apprentissage et de test, taux d’apprentissage et moment d’inertie) per-
met de juger de l’ajustement du modèle. Une fois le modèle ajusté à une série de
mesures, le RNA est utilisé pour construire les effets réversibles et irréversibles,
par analogie avec ceux du modèle HST. Pour ce faire, une technique de visualisa-
tion est utilisée, basée sur le principe de l’analyse de sensibilité: l’effet d’une des
charges est obtenu en faisant varier la variable associée à cette charge sur toute
la largeur de sa plage de variation, tandis que les autres variables explicatives sont
fixées à des valeurs de référence. Le modèle étant non-linéaire, cette représen-
tation s’interprète en considérant les valeurs de référence des autres variables
explicatives, et constitue un outil de visualisation pour aider à la compréhension.
Le nombre minimum d’observations pour utiliser les RNA dépend de leur com-
plexité: en règle générale, il faut disposer d’au minimum une dizaine d’observations
pour caler chaque paramètre (poids). Sur un réseau comme celui présenté sur
la figure 1 (39 paramètres), il faut donc disposer d’environ 400 observations. De
nombreux types de phénomènes d’auscultation peuvent être analysés grâce aux
RNA: ouverture de fissure, piézométrie, débit, dès lors que le comportement additif
linéaire d’HST n’est plus adapté. L’exemple ci-dessous concerne un barrage de type
voûte à double courbure. Sa hauteur sur fondations est de 130 m, pour une longueur
en crête de 425 m. L’épaisseur varie de 25 m (base) à 6 m en crête. Sous l’effet
des basses températures, en hiver, le béton se contracte, et du fait de sa forme
en arc, la voûte se déporte vers l’aval. L’importante épaisseur de béton en pied
d’ouvrage est à l’origine d’une inertie thermique qui conduit à des effets saison-
niers maximaux en fin de saison chaude ou froide (octobre ou avril). Le régime
hydraulique réversible à l’interface barrage-fondation est fortement influencé par les
mouvements de décompression ; ainsi le déport vers l’aval contribue à décomprimer
la zone de contact béton-rocher en pied amont, et induit une variation de perméa-
bilité qui influence à son tour les sous-pressions et les débits de drainage. En fin
de période froide, la zone de contact béton-rocher est dans son état d’ouverture
maximal et la charge hydrostatique est alors transmise plus facilement. A contrario,
en période chaude, la dilatation du béton tend à recomprimer la zone de contact, ce
qui limite alors la transmission de la charge et diminue les débits de drainage. L’effet
hydrostatique des phénomènes mesurés en pied des consoles centrales étant donc
en partie conditionné par l’état thermique dans cette partie de l’ouvrage, ces deux
effets ne sont donc plus indépendants mais couplés.
Ainsi, si on s’intéresse aux débits de drainage totalisés collectés en pied de

barrage et qui sont représentatifs de l’état de sollicitation du pied de la voûte, les pics
de débit sont observés au mois de juin lorsque l’ouvrage est dans un état froid et que
le niveau de la cote de retenue est élevé. L’application du modèle HST sur la période
2000–2017 à l’analyse de ce débit de drainage (Fig. 2) présente un coefficient de
corrélation faible (0,48) et donc un caractère explicatif faible. Une consultation du
238
Q.106 - R.16
Fig. 2
Débits de drainage bruts en bleu, modélisés par
HST (en rouge) par RNA (en vert) et cote de retenue (noir)
Water level (black) and flowrates: measurements (blue),
predictions with HST (red) and with RNA (green)
graphique des débits simulées avec HST montre les limites de la modélisation des
maxima de débits lorsque les sollicitations du pied de l’ouvrage sont les plus fortes
(en 2005 et 2006 sur la Fig. 2). Par ailleurs des débits simulés avec HST sont parfois
légèrement négatifs, ce qui illustre l’inadaptation d’HST.
Les RNA ont davantage de paramètres que le modèle HST (39 dans l’exemple
de Fig. 1 contre 10 pour HST) et il est intéressant de comprendre pourquoi ils per-
mettent une meilleure représentation des phénomènes mesurés. La Fig. 3 (gauche)
présente l’effet hydrostatique à différentes saisons, évalué à partir des mesures de
la période 2012–2017. Ainsi, à variations de cotes égales, les effets hydrostatiques
sur les débits de drainage sont plus importants en avril (fin de période froide) qu’en
octobre (fin de période chaude). Cet effet, non décrit par HST, (en gris) (lorsque
la cote passe de 415 m à 425 m, les débits de drainage modélisés par HST aug-
mentent de 16 L/min à 40 L/min quelle que soit la saison) est bien modélisé par les
RNA qui prévoient une augmentation de 14 L/min à 60 L/min en avril et de 14 L/min
à 18 L/min en octobre. C’est cette meilleure estimation par les RNA qui explique
leur meilleur coefficient de corrélation à 0,97. La Fig. 3 (droite) montre que pour les
cotes les plus élevées, les effets hydrostatiques des débits de drainage sont plus
élevés sur la période 2012–2017 que sur la période 2000–2005. Cette hausse des
débits sur la période 2012–2017 est due à une modification en 2011 du réseau de
239
Q.106 - R.16
Fig. 3
Effets hydrostatiques calculés par RNA variant avec la saison et la
date et par HST (courbe grise à gauche)
Hydrostatic effects by Artificial Neural Networks depend on season
and time and by the HST model (grey curve on the left side)
drainage vu par le totalisateur et cette modification se traduit ici par une hausse de
l’irréversible uniquement à cote haute. Cet effet ne peut pas être vu par HST qui
modélise un effet hydrostatique et irréversible moyens sur la période 2000–2017
et quelle que soit la saison. Les RNA ont la capacité de capter les non-linéarités
entre variables explicatives, là où HST ne modélisait qu’une combinaison linéaire
de celles-ci.
La cote maximale relevée au sein des mesures est d’environ 426 m sur la
période de janvier à avril entre 2011 et 2017, ce qui explique que la courbe verte
de la Fig. 3 obtenue avec les RNA n’est pas tracée au-delà de cette cote. En effet,
il faut s’affranchir de toute extrapolation vis-à-vis des variables explicatives (dont
la cote de retenue) car les RNA ne sont pas adaptés pour l’extrapolation. Ainsi,
moyennant ces précautions, une très bonne confiance est obtenue dans les effets
qui sont affichés et qui reposent systématiquement sur un nombre conséquent et
suffisant d’observations. Il est également possible d’afficher l’évolution de l’effet
hydrostatique avec le temps.
Les RNA apparaissent ainsi comme un outil puissant pour mieux préciser
le comportement de l’ouvrage et porter un diagnostic éclairé et quantifié sur son
comportement. Le gain est surtout important pour les phénomènes hydrauliques
(sous-pressions et débits) pour lesquels interviennent significativement des non-
linéarités et des phénomènes à seuil. Ces résultats étant directement valorisables
pour l’ingénierie, une application logicielle NOVAE [6][6] a été développée et
déployée afin de permettre l’analyse par RNA des mesures d’auscultation (quelle
que soit leur nature, hydraulique ou mécanique) du parc de barrages EDF. La seule
limite à l’application des RNA concerne le nombre d’observations nécessaires et
qui se chiffre en général à quelques centaines pour les RNA à l’architecture la
plus simple. Un modèle RNA nécessite davantage de temps qu’HST pour régler
les hyper-paramètres du calage puisqu’il faut en tester plusieurs combinaisons afin
d’être certain d’avoir un résultat consolidé. Ainsi, les RNA sont pour l’instant dédiés à
240
Q.106 - R.16
l’analyse des phénomènes les plus complexes pour lesquels les modèles classiques
ne donnent pas satisfaction.
2.2. LES COMPORTEMENTS NON-LINÉAIRES
Les modèles linéaires montrent parfois leurs limites, notamment pour l’étude
de phénomènes qui présentent des comportements non linéaires. L’exemple type
associé aux modèles non linéaires est l’évolution d’une ouverture de fissure en
fonction du niveau de sollicitation global (hydrostatique + thermique + irréversible).
L’ouverture du contact béton rocher des voûtes est un exemple de comportement
non-linéaire de la fondation: le contact béton-rocher ne commence à s’ouvrir qu’à
partir d’un certain seuil de sollicitation. Cette sollicitation peut tout à fait être définie
de manière identique à ce qui est fait dans la méthode HST (mêmes expres-
sions algébriques), mais elle sera transformée par une fonction non-linéaire pour
représenter de manière plus pertinente le phénomène mesuré. Plusieurs fonctions
peuvent être envisagées, selon le phénomène à représenter. Une fonction à deux
pentes est couramment utilisée dans les modèles non linéaires: l’effet sera différent
au-delà ou en-deçà d’une certaine valeur seuil de sollicitation qui est à caler. Il est
également possible d’utiliser une fonction sigmoïde (tangente hyperbolique). On
montrera au §2.4 que les fonctions de cette famille sont adaptées pour modéliser
des variations de perméabilité en fondation, sous l’effet des variations de l’état de
contrainte du rocher.
En prenant l’exemple d’une fonction avec une rupture de pente, pour une
certaine valeur seuil w, la réponse change de pente. Cette valeur seuil (w) ainsi que
la variation de pente (y) sont à caler:
Si MBmod > w, alors MBmodNL = MBmod (1)
Si MBmod < w, alors MBmodNL = MBmod + y∗ (w − MBmod ) (2)
Avec MBmod : Mesure brute modélisée par le modèle linéaire
Les phénomènes hydrauliques surveillés en auscultation (piézométrie et

débits) n’apparaissent qu’au-dessus d’un seuil de la cote de retenue permettant
la saturation de la fondation et du drainage. Il s’agit donc de phénomènes non
linéaires: dans le cas des débits de fuite on aura w = 0 et y = 1.
2.3. MODÉLISATION DES DÉBITS: MODÈLE HYDRAULIQUE LINÉAIRE PAR PARTIE
Les chapitres suivants décrivent des modèles simples, s’appuyant sur les lois
de l’écoulement, et qui vont permettre de mieux analyser ces phénomènes.
241
Q.106 - R.16
2.3.1. Approche par la loi de Darcy
Même si la perméabilité du milieu n’est pas homogène (présence de zonage

ou des organes d’étanchéité en fondation) et même si le drainage prélève une part
des écoulements surveillés, la loi de Darcy s’applique aux écoulements collectés
au drainage, ou aux débits de fuite, tant que la perméabilité de chaque partie du
parcours de l’eau reste constante, c’est-à-dire ne varie ni en fonction du temps, ni
de la saison, ni surtout en fonction de la charge appliquée sur le matériau. Pour tenir
compte du fait qu’on ne surveille pas des écoulements en régime établi, il faudra,
lorsque le temps de réponse du milieu est important (en milieu non saturé), prendre
comme variable explicative la charge retardée (cote de la retenue retardée Cr). S’il
n’y a qu’un seul cheminement à partir d’une source unique les débits n’apparaissent
qu’au-dessus d’un certain seuil: la cote d’apparition CA, on définit alors la charge
(variable explicative pertinente) comme:
Charge = Maximum (0; Cr − CA).
On obtient le modèle le plus simple suivant: (Qo est le débit qu’on observerait
barrage vide).
Influence en débit = Q − Qo = Charge ∗ K∗ S/L = Qm ∗ Charge
Ce modèle, linéaire au-delà du seuil CA, est déjà riche d’enseignements con-
cernant l’origine des circulations surveillées, il dépend en effet de 3 paramètres
physiques: la cote d’apparition (liée au seuil w défini dans le §2.2), le temps de
réponse, et un coefficient d’amplitude lié aux perméabilités.
2.3.2. Généralisation à des perméabilités faiblement variables
Si les perméabilités varient lentement dans le temps (colmatage par alluvion-

nement amont, efficience du drainage) on peut apporter en première approche un
correctif au coefficient d’amplitude en le multipliant par
Maximum (0; 1 + f(t)),
ce correctif est borné à zéro pour garder les débits positifs. La fonction f(t)
peut être choisie parmi les fonctions du temps classiques de HST. On obtient
Q − Qo = Qm ∗ Maximum (0; 1 + f(t)) ∗ Maximum (0; Cr − CA).
Cette formulation est alors assez souple pour modéliser de façon satisfaisante
sur le plan physique des débits présentant des influences diverses: apparition ou
modification dans le temps, tout en gardant une bonne vraisemblance y compris
242
Q.106 - R.16
en extrapolation, par exemple pour des abaissements de cote de retenue jamais

observés.
2.3.3. Le modèle HyLP: hydraulique linéaire par parties
Le modèle simple ci-dessus est déjà linéaire par parties en fonction de la

cote de retenue (la première partie de l’effet hydrostatique, au-dessous de la
cote d’apparition CA, étant de pente nulle). Si les débits collectés proviennent de
plusieurs sources distinctes dans l’espace, au travers d’un nombre limité de cir-
culations préférentielles localisées dans certaines faiblesses de l’ouvrage ou de la
fondation rocheuse, il est facile de généraliser le modèle ci-dessus en lui ajoutant
des sources supplémentaires avec leurs caractéristiques (cote d’apparition, temps
de réponse, coefficient d’amplitude, évolution dans le temps)

Q−Qo = Qmi ∗ Maximum (0; 1 + fi(t))∗ Maximum (0; Cri − CAi).
Le modèle obtenu (voir exemple ci-dessous au 2.3.4) est alors caractérisé

par un effet hydrostatique linéaire par parties (loi de Darcy pour chaque source).
Chaque source correspond à une circulation préférentielle différente (localisation
spatiale), qui peut évoluer, entrainant un pivotement des segments linéaires, dans
le sens horaire si la perméabilité diminue (compression, colmatage …) dans le sens
anti-horaire si la perméabilité augmente (décompression, dissolution, …) [7].
Un débit mesuré dans une collecte totalisatrice peut être étudié d’autant plus
facilement qu’on utilise les modèles issus des débits partiels mesurés, en y ajoutant
éventuellement des sources pour expliquer les apports intermédiaires. Ce modèle
devient alors très intéressant car il facilite la recherche de l’origine (au sens de la
localisation altimétrique) des débits, une partie pouvant provenir de circulations pro-
fondes, une autre partie de circulations plus hautes (contournement par les rives).
On peut ainsi plus facilement rechercher parmi les particularités de l’ouvrage, ou de
la fondation, celles qui peuvent être à l’origine d’évolutions indésirables, et orienter
les études et travaux pour traiter ces défauts si nécessaire.
2.3.4. Exemple d’application du modèle HyLP
Ces méthodes ont été employées pour interpréter le comportement à la mise

en eau de barrages en Amérique du Sud [7] ainsi que le comportement de plusieurs
barrages en Argentine dont celui de El Cadillal [8]: l’exemple des fuites de la «Presa
Lateral 3» par contournement du noyau dans une fondation comportant des couches
de gypse est présenté ci-dessous. L’analyse fine des mesures d’auscultation (débit
Afo3 Total) en utilisant les débits partiels mesurés a permis de construire un mod-
èle physique cohérent avec les mesures, (Fig. 4) et d’étayer les hypothèses de
circulations préférentielles en fondation:
243
Q.106 - R.16
Fig. 4
Débits de fuite (en noir, modèle HyLp en rouge) en l/min (A) dans le temps (B)
Leakages (in black, HyLp model in red) in l/min (A) over time (B)
On voit (Fig.5) que l’alimentation la plus profonde apparait à la cote 599,64,

elle est évolutive dans le temps, ainsi que celle à 603,62. Il y en a aussi deux autres
à 601,06 et à 606,71, mais qui sont de faible débit et non évolutives.
Ces informations permettent de mieux appréhender le comportement, son

évolution, de localiser les défauts les plus importants ou évolutifs, et donc de mieux
dimensionner le traitement de la fondation si nécessaire.
2.4. L’ANALYSE DE LA PIÉZOMÉTRIE EN FONDATION DES VOÛTES: MODÈLE NL-HST
Fort des enseignements tirés de l’analyse par RNA et de l’approche par Darcy,
un modèle d’interprétation des niveaux piézométriques (NP) plus simple de mise
en œuvre et d’utilisation que le RNA a été construit. Ce modèle intègre de façon
explicite les effets non-linéaires des charges de manière à figer la physique du
phénomène, en s’appuyant sur les observations précédentes et la connaissance de
la physique du phénomène contenue dans la littérature. Le principe de ce modèle
est de considérer qu’un écoulement est établi entre l’amont et l’aval, et que des
pertes de charge ont lieu le long de cet écoulement. Un piézomètre, positionné
dans ce profil à une distance x de l’extrémité amont du pied, enregistre le niveau
piézométrique (x) [m]. Ainsi, la charge hydraulique H(x) [m] vue par le piézomètre
244
Q.106 - R.16
Fig. 5
Effet Hydrostatique linéaire par parties (en 2019 en noir, en 2004 en bleu)
Hydrostatic effect linear by parts (in 2019 in black, in 2004 in blue)
en x est égale à une fraction (x) de la charge amont Ham. Ceci permet d’en déduire
l’expression des niveaux piézométriques (x): NP(x) = Nav + F(x) · Ham avec Nav
le niveau aval [m].
On attribue ensuite une résistance hydraulique équivalente à la zone de

l’interface située en amont et en aval du capteur, notées kam (resp. kav), et on
peut montrer que le facteur F(x) s’exprime en fonction du rapport de ces perméa-
bilités. Si on fait l’hypothèse que la perméabilité de la fondation est dépendante
des contraintes mécaniques (méca) que la voûte lui impose, on modélise ce facteur
comme une fonction de ces contraintes. Ce terme de contraintes est modélisé de
façon additive, comme il est d’usage de le faire pour les phénomènes mécaniques,
puis on tient compte du fait que ces contraintes mécaniques ont un effet non-linéaire
sur la perméabilité du milieu par l’introduction d’une fonction tangente hyperbolique
[9]. L’intérêt de cette fonction est qu’elle permet de modéliser le passage de l’état
fermé à ouvert, avec un effet de seuil sur le niveau de sollicitation global. Le mod-
èle NL HST ainsi obtenu s’exprime ci-dessous dans sa formulation la plus simple.
D’autres possibilités existent (cf. §4.1) pour la prise en compte de l’irréversible.
NP(x) = b0 + (h − b0 ) · (b1 + b2 · tanh(a0 + fther (S) + fhyd (Z) + firr (t))) + ε
Avec: fth (S) = a1 · cosS + a2 · sinS + a3 · cos2S + a4 · sin2S
fhyd (Z) = a5 · Z + a6 · Z2 + a7 · Z3 + a8 · Z4
245
Q.106 - R.16
Fig. 6
Débits Totaux (A) ramenés à conditions identiques, avec la tendance (en bleu),
et l’évolution dans le temps (B) des débits amenés par chaque alimentation
Total leakages (A) reduced to identical conditions, with the trend (in blue),
and the evolution over time (B) of the flows brought by each supply.
Fig. 7
Profil amont-aval et notations
Upstream-downstream view and notations
firr (t) = a9 · t
b 0 , b1 , b2 , a0 , . . . , 9 les coefficients de régression
Les 13 coefficients de régression du modèle sont ajustés par un procédé

d’optimisation itératif utilisant l’algorithme de Levenberg-Marquardt. Les résultats
246
Q.106 - R.16
Fig. 8
Linéarité de la piézométrie en fondation avec la cote:
perméabilité du milieu indépendante des variations de charge hydrostatique.
Linearity of the piezometry in the foundation with the water level:
permeability of the medium independent of the hydrostatic head variations.
d’application de ce modèle se sont révélés parfaitement comparables à ceux des

RNA, les effets calculés par le modèle s’interprétant de façon conforme à la physique
du phénomène. De plus, l’analyse de nombreux capteurs a montré que le modèle NL
HST est capable d’expliquer un ensemble varié de comportements, démontrant la
robustesse du modèle. Enfin, l’analyse de la piézométrie au contact des voûtes par le
modèle NL HST permet de poser un diagnostic sur l’état du contact, en particulier en
caractérisant l’extension spatiale de son ouverture. Ceci passe par une analyse fine
des effets hydrostatiques, dont les clés de lecture et d’interprétation sont décrites en
détail dans une cartographie des effets hydrostatiques. Un élément caractéristique
d’un graphe d’effet hydrostatique est la courbure décrite par chacune des courbes
de ce graphe. Cette courbure renseigne sur la sensibilité de la perméabilité à la
charge hydrostatique.
2.4.1. Cas linéaire: perméabilité invariable kH = f (x)
Lorsque la perméabilité du milieu ne varie pas avec la charge hydrostatique

et les conditions thermiques, (au moins sur une certaine plage: à cote basse par
exemple) et que la variation de piézométrie observée est uniquement due à une
augmentation de la charge à l’amont; autrement dit, pour une augmentation de la
condition limite amont de l’écoulement, l’effet de la cote est linéaire. Les variations
de contraintes mécaniques sur la voûte ne suffisent pas à faire varier les contraintes
mécaniques de l’interface de manière à en modifier la perméabilité du milieu qui peut
donc être modélisé en faisant appel à la loi de Darcy.
247
Q.106 - R.16
Fig. 9
Courbure de l’effet de la charge hydrostatique sur la piézométrie:
déformations élastiques de la fondation
Curvature of the effect of the hydrostatic load on the piezometry:
elastic deformations of the foundation
Ceci peut correspondre au comportement de la fondation d’un barrage poids,

tant que la perméabilité ne varie pas avec les variations de hauteur d’eau. Dans
le cas des voûtes, cette linéarité de l’effet hydrostatique s’interprète comme corre-
spondant à un état de l’interface invariant quelle que soit la charge hydrostatique
et thermique, qui ne s’observe en général que pour un certain intervalle de vari-
ations de cote (cotes basses). L’état du contact peut alors être soit constamment
comprimé pour cet intervalle de cote (cas d’un capteur situé très à l’aval), soit con-
stamment décomprimé, voire en traction (cas d’un capteur très proche de l’amont).
Ceci s’observe aussi bien pour des ouvrages dont l’interface présente une fissuration
que pour des ouvrages dont l’interface ne se déforme que de façon élastique.
2.4.2. Perméabilité variable en fonction de la charge: kH = f (x, meca), sans

fissuration du contact béton-rocher
Le second cas rencontré correspond au cas où l’effet hydrostatique présente

une courbure, mais où celle-ci reste de signe constant avec les variations de cote.
La présence d’une courbure indique que la perméabilité du contact varie avec

les variations de cote, donc que l’interface se comprime ou décomprime sous l’effet
des déplacements de la voûte, et donc sous l’effet des variations de sollicitations
extérieures. Les variations de perméabilité interviennent en réponse à une évolu-
tion des contraintes mécaniques de l’interface, donc à une déformation de celle-ci.
248
Q.106 - R.16
Fig. 10
Changement de courbure de l’effet hydrostatique (fissuration du contact)
Change in curvature of the hydrostatic effect
(opening of the rock-concrete interface)
Si cette déformation reste dans le domaine élastique, elle correspond à une simple
décompression élastique de la fondation sous le pied amont de la voûte, accompag-
née d’une compression de la fondation sous le pied aval. La piézométrie au contact
évolue donc sous l’effet de la variation du contraste de perméabilité entre l’amont
et l’aval. Ce genre d’évolution peut également correspondre au comportement des
fondations de barrages poids.
2.4.3. Changement de courbure: kH = f (x, meca) avec fissuration du contact
Enfin, la dernière configuration observée correspond au cas où un change-

ment de courbure franc est visible sur les courbes de l’effet hydrostatique.
La présence du changement de courbure témoigne d’une ouverture de la

zone de contact. La cote à partir de laquelle ce changement de courbure intervient
s’interprète comme correspondant à la charge hydrostatique à partir de laquelle
la zone du contact à l’amont du capteur est passée à l’état décomprimé, voire en
traction, ouvrant ainsi la fissure et permettant le transfert de la charge hydrostatique
vers l’aval. Cette ouverture de la fissure se traduit par une variation de perméa-
bilité brutale, qui produit une hausse de piézométrie supérieure à la hausse de la
cote de retenue, contrairement au cas de déformation élastique qui correspond à
des déformations continues et progressives. Afin de faciliter l’analyse des effets
249
Q.106 - R.16
Fig. 11
Effet hydrostatique (a) et sa dérivée seconde (b) modélisé par NL HST
Hydrostatic effect (a) and its second derivative (b) modeled by NL HST
hydrostatiques, un graphe supplémentaire qui présente la dérivée seconde1 de

l’effet hydrostatique en fonction de la cote permet d’apprécier de façon plus objec-
tive les changements de courbure et les points d’inflexion (annulation de la dérivée
seconde).
2.4.4. Présence ou non d’un point d’inflexion
Dans le cas où un changement de courbure est présent à partir d’une cote

seuil, et donc que la zone du contact est fissurée, les effets hydrostatiques peuvent
également présenter un point d’inflexion pour des cotes supérieures à la cote seuil.
Ces points d’inflexion s’interprètent comme traduisant le chargement mécanique
(remplissage et état thermique) à partir duquel la fissuration du contact atteint le
capteur.
Avant que la fissure n’atteigne le capteur, la perméabilité à l’aval de ce capteur

n’évolue pas, tandis que la perméabilité à l’amont augmente. La courbure de la fonc-
tion hydrostatique est donc positive (dérivée seconde positive). À l’inverse, une fois
que la fissure a atteint le capteur, si la charge hydrostatique continue d’augmenter,
l’ouverture du contact continue sa propagation vers l’aval, dépassant ainsi le cap-
teur. La perméabilité à l’aval du capteur se met alors à augmenter, tandis que la
perméabilité à l’amont du capteur augmente aussi mais cette augmentation est
1La dérivée seconde d’une fonction associée à une courbe définit la courbure de cette
courbe.
250
Q.106 - R.16
compensée par l’augmentation de la perméabilité à l’aval, ce qui se traduit par une

courbure négative (dérivée seconde négative). Le passage de la courbure positive
à la courbure négative définit alors un point d’inflexion (annulation de la dérivée
seconde), qui est interprété comme correspondant à la charge hydrostatique pour
laquelle la fissure atteint le capteur. Lorsque la cote continue d’augmenter, la fissure
continue à se propager vers l’aval. L’extrémité de la fissure est alors trop éloignée
du capteur pour l’influencer, et les pressions d’eau mesurées sont alors influencées
uniquement par la charge amont, et leur évolution avec la charge devient linéaire,
d’où une courbure de l’effet piézométrique qui s’annule. À l’extrême, il est possible
d’imaginer que pour un barrage dont le contact serait dégradé au point où la fissure
atteindrait le voile de drainage, cette mise en communication se traduirait par une
chute de la piézométrie pour les capteurs situés à proximité immédiate du voile de
drainage.
Ainsi, l’analyse des effets hydrostatiques fournit une description de la réponse

locale de la piézométrie aux variations de charges, mais aussi, et c’est là qu’est la
plus-value majeure, l’interprétation d’un seul capteur permet de déduire des élé-
ments de caractérisation spatiale de l’état du contact. En connaissant la position
du capteur étudié, il est possible d’établir jusqu’où la fissuration du contact s’est
propagée, et le cas échéant sous quelles conditions de charge, en distinguant la
contribution de chaque charge. Jusqu’à présent, la détermination de la propagation
spatiale de la dégradation du contact n’était possible que de façon qualitative via
les mesures d’auscultation brutes, ou grâce à l’utilisation de modèles numériques,
dont la mise en œuvre est longue, complexe, et inadaptée à la surveillance en temps
réel. La présence de plusieurs capteurs situés en différents points du contact permet
d’affiner la caractérisation de l’ouverture à l’interface béton-rocher non seulement
dans la direction amont-aval mais également dans la direction rive à rive.
2.5. IRRÉVERSIBILITÉ ET MODÉLISATION STATISTIQUE
Dans le modèle classique HST, l’évolution dans le temps est supposée être
un terme additif, indépendant des autres variables (cote et saison). En réalité ce
n’est pas toujours le cas notamment pour les phénomènes hydrauliques (exemples
des paragraphes 2.3.4 et 2.4) : si les débits de fuite augmentent dans le temps
(dégradation de l’étanchéité d’un masque) c’est bien visible quand le barrage est
plein, alors qu’à vide les fuites n’évoluent pas et restent nulles. Dans ce cas, HST
utilise une loi d’évolution dans le temps «moyenne» et sur les mesures corrigées
obtenues à conditions identiques «barrage plein» on retrouve cette vitesse moyenne
d’évolution irréversible. De plus les mesures sont assez mal corrigées. Un modèle
adapté aux phénomènes hydrauliques pourra prendre en compte cette réalité et
permettre une meilleure évaluation des vitesses d’évolution irréversible (qui seront
différentes à plein et à vide). Ces améliorations sont pertinentes dans beaucoup de
cas, cependant il faut éviter les extrapolations incertaines en dehors du domaine de
sollicitations couvert par l’échantillon de mesures dont on dispose.
251
Q.106 - R.16
De plus l’évolution irréversible d’un phénomène hydraulique (§2.4) peut

provenir de différentes causes:
Npbrut = Nαv (t) + RN − Rvide (1 − z)

(i) (ii)
∗ [ α 1 (t) +(α 2 ∗ tanh(α 3 + ξ (z) + ξ (S) + ξ (t))] + ∈

(iii)
L’évolution peut provenir soit (i) du niveau aval, soit (ii) de la perméabilité du
milieu (colmatage …), soit (iii) de la fissuration. Il est impossible de réaliser cette
séparation des origines d’une évolution dans la modélisation HST. La non-linéarité
d’un phénomène se traduit également par des évolutions irréversibles qui dépen-
dent de l’état de sollicitation: prenons par exemple un débit de fuite qui diminue
lorsqu’un joint se ferme lorsque le barrage se remplit. Ce débit peut ne pas évoluer
à RN (fermé, donc débit nul), alors que son apparition au cours de l’abaissement de
cote commence plus tôt et s’amplifie dans le temps. Les mesures corrigées mon-
treront un effet irréversible proportionnel à la pente de la fonction de non-linéarité
dans la zone des conditions identiques choisies: maximal dans la zone de fonc-
tionnement linéaire, il sera limité à cote haute par la non linéarité. Dans ce cas le
modèle HST aurait déterminé une évolution irréversible «moyenne». Avec le modèle
non-linéaire, la composante irréversible sera différente en fonction des conditions
identiques choisies, et les mesures ramenées à conditions identiques seront mieux
corrigées.
3. CONCLUSION
Les modèles historiques d’analyse des données d’auscultation ont montré leur
efficacité, et leur relative facilité d’utilisation, mais se sont avérés limités dans cer-
tains cas, ce qui explique la recherche très active au cours des dernières décennies.
Les chroniques de mesures d’auscultation contiennent les données du comporte-
ment des barrages patiemment et méticuleusement réalisées et accumulées par
nos prédécesseurs, souvent depuis la mise en eau des barrages
Les modèles statistiques de comportement qui s’appuient sur un historique de

plus en plus étoffé dans le temps, peuvent ainsi devenir plus précis. Avec l’aide des
outils statistiques modernes tels que les RNA, la recherche de modèles adaptés
aux phénomènes mesurés est plus efficace. Ces facteurs ont permis l’émergence
récente de nouvelles familles de modèles qui complètent et enrichissent désor-
mais les analyses classiques, même si leur industrialisation n’est pas totalement
terminée. Certains concepts sont riches car ils se rapprochent de la réalité des
comportements des ouvrages, et montrent leur complexité.
Les nouveaux modèles permettent la meilleure prise en compte des

phénomènes hydrauliques et des comportements non-linéaires. Ils réclament des
252
Q.106 - R.16
échantillons de mesures assez fournis, et les résultats en termes de comporte-

ment ne doivent pas être extrapolés hors du champ des situations observées. Il y
a donc possibilité de tirer des informations plus complètes, plus pertinentes pour
mieux comprendre le comportement particulier de certains ouvrages, à partir de
l’historique de leurs mesures d’auscultation, mais l’interprétation demandera une
attention et des compétences particulières, notamment une compréhension fine
des phénomènes physiques. Ces modèles complexes sont réservés au traitement
de cas spécifiques par l’ingénierie (quelques exemples sont donnés ci-dessus), pour
lesquels ils doivent être adaptés avec précaution afin de préserver le sens physique
des résultats. Les plus utiles mériteront d’être intégrés ultérieurement dans des
outils de surveillance quotidiens des ouvrages.
REMERCIEMENTS
Les auteurs tiennent à remercier Pauline Laviron-Petit, Thierry Guilloteau,

Emmanuel Robbe, Matthieu Castay et Ascensio Lara, pour leur relecture et leurs
conseils. Nos remerciements vont également à nos deux relectrices du CFBR,
Isabelle Penot et Nathalie Rosin-Core, pour leur relecture minutieuse et leurs
conseils de clarification du texte.
REFERENCES
[1] G. Willm, N. Beaujoint G. Willm, “ Les méthodes de surveillance des bar-

rages au sein de la production hydraulique d’Electricité de France. Problèmes
anciens, solutions nouvelles, 9ème Congrès des grands Barrages, Istanbul,
1967, Q 34, R 30, pp. 529–550.
[2] Penot, I. & Daumas, B. & Fabre, JP. «Monitoring behaviour», International
Water Power & Dam Construction, December 2005.
[3] Penot, I. & Fabre, JP. & Daumas, B. Analyse et modélisation du comportement
des ouvrages de génie civil par la prise en compte des températures de l’air:
méthode H.S.T. Thermique. 23ème Congrès des grands Barrages, Brasilia,
2009, Q 91, R 60.
[4] Fabre, J-P. & Mauris, F. & Simon, A. Nouvelles analyses du comportement des
fondations des barrages (non-linéarités, effets élastiques différés) Colloque
CFBR: Auscultation des barrages et des digues. Pratiques et perspectives,
Chambéry, 2012.
[5] Simon, AG. & Fabre, JP. «Analyse des mesures d’auscultation des barrages
avec les réseaux de neurones» Colloque CFBR: Auscultation des barrages et
des digues. Pratiques et perspectives, Chambéry, 2012.
253
Q.106 - R.16
[6] Simon, A., De Granrut, «Analysis and Interpretation of Dam monitoring: use
of artificial neural networks in an industrial context» Dam World Conference,
2018.
[7] Lara, A. «Hidro-auscultación: dos lecciones útiles» VIII Congreso Argentino

de Presas y Aprovechamientos Hidroeléctricos Santa Fe – Argentina, 2015.
[8] Paolantonio, R. & Fabre, JP. «Aprovechamiento hidroelectrico El Cadillal»

Hydroelectrica Tucuman S.A. Informe anual de auscultación, Périodo Abril
2019 Marzo 2020.
[9] De Granrut, M. «Analyse et interprétation de la pression d’eau en fondation

des barrages-voûtes à partir des mesures d’auscultation» thèse de l’Université
de Grenoble Alpes, soutenue le 28 octobre 2019.
254
Q.106 - R.17
-------
-------
ANALYSE DES DONNÉES D’AUSCULTATION DE BARRAGES

PAR DES MÉTHODES D’APPRENTISSAGE AUTOMATIQUE (∗)
Guillaume VEYLON
INRAE
Nathalie ROSIN-CORRE
TRACTEBEL ENGINEERING
Yifeng LIN
TRACTEBEL ENGINEERING
Claudio CARVAJAL, Antoine WAUTIER, Franck TAILLANDIER,

Li-Hua LUU & Laurent PEYRAS
INRAE
FRANCE
RÉSUMÉ
La sécurité d’un barrage repose sur le suivi de son comportement tout au long
de sa vie. De nombreuses méthodes existent pour établir un modèle statistique de
comportement d’un barrage. Elles consistent à établir un modèle prédictif d’une
mesure en fonction de plusieurs variables explicatives. L’approche la plus utilisée
par la profession, la régression multilinéaire, montre ses limites lorsque la popu-
lation n’est pas statistiquement homogène ou lorsque les phénomènes modélisés
sont fortement non linéaires (effets de seuil par exemple). L’article présente trois
types de modèle d’apprentissage automatiques et compare leurs performances
respectives sur la base de trois applications à la modélisation des déplacements
∗ Analysis of Dam Monitoring Data by Deep Learning Methods
DOI: 10.1201/9781003211808-113 255

Q.106 - R.17
d’un barrage poids, des débits de fuite d’un barrage voûte et des niveaux pié-
zométriques dans le noyau d’un barrage en remblai. Nos analyses montrent que les
méthodes d’apprentissage automatique sont susceptibles d’apporter d’intéressants
compléments à l’analyse, notamment lorsque les modèles classiques de régression
multilinéaire ont de faibles coefficients de détermination. Les modèles de régres-
sion non linéaire (SVR) se sont révélés faciles d’emploi et robustes. Les modèles
de réseaux de neurones (ANN) ont démontré leur efficacité à prendre en compte
les fortes non-linéarités. Les modèles de forêt aléatoire (RF) ont été capables
d’identifier les discontinuités dans l’évolution temporelle des mesures. Loin de con-
stituer des méthodes universelles pour traiter tous les problèmes d’analyse des
mesures d’auscultation des barrages, ces méthodes nécessitent une technicité et
une expertise importantes et doivent être utilisées avec rigueur et précaution.
SUMMARY
Dam safety largely relies on monitoring its behavior throughout its life. Many
methods exist to build a statistical model of this behavior which consists in estab-
lishing a predictive model of a measure based on several explanatory variables. The
approach most widely used by the profession is the multiple linear regression (MLR).
However, MLR shows its limits when the population is not statistically homogeneous
or when the modeled phenomena are strongly non-linear (such as threshold effects).
The article presents three types of machine learning models and compares their
respective performances on the basis of three applications: the modeling of the
displacements of a gravity dam, the leakage rates of an arch dam and the piezo-
metric levels in the core of an embankment dam. Our analyzes show that machine
learning methods are likely to provide useful contributions to the statistical analyses,
especially when classical multiple linear regression models have low coefficients of
determination. Non-linear regression models (SVR) have shown ease of use and
robustness. Artificial neural network models (ANN) have demonstrated their effi-
ciency in considering strong non-linearities. Random forest (RF) models have been
able to identify discontinuities over time. Far from constituting universal methods
able to deal with all issues of dam monitoring data analysis, these methods require
significant technical skills and expertise and must be used with rigor and caution.
1. INTRODUCTION
La sécurité d’un barrage repose sur le suivi de son comportement tout au

long de sa vie. Ce suivi se fonde sur des données d’auscultation qui sont analysées
afin de détecter les indices de son vieillissement ou l’apparition de pathologies. Les
sollicitations auxquelles l’ouvrage est soumis sont variables dans le temps (niveau
de la retenue, température, piézométrie, etc.) et peuvent avoir des effets réversibles
ou irréversibles. Il est particulièrement important de détecter assez tôt des derniers,
qui peuvent être la manifestation d’un mécanisme de défaillance, pour mettre en
256
Q.106 - R.17
œuvre d’éventuelles actions correctives. Il est donc nécessaire de disposer d’outils

statistiques capables de détecter des évolutions irréversibles du comportement du
barrage indépendamment des sollicitations auxquelles il est soumis.
De nombreuses méthodes existent pour établir un modèle statistique de com-

portement d’un barrage [1]. Les modèles de régression multilinéaires [2] permettent
de soustraire aux mesures brutes les influences de la charge hydrostatique, des
saisons, de la température ou de la pluviométrie pour mettre en évidence une
éventuelle dérive temporelle irréversible du comportement de l’ouvrage. Ces méth-
odes statistiques sont couramment mises en œuvre dans le domaine des barrages
et ont démontré leur pertinence et leur robustesse pour une majorité des situa-
tions. Néanmoins, les modèles de régression multilinéaire présentent certaines
limites intrinsèques à la méthode: i) les différentes composantes du modèle ne
sont généralement pas strictement indépendantes (niveau de la retenue et saison
par exemple), ii) les lois d’évolution des composantes sont postulées a priori, iii)
ces modèles ne prennent généralement pas en compte de couplage entre les dif-
férentes composantes explicatives et iv) cette approche prend mal en compte les
fortes non-linéarités ou les effets de seuil qui peuvent exister.
Afin de répondre à ces limites, nous mettons en œuvre trois méthodes

d’apprentissage supervisé basées respectivement sur les modèles de réseaux de
neurones, les modèles de régression à vecteurs supports et les modèles de forêts
aléatoires. Ces approches ont l’avantage de pouvoir prendre en compte la complex-
ité des relations entre les variables explicatives et les variables à expliquer sans faire
d’hypothèse supplémentaires sur les non-linéarités, les couplages ou les effets de
seuils.
Dans la littérature, cette approche a déjà été employée dans le cadre

d’analyses du comportement mécanique de barrages voûtes. L’essentiel de ces
applications concerne la modélisation de barrages voûtes en termes de déplace-
ments [3,4,1,5], de piézométrie [6] ou de débits de drainage [7,1]. Les applications à
des barrages poids ou en remblais sont rares [8,9,10] et ne concernent pas l’analyse
des débits de fuite ou de drainage. Notre article présente plusieurs exemples de
modèles à base de réseaux de neurones pour des types de barrages variés (poids,
voûte et remblais). Les performances de ces modèles sont comparées avec celles
d’un modèle de régression multilinéaire pour la prédiction de plusieurs types de
variables (déplacement, débit de fuite, piézométrie). Les difficultés liées à la mise
en œuvre de la méthode (conditionnement de la base d’apprentissage, choix de
l’architecture, méthodologie de validation croisée, calage des hyperparamètres,
etc.) sont discutées en partie 6. Une méthodologie d’implémentation dans un logiciel
d’analyse de données d’auscultation pour le diagnostic des mesures est proposée.
Nous soulignons ici que l’objet du présent travail est de comparer les méth-
odes pour un même jeu de données. Nous avons donc volontairement écarté les
techniques de découpage temporel a priori des périodes temporelles (pratiqué
couramment en analyse multilinéaire lorsque la population initiale n’est pas statis-
tiquement homogène) ou de ré-échantillonnage des données d’apprentissage. Par
ailleurs, bien que des effets retards puissent être pris en compte dans le choix des
257
Q.106 - R.17
variables explicatives, nous avons ici choisi de ne considérer que des variables
explicatives concomitantes à la variable à expliquer.
2. ETAT DE L’ART
Un modèle explicatif a pour objet de prédire la valeur d’une variable à expliquer

(Y ) en fonction d’une ou plusieurs variables explicatives (X1 , . . . Xp ). Du point de vue
formel:
Y = f (X1 , . . . , Xp ) + ε (1)
où f est une fonction représentative de la réalité physique du barrage et dépend

du phénomène à expliquer (déplacement, débit de fuite) et l’erreur non prise en
compte par le modèle. L’idée n’est pas ici de décrire précisément les phénomènes
physiques, qui sont seuls accessibles par le biais de modèles hydromécaniques
complexes nécessitant l’introduction d’hypothèses supplémentaires et un gros effort
computationnel, mais de construire un modèle statistique permettant de relier entre
elles les différentes variables.
La difficulté provient du fait que le problème est généralement non linéaire

et que les variables explicatives observables ne sont pas nécessairement indépen-
dantes. On justifie toutefois l’indépendance des variables explicatives en consid-
érant des variables représentatives de phénomènes physiques dont on admet
une indépendance suffisante. Les variables explicatives ayant un effet réversible
généralement considérées sont classiquement: la cote de la retenue (Z), l’angle
saisonnier (θ) et parfois la température de l’air extérieur (T) ou la pluviométrie (P).
L’ensemble des effets irréversibles dus au vieillissement du barrage sont pris en
compte à travers la variable temps (t).
2.1. MODÈLES DE RÉGRESSION LINÉAIRE MULTIPLE (MLR)
Dans les modèles de régression multilinéaire (Multiple Linear Regression), la

réponse du système est supposée être une superposition additive des réponses à
chacune des variables explicatives. La non-linéarité des phénomènes est traitée
en supposant que f est une somme de fonctions non linéaires des variables
explicatives. Ainsi, on a:
Y = EZ + Eθ + Et + ε (2)
258
Q.106 - R.17
avec
EZ = a1 Z + a2 Z 2 + a3 Z 3 + a4 Z 4
Eθ = b1 cos θ + b2 sin θ + b3 cos 2θ + b4 sin 2θ
Et = c1 t + c1 e−t/t0
où les dix paramètres ai , bi et ci sont des constantes à déterminer par régres-

sion linéaire multiple. Le paramètre t0 qui représente l’amortissement de la dérive
irréversible du phénomène à expliquer est généralement déduit de manière itérative.
La qualité de la régression est caractérisée par le biais du coefficient de détermina-
tion (R 2 ) – appelé aussi coefficient d’explication – qui représente la part de variance
expliquée par le modèle par rapport à la variance totale des mesures observées.
Il s’agit d’une mesure de la capacité d’apprentissage statistique du modèle. Un
coefficient R 2 = 0,8 signifie que 80% de la dispersion des mesures observées est
expliquée par le modèle. Le modèle fournit des valeurs de poids explicatif à chaque
variable.
2.2. MODÈLES DE RÉGRESSION À VECTEURS DE SUPPORT (SVR)
La méthode SVR (Support Vector Regression) [11] est une méthode adaptée
à la résolution de problèmes de régression dérivée des méthodes SVM (Support
Vector Machine) utilisées pour les problèmes de classification. Il s’agit de trans-
poser les données d’apprentissage dans un espace de plus grande dimension (par
l’intermédiaire d’une fonction noyau) et de construire la fonction de régression sur
certaines seulement des données transposées (vecteurs supports). La fonction de
prédiction (ou régression) postulée est de la forme:
n
f (x) = (αi − αi∗ )K(xi , x) + b (3)
i=1
où b est le terme de base, K est une fonction noyau et les coefficients αi , α∗i sont
déterminés en minimisant une fonction de coût. La mise en œuvre de l’algorithme
dépend de deux hyper-paramètres: un paramètre de sensibilité ε correspondant
à la marge de tolérance admise sur l’erreur de la régression et un paramètre de
régularisation C pénalisant les erreurs. L’effet de la tolérance aux erreurs (ε) est alors
plus ou moins accentué par le paramètre C au cours du processus d’apprentissage.
Une valeur trop élevée peut conduire à un sur-apprentissage alors qu’une valeur
trop faible rend difficile l’assimilation des données d’entrée.
Les p variables explicatives d’entrée et la base d’apprentissage de la méthode

mise en œuvre ici sont les mêmes que pour les modèles de réseaux de neurones. Le
noyau est de type radial (i.e. Gaussien), les autres hyperparamètres sont déterminés
par des techniques de validation croisée.
259
Q.106 - R.17
2.3. MODÈLES DE FORÊTS ALÉATOIRES (RF)
Les arbres de décision permettent de partitionner l’espace des variables

explicatives en groupes homogènes. Les forêts aléatoires [12] consistent à faire
croître un certain nombre d’arbres de décision. A chaque embranchement de l’arbre,
une fonction de segmentation fait intervenir une fonction non linéaire nv des p
variables initiales. Chaque arbre est optimisé au regard de la sélection des vari-
ables associées à chaque fonction de segmentation. Les prévisions obtenues par
chaque arbre sont ensuite moyennées, ce qui permet de réduire la variance des
prédictions.
Pour réduire la corrélation entre les arbres, chaque arbre est construit aléa-
toirement à partir d’une partie seulement des données. Les p variables explicatives
d’entrée et la base d’apprentissage de la méthode mise en œuvre ici sont les mêmes
que pour les modèles de réseaux de neurones.
Les hyperparamètres de cette méthode sont: le nombre d’arbres de la forêt

(na ) et le nombre de variables tirées au sort à chaque embranchement (nv ). Plus
le nombre d’arbres est important, plus le nombre de classes finales augmente. Il
est cependant important de limiter le nombre d’arbres pour avoir suffisamment de
points dans chaque classe.
2.4. MODÈLES DE RÉSEAUX DE NEURONES ARTIFICIELS (ANN)
Un modèle de réseau de neurones (Artificial Neural Network) est un modèle

mathématique constitué d’un assemblage d’éléments (perceptrons) interconnectés
et organisés en couches. Il permet de calculer les composantes d’un vecteur de
variables de sortie (couche de sortie) en fonction d’un vecteur de variables d’entrée
(couche d’entrée).
Chaque perceptron transforme un signal d’entrée (somme pondérée des

valeurs données par les perceptrons de la couche précédente plus un éventuel
biais) en un signal de sortie selon une fonction d’activation non linéaire. Si l’on note
(x1 , . . . , xd ) les variables d’entrée, bl la valeur du biais, l’ensemble des entrées d’un
perceptron affectées d’un poids (w1 , . . . , wd ), zl la valeur en sortie du perceptron et
g la fonction d’activation, on a:

d
zj = g wj xj + bl (4)
j=1
Un réseau de neurones est défini par son architecture (nombre de couches,

nombre de perceptrons par couche), par l’ensemble des poids affectés à chaque
260
Q.106 - R.17
connexion et des biais affectés à chaque neurone. Le calibrage des poids neuronaux
est réalisé par un algorithme de rétro-propagation de l’erreur à partir d’une base
d’apprentissage constituée d’une partie de la base de données. La validation du
modèle est réalisée sur la base de test constituée par l’autre partie de la base de
données.
Avant de procéder à la phase d’apprentissage, il convient de réaliser les étapes

préliminaires suivantes:
• Choisir les variables explicatives d’entrée pertinentes;

• Choisir une architecture du réseau de neurones adéquate (nombre de couches
cachées et de perceptrons par couche) en conciliant simplicité, vitesse de
convergence de l’apprentissage et interprétabilité;
• Conditionner la base d’apprentissage afin de nettoyer les données erronées
et d’homogénéiser les distributions des variables d’entrée et de sortie.
Choix des variables explicatives d’entrée. Les réseaux de neurones sont

capables de reproduire toute fonction non linéaire. Il est donc inutile d’introduire des
fonctions dérivées des variables explicatives pour tenir compte d’une éventuelle non-
linéarité. Nous nous limitons donc aux p variables explicatives suivantes: temps (t),
cote de retenue (Z) et angle saisonnier (sinθ et cosθ, afin de garantir la périodicité
de l’influence de la saison et la bijectivité). Cette parcimonie dans le choix des
variables d’entrée nous permettra d’améliorer la robustesse du modèle et de faciliter
la convergence de l’apprentissage.
Choix de l’architecture du réseau. Nous nous plaçons donc ici dans le cas
d’un réseau comprenant trois variables explicatives en entrée et une seule variable
à expliquer. Le choix de l’architecture consiste ainsi à choisir le nombre de couches
cachées et de neurones pour chacune d’entre-elles. Il convient de trouver le nom-
bre de perceptrons optimum pour concilier complexité de la réponse modélisée et
robustesse de la convergence de l’apprentissage. Un nombre trop grand de per-
ceptrons ne permet pas de retranscrire la relation entre observations et prédictions.
On obtient alors un modèle «ininterprétable». Un nombre de perceptrons trop petit
conduit à un sur-apprentissage non transposable à la population test (difficultés à
prédire la réponse en-dehors du domaine de la base d’apprentissage).
L’architecture du modèle de réseau de neurone a donc été déterminée par val-

idation croisée en maximisant le coefficient de détermination moyen sur 20 tirages
aléatoires de Napp vecteurs d’apprentissage et Ntest vecteurs de test. Les deux con-
traintes imposées au réseau ont été celles d’avoir entre 1 et 3 couches avec au moins
3 perceptrons par couche (hyperparamètre m) et de limiter le nombre maximal de
poids neuronaux à déterminer à Napp /3.
Base d’apprentissage. Il convient dans un premier temps de choisir la dimen-

sion de la base d’apprentissage Napp . Dans ce qui suit, nous avons choisi Napp = 3/4
261
Q.106 - R.17
Tableau 1
Principaux hyperparamètres associés à chaque méthode
d’apprentissage automatique
Main hyperparameters associated to each learning
machine method
HYPERPARAMÈTRES DÉPLACEMENTS DÉBITS DE FUITE NIVEAUX PIÉZOMÉTRIQUES
BARRAGE POIDS BARRAGE VOÛTE BARRAGE EN REMBLAI
Base de données
p 4 4 4
N 641 1953 486
Napp 450 1500 300
Ntest 150 300 100
Npred 41 153 86
SVR
K radial radial radial
ε 0,1 0,1 0,1
C 3 3 2
RF
na 100 100 80
nv 2 2 2
ANN
g sigmoïde sigmoïde sigmoïde
m (4,4) (6,6) (4,8,4)
nrep 6 5 10
N (avec N la dimension totale de la base). Contrairement aux problèmes plus clas-

siques de reconnaissance d’image, la taille de la population d’apprentissage est
assez limitée ce qui nécessite de repasser plusieurs fois la base d’apprentissage
lors de la descente de gradient stochastique afin d’atteindre les couches les plus
profondes. Le nombre de répétitions nrep est un des hyperparamètres à déterminer
pour chaque base de données considérée.
3. DÉPLACEMENTS D’UN BARRAGE POIDS
3.1. PRÉSENTATION SUCCINCTE DU BARRAGE ET DES MESURES
Le barrage considéré est de type poids, d’une centaine de mètres de hauteur

et achevé à la fin des années 1940. Il a été équipé à l’origine de pendules directs
sur la hauteur des plots, remplacés après une vingtaine d’années par des pend-
ules superposés. Les mesures considérées ici sont celles des déplacements dans
la direction amont-aval enregistrés par le pendule supérieur de l’un des plots cen-
traux de plus grande hauteur. On dispose de 48 années d’observations soit, après
nettoyage, 641 mesures à pas de temps mensuel.
262
Q.106 - R.17
Fig. 1
Séries chronologiques des mesures de déplacement amont-aval (D) et de la cote
de retenue (Z)
Time series of the upstream/downstream movements (D)
and of the reservoir level (Z)
3.2. ANALYSE DES MESURES DE DÉPLACEMENT
L’exploitation de la retenue se caractérise par un marnage modéré et des

abaissements significatifs périodiques pour favoriser la transparence sédimentaire.
Les mesures brutes mettent en évidence une alternance de périodes avec des
variations saisonnières autour d’une position stable et de périodes où l’ouvrage
tend à se déplacer légèrement vers l’amont, le passage d’une période à l’autre
coïncidant généralement avec une vidange partielle.
La calibration et la validation des modèles ont été effectuées sur la base des
45 premières années ( j ≈ 16400 j) et leurs capacités de prédiction ont été évaluées
sur les 3 dernières années (Figure 1). Les coefficients de détermination obtenus
par les différentes méthodes varient entre 0,90 et 0,92. Les ajustements obtenus
par l’ensemble des méthodes sont donc bons.
La Figure 2 représente les valeurs calculées par chacune des méthodes mises
en œuvre en fonction des valeurs mesurées et l’influence de chaque variable explica-
tive sur le déplacement, les autres variables explicatives étant constantes (égales
à leur valeur médiane). L’influence de la cote de la retenue perçue par les modèles
est assez modérée, inférieure à 0,5 mm pour 10 m d’augmentation du niveau d’eau.
Les modèles ANN, MLR et SVR s’accordent sur la même tendance, plus marquée
pour le modèle SVR. Le modèle RF identifie une influence négligeable du niveau de
la retenue jusqu’à la cote 327 m suivie d’une marche d’escalier d’amplitude 0,1 mm
environ.
263
Q.106 - R.17
L’influence saisonnière est plus marquée et varie en fonction de la méthode.

D’une amplitude de 0,5 mm pour le modèle SVR, elle passe à 1 mm pour le modèle
RF, puis à 1,5 mm pour le modèle MLR et enfin 2,5 mm pour le modèle ANN et en
phase pour ces trois modèles. Il est intéressant de noter que le modèle RF montre
des discontinuités significatives, suggérant l’intégration d’un effet temporel dans la
composante saisonnière (marnage saisonnier de la retenue).
Fig. 2
Résultats obtenus avec les différents modèles
Results obtained from the different models
264
Q.106 - R.17
Le modèle SVR ne perçoit pratiquement pas l’influence de la dérive temporelle

alors que les trois autres modèles montrent des tendances analogues et du même
ordre de grandeur (entre 2 et 3 mm en 40 ans). Le modèle RF montre une évolution
irrégulière dont certaines accélérations semblent corrélées avec des abaissements
de cote de retenue importants. Il est intéressant de noter (Figure 3) que les inflexions
de la courbe non explicables par l’abaissement de la retenue correspondent à des
étés plus chauds que la moyenne à l’été 1983 (t ≈ 5600 j), 1991 (t ≈ 8500 j) et 2003
(t ≈ 13 000 j).
Fig. 3
Dérive temporelle de D (diamants oranges) du modèle de forêts aléatoires (RF) et
cote de retenue Z (courbe bleue)
Time drift D (orange diamonds) from random forest model (RF) and time series of
the reservoir level Z (blue curve)
Les capacités prédictives des différentes méthodes sont évaluées sur la base
d’une prédiction des valeurs de déplacements sur les trois dernières années de la
série (Figure 2, en bas). Les résultats montrent des comportements analogues pour
les 4 modèles. Les modèles les plus performants semblent être les modèles SVR et
RF, mais aucun d’entre eux n’a été capable de reproduire l’amplitude des sommets
ni du troisième creux de la courbe de mesures.
4. DÉBITS DE FUITE D’UN BARRAGE VOÛTE
Le barrage d’une hauteur de 26 m a été achevé à la fin des années 1940. La

voûte à parement amont cylindrique est appuyée sur deux culées. Des fuites sont
265
Q.106 - R.17
mesurées au pied aval de la culée rive gauche qui proviennent d’une reprise de
bétonnage fuyarde. On dispose de 40 années d’observations soit, après nettoyage,
1953 mesures à pas de temps hebdomadaire.
4.2. ANALYSE DE MESURES DE DÉBITS DE FUITE
Les séries temporelles de la cote de la retenue (Z) et du débit de fuite en

rive gauche (Q) montrent des variations importantes au cours du temps (Figure 4).
Jusqu’à 1984 (i.e. t = 2200 j environ), les débits maximums sont compris entre 8
et 18 l/min et tendant à augmenter dans le temps. Ces observations ont conduit à
injecter la culée. Les débits de fuite ont diminué de manière importante les années
suivantes puis ont augmenté jusqu’en 1993 (i.e. t = 6000 j) pour se stabiliser autour
de valeurs de pic comprises entre 5 et 10 l/min.
Fig. 4
Mesures du débit de fuite en rive gauche (Q) en fonction du temps (t) et de la cote
de la retenue (Z)
Leakage flow rate on left bank (Q) as a function of time (t) and reservoir level (Z)
La représentation du débit de fuite en fonction de la cote de la retenue mon-

tre une corrélation marquée. Elle met également clairement en évidence un effet
de seuil autour de la cote d’une reprise de bétonnage située vers 256,5 m NGF.
Toutefois, il est difficile d’établir un modèle mathématique a priori sur la base des
mesures brutes. Il est donc nécessaire de recourir aux outils d’analyse statistique
des données.
sur les 3 dernières années (Figure 5). Les coefficients de détermination obtenus
par les différentes méthodes varient entre 0,80 et 0,92. Les meilleurs ajustements
sont obtenus par les méthodes ANN et RF (R 2 = 0,92) alors que les méthodes SVR
et MLR donnent des ajustements médiocres (R 2 = 0,81 et 0,80 respectivement).
266
Q.106 - R.17
La méthode MLR a tendance à surestimer le débit pour des débits faibles et

à le minimiser pour les valeurs fortes. La méthode SVR ne permet pas de rendre
compte des observations et a tendance à surestimer les débits. Les méthodes RF
et ANN permettent d’obtenir un nuage de points plus resserré centré sur la première
bissectrice mais la méthode ANN a l’inconvénient de calculer des valeurs de débit
négatives, contrairement à la méthode RF.
Fig. 5
267
Q.106 - R.17
Toutes les méthodes s’accordent sur le caractère faible de l’influence saison-

nière pour une cote égale à la valeur médiane 258 m NGF. Il est à noter que, dans le
cas approches par apprentissage, cette observation ne peut être extrapolée à des
cotes de retenue différentes. Par exemple, les analyses que nous avons menées
pour une cote de retenue égale à 256 m NGF montrent que l’influence saisonnière
devient négligeable.
L’effet de la cote de la retenue à saison moyenne sur le débit de fuite est égale-
ment bien retranscrit à travers les modèles. Un effet de seuil est bien observé entre
les cotes Z = 256 m et Z = 257 m (correspondant à la cote de la reprise de béton-
nage fuyarde). Il ne semble pas influencé par la saison. Le modèle MLR conduit
cependant à une relation parabolique peu physique pour Z < 255 m. Cela pourrait
s’expliquer par une compensation de l’effet saisonnier qui diminue en hiver à cote
de retenue basse. Les autres modèles donnent une relation très proche jusqu’à
Z = 259 m et divergent ensuite, la méthode ANN donnant la relation la plus majo-
rante. Il est intéressant de noter que les trois méthodes suggèrent l’existence d’un
palier au-delà duquel la cote de la retenue a un effet plus limité. Ce constat pourrait
s’expliquer par une refermeture de la fissure pour les cotes de retenue hautes.
Concernant la dérive temporelle, les méthodes ANN et RF mettent en évi-

dence deux parties distinctes délimitées par un artéfact, la première entre t = 0 et
t = 2000 j marquée par des variations importantes avant injection de l’ouvrage et la
seconde témoignant d’une tendance à la stabilité. La méthode SVR retranscrit ces
deux tendances, mais sans faire apparaître de rupture dans le comportement du
barrage.
La capacité prédictive de chaque modèle peut être appréciée sur la Figure 5

(en bas). On constate que tous les modèles reproduisent les observations qualita-
tivement. Le modèle MLR a cependant tendance à réduire l’amplitude des variations
en majorant les faibles débits. Le modèle ANN donne des résultats proches des
mesures, y compris en termes de prédiction des pics, mais la composante de dérive
temporelle à la baisse le conduit à extrapoler des valeurs de débits systématique-
ment supérieures aux valeurs mesurées. Les approches SVR et RF semblent quant
à elles plus robustes. Elles ne présentent pas de dérive temporelle et démontrent une
bonne capacité à reproduire les pics. Il est possible que l’aptitude de ces méthodes
à produire des classifications leur permette de distinguer clairement deux périodes
aux comportements différenciés. Par ailleurs, ces deux modèles ne prédisent pas
de valeurs de débits négatifs, contrairement au modèle ANN.
Pour la présente application, c’est la méthode RF qui conjugue les meilleurs

résultats en termes de capacités d’explication des mesures observées et de prédic-
tion des débits en extrapolation. Le modèle ANN permet d’identifier un changement
de comportement de l’ouvrage mais a une capacité prédictive en extrapolation
inférieure. Le modèle SVR possède de bonnes capacités prédictives mais s’avère
incapable de séparer clairement les deux périodes de comportement du barrage.
Le modèle MLR apparait moins performant pour intégrer le comportement à seuil
(observé en fonction du niveau de la retenue).
268
Q.106 - R.17
5. PIÉZOMÉTRIE DANS LE NOYAU D’UN BARRAGE EN REMBLAI
Le barrage considéré, construit à la fin des années 1970, est un barrage en

enrochements à noyau argileux étanche de 27 m de hauteur sur terrain naturel.
Les mesures piézométriques considérées sont celles enregistrées par une

cellule à corde vibrante située dans le tiers aval du noyau, environ 12 m sous le
niveau de la crête du barrage. La série chronologique s’étend sur environ 20 ans
avec un pas de temps compris entre 15 et 30 jours, soit 487 mesures au total après
nettoyage des données. Comme précisé précédemment, nous avons ici choisi de
ne considérer que des variables explicatives concomitantes à la variable à expliquer.
5.2. ANALYSE DE MESURES PIÉZOMÉTRIQUES
La cote de la retenue évolue autour de sa valeur normale (Z = 51 m), présen-

tant généralement un faible marnage, avec des abaissements plus ou moins
importants en fonction des années (Figure 6). Le niveau piézométrique varie entre
40,7 m et 42,7 m et montre une certaine corrélation entre les valeurs extrêmes
de cote de retenue et les valeurs extrêmes de niveau piézométrique. L’analyse
qualitative de ces courbes semble suggérer une légère dérive temporelle, mais les
Fig. 6
Séries chronologiques des mesures de cote de la retenue Z (courbe bleue)
et de la piézométrie mesurée par la cellule c11 (courbe noire).
Time series of the reservoir level Z (blue curve)
and of the piezometric c11 cell measurements (black curve).
269
Q.106 - R.17
fluctuations observées semblent plutôt associées à l’effet de la cote de la retenue.

Les effets saisonniers ne sont pas particulièrement prononcés.
sur les 3 dernières années. Les coefficients de détermination obtenus par les dif-
férentes méthodes sont assez variables (Figure 7). Les meilleurs ajustements sont
Fig. 7
270
Q.106 - R.17
obtenus par les méthodes SVR et RF (R 2 = 0,91) alors que les méthodes MLR et
ANN donnent des ajustements médiocres (R 2 = 0,78 et 0,79 respectivement).
Les effets saisonniers sont faibles pour l’ensemble des méthodes (moins de
20 cm). Les courbes représentant l’influence de la cote de la retenue présentent des
caractéristiques assez proches (avec toutefois une variation en marches d’escalier
pour RF). Elles sont quasiment confondues pour Z supérieur à 50 m. Les courbes
ANN, SVR et RF identifient un effet de seuil plus ou moins marqué aux-alentours de
Z = 46–48 m. Les courbes représentant la dérive temporelle des mesures présentent
toutes une tendance décroissante dont les pentes moyennes sont du même ordre
de grandeur. Les courbes de dérive temporelle obtenues par les méthodes SVR et
ANN sont très proches. La courbe correspondant à la méthode SVR montre une
inflexion aux alentours de t ≈ 4500 j dont la signification physique semble douteuse.
La méthode RF identifie une évolution discontinue, en marches d’escalier. Sur la
Figure 8, on constate que les ruptures de pente de la courbe sont concomitantes aux
deux vidanges de la retenue (t ≈ 500 j et t ≈ 2300 j) et aux abaissements importants
(Z ≤ 45 m) survenus au cours de la vie de l’ouvrage.
Fig. 8
Tendance irréversible obtenue par la méthode RF et série
chronologique des mesures de la cote de retenue (Z)
Time drift from RF method and time series of the reservoir level (Z)
La capacité prédictive des modèles est testée sur les trois dernières années de
la série (Figure 7, en bas). Qualitativement, on constate des tendances analogues
entre les mesures et l’ensemble des modèles (qui montrent des oscillations plus
ou moins marquées). Les modèles MLR et ANN prédisent en moyenne des valeurs
inférieures aux mesures. La méthode SVR suit la tendance moyenne des mesures
mais présente des oscillations plus importantes. La méthode RF permet quant à elle
271
Q.106 - R.17
de prédire l’évolution du niveau piézométrique avec une bonne pertinence. Cela sug-
gère que le modèle représentatif établi pour 3800 j < t < 5600 j reste assez pertinent
pour prédire le comportement de la piézométrie pour 5600 j < t < 6700 j.
6. DISCUSSION
Trois méthodes d’apprentissage automatique basées sur des modèles de

réseaux de neurones, de régression à vecteurs supports et de forêts aléatoires
ont pu être mises en œuvre et appliquées à l’analyse des mesures de déplace-
ment d’un barrage poids, de fuite d’un barrage voûte et de piézométrie dans le
noyau d’un barrage en remblai, chaque série chronologique comportant entre 500
et 2000 mesures. Les résultats ont été confrontés à ceux de la méthode classique
de régression linéaire MLR. Les modèles ont été construits à partir d’une base
d’apprentissage obtenue par tirage aléatoire d’une partie des mesures de la série
chronologique. La capacité des modèles à reproduire les mesures de la base de test
(partie des mesures non utilisées pour la construction des modèles) a été étudiée.
De plus, les capacités prédictives de chacune de méthodes ont été évaluées sur
les 3 dernières années des séries chronologiques étudiées.
Nos analyses mettent en évidence les limites des modèles MLR qui peinent à
reproduire les fortes non-linéarités. En particulier, la médiocrité de l’apprentissage
statistique de la méthode MLR s’explique principalement par la forme prédéfinie
et la nature continue des fonctions utilisées. Lorsque les effets non linéaires ont
une importance limitée et que la population initiale est suffisamment homogène
statistiquement, les capacités prédictives sont comparables aux autres modèles.
Les modèles SVR permettent de tenir compte d’effets non linéaires (effets de
seuil notamment) et ne nécessitent pas d’efforts d’implémentation importants.
Les modèles ANN sont aptes à tenir compte des effets non linéaires
mais également à identifier les périodes de comportement distinctes (popula-
tions statistiquement non homogènes). Nos résultats montrent cependant que
les modèles ANN ne sont pas nécessairement les plus performants lorsque la
dimension de la base d’apprentissage est insuffisante. Une amélioration sensible
de leur performance est cependant envisageable en utilisant des techniques de
ré-échantillonnage ignorées dans le cadre du présent travail.
Les modèles RF se sont avérés être particulièrement efficaces pour détecter

les changements de comportement brutaux. Ils permettent d’identifier clairement les
effets de seuil et tiennent naturellement compte des non-linéarités. Sur les appli-
cations présentées, ces modèles ne semblent pas être pénalisés en termes de
capacités prédictives.
272
Q.106 - R.17
Il est bon de rappeler que, quels que soient les modèles statistiques utilisés, ils
restent sensibles à la qualité des données et aux biais d’échantillonnages (erreurs
de mesures, populations trop petites, distributions statistiquement non homogènes).
La sélection des données d’entrée, le choix de variables explicatives ainsi que les
méthodes de ré-échantillonnage peuvent permettre de corriger nombre de ces biais
et d’améliorer la confiance de l’ingénieur dans les résultats de ses analyses.
7. CONCLUSIONS ET PERSPECTIVES
Notre travail a montré que les méthodes d’apprentissage automatique sont

susceptibles d’apporter des compléments intéressants pour l’analyse des données
d’auscultation des barrages, et cela notamment lorsque les modèles classiques de
régression linéaire ont de faibles coefficients de détermination. Il revient à l’ingénieur
en charge de l’interprétation des mesures d’exercer son jugement pour apprécier
la pertinence de l’utilisation de ces méthodes, qui nécessitent une technicité et une
expertise importantes.
Un prototype de plateforme logicielle est en cours de développement dans

laquelle les méthodes présentées ici sont préprogrammées. L’utilisateur pourra ainsi
appliquer ces méthodes à une série ou plusieurs séries de mesures et en apprécier
la pertinence comme aide à l’analyse en fonction des résultats (coefficient de déter-
mination, courbes d’influence des variables explicatives). Il sera également possible
de comparer les mesures réelles et les mesures prédites par les modèles en vue
d’un diagnostic des mesures.
REFERENCES
[1] SALAZAR F., TOLEDO M.A., ONATE E., MORAN R. (2016) An empirical
comparison of machine learning techniques for dam behaviour modelling,
Structural Safety, 56, pp. 9–17. Doi: 10.1016/j.strusafe.2015.05.001.
[2] Swiss Committee on Dams. ICOLD (2003). Methods of analysis for the
prediction and the verification of dam behaviour.
[3] MATA J. (2011). Interpretation of concrete dam behaviour with artificial neu-
ral network and multiple linear regression models, Engineering Structures,
Volume 33, Issue 3, Pages 903–910. Doi: 10.1016/j.engstruct.2010.12.011.
[4] RADOVANOVIC S., MILIVOJEVIC V., CIROVIC V., DIVAC V., MILIVOJEVIC
N. (2015). Prediction of Concrete Dam Deformation using Artificial Neural
Networks, In: Proceedings of the Fourth International Conference on Soft
273
Q.106 - R.17
Computing Technology in Civil, Structural and Environmental Engineering,

p. 1–11.
[5] BELMOKRE A., MIHOUBI M., SANTILLAN D. (2019). Analysis of Dam Behav-
ior by Statistical Models: Application of the Random Forest Approach, KSCE
Journal of Civil Engineering, 23(11), p. 4800–4811. Doi: 10.1007/s12205-019-
0339-0.
[6] SIMON A., ROYER M., MAURICE F., FABRE J.P. (2012). Analyse des mesures
d’auscultation des barrages avec les réseaux de neurones, In Colloque Aus-
cultation des barrages et des digues – Pratiques et perspectives. Colloque
CFBR, Chambéry, 27–28 novembre 2012, p. 314–328.
[7] SANTILLAN D., FRAILE-ARDANUY J., Toledo M.A. (2013). Dam seepage
analysis based on artificial neural networks: The hysteresis phenomenon, The
2013 International Joint Conference on Neural Networks (IJCNN), Dallas, TX,
2013, pp. 1–8. Doi: 10.1109/IJCNN.2013.6707110.
[8] KIM Y., KIM B. (2008). “Prediction of relative crest settlement of concrete-faced
rockfill dams analysed using an artificial neural network model.” Computers
and Geotechnics, Vol. 35, No. 3, pp. 313–322, Doi: 10.1016/j.compgeo.
2007.09.006.
[9] TAYFUR G., SWIATEK D., WITA A., and SINGH V. (2005). “Case study: Finite
element method and artificial neural network models for flow through Jeziorsko
earthfill dam in Poland.” Journal of Hydraulic Engineering, Vol. 131, No. 6,
pp. 431–440, Doi: 10.1061/(ASCE) 0733-9429(2005)131:6(431).
[10] SALAZAR F., MORAN R., TOLEDO M. (2017). Data-Based Models for the
Prediction of Dam Behaviour: A Review and Some Methodological Consid-
erations, Archives of Computational Methods in Engineering, 24(1), p. 1–21.
Doi: 10.1007/s11831-015-9157-9.
[11] VAPNIK V. (1995). The Nature of Statistical Learning Theory, Springer.
[12] BREIMAN L. (2001). “Random Forests”. In: Mach. Learn. 45.1 (2001),
pp. 5–32. DOI: 10.3390/rs10060911.
274
Q.106 - R.17
1. INTRODUCTION
The safety of a dam is based on monitoring its behaviour throughout its life.
This is based on monitoring data which is analysed to detect evidence of ageing
or the appearance of pathologies. The stresses to which the structure is subjected
are variable over time (reservoir level, temperature, piezometry, etc.) and can have
reversible or irreversible effects. Irreversible changes may signal a failure mech-
anism, so it is particularly important to detect them quite early so that possible
corrective actions can be implemented. It is therefore necessary to have statistical
tools available capable of detecting irreversible changes in the behaviour of the dam,
regardless of the stresses to which it is subjected.
There are many methods available to establish a statistical model of dam

behaviour [1]. Multi-linear regression models [2] enable the influence of hydrostatic
load, seasons, temperature or rainfall to be removed from raw measurements to
show a possible irreversible time drift in the behaviour of the structure. These statis-
tical methods are commonly used with dams and have demonstrated their relevance
and robustness for most situations. However, multi-linear regression models have
some limitations intrinsic to the method: (i) the different components of the model
are generally not strictly independent (e.g., reservoir level and season), (ii) the laws
of evolution of the components are applied a priori, (iii) these models do not gen-
erally take into account the coupling between the different explanatory components
and (iv) this approach does not take into account strong non-linearities or threshold
effects that may exist.
To address these limitations, we have implemented three supervised learning

methods based respectively on neural network models, support vector regression
models, and random forest models. These approaches have the advantage of being
able to take into account the complexity of the relationships between explanatory
variables and response variables without making additional assumptions about non-
linearities, couplings or threshold effects.
In the literature, this approach has already been used in analyses of the
mechanical behaviour of arch dams. Most of these applications concern the mod-
elling of arch dams in terms of movements [3,4,1,5], piezometry [6] or drainage
flows [7,1]. Applications to gravity dams or embankment dams are rare [8,9,10] and
do not concern the analysis of leakage or drainage flows. Our paper presents sev-
eral examples of models based on neural networks, support vector regression and
random forests for various types of dam (gravity, arch and embankment). The per-
formance of these models is compared with that of a multi-linear regression model
for predicting several types of variable (displacement, leakage rate, piezometry).
Difficulties in implementing the method (conditioning the learning base, choice of
architecture, cross-validation methodology, setting of hyperparameters, etc.) are
discussed in part 6. A methodology for implementation in monitoring data analysis
software for measurement diagnostics is proposed.
275
Q.106 - R.17
It is emphasised here that the purpose of this work is to compare methods for
the same data set. We have therefore deliberately discarded techniques involving
a priori time slicing of time periods (commonly used in multi-linear analysis when
the initial population is not statistically homogeneous) or re-sampling of learning
data. In addition, although delay effects can be taken into account in the choice of
explanatory variables, we have chosen to consider only explanatory variables that
are concomitant with the response variable.
2. STATE OF THE ART
An explanatory model is intended to predict the value of a response variable

(Y) based on one or more explanatory variables (X1 , . . . Xp ). Formally:
Y = f (X1 , . . . , Xp ) + ε (1)
where f is a function representative of the physical reality of the dam and depends on
the phenomenon to be explained (movement, leakage rate) and the error not taken
into account by the model. The idea here is not to precisely describe the physical
phenomena, which are only accessible through complex hydromechanical models
requiring the introduction of additional assumptions and a major computational effort,
but rather to construct a statistical model to link the different variables.
The difficulty arises from the fact that the problem is generally non-linear and
that observable explanatory variables are not necessarily independent. However,
the independence of the explanatory variables can be justified by considering vari-
ables representative of physical phenomena that are recognised as being sufficiently
independent. The explanatory variables with a reversible effect generally considered
are classically: the reservoir level (Z), the seasonal angle (θ) and sometimes the out-
door air temperature (T) or rainfall (P). All irreversible effects due to the ageing of
the dam are taken into account through the time variable (t).
2.1. MULTIPLE LINEAR REGRESSION MODELS (MLRS)
In multiple linear regression models, the system response is assumed to be

an additive superposition of responses to each of the explanatory variables. The
non-linearity of phenomena is dealt with by assuming that f is a sum of non-linear
functions of the explanatory variables. Thus, we have:
Y = EZ + Eθ + Et + ε (2)
276
Q.106 - R.17
where
EZ = a1 Z + a2 Z 2 + a3 Z 3 + a4 Z 4
Eθ = b1 cos θ + b2 sin θ + b3 cos 2θ + b4 sin 2θ
Et = c1 t + c1 e−t/t0
where the ten parameters ai , bi and ci are constants to be determined by mul-

tiple linear regression. The parameter t0 , which represents the damping of the
irreversible drift of the phenomenon to be explained, is usually inferred iteratively.
The quality of the regression is characterised by the coefficient of determination
(R 2 ) – also known as the coefficient of explanation – which represents the part of
the variance explained by the model in relation to the total variance of the observed
measurements. This is a measure of the statistical learning capacity of the model. A
coefficient R 2 = 0.8 means that 80% of the dispersion of the observed measurements
is explained by the model. The model provides explanatory weight values for each
variable.
2.2. SUPPORT VECTOR REGRESSION MODELS (SVR)
The Support Vector Regression (SVR) method [11] is an appropriate method

for resolving regression problems derived from the Support Vector Machine (SVM)
methods used for classification problems. This involves transposing learning data
into a higher-dimensional space (through a kernel function) and constructing the
regression function on only some of the transposed data (support vectors). The
postulated prediction (or regression) function is of the form:
n
f (x) = (αi − αi∗ )K(xi , x) + b (3)
i=1
where b is the base term, K is a kernel function and coefficients αi , αi * are determined
by minimising a cost function. The implementation of the algorithm depends on two
hyperparameters: a sensitivity parameter ε corresponding to the tolerance margin
allowed on the regression error and a regularisation parameter C penalising errors.
The effect of error tolerance (ε) is then more or less accentuated by the C parameter
during the learning process. Too high a value can lead to over-learning while too low
a value makes it difficult to assimilate input data.
The input explanatory variables p and the learning base of the method imple-
mented here are the same as for the neural network models. The kernel is radial
(i.e. Gaussian), the other hyperparameters are determined by cross-validation
techniques.
277
Q.106 - R.17
2.3. RANDOM FOREST MODELS (RF)
Decision trees allow the space of explanatory variables to be partitioned into

homogeneous groups. Random forests [12] come from growing a number of deci-
sion trees. At each branch of the tree, a segmentation function involves a non-linear
function nv of the initial p variables. Each tree is optimised for the selection of vari-
ables associated with each segmentation function. Forecasts obtained by each tree
are then averaged, which reduces the variance of predictions.
To reduce the correlation between trees, each tree is constructed randomly

from only part of the data. The input explanatory variables p and the learning base
of the method implemented here are the same as for the neural network models.
The hyperparameters of this method are: the number of trees in the forest (na )
and the number of variables drawn at random at each branch (nv ). The larger the
number of trees, the greater the number of final classes. However, it is important to
limit the number of trees to have enough points in each class.
2.4. ARTIFICIAL NEURAL NETWORK MODELS (ANN)
An Artificial Neural Network is a mathematical model consisting of an assembly

of interconnected and layered elements (perceptrons). It allows the components of
an output variable vector (output layer) to be calculated from an input variable vector
(input layer).
Each perceptron transforms an input signal (a weighted sum of the values

given by the perceptrons of the previous layer plus a possible bias) into an output
signal according to a non-linear activation function. If (x1 , . . . , xd ) are the input vari-
ables, bl the value of the bias, the set of inputs of a perceptron assigned a weight
(w1 , . . . , wd ), zl the output value of the perceptron and g the activation function,
we have:

d
zj = g wj xj + bl (4)
j=1
A neural network is defined by its architecture (number of layers, number of

perceptrons per layer), by all the weights assigned to each connection and the
bias assigned to each neuron. The calibration of neural weights is performed by an
algorithm to reverse-propagate error from a learning base made up of part of the
database. The validation of the model is performed on the basis of a test formed by
the other part of the database.
278
Q.106 - R.17
Table 1
Main hyperparameters associated with each machine learning method
HYPERPARAMETERS GRAVITY DAM ARCH DAM EMBANKMENT DAM
MOVEMENT LEAKAGE RATES PIEZOMETRIC LEVELS
Database
p 4 4 4
N 641 1953 486
Napp 450 1500 300
Ntest 150 300 100
Npred 41 153 86
SVR
K radial radial radial
ε 0.1 0.1 0.1
C 3 3 2
RF
na 100 100 80
nv 2 2 2
ANN
g sigmoid sigmoid sigmoid
m (4,4) (6,6) (4,8,4)
nrep 6 5 10
Before proceeding to the learning phase, the following preliminary steps should
be taken:
• Select relevant input explanatory variables;

• Choose an appropriate neural network architecture (number of hidden lay-
ers and perceptrons per layer), combining simplicity, speed of learning
convergence and interpretability;
• Condition the learning base to clean up erroneous data and to homogenise
distributions of input and output variables.
Choice of input explanatory variables. Neural networks are capable of repli-

cating any non-linear function. It is therefore unnecessary to introduce functions
derived from explanatory variables to take account of possible non-linearity. We
therefore limit ourselves to the following explanatory variables p: time (t), reservoir
level (Z) and seasonal angle (sinθ and cosθ, to guarantee the periodicity of the influ-
ence of the season and bijectivity). This parsimony in the choice of input variables will
allow us to improve the robustness of the model and facilitate learning convergence.
Choice of network architecture. We thus find ourselves with a network com-

prising three explanatory variables as the input and a single response variable. The
choice of architecture thus consists in choosing the number of hidden layers and
of neurons for each of them. It is necessary to find the optimum number of per-
ceptrons to reconcile the complexity of the modelled response and the robustness
of the learning convergence. Too many perceptrons do not allow translation of the
relationship between observations and predictions. We then get an “uninterpretable”
279
Q.106 - R.17
model. Too few perceptrons lead to over-learning that is not transferable to the test
population (difficulties in predicting the response outside the field of the learning
base).
The architecture of the neural network model was thus determined by cross-
validation by maximising the average coefficient of determination over 20 random
runs of Napp learning vectors and Ntest test vectors. The two constraints imposed on
the network were to have between 1 and 3 layers with at least 3 perceptrons per
layer (hyperparameter m) and to limit the maximum number of neural weights to be
determined to Napp /3.
Learning Base. The first step is to choose the dimension of the Napp learning
base. For what follows, we chose Napp = 3/4 N (with N the total dimension of the
base). Unlike more conventional image recognition problems, the size of the learning
population is quite limited, which requires a re-run of the learning base several times
during the descent of the stochastic gradient in order to reach the deepest layers.
The number of nrep repetitions is one of the hyperparameters to be determined for
each database.
3. MOVEMENT OF A GRAVITY DAM
3.1. BRIEF DESCRIPTION OF THE DAM AND THE MEASUREMENTS
The dam in question is of gravity type, about a hundred metres high and com-
pleted in the late 1940s. It was originally equipped with direct pendulums at the top
of the plots, replaced after twenty years by overlapping pendulums. The measures
considered here are those of the upstream-downstream movements recorded by
the upper pendulum of one of the highest central blocks. There are 48 years of
observations or, after cleaning, 641 measurements at monthly intervals.
3.2. ANALYSIS OF MOVEMENT MEASUREMENTS
The operation of the reservoir is characterised by moderate drawdown and

significant periodic level reductions to promote sedimentary transparency. Raw
measurements show an alternation of periods with seasonal variations around a
stable position and periods when the structure tends to move slightly upstream, the
transition from one period to the next usually coinciding with partial drawdown.
The calibration and validation of the models were performed on the basis of
the first 45 years ( j ≈ 16400 days) and their predictive capacities were evaluated over
280
Q.106 - R.17
the last 3 years (Figure 1). The coefficients of determination obtained by the different
methods vary between 0.90 and 0.92. The fits obtained by all methods are therefore
good.
Fig. 1
Time series of the upstream-downstream movements (D)
and of the reservoir level (Z)
Figure 2 shows the values calculated by each of the methods used in relation to
the measured values and the influence of each explanatory variable on movement,
with the other explanatory variables being constant (equal to their median value).
The influence of the reservoir level perceived by the models is quite moderate, less
than 0.5 mm per 10 m increase in water level. The ANN, MLR and SVR models
agree on the same trend, more marked for the SVR model. The RF model identifies
a negligible influence of the reservoir level up to 327 m followed by a step of about
0.1 mm amplitude.
Seasonal influence is more pronounced and varies by method. With an ampli-

tude of 0.5 mm for the SVR, it increases to 1 mm for the RF model, then to 1.5 mm for
the MLR model and finally to 2.5 mm for the ANN model and in phase for these three
models. It is interesting to note that the RF model shows significant discontinuities,
suggesting the integration of a time effect into the seasonal component (seasonal
drawdown of the reservoir).
The SVR model barely perceives the influence of time drift, while the other
three models show similar trends of the same magnitude (between 2 and 3 mm
in 40 years). The RF model shows an irregular evolution, some of which appears
to be correlated with significant reductions in the reservoir level. It is interesting to
note (Figure 3) that the curve inflections that are not explained by the reduction in
the reservoir level correspond to warmer summers than average in the summers of
1983 (t ≈ 5600 days), 1991 (t ≈ 8500 days) and 2003 (t ≈ 13 000 days).
281
Q.106 - R.17
Fig. 2
The predictive capacities of the different methods are evaluated on the basis of
a prediction of the movement values over the last three years of the series (Figure 2,
bottom). The results show similar behaviours for the 4 models. The best-performing
models appear to be the SVR and RF models, but none of them was able to replicate
the amplitude of the peaks or the third trough of the measurement curve.
282
Q.106 - R.17
Fig. 3
Time drift D (orange diamonds) from random forest model (RF) and time series of
the reservoir level Z (blue curve)
4. LEAKAGE RATES OF AN ARCH DAM
The 26 m high dam was completed in the late 1940s. The arch with a cylindrical
upstream face is supported on two piers. Leaks are measured at the downstream
foot of the left bank pier, coming from a leaking lift joint. There are 40 years of
observations or, after cleaning, 1953 weekly measurements.
4.2. ANALYSIS OF LEAKAGE RATE MEASUREMENTS
The time series of the reservoir level (Z) and the leakage rate on the
left bank (Q) shows significant variations over time (Figure 4). Until 1984 (i.e.
t = approximately 2200 days), maximum flows are between 8 and 18l/min and tend
to increase over time. These observations led to injection of the pier. Leakage
flows decreased significantly in subsequent years and then increased until 1993
(i.e. t = 6000 days) to stabilise around peak values of 5–10 l/min.
The representation of the leakage rate as a function of the reservoir level shows
a marked correlation. It also clearly shows a threshold effect around a concrete joint
283
Q.106 - R.17
Fig. 4
Leakage flow rate on left bank (Q) as a function of time (t) and reservoir level (Z)
located at around 256.5 m NGF. However, it is difficult to establish a mathemati-

cal model a priori on the basis of raw measurements. It is thus necessary to use
statistical data analysis tools.
the last 3 years (Figure 5). The coefficients of determination obtained by the different
methods vary between 0.80 and 0.92. The best fits are obtained by the ANN and RF
methods (R 2 = 0.92) while the SVR and MLR methods provide poor fits (R 2 = 0.81
and 0.80 respectively).
The MLR method tends to overestimate the flow for low flows and minimise
it for strong values. The SVR method does not account for the observations and
tends to overestimate flows. The RF and ANN methods provide a narrower cloud
of points centred on the first bisector, but the ANN method has the disadvantage of
calculating negative flow values, unlike the RF method.
All methods agree that the seasonal influence is weak at a level equal to
the median value of 258 m NGF. It should be noted that, in the case of learning
approaches, this observation cannot be extrapolated to different reservoir levels.
For example, the analyses we conducted for a reservoir level of 256 m NGF show
that seasonal influence becomes negligible.
The effect of the average season reservoir level on the leakage rate is also
well transcribed through the models. A threshold effect is clearly observed between
levels Z = 256 m and Z = 257 m (corresponding to the level of the leaking lift joint).
It does not seem to be influenced by the season. The MLR model, however, leads
to an unrealistic parabolic relationship for Z < 255 m. This could be explained by
compensation for the seasonal effect that decreases in winter with a low reservoir
level. The other models give a close relationship up to Z = 259 m and then diverge,
with the ANN method giving the relationship with the highest level. It is interesting to
284
Q.106 - R.17
Fig. 5
note that the three methods suggest that there is a stage beyond which the reservoir
level has a more limited effect. This could be explained by a closure of the crack at
higher reservoir levels.
For time drift, the ANN and RF methods show two distinct parts delimited by
an artefact, the first between t = 0 and t = 2000 days marked by significant variations
before injection of the structure and the second showing a tendency to stability. The
285
Q.106 - R.17
SVR method transcribes these two trends, but does not reveal a rupture in the dam’s
behaviour.
The predictive capacity of each model can be assessed in Figure 5 (bottom).

All models are found to replicate the observations qualitatively. The MLR model,
however, tends to reduce the amplitude of variations by exaggerating low flows. The
ANN model yields results close to the measurements, including predicting peaks,
but the downward time drift component leads to extrapolation of flow values that
consistently exceed the measured values. The SVR and RF approaches appear to
be more robust. They do not exhibit time drift and demonstrate a good capacity to
reproduce the peaks. It is possible that the ability of these methods to produce clas-
sifications allows them to clearly distinguish two periods of differentiated behaviour.
Moreover, these two models do not predict negative flow values, in contrast to the
ANN model.
For this application, the RF method combines the best results in terms of the
ability to explain the observed measurements and predict flows by extrapolation.
The ANN model identifies a change in the behaviour of the structure but has a lower
predictive capacity by extrapolation. The SVR model has good predictive capaci-
ties but is unable to clearly separate the two periods of dam behaviour. The MLR
model appears to be less effective in integrating threshold behaviour (observed as
a function of the reservoir level).
5. PIEZOMETRY IN THE CORE OF AN EMBANKMENT DAM
The dam in question, built at the end of the 1970s, is a 27 m high clay-core
rockfill dam on natural ground.
The piezometric measurements considered are those recorded by a vibrating

wire cell located in the downstream third of the core, about 12 m below the dam crest
level. The time series extends over about 20 years with a time interval of between
15 and 30 days, or a total of 487 measurements after data cleaning. As mentioned
above, we have chosen to consider only explanatory variables that are concomitant
with the response variable.
5.2. PIEZOMETRIC MEASUREMENT ANALYSIS
The reservoir level varies around its normal value (Z = 51 m), generally with a
low drawdown, with level reductions of varying magnitude depending on the years
(Figure 6). The piezometric level varies between 40.7 m and 42.7 m and shows a
286
Q.106 - R.17
Fig. 6
Time series of the reservoir level Z (blue curve) and of the piezometric c11 cell
measurements (black curve).
certain correlation between extreme reservoir level values and extreme piezometric
values. Qualitative analysis of these curves suggests a slight time drift, but the
observed fluctuations appear to be more associated with the effect of the reservoir
level. Seasonal effects are not particularly pronounced.
the last 3 years. The coefficients of determination obtained by the different methods
are quite variable (Figure 7). The best fits are obtained by the SVR and RF methods
(R 2 = 0.91) while the MLR and ANN methods provide poor fits (R 2 = 0.78 and 0.79
respectively).
Seasonal effects are low for all methods (less than 20 cm). The curves rep-
resenting the influence of the reservoir level have similar characteristics (although
there is a variation in steps for RF). They are almost the same for Z over 50 m. The
ANN, SVR, and RF curves identify a more or less marked threshold effect around
Z = 46–48 m. The curves representing the time drift of measurements all show a
decreasing trend with the average slopes of the same order of magnitude. The time
drift curves obtained by the SVR and ANN methods are very close. The curve cor-
responding to the SVR method shows an inflection around t ≈ 4500 days, its physical
significance appearing doubtful. The RF method identifies a discontinuous change,
in steps. In Figure 8, the slope breaks of the curve are concomitant with both draw-
downs of the reservoir (t ≈ 500 days and t ≈ 2300 days) and significant level reductions
(Z ≤ 45 m) over the lifetime of the structure.
The predictive capacity of the models is tested over the last three years
of the series (Figure 7, bottom). Qualitatively, there are similar trends between
287
Q.106 - R.17
Fig. 7
measurements and all models (which show more or less marked oscillations).
The MLR and ANN models predict, on average, values that are lower than the
measurements. The SVR method follows the average measurement trend but has
larger oscillations. The RF method allows for a good prediction of the evolution
of the piezometric level. This suggests that the representative model established
288
Q.106 - R.17
Fig. 8
Time drift from RF method and time series of the reservoir level (Z)
for 3800 days < t < 5600 days remains relevant enough to predict the behaviour of
piezometry for 5600 days < t < 6700 days.
6. DISCUSSION
Three machine learning methods based on neural network, support vector

regression, and random forests models were implemented and applied to the anal-
ysis of measurements of movement of a gravity dam, leakage rate of an arch dam,
and piezometry in the core of an embankment dam, with each time series com-
prising between 500 and 2000 measurements. The results were compared with
the classical MLR linear regression method. The models were constructed from a
learning base obtained by random drawing of some of the measurements in the time
series. The ability of the models to replicate the test base measurements (the part
of the measurements not used for model construction) was studied. In addition, the
predictive capacities of each method were evaluated over the last 3 years of the time
series studied.
Our analyses highlight the limitations of MLR models that struggle to repli-
cate strong non-linearities. In particular, the mediocrity of the statistical learning of
the MLR method is mainly due to the predefined form and the continuous nature
of the functions used. When non-linear effects are of limited importance and the
initial population is sufficiently statistically homogeneous, predictive capacities are
comparable to other models.
289
Q.106 - R.17
SVR models allow for non-linear effects (including threshold effects) and do
not require significant implementation efforts.
ANN models are capable of taking non-linear effects into account but also
of identifying distinct periods of behaviour (statistically non-homogeneous popula-
tions). Our results, however, show that the ANN models are not necessarily the
most effective when the learning base dimension is insufficient. However, a sig-
nificant improvement in their performance can be achieved by using re-sampling
techniques ignored in this work.
RF models have proved to be particularly effective in detecting abrupt changes

in behaviour. They clearly identify threshold effects and naturally take non-linearities
into account. For the applications presented, these models do not appear to be
disadvantaged in terms of predictive capacities.
It is worth remembering that, regardless of the statistical models used, they

remain sensitive to data quality and sampling bias (measurement errors, too small
populations, statistically non-homogeneous distributions). The selection of input
data, the choice of explanatory variables and the re-sampling methods can cor-
rect many of these biases and improve the engineer’s confidence in the results of
the analyses.
7. CONCLUSIONS AND PROSPECTS
Our work has shown that machine learning methods can provide interesting
additions to the analysis of dam monitoring data, especially when conventional linear
regression models have low coefficients of determination. Engineers in charge of
interpreting measurements are responsible for exercising their judgement in assess-
ing the appropriateness of using these methods, which require significant technical
skills and expertise.
A prototype software platform is being developed in which the methods pre-

sented here are pre-programmed. The user will be able to apply these methods
to one or more series of measurements and to assess their relevance as a tool
for analysis based on the results (coefficient of determination, influence curves of
explanatory variables). It will also be possible to compare the actual measurements
and those predicted by the models for diagnosis of the measurements.
REFERENCES
[1] SALAZAR F., TOLEDO M.A., ONATE E., MORAN R. (2016) An empirical
comparison of machine learning techniques for dam behaviour modelling,
Structural Safety, 56, pp. 9–17. Doi: 10.1016/j.strusafe.2015.05.001.
290
Q.106 - R.17
[2] Swiss Committee on Dams. ICOLD (2003). Methods of analysis for the
prediction and the verification of dam behaviour.
[3] MATA J. (2011). Interpretation of concrete dam behaviour with artificial neu-
ral network and multiple linear regression models, Engineering Structures,
Volume 33, Issue 3, Pages 903–910. Doi: 10.1016/j.engstruct.2010.12.011.
[4] RADOVANOVIC S., MILIVOJEVIC V., CIROVIC V., DIVAC V., MILIVOJEVIC
N. (2015). Prediction of Concrete Dam Deformation using Artificial Neural
Networks, In: Proceedings of the Fourth International Conference on Soft
Computing Technology in Civil, Structural and Environmental Engineering,
p. 1–11.
[5] BELMOKRE A., MIHOUBI M., SANTILLAN D. (2019). Analysis of Dam Behav-
ior by Statistical Models: Application of the Random Forest Approach, KSCE
Journal of Civil Engineering, 23(11), p. 4800–4811. Doi: 10.1007/s12205-
019-0339-0.
[6] SIMON A., ROYER M., MAURICE F., FABRE J.P. (2012). Analyse des
mesures d’auscultation des barrages avec les réseaux de neurones, In Col-
loque Auscultation des barrages et des digues – Pratiques et perspectives.
Colloque CFBR, Chambéry, 27-28 novembre 2012, p. 314–328.
[7] SANTILLAN D., FRAILE-ARDANUY J., Toledo M.A. (2013). Dam seepage
analysis based on artificial neural networks: The hysteresis phenomenon, The
2013 International Joint Conference on Neural Networks (IJCNN), Dallas, TX,
2013, pp. 1–8. Doi: 10.1109/IJCNN.2013.6707110.
[8] KIM Y., KIM B. (2008). “Prediction of relative crest settlement of concrete-faced
rockfill dams analysed using an artificial neural network model.” Computers
and Geotechnics, Vol. 35, No. 3, pp. 313–322, Doi: 10.1016/j.compgeo.
2007.09.006.
[9] TAYFUR G., SWIATEK D., WITA A., and SINGH V. (2005). “Case study: Finite
element method and artificial neural network models for flow through Jeziorsko
earthfill dam in poland.” Journal of Hydraulic Engineering, Vol. 131, No. 6,
pp. 431–440, Doi: 10.1061/(ASCE) 0733-9429(2005)131:6(431).
[10] SALAZAR F., MORAN R., TOLEDO M. (2017). Data-Based Models for the
Prediction of Dam Behaviour: A Review and Some Methodological Consid-
erations, Archives of Computational Methods in Engineering, 24(1), p. 1–21.
Doi: 10.1007/s11831-015-9157-9.
[11] VAPNIK V. (1995). The Nature of Statistical Learning Theory, Springer.
[12] BREIMAN L. (2001). “Random Forests”. In: Mach. Learn. 45.1 (2001),
pp. 5–32. DOI: 10.3390/rs10060911.
291
Q.106 - R.19
-------
-------
LESSONS LEARNED FROM LONG TERM SURVEILLANCE OF TWO

TAILINGS STORAGE FACILITIES IN THE CANADIAN ARCTIC(∗)
Mathieu DESJARDINS
Senior Geotechnical Engineer, DE BEERS
Sean WHITAKER
Principal Closure Projects, DE BEERS
Antony TROLLOPE
Principal Hydraulics Engineer, ANGLO AMERICAN
Caius PRISCU
Head of Mineral Residue Facilities and Water Mgmt.,
ANGLO AMERICAN
CANADA
SUMMARY
Two of De Beers Canada’s mining operations located in the Northwest Territory

and in Northern Ontario are preparing for long term closure, the Snap Lake Mine and
Victor Mine respectively. They have been in operation since the mid’ 2000s and are
now in an active decommissioning and closure implementation stage, including their
respective tailings dams (also called Fine Residue Deposits or FRDs). This paper
presents the monitoring and surveillance programs implemented at the two sites, in
alignment with the Canadian Dam Association guidelines, the Mining Association
of Canada guideline and the requirements of the internal Anglo American technical
standard. Considering the arctic / sub-arctic location of the operations, monitoring
and instrumentation challenges of the FRDs have been numerous, and the teams
∗ Leçons apprises de la surveillance à long terme de deux barrages à résidus miniers dans
l’arctique canadien
292 DOI: 10.1201/9781003211808-114

Q.106 - R.19
had to adapt and experiment with numerous types of equipment, instruments and
technologies. This paper presents the lessons learned from some success stories
of long term monitoring and recommendations for potential arctic conditions imple-
mentation, as well as some of the shortcomings of using certain surveillance and
monitoring technologies, including remote sensing. The critical importance of such
data gathering over many years, its quality, analysis and interpretation of results in
preparing the closure plans is also highlighted.
RÉSUMÉ
Deux des exploitations minières de De Beers Canada situées dans les Terri-
toires du Nord-Ouest et dans le Nord de l’Ontario se préparent à une fermeture à
long terme, respectivement la mine Snap Lake et la mine Victor. Ils sont en service
depuis le milieu des années 2000 et sont maintenant dans une phase active de
déclassement et de mise en œuvre de la fermeture, y compris leurs barrages de
résidus respectifs (également appelés dépôts de résidus fins ou FRD). Cet article
présente les programmes de suivi et de surveillance mis en œuvre sur les deux sites,
conformément aux lignes directrices de l’Association canadienne des barrages, aux
lignes directrices de l’Association minière du Canada et aux exigences de la norme
technique interne Anglo American. Compte tenu de l’emplacement arctique / sub-
arctique des opérations, les défis de surveillance et d’instrumentation des FRD ont
été nombreux et les équipes ont dû s’adapter et expérimenter de nombreux types
d’équipements, d’instruments et de technologies. Cet article présente les leçons
tirées de certaines réussites de surveillance à long terme et des recommandations
pour la mise en œuvre potentielle des conditions arctiques, ainsi que certaines des
lacunes de l’utilisation de certaines technologies de surveillance et de surveillance
qui compris la télédétection. L’importance critique d’une telle collecte de données
sur de nombreuses années, sa qualité, l’analyse et l’interprétation des résultats lors
de la préparation des plans de fermeture sont également soulignées.
1. INTRODUCTION
Tailings storage facilities (TSFs) and associated containment structures such

as tailings and water retaining dams have recently received much attention at a
global scale given the recent tailings dams incidents that took place in Canada (Mt.
Polley), Brazil (Samarco and Brumadinho), and Australia (Cadia) to name a few.
The criticality of good governance, high quality stewardship and implementation
of critical controls at TSFs from around the world have elevated the necessity for
leading practices in relation to the safety of tailings dams.
The criticality and importance of quality monitoring and surveillance of such

facilities, the selection of proper instrumentation, the definition of controls based on
293
Q.106 - R.19
manual and automated readings and the role they play in an effective stewardship
program has gained increasing recognition, and triggered much effort on the part
of mining companies, mining industry associations, international organizations, and
regulatory bodies.
Despite these efforts and their improvements, tailings management contin-

ues to represent one of the highest and most visible risks to the mining industry
and the industries’ social license to operate. The industry, however, also recognizes
the safety of tailings dams to be paramount in the long term, even after construc-
tion and operation of the facility have ceased. Sustainable tailings management
includes actions and activities that start long before the first shovel breaks ground
and go through the closure period and into post-closure, in some cases for decades
after the operations ceased.
The role that quality instrumentation monitoring and surveillance practice plays
in defining, fine-tuning and shaping the care and maintenance and closure concepts
and principles of TSFs is extremely important at the two De Beers operations in
northern Canada as presented and discussed in this paper. Instrumentation monitor-
ing combined and complemented with regular field inspections, ground and satellite
surveillance, and continuous field observations results in a knowledge database
that is of prime importance as an input into the progressive reclamation and closure
concepts to be adopted and implemented. In other words, post-closure risk manage-
ment is informed through quality monitoring data and strong surveillance programs,
completed by competent and trained professionals, and independently verified.
This paper presents the instrumentation monitoring and surveillance work

undertaken at two of De Beers Canada operations located in northern Canada,
which are at the initial stages of active closure: Victor Mine is in Northern Ontario,
and Snap Lake Mine is located in the Northwest Territories. This paper presents
details related to the mine sites, their location, the TSFs and the challenges faced
with respect to operation, monitoring and surveillance of each of the facilities in
difficult climatic and remote arctic conditions. The paper also discusses and doc-
uments the challenges and the lessons learned when it comes to instrumentation
and surveillance of tailings dams in an arctic environment.
2. STANDARDS AND GUIDANCE IN TAILINGS DAM

MONITORING AND SURVELLANCE
2.1. ANGLO AMERICAN TECHNICAL STANDARD
De Beers Canada (De Beers) is part of De Beers Group, the world’s lead-
ing diamond company and is owned 85% by Anglo American and 15% by the
Government of the Republic of Botswana. De Beers follows leading practices in
294
Q.106 - R.19
the area of instrumentation, monitoring and surveillance of TSFs, and has adapted
such practices to the demanding Canadian arctic environment.
Anglo American (AA), the parent company of De Beers Canada, has devel-
oped and deployed since 2014 a group-wide technical standard (AA Standard), and
supporting documents, drawing upon best practices worldwide [1]. The AA Standard
identifies and incorporates a series of minimum requirements for design, operation,
monitoring, surveillance and closure of mineral residue facilities, providing a solid
basis for consistent governance and quality stewardship. The AA Standard has
established governance processes [2], [3], and defines a list of requirements that
include frequency of inspections and assurance based on the Consequence Clas-
sification of Structures (CCS) rating, applicable to active, inactive, care and main-
tenance (C&M) and closed sites. In addition, the AA Standard has specific require-
ments for an Operation, Maintenance, and Surveillance (OMS) manual that includes
specifics for various surveillance and inspection levels, as well as instrumentation
monitoring, reporting, and requirements for establishing safe ranges and thresh-
olds, including definition and implementation of Trigger-Action-Response-Plans
(TARPs) – a key protocol in risk management of TSFs over their entire life cycle.
2.2. MINING ASSOCIATION OF CANADA GUIDELINES
The Mining Association of Canada (MAC) has prepared over the years sev-
eral guidance documents that are considered leading practice around the world,
with a few additional countries adopting them internally [4]. MAC documents make
clear reference to monitoring requirements and surveillance programs to ensure key
structural performance and water balance performance are aligned with the design
intent. MAC guidance on monitoring and surveillance are included specifically in the
following three documents:
• “A Guide to the Management of Tailings Facilities” [5]

• “Developing an Operation, Maintenance and Surveillance Manual for Tailings
and Water Management Facilities” [6], and
• Towards Sustainable Mining” [7] which is a program established in 2004 that
commits MAC members to implement responsible mining.
2.3. CANADIAN DAM ASSOCIATION GUIDELINES
The Canadian Dam Association (CDA) has issued several guidance docu-
ments [8], [9], referring to the leading practices in dam safety and the role of
instrumentation monitoring and surveillance in dam safety and risk management.
The guideline document referring directly to mining dam safety [9] discusses and
emphasizes the continuation of monitoring and surveillance through closure and
post-closure stages of a TSF (either in active or in passive care).
295
Q.106 - R.19
2.4. GLOBAL INDUSTRY STANDARD ON TAILINGS MANAGEMENT
The recently released Global Industry Standard on Tailings Management

(GISTM) presents the most recent example of leading practices in preventing catas-
trophic failures of TSFs [10], [11]. The GISTM refers to Requirement 6.3 for the
“Construction Record Report” that includes any changes in the monitoring system,
and to Requirement 6.4 to develop, implement, review and train the staff on the
OMS manual, which is to be updated at least annually during the TSF life cycle.
Resulting updates from the monitoring and surveillance programs need to be inter-
preted and must inform the risk management process for the facility. The design
and implementation requirements of a comprehensive and integrated engineering
monitoring program are additionally discussed in the requirements listed under Prin-
ciple 7 to inform the risk management process and provide support to verify design
assumptions and monitoring potential failure modes.
3. SITE LOCATIONS AND CLIMATIC CONDITIONS
3.1. VICTOR MINE
The Victor Mine is located in Northern Ontario, Canada. The former diamond
mine is approximately 90 km west of the First Nation community of Attawapiskat and
100 km west of the James Bay coast at a latitude of 52◦ 49 36 N and longitude
of 83◦ 53 7 W as shown on Figure 1. It is Ontario’s first diamond mine and the
second of three in Canada for De Beers. It is a remote fly-in/fly-out operation that
is supported by a 375 km winter ‘ice’ road available for 4–6 weeks which is used to
bring supplies and equipment to site.
Several kimberlite pipes have been discovered in the area since 1987, however
only three pipes were mined at the Victor operation, creating a single open pit. First
ore was mined in March 2008 and the last ore in March 2019.
Victor mine has a subarctic climate with long cold winters and short cool to mild
summers. Between November and April, the air temperature is continuously below
0◦ C with a mean annual temperature of about −1.5◦ C recorded during the 38-year
baseline period (1981–2019). The mean air temperature peaks in July at 16.1◦ C
however daily highs can exceed 30◦ C. The mean annual rainfall is about 581 mm
with peak precipitation occurring in July as rainfall. In a typical year there are approx-
imately 139 frost-free days. The current climate extreme indices are consistent with
the current climate trends, showing warming minimum temperature trends and likely
more extreme rainfall events [12].
296
Q.106 - R.19
Fig. 1
Location of De Beers Canada Offices and Operations in Canada.
Emplacement des bureaux et des opérations de De Beers Canada au Canada.
Fig. 2
Victor Mine Site overview.
Vue aérienne de la mine Victor.
297
Q.106 - R.19
3.2. SNAP LAKE MINE
The Snap Lake Mine is an underground diamond mine that operated from
2008 to 2015. In December of 2015, De Beers ceased diamond mining operations
and has since progressed closure design and regulatory applications. The mine is
located 220 km Northeast of Yellowknife, Northwest Territories, at a latitude of 63◦
36 19 N and a longitude of 110◦ 52 00 W as shown on Figure 1. The Mine is located
on the northeast peninsula of the western shore of Snap Lake, in an isolated area
with no permanent access roads. Access to the site is by aircraft or a winter ice road
that spurs off the main winter road from Yellowknife to the Lupin Mine.
Fig. 3
Snap Lake Mine overview.
Vue aérienne de la mine Snap Lake.
The average annual temperature is −5.9◦ C. Between October and April, the air
temperature at the Mine site is continuously below 0◦ C having average monthly tem-
peratures between about −32◦ C and −15.9◦ C. In the ‘open water’ season between
May 15 and October 15, the mean air temperature is above 0◦ C and peaks in July with
a mean temperature of 13.4◦ C. Observed climate normals (1981–2010) at Environ-
ment and Climate Change Canada’s Yellowknife monitoring station show an annual
average of 115 frost-free days [13]. The mine site is situated just north of the diffuse
border between the zones of discontinuous and continuous permafrost.
298
Q.106 - R.19
Annual average total precipitation is 356 mm and in a typical year, 67% of

precipitation occurs as snow, while 33% occurs as rain. The winter freeze begins in
October when the average temperature drops below 0◦ C. Snow accumulates over
the winter months, until late May. Spring freshet occurs in June when snow melt
releases high volumes of water [14]. Most of the annual runoff volume is recorded
during the months of June and July; flows steadily decrease through the summer,
and streams completely freeze over by October.
4. VICTOR TAILINGS FACILITY
4.1. DESCRIPTION AND INSTRUMENTATION
The tailings facility at Victor Mine is known as the PKC Facility and consists
of two impoundments, Cell 1 and Cell 2 (Figure 4). Cell 1 construction began in
2009 and the dykes have been raised four times since commissioning (2010, 2011,
Fig. 4
Victor Mine PKC Facility. Cell 1 to the north, Cell 2 to the south (foreground).
Installation PKC de la mine Victor. Cellule 1 au nord, cellule 2 au sud (devant).
299
Q.106 - R.19
2012, and 2013) to a maximum height of 12 m. Cell 2 was only constructed to the
1st stage, with extra capacity available to accommodate potential mining of other
nearby kimberlite deposits.
The coarse fraction of processed kimberlite (PK) was used as the primary
construction material for the dykes. The dykes are homogeneous fill embankments
and were progressively raised by the downstream method of construction to provide
suitable containment for the fine PK, in addition to providing the required storage
needs for water management. The coarse PK acts as a permeable material that
provides containment for the fine PK while allowing seepage to report to the collec-
tion ditches around the perimeter of the facility. The decant pipes in Cell 1 and Cell
2 control the water level within each cell by conveying excess water from the PKC
Facility to the Polishing Pond (former site quarry).
Processing of fine PK ceased in Q2 2019. Over the 11+ year mine life, 28.4
million metric tonnes of kimberlite was processed. Cell 1 has a surface area of about
79 hectares and stores 5.32 M-m3 of fine PK. Cell 2 shares the South Dyke of Cell
1 to the north and provides 1.71 M-m3 of fine PK storage [15].
Significant upgrades were made to the instrumentation and monitoring pro-

gram during the final year of operation in preparation for closure. The following
summarizes the current instrumentation around the PKC Facility:
• 10 standpipe piezometers around Cell 1 dykes. All have been instrumented

with vibrating wire piezometers (VWP) and are wirelessly connected to a base
station for remote monitoring;
• 5 vibrating wire piezometers around Cell 2. One of which is wirelessly con-
nected to the base station for remote monitoring, the others require manual
downloading of readings;
• 12 groundwater quality monitoring wells at the downstream toe of the North
Dyke in Cell 1;
• 6 settlement plates around both Cell 1 and 2;
• In addition, remote monitoring using lnSAR began in 2019.
The piezometers are installed within the dykes to monitor the phreatic sur-
face and long-term dyke performance. The settlement plates were installed in 2018
around the perimeter dykes and the first two years of data has indicated negligible
movement (i.e. <3 cm).
4.2. FACILITIES MONITORING AND SURVEILLANCE
An Operation, Maintenance, and Surveillance (OMS) Manual has been in

place since commissioning of the PKC Facility in 2009. This document was
300
Q.106 - R.19
developed by the Engineer of Record (EoR) and has been updated regularly through-
out the operational phase. The most recent version [15] has been updated for
active-closure, in general accordance with the latest MAC guidelines [6].
Regularly scheduled inspections (annual DSI, quarterly inspections by a

Geotechnical Engineer, and bi-weekly/monthly routine inspection depending on the
season completed by the on-site staff) are required to ensure that the facility is
preforming as intended and to identify areas where maintenance work is required.
Fig. 5
Victor PKC Facility main instrumentation data logger hub with satellite modem.
Centre d’instrumentation Victor PKC Facility principal avec modem satellite.
As of 2018 all the VWP’s for critical monitoring are outfitted with radio data
loggers that send data to a central instrumentation hub located on the south-east
corner of Cell 1. This central solar powered hub receives then transmits the data via
satellite modem to a cloud based server to avoid issues with changes to the mine
network during closure (Figure 5). Due to the power draw from the satellite modem
and limited sunlight in the winter, the modem is set to turn on daily at noon for 30
minutes to transmit the data files. Instrument data is now reviewed and assessed
against the TARP thresholds remotely by the Geotechnical Engineer.
301
Q.106 - R.19
4.3. KEY CHALLENGES AND LESSONS LEARNED
4.3.1. Freshet preparation
Freshet is a common occurrence at northern Canadian mines and refers to

the spring thaw where ice and snow melt suddenly. This event often results in signif-
icant surface water flow creating erosion channels on the downstream slopes of the
dykes, increased flow through water control structures, and increased water levels
in storage ponds.
This natural occurrence cannot be exactly predicted but has become a routine
part of tailings facility management. From experience it was found that to best man-
age the freshet requires action and planning the year prior, in advance of freeze-up.
All maintenance work is completed in late August or September repairing erosion
channels, armoring dyke shoulders and crests, ensuring drainage ways are not
blocked. Piping systems are drained of water to ensure they do not freeze up as
they typically will not thaw in time for freshet pumping. In the spring the monitoring
frequency is increased by onsite personnel in order to take quick action once freshet
arrives.
4.3.2. Winter inspections
During the operational phase, when slimes were being deposited, twice daily
inspections of the PKC Facility was critical to performance and safety. Since pro-
duction has ceased it has become evident that wintertime inspections provide little
value as the conditions of the tailings facility do not change, and any visible signs
of change are completely obscured by snow and ice. Additionally, the staffing and
equipment needed to keep access roads open constantly has become scarcer in
closure. This has resulted in the need to challenge some of the operational pro-
cedures and schedules that have been in place since the commissioning of the
tailings facility. The AA Standard [1] specifies the monitoring and surveillance fre-
quencies required, and a formal request based on the risk during closure to reduce
the frequencies during the winter months has been logged.
4.3.3. Remote sensing – InSAR
Interferometric Synthetic Aperture Radar (InSAR) has been in place since

2019 at Victor Mine. This technology was selected as it can typically monitor all
the assets on the property for deformation at sub-centimeter accuracy. In addi-
tion to monitoring the PKC Facility, the open pit long-term break-back area and
302
Q.106 - R.19
numerous stockpiles are monitored for stability. Closure related in-ground instrumen-
tation installation requirements were reduced as a result of InSAR implementation.
Selection of the specific InSAR product was based on risk level, physical and environ-
mental characteristics, and business need and may change as the mine transitions
from the active-closure to post-closure phase.
A significant limitation of InSAR monitoring in Northern Canada is the inability

for the radar signal to penetrate snow and ice. Due to the winter conditions at Victor
there is a limited monitoring window available between April and October, however
due to freezing conditions monitoring is not as critical during these winter months.
The upside of this seasonal monitoring is reduced cost as monitoring only occurs
for half the year. The downside is that because InSAR accuracy increases with the
number of subsequent images acquired this interrupted monitoring is also a potential
limitation to its use at the mine site. The annual break in the data means that every
spring re-baselining needs to take place to get reliable results.
A high-resolution satellite (TerraSAR-X) with 3 × 3 m pixels was selected to

be able to receive multiple signal returns from the narrow dykes, a monitoring area
constraint. The maximum possible frequency (11-day return) was selected to reduce
the time required to build the baseline data stack. The reporting frequency is slightly
less frequent (monthly during the snow free period) to reduce costs but allows staff
to target inspection areas based on the data during the summer and make decisions
on critical maintenance before freeze-up. Linking previous years’ InSAR results is
being investigated to provide continuity and determine long term deformation trends
year over year. Artificial corner reflectors installed at key areas of the dykes are
being considered for the future to improve results.
5. SNAP LAKE TAILINGS FACILITIES
5.1. DESCRIPTION AND INSTRUMENTATION
Snap Lake Mine has a co-disposal facility (tailings facility) called the North
Pile. The North Pile facility is the permanent surface location for the disposal of
processed kimberlite (PK), waste rock, and non-hazardous waste (landfill waste)
materials at the Mine and was to be developed in three cells: Starter Cell, East Cell,
and West Cell. The West Cell was never fully developed.
Instrumentation is in place to monitor embankment stability and deformation,

ground temperature, seepage, water quality and to monitor operational performance.
The Snap Lake monitoring system includes thermistors installed on the embankment
crests and downstream slopes as well as at key locations on original ground adjacent
303
Q.106 - R.19
to the North Pile. VWPs and standpipes have been installed both on the embankment
crests and downstream slopes of the North Pile as well as on the embankments of
the Water Management Pond (WMP). Survey monuments have been installed on
the North Pile and WMP embankments at select cross-sections for monitoring of
embankment stability.
In 2018 De Beers installed camera stations at each of the perimeter water

control structures to remotely monitor water levels within their safety thresholds
consistent with the OMS manual and legal response framework.
5.2. FACILITIES MONITORING AND SURVEILLANCE
A comprehensive OMS Manual has been developed and used for the Snap
Lake operation since the establishment of the facility in 2008. This has been a living
document updated annually and owned by the North Pile’s competent person as
per the AA Standard [1] and in close discussion with the EoR. The approach for the
North Pile monitoring is a combination of on-site and remote monitoring consistent
with the facility’s risk classification. This approach will continue into closure and
post-closure and will be adjusted based on field observations and performance of
the structures.
Regularly scheduled inspections are required to ensure that the North Pile
facility is performing as intended and to identify problems and issues so that nec-
essary corrective actions may be implemented in a timely manner. Surveillance
tasks include visual inspections, monitoring of instrumentation and water levels and
preparation of written and photographic documentation. De Beers and the EoR have
developed a TARP framework that includes the threshold values for four trigger levels
for a range of monitoring criteria.
In 2017 the Snap Lake systems began to be automated focusing on the

geotechnical instrumentation data collection for the East Cell embankment dam,
which included connecting existing instruments to data loggers installed along the
facility that collect information and transmit it through satellite modem to a cloud
based server accessible to De Beers for evaluation either on or off-site.
5.3. KEY CHALLENGES AND LESSONS LEARNED
5.3.1. Preparing for closure
The Snap Lake team have looked to implement a robust monitoring strategy
throughout the lifecycle of the mine. The mine installed all the instrumentation at the
304
Q.106 - R.19
commencement of operations under the premise of a continuous on-site presence

for the operation of the mine. However, in 2015 this posed a challenge as the mine
ceased operations prematurely and remote monitoring at the time was not a design
consideration. The Snap Lake Mine is now in a period of extended care and mainte-
nance as it prepares for final closure of the Site. FutureSmart™is Anglo American’s
platform for redefining the mining innovation cycle by creating shared value through
the smart adaption and application of technology. The closure sites are now looking
to FutureSmart™their monitoring strategies.
5.3.2. Remote operation
The remote and extreme environment posed the greatest challenge for both
on-site monitoring of the embankments and damage to the instrument infrastructure
due to routine maintenance. Access by the team to the vacant site during the winter
was minimized to lower the operational risks with a reduced workforce. In 2017 De
Beers began to implement an automation strategy to limit the need for continuous
on-site monitoring.
The geotechnical instrumentation data collection for the East Cell embankment
dam was identified as critical for monitoring by the EoR and De Beers’ competent
person. The team then targeted these instruments for remote monitoring retrofitting.
This was accomplished by connecting the existing instruments (thermistors, stand-
pipes and vibrating wire piezometers) to data loggers and radio receivers placed
along the facility to collect information and transmit it via satellite modem to a cloud-
based server accessible to De Beers for evaluation. The data loggers collect and
send information on a set schedule, and are routinely checked against the Golder
monitoring response framework (TARP thresholds) for the North Pile & WMP. The ini-
tial learnings from our implementation of this strategy was the need for a more robust
set-up of the systems and ensuring on-going support by information technology
professionals to troubleshoot uploads to central databases.
5.3.3. Continuous improvement and ongoing support
The solutions are never “plug and play” and require initial and on-going support
to ensure long term function. Due to the topography and size of the North Pile the
pre-existing instrument cables could not be hard wired to a single location. Instead
cables were extended and routed to five separate data logger hub locations installed
in small huts to protect the equipment from the harsh environment (Figure 6). These
numerous huts on the top of the facility connected to each other by radio receivers
ultimately increase the risk of maintenance issues (damage during grading/snow
clearing) and surge damage from lightning strikes.
305
Q.106 - R.19
Fig. 6
Snap Lake instrumentation hut and datalogger / satellite modem set-up.
Cabane d’instrumentation à Snap Lake avec enregistreur de données et
configuration du modem satellite.
Electrical storms have led to damage of several instruments and cables

increasing repair costs. Based on the success at Victor Mine, to mitigate some
of the maintenance issues experienced, in the future wireless radio-transmitting
data loggers should be used instead of hardwired instrumentation. De Beers will
look to improve upon the currently installed remote monitoring systems in closure,
harmonize systems across sites, ensure technical training, and ensure ongoing IT
support.
5.3.4. Cameras vs pressure transducers
In 2018 De Beers installed camera stations at each of the perimeter water

control structures to remotely monitor water levels within each of the appurtenant
structures and to act as an indicator system in the case water levels exceed safety
thresholds as per the North Pile TARP [16], (Figure 7). Initially De Beers looked to
install pressure transducers in the perimeter water control structures. Due to the
extreme environment and the lack of water flow during extended care and main-
tenance (ECM) and Closure the sumps freeze to the bottom which can damage
306
Q.106 - R.19
the transducers and make them ineffective. This led to the installation of the cam-
era and staff gauge system at the perimeter sumps and WMP. Using the available
camera technology on the market the sumps can be evaluated, however additional
complications arose with the frosting of the camera lenses.
The site team then worked with the camera suppliers to re-design the system
to limit frosting with a heating system that pre-warms the camera lens before the
image is taken. This did result in an increase in mechanical utilization of the cameras,
however there are occasional ongoing issues when extreme cold temperatures are
reached. Figure 8 shows the remote camera dashboard as well as the manual
photographs taken during the monthly site inspection.
Fig. 7
Response framework at Snap Lake Mine.
Cadre de réponse à la mine de Snap Lake.
5.3.5. Long term power requirements
While all the above critical challenges have been resolved the long-term power
requirements for any type of instrumentation is a main driver for instrument selec-
tion and strategy. Systems that are installed in the remote arctic and sub-arctic
environments are subject to several climatic factors not encountered elsewhere.
As such the power demand cannot simply be solved by solar panels and bat-
teries. A combination of wind, solar and battery technologies coupled to other green
technologies are necessary to ensure the sustainability of any remote monitoring
approach.
307
Q.106 - R.19
Fig. 8
Remote monitoring of sumps and water management pond with cameras.
Surveillance à distance des puisards avec des caméras.
While there is no “silver bullet” each site has its own unique opportunities that
requires out of the box thinking. Technologies like the Green Cube™, a small trans-
portable and adaptable energy system have been used successfully at our other
advanced exportation Sites and for our closing operations these types of power
units will be necessary until longer term fully remote options are available like opti-
mized LIDAR and Satellite systems (Low-earth orbiting satellites) can be used to
capture, supplement or replace our current technologies.
6. RECOMMENDED APPROACHES AND PRACTICAL

CONSIDERATIONS FOR ARCTIC ENVIRONMENTS
While there is no single solution on the market for the monitoring and man-
agement of our tailings assets there are many considerations when designing,
modifying or upgrading systems to be FutureSmart™. It is critical to ensure that
our facilities are run in accordance with the guidance documents and our internal
Anglo American Standards and use robust FutureSmart™technology such as low
maintenance wireless data loggers, InSAR, drones, and fiber optic cables. The
308
Q.106 - R.19
below recommendations are provided to enable a successful monitoring system for

closure.
6.1. INSTRUMENTATION SET-UP
Datalogger and telemetry implementation at Victor was able to capitalize on

the learnings from Snap Lake. Damage of the instrumentation cables has occurred
on numerous occasions at Snap Lake due to snow clearing and maintenance work
and has been mitigated as best as possible with careful survey of the instrumentation
lines as well as protection and marking of the cables. This was largely avoided at
Victor mine with the implementation or wireless radio dataloggers used to transmit
the data to the main datalogger hub.
The advantage of the wireless system will also be realized during the con-
struction of the final closure configuration. At Snap Lake a decision will need to
be made on whether to convert to wireless or use caution when re-grading and
placing cover material to avoid severing instrument lead cables needed for post-
closure monitoring. The lithium batteries used in the wireless dataloggers have
been tested and proven in arctic climates and are the preferred method for new
installations.
6.2. INSTRUMENT AND REMOTE SENSING LIMITATIONS
Understanding the limitations of both point locations instruments and remote

sensing methods is critical. At Victor mine the in-ground instruments are spread
thin and this limitation is addressed with the co-use of InSAR monitoring. InSAR
monitoring does not solve this issue entirely as excellent monitoring conditions are
rarely met. For the time being the best solution is to take a balanced approach
using both in-ground instruments and remote sensing solutions and continue vigilant
visual monitoring to ground truth the results and identify issues not detectible with
the current technology. Currently under testing at multiple Anglo American mine
sites are real-time fiber optic deformation and leakage detection systems that could
address the shortfalls on the existing methods.
6.3. INSTRUMENTATION DATA VISIBILITY
Going forward a unified solution for monitoring of the closed mine sites is under
development. To improve the response to changing conditions of the tailings facilities
309
Q.106 - R.19
visualization software will be implemented. This will change how the data currently
being collected is being shared, viewed, reported, and acted on and only becomes
more important as presence at the closed mine sites diminishes.
6.4. ON-GOING IT SUPPORT
For closed sites the goal is to relinquish our operations to the respec-
tive government bodies. However, some of our tailings facilities will have long
term monitoring requirements and any monitoring approach should be able to be
upgraded or handed over in the future. While there are always initial growing pains,
the on-going IT support is critical to ensure a sustainable monitoring approach is
handed over after the completion of closure exercises. Training and establishment
of evergreen strategies is needed to ensure a successful transition of assets back
to key stakeholders.
6.5. LONG TERM POWER REQUIREMENTS
A critical component for arctic operations and especially on closed operations

is the requirement of power for any solution selected. Focus should be on minimiz-
ing power demands, long term energy storage solutions and options that allow for
the least amount of maintenance to improve overall sustainability of the monitoring
regime. This is a focus of De Beers and Anglo American as we look to become
a carbon neutral company but is more significant for arctic operations that require
special logistics and extreme weather considerations. Robust proven technologies
are needed but they also must not rely on human intervention.
7. CONCLUSIONS
The role that quality instrumentation monitoring and surveillance practice plays
in defining, fine-tuning and shaping the care and maintenance and closure concepts
and principles is extremely important at the two De Beers operations in northern
Canada. As we look to FutureSmart™our global operations by redefining the min-
ing innovation cycle through the smart adaption and application of technology the
lessons learned from some of our success and challenges will drive us into a sus-
tainable and robust long term monitoring solution. The critical importance of such
data gathering over many years, its quality, analysis and interpretation of results in
preparing the closure plans will be critical as we transition our assets to their next
phase of project life cycle.
310
Q.106 - R.19
ACKNOWLEDGEMENTS
The authors wish to acknowledge the support of De Beers Canada and Anglo
American for allowing to publish this information. Special note to our colleagues at
the Victor and Snap Lake Mine for the hard work and dedication to the ongoing man-
agement of our operations. In addition, the authors wish to thank Golder Associates
for the technical support throughout the years and Engineer of Record role at the
Victor and Snap Lake facilities.
REFERENCES
[1] Anglo American (2019): “Anglo American Technical Standard AA TS 602

001 Mineral Residue Facilities and Water Management Structures Stan-
dard and Technical Specifications”. Original Version released February 2014.
https://www.angloamerican.com
[2] Anglo American (2019): “Managing Mine Tailings Safely” https://www.

angloamerican.com/sustainability/tailings-summary/managing-tailings-safely
[3] Priscu, C (2020): “Tailings Dam Safety Management and Governance Frame-
work at Anglo American”, in Society for Mining, Metallurgy & Exploration
(SME)’s Mine Tailings Symposium at the SME 2020 Annual Conference and
Expo, Phoenix, AZ.
[4] Dumaresq, C., and Davies, M (2019): “Responsible tailings management –

global best practice guidance”, in Sustainable and Safe Dams Around the
World / Un monde de barrages durables et sécuritaires, Proceedings of the
ICOLD 2019 Symposium, (ICOLD 2019), June 9-14, 2019, Ottawa, Canada
/ Publications du symposium CIGB 2019, juin 9-14, 2019, Ottawa, Canada
Edited By Jean-Pierre Tournier, Tony Bennett, Johanne Bibeau.
[5] Mining Association of Canada (2019): “A Guide to the Management of Tailings

Facilities’, Version 3.1, February. https://mining.ca/
[6] Mining Association of Canada (2019): “Developing an Operation, Mainte-

nance, and Surveillance Manual for Tailings and Water Management Facilities,
Second Edition. February. https://mining.ca/
[7] Mining Association of Canada (2019): “Towards Sustainable Mining Tailings

Management Protocol”, February. https://mining.ca/
[8] Canadian Dam Association (2013):” Dam Safety Guidelines”. www.cda.ca
[9] Canadian Dam Association (2019): “Application of Dam Safety Guidelines to

Mining Dams”, 2019 Edition. www.cda.ca
311
Q.106 - R.19
[10] Global Tailings Review initiative (2019-20). https://globaltailingsreview.org
[11] ICMM (International Council on Mining and Metals). 2016. Tailings gover-
nance framework: Position statement on preventing catastrophic failure of
tailings storage facilities. https://www.icmm.com/position-statements/tailings-
governance
[12] Golder Associates (2020): “De Beers Victor Diamond Mine Detailed Current
Climate Baseline.”
[13] ECCC (2019): Canadian Climate Normals 1981-2010 Station Data for Yel-
lowknife A. Accessed April 24 2019 from http://climate.weather.gc.ca/climate_
normals/results_1981_2010_e.html?stnID=1706
[14] De Beers (2002): Developer’s Environmental Assessment Report, Snap Lake

Diamond Project. Appendix XI.1 - Environmental Health Methods. Submitted
to the Mackenzie Valley Environmental Impact Review Board.
[15] Golder Associates (2020); “Operation Maintenance and Surveillance Manual

Victor Mine PKC Facility”.
[16] Golder Associates (2018): “Monitoring Response Frameworks for North Pile
Facility and Water Management Pond Dams, Snap Lake Mine – Revised for
Extended Care and Maintenance Phase. Technical Memorandum. Internal
Report.
312
Q.106 - R.20
-------
-------
BUT WHERE EXACTLY ARE WE?”: POSITIONING AND NAVIGATION

ACCURACY IN REMOTE UNDERWATER SURVEYS (∗)
Bill SHERWOOD
Manager, International Marine Development, ASI MARINE
Ryan BUCHOON
Project Manager, CH, Certified Hydrographer, ASI MARINE
Robin HOULIK
Project Manager, ROV, ASI MARINE
CANADA
SUMMARY
This paper will look at processes for remote acquisition, storage, and post
processing of sonar data from underwater dam surveys. This will include tie-ins to
existing data such as previous inspections (e.g., to determine rates of erosion over
time) and above water LIDAR. The paper will be of interest to dam owners, operators
and consulting engineers who wish to understand:
what sonar technologies/techniques have a proven record?

what levels of accuracy and repeatability can be expected from underwater
surveys?
how is data ultimately presented in reports?
what are the critical factors affecting data quality during field operations?
what can go wrong, and what can be done during the post-processing stage
to mitigate errors imbedded in the data?
∗ ‘‘Mais où en sommes-nous exactement?”: Précision du positionnement et de la navigation
lors d’inspections sous-marine
DOI: 10.1201/9781003211808-115 313

Q.106 - R.20
what’s new in sonar technology and how has existing technology been adapted
to new purposes?
Examples from inspection and survey reports of dams and surrounding struc-
tures, including spillways and bypass tunnels will be used. These will show how
sonar can be an effective tool for metrology, establishment of baseline structural
conditions, and most importantly, acquisition of quantifiable data for both immediate
and long-term risk management.
RÉSUMÉ
Ce rapport se penche sur les processus d’acquisition, de stockage et de post-

traitement à distance des données sonar issues des levés de barrages sous-marins.
Le tout tient compte des liens avec les données existantes telles que les inspections
précédentes (exemple: les taux d’érosion au fil du temps ou autres) et le LIDAR pour
la portion au-dessus de l’eau. Ce document intéresse les propriétaires de barrages,
les exploitants et les ingénieurs-conseils qui souhaitent comprendre:
• Quelles technologies / techniques de sonar ont fait leurs preuves ?

• Quels niveaux de précision et de répétabilité peut-on attendre des levés sous-
marins ?
• Comment les données sont-elles finalement présentées dans les rapports ?
• Quels sont les facteurs critiques affectant la qualité des données pendant les
opérations sur le terrain ?
• Qu’est-ce qui peut mal tourner et que peut-on faire pendant la phase de post-
traitement pour atténuer les erreurs incrustées dans les données ?
• Quoi de neuf dans la technologie du sondeur et comment la technologie
existante a-t-elle été adaptée à de nouvelles fins ?
Le rapport présente également des exemples tirés de rapports d’inspection

et d’enquête des barrages et des structures environnantes réelles, y compris des
déversoirs et des tunnels de contournement ou de dérivation. Ceux-ci montrent com-
ment le sonar peut être un outil efficace pour la métrologie, l’établissement des con-
ditions structurelles de base et, surtout pour l’acquisition de données quantifiables
pour la gestion et l’identifications des risques immédiats et à long terme.
1. DAM SURVEYS – THE INDISPENSABILITY OF SONAR
Comprehensive hydrographic underwater assessments of dams now include

sonar surveys (SOund NAvigation and Ranging) of headwaters, reservoirs, supply
canals, intakes, penstocks, headrace and tailrace tunnels, splash pools, spillways
and receiving water bodies. This data is fast becoming an essential part of asset risk
management, especially for older structures or those exhibiting anomalous behavior.
Even new projects now require post watering-up sonar surveys both as a contrac-
tual condition of substantial completion and, of equal importance, for owners to
314
Q.106 - R.20
establish an ‘apples-to-apples’ baseline against which future sonar survey data can
be compared.
Hydrography is the measurement and description of physical features found

underwater. This general marine survey term includes bathymetry (bottom mapping
or submarine topography) and scanning of any sort in the water column, so long as
it produces a measurable result.
Sonar is the primary tool used by modern hydrographers. The sister science
of terrestrial lidar (LIght Detection And Ranging) provides above water survey data.
Because neither sonar nor terrestrial lidar are effective in shallow water, hydrogra-
phers also use bathymetric lidar, which can penetrate and scan the shallow water
bottom areas found along shorelines. Taken together, these technologies can pro-
vide survey coverage of all critical areas and greatly enhance the value of the entire
knowledge base for dams and their environs.
Terrestrial lidar imaging does not necessarily need to be of survey grade to be

extremely useful for both project planning and reporting. Just as site operations pho-
tographs can help those who might not have been present quickly grasp the layout
and essential working parameters of a project, even a non-survey-grade lidar fly-
over can provide a useful general reference. In any case, structural issues observed
above the water line are often found to have their origins below it. Dam risk man-
agement programs are therefore only comprehensive if they also emphasize the
importance of regular underwater surveys.
Data bases are rapidly becoming virtually all-encompassing, and data man-
agement issues are critical. Having a comprehensive archive of historical data is
critical, but ensuring that data subsets are reliably correlated and, most importantly,
easily accessible, is an essential part of risk management. Figure 1 shows how
disparate data types have been effectively integrated to show the entire project
scope via a virtual ‘flythrough’.
Fig. 1
Screen captures of flythrough: (top l) aerial lidar of site; above-water lidar survey of
pumphouse shaft and, beneath the water line, multibeam sonar, including a portion
of the intake tunnel.
Captures d’écran du survol: (haut l) lidar aérien du site; levé lidar au-dessus de
l’eau du puits de la station de pompage et, sous la conduite d’eau, sonar
multifaisceaux, y compris une partie du tunnel d’admission
315
Q.106 - R.20
1.1. ACCURACY OF SONAR VS LIDAR
Owners contracting an underwater inspection understandably want to know

what level of sonar survey accuracy can be expected. Will small cracks and dis-
placements be apparent and if so, measured to what tolerances? The impressive
results produced by terrestrial lidar may lead to the mistaken belief that a similar
degree of fine measurement is easily duplicated underwater. The fact is that sub-
surface surveys are subject to a broad range of factors and influences that have an
impact on data quality and reliability.
Though there are similarities, lidar and sonar differ fundamentally in the type
and behaviour of the energy waveforms they employ. Lidar uses pulsed laser,
whereas sonar uses pulsed sound. While both establish target range by measuring
the time it takes for each pulse to travel from and return to a source, laser is prop-
agated as a parallel beam, while sound propagates as a wave that both dissipates
and fans out as it travels. Sonar ranging accuracy can therefore be generally under-
stood as a direct function of distance between the sensor and the target, whereas
lidar is not, so its inherent ranging accuracy is greater.
Lidar typically has an additional advantage over sonar since it can easily incor-
porates independent pre-existing coordinates across the survey field derived from
monuments, maps or drawings, all of which can be used as points of triangulation
for data confirmation and correction. Taken together, these refinements provide a
high level of confidence in the overall result. For sonar surveys, pre-existing under-
water data e.g., from design and as-built drawings, though useful, is generally not
as valuable for position truthing. And even long-range sonar rarely if ever achieves
the panoramic birds-eye view of a site that lidar can, limiting direct line-of-sight
measurements to other targets that can be used to correlate position.
To increase accuracy in underwater surveys, sonar manufacturers have

made great progress in developing higher frequency transducers with very high
resolutions. Surveyors have similarly adapted these technologies in inventive ways
that yield more and better hydrographic data. However, confusion can sometimes
be caused when there is only focus on the maximum performance specifications of
the type of sonar on offer. While sonar resolution specifications are a good starting
point for those concerned with identifying and detailing small features, the accurate
positioning of features is of equal if not greater importance.
1.2. REPEATABILITY
The discipline of underwater survey is governed by the same immutable sci-

entific rule that underpins all geoscience: each result, i.e., the position of a sonar
target in either two-or-three-dimensional space must be able to be duplicated under
the same conditions independently. To achieve this in and around the underwater
316
Q.106 - R.20
environs of a dam requires, at the very least, triangulation between an underwater

target to a known external point, such as a local monument. The steps might be as
follows: say a feature is determined to be of interest – position ‘f’, has been detected
by a sonar mounted on a free swimming ROV – position ‘r’, and the ROV position
is itself transmitted acoustically to a distant sensor on a surface support vessel –
position ‘s’, and this position is then correlated to the terrestrial monument in the
dam, or GPS ‘t’. Considering that there are inherent accuracy tolerances within each
of these systems and that many other factors can further downgrade the certainty
of positional data underwater, the actual position of the original feature might in fact
be quite far from deduced. Therefore, just knowing that a specific feature (e.g., a
large boulder) has been measured to a certain tolerance of dimension or volume
based on sonar equipment specifications has significantly less meaning if one isn’t
confident of that feature’s actual location relative to the structure, especially if it is
near a deep-water intake. More important, has it moved since the date of the last
sonar survey, and if so, by how much?
Fig. 2
Comparative analysis (l) of two multibeam sonar scans of the same concrete
erosion feature taken two years apart.
Captures d’écran du survol: (haut l) lidar aérien du site; levé lidar au-dessus de
l’eau du puits de la station de pompage et, sous la conduite d’eau, sonar
multifaisceaux, y compris une partie du tunnel d’admission.
1.3. THE CRAFT OF DATA STITCHING
The comparative difference in range between lidar and sonar is significant,

for it means that over the course of a single survey the sonar transducer must be
repositioned, often many times, to achieve the overlap required for full coverage.
317
Q.106 - R.20
Data for each of these separate patches or strips must then be stitched together,
quilt like, for final assembly in post processing. After that, disparate sets of data (e.g.,
a bypass tunnel vs. an open spillway), acquired at various times from different types
of sonar devices – often with different operating frequencies and ranges – must be
rendered correlatively both to each other and to terrestrial coordinates.
Finally, underwater navigation and positioning not only relies on combined

data from multiple sonar devices but on other critical sensors measuring depth and
platform orientation. Again, though analogous to altitude and heading sensors used
in aerial lidar, correcting the inevitable errors in any data set becomes much more
difficult when there are few to no independently verifiable coordinates within range.
This becomes more of an issue in confined space ROV penetrations into pen-
stocks and tunnels, particularly long ones, since GPS is unavailable. Unlike above
water positioning, navigating and making real time underwater observations in the
field, confusion can sometimes occur about even the simple question of one’s cur-
rent position and orientation, making the operational skill sets of ROV pilots vital to
the acquisition of quality data. The result, namely the final underwater survey report,
is pure hydrography combined with expert technical craftmanship, both in the field
and during post processing. A geoscientific model derived from an underwater sonar
survey can rival the quality and usefulness of terrestrial lidar, although there can be
significant challenges when acquiring data, stitching scans together and, in many
cases, integrating multiple data sets.
Fig. 3
Imaging sonar scans of the upstream face of a dam. Circular patches indicate
locations of more than a dozen stationary sonar scans.
Imagerie par balayage sonar de la face amont d’un barrage. Les patchs circulaires
indiquent les emplacements de plus d’une douzaine de scans sonar stationnaires.
318
Q.106 - R.20
2. SONAR TYPES AND THEIR APPLICATIONS
As mentioned, surveys of different underwater zones often require the appli-

cation of quite different subsets of sonar technology and/or modes of deployment.
Each type of sonar possesses intrinsic capabilities and limitations for accurately
delineating a target and determining its position. Before looking at these subsets,
however, there are basic commonalities that should be considered.
All sonars use a transducer or transducer array, although these can be

deployed in many ways; mounted on a boat, a free swimming ROV, a pole or frame
fixed to a side wall or a trolley that rolls by gravity down the upstream slope of a dam –
any platform that can place the sensor at desired (and known) locations in the water
column. A transducer, as its name suggests, both sends sound pulses and receives
any that happen to be reflected (echoed) back again. A clock system measures the
total time from pulse propagation to return. Because the speed of sound in water
is constant, the distance to the target (range) is computed by multiplying that time
interval by the velocity of sound and then dividing by two; d = (v x t) /2. Water tem-
perature affects the speed of sound and sonar equipment must be field calibrated
accordingly (see section 3.1 – CTD probe). Thermoclines – the boundaries between
water layers of different temperatures, can also act as refractors, altering the vector
of the sonar signal and consequently the reliability of data.
Targets can be anything within range in the water column that successfully
reflects pulses, such as an abutment or rock surface. Unlike visible light, sonar is
not affected by turbidity, making it an ideal sensor in low to no visibility circum-
stances. Lower frequency sonars can penetrate greater distances underwater, but
the trade-off is that they do not provide the higher resolutions that high frequency
sonars can. Conversely, as with visible light, higher frequencies are capable of much
finer resolution but are more susceptible to attenuation as they travel through a fluid
medium, and therefore have a more limited range. Hard, angular surfaces are good
reflectors, rounded ones less so (although large fish can be detected). Soft surfaces
such as very light silt and marine growth and surfaces covered with severe bioac-
cumulation will absorb much (and occasionally all) of the pulse energy. The desired
return of a thin bright line produced by hard surfaces instead becomes a thick fuzzy
band, making it difficult to accurately determine the true boundary and therefore the
range to the target surface. This introduces a larger margin of error for bathymetric
calculations.
Sonar cannot penetrate solid objects and as with a flashlight in the dark, each
target casts a shadow behind it. A sonar shadow will obscure any feature contained
within it, meaning that for full coverage of a zone that contains obstructions to the
line-of-sight, (e.g., a row of dissipator piers) the transducer needs to be repositioned,
sometimes many times, to fill in the ‘acoustic picture’. Though they do obscure, shad-
ows can also help immensely with image interpretation. Some of the best sonar
imagery is comprised of both the bright return of each target surface and the dark
shadows emanating behind. Depending on shear angle, this light/dark combination
319
Q.106 - R.20
can reveal aspects of the target that might otherwise be missed, including a
remarkable degree of texture such as construction joints, debris, exposed aggregate
or rebar on a spillway slab.
Although sonar systems, regardless of type, all have critical real-time viewing
capabilities that can help inform and refine the data-gathering process as it unfolds in
the field, raw sonar data is typically also stored for post-processing. This is primarily
for three reasons. The first is that there is always anomalous data that must be
‘cleaned up’, then filtered to select the best material for assembly. Second, most field
data gathering exercises store a tremendous amount of information that includes
much that cannot reasonably be understood in real time. The importance of this
data may not have been apparent during site work but nonetheless can greatly
enhance the final product. This includes applying quantitative calculations as well
as correlative data for corrections in survey positioning as presented in the final
report. This is aside from duplicate data or data that or is not relevant or of low
quality all of which is filtered. Third, as stated previously, once processed, the entire
record must be evaluated to select the best quality image sets, stitch these together
and then sometimes further combine this assemblage with other sonar data sets
rendered according to their own formats e.g., 2D and 3D.
Archiving raw data also provides a practical option because it may have been
obvious during field operations and/or the preliminary report that post-processing
efforts should be prioritized to concentrate on smaller sectors, e.g., those exhibit-
ing obvious structural anomalies. The remaining data is of lower priority and can
be earmarked for future analysis. Post-processing requires trained, experienced
technicians and is time consuming, so this piece meal approach is an economical
means for dam owners to access the most important survey information quickly
while still maintaining perpetual access to lower priority data in a pay-as-you-go
manner.
The ultimate power of sonar is measurability. Appropriate selection of sonar

technology suitable for task, when properly calibrated and skillfully deployed, can
yield data that contains far more information than is obvious from the beautifully
rendered imagery shown here. Though renderings already tell us a lot about an
underwater structure or environment, it is the quantitative quality of data that forms
the basis of good hydrographic reporting.
2.1. IMAGING SONAR
Also referred to as navigation sonar, this sensor is typically the one ROV pilots
rely on when maneuvering around the survey field. Effective at a range of 30 m
(100 ft) while still providing quite high resolution, this sonar is used for scanning
larger areas in order to identify potential targets. It is also a powerful tool when used
320
Q.106 - R.20
in conjunction with other sensors for closer examination and detailed delineation of
targets. Depending on shear angle, imaging sonar is particularly good at revealing
planar displacement (e.g., between concrete spillway slabs), construction joints (Fig.
XX) and even textures of sediments or surface erosion.
Though standard on most ROVs for navigation, imaging sonar is also a highly
effective survey sensor when pole-mounted and pointed vertically downwards from
just beneath the water’s surface. This technique doesn’t require that vessels or
ROVs enter the water at all – so long as the water surface is directly accessible
from the structure, be it the dam crest, sidewall or abutment. Because deployment
survey points on the structure can be plotted terrestrially, accurate positioning of
underwater targets is relatively easy. Again, adjusting the graze angle can reveal a
great deal about the alignment of targets in this plane.
Fig. 4
Imaging sonar of intake grating (top) and details (bottom).
Sonar d’imagerie du réseau d’admission (en haut) et détails (en bas).
321
Q.106 - R.20
2.2. PROFILING SONAR
Profiling sonars are, as the name implies, used for scanning cross sections in
any plane that the sensor can be positioned perpendicular to. These high frequency
transducers provide excellent ranging accuracy (0.5 percent of range) but are limited
to shorter ranges due to higher attenuation through water.
Fig. 5
Conduit sediment mapping by ROV. The distance from the sonar sensor to the
sediment boundary layer is 1.22 m (4 ft.).
Cartographie des sédiments des conduits par ROV. La distance entre le capteur
sonar et la couche limite des sédiments est de 1,22 m (4 pi).
2.3. SIDESCAN SONAR
The two integrated transducers for this device, each scanning opposite sides,
are deployed using a torpedo shaped ‘fish’ that is towed behind a moving vessel
(Figure 8). As with imaging sonar, shear angle is critical for best results, and oper-
ators must maintain a constant speed and heading to attain this. Heights of targets
above bottom can be calculated using simple geometry, since distances (transducer
to top and bottom of water column), and angle (to a shadow’s far edge) can be
322
Q.106 - R.20
Fig. 6
Gate gain sonar profiles. Imaging sonar in profiling mode (l) and profiling sonar (r).
Profils de rainures de vannes portes. Sonar d’imagerie en mode de profilage (l) et
sonar de profilage (r).
obtained from the sonar record. The scan width on-bottom of a single track, called
swath, decreases (bands gets thinner) in shallower water, requiring a tighter survey
pattern for sufficient overlap between swaths. For surveys from a surface vessel,
water exceeding 100 m (328 ft) in depth will result in a less accurate survey, so a
towed ‘fish’ operating at depth is used.
Fig. 7
Sidescan beam angle settings for different areas of the dam. Sidescan of upstream
and downstream vertical faces (especailly useful for large dams) will appear as
imaging sonar renderings as in Fig. 4.
Paramètres d’angle du faisceau de balayage latéral pour différentes zones du
barrage. Le balayage latéral des faces verticales en amont et en aval
(particulièrement utile pour les grands barrages) apparaîtra sous forme de rendus
de sonar d’imagerie comme sur la figure 4.
323
Q.106 - R.20
Fig. 8
Side-scan image of a spillway showing dissipator piers, some badly eroded.
Image de balayage latéral d’un déversoir montrant des piliers de dissipateur,
certains très érodés.
2.4. MULTIBEAM SONAR
Multibeam sonar is unique in that it uses not one transducer but an array of
many, so instead of scanning an area by sequentially stepping through an arc, it
takes in a complete profile slice of topography simultaneously. As the transducer
array moves across a surface e.g., fixed from a survey vessel, it builds a 3D model,
called a point cloud, often rendered as an elevation heat map. A high frequency
multibeam sensor deployed at the surface has an effective range of 100m (328 ft.).
Within this range, accuracy is about 5 cm (2 in.) and this decreases beyond 100
meters depth, at which point an ROV is typically deployed with navigational tie-ins
to the support vessel, dam structure or shore.
324
Q.106 - R.20
Fig. 9
Combined rendering of dissipation piers: (centre)
multibeam point cloud data surrounded by the same data presented
as an elevation heat map.
Rendu combiné des piles de dissipation: données de
nuages de points multifaisceaux (au centre) entourées des mêmes données
présentées comme une carte de chaleur d’élévation.
Fig. 10
Multibeam sonar images of powerhouse internal areas.
Images sonar multifaisceaux des zones internes de la centrale.
325
Q.106 - R.20
2.5. OTHER EXAMPLES OF SONAR DATA INTEGRATION
Where visibility permits, video data is invaluable for providing detail and context
to the sonar record and video stills can even be mapped to sonar-deduced surfaces.
For leak mapping, sonar often points the way to suspect locations through large
field identification of structural anomalies such as offsets and voids that can then be
investigated more closely with dye injection, telltails or acoustic leak detection.
Fig. 11
(r) Dam inlet channel (multibeam bathymetry/contours), sidewall (imaging),
and internal areas (ROV/multibeam). (l) Detail of inlets, gate slots, and
turbine galleries.
(r) Canal d’entrée du barrage (bathymétrie multifaisceaux / contours),
paroi latérale (imagerie) et zones internes (ROV / multifaisceaux). (l) Détail
des entrées, des rainures des vannes et des galeries de turbine.
3. OTHER POSITIONING SENSORS
To enhance the survey result, sonar is almost always used in conjunction with
other sensors, especially when deployed by ROV. These can include depth, CTD,
INS, as well as other sonars and acoustic beacons, plus an array of cameras for
follow up spot investigation. Dam structures can vary greatly in design, size and
overall layout, each presenting unique challenges to the acquisition of quality data
that is both verifiable and repeatable.
3.1. CONDUCTIVITY, TEMPERATURE AND DEPTH (CTD) PROBE
Changes in water temperature, salinity, and depth (pressure) matter because

each affect the speed of sound. Anything that affects water density must be mea-
sured and compensated for through calibration of sonar sensors, or inaccurate
326
Q.106 - R.20
readings will result. Conductivity is the basis for most total dissolved solids cal-
culations, and these are critical for hydrographic surveys, especially in seawater
because of large variations in salinity. Most large dams in the world are found in
bodies of freshwater, which has relatively low conductivity compared to seawater.
Within these lower parameters, however, freshwater conductivity can vary widely
due to geology effects. Dissolved minerals from runoff and the type of rock or soil
that groundwater passes through can cause significant differences in conductivity.
Seasonal variations in precipitation can also have an affect through dilution in wet
weather and concentration in dry. Temperature directly affects conductivity as well
as water density (pressure), and pressure increases with depth. CTD probes are
standard equipment in all underwater surveys.
3.2. INERTIAL NAVIGATION SYSTEMS (INS)
Inertial Navigation Systems (INS) are navigational systems capable of calcu-

lating position relative to some reference point or absolute coordinates. In navigation
and positioning they are used for virtually everything that moves while it scans, either
on the water surface or beneath. Their purpose is to correct for sensor movement,
in three planes – pitch, yaw and roll, or more – surge, heave, and sway. Imagine
a flashlight floating at night in perfectly calm water, with it’s beam pointing straight
downward to illuminate a circular area on the bottom. If a wind picked up, the flash-
light would start drifting with the air and water currents, bobbing around in the choppy
water, so that even small surface movements would become significant at depth and
the beam would not be pointing to one spot but continually dancing around a much
larger area. Rather than trying to keep the flashlight always pointed absolutely ver-
tical, the INS compensates for off-axis data by continually correcting sensor data
records against the movements that caused it. This paints an overall data ‘picture’
consistent with reality, regardless of sensor position and movement.
3.3. EXTERNAL TRACKING AND POSITIONING SYSTEMS
There are inherent inaccuracies of any inertial sensor, and these are magnified
over time. In order to get around this positional drift, INS systems are augmented by
some sort of reference device. The purpose of these navigation aids is to decrease
the inherent drift in INS calculated position. Global Positioning System (GPS) is
common for surface applications. However, a significant number of dam sites rely
on other local references such as traditional terrestrial surveys that are tied to the
structure as benchmarks. If GPS is used, it is common to set up temporary substa-
tions around the dam that correct for temporal errors in the satellite signal, which is
subject to fluctuations due to atmospheric and other conditions. Subsea reference
devices are usually used to plot the position of an ROV, which carries a pinger that
allows its position to be triangulated from a vessel or shore using a dual beacon.
327
Q.106 - R.20
The accuracy of survey sensors on a vessel decreases as depth increases. In

very deep water i.e., over 100 m (328 ft.), the ROV position, and (by extrapolation)
all sonar targets within the field of view, can only be accurately determined by the
aid of a local reference, called a long base line (LBL) system. Beacons are set at
known locations on the bottom and used for triangulation of points between them.
Depending on the area and layout, accuracy is typically sub-centimeter (0.4 in.).
Common in offshore petroleum surveys, this level of accuracy could be required for
modern inland deep-water interventions. ROVs have moved beyond straightforward
inspection and survey tasks and are now conducting or assisting with remote con-
struction projects: bulkhead installations, trash rack cutting and large debris removal.
The close tolerances in measurements required for some of these tasks makes this
offshore deep-water positioning technology ideal for such inland applications.
4. ACCESSING AND SURVEYING CONFINED SPACES
Even very small ROVs can now carry sonar sensors into confined spaces in
and around a dam structure. This allows for an even more comprehensive survey
picture, although it brings another set of unique challenges. No GPS signal means
that the onboard INS must be relied upon for positioning and navigation. Umbilical
payout adds a level of verification for ROV penetration distance from the access
point. Impressively long penetrations of headrace tunnels of up to 14 km (8.7 miles)
are inherently subject to increased error in INS accuracy. Locally observed features
such as adits and overhead shafts can help correct the
5. BRINGING BACK QUALITY DATA FROM THE FIELD – FACTORS
Starting with field work, many factors and variables can affect site conditions
and have a negative impact on survey results. These include water levels and flow
velocities; weather, (especially when vessels are to be operating in open water);
and plant processes, most notably the requirements to be able to suddenly abort
shutdowns and restore water flow. Each can have a negative bearing on the survey
outcome. For marine surveyors, the key is to be forearmed with as much reliable
plant knowledge as possible and to plan accordingly.
5.1. ORDER OF OPERATIONS
Hydrographic surveys are best done after completing terrestrial survey work.
If this has not yet been done, consult the hydrographer beforehand who, based on
drawings and a site visit, will be able to recommend zones and/or specific locations
for essential monuments. The land survey plan contains the necessary reference
stations for subsea tie-ins and forms the basis for what follows. Lidar bathymetry
328
Q.106 - R.20
is next, for mapping shallow water zones less than 0.5 m (20 in.) in depth. This is
correlated with above water data and followed by any order or combination of sonar
surveys:
• shoreside fixed scans e.g., on pole or trolley from the dam structure, walls or
embankments
• live vessel open water bathymetry
• ROV open water, vessel or land supported
• ROV internal conduits and confined spaces
5.2. PROJECT PLANNING
For marine survey fieldwork, the primary goal beyond worker safety is the
acquisition of quality data. Gaps in survey data can’t be ‘fixed in the edit’. For
any large scale or critical underwater survey projects, comprehensive site visits are
essential to developing an effective work plan. Hydrographic and ROV field opera-
tors find that many survey problems encountered on site are avoidable. Unexpected
site conditions can often be adapted to, but anything that interrupts the planned
workflow can have a negative impact on survey results and/or add costs. Relevant
advance information about the dam structure is not always easily accessible, and
in many cases plan or construction drawings, even as-builts, are not accurate. Ven-
dors appreciate information about actual (monitored/measured) vs. assumed site
conditions. This ensures there are no unwelcome surprises when field work starts.
These can include:
• active water level monitoring vs. by historical assumption

• for ROV operations, turbidity monitoring; though visibility is not required for
sonar, video can be extremely valuable for detailed observations.
• measured flow velocity vs. velocity estimated by volume calculations based
on spillway, channel or conduit design.
• where temporary flow diversion, reduction or shutdown is required, establish
realistic windows of operation, especially in unpredictable weather, and plan
for contingencies.
• measured access dimensions for confined spaces, vs. access dimensions
taken from drawings
• general site access, road conditions and road conditioning if necessary, since
cranes or other heavy equipment are sometimes needed to launch survey
vessels
Owners know best the environmental conditions into which the vendor’s tech-
nology will be introduced. Providing vendors with prior survey or inspection reports
can also help produce a better reporting result, especially when establishing trends
of erosion or sedimentation.
329
Q.106 - R.20
SUMMARY
We now live in a time when the knowledge base for underwater portions of a
dam structure derived from sonar surveys can now be relied upon for very accurate
measurement purposes. Trending for volume of structural erosion or accumulation
of sediment is a significant development and will only improve as the cumulative
survey record is added to over the long term. Disparate data sets (including non-
sonar data such as lidar, acoustic leak detection and video) are being effectively
correlated. Sonar fine measurement of underwater features e.g., for the purpose
of future installations of sealing bulkheads is now a common task. Adaptation of
sonar for specific tasks such as gate gain surveys and the creative use of side
scan technology for rapid imaging of vertical walls are just two examples of recent
innovations, which will undoubtedly continue.
Reliable underwater metrology data is neither easily acquired nor easily

post-processed except with skilled ROV and survey vessel operators deploying
appropriate sonar technology. The result, especially when correlated to historical
records and other survey data, adds value to the overall risk management data
base for dams in ways previously unimaginable.
Bill Sherwood has worked in the ROV industry for over 30 years, has indepen-
dently consulted to the New York City DEP and served in management positions for
Oceaneering International in both the Gulf of Mexico and the Canadian East Coast.
He is currently Manager of International Business Development at ASI Marine’s
head office in Stoney Creek, Ontario.
330
Q.106 - R.21
-------
-------
NUMERICAL ANALYSIS TO SUPPORT THE REHABILITATION AND

LONG-TERM MONITORING OF THE WABA DAM (∗)
Eric TIEDJE
Geotechnical Engineer,
HATCH, Mississauga
C. Richard DONNELLY
Global Principal Consultant, Dams and Waterpower,
HATCH, Niagara Falls
Kai-Sing HO
Global Geotechnical Director,
HATCH, Markham
Paul TOTH
Senior Engineer, ONTARIO POWER GENERATION,
Niagara on the Lake
Richard A. DALE
Renfrew
Ramy SAADELDIN
Niagara on the Lake
Luigi PERRA
Project Manager, ONTARIO POWER GENERATION,
Renfrew
CANADA
∗ Modélisation numérique en appui à la réhabilitation et la surveillance à long terme du
barrage Waba
DOI: 10.1201/9781003211808-116 331

Q.106 - R.21
SUMMARY
This paper outlines a case study on the analysis, design, and long-term mon-
itoring program for the rehabilitation of the 18m high Waba Dam, one of the largest
embankment dams built on sensitive Champlain Sea clay. Since construction in
1976, the dam has experienced extensive settlement and lateral deformation while
porewater pressures in areas of the foundation remain significantly higher than
hydrostatic conditions would suggest. Recent geotechnical investigations indicate
that portions of the foundation have experienced a decrease in shear strength due
to a disturbance to the clay fabric. While the dam is not considered an immediate
risk of instability, a rehabilitation program was undertaken to ensure the continued
safe operation of the dam until 2176. As part of the program, a detailed numerical
deformation analysis was undertaken of the dam and foundation. A two-dimensional
deformation analysis was undertaken with the finite element method using advanced
material models to describe the visco-plastic consolidation and strain softening
nature of the foundation clay. The analysis provided predictions of future settle-
ment and insights on the risk of further disturbance to the sensitive foundation.
Based in part on these predictions, appropriate methods for monitoring the dam
were developed.
RÉSUMÉ
Cet article présente une étude de cas sur l’analyse, la conception et le pro-
gramme de surveillance à long terme pour la réhabilitation du barrage Waba de 18 m
de haut, l’un des plus grands barrages en remblai construits sur de l’argile sensible
de la mer de Champlain. Depuis sa construction en 1976, le barrage a subi un tasse-
ment et une déformation latérale importants, tandis que les pressions interstitielles
dans les zones de la fondation restent considérablement plus élevées que les condi-
tions hydrostatiques ne le suggèrent. Des études géotechniques récentes indiquent
que certaines parties de la fondation ont subi une diminution de la résistance au
cisaillement en raison d’une perturbation du tissu d’argile. Bien que le barrage ne
soit pas considéré comme un risque immédiat d’instabilité, un programme de réha-
bilitation a été entrepris pour assurer la poursuite de l’exploitation sûre du barrage
jusqu’en 2176. Dans le cadre de ce programme, une analyse numérique détaillée
de la déformation du barrage et des fondations a été entreprise. Une analyse de
déformation bidimensionnelle a été entreprise avec la méthode des éléments finis
en utilisant des modèles de matériaux avancés pour décrire la consolidation visco-
plastique et la nature de l’adoucissement des déformations de l’argile de fondation.
L’analyse a fourni des prédictions sur les établissements futurs et des informations
sur le risque de nouvelles perturbations pour la fondation sensible. Sur la base en
partie de ces prévisions, des méthodes appropriées de surveillance du barrage ont
été développées.
332
Q.106 - R.21
1. INTRODUCTION
At the time of its construction in 1976, the Waba Dam was the largest embank-
ment dam built on a Champlain Sea clay deposit (Leda clay), a notoriously sensitive
and soft marine clay found in eastern Canada and parts of the United States. In
close agreement with the original design predictions, the 18m high dam settled a
maximum of 1.8m between 1976 and 2020 and is expected to continue to ettle for
decades to come. Recent geotechnical investigations indicated that portions of the
clay foundation have experienced a loss in shear strength as a result of the over
loading, which has led to a slow rate of dissipation of excess porewater pressure,
lateral movement, and de-structuring in the clay fabric.
This paper presents a review and interpretation of the historic behaviour of

the dam and foundation and provides details on a numerical analysis program
undertaken to support the rehabilitation and long-term monitoring program.
1.1. BACKGROUND AND ORIGINAL DESIGN
The Waba Dam is located on the Madawaska River, near the 83 MW Arn-
prior Generating Station, located 60 km west of Ottawa, Ontario, Canada. Owned
and operated by Ontario Power Generation (OPG), it was designed by Hatch Ltd.
(formerly Acres International Ltd.) and was constructed between 1973 and 1976.
Fig. 1
Original design of Waba Dam, plan view.
Conception originale du barrage de Waba.
333
Q.106 - R.21
The dam is approximately 1100 m long, with a maximum height of 18 m. The

dam consists of a compacted clay core with downstream inclined chimney filter and
transition zones, as well as an upstream filter, with associated transition layers, rock-
fill and riprap. Two large weighting berms, approximately 5 m to 10 m high, extend
upstream and downstream from the core to a maximum distance of approximately
180 m and 100 m, respectively. A key trench, approximately 6m wide and 2m deep
extends through the surficial weathered and fissured clay crust into the marine clay
deposit. The sides slopes of the dam core and weighting berms are 3H:1V and
4H:1V respectively. As is illustrated in Fig 1, the Waba Dam was discretized into
three areas in recognition of expected differences in settlement due to the variable
height of fill with each area built to a different crest elevation.
Fig. 2
Original design of Waba Dam, typical Area 1 section.
Conception originale du barrage de Waba, section typique de la zone 1.
1.2. GEOTECHNICAL CONSIDERATIONS
Due to the nature of the foundation soils, extensive geotechnical investiga-

tions have been performed prior to the construction of the dam and in the decades
following [1]. The foundation consists of a surficial weathered clay crust overtop of
an approximately 75m deep deposit of Champlain Sea clay, underlain by varying
thickness of sand and gravel and till, which in turn are underlain by crystalline lime-
stone bedrock. The total thickness of overburden under the dam is approximately
80m to 90m. The abutments are also composed of Champlain Sea clay, with an
average slope prior to construction of between approximately 10H:1V to 40H:1V.
The foundation clay is discretized into an Upper Clay unit and Lower Clay unit, with
slight variations in properties, illustrated in Tables 1 and 2.
Champlain Sea clay can be characterized by a highly disturbed and open-

bonded structure with significant salinity which gives it its unique engineering
334
Q.106 - R.21
Table 1
Typical stratigraphy under the centerline of the dam, Area 1.
MATERIAL TYPICAL DEPTH, M TYPICAL ELEVATION, M TYPICAL THICKNESS, M
Dam Fill 0 to 18 101 to 83 18
Upper Clay 18 to 44 83 to 57 26
Lower Clay 44 to 76 57 to 25 32
Sand, Sand and Gravel, 76 to 79–88 25 to 22–13 3–12
Glacial Till (varies) (varies) (varies)
Limestone Bedrock 79–88 (varies) 22–13 (varies)
Table 2
Typical Range in Material Properties for the Foundation Clay.
UPPER CLAY LOWER CLAY
Bulk Unit weight, kN/m3 14–18 16–19
(increasing with depth) (increasing with depth)
Moisture Content, % 48–93 46–81
Plasticity Index 25–56 19–39
Clay Fraction, % 65–78 51–69
Salinity, ppm 4–24 7–21
Over Consolidation Ratio (Preconstruction) 1.1–1.5 1.1–1.5
Critical State Friction Angle, ◦ 21–29 26–28
Undrained Shear Strength (Undisturbed), kPa 15–135 20–150
(increasing with depth) (increasing with depth)
Virgin Compression Index 0.4–4.0 0.3–0.6
properties. These clays typically have a relatively low density, a high compress-
ibility, a firm to stiff undisturbed shear strength, and a very low shear strength when
the soil fabric is disturbed (high sensitivity). The undisturbed shear strength of is typi-
cally proportional to the apparent preconsolidation pressure of the clay. Importantly,
the clay is strain-softening in nature. Figures 3 illustrates the typical mechanical
behaviour for Champlain Sea clay.
2. HISTORICAL BEHAVIOUR
The shear strength of Champlain Sea clays is highly impacted by the stress
and deformation history of the soil. Properly interpreting and studying the historic
behaviour of the foundation and the dam were critical to the understanding of the cur-
rent conditions and expected future conditions. The Waba Dam has been extensively
instrumented and monitored by OPG and various consultants and researchers since
its construction [1]. A detailed description of this monitoring and historic behaviour
of the foundation is presented under separate cover [4, 5]. A summary of the key
observations of the dam’s behaviour are presented in following discussions.
335
Q.106 - R.21
Fig. 3
(A) Typical consolidation curves for undisturbed and remolded Champlain Sea
clays, figure from [2], (B) typical stress-strain and pore pressure-strain response
for Champlain Sea clays, figure from [3].
(A) Courbes de consolidation typiques pour les argiles de la mer de Champlain
non perturbées et remoulées, figure de [2], (B) réponse typique de
contrainte-déformation et de pression interstitielle pour les argiles de la mer de
Champlain, figure de [3].
2.1. SETTLEMENT
The original design predicted the dam would settle between 1.9 and 4.0 m by
2026, and as of 2018, observed settlement of the dam agrees well with the lower
bound of the original design estimate. However, early in the operation of the dam, the
observed settlement was significantly less than predicted. As illustrated in Figure 4,
there has been negligible settlement of the embankment itself with crest settlement
closely following measured foundation settlement.
2.2. LATERAL DEFORMATION
The inclinometer data shows that the foundation is laterally deforming at a

specific elevation within the Upper Clay, occurring at a ratio of approximately 0.2
with respect to the vertical settlement [5]. The results of the investigation program,
including Cone Penetration Testing (CPTu) and insitu vane shear testing in 2018
showed that the clays within this relatively narrow zone had destructured, forming a
weak shear plane. Testing beneath the weighting berm including CPTu, vane shear
testing, cross-hole seismic and cross-hole tomography tests showed that this zone
of destructuring does not extend beyond the toe of the main dam into the foundation
of the weighting berms.
336
Q.106 - R.21
Fig. 4
Predicted maximum crest settlement in Area 1 versus observed behaviour.
Tassement maximal en crête prévu dans la zone 1
par rapport au comportement observé.
Fig. 5
(A) Cumulative lateral deformation measured in the foundation clay near the
centerline in Area 1(Inclinometer SI2), 1976 to 2017 and (B) change in the
maximum lateral deformation with time.
(A) Mesure cumulative de la déformation latérale dans l’argile de fondation près de
la ligne médiane dans la zone 1, 1976 à 2017 et (B) variation de la déformation
latérale maximale avec le temps.
337
Q.106 - R.21
2.3. POREWATER PRESSURE
The long-term trends in the changes in porewater pressures were interpreted

from multiple generations of piezometers of varying types installed in similar loca-
tions and elevations. All the measured foundation piezometric pressures remain well
above the steady-state hydrostatic pressures with the highest pressures located just
below the base of the dam in an area corresponding to the location of the destruc-
tured shear plane. As illustrated in Figure 7, despite significant settlement, there has
been only limited pore pressure dissipation. The extrapolated trend suggests that
primary and secondary consolidation is occurring simultaneously and excess pore
Fig. 6
The piezometric head profile in the foundation under the central core at select
points of time.
Profil de la tête piézométrique dans la fondation sous le noyau central à certaines
dates choisies.
338
Q.106 - R.21
pressures in the foundation beneath the main dam (above the normal water level in
the reservoir) will be present for decades to come.
Fig. 7
Observed changed in the piezometric head between El. 80.2 and 81.7
under the central core.
Changements observés dans la tête piézométrique entre El. 80,2 et 81,7
sous le noyau central.
2.4. PRECONSOLIDATION PRESSURE
The preconsolidation pressure in the foundation was measured prior to the

construction of the dam and during the rehabilitation design period in 2018. Recent
measurements in the Upper Clay beneath the highest portion of the main dam
indicate there has been no increase in preconsolidation pressure, as shown in
Fig 8. Instead, the measured preconsolidation pressure is below the current effec-
tive stress, indicating an Over Consolidation Ratio (OCR) of less than one in the
elevation range corresponding to 1) the highest level of shear deformation, 2) where
the applied stress from the original dam construction exceeded the preconsolidation
pressure of the foundation, and 3) where porewater pressures remain significantly
above hydrostatic conditions. While such an OCR is generally inconsistent with cur-
rent the practice of geomechanics, it likely reflects destructuring of the clay fabric that
has occurred in this portion of the foundation. While not illustrated, the investigation
also showed that outside of the footprint of the central core, approximately 30 m
downstream from the centerline, the present-day OCR in the Upper Clay is greater
than one. As the undrained shear strength of the Champlain Sea clay is closely cor-
related to the preconsolidation pressure, the measured undrained shear strength
of the foundation within this zone has likewise decreased from preconstruction
values.
339
Q.106 - R.21
Fig. 8
Change in the measured preconsolidation pressure since construction and the
current measured Over Consolidation Ratio in the foundation under the centerline
of the dam in Area 1.
Changement de la pression de préconsolidation mesurée depuis la construction et
du rapport de sur-consolidation mesuré actuel dans la fondation sous l’axe du
barrage dans la zone 1.
2.5. INTERPRETATION
The original design expected the foundation clay would ultimately consolidate
under weight of the embankment, leading to an expected long-term increase in
shear strength once excess pore pressures had dissipated. However, as of 2018,
no measurable increase in shear strength has materialized. The measured precon-
solidation pressure and undrained shear strength strongly indicates that the upper
zone of the Upper Clay layer under the highest portion of the main dam has destruc-
tured. While total settlements are closely in line with the design predictions, the
mechanism for settlement did not follow the original assumptions. Settlement of the
dam cannot be solely attributed to changes in effective stress (i.e. primary consoli-
dation) and most of the settlement is attributed to secondary consolidation. This form
of consolidation is visco-plastic in nature [6,7,8], and, importantly, is not expected
to lead to future increases in shear strength of the clay, at least for the foreseeable
future.
340
Q.106 - R.21
3. REHABILIATION DESIGN
In 2018, Hatch was retained to prepare a rehabilitation design in order to

ensure the safe operation of Waba Dam until 2076. Conceptually, the rehabilitation
design consisted of the following components:
• Reconstruct the top of the dam to ensure an adequate crest and core elevation,
while at the same time minimizing the stress applied to the disturbed portion
of the foundation;
• Provide a filter on the downstream face of the dam, to account for previous
settlement of the original inclined chimney filter;
• Reinforce the downstream weighting berm with an earth fill buttress in order
to improve the stability of the dam and address local surface drainage
issues;
• Improve drainage elsewhere on the dam and address various minor opera-
tional concerns;
• Update the instrumentation and long-term monitoring program.
Figures 9 and 10 illustrate the conceptual design for the rehabilitation.
Fig. 9
Cross section of the rehabilitated dam crest, downstream is to the right.
Coupe transversale de la crête du barrage réhabilité, l’aval est à droite.
Fig. 10
Location and size of the downstream buttress berm, downstream is to the left.
Emplacement et taille de la berme de contrefort en aval, l’aval est à gauche.
341
Q.106 - R.21
4. STABILITY ANALYSIS
As part of the rehabilitation design, a stability analysis was performed consid-

ering the changes in the foundation shear strength since the original construction,
in accordance with Canadian dam safety standards. The stability of the dam was
assessed using a two-dimensional limit equilibrium analysis with the Morgenstern-
Price method, implemented in the commercial software Slope/W. Due to the
sustained excess porewater pressures observed in the foundation, undrained shear
strength conditions were adopted for the clay foundation which yielded a conser-
vative, but likely more realistic, assessment of stability. Importantly, non-circular
or ‘optimized’ failure surfaces were considered in order to capture the expected
‘block-type’ failure pattern along the known weak destructured shear plane. The
general shape and location of the critical failure surface ultimately agreed well with
the findings from the deformation analysis presented in Section 5.2.3.
While extension of the destructured zone beyond the limits of the main dam
underneath the weighting berm has not occurred in the nearly 50 years since the
Waba Dam was constructed, the current design is intended to ensure that the struc-
ture continues to satisfy the current CDA dam safety standards for the next 57 years
(i.e., until the dam reached 100 years in age). As would be expected, the pre-
dicted factor of safety is highly dependent on the lateral extent of the destructured
zone under the downstream weighting berm. To assess the potential for instability
assessments were performed for various assumptions regarding the extension of
the destructured zone including the highly unlikely assumption that the entire Upper
Clay deposit could destructure underneath the dam and weighting berm (Figure 11).
Fig. 11
Conceptual illustration of the critical failure surface and the influence of the
potential lateral extent of the destructured (remolded) zone.
Illustration conceptuelle de la surface critique de rupture et de l’influence de
l’étendue latérale potentielle de la zone remodelée.
The results of this analyses∗ , as illustrated in Figure 12, show that the reha-
bilitated dam will meet currently safety guidelines for normal conditions, so long as
the extent of the destructred zone does not extend approximately 30 m from the
centerline of the dam and that the dam will remain stable even if the entire Upper
Clay deposit destructures to a distance of 120 m from the centerline of the dam.
∗ The results are shown for the normal loading condition for illustrative purposes
342
Q.106 - R.21
Fig. 12
Variation in the static Factor of Safety with the lateral extent of the disturbed
‘remolded zone’, i.e. the destructured zone.
Variation du facteur de sécurité statique avec l’étendue latérale de la “zone
remodelée” perturbée.
5. DEFORMATION ANALYSIS
The objective of the deformation analysis was: 1) to estimate the future change
in crest and core elevations, and 2) to assess the future strain deformation and
stress concentration in the clay foundation such that the risk for future disturbance
and strain-softening can be understood. The deformation analysis was undertaken
by means of the Finite Element Method in a 2D plane-strain framework, using the
commercial software PLAXIS 2D and the newly developed Creep-SCLAY1S soil
model, which can describe both the visco-plastic secondary consolidation behavior,
as well as the destructuration-induce strain softening behavior. The analysis was
time-dependent in nature (transient with implicit integration) and assessed both the
deformations in the dam and foundation, along with the porewater pressures in the
foundation. The full construction of the dam was simulated to properly represent the
stress and deformation history throughout the foundation.
The model consisted of 3910, 15-noded triangular elements. The lower bound-
ary was selected as the bottom of the Lower Clay layer as El. 21 m. No groundwater
flow was considered between the Lower Clay and the underlying sand and till due to
the low permeability of the clay. Time stepping was performed using an implicit inte-
gration scheme. The 2D deformation analysis was calibrated by a more simplified,
343
Q.106 - R.21
but well-established, model for settlement of embankments on soft clays, the 1D

Gibson-Lo model for secondary consolidation [12]. The findings from the Gibson-Lo
analysis were used to confirm the validity of finite element analysis as an exercise
in due diligence. While not presented in this report, good agreement was obtained
between the two models with regards to total crest settlement.
It should be noted that a similar deformation analyses was performed for aca-
demic purposes between 2015 and 2017 [11], which provided valuable insight for the
development of the modelling plan for the rehabilitation works. The seismic deforma-
tion of the dam is not presented herein, however other parties have demonstrated
the stability of the dam under seismic conditions to be adequate [9,10].
5.1. MATERIAL MODELS
Two soil constitutive models were adopted: “Creep-SCLAY1S” was used for
the foundation clays, except for the surficial curst, and “Hardening Soil” was used
for all other materials. The “Creep-SCLAY1S” model was developed by PLAXIS and
released in 2018. The model can simulate rate-dependent elasto-visco-plastic defor-
mation behaviour, while also capturing the inherent anisotropy and destructuration
potential observed in Champlain Sea clays [13]. Destructuration of the clay is a
particular concern, as it is associated with the development excess pore pressure
and strain-softening behaviour. This constitutive model allows for an explicit deter-
mination of porewater pressure at all stages of the analysis. The “Hardening Soil”
constitutive model was based on the Mohr-Coulomb failure criteria with a hyper-
bolic stress-strain response, a function of the stress conditions. The model is well
established for geomaterials and has a long track record of use.
One of the challenges with utilizing complex constitutive models such as

Creep-SCLAY1S is properly identifying material parameters. Typically, most param-
eters would be directly measured, or indirectly and reasonably assumed, based
on an extensive laboratory testing program. In this case, for an existing dam with
sufficient monitoring data, there was an opportunity to confirm laboratory-derived
parameters by reproducing real-world results. This was accomplished using 43
years of available monitoring data to support selection of model parameters. This
calibrated method was considered to significantly reduce the uncertainty associ-
ated with future predictions from such numerical analysis. Ultimately the material
parameters were determined from three sources:
• Direct or indirect measurements from laboratory and field tests conducted

both prior to and follow dam construction, specifically: conventional and 60-
day oedometer tests, vane shear tests, CU triaxial compression test and CU
triaxial extension tests;
• Reproducing historic behaviour to back-calculate material properties; and
• Reported values for similar materials.
344
Q.106 - R.21
Table 3
Creep-SCLAY1S Material Properties Adopted for the
Foundation Clays.
UPPER CLAY, UPPER CLAY, UPPER CLAY, LOWER
SOIL ZONE 1 ZONE 2 ZONE 3 CLAY
Saturated Unit 15.3 15.5 16.0 16.4
Weight, kN/m3
eint 2.21 2.00 1.78 1.62
POP 0.0 0.0 0.0 0.0
OCR 1.0 1.0 1.0 1.0
φcv , ◦ 23 28 28 28
C ci 0.483 0.465 0.480 0.400
CS 0.057 0.074 0.055 0.049
Cα 0.025 0.026 0.027 0.043
St 5 5 5 5
ν 0.2 0.2 0.2 0.2
τ 1 1 1 1
ω 50 70 40 30
ξ 15 15 11 10
ξf 0.2 0.3 0.3 0.3
K 0−NC 0.7 0.7 0.7 0.6
Table 4
Hardening Soil Material Properties Adopted for Fill Materials.
GRANULAR IMPERVIOUS RANDOM UPPER CLAY
MATERIAL FILL FILL FILL CRUST
Drainage Type Drained Drained Drained Drained
Saturated Unit 20.0 18.6 17.5 16.5
Weight, kN/m3
Phi, ◦ 35 28 25 28
cref , kPa 1 5 5 5
ref
E 50 , MPa 40 20 15 20
ref
E oed , MPa 40 20 15 20
ref
E ur , MPa 120 60 45 60
m 0.5 0.5 0.5 0.5
5.2. DEFORMATION ANALYSIS RESULTS
5.2.1. Crest settlement
Figure 13 illustrates the predicted and observed settlement of the crest of the
dam (SM14), the top of the foundation directly underlying the crest (FSG12) and the
top of the foundation approximately 30 m downstream of the dam crest (FSG6) in
Area 1.
345
Q.106 - R.21
Fig. 13
Predicted and observed settlement, Area 1.
Habitat prévu et observé, zone 1.
5.2.2. Porewater pressures
Fig 14 illustrates the predicted and observed piezometric head in the founda-
tion clay in the destructured zone. Porewater pressure in excess of the hydrostatic
conditions are expected to persist within the destructured zone for the full service-life
of the dam.
5.2.3. Shear strain levels
Fig 15 illustrates the predicted development of deviatoric strain within the dam
and foundation. Deviatoric strains are expected to be highest within the upmost
portion of the Upper Clay unit and higher downstream of the core as compared to
upstream. These observations are consistent with findings of the limit-equilibrium
analysis and indicate the critical failure surface for the dam will be a block-type
pattern, including a long sub-horizontal plane through the weakest portion of the
Upper Clay. The analysis also identifies a potential high-strain zone in the foundation
near the toe of the downstream weighting berm. This specific area has not been
extensively investigated or monitored previously.
346
Q.106 - R.21
Fig. 14
Predicted and observed piezometric head under the centerline of the dam (Area 1)
between El. 75.7 and 72.8, from prior to construction to the end of the current
service life.
Tête piézométrique prévue et observée sous la ligne médiane du barrage (zone 1)
entre El. 75,7 et 72,8, de la période précédant la construction jusqu’à la fin de la
durée de vie actuelle.
Fig. 15
Predicted deviatoric strain levels in the dam and foundation (Area 1) in 2020 prior
to rehabilitation and in 2076 at the end of the design service life.
Niveaux de déformation déviatoriques prévus dans le barrage et la fondation (zone
1) en 2020 avant la réhabilitation et en 2076 à la fin
de la durée de vie utile de conception.
347
Q.106 - R.21
Fig. 16
Predictions of the deformation analysis with highlighted green areas corresponding
to deviatoric strains in excess of 8%.
Prédictions de l’analyse de déformation avec des zones vertes en surbrillance
correspondant à des déformations déviatoriques supérieures à 8%.
6. CONCLUSIONS
Good agreement was ultimately obtained between the deformation analysis

and the historic monitoring observations. The deformation analysis provided valu-
able insight of the expected behaviour of the foundation, specifically helping to
confirm the interpretation of the conditions of the foundation. Overall, the following
conclusions from the analysis and interpretations are noted:
1. The settlement and deformation of the dam will continue up to, and beyond,
the end of its current service life, albeit at a slightly decreasing rate;
2. Porewater pressures in excess of hydrostatic conditions will be present in the
foundation up to, and beyond, the current service life of the dam;
3. Where the foundation clay was loaded significantly beyond its characteristic
preconsolidation pressure, destructuring of the sensitive particle fabric has
likely occurred, leading to a decrease in clay shear strength;
348
Q.106 - R.21
4. With ongoing deformation, the disturbed zone in the foundation may potentially
spread downstream, reducing overall stability;
5. The highest levels of shears strains within the foundation are expected to occur
directly under and slightly downstream of the central clay core;
6. Strain softening may become a concern within the upper zones of the
foundation as the dam approaches the end of its current design life;
Understanding the stress and deformation history of the clay foundation was
critical for the rehabilitation design and safe continued operation of the Waba dam.
For rehabilitation projects, it is important understand and critically evaluate the
original design process wherever possible.
6.1. MONITORING PROGRAM
There are two foreseeable scenarios for the foundation. First, the destructured
zone will eventually consolidate, thereby increasing in shear strength and leading
greater stability for the dam. The second possible scenario, the foundation clay
continues to deform and destructure, likely leading to an expansion of the disturbed
zone under the dam and reduced overall stability. As there is little precedence for the
situation, it is difficult to determine how the foundation will behave and worst-case
conditions should be assumed.
Settlement will continue to be monitored with conventional survey monuments,

installed both at the ground surface and on the foundation-fill interface. A single
automated in-line vibrating wire extensometer under the centerline of the dam is
included to assess the differential settlement across the various zones of the foun-
dation clay. The resilience of the extensometer will be tested under the substantial
future deformations expected for the dam.
The known shear plane in the Upper Clay, around elevation 80m, will be closely
monitored for lateral deformation using both fully automated, fixed in place ‘shape
arrays’ and supplemented by manual inclinometers. The shape arrays and incli-
nometers are located mostly in Area 1 along the downstream weighting berm and
at prescribed distances from the crest, such that the factor of safety of the dam for
global stability can be periodically assessed from the known strain softening behav-
ior of the foundation. The lateral deformation will be a key parameter in future alert
and alarm levels for the dam.
Automated vibrating wire piezometers will continue to monitor the slow dissi-
pation in excess pore pressure within the destructured zone. Under the new fill of
the buttress berm, piezometers will monitor the future consolidation of this material.
Additional piezometers will target the elevation of the expected failure plane. Pore-
water pressures may act as a lagging indicator of destructuring within the foundation,
as Champlain Sea clays are known to generate pore pressures upon shearing and
destructuring.
349
Q.106 - R.21
A key component of the long-term monitoring of Waba Dam will be periodic

measurement of the preconsolidation pressure in the foundation clay downstream of
the destructured zone under the crest. Experience with the site indicates that phys-
ically sampling the sensitive clay while there is excess pore pressure is generally
unreliable. In-situ measurements of the preconsolidation pressure is the preferred
approach, specifically piezocone penetration tests (CPTu). Geophysical testing,
including cross-hole seismic tests and seismic CPTu, have been performed at the
site, however no conclusion on their efficacy for the site can be made at this time.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of other individuals who
have made various contributions to the site interpretation and general analysis. From
Hatch: Warren Hoyle, Breighann Merry, Samantha Taylor, David Parkes, Chris Han-
non, Steve Perkins, and Ralph Serluca, with Prof. K.Y. Lo serving as an external
reviewer on behalf of Hatch. From OPG: Dr. Tareq Salloum and Tony Bennett, with
Dr. Kaare Hoeg, Joseph Ehasz, and Dr. Kerry Rowe serving as external reviewers
on behalf of OPG.
REFERENCES
[1] ONTARIO POWER GENERATION, Madawaska River System-Arnprior GS –

Main Dam & Waba Dam: Dam Technical History & Performance Monitoring
Summary Report, Report No. R-DSP08410.1-27.0303—0025, 2008.
[2] MITCHELL J.K., SOGA K. Fundamentals of Soil Behaviour, Vol. 3., 2005.
[3] LEROUEIL S., TAVENAS F., BRUCY F., LA ROCHELLE P. Behaviour of

destructured natural clays, Journal of Geotechnical and Geoenvironmetnal
Engineering, Vol. 105, 1979.
[4] TAYLOR S., DONNELLY, C.R., TIEDJE, E., MERRY B., HOYLE W., HO K.S.,
PERRA L., SALLOUM T., DALE R.A. The performance of the Waba Dam under
changing foundation conditions, CDA 2019 Annual Conference, 2019.
[5] SALLOUM T., BHARDWAJ V., HASSAN P., CRAGG C. Deformation perfor-
mance of Waba Dam, GeoHalifax, 2009.
[6] LAW K.T. Analysis of embankments on sensitive clays,Doctoral Thesis, The

University of Western Ontario, 1974.
350
Q.106 - R.21
[7] HINCHBERGER S.D., ROWE R.K. Modelling the rate-sensitive characteristics

of the Gloucester Foundation Soil., Canadian Geotechnical Journal Vol. 35,
1998.
[8] HINCHBERGER S.D., QU G., LO K.Y. Constitutive approach for rate-sensitive

anisotropic structured clays, International Journal for Numerical and Analytical
Methods in Geomechanics, Vol 34, 2010.
[9] LAW K.T., REFAHI K., LAM T., KO P., HASSAN P. Seismic Deformation of
Waba Dam, CDA Annual Conference, 2005.
[10] LAW K.T., REFAHI K., CHAN P., LAM T., TANG T., HASSAN P. Instantaneous
Factor of Safety of Waba Dam During Earthquakes, GeoSaskatoon, 2005.
[11] LIU J., SHI C., AFROZ M., KIRSTEIN A. Numerical Investigation of Long-Term
Settlement of Waba Dam, Technical Report, Ryerson University, 2017
[12] GIBSON R.E., LO K.Y. A Theory of Consolidation for Soils Exhibiting Sec-
ondary Compression, Acta Polytechnica Scandinavica, 296, 1961.
[13] PETALAS A.L., KARLSSON M., KARTUNEN M. Modelling of Undrained

Shearing of Soft Natural Clays, 7th International Symposium on Deformation
Characteristics of Geomaterials, 2019
351
Q.106 - R.22
-------
-------
DISTRIBUTED ACOUSTIC SENSING FOR DETECTION OF CHANGES IN

EMBANKMENT DAMS RELATED TO SEEPAGE AND INTERNAL EROSION (∗)
Sam JOHANSSON
HYDRORESEARCH AB, Sweden
Ari DAVID
SILIXA Ltd, United Kingdom
Michael MONDANOS
Anna STORK
Aurélien MORDRET
SISPROBE UGA, ISTerre, France
SWEDEN
SUMMARY
In Sweden, about 100 dams are already equipped with optical fibres for detec-
tion of seepage (based on Distributed Temperature Sensing, DTS) or movement
(based on Distributed Strain Sensing, DSS). With complementary measurements
using Distributed Acoustic Sensing (DAS) it may be possible to detect anomalous
seepage flow in dams without a drainage system, to detect this flow at some dis-
tance from the cable, and furthermore, provide information on the dam integrity.
∗ Mesures acoustiques par fibres optiques pour la détection de changements dans les
barrages en remblai liés à l’infiltration et à l’érosion interne.
352 DOI: 10.1201/9781003211808-117

Q.106 - R.22
The potential capabilities of the technology were therefore tested with two research
projects in two embankment dams in Sweden.
A passive seismic technique, known as ambient noise interferometry (ANI),

is applied to DAS data to image the dams and obtain absolute seismic velocities in
the dam and foundations, providing a snapshot of the structure. An extension of the
method to dam monitoring gives relative temporal changes of the seismic velocities
in the dam with respect to a pre-defined baseline. Temporal changes in seismic
velocities could be a result of material changes in the dam, for example caused by
new seepage locations or ongoing internal erosion.
This feasibility study, combining DAS technology and ANI techniques, shows
significant potential for the method to image and monitor dams and further mea-
surements are required to fully understand the advantages and limitations of the
techniques. Absolute seismic velocities have been determined to image the dam
and highlight areas of inhomogeneity or seepage in a dam. Some areas of seepage
identified through temperature measurements coincide with areas of anomalous
seismic velocities. This assessment of seismic velocities allows baseline measure-
ments to be made for comparison with later monitoring surveys. Monitoring at the
Luossajärvi dam shows areas of the dam with variable seismic velocities correlate
with areas of known seepage. To fully understand the relationship between seepage
and seismic velocity information further studies are required.
RÉSUMÉ
En Suède, une centaine de barrages sont déjà équipés de fibres optiques pour
la détection des infiltrations (sur la base de mesures de température distribuées,
DTS) ou des mouvements (sur la base de mesures de déformations distribuées,
DSS). Avec des mesures complémentaires utilisant la détection acoustique dis-
tribuée (DAS), il peut être possible de détecter un débit d’infiltration anormal dans
les barrages sans système de drainage, de détecter ce débit à une certaine dis-
tance du câble et, en outre, de fournir des informations sur l’intégrité du barrage.
Les capacités potentielles de cette technologie ont donc été testées dans le cadre
de deux projets de recherche dans deux barrages en Suède
Une technique sismique passive, connue sous le nom d’interférométrie de

bruit ambiant (ANI), est appliquée aux données DAS pour imager les barrages et
obtenir des vitesses sismiques absolues dans le barrage et les fondations, four-
nissant un instantané de la structure. Une extension de la méthode à la surveillance
des barrages donne les changements temporels relatifs des vitesses sismiques
dans le barrage par rapport à une ligne de base prédéfinie. Les changements tem-
porels des vitesses sismiques peuvent être le résultat de changements matériels
dans le barrage, par exemple causés par de nouveaux points d’infiltration ou par
une érosion interne continue.
353
Q.106 - R.22
Cette étude de faisabilité, qui combine la technologie DAS et les techniques

ANI, montre que la méthode présente un potentiel important pour l’imagerie et la
surveillance des barrages et que des mesures supplémentaires sont nécessaires
pour comprendre pleinement les avantages et les limites des techniques. Des
vitesses sismiques absolues ont été déterminées pour imager le barrage et mettre
en évidence les zones d’inhomogénéité ou d’infiltration dans un barrage. Certaines
zones de suintement identifiées par les mesures de température coïncident avec des
zones de vitesses sismiques anormales. Cette évaluation des vitesses sismiques
permet d’effectuer des mesures de référence pour les comparer aux relevés de
surveillance ultérieurs. La surveillance du barrage de Luossajärvi montre que les
zones du barrage présentant des vitesses sismiques variables sont en corrélation
avec les zones de suintement connues. Pour bien comprendre la relation entre les
infiltrations et les vitesses sismiques, des études supplémentaires sont nécessaires.
1. INTRODUCTION
The performance of Distributed Acoustic Sensing (DAS) using optical fibers

has increased significantly in recent years. The incidence of acoustic or seismic
waves on a fiber-optic cable results in strain on the fiber, measured using a DAS
interrogator. New developments in the monitoring technology and geophysical anal-
ysis techniques may facilitate studies of seepage/internal erosion phenomena in
embankment and tailings dams and give information on dam integrity.
In Sweden, about 100 dams are already equipped with optical fibers for detec-
tion of seepage (based on Distributed Temperature Sensing, DTS) or movement
(based on Distributed Strain Sensing, DSS). With complementary measurements
using DAS it may be possible to detect anomalous seepage flow also at some dis-
tance from the cable, and furthermore, provide information on the dam integrity.
The potential capabilities of the technology were therefore tested with two research
projects in two embankment dams in Sweden.
Two passive methods can be applied to dams using DAS: 1) seismic monitor-
ing/imaging using correlations of recordings of the ambient seismic noise field and
2) measuring the sound intensity of water flowing through the dam. Here we test
the first method using a technique known as ambient noise interferometry (ANI) to
image a dam and obtain absolute seismic velocities in the dam and foundations as
a fixed snapshot of the structure. An extension of the method to monitoring can pro-
vide relative temporal changes in seismic velocities within the dam with respect to a
pre-defined baseline. Any temporal changes in seismic velocities could be a result
of material changes in the dam, for example caused by ongoing internal erosion.
This study describes tests of DAS-ANI techniques conducted at two dams

in Sweden. We report results from an imaging survey at the Näs Dam, funded by
354
Q.106 - R.22
Energiforsk [1] and a monitoring survey at the Luossajärvi Dam. The tests were
conducted with the objective to:
• Evaluate the potential for using existing fibre-optic installations and DAS tech-
nology to gain further information regarding changes in material properties
and in seepage flow (both flow rate and location).
• Evaluate the capability of a combined DAS-ANI study to image dams and
detect concentrated or anomalous seepage flow in dams.
• Test the capabilities of DAS for general dam monitoring and the detection of
changes in a dam over time, with the ultimate goal to provide an Early Warning
System.
2. TECHNOLOGY AND METHODS
2.1. REQUIREMENTS FOR DAM MONITORING
Internal erosion is the cause of almost 50% of embankment dam failures ([2]
and [3]) and a large body of research has been performed to understand, detect and
prevent this mechanism and consider potential failure modes [4]. Where internal
erosion occurs, there is an expected change in leakage flow, porosity, and water
saturation (due to higher water levels/pressure). To detect damage in dams, any
suitable method must be able to measure deviations in those parameters either
directly or by quantifying a secondary parameter caused by changes in the primary
parameters.
Previous studies have shown that methods to record parameters related to

leakage flow or changes in leakage flow are more sensitive than those that record
material properties/material changes ([5] and [6]. Methods that focus on flow-
dependent parameters such as temperature, self-potential and resistivity (using the
temperature influence) are generally better able to detect changes than material
dependent methods such as seismic methods. Material seismic velocities are sen-
sitive to water content and/or pressure changes and, in the past, active seismic
methods have been tested to image dams and detect anomalous areas. However,
it is difficult to obtain the spatial resolution and repeatability of surveys required to
monitor time-dependent effects such as small leakages using conventional seismic
monitoring equipment.
This study reports the results testing a new seismic approach to dam imaging
and monitoring, applying ambient seismic noise interferometry. This is combined
with DAS technology to provide dense spatial sampling capabilities and repeat-
able measurements with permanently installed fibre-optic cables in dams. Surveys
can be conducted on demand using the same cables for tens of years. Passive
seismic methods, specifically seismic interferometry, removes the need to mobilise
personnel and resources, paving the way for remote data collection. The aim of
355
Q.106 - R.22
the surveys are to determine whether seismic velocities can be determined with
sufficient precision and accuracy to enable the effect of seepage to be resolved
and whether the combination of DAS technology and ANI techniques provides the
desired spatial resolution in measurements.
2.2. DISTRIBUTED ACOUSTIC SENSING (DAS)
DAS technology has developed and improved significantly in the last decade.
The technique measure effectively the speed of sound and seismic waves in soils,
rocks and fluids. Like DTS, the DAS only requires an interrogator unit, a power
source and a fibre optic cable for data collection. DAS systems measure dynamic
strain [7] and measurements can be made on the same fibre-optic cables installed
for temperature sensing. The technology has been deployed for borehole monitor-
ing, ground movement detection, and in industrial environments (oil and gas and
geothermal exploration, e.g. [8], [9] and [10].
A DAS system provides a sensor array that is unmatched in spatial extent,

resolution and sampling by traditional seismic instruments, while allowing the appli-
cation of adapted traditional seismic processing methods (e.g. [11] and [12]. The
propagation of active or passive seismic energy in the ground can be analysed to
determine seismic velocities and assess ground composition. The aim of this study
is to develop this for application to earthen dam assessment and monitoring. The
operating principle for DAS is illustrated in Fig. 1.
Fig. 1
Silixa intelligent Distributed Acoustic Sensing (iDAS) interrogators, iDAS v2 (left)
and iDAS v3 (Carina) (right).
Silixa Interrogateurs acoustiques intelligents pour mesures sur fibres optiques
(iDAS) iDAS v2 (à gauche), version iDAS v3 (Carina) à droite.
356
Q.106 - R.22
A pulse of light is emitted from the DAS interrogator and travels down an optical
fibre. Measurements can be made on fibres up to tens of km long and a single-mode
or multimode cable can be used. A small amount of the light is naturally scattered in
the fibre and returns to the interrogator, measuring the amplitude and phase of the
backscattered light. When an acoustic or seismic wave exerts tiny pressure/strain
changes on the fibre this changes characteristics of the backscattered light and
the relative phase and amplitude can be used to determine the strain rate on the
fibre.
2.3. AMBIENT NOISE INTERFEROMETRY (ANI)
ANI makes use of naturally occurring and anthropogenic background noise to

calculate seismic velocities and hence image the subsurface. By cross-correlating
noise records between two sensors, it is possible to reconstruct the Earth’s response
as if one sensor were a seismic source and the other a receiver. Thus, a seismogram
is produced. Since ambient seismic noise travels through the subsurface the same
way that active source energy does, it is possible to extract some of the same
information that could be gathered from an active survey. The theory of the method
is set out in early papers, e.g. [12] and [13].
The output from ANI are estimates of seismic wave velocities in the
dam/subsurface, with the resolution and depth of investigation dependent on the
type of waves investigated and the available frequency bands. Seismic velocities give
information on the characteristics of material in the dam/subsurface. For the case of
surface waves, lower frequency studies provide information on deeper structures.
Additionally, by studying seismic velocity changes through time, it may be possible
to detect changes in water saturation or structure.
Different parts of the seismic wavefield can be used for ANI, i.e. body waves,
surface waves, and coda waves (Fig. 2). Travelling through the interior of the dam,
body waves (P- and S-waves) are direct arrivals and P-waves arrive first in the seis-
mograms. Surface waves travel slower, tend to be lower in frequency and travel along
the surface of the dam. High frequency surface waves sample only very shallow
depths, with lower frequencies penetrating deeper. Surface waves are dispersive,
meaning that their phase velocity depends on frequency. Additionally, as seismic
waves travel through the subsurface some energy is scattered by inhomogeneities.
These randomly scattered waves that arrive later than body or surface waves are
known as coda waves. The later in the coda we look, the more scattered the waves
are and, for a noise cross-correlation, that means that waves in the coda of the cor-
relation have stayed in the medium between and around the two stations used for
the correlation much longer than the direct waves. Coda waves are therefore more
sensitive to changes in the medium than direct waves and tend to be more stable
with a varying noise sources than body or surface waves [14] a result, coda waves
357
Q.106 - R.22
Fig. 2
Example of ambient seismic noise cross-correlation with the direct P-wave
highlighted in blue, the direct surface-wave highlighted in red and the coda waves
highlighted in green.
Exemple d’une corrélation bruit sismique ambiant avec l’onde P surlignée en bleu,
l’onde de surface en rouge, et les ondes de la coda en vert.
are often used in monitoring or time-lapse surveys to improve repeatability when

using a permanent array of sensors.
Ambient seismic noise is continuous and therefore can be used to continually

monitor the subsurface. Ideally the measured noise field should consist of evenly
distributed noise sources completely surrounding the measurement area. For stable
cross-correlated waveforms, ANI requires noise sources to be unchanging through
time.
2.3.1. Imaging with ANI
Using ANI we can build a model of the lateral variations of seismic wave speed
along the cable and the dam. To take advantage of the density of sensors along the
cable and the associated coherency of the wavefield between neighbouring sensors,
we chose a specific type of imaging technique called eikonal tomography ([15], [16]
and [17]). The results give a measurement of the local seismic velocity between
each sensor.
2.3.2. Continuous monitoring with ANI
Ambient seismic vibrations occur continuously so one can regularly compute

cross-correlations through time, on an hourly basis for instance. Analysing the tem-
poral variations of the waveforms gives access to the temporal changes of the
358
Q.106 - R.22
seismic velocity of the material crossed by the waves. These changes in velocities
can, in turn, provide information on the temporal changes in the material properties
of the dam or the near-surface, such as new seepage or internal erosion.
The different parts of the seismograms can be used to monitor these changes,
and each part has its drawbacks and advantages. Using the direct waves (body- or
surface-waves), when possible, can allow more accurate localisation of the change.
However, direct waves are less sensitive to small changes in the structure because
they spend a small amount of time sampling the dam material. Here Coda-Wave
Interferometry [18] applied because coda-waves are useful to identify and monitor
small changes in the propagation medium. Coda-waves propagate for a long time
within the medium and they sample any potential changes repeatedly, and so are
very sensitive to changes in the dam.
3. SEISMIC IMAGING AT THE NÄS POWER PLANT DAM
3.1. SITE DESCRIPTION
The Näs power plant, in the Dalälven river, is owned by Vattenfall. The plant
has two main embankment dams on either side of the power station and the spillway
and an additional embankment dam was built in 2007, the New Dam. This is the
dam studied here. The dam is founded on natural soil (mainly moraine) and the
bedrock level is approximately 10 m below foundation level. The dam has a central
core of till. The core is surrounded by a filter of fine sand. Gravel and sand are used
for the main support fill. The general flow in the dam has been estimated by earlier
temperature measurements to be about 0.010 l/s and m, with a minor local increase,
about 2-3 times higher in a smaller area close to the right abutment [1].
Fig. 3
Cross section of the dam showing installed cables [1].
Coupe du barrage montrant la position des câbles [1].
The cable installations in the New Dam are designed for research purposes
(initially designed to be used for temperature and strain measurements) and are
359
Q.106 - R.22
more extensive than normal cable installations in embankment dams. Fibre optic
cables are located both in the upstream and downstream filter and also on top of
the core crest (Fig. 3). In this study DAS data is recorded on the crest cable.
3.2. MEASUREMENTS
Passive seismic DAS data was collected continuously for more than 1 month,
between 20 September and 29 October 2019, from the crest cable in the New Dam.
A sampling frequency of 1000 Hz was used with a spatial sampling (channel spacing)
of 1m. The raw data were decimated to 100Hz and then cross-correlated. Every fifth
channel in the fibre was used and the data cross-correlated with the data from all
the other channels selected. The cross-correlated data show clearly propagating
seismic waves along the crest of the dam with arrivals consistent with body waves
between 20 and 45 Hz (Fig. 4a). At lower frequencies (5–20 Hz) waves believed to
be surface waves are visible (Fig. 4b).
Fig. 4
Left: Cross-correlated waveforms along the dam showing body-wave arrivals.
Bandpass filtered 20-45 Hz. Right: Cross-correlated waveforms (5 – 20 Hz
bandpass filtered) showing arrivals believed to be surface waves.
Gauche : Corrélations filtrées entre 20 et 45 Hz montrant des ondes de volume se
propageant le long du barrage. Droite : Correlations filtrées entre 5 et 20 Hz,
montrant la propagation d’ondes de surface.
ANI methods are tested on the DAS data to determine the possibilities of the
technique to establish the location of inhomogeneities in the dam. Inhomogeneities
may be due to differences in composition or compaction in the dam or differences
in the water saturation.
Seismic velocities in the dam are computed using different methods to deter-
mine the most consistent and stable method. The calculation of velocities using
various methods also provides confidence in the results. The imaging tests car-
ried out include the use of direct body-wave, using the cross-correlation method,
360
Q.106 - R.22
and direct surface-wave (using the cross-correlation, the spectral phase and the
cross-wavelet transform).
3.3. RESULTS AND DISCUSSIONS
The velocity of body waves identified in the cross-correlated waveforms varies

across the dam. Two velocity zones are observed, one with velocities 500–650 m/s
in the southern part of the dam and one with lower velocities, 350–500 m/s in the
northern part of the dam (Fig. 5). With a wavelength of 20 m the minimum resolution
is ∼5 m.
Fig. 5
Body-wave velocity along the cable and its surroundings [1].
Vitesse des ondes de volume le long du câble et de ses environs [1].
The difference in velocities observed in the dam could be due to changes in

compaction or differences in water saturation. Without further investigation it is not
possible to determine the actual cause of variation in the dam but the measurements
in this study could form a baseline measurements and any future changes in velocity
would help provide information on physical changes in the dam.
Absolute surface wave velocities also show some variation along the dam with
velocities between 400 m/s and 750 m/s (Fig. 6). These waves are not only sampling
the dam, but also the bedrock beneath it, up to approximately 40m below the dam
crest, and these variations may be due to variations in the dam or in the geology
below the dam. Velocities are estimated using three methods and all have the same
pattern in velocity variation (Fig. 7), thereby providing confidence that this variation is
a real phenomenon. Lower velocities on the right side of the dam near the abutment
coincide with an area of seepage flow, evaluated to be 0.02 litres/second based on
long-term temperature measurements.
361
Q.106 - R.22
Fig. 6
Velocity model of the New Dam calculated with the surface wave [1].
Vitesses des ondes de surface le long du barrage [1].
Fig. 7
Comparison of the surface-wave velocity measurements from three different
methods [1]
Comparaison des mesures de vitesses des ondes de surfaces en utilisant trois
méthodes [1].
4. SEISMIC MONITORING AT THE LUOSSAJÄRVI DAM
4.1. SITE DESCRIPTION
Following the promising results obtained from measurements at the Näs Dam
a larger scale test was performed collecting passive seismic data at the Luossajärvi
Dam, part of the LKAB Kiruna mine site (Fig. 8). The objective of this data collection
362
Q.106 - R.22
is to further assess the potential for the use of DAS data to image and monitor
dams using ANI. This dam was chosen because the longer length of the dam allows
different spatial sampling and a range of frequencies to be tested.
To this end, a passive seismic survey was conducted from 24 June to 7 August
2019 on a newly constructed dam that bisects Lake Luossajärvi. An iDAS was set
up to continuously record ambient noise data at 1000 Hz with a spatial sampling of
1m on three loops of fibre with a total length of 5585m in both the downstream and
upstream parts of the dam (Fig. 8).
Fig. 8
Map view of the Lake Luossajärvi Dam (Google Earth). Satellite image of the site
showing 99 DAS channels (sensor positions) used for the tomography study with
red points.
Carte du barrage du lac Luossajärvi (Google Earth). Image satellite du site
montrant 99 canaux DAS (positions des capteurs, points rouges) utilisés pour
l’étude tomographique.
4.2. METHODS
The fibre array used for this study is composed of three fibre loops installed in
the Luossajärvi Dam. Two are installed at the base of the dam on the downstream
side (the “wetland” and the “core” cables) and one is installed in the drainage sys-
tem (the “drainage” cable). The wetland fibre is sampled every 40 m, the core and
drainage fibres every 15 m with a central section sampled every 5 m (Fig. 8). Vary-
ing sampling distances were tested to determine the density of measurement points
363
Q.106 - R.22
required for different targets. For the monitoring tests (observations of variations in
seismic velocity), data from the core and drainage fibres are used. The raw data are
decimated to 100 Hz and then data from each pair of sensors are cross-correlated.
Monitoring (relative velocity measurements) is performed using data from

74 core fibre channels and 74 drainage fibre channels from the same section of the
dam. Cross-correlation pairs are formed between fibres with channels 15 to 25 m
apart. An average reference cross-correlation is calculated and it is then compared
to a waveform from a given time. Coda-wave interferometry is used to determine any
changes in velocities with time and a frequency band 22–37 Hz is found to provide
good results with a resolution similar to the channel spacing.
Significant changes in seismic velocities can be caused by large rainfall events

inducing changes in groundwater levels or large changes in loading on a dam. To
understand velocity variations in the monitoring results it is necessary to understand
and remove the effect of environmental factors, such as weather and water level,
on seismic velocities. We test this with a model for groundwater level following [19]
using local weather data (rainfall and atmospheric pressure) and assuming a ground
porosity of 15%. Using a Monte Carlo simulation the best equilibrium ground water
level, hydraulic diffusivity and velocity changes between saturated and unsaturated
ground are determined to explain the seismic velocity changes observed.
4.3. RESULTS
Velocity variations relative to the refence seismic velocity are displayed in

Fig. 9. Over the monitoring period there is a general increase in velocities with time
and variation between days. Lateral variations are also observed along the dam.
Fig. 9
Velocity changes in the dam over the monitoring period, relative to the reference.
The red circle indicates the approximate area of sensitivity of the waves used in
the analysis.
Variations de la vitesse du barrage au cours de la période de surveillance, par
rapport á la référence. Le cercle rouge indique la zone approximative de sensibilité
des ondes utilisées dans l’analyse.
364
Q.106 - R.22
Using the modelling method described in Section 4.2.2, the variations in veloc-
ity due to changes in ground water level and atmospheric pressure are estimated.
The median velocity changes in the dam are well correlated with the modelled veloc-
ity changes (Fig. 10). These variations due to environmental effects are removed
from the data to obtain residual velocity changes (Fig. 11). These residual changes
are up to ±1% and highlight anomalous areas in the dam for further investigation.
Fig. 10
Spatially-averaged seismic velocity changes (blue), with uncertainties, compared
with synthetic models of seismic velocity changes based on groundwater level
changes alone (yellow curve) and groundwater level plus atmospheric pressure
variations (red curve).
Variations de la vitesse sismique moyennées spatialement (en bleu), avec les
incertitudes, par rapport aux modèles synthétiques de variations de la vitesse
sismique basés sur les seules variations du niveau des eaux souterraines (courbe
jaune) et du niveau des eaux souterraines plus les variations de la pression
atmosphérique (courbe rouge).
Fig. 11
Velocity changes residuals after correction from the environmental variations.
Résidus des changements de Vitesse après corrections des variations
environnementales.
365
Q.106 - R.22
DISCUSSION
For dam imaging and monitoring purposes, it is not necessarily obvious how
leakage in a dam would affect seismic velocity measurements. An increase in water
content and removal of material from within the dam could be expected to decrease
velocities but if there is an increase in pressure in parts of the dam, this may lead to
the opposite effect. These competing effects may lead to a high variance in seismic
velocities.
Both the Näs Dam and the Luossajärvi dam are monitored for seepage using
DTS technology and areas of seepage have been identified. Areas of low seismic
velocities within the dam and or the foundation close to the right abutment at the
Näs Dam correlate with an area of seepage identified in the DTS data.
To test whether regions of high seismic velocity variance are correlated with
seepage, the monitoring results from Luossajärvi dam with environmental effects
removed are compared to seepage data derived from DTS monitoring. A comparison
with the seepage flow data inferred from the temperature data at the dam shows good
correlation between the magnitude of temporal fluctuations in velocity measured
from DAS and the seepage values from DTS. The seismic velocity changes correlate
with both the core and drainage seepage data (Fig. 12). In particular, a high variance
in seismic velocity was found at 1350m at the Loussajärvi dam. This coincides with
an area of known leakage in the drainage cable.
Fig. 12
Comparison between the seepage data inferred from the fibre temperature
measurements (raw data as fine curves and smoothed data as thick curves) in the
core (orange) and in the drainage (blue), and the standard deviation of the seismic
velocity changes measured along the calendar time (black curve).
Comparaison entre les données d’infiltration déduites des mesures de
température des fibres (données brutes sous forme de courbes fines et données
lissées sous forme de courbes épaisses) dans le cœur (orange) et dans le
drainage (bleu), et l’écart-type des résidus des variations de vitesse sismique
mesurées au cours du temps calendaire (courbe noire).
366
Q.106 - R.22
The results from these initial studies show promise for the combination of
DAS technology and ANI techniques to highlight anomalous areas in dams and
their surrounding materials. The measurement technique provides complementary
information to DTS because DTS provides accurate spatial measurements at the
cable, whereas DAS provides the opportunity to detect anomalies throughout the
dam but likely with a lower spatial resolution.
Further studies over longer timeframes and testing different spatial resolutions
are required to determine the limits of high frequency imaging with DAS and to
confirm or disprove environmental effects on seismic velocities.
CONCLUSIONS
In conclusion, a feasibility study combining DAS technology and ANI tech-

niques to identify anomalies or areas of seepage/leakage in earthen dams has been
conducted using data from two dams. The results show significant potential for the
method and further measurements are required to fully understand the advantages
and limitations of the techniques.
Absolute seismic velocities have been determined from body waves, surface
waves and coda waves to highlight areas of inhomogeneities or seepage in a dam
and some areas of seepage identified through temperature measurements coincide
with areas of anomalous seismic velocities. This assessment of seismic velocities
allows baseline measurements to be made for comparison with later monitoring
surveys.
To enable the use of the technology as a monitoring tool, external effects,

such as weather effects must be removed from the data. A method to model the
effects of rainfall and atmospheric pressure on seismic velocities has been devel-
oped and shows good correlation with the recorded data. Following removal of these
effects from the data collected at the Luossajärvi dam, areas of the dam with variable
velocities correlate with areas of known seepage. To fully understand the relationship
between seepage and seismic velocity information further studies are required, par-
ticularly where seepage or defect information is known, and longer-period studies
are needed to fully resolve environmental effects.
ACKNOWLEDGEMENTS
We would like to express our gratitude to Swedish Energy Agency for funding
the initial tests at Näs dam. We are also grateful for the dam owners, Vattenfall
Vattenkraft AB and LKAB for support and allowing us to publish these results.
367
Q.106 - R.22
REFERENCES
[1] Johansson, S., Stork, A., David A., Mondanos, M. & Nygren, C. Fibre-Optic
Distributed Acoustic Sensing for Detection of Seepage and Internal Erosion,
Energiforsk AB, REPORT 2020:682, ISBN 978-91-7673-682-1
[2] Foster M., Fell R. & Spannagle M. (2000) The statistics of embankment dam
failures and accidents. Canadian Geotechnical Journal 37: 1000–1024.
[3] Fell R., MacGregor P., Stapledon D. & Bell G. (2005) Geotechnical engineering
of embankment dams, A.A. Balkema, ISBN: 04-1536-440-x, 912p.
[4] ICOLD (2017), Dam safety management: Operational phase of the dam life
cycle, Bulletin No. 154, International Commission on large dams, Paris.
[5] Johansson S. (1997) Seepage Monitoring in Embankment Dams, Doctoral

Thesis, TRITA-AMI PHD 1014, ISBN 91-7170-792-1, Royal Institute of
Technology, Stockholm.
[6] Sjödahl, P., Dahlin, T., Johansson, S. (2019). Geofysiska metoder inom damm-
säkerhetsområdet. Energiforsk rapport 2019:571 ISBN 978-91-7673-571-8.
[7] Parker, T., Shatalin, S., & Farhadiroushan, M. (2014). Distributed Acoustic
Sensing – a new tool for seismic applications. First Break, 32(2010). doi:
10.3997/1365-2397.2013034.
[8] Jousset, P., Reinsch, T., Ryberg, T., Blanck, H., Clarke, A., Aghayev, R.,
Krawczyk, C.M. (2018). Dynamic strain determination using fibre-optic cables
allows imaging of seismological and structural features. Nature Communica-
tions, 9(1).
[9] Mondanos, M. and T. Coleman (2019) Application of distributed fibre-optic

sensing to geothermal, First Break, 37(7), 51–56.
[10] Harris, K., D. White, D. Melanson, C. Samson and T.M. Daley. (2016). Feasi-
bility of time-lapse VSP monitoring at the Aquistore CO2 storage site using a
distributed acoustic sensing system, International Journal of Greenhouse Gas
Control, 50, 248–260.
[11] Correa, J., Pevzner, R., Bona, A., Tertyshnikov, K., Freifeld, B., Robertson,
M. and Daley, T. (2019). 3D vertical seismic profile acquired with distributed
acoustic sensing on tubing installation: A case study from the CO2CRC Otway
Project, Interpretation, 7, SA11–SA19.
[12] Campillo, M. (2006). Phase and correlation in ’random’ seismic fields and the
reconstruction of the green function. Pure and applied geophysics, 163(2–3),
475–502.
368
Q.106 - R.22
[13] Wapenaar, K., & Fokkema, J. (2006). Green’s function representations for
seismic interferometry. Geophysics, 71(4), SI33–SI46.
[14] Colombi, A., Chaput, J., Brenguier, F., Hillers, G., Roux, P., & Campillo, M.
(2014). On the temporal stability of the coda of ambient noise correlations.
Comptes Rendus Geoscience, 346(11-12), 307–316.
[15] Lin, F.-C., Ritzwoller, M. H., & Snieder, R., 2009. Eikonal tomography: sur-
face wave tomography by phase front tracking across a regional broad-band
seismic array, Geophysical Journal International, 177(3), 1091–1110.
[16] Lin, F.-C., Li, D., Clayton, R. W., & Hollis, D., 2013. High-resolution 3D shal-
low crustal structure in Long Beach, California: Application of ambient noise
tomography on a dense seismic array, Geophysics, 78(4), Q45–Q56.
[17] Mordret, A., Landès, M., Shapiro, N., Singh, S., Roux, P., & Barkved, O.,
(2013). Near-surface study at the Valhall oil field from ambient noise surface
wave tomography, Geophysical Journal International, 193(3), 1627–1643.
[18] Snieder, R., Grêt, A., Douma, H., & Scales, J., (2002). Coda wave interferom-
etry for estimating nonlinear behavior in seismic velocity, Science, 295(5563),
2253–2255.
[19] Sens-Schönfelder, C. & Wegler, U. (2006). Passive image interferometry

and seasonal variations of seismic velocities at Merapi volcano, Indonesia,
Geophys. Res. Lett., 33(21).
369
Q.106 - R.23
-------
-------
OBSERVATIONS FROM INITIAL IMPOUNDMENT OF AN EXPERIMENTAL

EMBANKMENT DAM – FIELD DATA AND MODELLING (∗)
Jasmina TOROMANOVIC
LULEÅ UNIVERSITY OF TECHNOLOGY
Jan LAUE
Hans MATTSSON
Sven KNUTSSON
Peter VIKLANDER
VATTENFALL AB, VATTENFALL R&D, LULEÅ UNIVERSITY OF TECHNOLOGY
Christian BERNSTONE
VATTENFALL AB, VATTENFALL R&D
SWEDEN
SUMMARY
In 2019, an experimental embankment dam was built as a collaboration

between Sweden’s largest dam owner Vattenfall and several groups of researchers.
The purpose was to study the mechanical behaviour of the dam over time and to
detect built-in damages, which are related to ageing effects in real dams. Research
∗ Observations de la mise en place initiale d’un barrage en remblai expérimental – données
de terrain et modélisation
370 DOI: 10.1201/9781003211808-118

Q.106 - R.23
connected to the experimental dam at Luleå University of Technology is focusing on

mechanical behaviour. In this paper, the pore pressure and deformation response
in the dam is reviewed during impoundment, both from a physical and numerical
point of view. Predictions are performed of the pore pressure and deformations for
further operation until June 2021, which is the expected end of the experimental
dam project.
In field, the pore pressure is observed by 12 vibrating wire piezometers placed

in the filter zones and in the core of the dam. These transducers allow for measure-
ment of both negative and positive pore pressure. The deformations are monitored
by three shape accelerometer arrays, which are placed in the upstream, middle and
downstream part of the dam. All of these measurements are continuously logged
and presented online, providing access to the behaviour of the dam in real-time.
The numerical modelling is performed in the finite-element software PLAXIS

2D. A fully-coupled analysis is done in order to assess the pore pressure develop-
ment and deformations caused by the impoundment. This type of analysis allows
for development of flow and deformations in the same computational phase. The
majority of the values of parameters used in modelling are from the design phase
of the dam, with some preliminary laboratory results incorporated.
From the field observations, the pore pressure development during impound-
ment has been followed. There has been a gradual development of the pore
pressure, with quick increase in the coarser parts of the dam and slower increase
in the core of the dam. In order to better understand the saturation process of the
core, the initial saturation ratio from construction of the dam was taken into con-
sideration when performing the numerical modelling. From laboratory testing, a soil
water characteristic curve was available, thus providing the opportunity to model
the relation between saturation and suction. As the water in the dam was filled and
lowered, the loading-unloading response of the dam could be observed in terms of
deformations. Not all deformations were recovered when lowering the level, noting
that the elastic region is exceeded. This is also captured by the numerical modelling.
In on-going work at Luleå University of Technology, the material from the

experimental dam is being extensively characterized by laboratory testing. Future
numerical modelling will have a better base and the data will be corrected for poten-
tial affecting factors. Pore pressures, deformations and strains will be looked into in
future studies, both from a physical and numerical point of view.
RÉSUMÉ
En 2019, un barrage expérimental en remblai a été construit dans le cadre

d’une collaboration entre le plus grand exploitant de barrage suédois Vattenfall
371
Q.106 - R.23
et plusieurs groupes de recherche. L’objectif était d’étudier le comportement

mécanique des barrages dans le temps et de détecter les dommages causés durant
la construction, qui sont liés aux effets du vieillissement dans les barrages réels. Les
recherches de l’Université de Technologie de Luleå se concentrent sur le comporte-
ment mécanique du barrage expérimental. Dans cet article, la pression interstitielle
et la réponse à la déformation dans le barrage expérimental sont examinées pendant
la mise en eau, à la fois d’un point de vue physique et numérique. Des prévisions
sont effectuées sur la pression interstitielle et les déformations pour une exploitation
ultérieure jusqu’en juin 2021, qui est la fin prévue du projet de barrage expérimental.
Sur le terrain, les pressions interstitielles sont mesurées par 12 piézomètres

à fil vibrant placés dans les zones filtrantes et au cœur du barrage. Ces transduc-
teurs permettent de mesurer à la fois la pression interstitielle négative et positive.
Les déformations sont surveillées par trois réseaux d’accéléromètres de forme,
situés dans la partie amont, milieu et aval du barrage. Toutes ces mesures sont
enregistrées en continu et présentées en ligne, donnant accès au comportement
du barrage en temps réel.
La modélisation numérique est réalisée dans le logiciel d’éléments finis

PLAXIS 2D. Une analyse entièrement couplée est effectuée afin d’évaluer le
développement de la pression interstitielle et les déformations causées par la
retenue. Ce type d’analyse permet le développement de simulations d’écoulement
et de déformations dans la même phase de calcul. La majorité des valeurs des
paramètres utilisés dans la modélisation proviennent de la phase de conception du
barrage, avec quelques résultats préliminaires de laboratoire incorporés.
À partir des observations sur le terrain, le développement de la pression inter-

stitielle pendant la mise en eau a été suivi. Il y a eu un développement progressif de
la pression interstitielle, avec une augmentation rapide des parties plus grossières
du barrage et une augmentation plus lente du cœur du barrage. Afin de mieux com-
prendre le processus de saturation du cœur, le taux de saturation initial durant de la
construction du barrage a été pris en compte lors de la réalisation de la modélisation
numérique. À partir d’essais en laboratoire, une courbe caractéristique de la teneur
en eau du sol était disponible, offrant ainsi l’opportunité de modéliser la relation
entre la saturation et la succion. Au fur et à mesure du remplissage et de la vidange
du barrage, la réponse chargement-déchargement du barrage a pu être observée
en termes de déformations. Toutes les déformations ne sont pas résorbées lors de
l’abaissement du niveau, notant que les limites élastiques étaient dépassées.
Dans le cadre des travaux en cours à l’Université de Technologie de Luleå

, les matériaux du barrage expérimental sont extensivement caractérisés par des
tests en laboratoire. La future modélisation numérique aura une meilleure base et
les données seront corrigées pour les facteurs ayant potentiellement un effet. La
pression interstitielle, les déformations et les contraintes seront examinées dans de
futures études, à la fois d’un point de vue physique et numérique.
372
Q.106 - R.23
1. INTRODUCTION
In 2019 an experimental embankment dam was built as a collaboration

between Sweden’s largest dam owner Vattenfall AB and research groups from Luleå
University of Technology (LTU), Uppsala University, Lund University of Technology
and HydroResearch AB in order to study the behaviour of the dam. There are
two main purposes of the experimental dam research. One part is to study the
mechanical behaviour of the embankment dam during first impoundment as well as
throughout the continued operation. Another part is to localise by geophysical meth-
ods several embedded damages that were built into the dam, see [1] for an overview
of the detection methods. Monitoring by geotechnical equipment will be performed
before and during reservoir filling as well as during the operational phase. Vattenfall
AB is funding the construction if the experimental dam at the sites of its research
and development laboratory facilities in Älvkarleby, Sweden.
In this paper, the field data consisting of pore pressure and deformations from
the initial impoundment of the dam is reviewed. To understand the saturation pro-
cess and the flow net development, the interpretation of the field data is supported
by numerical modelling. In the modelling, unsaturated conditions are taken into con-
sideration as well as an interval for the values of the hydraulic conductivity of the
central core in the dam.
2. THE EMBANKMENT DAM
To be able to construct the experimental dam on dry land and have the ability
for a controlled operation, a surrounding supportive structure of concrete was built,
see Fig. 1. The dimensions are 20 m x 16 m, with the foundation level 4 m under the
Fig. 1
The experimental dam inside the supportive structure.
Le barrage expérimental à l’intérieur de la structure de soutien.
373
Q.106 - R.23
ground surface. In order to achieve a controlled environment, a tent was raised over
the supportive structure, in order to protect from precipitation. The closely located
river Dalälven is the water supply.
The constructed embankment dam is four metres high. The dam is zoned,
consisting of a central glacial till core which is surrounded by fine- and coarse filters,
see Fig. 2. Filter zones are typically consisting of sand and/or gravel, where their
task is to protect the core i.e. to prevent internal erosion. Outside the filter zones, the
shoulder material is placed. This is the coarsest part of the dam that is quite large in
volume and is providing stability to the structure. The embankment dam is designed
according to the Swedish dam safety guidelines, RIDAS [2]. The design is based
on the core material and other fractions are matching in order to prevent internal
transport. For details on the particle size distribution, see [3]. It took four weeks to
complete the construction of the embankment dam. The dam was constructed in
20 cm thick horizontal layers. A detailed description of the construction work can be
found in [3].
Fig. 2
Cross-section of the embankment dam.
Coupe transversale du barrage en remblai.
The geotechnical instrumentation in the dam consists of deformation, pore

pressure and strain monitoring devices, where all measurements are performed
continuously. These sensors are concentrated to one section that is set aside for
the purposes of the geotechnical investigation, as exemplified by the cross-section in
Fig. 2. To measure pore pressure, vibrating wire piezometers were placed in the core
and in the filters of the dam. The piezometers were installed during construction of
the dam, where they were placed directly on a compacted layer. The pore pressure
transducers are of the type Geosense VWP-3400, with a measuring range from
−70 kPa to 345 kPa. The vibrating wire within the sensor is in a closed system, which
374
Q.106 - R.23
will react to the pressure changes outside the transducer. This also means that
the measurements are affected by the atmospheric pressure which needs to be
corrected for.
The deformations in the dam are monitored by a type of inclinometers, shape

accelerometer arrays (SAA) of type Measurand SAAV250. The system allows for
deformation monitoring in x-, y- and z-directions. There is one sensor in each seg-
ment of 0.25 m, in order to fit the scale of the experimental dam. Three inclinometers
have been installed, with position and length as given in Fig. 2; 3.5 m upstream,
4.0 m in the centre and 3.0 m downstream. The x-direction is from upstream to
downstream. The y-direction is along with the dam axis and z-direction vertical.
When installing the SAA’s, they were guided down into plastic tubes that were built
in during construction. Continuous measurements of the deformations are on-going.
The data is available online, providing the possibility to follow the development in
real-time. Strain measurements with fibre optics are performed in the bottom of the
embankment dam. These measurements are not evaluated in this paper, but will be
analysed in our continued research.
The water level in the experimental dam is continuously logged on the

upstream side. In Fig. 3 the measured levels during the impoundment are shown.
The impoundment was started the 16th March 2020. When the water level reached
2 metres, the impoundment was interrupted due to a need to improve the seep-
age water collecting downstream, where after the impoundment was continued.
Over 39 days, between the 8th May 2020 and the 6th June 2020, the dam was
impounded to the maximum operation level at almost 3.5 metres (this corresponds
to the stair-looking part of the graph). Thereafter, the water level was slightly lowered
to 3.2 metres and kept constant.
Fig. 3
Water levels in the reservoir.
Niveaux d’eau dans le réservoir.
375
Q.106 - R.23
3. NUMERICAL MODELLING
3.1. CONSTITUTIVE MODELLING
The soil behaviour of the dam body was found to be reasonably well described
by the Hardening soil model, see [4]-[7]. The Hardening soil model is an elasto-
plastic strain hardening constitutive model, which takes plasticity into account. The
failure criterion of the Hardening soil model is equal to the one of the Mohr-Coulomb
model. The values of the parameters included in the numerical model are based on
literature studies and previous research, [3] and [8], as the laboratory characteri-
zation of the material from the experimental dam is ongoing and not yet finalized.
The values of the parameters used for modelling the constitutive behaviour of the
dam body are presented in Table 1. A linear elastic description is chosen for the
foundation. A simple model is considered sufficient since the foundation is stiff. The
values are found in Table 2.
Table 1
Values for the constitutive model Hardening soil; dam body.
Parameter Core Fine filter Coarse filter Shoulder
Unsaturated unit weight γunsat (kN/m3 ) 21 19 17 18
Saturated unit weight γsat (kN/m3 ) 23 21 19 20
Cohesion c(kPa) 1 0 0 0
Friction angle ϕ( ◦ ) 42 36 40 41
Dilatancy angle ψ( ◦ ) 0 6 10 11
Moduli E50ref = E ref = 1 · E ref (MPa) 70 50 50 20
oed 3 ur
Power for stress level dependency m( − ) 0.7 0.5 0.5 0.5
Table 2
Values for the linear elastic constitutive model; foundation.
Unit weight Young’s modulus Poisson’s ratio
γ(kN/m3 ) E(GPa) ν( − )
27 10 0.2
3.2. HYDRAULIC MODELLING
The core of the dam is modelled to allow for unsaturated behaviour, and will
allow for different values of the hydraulic conductivity as the core material saturates.
The till material from the core was tested in order to obtain values of parameters for a
soil water retention curve. Values for the van Genuchten hydraulic model [7] are pre-
sented in Table 3, these values should be considered as preliminary evaluations of
laboratory results. Different hydraulic conductivities are considered in the modelling
of the core, as laboratory results are not yet available. Cases A-E are presented in
Table 3, with values corresponding to a fully saturated soil. For the rest of the dam
376
Q.106 - R.23
Table 3
Values for the van Genuchten hydraulic
model.
Model parameter Value
Residual saturation 0.5
Saturated conditions 100
Fitting parameter related to the air entry 1.55
Fitting parameter related to water extraction 0.2
Table 4
Values of the hydraulic conductivity for the
modelling cases; core.
Case Hydraulic conductivity [m/s]
A 10−9
B 10−8
C 2 · 10−8
D 5 · 10−8
E 10−7
Table 5
Values of the hydraulic conductivity; fine
filter, coarse filter and shoulder.
Zone Hydraulic conductivity [m/s]
Fine filter 10−5
Coarse filter 10−2
Shoulder 10−1
body, values of the hydraulic conductivity are found in Table 4, with values that are
from the design stage of the embankment dam. The same value of the hydraulic
conductivity in horizontal and vertical directions is assumed for the material of each
dam zone.
3.3. FINITE-ELEMENT MODELLING
Since the analysed section of the dam is at such a large distance from
the supportive concrete structure, seen in Fig. 1, the effects of the wall can be
excluded. Then, it is reasonable to assume plane-strain conditions and conduct
two-dimensional modelling. The used finite-element software is PLAXIS 2D 2019
[7]. The core is expected to show undrained behaviour, as the glacial till has a low
hydraulic conductivity. The undrained behaviour of the core is modelled by using
effective parameters, option “Undrained A” in PLAXIS. All other dam zones are
set to represent drained behaviour. The foundation is modelled as non-porous, not
allowing any pore pressure response. The geometry of the dam in the finite element
program is set according to the cross-section, presented in Fig. 2. This finite ele-
ment model, presented in Fig. 4, is a further development of a model created in a
377
Q.106 - R.23
Fig. 4
Finite element model of the experimental dam.
Modèle par éléments finis du barrage expérimental.
student project at LTU, see [9]. Following the real construction sequence of the dam,
modelling is performed in several steps, as:
• Construction: The dam is constructed in horizontal layers. The horizontal

lines in Fig. 4 are representing the layers built. Approximately one layer is built
every day. The water content of the core from field measurements is applied.
Development and dissipation of pore pressure are allowed.
• Rest: After the construction work was finished in November 2019, it took
until March before the impoundment could be started. This “resting phase” is
included. The computation phase allows for the saturation to develop, as there
is some water from the construction in the core.
• Impoundment and constant level:
The water levels presented in Fig. 3 are implemented. During these compu-
tations, the pore pressures and deformations are allowed to simultaneously
develop, running a fully-coupled flow-deformation analysis.
4. OBSERVATIONS FROM IMPOUNDMENT – RESULTS AND ANALYSIS
4.1. SATURATION PROCESS OF THE CORE
The gradual saturation of the core is captured by the numerical modelling, the
results are presented in Fig.5. Different values of the permeability of the core have
been modelled in Case A to E, see Table 5. Notice, for Cases A to E in Fig. 5,
the values of the permeability are increasing downwards. The number of days of
378
Q.106 - R.23
impoundment that have been analysed, are presented as “dots” in Fig. 3, except
445 days which corresponds to June 1st 2021.
With increasing permeability of the core, the saturation process becomes

quicker, as could be expected. As suction is allowed in the numerical model, the
till in the core is partially saturated in zones above the phreatic line. The degree of
saturation at the very beginning of the impoundment in March is presented in the
leftmost part of Fig. 5. In November, at the end of construction, the degree of satura-
tion in the core was determined to 82% (based on field data) and this is assigned to
the numerical model for the construction phase. In March, according to the numeri-
cal model, the degree of saturation has decreased in the upper part and the middle
part of the core. This is also supported by the field observation of seepage water
leaving the dam structure before impoundment.
The pore pressure in all cases reaches steady-state before June 1st 2021.
The position of the theoretical phreatic line in the dam body will differ depending on
how permeable the core is. When comparing Cases A and E, a distinct difference
can be noted in the position of the phreatic lines in the central part of the dam body
after 445 days in Fig. 5. Case E will lead to more water in the core, compared to
Case A, see Fig. 5.
Fig. 5
Development of the degree of saturation.
Développement du degré de saturation.
379
Q.106 - R.23
4.2. PORE PRESSURE
The pore pressure response in the dam has been logged since the end of
November 2019. The portion of data that is studied in this paper covers the start
of impoundment at March 13th 2020 until September 2nd 2020. All of these mea-
surements need to be corrected for the variations in the atmospheric pressure, this
effect has been assessed to variations of +/−1.5kPa. Values of the pore pressures,
from field and modelling, are presented in Figs. 6-8. In Figs. 6-8, field observations
are presented as “PX” and the associated results from the numerical modelling as
“PXM”, where the number of the pore pressure transducer is “X” and is presented
in Fig. 2. In the numerical modelling, predictions until June 1st 2021 are also per-
formed. The pore pressure response from the start of impoundment between 13th
March and 5th April (0–20 days) can be summarized as follows:
• Increased pressure is observed in readings from P5 and P6, the pressure

increase is following the upstream water level almost immediately as these
transducers are measuring the response in coarse material.
• The bottom of the core, meaning transducers P4 and P8 as well as P3 and P7
are reacting with increased pressure. P3, P4 and P5 reacted first, immediately
after the impoundment started.
• The most downstream transducers P1 and P2 are not showing any reactions,
same as for transducers P11 and P12 in the top of the core.
• Positive pore pressures have been measured in some of the transducers. The
pore pressures should not be positive, when there is no water in the in the
reservoir. Either the pressures should be zero or below zero, meaning that it is
only suction that should be occurring which would be shown as negative pore
pressures. The pore pressures measured in the transducers at the start of the
impoundment are probably not correct.
The second try to impound between 8th May and 13th June (53–89 days) can
be summarized as follows:
• P4, P5 and P6 are reacting immediately. P7 and P8 are showing a delayed

response, P11 and P12 are showing very little response, although reacting.
• P3 is deviating from the expected behaviour and shows similar pressure levels
as the readings from the most downstream transducers P1 and P2, meaning
a very low pressure. This could be interpreted as that the saturation has not
yet being finished and that the pressures have not reached steady-state condi-
tions. Alternatively, there is a simpler path for the water to take through the dam
structure than the theoretical flow path retrieved from seepage computations.
This could be the reason for the very low pore pressures observed.
In Figs. 6-8, all dashed lines are representing the modelled results with the
hydraulic conductivity of the core according to Case C. When comparing the pore
380
Q.106 - R.23
pressure from field with the corresponding results of the numerical modelling, the
following can be noted:
• The trend of the pore pressure development is captured by the numerical

modelling. However, negative pore pressures are higher than in reality.
• At all times the pore pressures are underestimated in the model. This means
that the phreatic line in the dam is not exactly following the position according
to the theoretical one. However, if the measured pore pressures are adjusted
with the value of the pore pressures measured directly after installation there
is a far better agreement with the results of the numerical model.
• All peak values of pore pressure, corresponding to the pressures measured
at the highest water level, are even more underestimated by the model than
the rest of the values.
• Regarding time, there seems to be a delay in response of the pore pressure
in the numerical data presented, compared to the field measurements. The
delay is less in Case E, where the hydraulic conductivity is higher and thus
allowing a quicker development of the pore pressures.
• There is no significant pore pressure change in the results of the numerical
model after September 2020 to June 2021.
In Fig. 6 the data from the pore pressure readings for the transducers P1, P2,
P5 and P6 are presented. These are the transducers located in the filter zones.
P1 and P2 are theoretically in suction all the time, but not according to the mea-
surements. Although, if the very first readings in November are considered as a
Fig. 6
Pore pressure development for P1-P2 and P5-P6, field and modelling.
Développement de la pression interstitielle pour P1-P2 et P5-P6, terrain et
modélisation.
381
Q.106 - R.23
zero readings and the measured pore pressures are adjusted by subtracting the
associated zero reading, the value of pore pressures from field will be under zero
thus having suction also. Pore pressures in P5 and P6 in the model are developing
according to the field, but here also the values of the pore pressures are underesti-
mated. On the long term, the values of the pore pressures upstream in the filter are
well predicted. However, these are positive pressures. It is more difficult to predict
pore pressures when suction is occurring. If wanting to capture the suction behaviour
from field, the unsaturated properties need to be given more attention in the future
modelling in this project.
Fig. 7
Pore pressure development for P3-P7-P9-P11, field and modelling.
Développement de la pression interstitielle pour P3-P7-P9-P11, terrain et
modélisation.
In Fig. 7 the pore pressures are presented for the transducers P3, P4, P5 and
P6 that are located in the downstream part of the core. The measurements from
P3 are not following any trend similar to the associated results from the modelling.
The pore pressures in P7 are underestimated by almost 20 kPa in the model, com-
pared to field measurements, when the pore pressures have stabilised. According
to the theoretical phreatic line, the top transducers P11 and P9 should always be
measuring negative pore pressure. P9 is showing positive pore pressure. Even if
the values are shifted according to what is initially measured, in November, the pore
pressure will still be positive. P11 shows negative values, this is interpreted as the
phreatic line drops somewhere between P9 and P11 in reality. This is somewhat
better captured numerically in Cases D-E, which have higher values of the hydraulic
conductivity.
382
Q.106 - R.23
Fig. 8
Pore pressure development for P4-P8-P10-P12, field and modelling.
Développement de la pression interstitielle pour P4-P8-P10-P12, terrain et
modélisation.
In Fig. 8, the response of pore pressures from transducers P4, P8, P10 and
P12 is presented along with corresponding values from the numerical modelling.
P12 shows suction both in field and from the modelling. P4, P8 and P10 are all
showing the same trend in pore pressure during impoundment as the modelling, but
with differences in magnitudes.
4.3. DEFORMATIONS
In this study, the horizontal deformations in the dam, caused by the impound-
ment have been preliminarily assessed, in the positions of the inclinometers from
Fig. 2. The dam moved downstream, during first part of impoundment and there-
after in the upstream direction as a response to the subsequent lowering of the water
level – recovering some deformations and showing the loading-unloading behaviour.
Some deformations were permanent, showing that the elastic range is exceeded
during the first impoundment stage. This is observed both by the field data and the
modelling. In Fig. 9 the horizontal deformations at the end of second stage of the
impoundment (89 days in Fig. 3) are shown for the inclinometers in the upstream
(left), central (middle) and downstream (right) positions. The solid line represents the
field measurements and the dotted represents the corresponding modelling results.
The zero level represents the foundation level. In all graphs, the spacing between
the lines on the horizontal axes correspond to 5 mm.
383
Q.106 - R.23
From the modelling, it is noted that the main deformations for the shoulder
zones, both upstream and downstream, are settlements with approximately the
same inclination as the slopes. Settlement in these parts would then cause the
deformations in the horizontal directions outward from the dam. The magnitude of
the deformations in the dam is underestimated by the numerical modelling. This is
related to the choice of the stiffness parameter, which are only taken from literature.
The bottoms of the inclinometer are not fixed to the concrete structure and
could therefore have moved without the readings showing it. This means that the
deformations from field in Fig. 9 are relative for the impoundment phase.
Fig. 9
Horizontal deformations caused by the impoundment.
Déformations horizontales causées par le bassin.
The deformations from the long-term computation are not presented, as these
are almost identical to the ones after 89 days of impoundment.
In this study, the pore pressures and the deformations from the initial impound-
ment of an experimental embankment dam have been assessed. The measured
pore pressure and deformations are compared to the associated results from
finite-element modelling of the impoundment of the dam.
384
Q.106 - R.23
Pore pressure from field is showing a gradual pore pressure development,

with what seems by the initially processed data to have reached stable conditions.
In general, the transducers react in a logical order that is well corresponding to the
changes of the water level. Some transducers in the top of the core and in the down-
stream filters are showing suction (negative pore pressure). Under the assumption
the there is no positive pore pressure in the dam directly after finished construction,
the curves representing the pore pressure can be shifted with the initially measured
value and will correspond better to a theoretical steady-state phreatic line. If not
shifted, the field measurements indicate that there is more water in the core than
predicted by theory.
The numerical model is capturing the saturation process in the core. As

expected, the rates of saturation are higher with increasing values of the hydraulic
permeability of the core. For the filter zones, consisting of more coarse material,
the pore pressures are quite well predicted. The pore pressures in the core during
impoundment are the most difficult to predict. The material of the core is more
complex to describe, since it includes an unsaturated aspect. Even though the
data representing the soil water characteristic curve is preliminarily evaluated, the
saturation process of the core and the flow net development of the core is captured.
From both field and numerical modelling of deformations, the loading-

unloading behaviour is observed. When lowering the water level, some deformations
are recovered. The dam moves towards downstream during first part of impound-
ment and thereafter upstream again when the water level has been lowered. From
the modelling, in general, the deformations are underestimated.
In on-going work, the material from the zones of the experimental dam is being
extensively characterized. Future numerical modelling will include fewer uncertain-
ties, compared to what has been the case up to now. The geotechnical field data
will be corrected for potential affecting factors, e.g. the pore pressure measurement
will be corrected for the variations in the atmospheric pressure. Pore pressures,
deformations and strains will be looked into in future studies, both from a physical
and numerical point of view.
ACKNOWLEDGEMENTS
The research presented in this paper was carried out as a part of ‘Swedish
Hydropower Centre – SVC’. SVC was established by the Swedish Energy Agency,
Energiforsk and Svenska kraftnät together with Luleå University of Technology,
KTH Royal Institute of Technology, Chalmers University of Technology and Uppsala
University. Participating companies and industry associations are: Andritz Hydro,
Boliden, Fortum Generation, Holmen Energi, Jämtkraft, Karlstads Energi, LKAB,
Mälarenergi, Norconsult, Rainpower, Skellefteå Kraft, Sollefteåforsens, Statkraft
Sverige, Sweco Energuide, Sweco Infrastructure, Tekniska verken i Linköping,
Uniper, Vattenfall R&D, Vattenfall Vattenkraft, Voith Hydro, WSP Sverige, Zinkgruvan
385
Q.106 - R.23
and AFRY. The authors would like to thank Vattenfall AB for their contribution in build-
ing the experimental facility and to care for its proper operation. Energiforsk’s Dam
safety program is also acknowledged for support. Cautus Geo AS and Marmorta
Engineering AG are acknowledged for the support in installing the measurement
system.
REFERENCES
[1] LAGERLUND, J., TOROMANOVIC, J., DAHLIN, T., JUHLIN, C., JOHANS-
SON, S. (2020). ”Testdamm i Älvkarleby för skadedetektering”. Bygg & teknik,
1, p. 31–34. (In Swedish)
[2] SWEDENERGY. (2019). “RIDAS 2019: Energiföretagens riktlinjer för damm-

säkerhet.” Swedenergy, Stockholm, Sweden. (In Swedish)
[3] TOROMANOVIC, J., LAGERLUND, J., VIKLANDER, P., LAUE, J. (2020).

”Geotechnical instrumentation of an experimental embankment dam.” Euro-
pean Conference on Physical Modelling in Geotechnics (Laue, J., Bansal, T.
(ed.)). Luleå University of Technology, Luleå, Sweden, p 171–176.
[4] TOROMANOVIC, J. (2018). “On Parameter Identification for Better Predictions

of Dam Behaviour.” Licentiate Thesis, Luleå University of Technology, Luleå,
Sweden.
[5] VAHDATI, P. (2014). “Identification of Soil Parameters in Embankment Dams by

Mathematical Optimization.” Licentiate thesis, Luleå University of Technology,
Luleå, Sweden.
[6] SCHANZ, T., VERMEER, A., BONNIER, P. (1999). “The hardening soil model:
formulation and verification.” Beyond 2000 in Computational Geotechnics – 10
Years of Plaxis International. (Brinkgreve, R.B.J. (ed.)). Balkema, Rotterdam,
the Netherlands, p. 281–296.
[7] BRINKGREVE, R.B.J., KUMARSWAMY, S., SWOLFS, W.M., ZAMPICH, L.M.,

RAGI MANOJ, N. (2019). “PLAXIS 2019”. Plaxis bv, Delft, the Netherlands.
[8] TOROMANOVIC, J., MATTSSON, H., KNUTSSON, S., LAUE, J. (2020).

“Parameter identification for an embankment dam using noisy field data.”
Proceedings of the Institution of Civil Engineers - Geotechnical Engineering,
Available from: https://doi.org/10.1680/jgeen.19.00163.
[9] SJÖDIN, A., SJÖLI, E. (2020). “Initial development of flow through an embank-
ment dam.” Student paper, Division of Mining and Geotechnical Engineering,
Luleå University of Technology, Luleå, Sweden.
386
Q.106 - R.24
-------
-------
SAA MEASUREMENT FOR POSITIONING FILTER TIPS OF STANDPIPES IN

EMBANKMENT DAMS (∗)
Robert TORNBERG and Christian BERNSTONE

VATTENFALL AB
Peter VIKLANDER
VATTENFALL VATTENKRAFT AB and LULEÅ UNIVERSITY OF TECHNOLOGY
Hedwig HAAS
UPPSALA UNIVERSITY
SWEDEN
SUMMARY
As dam owner, Vattenfall has hundreds of standpipes installed in embankment

dams. When installing (drilling) these, there can be uncertainties about whether the
filter tip becomes correctly positioned in the ground in relation to as designed. In
dam safety assessments, the actual filter position and possible pipe deviation is
essential (e.g. placed into impermeable core instead of the permeable filter). This
paper describes how Shape Acceleration Array (SAA) measurements can be used
to determine the real position by depth of open standpipes including filter tips. The
method has up to now not been used in dams in Sweden. The instrument com-
prises of a continuous strand of segments having an array of three micro-electrical
mechanical sensors that makes it possible to track magnitude and direction of lat-
eral deviations similar to an inclinometer. Overall, results from laboratory and in-situ
testing show that it is possible to obtain the position by depth of the standpipe. To
conclude, this technique offers an extremely useful tool for quality control at installa-
tion as well as monitoring of already installed standpipes, even though some limiting
factors are to be improved.
∗ Mesure inclinométriques pour le positionnement des pointes poreuses de piézomètres
dans les barrages en remblai
DOI: 10.1201/9781003211808-119 387

Q.106 - R.24
RÉSUMÉ
En tant que propriétaire de barrages, Vattenfall a des centaines de pié-

zomètres installés dans des barrages en remblai. Lors de l’installation (par forage)
de ceux-ci, il peut y avoir une incertitude sur le fait que la pointe poreuse soit cor-
rectement positionnée dans le sol, conformément au plan. Lors de l’évaluation de
la sûreté d’un barrage, la position réelle du filtre et la déviation potentielle du pié-
zomètre sont essentielles (par exemple, en cas de positionnement dans le noyau
imperméable au lieu du filtre perméable). Cette publication décrit comment le Shape
Acceleration Array (ou chaîne inclinométrique) (SAA) peut être utilisé pour déter-
miner la position réelle du piézomètre en fonction de sa profondeur, pointe poreuse
incluse. Le SAA n’a jusqu’à présent jamais été utilisé dans des barrages suédois.
L’instrument est constitué de segments rigides reliés entre eux et contenant cha-
cun un réseau de trois capteurs micro-électriques mécaniques qui permettent de
déterminer l’amplitude et la direction de la déformation latérale à la manière d’un
inclinomètre. Globalement, les résultats de tests en laboratoire et sur site démon-
trent qu’il est possible d’obtenir la position en fonction de la profondeur du piézomètre
et de sa pointe poreuse. En conclusion, cette technique se révèle un outil très
utile tant pour le contrôle qualité de nouveaux piézomètres que pour la surveil-
lance de ceux déjà installés. Cependant, certains facteurs limitants restent encore
à améliorer.
1. INTRODUCTION
Large-scale hydropower development in Sweden started in the beginning of

the 20th century, with the largest dams built between 1950-1970 [1]. The average age
of hydropower embankment dams is in the order of 60 years [2]. Vattenfall, as one
of the major dam owners, manages a large number of dams, including embankment
dams. These are commonly zoned, with a central core of low permeability fine-
grained moraine flanked by zones of considerably more pervious sand-gravel filter
material and supporting fill with earth or rockfill shoulders. An example is shown
in Fig. 1. Several dams built later were originally built with a too coarse filter thus
susceptible to internal erosion [3]. This deficit has commonly been mitigated by a
downstream supporting berm together with dam instrumentation.
Swedish dam owners shall comply with the national dam safety guideline
RIDAS [5]. It is there stipulated that, for high consequence embankment dams,
the seepage through them shall be assessed together with data from dam instru-
mentation, i.e. typically installations of piezometers in open standpipes. Thus, by
monitoring the phreatic surface at the dam, the fulfillment of the dam safety can be
assessed over time [3].
388
Q.106 - R.24
Fig. 1
A cross section of av typical rock-filled embankment dam in Sweden. Modified
from reference [4].
Coupe d’un barrage en remblai rocheux caractéristique suédois. Modification de
référence [4].
Open standpipe piezometers are thus a common method to check the level of
the groundwater table and phreatic surface in embankment dams. This water level
is referenced to the position of the filter tip, where water is free to enter. Given the
different material properties of the core and its adjacent filters, it becomes essential
that the filter tip can be positioned correctly in the ground in relation to as planned. If
the standpipe, due to poor control and drill equipment, is placed in a low permeable
material (e.g. the core) instead of permeable material, then the interpretation of the
water level becomes wrong. This can possibly lead to false conclusions about the
safety status as stability. In addition, a leakage in the vicinity of a such standpipe
would not be detected.
The purpose of the drilling is essential for selecting the right drilling method
and for having appropriate drilling equipment at hand for any possible situation while
drilling. For open standpipe installations, the method used is typically conventional
percussion drilling based on freefalling weight or compression driven hammers with
drill bits. The trajectory of the borehole is set by the intended position of the piezome-
ter tip, the potential need to avoid already installed equipment, and the accessibility
for the drill rig at the surface. The position of the entrance and the inclination of the
drill string can be well controlled by the operator. However, there may be deviations
from the theoretical trajectory by depth and it is known that such unwanted differ-
ences may not be constant throughout the length of the borehole. This is not good for
the drill rig itself since excessive borehole deviations will cause rapid external wear
and certain types of drill string failures [6], nor for the dam owner. One reason for
this to happen might be due to drilling into coarse inhomogeneous material, stones
and boulders in the dam.
While drilling it is common practice that the drill operator throughout the work
checks the borehole trajectory by simply lowering a flashlight attached to a rope
389
Q.106 - R.24
into the borehole. In this way a significant deviation from straightness is detected
by losing sight of the lightbulb. Such a deviation would normally lead to disapproval,
abandon work, and a re-drill from a new starting point. This method of quality control
is quite inaccurate and is applied for roughly the top 10 meters of the borehole only.
There are examples when advanced drill rigs are used for dam safety pur-
poses. In 2019 Sonic drilling was used to collect undisturbed soil samples of the
shoulder material with limited impact on the dam itself [7]. Such work could be done
in combination with standpipe installations, but as a general tool for this would be
too costly.
In Spring 2020, Vattenfall in cooperation with the consulting company Sweco,

conducted tests with a shape acceleration array (SAA) on several embankment
dams with existing open standpipes [8]. The objective was to gain data on the actual
position by depth of the standpipes. This equipment is different from other tools for
measuring the straightness of boreholes in that it is not a logging tool where a sensor
is active while lowered into a borehole. The results from this test is presented in this
paper.
2. EQUIPMENT AND PROCEDURE
2.1. THE SAA EQUIPMENT
The equipment (a SAAscan by Measurand) is a strand assembly with a num-

ber of rigid segments with sensors connected with flexible joints (see Fig. 2). Each
segment contains of an array of three micro-electrical mechanical sensors. These
sensors register magnitude and direction of lateral deviation (tilt), similar to what
is done by conventional inclinometers. When applied into an open standpipe, the
strand is placed temporarily in the pipe∗ . When starting a measurement, each seg-
ment gives data on its position, which represent the standpipe position in space
referenced to a fixed origin. This can be done as long as the vertical deviation is
less than 60◦ .
The specific instrument that was used in the field tests is designed for use in
up to 40 m long standpipes. It has 80 segments of 0.5 meters, and thus maximum 80
sampling points (depending on how much of the strand that is needed for a standpipe
of a certain length). For the practical handling, these segments are rolled up on a
0.9 meter diameter cradled drum. Its total weight is 40 kg.
∗ There are as well instruments adopted for permanent installations.
390
Q.106 - R.24
Fig. 2
Components of the SAA scan.
Composants du scan SAA.
Before starting a standpipe examination all existing installations must be

removed. This is followed by positioning the cradle and the drum close to the stand-
pipe. The strand is then lowered into the pipe while assuring that its bottom segment
does not touch or rest on the lower end of the pipe (which would lead to zigzagging).
As the strand is in position its top segment must be clamped into a fixed position.
The azimuth of the reference mark on the top segment is set with the help of a com-
pass (this is needed for later data referencing to the magnetic north). When having
this in place, the instrument can be connected to a controlling unit (a computer) for
a check-up analysis. The procedure continues with starting the recording. A quick
view of the recorded data can be done in the field, but an in-depth analysis requires
further processing. The survey setup is shown in Fig. 3.
2.2. LIMITATIONS
There are some limiting factors when using the equipment in the sense of how
narrow or bent boreholes or standpipes are, that can be surveyed:
391
Q.106 - R.24
Fig. 3
Sketch of a SAA scan on a drum installed into an open standpipe.
Schéma d’un scan SAA sur enrouleur installé dans un piézomètre.
• Precautious installation should be taken if the standpipe is sharply bent. The

connection hose between segments should not be bent more than 70◦ (bending
it more may result in permanent damage).
• The strand with segments has a diameter of 23 mm, therefore, open stand-
pipes should have an inner diameter of at least 27 mm. The risk of having the
strand stuck in the standpipe increases if the diameter of the pipe is too small.
• The instrument should be placed as close as possible to the survey object. If
a distant position is needed, e.g. due to poor accessibility at a downstream
slope, then extra assistance is needed during the survey to make sure that
excessive bending between segments or other damages to the equipment
does not occur.
3. METHOD
3.1. LEARNING TO USE THE EQUIPMENT AND LABORATORY TESTING
In the initial stages of the project, the SAA equipment were tested in a lab-
oratory environment to verify the concept. Inclinometer pipes with a diameter of
50.8 mm (2 ) and a length of 8 m were used to do tests on both unaffected and
deformed pipes. In the latter case, a 255 mm deformation were applied 4.9 m below
its top. The pipe was positioned vertically and clamped to a stairway railing. The test
gave practical experiences on conditions that could be expected in the field, as well
as on the results that could be expected based on precise deviations due to forces
exerted on the test pipe.
392
Q.106 - R.24
The result of the verification test conducted by SAA measurement of an

inclinometer pipe with applied deformation in the laboratory is shown in Fig. 4.
Fig. 4
Measured deviation by depth with SAA (black) compared to applied deformation
(green) at the reference level (red).
Déviation mesurée en fonction de la profondeur avec SAA (noir) comparée à la
déformation appliquée (vert) au niveau de référence (rouge).
The results show that at the point where deformation is applied the difference
is 12 mm between the SAA measurement and the applied deformation after com-
pensating for the fact that the strand due to gravity is following one side of the pipe
wall. The result is satisfactory considering that the strand is not centered in the pipe
and that there might be some handling errors during setup and measurement.
3.2. FIELD SURVEYS
The surveys were conducted in three field campaigns during February to May
2020 to nine embankment dams in different parts of Sweden. 3-8 open standpipes
were measured at each site. At each site the measurements were done in three to
393
Q.106 - R.24
eight standpipes, thus covering a wide range of lengths, design and placement in
relation to the core.
The selection of which standpipes to survey was based on desktop work in

office. However, some of the cases that were identified as suitable turned out to be
less so at site, for example due to that permanently monitoring equipment could not
be temporarily removed. The sometimes challenging weather conditions with snow
in the terrain and in some cases ice inside the pipes led to that easy access pipes
needed to be prioritized.
The main objective of the three survey campaigns were to test the equipment’s
suitability for obtaining the true standpipe filter tip positions. The preference was to
investigate pipes situated close to the downstream part of the core, as illustrated
in Fig. 1. However, pipes in the downstream support fill were also investigated. It
is generally a coarser material that can be more difficult to drill in (as well as the
operator’s deviation control).
4. RESULTS AND ANALYSIS
4.1. INTRODUCTION
The field campaigns have resulted in an extensive data of SAA results. In this
paper we present results from two representative measurements: “Open Standpipe
1” and “Open Standpipe 2”.
4.2. OPEN STANDPIPE 1
The dam holding Standpipe 1 is 60 years old. The pipe was installed in the late
1960s, with the as designed position illustrated in Fig. 5 (dotted line) where the actual
measured position is displayed as well. There exists no detailed documentation from
the installation works.
The data from the SAA-measurements are presented in Fig. 6. The straight
line represents the intended trajectory. In this case the true position of the filter tip is
1 meter away in the downstream direction compared to as designed. There is also
a small (0.3 meter) perpendicular deviation. Moreover, the pipe has not been drilled
to the intended depth (i.e. 2.9 meter difference), even though it is still in the intended
material zone, giving data on the ground water table of the soil foundation.
394
Q.106 - R.24
Fig. 5
Section of the embankment dam including open standpipe 1.
Coupe du barrage en remblai avec le piézomètre 1.
Fig. 6
Results from open standpipe 1.
Résultats du piézomètre 1.
395
Q.106 - R.24
Fig. 7
Results from open standpipe 2.
Résultats du piézomètre 2.
4.3. OPEN STANDPIPE 2
The standpipe 2 was installed in the construction phase of an embankment

dam that was commissioned in the 1950s. A principal cross section of the dam is
shown in Fig. 8, together with the intended trajectory of the open standpipe (dotted
line). The dam has an unusual design where the impermeable core is constructed of
a combined till and concrete wall structure, with the concrete part at its downstream
side. The standpipe shall measure the ground water level in the filter zone next to
the concrete wall
By inspection at site of its above-the-surface visual part it is known that the pipe
has been impaired, whereby it has become inclined in the downstream direction.
Results from the survey is shown in Fig. 7. The data shows that the standpipe
was installed in the support fill (a mix of glacial till, gravel and rockfill), downstream
396
Q.106 - R.24
the targeted filter. The shape of the standpipe deviates from a straight line, which
raises questions about the causes for this to happen. This needs to be further inves-
tigated. The SAA data gives that the maximum downstream directed deviation is
approximately 1.7 meter. The corresponding dam-line deviation is small (0.1 meter)
Fig. 8.
Fig. 8
Section of the embankment including open standpipe 2.
Coupe du remblai avec le piézomètre 2.
5. DISCUSSION AND CONCLUSIONS
From the field-work practical perspective, the limiting factors for standpipe
surveys on embankment dams with the SAA-equipment are:
• that the in-place equipment for measuring the water table must be temporarily
removed (though this is needed irrespective of method for post-installation
straightness control),
• its size and weight,
• that the equipment can be challenging to install because of the limited inside
space of the standpipe itself.
There are a number of possible sources of error that affects the accuracy of
the measurements, e.g. that the top segment is not clamped to the pipe casing, that
the segments cannot be centered in the pipe, and the accuracy of the sensor itself.
The surveys were conducted from February to May, which was not optimal
due to freezing temperature and few hours with daylight. Surveying of embankment
dams lead to that steep terrains must be managed which led to sometimes difficult
and time consuming handling of the heavy equipment.
397
Q.106 - R.24
We have not benchmarked the SAA-equipment with other down the hole bore-
hole logging instruments. A product data sheet comparison of size and sensor
diameter shows that the SAA with its relatively small sized probe fits in 2” pipes
whereas other logging tools do not.
Our measurements in the laboratory and at the nine embankment dams on

both short and long standpipes shows that the method can be a useful tool for
improving the knowledge of existing embankment dam standpipes.
Other use of the equipment can be for quality control of the drill works when
installing new standpipes, and for repeated surveys for the quantification of ongoing
embankment dam movements (similar to conventional inclinometers but with a less
accurate reading).
The dam safety aspects of the cases where our measurements detected
sizeable in-situ horizontal deviations needs to be further evaluated.
REFERENCES
[1] Svenska kraftnät. (2019). “Dammar och Dammteknik: En introduktion”. [In

Swedish].
[2] Bernstone, C. (2006).”Automated performance monitoring of concrete dams”.

Doctoral Thesis, Lund University, ISBN 91-628-6982-5, 202 p.
[3] ICOLD. (2017). “Internal Erosion of Existing Dams, Levees and Dikes, and
their foundations”. Bulletin No 164. International Commision on Large Dams,
Paris.
[4] Vattenfall. (1988). ”Jord- och stenfyllningsdammar”. ISBN 91-7186-271-4. [In

Swedish].
[5] Swedenergy. (2017). “RIDAS – Energiföretagens riktlinjer för dammsäkerhet”.

[In Swedish].
[6] Drenth, C. (2019). Bulletin CDDA technical committee borehole deviation

guidelines, in 2019 CDDA AGM, St. John’s, Newfoundland, May 2020, 4 p.
[7] Ekström, I., Ljunggren, M. and Viklander, P. (2019). ”Large-Scale Sonic Drilling
To Determine Supporting Fill Parameters”. Association of State Dam Officials
Annual Conference(Dam Safety 2019), Orlando, USA, 8-12 September 2019.
[8] Haas, H. (2020). “Measurement of existing open standpipes in embankment

dams with new method for positioning”. Master thesis Uppsala University. [In
Swedish].
398
Q.106 - R.25
-------
-------
TEMPERATURE MODELLING AND DISTRIBUTED TEMPERATURE SENSING

USING OPTICAL FIBRES IN A TEST DAM (∗)
Sam JOHANSSON
Ph. D., M. Sc. (Civ. Eng.), Dam Safety Monitoring HydroResearch Sam J AB
Christian BERNSTONE
Ph.D., M. Sc. (Civ. Eng.), Dam Safety Monitoring, VATTENFALL, AB
SWEDEN
SUMMARY
Seepage monitoring in embankment dams based on temperature measure-

ment is a well-established method in Sweden. Measurements are performed by
manual measurements, point sensors or, by distributed measurements using opti-
cal fibres. The latter ways have so far been applied in about 100 hydropower and
tailings dams, starting in 1998. Significant development has been seen both regard-
ing monitoring and evaluation. The experimental-scale earth embankment dam, at
Vattenfall AB’s laboratory facilities in Älvkarleby, Sweden, was equipped with optical
fibres. The tests at this dam offers a unique opportunity to evaluate the use of tem-
perature measurements to detect the built-in unknown defects. Result from the first
filling exhibit eight locations of concentrated outflow were identified, i.e. two more
than the built-in defects during filling. Seepage flow exiting at one of the detected
outflow areas was simulated, to compare measured and simulated temperature.
The defect location, as well as its vertical extension and hydraulic conductivity, was
found. Further data evaluation using all temperature data, and more detailed sim-
ulations are planned in upcoming reports, when also other evaluation methods will
be used.
∗ Modélisation de la température et détection de température distribuée dans un barrage
d’essai
DOI: 10.1201/9781003211808-120 399

Q.106 - R.25
RÉSUMÉ
Le contrôle des infiltrations dans les barrages en remblai basé sur la mesure de
la température est une méthode bien établie en Suède. Les mesures sont effectuées
par des mesures manuelles, des capteurs ponctuels ou, par des mesures dis-
tribuées à l’aide de fibres optiques. Ces dernières méthodes ont jusqu’à présent été
appliquées dans une centaine de barrages hydroélectriques et de résidus miniers,
à partir de 1998. Des progrès significatifs ont été observés en matière de suivi et
d’évaluation. Le barrage en remblai de terre à l’échelle expérimentale, dans les
installations du laboratoire de Vattenfall AB à Älvkarleby, en Suède, était équipé
de fibres optiques. Les tests de ce barrage offrent une occasion unique d’évaluer
l’utilisation des mesures de température pour détecter les défauts inconnus inté-
grés. Le résultat du premier remplissage présente huit emplacements de sortie
concentrée, c’est-à-dire deux de plus que les défauts intégrés lors du remplis-
sage. L’écoulement d’infiltration sortant à l’une des zones d’écoulement détecté
a été simulé, pour comparer la température mesurée et simulée. L’emplacement du
défaut, ainsi que son extension verticale et sa conductivité hydraulique, ont été trou-
vés. D’autres évaluations des données utilisant toutes les données de température
et des simulations plus détaillées sont prévues dans les prochains rapports, lorsque
d’autres méthodes d’évaluation seront également utilisées.
1. INTRODUCTION
Vattenfall has, in accordance with the Swedish dam safety guideline [1],
designed and built an experimental embankment dam at their laboratory facilities in
Älvkarleby. The dam has six built-in defects, that all represents damages that even-
tually could evolve to a dam break. These defects are sized to be realistic, while
still large enough to be realistically detected in blind-tests by suitable geophysical
methods. These investigations/monitoring started in Spring 2020 and will continue
until Summer/Autumn 2021. The location and size of the defects will be revealed at
the end of 2021, after the performers have presented their reports.
Several methods will be applied by different performers both for short time mea-
surements (in terms of campaigns) and long-term measurement as shown Fig. 1.
This paper describes only the use of seasonal temperature variations for the detec-
tion of seepage flow and internal erosion (the method is sometimes called “Passive
Method”, in opposite to “Active Method” or “Heat Pulse Method” where heat/cold is
added). Measurements using the latter method will be done at the end of 2020.
Temperature measurements for seepage detection is widely used in Sweden
since the 1990-ties, and the use grow significantly when Distributed Temperature
Sensing in optical fibres was introduced. Today, around 100 embankment or tailings
dams are equipped with optical fibres of which 14 have permanent monitoring and
evaluation [2].
400
Q.106 - R.25
Fig. 1
Project time plan regarding methods and monitoring teams.
Calendrier du projet concernant les méthodes et les équipes de suivi.
2. TEST EMBANKMENT DAM
2.1. BACKGROUND – DESIGN
Our research builds on a previous blind test in Norway 2003 where the potential
of resistivity, self-potential and temperature surveying was undertaken on a rockfill
embankment dam [3]. The dam, with a height of 5.25 m and a length of 37 m,
was built with a central core of moraine with supporting rock fill and with six in-built
defects consisting of permeable material at various depths and locations unknown to
the personnel carrying out the measurements and data interpretation. It was from a
beforehand desktop sensitivity study based on the material properties of the building
materials known that it would not be possible to detect small defects using single
campaign surveys. This was also confirmed, but the geophysicist still managed to
indicate the locations by testing a monitoring approach where changes over time
could be taken into account.
The Norwegian blind-test dam was not built as a conventional earth embank-
ment dam, where there would normally be filters separating the impermeable core
401
Q.106 - R.25
from the outer coarse pervious shell (rock fill). In our case, the experimental dam
has been designed in accordance with the Swedish dam safety guideline.
The geophysical and geotechnical dam instrumentation installed all support

the control of relevant failure modes. This means that any discrepancies found in
analyses must be attributable to one of eight threats and twelve errors [4]. The built-in
defects represent damages that eventually could evolve to a dam break. These were
sized to be realistic, while still large enough to be realistically detected by the applied
methods. The final choice of sensor technologies is based on synthetic modelling
studies in 2018 to ascertain that the defects could be expected to be detectable.
Temperature finite element modelling gave promising results, and thus optic fiber
sensors passed to be installed. The results, which are outside the scope to present
in this paper, have been used for the placing of sensor cables in the dam.
The results from the modelling of resistivity and seismic methods gave also
promising results, and monitoring equipment were installed. Monitoring equipment
for the geotechnical research were also installed in the dam body.
2.2. GEOMETRY AND MATERIAL AND DEFECT
The chosen embankment dam design is a 20-meter-long dam where the imper-
meable core is supported by two connecting filter zones and support fills. The height
is 4 meters, and its base against the foundation perpendicular to the dam line is
15 meters.(Fig. 2). The two sidewalls which the dam connects to are slightly angular
(12,5 %). The bottom slab is inclined 1% towards the downstream side, where a
ditch collects the seepage water through the dam body. This seepage is divided by
ribs into eight sections which are continuously monitored by flow, but this data set
is not made available to the investigators (but will be revealed later).
The dam rests on a rigid concrete support structure submerged into a river-
bank of 12 meters of sand overlying bedrock (drained conditions). The experimental
dam thus has its crest in line with the surrounding ground. To the extent possible,
the design is made to avoid structural members with negative impact on geophys-
ical measurements. The bottom concrete slab of the containment has therefore
been reinforced with fiberglass reinforcement bars. Since there is no electrically
conductive reinforcement this resembles foundation on rock.
The built-in defects, that all represents damages that eventually could evolve
to a dam break, are:
1. Cavity in the core (wood cube 0.4 m × 0.4 m × 0.4 m).

2. Horizontal permeable zone passing through the core (rectangular cross
section 0.1 m high and 0.5 m wide).
3. Vertically loose zone (elongated zone with square cross section with side
0.3 m).
402
Q.106 - R.25
Fig. 2
Embankment test dam at Älvkarleby, with its 3 m wide crest (holding short pipes for
the use of seismic energy sources) and the support fill of its downstream side (left).
Barrage d’essai en remblai à Älvkarleby, avec sa crête de 3 m de large (contenant
des tuyaux courts pour l’utilisation de sources d’énergie sismique) et le remblai de
soutien de son côté aval (à gauche).
4. Lump of concrete or large stone (cube 0.5 m × 0.5 m × 0.5 m).

5. Permeable horizontal zone at side (0.1 m × 0.1 m).
6. Filter defect on the upstream side.
Being a blind-test, the position of these defects are not known to the
investigators (in fact kept secret to a few persons only). They will be revealed later.
2.3. CABLE INSTALLATIONS
The cable installation is extensive compared to typical installations in large

dams, where for post-dam construction practical reasons normally just one or a few
cables are installed along the dam. All cables contain both multimode and single
mode fibres (see Table 2). This setup also allows measurements of acoustic signals
(Distributed Acoustic Sensing, DAS). Furthermore, the cable in the downstream toe
contains a copper wire that allows for heat-pulse measurements. This paper will only
present data from initial temperature measurements in the filter.
403
Q.106 - R.25
Table 1
Instrument and cable specifications.
Specifications Type Description
Optic fibre cables
Passive (Ø 6,5 mm) Silixa PS-2S4M-1PU065-01-Y 400 m. 6-channel tight buffered fibre optic cable
(4 multimode & 2 single mode).
Active (Ø 9,1 mm) Silixa HS-2S2M2C-1PU091-01-B 80 m. 2-channel composite fibre optic cable.
Instrument
Rugged distributed Silixa XT-DTS Sample resolution 25 cm.
temperature sensor
In the test dam there is one cable, placed at four levels in the upstream filter,
and one on the core crest (Fig. 3). Some parts of this cable are also placed in
water, allowing the water temperature to be measured. A second cable was used to
measure at five levels in the downstream filter, and a third cable is placed closed to
the dam toe.
Fig. 3
Illustration of the dam showing installed cables at different levels on the upstream
(US), core crest, downstream side (DS), and dam toe.
Coupe transversale du barrage montrant les câbles installés à différents niveaux
en amont (US), en crête du noyau, en aval (DS) et au pied du barrage.
The dam chainage (CH) along the dam starts on top of the dam at the con-
nection to the sloping concrete wall at the left side (CH 0 m) and ends at the right
abutment at CH 20.6 m on the crest. The bottom starts at CH 0.3 m and ends at
CH 20.3. All temperature data is presented along the dam using dam chainage.
2.4. OPERATION AND BASIC MEASUREMENTS
Filling up to the retention water level started on May 04 and was due to onload-
ing concerns performed slowly during working days and with a kept constant water
404
Q.106 - R.25
level during weekends. Continuous measurements of the reservoir and air tem-
peratures (Fig. 4), and the seepage flow rate are carried out (but not available
for the investigators). There are also temperature sensors in the pore pressure
sensors. These point temperature measurements can be used for evaluation and
compared with the line information that is achieved from the distributed temperature
measurements.
Fig. 4
Temperature in water and air, and water levels at filling.
Température de l’eau et de l’air et niveau d’eau lors du remplissage.
(1) Date [−] (1) Date [−]
(2) Temperature [◦ C] (2) Température [◦ C]
(3) Water level [m] (3) Niveaux [m]
(4) Water temperature (4) Température de l’eau [◦ C]
(5) Air temperature (5) Température de l’air [◦ C]
(6) Water level in reservoir (6) Niveau de l’eau
3. INITIAL TEMPERATURE SIMULATIONS
3.1. BACKGROUND
The background for using temperature measurements for seepage detections

in dams is the natural seasonal temperature variations in the reservoir water and the
ambient air that gives the temperature distribution in an embankment dam [5] and
[6]. This is also valid for the experimental dam, although the scale is smaller. The
effect of heat conduction from the surface will of course be much larger. On the other
405
Q.106 - R.25
hand, the streaming path lengths (i.e., the distance from the entrance dam surface to
the measuring point) will be shorter than in a conventional larger hydropower earth
embankment dam.
The basic principle for using temperature to measure seepage is the same
irrespective of scale: water enters into the upstream fill and filter, through the dam
core and further into the downstream filter and fill, and finally exit at the dam toe.
Cold water will enter during winter and warm water during summer. The seasonal
temperature response in the dam is dependent on the seepage flow. Also, short time
temperature variations (colder/warmer weeks) may cause temperature variations at
large flow, and tentatively as well day/night temperature variations at high seepage
flow.
Evaluation of temperature measurements during filling, or at large reservoir
water level variations, have been found useful to better understand inflow at dif-
ferent water levels when performed in large dams [7]. Such measurements have
revealed local areas with seepage outflow. The approach was also successful at the
previously described Norwegian blind test site [3].
Evaluation of temperature for seepage detection is often based on continuous
measurements using the seasonal pulse only. This approach will also be used at
the test dam based on a longer measurement period, about one year.
Not all the built-in defect will cause increasing seepage. Increased seepage
flow is expected at “Horizonal zone through the core”, “Vertical loose zone”, “Per-
meable zone at side”, and maybe at “Filter defect”. The remaining defects will cause
seepage flow decrease, unless concentrated seepage flow will appear around the
“concrete and wood cubes”. Local areas with lower permeabilities than the original
core (i.e., in this case the “concrete and the wood cubes”) will be difficult to detect
when the thermal transport with the water flow is of similar to the heat conduction.
The temperature distribution in the dam is a result of the coupled processes of
heat flow and seepage water flow through the dam. The resulting distribution can be
modelled using Richards equation which enables solving the transport in both the
saturated and unsaturated regions. The thermal process is described by heat con-
duction and convection. The boundary conditions are pressure (head) as described
by the reservoir water level and the corresponding downstream water level, and tem-
peratures in water and air. Since those conditions are time dependent a transient
solution is needed. The numerical modelling for the cases presented in this paper
has been performed using the sub-surface module of the finite-element software
COMSOL Multiphysics. In this way, the coupled transient problem for unsaturated
flow (i.e., the Richard equation) and heat flow can be solved.
3.2. SIMULATION MODEL FOR DESIGN SUPPORT
Simulations were made in the design phase in order to determine the detection
ability for the methods to be tested. The result provided guidance for the project team
406
Q.106 - R.25
on the selection of appropriate size and properties of the defects. Basic material data
were provided (Table 2). Remaining data needed for the simulations (using Richards
equation for un-saturated flow) were either calculated or assumed based on available
information and experience from similar simulations. The design geometry was also
given showing the central core, surrounded by two filters upstream and downstream,
supported by fill (Fig. 5).
Table 2
Input values for the simulations for the different material as shown in Fig 5.
Données d’entrée pour les simulations pour les différents matériaux, tel que
montré à la Fig 5.
Given data Mat A Mat B Mat C Mat D
Compact density [kg/m3 ]: 2600 2600 2600 2600
Wet density [kg/m3 ] 2250
Dry density [kg/m3 ] 1900 1700 1800
Hydraulic conductivity K [m/s] 1.00E-07 1.00E-05 1.00E-02 1.00E-01
Calculated/Assumed data Mat A Mat B Mat C Mat D
Permeability, k [m2 ] 1.00E-14 1.00E-12 1.00E-09 1.00E-08
Volumetric heat capacity [MJ/(m3 K] 2.1 2.1 2.1 2.1
Heat capacity [kJ/[kgK)] 808 808 808 808
Thermal Conductivity [W/mK] 2.9 2.9 2.9 2.9
Van Genuchten (unsaturated zone), [8]
alfa [1/m] 1.5 2 3.5 3.5
n 2 2.5 2.5 2.5
I 0.5 1.5 2.5 3.5
Formation compressibily [1/Pa], [9] 1.00E-07 1.00E-09 1.00E-09 1.00E-09
Fig. 5
Cross section of the test dam. The materials are defined in Table 2.
Coupe transversale du barrage avec les matériaux définis au Table 2.
The transient temperature simulations were made using the stationary pres-
sure solution for a high-water level (+3.5 m) and +0.25 m on the downstream side. All
simulations were made in 3D, where the defects were placed in the centre (x = 0),
which also acted as a symmetry plane, i.e. only half of the dam needed to be
included.
407
Q.106 - R.25
3.3. SIMULATION RESULTS
Simulations were made for the first five defects (No 1 – No 5, see above). The
material B, intended to be used for the construction of the defects No 2, 3 and 5,
was just 100 times larger in permeability than the core permeability. The simulations
gave that the flow rate for through the core without defects was about 6.1 · 10−6 m3 /s.
Damage No 2 gave the largest flow rate (?) increase, about 10% when placed 1 or
2 m above the bottom. The other defects gave seepage flow changes of about ±1%.
The thermal simulations showed that neither of those small flow changes
(<0.7 · 10−6 m3 /s) would be possible to detect with temperature measurements.
Simulations indicated that about ten times higher permeability would give such tem-
perature changes that it could be detected by continuous measurements. A more
permeable material was therefore suggested to be used in the defects in the test
dam.
4. MEASUREMENTS DURING FIRST FULL FILLING
4.1. TEMPERATURE MEASUREMENTS ALONG THE DAM
The temperature measurements were made using a Silixa XT-DTS, with a

sampling resolution of 0.25 m (Fig. 6). The measurement time was 30 min for each
channel, with a repetition time of 2 hours. Further information about temperature
monitoring using optical fibres can be found in [10] and [11].
The measurements were made channel by channel for the upstream lines
and crest, downstream lines and for the cable in the dam toe. The DTS-unit was
placed in a monitoring container, where diurnal temperature variations occur. This
effect is seen in the temperature data versus time as small variations (<0.1◦ C). Only
measurement for the first filling to the reservoir retention level is presented here,
and just for one line downstream (DS L2 in Fig. 6). This cable location represents a
preferred location for installations in new dams (whereas post-installations in existing
dams are relying on easy-access, e.g. close to the dam toe).
4.2. TEMPERATURE RELATED OBSERVATIONS AT RESERVOIR FILLING
The temperature in the dam body before filling started was between 4 and 5◦ C
(Fig. 6). A first temperature increase with a width of 2 m was observed in the DS L2
at CH 8 m, already after about 3 days after the filling started. A second temperature
408
Q.106 - R.25
change, about 1 m wide, followed about 4 days after filling at CH 12 m. These are
quick responses, indicating relatively high-water flow through deeply located defects.
Fig. 6
Temperature measurements in cable DS L2 at selected times during filling.
Mesures de température dans le câble DS L2 à des moments choisis pendant le
remplissage.
(1) Dam chainage [m] (1) Longueur [m]
(2) Date [−] (2) Date [−]
A third temperature change was observed after about 7 days at CH 3 m. This

water outflow is quite concentrated.
Furthermore, two temperature changes were detected after 3 weeks (May 25)
at CH 10 m, and at CH 15 m. Another one is seen about one day later at CH 11 m.
There are signs of two additional outflows in early June at CH 0.5, close to
the left abutment, and at CH 19 close to the right abutment. These all in all eight
outflows exceeds the number of built-in defects (six).
4.3. NUMERICAL SIMULATIONS
To illustrate what kind of information that can be achieved at increasing the

water level, we will focus on the temperature deviation at CH 3 m, studying data from
one line (DS L2) only. The sampling resolution indicates a concentrated seepage
outflow with a width of about 0.5 m, or probably less in the core. We also see from
Fig. 6 that the temperature is spreading wider with time.
409
Q.106 - R.25
Numerical simulations (forward modelling) were made to analyse the cause of

this temperature anomaly (at CH 3 m). This is done by testing a number of material
property changes (both by extent and by resulting water flow increases) that would
give the same temperature response as registered. The simulations were made in
a similar way as was used for the design, as described above. Basic data were also
the same. All simulations were made in 3D. The model width was one meter only,
with the defect in the symmetry plane.
From the measurements in DS L2 we have some idea about the width of the
defect (0.4 m), but no information about the flow rate, the level of the defect nor
its vertical extension. Simulations were made for defects with different heights and
level. However, the permeability was assumed to be constant within in the entire
defect.
Several combinations of those parameters will be possible of which two are

showed in (Fig. 7). In the first case (C1) we assume a defect between level 22.7 m
Fig. 7
Simulated and measured temperature in DS L2.
Température simulée et mesurée dans le câble DS L2.
(1) Date [−] (1) Date [−]
(3) Water level [m] (3) Niveau de l’eau [m]
(4) Water temperature (4) Température de l’eau
(5) Calculated temperature, C1 (5) Température calculé, C1
(6) Calculated temperature, C2 (6) Température calculé, C2
(7) Measured temperature (7) Température measuré
(8) Water level in reservoir (8) Niveau de l’eau
410
Q.106 - R.25
and 23.4 m, with permeability of the damaged zone that is 10000 higher than its sur-
roundings (i.e. 1 · 10−3 m/s). The modelling results based on these parameters are
in good agreement with measured temperatures for the first two weeks of measure-
ments. The temperature rise, starting on May 19, occurs later in the simulations. A
better agreement with the second temperature rise is achieved by a defect (C2) posi-
tioned at 23.4 m and 24.4 m in height, with 20000 higher hydraulic conductivity (i.e.
2 · 10−3 m/s). However, the calculated temperature will give an earlier temperature
response than what was measured. It is probable that these simplified defect models
(square shaped cross sections with constant hydraulic conductivity) will not describe
the defect perfectly. A one-meter high defect at level at level 23.3 m with about 15000
times higher hydraulic conductivity seems probable based on measurements in just
one line and simulations.
Further simulations will be performed also integrating the seasonal tempera-

ture variation from measurements at retention level for the coming year. Information
from other measuring lines, and point sensors can also be included.
This evaluation includes only a limited amount of all data collected. Further
evaluation will be performed analyzing temperatures upstream and downstream the
core on different levels, as well as temporal temperature variations at retention level.
Further simulations are also planned.
5. DISCUSSION AND CONCLUSIONS
The temperature distribution in embankment dams as well as in tailing dams is

dependent on the seepage regime, which is the reason why temperature measure-
ments are widely used in Sweden for detection of seepage flow changes and internal
erosion detection. Measurements can preferably be performed using distributed
temperature measurements using fibre optics. The installation of such cables in the
experimental dam is very extensive, allowing the method to be tested.
The experimental dam has six in-built defects, and several methods will be
applied for finding the location for those defects, which are unknown for the monitor-
ing teams. Measurements started in April 2020 and will continue until the middle of
2021, followed by reports from all monitoring teams. However, some evaluation of
temperature data can be made based the data from filling the reservoir, a method
that has been successfully used at real large dams.
Evaluation of data from one cable location only was chosen to describe an
evaluation situation based on information from increasing water levels.
Eight locations of concentrated outflow were identified, i.e. two more than the
built-in defects. Seepage flow exiting at one of the detected outflow areas was simu-
lated, to compare measured and simulated temperature. A possible defect location
was found as well as its vertical extension and hydraulic conductivity/flow rate.
411
Q.106 - R.25
Further data evaluation using all temperature data, and more detailed simula-
tions are planned in the coming final reports. Other evaluation methods will also be
used.
REFERENCES
[1] RIDAS (2017). Kraftföretagens riktlinjer för dammsäkerhet, Swedenergy AB.
[2] Johansson S. and Sjödahl P. (2017) Fibre-optic temperature monitoring for

seepage detection in embankment and tailings dams, in ICOLD 85th Annual
Meeting, Prague.
[3] Johansson, S. and Nilsson, Å. (2005). Internal Erosion Detection at the Røs-
vatn Test Site – Experiences from blind test using Resistivity, Self-Potential,
Temperature, and Visual inspection. Part A – Assessment report. Elforsk
rapport 05:42, Stockholm.
[4] ICOLD (2017), Dam safety management: Operational phase of the dam life
cycle, Bulletin No. 154, International Commission on large dams, Paris.
[5] Johansson, S. (1997) Seepage Monitoring in Embankment Dams, Doc-

toral Thesis, TRITA-AMI PHD 1014, ISBN 91-7170-792-1, Royal Institute of
Technology, Stockholm.
[6] CEATI REPORT No. T062700-0214, A Guide for Seepage Monitoring of

Embankment Dams using Temperature Measurements, 2009, 62 pages.
[7] Johansson, S., Sjödahl, P. Viklander, P. (2012) Upgrading seepage monitoring

system using temperature measurements – experience from measurements
and modelling at Seitevare dam. ICOLD Congress Q95, R.18, Kyoto.
[8] van Genuchten M Th (1980) A Closed-form Equation for Predicting the

Hydraulic Conductivity of Unsaturated Soils, Soil Sci. Soc. Am. J., vol. 44.
[9] Freeze R A, Cherry J A (1979) Groundwater. ISBN 0-13-365312-9, Prentice

Hall N.J., 604.
[10] Dakin, J.P., Pratt, D.J., Bibby, G.W., and Ross, J.N. (1985), “Distributed opti-
cal fiber Raman temperature sensor using a semiconductor light source and
detector”, Electron. Lett., vol. 21, no. 13, pp. 569-570, June 1985.
[11] Tyler, S. W., Selker J. S., Hausner, M. B., Hatch, C. E., Torgersen, T., Tho-
dal, C. E., Schladow, S. G. (2009) Environmental temperature sensing using
Raman spectra DTS fiber-optic methods, Water Resources Research, Vol. 45,
W00D23, 11 pages.
412
Q.106 - R.26
-------
-------
DAM BODY BEHAVIOR MONITORING BY GNSS IN HACHISU DAM AND

APPLICABILITY (∗)
Kunihiro TOMITA
President, CIVIL ENGINEERING & ECO-TECHNOLOGY CONSULTANTS
CO., LTD.
Tetsuya SUMI
Professor, Disaster Prevention Research Institute, KYOTO UNIVERSITY
Akira SUZUKI
Director, Hachisu Dam Management Office, Chubu Regional Development Bureau,
MINISTRY OF LAND, INFRASTRUCTURE, TRANSPORT AND TOURISM
Shigeharu JIKAN Shigeyoshi NOYORI

CIVIL ENGINEERING & ECO-TECHNOLOGY CONSULTANTS CO., LTD.
Nobuteru SATO Chikako, ARAYA

KOKUSAI KOGYO CO., LTD.,
JAPAN
SUMMARY
Hachisu Dam is a concrete gravity dam, 78 m in dam height. Nearly 30 years

have passed since the completion and the dam comprehensive inspection has been
implemented. As part of the dam comprehensive inspection, a dam body deforma-
tion monitoring system using GNSS was introduced in Hachisu Dam in order to
clarify the factors of dam body deformation measured in the plumb line.
This paper first reviewed the dam safety management system in Japan and the
status of technical development in GNSS used in existing dams. Then, this paper
∗ Contrôle du comportement du corps du barrage de Hachisu par GNSS et applicabilité
DOI: 10.1201/9781003211808-121 413

Q.106 - R.26
studied the measurement values of GNSS, which is being used for measurement
in Hachisu Dam, and the existing plumb line using multiple regression analysis and
verified the applicability of GNSS in the dam body deformation measurement of
Hachisu Dam and reported issues to address in the future concerning GNSS.
RÉSUMÉ
Le barrage de Hachisu est un barrage-poids en béton de 78 m de haut. Il a été

terminé il y a presque 30 ans et un système de visites détaillées a été mis en place.
Dans le cadre du programme des visites détaillées, un système de surveillance de
la déformation du corps du barrage par GNSS a été installé pour le barrage de
Hachisu afin de clarifier les facteurs de déformation du corps du barrage mesurés
au fil à plomb.
Ce rapport examine dans un premier temps les systèmes de gestion de la

sécurité des barrages au Japon et le statut des développements techniques dans
la géolocalisation par satellites (GNSS) mise en œuvre pour les barrages existants.
L’étude s’intéresse ensuite aux valeurs de mesure par la GNSS dans le cadre du
barrage de Hachisu, aux valeurs de mesure par pendules soumises à analyse de
régression multiple et elle procède à un examen de l’applicabilité de la GNSS dans
la mesure de la déformation du corps du barrage de Hachisu et des problèmes
connus à résoudre dans le futur concernant la GNSS.
1. INTRODUCTION
Hachisu Dam is a concrete gravity dam, 78 m in dam height, located in the

upstream area of the Kushida River system, the central region of Japan’s main
island. Nearly 30 years have passed since the management of Hachisu Dam started
in 1991, and “Dam comprehensive inspection” was implemented from 2019 to 2020
in view of the deterioration due to aging to survey/evaluate the soundness of the dam
and reflect those results in the future maintenance policy [1]. As part of this dam
comprehensive inspection in Hachisu Dam, the dam body deformation monitoring
system using the Global Navigation Satellite System (GNSS) was introduced in order
to clarify the factors of dam body deformation measured in the existing plumb line.
This paper first reviews the structure of dam safety management system imple-
mented in Japan and the status of GNSS technology as a method of measuring dam
body deformation. Then, this paper analyzes/compares the measurement values of
GNSS and the existing plumb line in Hachisu Dam and reports the applicability of
GNSS in the dam body deformation measurement of concrete gravity dams as well
as issues to address in the future.
414
Q.106 - R.26
2. APPLICATION OF GNSS TO THE DAM BODY DEFORMATION

MEASUREMENT TECHNOLOGY
2.1. AM SAFETY MANAGEMENT SYSTEM AND DAM BODY MEASUREMENT IN JAPAN
The dam safety management system in Japan consists of daily inspection,

temporal inspection, periodic examination, and comprehensive inspection, as shown
in Fig. 1 [2][3][4]. Daily inspection is basic inspection that is implemented by the dam
administrator to grasp the daily condition of dam facilities, while temporal inspection
is implemented in the event of an earthquake or flood exceeding a certain extent.
Periodic examination is implemented by experts other than dam administrators gen-
erally every three years to evaluate soundness based on change of state in a certain
period and comparison with allowable values / general values. In contrast, dam com-
prehensive inspection is implemented to investigate and evaluate comprehensively
the soundness of dams that have been in operation for every 30 years, with focus on
long-term changes due to aging, internal state of structures, etc. and aims to realize
effective and efficient maintenance by reflecting the maintenance policy obtained as
a result of such investigation/evaluation in maintenance, daily inspection, etc.
Fig. 1
Structure of the dam safety management system in japan.
Organigramme du système de gestion de la sécurité des barrages au Japon.
Measurement items required for the monitoring conditions of dam safety man-
agement system are provided by the Cabinet Order Concerning Structural Standards
for River Administration Facilities, as shown in Table 1 [5]. Note that these measure-
ment items for each dam type are the minimum requirements and stress, strain,
temperature, etc. are also generally measured in individual dams.
415
Q.106 - R.26
Table 1
Measurement items at dam safety management.
Dam type Dam height Measurement item
Concrete gravity dam Less than 50m Leakage/seepage, uplift
50m or higher Leakage/seepage, uplift, Deformation amount
Arch dam Less than 30m Leakage/seepage, uplift
30m or higher Leakage/seepage, uplift, Deformation amount
Embankment dam Homogeneous Leakage/seepage, Deformation amount,
Phreatic line
Other* Leakage/seepage, Deformation amount
(*) Zoned embankment dam, etc.
At Hachisu Dam (PG, H = 78 m), described below, data is collected and

stored through automatic measurement, including of leakage/seepage, uplift, defor-
mation amount, and other items. These data comprise the safety management
measurement items.
2.2. APPLICATION OF GNSS TO DAM BODY DEFORMATION MEASUREMENT IN JAPAN
In Japan, the displacement measurement of ground and structures using

GNSS started in the 1980’s and then GNSS also began to be used for the displace-
ment measurement of large-scale slopes vulnerable to landslide from around 1990.
As studies advanced in the latter half of the 1990’s on the development and accuracy
enhancement of GNSS receivers and automatic measurement system, it became
possible to measure three-dimensional displacement by millimeter using GNSS [6].
In the field of dam body deformation measurement for dam safety management, the
number of dams where GNSS is introduced in parallel with the existing instrumen-
tation system for monitoring dam body deformation has been also increasing since
the beginning of the 2000’s.
The positioning method of GNSS is roughly classified into the point positioning
method and the relative positioning method, as shown in Fig. 2. The point positioning
method has been used for navigation of cars, airplanes, ships, etc. and also incorpo-
rated in mobile phones, and the accuracy of point measurement is several to several
tens of meters. The relative positioning method obtains relative coordinates of two
points with high accuracy by receiving electric waves with two receivers. Particularly,
the static positioning method, one of the relative positioning methods, is appropri-
ate when highly accurate measurement by millimeter, e.g. dam body deformation
amount, is required [6]. In addition, errors of measurement values in GNSS are
classified into those related to the GNSS satellite, transmission path, and receiver.
Of these errors, the effects of clock errors, orbit errors, and ionosphere are mostly
removed by the static positioning method. It is difficult to remove errors caused by the
reflection of electrical GNSS waves against buildings, etc., such as multipath errors,
receiver noise, and the effects of satellite constellations [6]. For such remaining
random errors, it is necessary to perform smoothing, and there is a method of using
416
Q.106 - R.26
Fig. 2
Positioning methods of GNSS.
Méthodes de positionnement par GNSS.
1 Point positioning method 1 Méthode de positionnement par point
2 Relative positioning method 2 Méthode de positionnement relative
a trend model for this smoothing [7] [8]. This method estimates a true value from
uneven measurement values using a time-series analysis model with probability
structure. In GNSS deformation measurement of an embankment dam body using
this method, a calculation accuracy of about 1–2 mm was confirmed [9].
Introduction of GNSS to dam body deformation measurement in Japan began
from embankment dams. The existing method in dam body deformation mea-
surement of an embankment dam periodically measures the measurement points
prepared on the dam body surface by electro-optical distance measurement, while
GNSS is used not only at ordinary times but for emergency measurement in
case of an earthquake or flood utilizing the merit of GNSS, which has the func-
tion of automatic and continuous measurement of three-dimensional displacement
with high accuracy. During the 2011 off the Pacific coast of Tohoku Earthquake
(Magnitude = 9.0), the dam body deformation amount in Ishibuchi Dam (CFRD,
H = 53 m, 1953) was promptly measured with GNSS, so that it was confirmed that
GNSS is available for temporal inspection of dams in the event of an earthquake [10].
Further, during the 2018 West Japan Heavy Rain, dam body deformation in Iwaya
Dam (ER, H = 127 m, 1976) due to a rapid water level rise when a flood occurred
was continuously measured in real time using GNSS [11].
In addition, introduction of GNSS is proceeding in not only in dams
under construction but in dams during operation partially because “Engineering
Manual on External Displacement Measurement of Embankment Dams Using GPS”,
a guideline for GPS measurement of embankment dams, was formulated in 2014 [6].
417
Q.106 - R.26
Fig. 3
GNSS for a rockfill dam.
GNSS pour barrages en enrochement.
On the other hand, plumb lines are mainly used for dam body deformation
measurement of concrete dams. Deformation measurement using the plumb line is
highly accurate but the plumb line is generally installed only on the representative
cross section for the reasons of impact on dam body construction, construction
expense, etc. resulting from plumb line installation. In addition, if a plumb line needs
to be newly installed in a dam during operation, it is necessary to consider the impact
of drilling a new plumb line hole on dam body. In contrast, installation of GNSS is
advantageous in that it can be installed with common-use tools, that increase of
measurement points is easy, and that there is little impact of installation to dam body.
At present, GNSS has been installed in some concrete dams including dams
in operation, where measurement of dam body deformation is proceeding includ-
ing analysis and evaluation of measurement results. The measurement accuracy
of GNSS is about 1–2 mm, less accurate than a plumb line, which enables highly
accurate measurement of “detection accuracy of ±0.1 mm” but a result that a mea-
surement equivalent to plumb line is possible with regard to the behavior monitoring
of dam body deformation to evaluate relative difference between measurement
points [12].
As described above, in comparison with conventional measurement methods,

such as electro-optical distance measurement in embankment dams and plumb
line measurement in concrete dams, the GNSS measurement system in dam body
deformation is highly useful, because it secures the continuity of measurement fre-
quency / real-time data, it can be expanded to enable automatic measurement of
multiple points in three dimensions, and can be easily installed at low cost not only
in dams under construction, but in dams during operation, among other ways.
GNSS was also introduced in addition to conventional plumb line measurement

to Hachisu Dam, a dam in operation for 30 years, to measure dam body deformation.
418
Q.106 - R.26
3. UTILIZATION OF GNSS IN HACHISU DAM
3.1. OUTLINE OF HACHISU DAM
Table 2, Fig. 4, and Fig. 5 show the specifications, location, and

plan/front/sectional views of Hachisu Dam, respectively [13]. The plumb line was
installed in the 12BL of the largest cross section and dam body deformation has
been measured since the start of first filling.
Table 2
Specifications of Hachisu Dam.
Dam type Concrete gravity dam
Dam height 78 m
Crest length 280 m
Dam volume 484,000 m3
Reservoir storage 32,600,000 m3
Objective F,N,W,P
Completion 1991
F: Flood control
N: Maintenance of normal function of river water
W: Water supply, P: Power generation
Fig. 4
Location of Hachisu Dam.
Location of Hachisu Dam.
419
Q.106 - R.26
Fig. 5
Structure of Hachisu Dam and arrangement of plumb line and GNSS.
Structure du barrage de Hachisu et ordonnance du fil à aplomb et de la GNSS.
1 Plumb line 1 Fil à plomb
2 Measurement points of GNSS 2 Points de mesure du GNSS
3 Control point of GNSS 3 Point de contrôle du GNSS
3.2. OUTLINE OF THE HACHISU DAM COMPREHENSIVE INSPECTION
Fig. 6 shows the structure of dam comprehensive inspection in Hachisu Dam

and the following outlines the inspection results of dam body, etc. [1].
420
Q.106 - R.26
Although such phenomena as minor cracks and efflorescences are found on

the surface of dam body concrete, they are seen only on the surface and the dam
functions are maintained. Although there was much leakage/seepage at the time
of first filling of reservoir, but it decreased thereafter and is stable at present. Uplift
is stable on the linear relationship with reservoir level and the stability of dam body
was also confirmed in the verification based on actually measured uplift. As for dam
body deformation, the value of dam body deformation measured in accordance with
the plumb line fluctuates yearly in the upstream and downstream directions due to
the effect of reservoir level and temperature, but measurement values are found
remaining over time about 20 mm in the directions of upstream and right bank. For
these remaining measurement values of the plumb line, however, no phenomena
representing the effect of these values were found in the visual observation of dam
body concrete.
From the facts described above, Hachisu Dam is considered to have no prob-
lems with safety and functionality. Meanwhile, since residual displacement value
was found in the plumb line measurement, displacement of the dam body and foun-
dation has been measured with GNSS to monitor their state since February 2020 in
order to analyze the factors of remaining.
Fig. 6
Outline of the structure of dam comprehensive inspection.
Généralités relatives à l’organisation d’une visite détaillée du barrage.
3.3. INSTALLATION OF GNSS
Fig. 5 also shows the arrangement of the plumb line and GNSS in Hachisu
Dam. The plumb line is a normal plumb line type that measures the relative
421
Q.106 - R.26
deformation in the upper and lower parts of the dam body and is installed in the
12BL of the largest cross section. On the other hand, since the measurement value
of GNSS represents relative displacement to GK1, which is the control point in the
static positioning method, in order to find such relative deformation of the dam body
as found in the plumb line, we decided to measure relative deformation of the dam
body with the difference in GNSS measurement values of the crest and foundation
by setting GNSS measurement points of G1 on the crest of 12BL, G2 on the crest of
14BL, and G3 on the foundation of 14BL [12]. Comparison with the plumb line mea-
surement values was made using the difference values of “G1-G3” and “G2-G3”,
which represent difference between the crest of 12 BL/14BL and the foundation of
14BL. Fig. 7 shows the state of GNSS installation in Hachisu Dam.
Fig. 7
GNSS installation in Hachisu Dam.
Installation de la GNSS pour le barrage de Hachisu.
3.4. ANALYSIS OF GNSS MEASUREMENT RESULTS
Of the GNSS measurement values in Hachisu Dam, Fig. 8 shows measure-

ment values at G1 of the crest and trend values. The measurement value of G1
was calculated by the static positioning method using GK1 as a control point. A
characteristic of GNSS measurement values is that there is little variation of errors
in the horizontal direction, but there is a tendency for errors to increase / to become
larger in the vertical direction. This tendency was observed in measuring Hachisu
Dam. Trend values were calculated by smoothing the measurement values with such
variation using a trend model.
For the trend values of GNSS, Fig. 9 shows the displacement vectors for the
period from the start of GNSS measurement (Feb. 15, 2020) to the end of July
2020. G1 and G2 on the crest were displaced to the upstream side, and displace-
ment of G1, with a larger cross section of dam body, was larger than G2. In addition,
G3 on the dam body foundation was displaced to the upstream side, but displace-
ment of dam body foundation is considered small in itself. The main factor in this
422
Q.106 - R.26
Fig. 8
GNSS (G1) measurement value and trend value.
Valeur de mesure et valeur tendancielle de la GNSS (G1).
1. GNSS measurement value 1. Valeur de mesure de la GNSS
2. GNSS trend value 2. Valeur tendancielle de la GNSS
3. Displacement in the upstream/ 3. Déplacement dans la direction
downstream directions (mm) amont/aval (mm)
4. + Upstream − Downstream 4. + Amont − Aval
5. Displacement in the right/left bank 5. Déplacement dans la direction rive
directions (mm) droite/gauche (mm)
6. + Left bank − Right bank 6. + Rive droite − Rive gauche
7. Perpendicular direction 7. Déplacement dans la direction
displacement (mm) perpendiculaire (mm)
8. + Uplift − Subsidence 8. + Soulèvement − Subsidence
displacement is considered to be as follows. Since control point GK1 is set to the

end of the downstream side of the concrete structure, which is long in the upstream
and downstream directions, it swells under the effect of temperature as air tem-
perature rises from the winter to summer. As a result, GK1 was displaced to the
downstream side and the measurement value showed the displacement of G3 to
the upstream side in the calculation of relative displacement by the static positioning
method. This effect of temperature on GK1 is considered to be included in the mea-
surement values of G1 and G2 on the crest. However, since this effect is removed
in the difference value of GNSS in the crest and foundation, there is no particular
problem in comparison with the plumb line measurement value showing dam body
deformation amount.
423
Q.106 - R.26
Fig. 9
Displacement vector of GNSS.
Vecteur de déplacement de la GNSS.
1. Scale of displacement vector 1. Échelle du vecteur de déplacement

2. Covered period of displacement vector 2. Période couverte par le vecteur de
déplacement
Fig. 10 provides an example of the result of analyzing the plumb line measure-
ment value and the trend value of GNSS. The figure is shown with the difference
value (G1-G3) between G1, which is set to the crest of 12BL, where the plumb line
is also installed, and G3 on the foundation of 14BL and deformation in the upstream
and downstream directions, where deformation amount is large. The Figure is also
shown with the amount of deformation based on the date of February 15, 2020 as
zero reference since GNSS measurement started on the same date at midwinter.
When the reservoir level begins to decline from April, combined with deformation
due to the rise of outside air temperature in the same period, both the plumb line
measurement values and the trend values of GNSS show the tendency of defor-
mation to the upstream side but the latter (G1-G3) is quantitatively larger than the
former.
3.5. MULTIPLE REGRESSION ANALYSIS OF DAM BODY DEFORMATION
For the measurement data from February to July 2020, it is difficult to analyze
the factors of the quantitative difference mentioned above since the measurement
period is as short as about 6 months and the reservoir level and outside air tem-
perature vary substantially in this period. Then, focusing on the fact that dam
body deformation in a concrete dam is mainly subject to the effect of reservoir
level and outside air temperature, we decided to conduct a multiple regression
424
Q.106 - R.26
Fig. 10
Hachisu Dam comparison of GNSS and plumb line measurement values.
Comparaison des valeurs de mesure du fil à plomb aux valeurs de mesure de la
GNSS pour le barrage de Hachisu.
1. Reservoir level (EL.m) 1. Niveau du réservoir (EL.m)
2. Outside air temperature (◦ C) 2. Température de l’air extérieur (◦ C)
3. Displacement in the upstream/ down- 3. éformation de la direction dans le sens
stream directions (mm) haut-bas (mm)
4. + Downstream − Upstream 4. + Aval − Amont
5. Plumb line measurement value 5. Valeur de mesure du fil à plomb
6. GNSS (G1) trend value 6. Valeur tendancielle de la GNSS (G1)
7. GNSS (G3) trend value 7. Valeur tendancielle de la GNSS (G3)
8. GNSS (G1-G3) trend value 8. Valeur tendancielle de la GNSS (G1-G3)
analysis on dam body deformation using reservoir level and outside air temperature
as parameters. The main factors for this difference were analyzed.
Measurement values of the plumb line in Hachisu Dam in the upstream/

downstream direction are found remaining gradually toward the upstream side over
time. Then, in the multiple regression analysis, measurement values of the plumb
line and data of reservoir level and outside air temperature were used for the period
of about 1,600 days from June 1, 1991, when first filling of water was finished, to
December 31, 1995, in which the behavior of dam body deformation was being
stabilized. Remaining deformation in this period is indicated as a residual of mul-
tiple regression analysis. Further, using the parameters of the multiple regression
analysis identified in this period, predictive value of the multiple regression analysis
on dam body deformation was calculated using the reservoir level and outside air
temperature in and after February 2020 as input values. Then, a factor analysis
425
Q.106 - R.26
was conducted by comparing the predictive value obtained with the plumb line and
GNSS measurement values.
Multiple regression analysis determines a partial regression coefficient (a) with

the highest correlativity, using dam body deformation as an objective variable (Y)
and reservoir level and outside air temperature as an explanatory variable (X). As
explanatory variables other than the above, solar radiation, the inside and outside
difference of dam body concrete temperature, water storage temperature distribu-
tion, etc. are considered as factors affecting dam body deformation, but they have
been limited to reservoir level and outside air temperature since their records are
available for them for the entire period and their effect is more dominant. For outside
air temperature, in consideration of the effect of heat conduction to and heat accu-
mulation in dam body, in addition to the outside air temperature on the relevant day,
a group of average outside air temperatures for a certain period that had elapsed up
to the relevant day was set as multiple explanatory variables using the stepwise pro-
cedure, so that a partial regression coefficient with high correlativity could be easily
identified. In the equation of multiple regression analysis Eq. [1], objective variable
(Y) is dam body deformation, explanatory variable (Xi) is the reservoir level of the
relevant day or a group of average outside air temperatures for a certain period,
“ai” is partial regression coefficient, and “b” is the constant term of residual. Table 3
provides a list of variables and coefficients.

11
Y= (ai.Xi) + b (1)
i=1
Table 3
Parameters of multiple regression analysis.
Partial
regression Standard
Factor Explanatory variable coefficient error
Water level X1 Water depth (m) a1 −0.9407 0.0319
X2 Square term of water depth (m2 ) a2 0.0094 0.0003
Temperature X3 Average outside air temperature of 1 day before (◦ C) a3 0.0000 0.0000
X4 Average outside air temperature of 2–5 days before (◦ C) a4 −0.0362 0.0097
X7 Average outside air temperature of 21–30 days before (◦ C) a7 0.0000 0.0000
Fig. 11 shows the result of multiple regression analysis. In the period covered,
the measurement value of plumb line and the analysis value of multiple regression
analysis are very close for changes in reservoir level and outside air temperature,
and a parameter with high correlativity was identified as shown by the coefficient of
determination (R 2 ) of 0.896.
426
Q.106 - R.26
Fig. 11
Multiple regression analysis of plumb line measurement values.
Analyse de régression multiple des valeurs de mesure au fil à plomb.
6. Multiple regression analysis value 6. Valeur de l’analyse de régression multiple
7. Residual (5–6) (mm) 7. Résiduel (5–6) (mm)
8. Period of multiple regression analysis 8. Période de l’analyse de régression multiple
Next, the predictive value of multiple regression analysis was calculated using
the parameters of Table 3, with the input values being water level and air temperature
in February 2020 and thereafter. As the result of calculation is shown in Fig. 12, the
predictive value of multiple regression analysis shows a tendency of deformation to
the downstream side until March and to the upstream side in and after April. The
measurement value of plumb line shows a tendency of deformation to the upstream
side in April and thereafter but, as compared with the predictive value of multiple
regression analysis, deformation amount and reduction tendency are small. The
trend value of GNSS (G1-G3) showed a tendency of deformation to the upstream
side in April and thereafter and the reduction tendency was similar to the predictive
value of multiple regression analysis. Note that the trend value of GNSS varied a
little in a short term in the upstream/downstream direction in June and thereafter. In
addition, no remaining dam body deformation was observed in the GNSS measure-
ment because the measurement period was as short as about 6 months.
427
Q.106 - R.26
Fig. 12
Comparison of the deformation amount of plumb line, GNSS, and multiple
regression analysis.
Comparaison de la quantité de déformation mesurée au fil à plomb, par GNSS et
par analyse de régression multiple.
5. Predictive value of multiple 5. Valeur prédictive de l’analyse de
regression analysis régression multiple
7. GNSS (G1-G3) trend value 7. Valeur tendancielle de la GNSS (G1-G3)
3.6. APPLICABILITY AND ISSUES OF GNSS IN HACHISU DAM
The applicability of the GNSS (G1-G3) trend value in Hachisu Dam was studied
by comparing it with the predictive value of multiple regression analysis using reser-
voir level and outside air temperature as parameters. It was consequently confirmed
with the GNSS (G1-G3) trend value that the tendency of dam body deformation to
the upstream side when the water level declines in April and thereafter is gener-
ally similar to the predictive value of multiple regression analysis. The difference of
deformation amount between GNSS and predictive values was caused from Febru-
ary to March, during which the water level rises, and it is necessary to see how this
difference changes in the measurement period of about one year. It is also neces-
sary to identify, from the standpoint of improving accuracy, the main cause of the
phenomenon of slight variation in GNSS trend value (G1-G3) over a short period in
the upstream/downstream direction in June and thereafter.
428
Q.106 - R.26
For the difference of deformation amount and reduction tendency in the plumb
line measurement value as a result of comparison with GNSS (G1-G3) trend value
and the predictive value of multiple regression analysis, it is necessary to see
how this difference changes, including the seasonal variation of reservoir level and
outside air temperature, in the measurement period of at least one year.
In addition, for the phenomenon of the past plumb line measurement value
remaining to the upstream side, it has not been confirmed at present in the GNSS
measurement value; it is necessary to continue GNSS measurement for several
years to see whether dam body deformation remains.
About 6 months have passed since GNSS measurement started at Hachisu

Dam and the behavior of dam body deformation has generally been confirmed in
the comparison of GNSS trend value and the predictive value of multiple regression
analysis. It is thought at present that GNSS is applicable in the measurement of dam
body deformation in Hachisu Dam.
Meanwhile, there is no particular problem in dam body deformation measure-

ment by the static positioning method. However, if it is possible to secure the fixed
point against the absolute coordinate in arrangement selection of control point GK1,
a measurement value as an absolute displacement amount including the dam sur-
rounding foundation will be obtained and it will be possible to monitor landslide, right
and left bank abutment, reservoir revetment erosion, etc. Accordingly, it would be
beneficial to study further application of the GNSS system to landslide monitoring,
etc. in Hachisu Dam.
4. CONCLUSIONS
As regards the applicability of GNSS in the measurement of dam body defor-

mation in a concrete dam, it is thought that the applicability was demonstrated in
the preceding verification of GNSS in a concrete dam and that it was also verified
at present in the GNSS measurement at Hachisu Dam.
We anticipate that the introduction of GNSS at Hachisu Dam will result in the
accumulation of GNSS measurement data and will contribute to the advancement
and rationalization of dam safety management.
REFERENCES
[1] Ministry of Land, Infrastructure, Transport and Tourism (MLIT), Chubu

Regional Development Bureau, Hachisu Dam Administration Office, Hachisu
Dam Comprehensive Inspection Data, 2020. (in Japanese).
429
Q.106 - R.26
[2] MLIT, Technical Criteria for River Works, Part of Maintenance (Part of Dam),
Apr. 2014. (in Japanese).
[3] MLIT, Water and Disaster Management Bureau, Guide to Dam Regular
Inspection, Oct. 2016. (in Japanese).
[4] MLIT, Water and Disaster Management Bureau, Dam Comprehensive Inspec-
tion Manual and Commentary, Oct. 2013. (in Japanese).
[5] Japan Institute of Construction Engineering, Revised Commentary of the Cab-

inet Order Concerning Structural Standards for River Administration Facilities,
Sankai-Do, 2000 . (in Japanese).
[6] Japan Society of Dam Engineers, Manual for Using GPS in Displacement
Measurement of an Embankment Dam, June 2014. (in Japanese).
[7] Norikazu SHIMIZU, Hiroshi ADACHI, Shuji KOYAMA, Application of a Smooth-

ing Method to Measured Displacements of Large Slopes by Using the GPS
Monitoring System, Journal of the Mining and Materials Processing Institute
of Japan ,1998. (in Japanese).
[8] Takuro MATSUDA, Hiroshi ADACHI, Yoshie NISHIMURA, Nobukazu SHIMIZU,

Applicability of the Trend Model for Smoothing Measured Displacements
by Using Global Positioning System and Methods Predicting Displacement
Behavior, Journal of Japan Society of Civil Engineers, No.715/III-60,333-343,
2002. (in Japanese).
[9] Yoshikazu YAMAGUCHI, Toshihide KOBORI, Makoto OONO, Tomoharu

IWASAKI, A Consideration on Exterior Deformation of Rockfill Dams Using
Global Positioning System (GPS), Dam Engineering, 15(2),120-136, 2005. (in
Japanese).
[10] Toshihide KOBORI, Yoshikazu YAMAGUCHI, Shinichiro NAKASHIMA,

Norikazu SHIMIZU, Continuous Displacement Monitoring of Rockfill Dam
during Earthquakes by Using Global Positioning System, Dam Engineering,
Vol.25, No.1, 2015. (in Japanese).
[11] Nobuteru SATO, Shigeki ICHIKAWA, Morimasa TSUDA, New Knowledge and
Approach concerning Deformation Behavior of Rockfill Dam Body Using GPS
Measurement, The Japanese Geotechnical Society Research Presentation
Conference, July 2020. (in Japanese).
[12] Yamaguchi University, Measurement Management Research Subcommittee of

the Japan Society of Dam Engineers , Deformation Measurement of Concrete
Dams Using GPS (I) -(IV), Dam Engineering, 30(1)-, 2020- . (in Japanese).
[13] MLIT, Chubu Regional Development Bureau, Hachisu Dam Administration

Office, Hachisu Dam Comprehensive Record of Construction, 1991. (in
Japanese).
430
Q.106 - R.27
-------
-------
BENEFITS OF INTRODUCING AN ICT-BASED STAFF SUPPORT SYSTEM FOR

DISASTER RESPONSE WORK ALONG THE LAKE BIWA SHORELINE AND
ITS EXTENDED USES (∗)
Megumi NAKASHIMA, Satoshi HARUNA, Yuji IWAMATSU, Sota UCHIDA,

Atsushi GOTO, Keisuke HATANO
Lake Biwa Development Integrated Operation & Maintenance Office
Incorporated Administrative Agency Japan Water Agency, Japan
JAPAN
SUMMARY
The Lake Biwa Development Integrated Operation & Maintenance Office

(hereinafter “Lake Biwa O&M Office”) of the Incorporated Administrative Agency
Japan Water Agency faced the challenge of “to establishing a quick, safe and reliable
operational system for all staff, regardless of job type” for the operation of drainage
pump stations for disaster response work in 2013 during the incidents such as the
total paralysis of public transport and road closures caused by flooding. Given the
small number of specialized technical staff, there was a limit to how many of the
staff can respond to pump and other equipment problems during disaster response
work. Thus, it became necessary to make it possible for “all staff to be competent
to handle faults at least at a minimum level.”
To solve these problems, the Lake Biwa O&M Office developed and introduced
two systems that utilize ICT (Information and Communication Technology).
∗ Effets de l’adoption d’un système d’appui aux agents basé sur les technologies de
l’information et de la communication pour le travail de prévention des accidents sur les

rives du lac Biwa, et utilisations supplémentaires de ce système.
DOI: 10.1201/9781003211808-122 431

Q.106 - R.27
One system is the “Drainage Pump Station Operational Support System”. It

makes it possible to operate equipment/facilities by following the image and auto-
mated audio navigation steps displayed on a tablet screen, which shows the steps
required to operate the drainage pump station in the form of a ’mission card’.
The other system is the “Fault Response Support System”. It allows minor
faults to be addressed. It uses head-mounted display (HMD) that allows the wearer
to communicate with audio and video with support staff using a PC or such from a
remote location and receive instructions from them.
The introduction of these two systems (hereinafter collectively called “Staff

Support System”) has made it possible for non-technical staff (i.e., staff not normally
involved in facilities operation/maintenance) who are unfamiliar with procedures to
quickly, safely, and reliably operate drainage pump stations at the time of actual
disaster response work. These systems have also made it possible to restore the
drainage pump station back to normal operation by an on-site operator sending
information such as images to a remote location which responds with instructions
on what actions to take.
Furthermore, it has been confirmed that the Staff Support System is not only
capable of solving problems in disaster response work, but can also be utilized to
improve efficiency in a wide variety of applications.
Although the Staff Support System was developed and introduced as a result
of the issues that arose in disaster management operations, thanks to its wide range
of applications that extend beyond disaster response work, We can expect its use
and application in a wide range of fields as a means to realize efficient and effective
operation and maintenance using a limited number of staff.
RÉSUMÉ
Le Bureau d’exploitation et de maintenance intégrées du développement du

lac Biwa (ci-après désigné le “ Bureau d’exploitation et de maintenance du lac Biwa ”)
de l’Agence japonaise de l’eau s’est trouvé face au défi d’“ établir un système opéra-
tionnel rapide, sûr et fiable, mis en œuvre par l’ensemble de son personnel sans
distinction de fonction ”. Il s’agissait d’assurer la manœuvre de ses stations de
pompage de drainage après la survenue de différents incidents – paralysie des
transports en commun, fermetures à la circulation des routes inondées, etc. –, lors
de son travail de prévention des désastres de 2013. En outre, le nombre réduit
de techniciens spécialisés limitait les possibilités de réponse en cas de défaillance
d’équipements de pompage lors de ce travail de prévention, impliquant la néces-
sité d’apporter à “ l’ensemble du personnel une capacité minimum de réponse aux
défaillances ”.
432
Q.106 - R.27
Pour résoudre ces défis, le Bureau d’exploitation et de maintenance du lac

Biwa a développé et adopté deux systèmes basés sur les TIC (technologies de
l’information et de la communication).
Le “ système d’appui opérationnel aux stations de pompages ” affiche les

procédures nécessaires à la manœuvre d’une station sur l’écran d’une tablette,
sous la forme d’une “ fiche de mission ”, et apporte à un opérateur la capacité de
manœuvrer les équipements/installations en naviguant à travers des images et des
messages vocaux automatisés.
Le “ système d’appui à la réponse aux défaillances ” permet de résoudre des

défaillances minimes au moyen d’un système d’affichage monté sur la tête (HMD),
qui donne à son porteur la capacité de communiquer par audio et vidéo avec le
personnel d’appui situé à distance, et de recevoir ses instructions, via un PC ou
autre dispositif.
L’adoption de ces deux systèmes (ci-après désignés le “ système d’appui au

personnel ”) a permis au personnel non technique, qui n’était pas familier de ces
procédures, de manœuvrer les stations de pompage de drainage avec rapidité,
sûreté et fiabilité lors d’un travail réel de prévention des désastres. En présence
d’une défaillance sur une de ces stations, ces systèmes ont aussi autorisé le rétab-
lissement de la station par un opérateur situé sur le terrain, lorsque celui-ci a transmis
des images ou autres informations à un site distant qui a alors indiqué la réponse à
apporter.
De plus, il été vérifié que le système d’appui au personnel pouvait non seule-
ment résoudre les problèmes du travail de prévention des désastres, mais aussi être
utilisé pour améliorer l’efficacité des opérations dans un vaste champ d’applications.
Bien que ce système ait été développé et adopté en réponse à des défis
rencontrés dans le travail de prévention des désastres, la grande variété de ses
applications, qui va bien au-delà de ce seul type de travail, promet des utilisations
dans un large éventail de domaines, comme moyen d’assurer des opérations de
manœuvre et de maintenance efficaces et précises avec un personnel réduit.
1. INTRODUCTION
Lake Biwa is the largest lake in Japan, with a catchment area of 3,848 km2 , a
lake surface area of 674 km2 and a total shore length of 235 km.
The Japan Water Agency implemented Lake Biwa development project from
1973 to 1992 with the aim of implementing flood control for Lake Biwa’s shoreline
and its downstream area of the Yodo River and ensuring a stable supply of urban
water. The project involved the construction of facilities scattered around Lake Biwa
(See Fig. 1), including roughly 50.4 km maintenance roads built on the embankment
433
Q.106 - R.27
Fig. 1
Locations of Facilities Developed around Lake Biwa.
Carte des installations développées autour du lac Biwa.
1 Lake Biwa Development Facilities 1 Installations développées autour du lac Biwa
2 Lake Biwa Development Integrated 2 Bureau d’exploitation et de aintenance
Operation & Maintenance Office intégrées du développement du lac Biwa
3 Kohoku Operation Office 3 Bureau d’exploitation et de maintenance du
lac Nord
4 Kosei Operation Office 4 Bureau d’exploitation et de maintenance du
lac Ouest
5 Konan Operation Office 5 Bureau d’exploitation et de maintenance du
lac Sud
6 Lake Biwa Surface area 674 km2 6 Superficie du lac Biwa 674 km2
7 Lake Biwa Shore length 235 km 7 Longueur des rives du lac Biwa 235 km
8 Lake Biwa Maintenance Road 50.4 km 8 Digue du lac et sa route de service 50.4 km
9 O&M Office 9 Bureau d’exploitation et de maintenance
10 Maintenance roads built on the 10 Digue du lac et sa route de service
embankment surrounding the lake
11 Drainage pump station 11 Station de pompage de drainage
12 Water supply station 12 Station d’approvisionnement en eau
13 Sluice gate, gates, etc. 13 Vanne murale, vanne anti-crue, etc
434
Q.106 - R.27
surrounding the lake for maintenance, 14 drainage pump stations, 4 water-supply

pump stations and a total of 158 gates/sluice gate/sluice pipe locations (hereinafter
collectively called “gates, etc.”) that comprise 263 gates.
The Lake Biwa Development Integrated Operation & Maintenance Office

(hereinafter “Lake Biwa O&M Office”) is engaged in operation, maintenance and
repair of these various facilities with a limited number of staff of only slightly more
than 40 employees, including regular employees and post-retirement employees
(hereinafter collectively called “staff”). At the time of a heavy rain, as part of the
flood management work for lakeshore areas, we are endeavoring to prevent Lake
Biwa backflow from inundating inland areas and minimize flood damage by patrolling
the facilities and operating gates, etc. and drainage pump stations
Table 1 shows a summary of personnel at the Lake Biwa O&M Office. Com-
paring job types, you can see that there is a drastically lower percentage of staff
in charge of machinery/telecommunications-related equipment (hereinafter “tech-
nical staff”) compared to the percentage of staff involved in administrative or civil
engineering work.
Table 1
Breakdown of Staff by Job Category.
Job type No. of staff Percentage
Administrative work 15 (2)* 37%
Civil engineering 19 (2) 46%
Mechanical work 5 (1) 12%
Telecommunications work 2 5%
Total 41 (5) 100%
(*) Number in parentheses is post- retirement employees and

is included in the total.
To perform disaster response work, the staff shown in Table 1 are divided into a
Disaster Response Headquarters Team (Lake Biwa Integrated O&M Office), Oper-
ation Office teams (Kohoku Operation Office, Kosei Operation Office and Konan
Operation Office) and nine field teams (See Table 2). Facility patrols and the oper-
ation of gates, etc. are carried out by field teams, while the operation of drainage
pump stations is carried out by contracted operators.
However, in disaster response work during Typhoon No.18 in 2013, the initial
deployment of contracted operators was delayed due to a total paralysis of public
transport and road closures caused by flooding at that time. Due to such a circum-
stance, staff needed to operate the pumping stations by themselves. Although the
Lake Biwa O&M Office conducts training for all staff several times a year on the
operation of drainage pump stations, the infrequency of disaster response work
makes it a big challenge “to establish a quick, safe and reliable operation for all staff,
regardless of job type” in order to operate facilities reliably in the event of an actual
emergency.
435
Q.106 - R.27
Table 2
Desaster Response Team List.
Composition Role No. of staff
Headquarters Team (Integrated O & M Office) Head of Headquarters Commander 1
Vice Head of Hdqtrs Deputy Commander 1
Admin. Team Chief Administration Team 1
Administration Team 3
Managemt Team Chief Management Team 1
Management Team 3
Kohoku Operation Office Team Chief/Sub Chief Oper. Office Team 3
No. 1 Team Field Team 3
No. 2 Team 2
No. 3 Team 3
Kosei Operation Office Team Chief/Sub Chief Oper. Office Team 2
No. 5 Team 3
Konan Operation Office Team Chief/Sub Chief Oper. Office Team 2
No. 7 Team 3
No. 8 Team 3
No. 9 Team 3
Total 41
The average duration of disaster response work in the Lake Biwa O&M Office
is around 10 days, thus the risk of pump failures and malfunctions rises due to the
prolonged operating time. Given the long distances between facilities due to them
being spread around Lake Biwa, it is unrealistic for the small number of technical
staff to deal with every fault and failure. This emphasized the need for “all staff to be
competent to handle faults at least at a minimum level.”
To solve the two problems above, the Lake Biwa O&M Office developed and
introduced two systems that took advantage of Information and Communication
Technology (ICT) [1].
2. SYSTEM OVERVIEW
The system developed by the Lake Biwa O&M Office is as described below.
2.1. DRAINAGE PUMP STATION OPERATIONAL SUPPORT SYSTEM
Until now, the operation of drainage pump stations has been performed using
paper-based operating manuals, which use minimal technical terms and a lot of
436
Q.106 - R.27
pictures and diagrams of equipment/facilities to ensure that anyone, regardless of job

type, can operate them. However, the operator needs to check numerous items for
the installed equipment and requires travel time and actual operation time combined,
all of which make it difficult for him to recognize the state of progress. Although
the training was conducted several times a year, since non-technical staff did not
normally operate the equipment, it took them a long time to get equipment up-and-
running because of the need to check procedures so many times. The paper-based
operating manuals also had the disadvantage of getting wet when it rained.
To solve this problem, the Lake Biwa O&M Office developed a “Drainage Pump
Station Operational Support System,” using Augmented Reality (AR) to facilitate
the operation of drainage pump stations, allowing the operator to use a tablet with
images and audio to navigate operating locations and operational methods as well
as checking tasks, photo records and more.
In this system, the operating procedures for the drainage pump station are
organized into scenarios and displayed on the tablet screen in the form of a ’mission
card’. The operator then carries out the work at each equipment by following the
respective mission card (See Fig. 2). AR markers are used to identify equipment to
be operated. The system is designed to prevent a human error by not allowing the
operator to proceed to the next step even if they read an AR marker of a different
equipment. The operator selects the appropriate option for each operation by clicking
“Yes” or “No” on the tablet screen. If any abnormality or problem occurs in the field,
the operator can also perform repair work by receiving instructions from a technical
staff through the Fault Response Support System which will be described later.
After repair work is completed, the drainage pump station can be operated again by
following the instructions on the tablet screen and continuing to carry out the work.
This system makes it possible for even a new hire or recently transferred staff
who is unfamiliar with procedures to operate a drainage pump station safely and
reliably without confusion over operating steps or worrying about mistakes, by simply
following the on-screen instructions. After operating, the result of the operator’s
work is uploaded to a cloud server and an operation record is automatically created.
This greatly reduces the amount of time and an effort operator needs to spend on
the computer after returning to the office, which leads to better work efficiency. Its
navigation abilities, which assist in operation, can also be used offline by being
downloaded beforehand.
At the Lake Biwa O&M Office, not only technical staff but also administrative
and civil engineering staff with no specialized knowledge of the equipment were
involved in the development of this system from the planning stage. The system was
designed to be easy to visually intuitive; for example, by using just one mission card
for each operation with photos and diagrams of the site to ensure that the system
can be easily operated by anyone. For the tablet, a dust-, drip-, and impact-resistant
model was selected in view of its anticipated use outdoors and in bad weather.
437
Q.106 - R.27
Fig. 2
Drainage Pump Station Operational Support System.
Système d’appui opérationnel aux stations de pompage de drainage.
1 a mission card 1 fiche de mission
2 instructions message 2 instructions
2.2. FAULT RESPONSE SUPPORT SYSTEM
The Drainage Pump Station Operational Support System, which uses AR, is
highly effective for tasks that can be predicted since the necessary procedures are
prepared in advance in the form of scenario.
438
Q.106 - R.27
However, for unexpected equipment failures and malfunctions during disaster

response work, there was no alternative to the conventional method of relying on
specialized technical staff. But since drainage pump stations and other facilities are
scattered around Lake Biwa, it not only takes time to fix problems, but also has the
problem of travelling to the locations. Therefore, “Fault Response Support System”
was decided to create.
Using the internet, this system allows support staff (the person giving instruc-
tions to the operator) to understand the situation at the failure site in real time through
images sent by the operator (person on site requiring instruction) and by commu-
nicating bidirectionally. A head-mounted display (hereinafter “HMD”), which can be
attached to the operator’s helmet, is equipped with a camera that captures what
operators see, a display (with a built-in microphone) mounted in front of the opera-
tor’s line-of-sight, and an earphone (See Fig. 3). This system has made it possible
to provide failure details and other information from a remote location, which pre-
viously required the field dispatch of technical staff with specialized knowledge or
experienced engineers from manufacturers. This system allows up to six people to
communicate with each other at the same time by simply logging into the dedicated
Fig. 3
Wearing a HMD.
Port de l’affichage monté sur la tête (HMD).
1 Camera 1 Caméra
2 Display unit 2 Écran
3 Wearable key board 3 Clavier portable
4 Earphone 4 Écouteur
439
Q.106 - R.27
website from their computers. Also, since the HMD enables hands-free recovery
work, it is much safer than using a video feed from a smartphone or such.
As described above, this system shares information on malfunctions and

defects and gives repair instructions via video and voice prompts, but assuming
situations where the operator cannot hear the voice of support staff?for example,
due to loud noise from operating pumps, it is also able to convey Text-based instruc-
tions (See Fig. 4) to an operator using the wearable keyboard. Support staff can
also provide hand-written instructions by hand-writing remarks on still images taken
by the HMD camera and sent them over the communication network (See Fig. 5). In
addition, the HMD also has a feature that allows support staff to remotely operate the
HMD and take images with the HMD camera. The system also allows support staff
to send drawings or other materials to the HMD. By giving support staff the ability to
send necessary materials to the HMD when the operator does not have them, this
improves reliability in both identifying the cause of failure as well as responding to
the failure.
Fig. 4
Text-based instructions (operator display screen).
Instructions de type texte (écran d’affichage de l’opérateur).
1 Instructions from the support staff 1 Instructions du personnel d’appui
2 Lift the breaker lever in the red frame 2 Relevez le levier du disjoncteur à l’intérieur
du cadre rouge
As with the tablet, the HMD selected is a dust- and drip-resistant model suitable
for use outdoors and in bad weather.
440
Q.106 - R.27
Fig. 5
Manually-written instructions (operator display screen).
Instructions annotées à la main (écran d’affichage de l’opérateur).
1 Lift the lever 1 Relever le levier
3. OPERATIONAL STATUS OF THE STAFF SUPPORT SYSTEM AND ITS

EFFECTS
These two systems were developed in two years since 2015. Collectively called
the “Staff Support System” by the Lake Biwa O&M Office, the systems went into full
operation in April 2017.
Status of the Staff Support System and the effects of its introduction are
described below.
3.1. DRAINAGE PUMP STATION OPERATIONAL SUPPORT SYSTEM
The Drainage Pump Station Operational Support System detected a breaker

that had not been activated at a drainage pump station during emergency operations
in 2017, making it possible to operate the drainage pump station safely and reliably.
At another drainage pump station, the system made it possible for a non-technical
staff unfamiliar with the facility to quickly, safely, and reliably operate the station (See
Fig. 6).
441
Q.106 - R.27
Fig. 6
Administrative staff is operating equipment.
Administrative staff is operating equipment.
A comparative test on operating time and report writing time was conducted
to compare the difference between using the conventional paper-based operating
manual and this system when used by administrative staff who have never operated a
drainage pump station. Results showed that using this system reduced the working-
time to get pumps running by 31% and reduced report writing time by 89%.
3.2. FAULT RESPONSE SUPPORT SYSTEM
When a large amount of driftwood accumulated in front of the suction tank

screen of the drainage pump station at the time of disaster response work in 2017,
the accumulation problem was resolved by providing proper instructions remotely
without dispatching technical experts to the site, and without interrupting operation
of the drainage pump station (See Fig. 7 8).
This system allowed the Disaster Response Headquarters of the Lake Biwa
O&M Office as well as the team leaders of each operation office, etc. to acknowledge
inundation and flooding conditions at the site and give proper instructions. This
process led to not only reducing the time normally needed to travel to the site but
also the risk of possible accidents during the travelling time.
442
Q.106 - R.27
Fig. 7
Driftwood accumulation status (HMD camera image).
Réponse à l’accumulation de bois dérivant.
Fig. 8
The staff at the Disaster Response Headquarters is giving instructions,
using the system.
Un opérateur au Centre de réponse aux catastrophes communique
au moyen du système.
443
Q.106 - R.27
4. EXTENDED USES
The Staff Support System was designed to support the operation and repair
of drainage pump stations when disaster response work was being carried out.
However, in the past 28 years since the Lake Biwa O&M Office’s disaster man-
agement activities began in April 1992, the system has only been used nine times
(as of March 31, 2020), leading to our concern over whether the staff’s proficiency
level concerning the Staff Support System could well be maintained. Therefore, it
was decided to establish a Staff Support System Applications Team (hereinafter
“Applications Team”) in the Lake Biwa O&M Office, which played a leading role in
system operation training, and also studied various extended uses for the system.
As a result, it was decided to use the Fault Response Support System for as many
tasks as possible with an emphasis on its application on construction management
field. Additional operation scenarios were for the Drainage Pump Station Operational
Support System, and the scenarios were modified as facilities were updated.
4.1. CONSTRUCTION WORK
The Lake Biwa O&M Office has been without staff in charge of construction
(hereinafter “construction work staff”) since April 2017. Whenever construction work
is required, construction work staff from the regional headquarters are dispatched
and handles the estimation and supervision of work. While the General Affairs Divi-
sion of the Lake Biwa O&M Office usually oversees simple repairs to the lodging
facility for the staff, the large-scale renovation work ordered in September 2017
required expert knowledge of materials as well as multi-stage check-ups. When the
work was ordered, it was assumed that a construction work staff would visit the sites
to supervise the work, however, considering the time required to travel to the sites
and check-up work, it was decided to use the Fault Response Support System on
a trial basis.
In the cases where construction required checking by construction work staff

or they were requested to be present by the work orderer for supervision, the staff
from the General Affairs Division of the Lake Biwa O&M Office, wore an HMD and
went to the site so that the regional headquarters could grasp on-site conditions
and check work, rather than having construction work staff be dispatched to sites.
By using the Fault Response Support System for field confirmation when someone
was required to be present on site, it became possible to not only avoid having the
contractor return for rework, but also receive instructions remotely by adding written
instructions on the drawings and sending the information to the sites so that decision-
making on matters that were difficult to judge by anyone other than a construction
work staff became feasible. (See Fig. 9).
The result of the trial work led to the frequent use of the system, namely 8
times out of 12 field confirmations and on-site inspections in this project, which has
444
Q.106 - R.27
Fig. 9
Written instructions from a remote location.
Instructions fournies par annotation d’une image à distance.
1 Instructions (information to the sites) 1 Instructions (informations sur les sites)
turned out to be a success in improving work efficiency in terms of the travel time
and labor cost reduction of construction work staff.
Like the Lake Biwa O&M Office, the same situation applies to other offices.
Due to the long travel time to get from one place to another, this system is considered
effective to use in construction work that requires the presence of the construction
work staff.
4.2. MULTI-STAGE CHECK-UPS AT THE FACTORY
When equipment updating projects involve factory-built mechanical or elec-

trical components, multi-stage check-ups are performed at the factory for critical
equipment and for contracts made under low cost bids and quality assurance is not
certain. Previously several supervisors visited such factories to conduct inspections,
but as a trial, the Fault Response Support System was used to perform the inspec-
tion work with just one supervisor wearing an HMD to visit such a factory while the
rest of the supervisors checked the inspections from a remote location.
In 2017, multi-stage check-ups were conducted for the tentative assembly of

opening/closing equipment for the gates. As a result, it was proved that adequate
445
Q.106 - R.27
inspections at the same level as the previous method was possible, checking the
value readings of measurement devices through images from the camera. (See
Fig. 10).
Fig. 10
Reading texts/figures using an HMD in a factory inspection.
Lecture de texte avec l’affichage monté sur la tête (HMD)
pendant l’inspection d’une usine.
4.3. SAFETY PATROL
As one of the ways to utilize the system to manage safety at construction sites,
the safety council of the Lake Biwa O&M Office used the Fault Response Support
System to patrol sites. HMD-wearing staff went to the site, and a safety council
member acting as support staff at the operation office pointed out unsafe contracted
operator practices by using images from the HMD.
Despite the patrol being conducted by a small number of people, this system
was highly efficient as it enabled many eyes to check the site and point out safety
management concerns of the contracted operators. It is anticipated that such patrols
can also serve as an educational tool for safety management for younger staff by
having them wear an HMD so that support staff can review and point out any unsafe
practices.
446
Q.106 - R.27
4.4. CHECKING OF INCIDENTS AND ACCIDENTS
As the Lake Biwa O&M Office manages a large area surrounding Lake Biwa, it
works closely with jurisdictional police and fire authorities, etc., so that they agreed to
inform the office in the event of a traffic accident, fire or other emergency (hereinafter
collectively called “emergencies”) taking place near a managed facility. Once such a
case is informed, staff of the responsible operation office are dispatched to the site,
but in many cases, it is not clear from the information received from such authorities
as the police or fire department whether the emergency falls under our administrative
scope. Normally, staff made it a practice of physically bringing related documents
related to the site of the emergency and check the situation at the site. However,
with the Fault Response Support System, staff patrolling the area went directly to
the site and communicated with the operation office, with using HMD. Thus, staff
at the operation office understood the situation, specific location of the site, etc.,
by using video and audio information from the HMD. Staff in the operation office,
away from the site, were also able to check documents to determine whether the site
was under its management and provide the staff on site with instructions on how to
respond to the situation. In addition, still photos were extracted from the HMD video
footage recorded by the staff on site, allowing a preliminary report to be prepared in
a highly efficient process
5. CONCLUSIONS
The Lake Biwa O&M Office has originally developed and introduced a Staff
Support System to solve the problems encountered in carrying out disaster response
work in the past. This system has enabled “to establish a quick, safe and reliable
operation of drainage pump stations by all staff regardless of job type” and made it
possible for “all staff to be competent to handle faults at least at a minimum level. The
Drainage Pump Station Operational Support System provides staff who are unfa-
miliar with the operation of facilities and equipment with the support which includes
prepared scenarios for such operation. After the system start, the opening and
closing of gates, etc. are also supported with prepared scenarios. Additionally, by
incorporating scenarios such as the operating parameters of each pump and oper-
ating conditions for gates, etc., this system also supports the operational reliability of
personnel in a way that reflects the unique characteristics of each facility/equipment.
One of the key features of the Fault Response Support System is that an HMD cam-
era is mounted on a helmet to allow the operator on site to work hands-free while
receiving instructions from the support staff in the remote office. The operator’s
display shows what they are seeing, and the support staff can also view images
while giving instructions. This system differs from other systems in that it allows
both parties to check and respond with each other in real time, making it dramati-
cally more effective. In addition, the system differs from other similar systems in that
it not only supports the sharing of typed instructions, photos, drawings and such,
447
Q.106 - R.27
but also allows support staff to send instructions with hand-writtenletters on the
images. Yet another benefit is that in today’s world of frequent and extreme weather
events, earthquakes, storms, floods, and other natural disasters, the use of this
Fault Response Support System enables real-time assessment and confirmation
of disaster conditions through video and audio communications between the site
and office, or even between other locations such as a regional office and the head
office. Therefore, this system has many advantages and is expected to have wide
application potential in terms of manpower, cost and efficiency for managing dams
or storage reservoirs.
REFERENCES
[1] UCHIDA S., IWAMATSU Y. Development of an ICT based drainage pump sta-
tion operational support and fault response support system. National Research
Group for Land Technology; Ministry of Land, Infrastructure and Transport,
2017. (in Japanese).
448
Q.106 - R.28
-------
-------
SEISMIC PERFORMANCE VERIFICATION OF A ROCKFILL DAM AGAINST

LARGE DOUBLET EARTHQUAKES (∗)
Nario YASUDA
Head of Eng. Division 1,
JAPAN DAM ENGINEERING CENTER
Zengyan CAO
Senior Engineer,
J-POWER BUSINESS SERVICE CORPORATION,
JAPAN
SUMMARY
The purpose of this study is to clarify the behavior of a rockfill dam during large
doublet earthquakes and to propose a procedure for verifying the seismic perfor-
mance of rockfill dams during such earthquakes. Impacted by a strong earthquake,
the dynamic characteristics of the Aratozawa Dam significantly changed temporarily.
The behavior of the dam subject to another strong earthquake before its dynamic
characteristics had recovered was predicted. Based on the analysis results, the seis-
mic performance of the dam was evaluated. The results clarified that the change in
the dynamic characteristics of the dam in response to the first event greatly affected
the behavior of the dam during the second event. In addition, a method was pro-
posed for predicting the final residual subsidence of the dam by summing the sliding
displacement caused by each earthquake and the subsidence related to the shaking
and rearrangement of soil particles. It is recommended that after a dam experiences
a strong earthquake, the reservoir is adjusted to a safe water level and maintained
at that level for a minimum of one week as an emergency response measure.
∗ Vérification des performances sismiques d’un barrage en enrochement contre des doublets
de séismes
DOI: 10.1201/9781003211808-123 449

Q.106 - R.28
RÉSUMÉ
Le but de ce rapport est de clarifier le comportement d’un barrage en enroche-

ment lors de doublet de grands tremblements de terre et de proposer une procédure
de vérification des performances sismiques des barrages en enrochement lors de
tels séismes. Impactées par le tremblement de terre d’Iwate-Miyagi Nairiku en 2008,
les caractéristiques dynamiques du barrage d’Aratozawa ont changé temporaire-
ment de manière significative. Le comportement du barrage soumis à un autre
puissant tremblement de terre avant que ses caractéristiques dynamiques se soient
rétablies a été prévu. Sur la base des résultats de l’analyse, la performance sismique
du barrage a été évaluée. Les résultats expliquent que le changement des carac-
téristiques dynamiques du barrage en réponse au premier événement a grandement
affecté le comportement dynamique du barrage lors du deuxième événement. De
plus, une méthode a été proposée afin de prévoir l’affaissement résiduel final du
barrage en additionnant le déplacement glissant provoqué par chaque séisme et
l’affaissement lié aux secousses et au réarrangement des particules du sol. Il est
recommandé, après qu’un barrage ait subi un fort tremblement de terre, que le
réservoir soit ajusté à un niveau d’eau sécurisé et maintenu à ce niveau pendant au
moins une semaine comme mesure d’intervention d’urgence.
1. INTRODUCTION
In Japan, seismic performance verification of existing dams against large

earthquakes is being carried out via dynamic response analysis. In the dynamic
response analysis of these dams, engineers use the dam’s normal operating condi-
tions as the initial conditions for the analyses and suppose the dam subject to a single
earthquake. However, after the Iwate-Miyagi Nairiku Earthquake in 2008 (Mj 7.2,
where Mj is the Japan Meteorological Agency (JMA) magnitude scale), the dynamic
characteristics of the Aratozawa Dam, a 74.4 m high rockfill dam, considerably dete-
riorated [1][2]. Approximately one week after the main shock, the natural frequencies
of the dam were found to have nearly recovered. On the other hand, during the 2016
Kumamoto earthquake of April 14 (21:26, Japan time) and 16 (01:25), (Mj 6.5 and
Mj 7.3), Mashiki Town experienced ground motions of JMA seismic intensity 7 twice
within 28 hours [3]. Generally, large aftershocks often occur after a strong inland
earthquake. In the future, a dam may experience multiple large seismic events in a
short period. Thus, it is necessary and beneficial to dam engineering to investigate
the behavior of the Aratozawa Dam subject to a ground motion of JMA seismic inten-
sity 7 before its dynamic characteristics fully recover from an earlier earthquake. The
situation mentioned here has not been addressed in the current “Guidelines of the
Seismic Performance Verification of Dams for Large-scale Earthquakes (draft)” [4].
Thus, this study clarifies the behavior of rockfill dams subject to large doublet earth-
quakes and proposes a seismic performance verification method and emergency
450
Q.106 - R.28
response measures in preparation for the occurrence of large doublet earthquakes

in the future.
In this study, the authors modeled the response of the Aratozawa Dam to
two large earthquakes. The dynamic response analyses of the dam are performed
considering the variation in the dynamic characteristics of the dam during each
earthquake. The calculations of slip deformation and the deformation related to
the shaking and rearrangement of soil particles are carried out considering the
change in the material properties of the dam. The final residual deformation of
the dam after the doublet earthquakes is predicted by summing the deforma-
tion induced by each earthquake, and the seismic safety of the dam is verified
via the residual deformation. In addition, in preparation for large doublet earth-
quakes, emergency response measures after large earthquakes are investigated for
rockfill dams.
2. DAM SPECIFICATIONS AND EARTHQUAKE MONITORING
2.1. DAM SPECIFICATIONS
The Aratozawa Dam is a 74.4 m high rockfill dam with a central clay core on the
Kitakami River, Kurihara City, Miyagi Prefecture, and the dam body was completed
in 1992. Table 1 shows the specifications of the dam. The location, the plan view and
the standard cross-section of the dam and the locations of seismographs are shown
in Fig. 1. The seismographs with three components (parallel and perpendicular to
the dam axis and vertical directions) are installed at the dam base (F), mid-height
of the core (M) and dam crest (T).
Table 1
Specifications of the Aratozawa Dam.
Dam location Kurihara City, Kurihara district, Miyagi Prefecture
Dam type Rockfill dam with central clay core
Dam height 74.4 m (minimum ground elevation, EL. 205.0 m)
Crest length 413.7 m
Crest width 10.0 m
Crest elevation EL. 279.4 m
Nonoverflow section elevation EL. 278.9 m
Slope gradient Upstream: 1: 2.7 downstream: 1: 2.1
Embankment volume 3048000 m3
Basin area 20.4 km2
Reservoir storage volume 13850000 m3
Design seismic coefficient 0.15 (dam body), 0.18 (intake tower), 0.16 (spillway)
451
Q.106 - R.28
Fig. 1
Location of the dam and the locations of seismographs.
Localisation du barrage et emplacements des sismographes.
2.2. NONLINEAR PROPERTIES OF EMBANKMENT MATERIALS EVALUATED FROM

EARTHQUAKE RECORDS
A large number of earthquakes since 1992, immediately after the dam body
completion, have been recorded. Of these, there were approximately 500 earth-
quakes with a maximum acceleration over 10 cm/s2 and 37 earthquakes over
100 cm/s2 at the dam base (F). On June 14, 2008, the Iwate-Miyagi Nairiku Earth-
quake in 2008 (Mj 7.2) occurred. At the Aratozawa Dam located approximately
15 km from the epicenter, the maximum acceleration of 1024 cm/s2 in the direc-
tion perpendicular to the dam axis was recorded at the dam base. This is the largest
earthquake magnitude ever recorded at the base of a dam in Japan. During the earth-
quake, the natural frequency and acceleration amplification factor of the dam sharply
decreased. Approximately 1 week after the main shock, the natural frequencies of
the dam nearly fully recovered to their preearthquake values.
The decrease in the natural frequency of the dam implies a reduction in the
stiffness of the dam materials. In a single degree of freedom system, the relationship
of the natural circular frequency ω, the stiffness k and the mass m are expressed as
Eq. [1].
k
ω2 = (1)
m
The relationship between the three physical quantities in Eq. [1] is also gener-
ally satisfied in civil engineering structures such as rockfill dams. Because the mass
of the dam body does not change, considering that the stiffness is proportional to
the shear modulus G of the dam material through Poisson’s ratio, the relationship
between the shear modulus G and the fundamental frequency is shown in Eq. [2].
f12 G
2
= (2)
f10 G0
452
Q.106 - R.28
Here, and are the initial shear modulus of the material in the small strain status
and the initial fundamental frequency of the dam, respectively. Therefore, the rate of
change in the shear modulus of the materials can be obtained from the rate of change
in the fundamental frequency of the dam using Eq. [2] [5]. In addition, the transfer
function of the dam can be obtained from the earthquake records at the crest of the
dam (T) and the dam base (F), and the frequency corresponding to the first peak
of the transfer function can be regarded as the fundamental frequency of the dam.
From the analysis of the records of the past small earthquakes, the fundamental
frequency of the Aratozawa Dam has been found to be 3.2 Hz [2]. To determine the
fundamental frequency of the dam for each earthquake, the earthquake records of
relatively large acceleration amplitude are extracted, focusing on the Iwate-Miyagi
Nairiku Earthquake in 2008 and its aftershocks. The list of the extracted earthquake
records is shown in Table 2. These earthquake records (Nos. 1-19) are numbered
in the order of occurrence.
Table 2
Earthquake records and the obtained results.
Focal Max. acc. at the Fundamental Average
Time of depth dam base1 freq. of the shear strain
No. occurrence M (km) (cm/s2 ) dam (Hz) of dam body
1 1996/08/11 03:12 6.1 9 28 2.930 5.320E-05
2 1996/08/11 08:10 5.8 10 36 2.830 3.920E-05
3 1996/08/11 15:01 4.9 10 30 3.027 6.600E-06
4 2003/05/26 18:24 7.1 72 114 2.637 1.727E-04
5 2008/06/14 08:43 7.2 8 1024 1.514 1.566E-03
6 2008/06/14 09:00 4.2 11 99 2.344 3.700E-06
7 2008/06/14 09:01 4.0 7 482 1.953 7.760E-05
8 2008/06/14 09:14 3.6 4 151 2.246 1.660E-05
9 2008/06/14 09:20 5.7 7 76 2.051 7.980E-05
10 2008/06/14 10:40 4.8 7 120 2.148 5.960E-05
11 2008/06/14 12:09 4.1 8 92 2.246 2.840E-05
12 2008/06/14 12:10 4.8 9 79 2.344 2.480E-05
13 2008/06/14 19:11 4.1 8 229 2.148 5.170E-05
14 2008/06/16 23:14 5.3 7 76 2.246 3.760E-05
15 2008/07/24 00:26 6.8 108 27 2.600 6.530E-05
16 2008/09/25 15:04 4.1 6 119 2.539 7.000E-05
17 2011/03/11 14:46 9.0 24 102 2.307 1.858E-04
18 2011/04/07 23:32 7.2 66 120 2.344 2.405E-04
19 2015/05/1306:13 6.8 46 18 2.942 1.350E-05
Note: 1 The maximum value of the three directional components of the dam base (F).
On the other hand, using the earthquake records of the dam, it is possible to
calculate the average shear strain of the dam according to the procedure shown in
Fig. 2.
From each earthquake record, the fundamental frequency of the dam in the
direction perpendicular to the dam axis and the average shear strain are calculated
and summarized in Table 2. The value of G/G0 is calculated according to Eq. [2]
and plotted with the average shear strain γ in Fig 3. Depending on the G/G0 -γ
453
Q.106 - R.28
Fig. 2
Procedure for calculating the shear strain of the dam from the recordings.
Procédure de calcul de la déformation de cisaillement du barrage à partir des
enregistrements.
Fig. 3
G/G0 -γ relationship obtained from the records of relatively strong earthquakes.
G/G0 -γ relation obtenue à partir des enregistrements de tremblements de terre
relativement forts.
relationship and the occurrence time of the earthquakes, this seismicity can be
divided into the following 3 periods.
Period I. Earthquake 1 to earthquake 5

Period II. Earthquake 6 to earthquake 14
Period III. Earthquake 15 to earthquake 19
The G/G0 -γ relationship in Period I can be approximated by a curve and is

clearly different from that in Period II. In response to the Iwate-Miyagi Nairiku Earth-
quake in 2008 (No. 5), it is clear that the material properties of the Aratozawa
Dam had significantly changed temporarily. Thus, during the earthquake response
analysis and seismic performance evaluation of the dam in Period II, the dynamic
454
Q.106 - R.28
characteristics and material properties of the dam during this period must be con-
sidered. In addition, the G/G0 -γ relationship in Period III can be estimated to have
nearly recovered to the relationship exhibited in Period I. The solid line showing the
G/G0 -γ relationship of the dam in Period I and Period III is very similar to that of
ordinary soil materials. The shear strain corresponding to G/G0 = 0.5 is 2.9 × 10−4 ,
which substantially coincides with the reference shear strain (3.0 × 10−4 ) identified
by a reproduction analysis of a past earthquake of the Aratozawa Dam [6]. In Period
II, there are no data reflecting a shear strain of 10−4 or more, but when the existing
data are fitted with a hyperbolic line, the relationship indicated by the broken line is
obtained.
3. CONSIDERED EVENTS AND STUDY METHODS
3.1. CONSIDERED EVENTS
In this study, it is assumed that the Aratozawa Dam will subject to a doublet
earthquake similar to the 2016 Kumamoto Earthquake. The foreshock (Mj 6.5) of
the 2016 Kumamoto Earthquake occurred on April 14, and the main shock (Mj 7.3)
occurred on April 16 of the same year. Mashiki Town, Kumamoto Prefecture, experi-
enced the ground motion of JMA seismic intensity 7 (the maximum seismic intensity
ever observed in Japan) twice within 28 hours. In this study, the terms “Earthquake
A” and “Earthquake B” are used instead of the terms “foreshock” and “main shock”,
respectively. For the Aratozawa Dam, the Iwate-Miyagi Nairiku Earthquake in 2008
is considered Earthquake A in the following analysis. Before the dynamic charac-
teristics of the dam recover, an earthquake of Mj 6.5 is modeled to occur directly
underneath the dam (Earthquake B).
The acceleration records obtained at the dam base will be used directly as the
input ground motions of Earthquake A. For Earthquake B, artificial seismic waves
will be created with the earthquake records obtained at the dam base during the
aftershocks immediately after Earthquake A. The ground motions at the dam base
for Earthquake B are created by fitting the earthquake records of an aftershock
equivalent to an Mj 6.5 earthquake to the lower limit spectra defined in the verification
guidelines of dams during large earthquakes [4].
3.2. STUDY METHODS
The earthquake response analysis of the dam will be performed with an equiv-
alent linearization method, which is a commonly available analysis method, instead
of nonlinear analysis. The main influencing factors for the seismic behavior and seis-
mic safety verification of dams include the ground motion, material properties such
455
Q.106 - R.28
as strength and elastic modulus, and reservoir conditions. If the material properties
of the dam have changed due to Earthquake A, the variation should be reflected
in the response analysis of the dam to Earthquake B. As shown in Fig. 3, the ini-
tial shear modulus and the G/G0 -γ relationship in Period II are clearly different from
those of the other periods. Therefore, to predict the earthquake responses of the dam
to Earthquake B within approximately one week after Earthquake A, the initial shear
modulus G0 and the shear strain dependence of the materials during the period are
considered. The residual deformation of the dam due to the doublet earthquake is
Fig. 4
Verification of the seismic performance of rockfill dams for doublet earthquakes.
Vérification de la performance sismique des barrages en enrochement pour les
tremblements de terre “doublet”.
456
Q.106 - R.28
predicted by summing the deformations calculated separately for each earthquake.

The deformation includes the sliding component and the subsidence associated
with soil particle oscillation. Figure 4 shows the earthquake response analysis and
the seismic performance verification procedure for rockfill dams subject to a large
doublet earthquake used in this study.
4. RESPONSE OF THE DAM TO EARTHQUAKE A
4.1. REPRODUCTION ANALYSIS
Figure 5 shows the 3D analysis model of the dam-reservoir-foundation sys-

tem. Considering the influence of the vibration of the free fields outside the analysis
region, viscous boundary conditions are used at the perimeter and bottom bound-
aries of the foundation rock and the upstream end of the reservoir model. In the static
stress analysis before the earthquake, the influence of the reservoir water on the
water level (EL 270.1 m) during the earthquake and that of the pore water pressure
inside the dam are considered. However, the influence of the hydrodynamic pres-
sure is neglected in the earthquake response analysis since the interaction between
embankment material and reservoir water is not yet fully understood.
Fig. 5
Analysis model.
Modèle d’analyse.
The equivalent linearization method is used in the analysis to consider the

influence of nonlinear material properties. The initial shear moduli of the embank-
ment materials are set according to Sawada’s equation [7], which is based on the
457
Q.106 - R.28
compressional and shear wave logging (P-S logging) of a number of rockfill dams. To
improve the accuracy of the initial shear moduli, they are adjusted slightly so that the
fundamental frequency of the dam obtained from eigenvalue analysis is consistent
with the result estimated for past small earthquakes. The approximate curve in the
normal status shown in Fig. 3 is used for considering the shear strain dependence
of the shear moduli.
The damping is evaluated by Eq. [3] and considered to be classical Rayleigh

damping.
[C]e = α[M ]e + β[K]e (3)
where [C], [M], and [K] are the damping matrix, mass matrix and stiffness matrix,
respectively. Superscript e indicates that these matrices belong to each element. α
and β are the parameters defined by Eq. [4].
α = 1.4hω1 β = 0.6h/ω1 (4)
where ω1 is the fundamental angular frequency of the dam; and the damping ratio
h is evaluated by Eq. [5], which considers its shear strain dependence.
γ
h = hmax + h0 (5)
γ + γr
where h, h0 and hmax are the damping ratio, its initial value and its maximum value,
respectively. The values of h0 and hmax are listed in Table 3 and Fig. 6 are identi-
fied by the reproduction analysis of the dynamic behavior of the dam during past
earthquakes [6]. γ is the effective shear strain (2/3 of its maximum value). γr is
the reference shear strain, which can be found from Fig. 3 to be approximately
2.9 × 10−4 when G/G0 = 0.5. The Poisson’s ratios of the embankment materials are
set according to Sawada’s equation [7]. The densities of the embankment materials,
foundation rock and spillway concrete from construction quality control tests of the
dam are listed in Table 4.
Table 3
Damping factors identified by the reproduction analysis of dynamic
behavior of the dam during past earthquakes.
Classification Maximum damping ratio hmax Initial damping ratio h0
A: Lower part of the core 20% 5%
B: Upper part of the core 30%
C: Filter: Transition 30%
E: Rock (inner): Rock (outer) 20%
Spillway concrete is treated as a linear material with an elastic modulus of

30000 N/mm2 and a Poisson’s ratio of 0.2. The foundation rock is also modeled as
458
Q.106 - R.28
Fig. 6
Zoning of the properties (refer to Table 3).
Zonage des propriétés (se reporter au Tableau 3).
Table 4
Density of the materials used in the analysis.
Density (g/cm3 )
Classification Wet Saturated
Core 2.04 2.10
Filter 2.34 2.43
Transition 2.24 2.33
Upstream rock Inner 2.15 2.29
Outer 2.15 2.32
Downstream rock Inner 2.18 2.32
Outer 2.13 2.30
Spillway concrete 2.40
Bedrock 2.60
a linear material. Its shear modulus is 5500 N/mm2 , which is obtained by converting
from the average shear wave velocity of the rock (Vs = 1440 m/s). The Poisson’s
ratio of the foundation rock is 0.25, which is estimated to be the average value of
the rock.
As for the input ground motions, since Earthquake A is a real earthquake, the
earthquake records (shown by the solid line in Fig. 7(c)) obtained at the dam base
(F) during the Iwate-Miyagi Nairiku Earthquake in 2008 are used. The input ground
motions at the bottom boundary of the model (as shown in Fig. 5) are created by
inverse analysis of the earthquake records at the dam base.
Figure 7 shows the acceleration response histories of the earthquake monitor-

ing stations. Although there is some difference in the maximum acceleration values
between the analysis results and earthquake records, the response waveforms are
generally coincident. Good reproducibility is obtained in the acceleration response
histories. The waveforms from 0 sec to 4 sec in the acceleration response histories
of the crest have approximately the same phase. In contrast, in the waveforms after
4 sec, the peak values of the earthquake records appear slightly later than those
of the analysis results. This is thought to be due to the fact that in the equivalent
459
Q.106 - R.28
Fig. 7
Acceleration response of Earthquake A.
Réponse d’accélération du tremblement de terre A.
460
Q.106 - R.28
linearization method, the stiffness of the embankment materials is a constant value

for the whole history, but in the actual dam, the rigidity of the dam decreased after
the principal motion of the earthquake, which resulted in a decelerating wave prop-
agation. Figure 8 shows the comparison of the Fourier spectra and the transfer
Fig. 8
Fourier spectra and transfer function of the acceleration response in the direction
perpendicular to the dam axis.
Spectre de Fourier et fonction de transfert de la réponse d’accélération
perpendiculairement à l’axe du barrage.
461
Q.106 - R.28
functions of the acceleration responses at the dam base and crest. In the frequency
range up to 2 Hz, the Fourier spectra and transfer functions between the earthquake
records and analysis results are similar. In the higher frequency range, the deviation
between the analysis results and earthquake records gradually increases as the
frequency increases. This finding may also be due to the continuous change in the
physical properties of the real dam, while the physical properties in the analysis
are treated as constant values. As a result, the physical properties suitable for the
principle motion, which mainly include the low-frequency components, lead to the
differences in the high-frequency seismic responses in the analysis.
4.2. RESIDUAL DEFORMATION CAUSED BY EARTHQUAKE A
The seismic performance of rockfill dams is evaluated via residual deformation

under the current guidelines [4]. In the previous section, the seismic behavior of
the Aratozawa Dam in response to Earthquake A was reproduced by numerical
analysis. Using the analysis results, the residual deformation of the dam caused
by Earthquake A is calculated. The residual deformation of the dam caused by
the earthquake will be calculated separately via two methods according to the two
deformation mechanisms, sliding and subsidence.
First, the deformation due to sliding is calculated. Here, the method proposed
by Watanabe and Baba [8] is used, which is based on the sliding block assumption
and the stress results of the earthquake response analysis. A 2D calculation is
performed with the standard cross-section extracted from the 3D model shown in
Fig. 5. In total, 1235 arcs are set up, covering all the possible sliding zones. To fully
investigate the stability of the dam, the arcs sliding toward the upstream side are
divided into 3 groups. The arcs for which both ends pass through only the upstream
surface of the dam are referred to as the “U1 Group”. Those with their upper end at
the crest and lower end at the upstream surface are categorized into the “U2 Group”.
In addition, the arcs that cut the core zone are referred to as the “U3 Group”. Similarly,
the arcs sliding downstream are divided into the “D1 Group”, “D2 Group” and “D3
Group”. Fig 9(a) shows the arcs of the U2 Group as an example. Table 5 shows the
strength parameters of the embankment materials based on the results of material
tests during dam construction.
As a result, except for one arc each near the upstream and downstream toes
of the dam, the safety factors of all of the other arcs are greater than 1. The arc
of the minimum safety factor in the 6 arc groups is summarized in Table 6, and
the positions of the corresponding arcs are shown in Fig. 10. For the sliding arcs
with safety factors lower than 1, the sliding displacement is calculated and shown
in Table 6. In the local surface layer of the downstream slope of the dam (the No.
4 arc), the sliding safety factor is 0.847, but nearly no sliding displacement occurs
(0.04 cm). This result suggests that the sliding in the Aratozawa Dam did not occur
during Earthquake A (the Iwate-Miyagi Nairiku Earthquake in 2008). This result is
462
Q.106 - R.28
Fig. 9
Arcs of the U2 Group (as an example of sliding arcs).
Arcs du groupe U2 (comme exemple d’arcs glissants).
Table 5
Strength parameters of the embankment materials.
Cohesion Friction angle
Division (MPa) (degrees) Evidence
Core 0.049 33.2 Based on triaxial compression test
Filter 0.078 42.2
Transition 0.039 39.9
Upstream rock zone Inner 0.049 42.7
Outer 0.049 43.4
Downstream rock zone Inner 0.049 40.2
Outer 0.049 42.7
Table 6
The results of sliding deformation for Earthquake A.
Slip direction Arc group Arc No. Safety factor Slip disp. (cm)
To the upstream side U1 1 0.994 0.0001
U2 2 1.469 –
U3 3 1.478 -
To the downstream side D1 4 0.847 0.04
D2 5 1.239 –
D3 6 1.407 –
consistent with the field survey after the earthquake [9], which reported that no slip
phenomenon was found to be due to the earthquake.
It has been noted that, in addition to sliding, subsidence may be caused by

the shaking and rearrangement of the soil particles in fill dams due to strong earth-
quakes. Such deformation can be regarded as a result of the decrease in the rigidity
of the dam. Subsidence can be conveniently evaluated by deformation calculation
463
Q.106 - R.28
Fig. 10
Arcs with the minimum safety factor of each arc group for Earthquake A.
Arcs avec un facteur de sécurité minimum de chaque groupe d’arc pour le
tremblement de terre A.
due to dead load, using the shear moduli of the embankment materials before and
after the earthquake (Railway Technical Research Institute 1997). Here, the rigidi-
ties of the embankment materials before Earthquake A are evaluated by the initial
shear moduli used in the earthquake response analysis, and those after the earth-
quake are evaluated by the converged shear moduli with the equivalent linearization
method. The 3D model shown in Fig. 5 is used for this analysis. The difference in
the deformation before and after the earthquake can be regarded as the residual
deformation caused by the earthquake. That is,
U r = U a − Ub (6)
where Ur is the residual deformation caused by the earthquake and Ub and Ua are
the deformations due to dead load before and after the earthquake, respectively.
Figure 11 shows the deformation due to this mechanism after Earthquake

A. In the standard cross-section of the dam, the maximum value of subsidence is
5.44 cm. The maximum subsidence of 5.53 cm occurs at the crest of the dam in the
dam axial section.
464
Q.106 - R.28
Fig. 11
Subsidence related to the shaking and rearrangement of the soil particles caused
by Earthquake A.
Affaissement lié aux secousses et au réarrangement des particules du sol
provoqués par le tremblement de terre A.
During an earthquake, sliding and subsidence related to the shaking and

rearrangement of soil particles may occur simultaneously. However, the sliding dis-
placement of a rigid block cannot be directly added to the calculated subsidence
results. As an estimation of the possible maximum subsidence at the crest, the ver-
tical component of the slippage at the crest can be summed with the subsidence
related to the shaking and rearrangement of the soil particles. Since no sliding
behavior was observed during Earthquake A, the residual deformation of the dam
caused by Earthquake A equals that related to the shaking and rearrangement of
soil particles, with a maximum value of 5.53 cm at the crest of the dam. According
to the field survey after Earthquake A, the subsidence at the crest of the Aratozawa
Dam was approximately 20 cm (Shimamoto et al. 2008). However, since the survey
was conducted 16 days after the mainshock, the measured subsidence includes
the contributions of many aftershocks, and other events since December 4, 2007.
Therefore, the exact residual deformation caused by Earthquake A itself cannot be
confirmed.
5. RESPONSE OF THE DAM TO EARTHQUAKE B
5.1. EARTHQUAKE RESPONSE ANALYSIS
Using the same model shown in Fig. 5 but with the material properties exhib-
ited after Earthquake A, the response of the Aratozawa Dam to the Earthquake
B is analyzed with the equivalent linearization method. As shown in Fig. 3, after
Earthquake A, the shear moduli of the embankment materials dropped to approxi-
mately 58% of those before Earthquake A (G/G0 = 0.58 corresponding to the shear
strain 1.0 × 10−6 ) for approximately one week. In consideration of the nonlinearity of
the materials, the earthquake response analysis is performed by the equivalent lin-
earization method using the approximate curve in the aftershock period, as shown in
465
Q.106 - R.28
Fig. 12
Synthetic ground motion at the dam base during Earthquake B (by fitting the
earthquake record of an aftershock to the lower limit response spectrum).
Mouvement synthétique du sol à la base du barrage pendant le tremblement de
terre B (en ajustant l’enregistrement du tremblement de terre d’une réplique au
spectre de réponse de limite inférieure).
Fig. 3. The damping is thought to increase with the decrease in the material rigidity.
Because the quantitative variation is unknown, the values shown in Table 3 and Eq.
[5] are used to obtain a conservative evaluation.
For setting Earthquake B, it is assumed that a potential fault exists directly

under the dam, causing an Mj 6.5 earthquake. The ground motion at the dam
base is created by fitting an earthquake record of a past earthquake to the lower
limit response spectrum specified in the current guidelines (MLIT 2005). Here, the
earthquake record at the dam base (F) in the aftershock that occurred 62 hours
after the Iwate-Miyagi Nairiku Earthquake in 2008 is used. Figure 12 shows the
acceleration waveforms and the spectra of the created ground motions.
5.2. RESIDUAL DEFORMATION CAUSED BY EARTHQUAKE B
As in the case of Earthquake A, the sliding deformation is calculated for the

standard cross-section. The stiffness of the embankment materials was greatly
reduced due to Earthquake A. Therefore, the residual strength of the embankment
materials should be used in the calculation due to Earthquake B. The results of
triaxial compression tests of Shichigasyuku Dam materials [10], which are similar
to the materials of the Aratozawa Dam, are used here. The residual strength fac-
tors used in the slippage calculation are shown in Table 7. The calculation results
are summarized in Table 8 and Fig. 13. Compared with the upstream side, the slip
466
Q.106 - R.28
Table 7
Residual strength parameters of embankment materials for the calculation of
sliding deformation caused by Earthquake B.
Cohesion Friction angle
Classification (MPa) (deg.) Notes
Core 0. 22.1 As the residual strength parameters, cohesion
is set to be zero, and frictional angle is 2/3 of
the peak value.
Filter 28.1
Transition 26.6
Upstream rock zone Inner 28.5
Outer 28.9
Downstream rock zone Inner 26.8
Outer 28.5
Table 8
Results of sliding calculations for Earthquake B.
Sliding dir. Group No. Safety factor Max. sliding disp. (cm) Max.subsidence (cm)
To the upstream side U1 1 0.384 2.80 1.99
U2 2 0.935 0.04 0.02
U3 3 1.279 - -
To the downstream side D1 4 0.700 46.20 32.91
D2 5 0.789 76.76 49.80
D3 6 0.950 1.71 1.10
Fig. 13
The arcs with the maximum slippage in each group caused by Earthquake B.
Les arcs avec le glissement maximum dans chaque groupe provoqué par le
tremblement de terre B.
467
Q.106 - R.28
displacement on the downstream side is larger overall. This is presumed to be due to

the difference in the slope angle of the upstream and downstream and the strength
of the materials. The slope of the upstream surface is 1:2.7, while that of the down-
stream surface is 1:2.1. In addition, as shown in Table 7, the frictional angles of the
downstream rock zones are slightly smaller than those of the upstream side.
The sliding deformations to the upstream side are generally judged to be per-
missible. To the downstream side, although the sliding section of the No. 4 arc
exhibits a displacement of 46.20 cm, it is very localized in the vicinity of the down-
stream toe. The sliding section of the No. 5 arc passes through the crest with a
sliding displacement of 76.76 cm. The subsidence at the crest accompanying the
sliding reaches 49.80 cm.
Figure 14 shows the residual deformation of the dam associated with the
shaking and rearrangement of soil particles caused by Earthquake B. Regarding
the maximum potential subsidence of the dam crest caused by Earthquake B, it is
possible to sum the vertical components calculated by the above two methods. This
sum is estimated to be 76.20 cm (49.80 cm + 26.40 cm) at the crest.
Fig. 14
Residual deformation related to the shaking and rearrangement of the soil
particles caused by Earthquake B.
Déformation résiduelle liée aux secousses et au réarrangement des particules du
sol provoqués par le tremblement de terre B.
5.3. RESIDUAL DEFORMATION CAUSED BY THE DOUBLET EARTHQUAKE
The deformation caused by the doublet earthquake is calculated by summing

the deformation induced by Earthquake A and Earthquake B. As the maximum
possible subsidence at the dam crest, the vertical components of the slip deforma-
tion and that related to the shaking and rearrangement of the soil particles can be
summed directly as long as the location used for each estimate remains the same.
The maximum possible subsidence at the dam crest caused by the doublet earth-
quake is predicted to be 81.27 cm. By comparing the subsidence calculated above
with the dam freeboard (200 cm), it is considered that the earthquake may not induce
468
Q.106 - R.28
immediate overflow. However, the subsidence of the crest reaches an amount that
requires detailed investigation, for example, via elastic-plastic analysis.
5.4. PROPOSAL OF EMERGENCY MEASURES AFTER A LARGE EARTHQUAKE
Even rockfill dams constructed by modern construction techniques are not

likely to suffer from liquefaction due to a large earthquake, but the following events
are considered to occur.
1) Overflow due to the excessive subsidence of the crest

2) Concentrated seepage along the slip surface cross the core
3) Water penetration through cracks
4) Increment of the amount of seepage flow and the generation of turbidity of
seepage water
Since these phenomena may cause dam failure, emergency inspection imme-
diately after a large earthquake is essential. As revealed in this study, there is
potential risk that the seismic performance of a rockfill dam may temporarily decline
due to the change in the dynamic characteristics and material properties after a large
earthquake. Therefore, in preparation for a large earthquake that may occur later, it
is proposed to lower the reservoir water to a safe level and keep it for a week. The
counter measure is also needed to prevent the infiltration of water into the cracks.
In addition, careful monitoring of the dam body and timely analysis of records are
recommended.
6. CONCLUSIONS
The seismic performance verification of a rockfill dam for a large doublet

earthquake was tested. A practical verification procedure was proposed.
The variation history of the material properties of the Aratozawa Dam is clar-
ified by analyzing 19 earthquake records. The material properties of rockfill dams
may change temporarily in response to strong seismic motion. If another large earth-
quake occurs before the dynamic characteristics of the dam and the properties of
the embankment materials recover, the subsidence of the dam may increase sig-
nificantly. From the viewpoint of dam safety management, it is necessary to quickly
lower the reservoir water to a safe level and keep it at the lower level for at least 1
week in preparation for a large doublet earthquake.
The stiffness of the rockfill dam that experiences a strong earthquake may
recover to its normal status after approximately 1 week. During this period, careful
469
Q.106 - R.28
frequent monitoring is necessary. Moreover, early remedial action to fix the damaged
portion of the dam body is needed.
ACKNOWLEDGEMENTS
Many thanks to the Miyagi Prefecture Kurihara Dam Management Office and
the Japan Commission on Large Dams for providing us with the earthquake records
and the information about the dam used in the present study.
REFERENCES
[1] OHMACHI T. and TAHARA T. Nonlinear earthquake response characteris-

tics of a central clay rockfill dam, Soils and Foundations, Vol.51, No.2, 2011,
227–238.
[2] MATSUMOTO N., YASUDA N., and CAO Z. Investigation on the long-term
fluctuation of the dynamic characteristics and transmitting behaviors of seismic
waves in a rockfill dam – foundation system, Journal of Structural Mechanics and
earthquake Engineering, JSCE, No.74 / I-3, 2018, 319–329.
[3] Earthquake Engineering Committee, Japan Society of Civil Engineers (JSCE).

Report on the damage surveys and investigations following the 2016
Kumamoto Earthquake, Maruzen Publishing Co., Ltd., 2017 (in Japanese).
[4] Ministry of Land, Infrastructure, Transport and Tourism (MLIT), River Bureau.
Guideline for seismic performance verification of dams against large earth-
quakes in Japan (draft), 2005 (in Japanese).
[5] YASUDA N. and MATSUMOTO N. Comparisons of deformation characteristics

of rockfill materials using monotonic, cyclic loading and in-situ tests, Canadian
Geotechnical Journal, Vol.31, No.2, 1994, 162–174.
[6] YASUDA N., MATSUMOTO N., and CAO Z. Study on the Mechanism of the
Peculiar Behaviors of the Aratozawa Dam During the 2008 Earthquake, Journal
of Disaster Research, Vol.13 No.1, 2018, 205–315.
[7] Railway Technical Research Institute. Design standards of railway structures,

Maruzen Publishing Co., Ltd., 1997 (in Japanese).
[8] SAWADA Y. and TAKAHASHI T. Study on the material properties and the
earthquake behaviors of rockfill dams, 4th Japan Earthquake Engineering
Symposium, 1975, 695-702 (in Japanese).
470
Q.106 - R.28
[9] WATANABE H and BABA K. A Study of sliding stability evaluation method

based on the dynamic analysis of fill dams,Large Dam, No.97, 1981, 25–38.
[10] SHIMAMOTO K., SATO N., OMACHI T., et al. Report on damage to dams
caused by the 2008 Iwate/Miyagi inland earthquake, 2008.
[11] HOJO K., HARITA K. and OGAWA M. Large-scale tri-axial test for the char-
acteristics of filldam materials, Technical memorandum of PWRI, No.1507,
Public Works Research Institute, Ministry of Construction, 1979.
471
Q.106 - R.29
-------
-------
INVESTIGATION OF DAMPING CHARACTERISTICS OF DAMS EVALUATED

BY DE/TFM METHOD(∗)
Masayuki KASHIWAYANAGI
Chigasaki Research Institute, Research &T Development Dept.,
ELECTIC POWER DEVELOPMET CO., LTD.
Zengyan CAO
Senior Engineer, Social & Environmental Engineering Department,
JP BUSINESS SERVICE CORPORATION,
JAPAN
SUMMARY
This paper examined the damping characteristics during earthquakes of 100 m

high dams based on the monitored earthquake behavior using the DE/TFM method,
which is newly invented by authors and uniquely relay on a free vibration extracted
from the earthquake response behavior of the dam. The examination focused on
the internal damping and dissipation (external damping) due to interaction among
dam, foundation and reservoir of a concrete gravity dam, an arch dam and a rockfill
dam. Following key findings are captured on the damping characteristics based on
the estimated damping ratios of dams examined. The internal damping of concrete
dams that is essential parameter for the dynamic analysis is identified as 3% and
1.5 % for the concrete gravity dam and the arch dam examined, respectively. An
additional internal damping proportional to the acceleration degree is inevitable for
the arch dam to incorporate the non-elastic behavior such as displacement of trans-
verse joints. Regarding the external damping of concrete dams, ones due to the
interaction with the reservoir are identified proportional to the water depth and ones
due to the interaction with the foundation are expected as similar degree as other
damping. A non-linear dependency on an acceleration amplitude is the outstanding
characteristics of the damping of the rockfill dam. It is interpreted that the non-linear
∗ Étude des caractéristiques d’amortissement des barrages à l’aide de la méthode DE/TFM
472 DOI: 10.1201/9781003211808-124

Q.106 - R.29
properties of dam material depending on shear strain of the dam have emerged. The
infiltration of the water into the rockfill dam absorbs the interaction with the reservoir.
RÉSUMÉ
Ce rapport porte sur les caractéristiques d’amortissement des barrages de

100 mètres de haut durant les séismes. Elle s’appuie sur l’évaluation du comporte-
ment observé des séismes à l’aide de la méthode DE/TFM. Cette méthode, qui a
été inventée par les auteurs, a pour spécificité de se concentrer sur une vibration
libre extraite du comportement du barrage en cas de séisme. L’examen a porté
en particulier sur l’amortissement interne et la dissipation (amortissement externe)
causés par les interactions entre le barrage (barrage-poids en béton, barrage-
voûte et barrage en enrochement), les fondations et le réservoir. Les principaux
résultats suivants ont été obtenus pour les caractéristiques d’amortissement, qui
ont été établies à partir des taux d’amortissement estimés des barrages étudiés.
L’amortissement interne des barrages en béton, qui est un paramètre essentiel
de l’analyse dynamique, était de 3% et de 1,5% respectivement pour le barrage-
poids en béton et le barrage-voûte. Un accroissement de l’amortissement interne
proportionnel au degré d’accélération est indispensable pour le barrage-voûte afin
qu’il puisse avoir un comportement non-élastique, comme le déplacement des
joints transversaux. En ce qui concerne l’amortissement externe des barrages en
béton, on a constaté que l’amortissement externe causé par l’interaction avec le
réservoir est proportionnel à la profondeur du réservoir, et que celui causé par
l’interaction avec les fondations est d’un degré comparable à celui des autres types
d’amortissement. Une dépendance non-linéaire avec l’amplitude de l’accélération
est la principale caractéristique de l’amortissement du barrage en enrochement. On
estime que les propriétés non-linéaires des matériaux du barrage selon la contrainte
de cisaillement ont émergé. L’infiltration de l’eau dans le barrage en enrochement
absorbe l’interaction avec le réservoir.
1. INTRODUCTION
Damping characteristics are one of the dominating parameters in seismic

responses of structures. Damping in structures can be categorized into those due to
structural reasons and material reasons according to the mechanism of their occur-
rences. Damping due to structural reasons is characterized by a variety of these
mechanisms. These are caused in dams as energy dissipation due to shear move-
ment in transverse joints associated with the dam behavior and due to the interaction
among a dam, a foundation and a reservoir (referred to as the dam system in this
paper). The damping is grouped as these due to material reasons and dissipation
in the dam and as dissipation due to interaction between a dam and a reservoir or
a foundation in this paper. The former is referred to as an internal damping and the
latter is as external dissipation in this paper.
473
Q.106 - R.29
A selection of damping characteristics is of great interest in order to match the

numerical model and the dynamic method used in the design or the safety assess-
ment of dams. The dynamic analysis of dams has been recently conducted with a
three-dimensional model consisting of a dam system associated with viscos bound-
ary simulating infinite foundation. With adequate numerical model, only the internal
damping is necessary for the analysis, while the dissipation due to interaction among
the dam system is naturally considered in the analysis. The internal damping is usu-
ally estimated by the laboratory tests of dam material or the reproduction analysis
of the dam behavior during earthquakes. These estimations involve problems which
are scale effect of the tests and dependency of the reproduction analysis on the char-
acteristics of the dam foundation and seismicity. Ohmachi & Kataoka [1] indicated
that the dissipation characteristics due to interaction in the dam system depended
on the stiffness of the dam foundation.
In dam engineering, studies have been done on damping evaluation. There

are several results for the estimation of the material damping of fill material, such
as Matsumoto et al. [2]. Sato et al. [3] investigated the rockfill dam behavior dur-
ing earthquakes which were attenuated by the damping according to the growth of
shear strain in the dam. For concrete dams, authors are unknown about experimental
results of damping of concrete for dams. Ueda et al. [4] investigated the relationship
between damping ratios and vibration amplitudes of two arch dams based on the in-
situ vibration tests and earthquake records where they applied a half-power method
[5] for the estimation of damping. JCOLD [6] has complied monitored earthquake
records broadly in Japan and examined these records in terms of the damping of
dams by a half-power method. In this research, the damping ratios of concrete dams
showed wide scatter ranging from 0 to 14% which provided unclear characteristics.
Authors think that its challenge exposes the limitation in accuracy of the half-power
method in damping estimation. It is often reported that damping ratios are identi-
fied by reproduction analysis of the seismic behavior of dams [7, 8]. Such results
have less applicability to other dams or higher intensity of earthquakes. Ohmachi &
Kataoka [1] investigated the dissipation due to interaction among the dam system
by the numerical simulation. The degree of the dissipation depended on the param-
eter of impedance ratio, which was characterized by the ratio of the shear modulus
of a dam and a foundation. Almost existing dams have been constructed on the
foundation with impedance of 2/3 to 2.0. The higher impedance the site has, the
larger dissipation the dam is shared. Authors consider that these challenges for the
internal damping fall short in providing versatile characteristics applicable to many
dams.
In this paper, authors aim to extract a typical characteristic of damping of dams

from the damping ratios estimated by a damping evaluation method aided by transfer
function matrix (DE/TFM method) using monitored earthquake responses of existing
dams in Japan. The investigation involves not only the estimation of damping ratios
of a concrete gravity dam, an arch dam and a rockfill dam of approximately 100 m
high but also the examination of the internal damping and the external dissipation of
dams. The results and these challenges are applicable to the seismic performance
474
Q.106 - R.29
examination of existing dams in a wider range of reservoir water depth or accelera-

tion amplitude in dams with many earthquake records accumulated. For the similar
examination in dams without earthquake monitoring, the damping characteristics
evaluated with similar challenges can be a substitute to the dam of similar scale and
same type.
2. OVERVIEW OF DE/TFM METHOD
2.1. ESTIMATION OF DAMPING RATIO BY THE DE/TFM METHOD
Authors have invented a DE/TFM method [9] for the estimation of damping
ratios of dams. The DE/TFM method is consist of two processes of Process I and
Process II. A free vibration of a dam is extracted from the earthquake response
record of the dam by utilizing a transfer function matrix (TFM) method [10] in Process
I. The damping ratio of the dam is estimated for the extracted free vibration by a
logarithmic decrement method (LDM) in Process II. Referring to Fig. 1, the concept
of the DE/TFM method is given below.
It is assumed that there are three-directional earthquake records A and B at

the dam crest, a and the foundation, b of the dam, as shown in Fig. 1 (a). For
simplicity, Fig. 1 (a) illustrates only 1-directional component of an earthquake record
as an example. These earthquake records are divided into two time-zones before
and after the time of occurrence of the maximum acceleration value at the dam
crest. As shown in Fig. 1 (b) (left), these waves are termed as A1 and A2 at the dam
crest and B1 and B2 at the foundation. Considering the propagation time t of the
wave from the foundation to the dam crest, strictly speaking, the start times of B2
and A2 are different. But here, it is supposed that A2 and B2 start simultaneously.
The response A2 at the dam crest is the combination of the free vibration (F2 ) of the
dam excited by the input ground motion of B1 and the forced vibration (R2 ) due to
the input ground motion of B2 . If R2 can be accurately obtained by an appropriate
method, F2 at the dam crest can be obtained by excluding R2 from A2 . That is,
F2 = A2 − R2 (1)
Then, the damping ratio of the dam can be obtained when the LDM (Eq. [2])
is applied to F2 . Therefore, the fundamental problem of the DE/TFM method is how
to find R2 with high accuracy.

1 pi
h= ln (2)
2π n pi + n
where, h: Damping ratio, pi , pi+n : Amplitude of a free vibration at ith wave
475
Q.106 - R.29
Fig. 1
Schematic view of dam response due to earthquakes.
Vue schématique de la réaction du barrage aux séismes.
1 Monitored behavior during earthquakes, 1 Comportement observé durant les
3-directional components séismes, éléments à 3 directions
2 Transfer function matrix 2 Matrice de la fonction de transfert
3 Seismograph 3 Séismographe
4 Time domains 4 Domaines temps
5 Frequency domains, {SiA2 } denotes a 5 Domaines fréquence, {SiA2 } dénote
spectrum of A2 (i = X , Y , Z) etc. un spectre de A2(i = X , Y , Z) etc.
6 Free vibration 6 Vibration libre
7 Forced vibration 7 Vibration forcée
The forced vibration (R2 ) can be simulated using the TFM of the dam [10]. The
TFM consists of nine components of transfer functions as shown in Eq. [3] and [T ]
is obtained using three sets of past earthquake records of the dam.
⎡ ⎤
Txx Txy Txz
[T ] = ⎣ Tyx Tyy Tyz ⎦ (3)
Tzx Tzy Tzz
476
Q.106 - R.29
Where, each component Tij (i, j = X , Y , Z) of the matrix is a transfer function that
considers mutual interference between directions.
The Fourier spectrum {SiR2 } (i = X , Y , Z) of the forced vibration at the dam crest,
a of the dam is obtained by the production of the Fourier spectrum, {SiB2 } (i = X , Y , Z)
of the input ground motion B2 at the foundation b of the dam and [T ], as shown in
Eq. [4] and Fig. 1 (b).
{SiR2 } = [T ]{SiB2 } (i = X , Y , Z) (4)
Each variable in Eq. [4] is a function of angular frequency, and Eq. [4] holds for
each frequency. By inversing Fourier transformation of the {SiR2 } obtained from Eq.
[4], the forced vibration R2 due to the input ground motion B2 can be obtained. Fur-
thermore, the free vibration F2 of the dam is estimated by Eq. [1]. As the practical
application of the DE/TFM method, above procedure is easier to conduct in the fre-
quency domain. To obtain F2 , the inversing Fourier transformation is conducted for
the deference between the Fourier spectra of {SiA2 } and {SiR2 }. In addition, to obtain
a free vibration with higher accuracy, it is better to cut high frequency components
sufficiently separated from the primary natural frequency of the dam in each vibra-
tion direction. Since the contributions of the input earthquake motion in the three
directional components are included in the method shown in Eq. [4], a high accurate
R2 can be obtained.
The application of the LDM to the free vibration, F2 was sometimes not
straightforward depending on the characteristics of F2 . The accuracy of TFM may
be impaired by the multiple excitation along the dam abutment, while the TFM
is practically estimated by the excitation at one fixed location of the dam founda-
tion where earthquakes are monitored. It is considered that waves in the reservoir
during an earthquake could reflect the dam response at the crest, resulting in
compromising F2 . Therefore, the following treatment is additionally provided before
applying the LDM to F2 The peak values of the positive half amplitude of the wave-
form shown in Fig. 2 (left) are denoted as A1+ , A2+ , etc. Similarly, those of the
negative half amplitude are denoted to be A1− , A2− , and so on. Plotting the absolute
peak values and the corresponding cycle numbers gives the relationship between
the amplitude and the cycle number indicated by • in Fig. 2 (right). Furthermore, an
logarithmic approximation curve is taken for this relationship. Instead of the original
data, the damping ratio of the dam is estimated by applying LDM (Eq. [1]) to this log-
arithmic approximation curve. Here, it is desirable that the logarithmic approximation
process has peak values of at least three cycles.
2.2. FEATURES OF DE/TFM METHOD
To make clear the feature of the DE/TFM method, the comparison is conducted
between the DE/TFM method and the half-power method. The half-power method
is applicable to not only individual earthquake record but also a response function
obtained by in-situ vibration tests. In case of using the earthquake data, it requires
477
Q.106 - R.29
Fig. 2
Processing of damping estimation for extracted free vibration
Traitement de l’estimation de l’amortissement pour la vibration libre extraite.
1 Acceleration 1 Accéleration
2 Elapsed time 2 Temps écoulé
3 Extracted free vibration 3 Vibration libre extraite
4 Cycles 4 Cycles
5 Logarithmic approximation 5 Approximation logarithmique
only the Fourier spectrum analysis, which is easily conducted by Fourier transforma-
tion technique. However, once the frequency distribution of the spectrum is obtained,
there comes some difficulties in the estimation of damping, which involve that identi-
fication of the dominant frequency is interfered by many spikes of the spectrum, and
that the degradation of the spectrum could be caused due to the dependency on the
sampling length and/or interval of earthquake data, and due to filtering and smooth-
ing procedure to highlight the necessary frequency, etc. Such difficulties incur the
arbitrary estimation as well as the skilled background of the person to estimate.
The DE/TFM method uses only the monitored earthquake records. These
records are processed by the method of Fourier transformation and simple arithmetic
operations in the frequency domain. These procedure of the DE/TFM method is
straightforward and simple, thus keeping away the difficulties encounter in the half-
power method. Thanks to such feature of the DE/TFM method, it is particularly
effective to treat a lot of accumulated records aiming the identification of damping
characteristics of existing dams. The disadvantage of the method is indicated that the
method requires at least suitable three sets of earthquake records for the estimation
of TFM. For this reason, the application of the method may be limited to dams
conducting earthquake monitoring more than several years.
3. DAMPING CHARACTERISTCS EVALUATED BY DE/TFM MTHOD
The damping characteristics of existing three dams are examined based on

the damping ratios estimated by the DE/TFM method. Major dam types with similar
478
Q.106 - R.29
Table 1
Dams for Examination
Dam name Dam type Dimension Completion Earthquake
(Height, Crest length, Volume) monitoring, from
Kazeya PG 101 m, 329.5 m, 592000 m3 1960 2007
Sakamoto VA 103 m, 256.3 m, 183000 m3 1962 2007
Ouchi ER 102 m, 340 m, 4459000 m3 1985 1985
(∗) All dams are owned by Electric Power Development Co., Ltd.
dam height are selected for the examination to demonstrate specific characteris-
tics of damping. These dams are listed in Table 1. The earthquake monitoring for
these dams are conducted in three directions of dam axis, stream and vertical direc-
tion. The monitored records have been accumulated as 35, 30 and 200 events for
Kazeya, Sakamoto and Ouchi dams, respectively as of 2019. The examinations of
the damping are conducted with each 30 events selected.
3.1. DAMPING OF A CONCRETE GRAVITY DAM
The damping ratios are successfully evaluated for 15 events and failed in 15
events for Kazeya, a concrete gravity dam. These processes of the evaluation by
the DE/TFM method are shown in below as an example. Three sets of the events
are necessary for the estimation of TFM in Process I. It is known that the behavior
during earthquakes depends on the interaction between a dam and a reservoir for
concrete dams and behavior of transverse joints for arch dams. Taking such behavior
of dams into consideration, three sets of events are adopted to the estimation of
TFM to consider such dependencies with respecting to the water depth and the
acceleration amplitude of the dam during earthquakes. It results 10 TFMs which are
estimated by each 3 events out from 30 events. One of the TFMs are shown in Fig. 3
as an example.
The free vibration, F2 is extracted from observed wave, A2 and forced vibra-
tion, R2 which is simulated using relevant TFM and Eq. [4]. Waves of A2 and F2
are compared in a case of 2016/11/19 earthquake in Fig. 4. By processing log-
arithm approximation of the peak values of F2 , the damping ratios are estimated
by Eq. [2]. The results of a concrete gravity dam are illustrated respect to the
reservoir water depth and the acceleration amplitude of the dam crest as shown in
Fig. 5.
The damping ratios estimated are ranging from 2.7% to 6% and 4% in aver-
age. While the maximum acceleration on the dam crest (Ac ) are almost less than
100 cm/s2 and 370 cm/s2 for one earthquake, the dependency of the damping ratios
479
Q.106 - R.29
Fig. 3
Example of TFM (Kazeya dam).
Exemple de TFM (barrage de Kazeya).
1 Amplitude of transfer function 1 Amplitude de la fonction de transfert
2 Frequency 2 Fréquence
Fig. 4
Comparison between observed wave and extracted free vibration in Kazeya dam.
Comparaison entre l’onde observée et la vibration libre extraite
(barrage de Kazeya).
1 Acceleration (cm/s2 ) 1 Accélération max. (cm/s2 )
2 Elapsed time 2 Temps écoulé
3 Observed at the dam crest 3 Observée à la crête du barrage
4 Extracted free vibration 4 Vibration libre extraite
5 Observed at the dam foundation 5 Observée aux fondations du barrage
480
Q.106 - R.29
Fig. 5
Damping ratio of a concrete gravity dam.
Taux d’amortissement d’un barrage-poids en béton.
1 Damping ratio (%) 1 Taux d’amortissement (%)

2 Max. acceleration (cm/s2 ), Ac 2 Accélération max. (cm/s2 ), Ac
3 Estimated damping ratio 3 Taux d’amortissement estimé
4 Average of the estimation 4 Moyenne de l’estimation
5 Non-dimensional water depth, H ∗ 5 Profondeur du réservoir non-dimensionnelle, H*
6 Regression to the water depth 6 Régression / profondeur
on Ac is not clear in Fig.5 (Left). However, one on the water depth is identified in the
range over 0.4 of non-dimensional water depth (H ∗). A linear relation is considered
for these dependencies as Eq. [5].
h = a + bAc + cH ∗ (5)
where, h: damping ratio, Ac : maximum acceleration on the dam crest, H ∗: non-

dimensional water depth which is respect to the maximum water depth of the
reservoir, a, b, c: coefficients
Examining the data shown in Fig. 5 based on the above equation, the water
effect makes the damping up to 1% which is illustrated by dotted line in Fig. 5 (Right).
It is reasonable that the deeper the water depth exist, the higher the interaction
between the dam and the reservoir is caused, resulting a certain damping in the
dam behavior. However, the acceleration degree has very small influence on the
damping of concrete gravity dams.
3.2. DAMPING OF AN ARCH DAM
By the similar examination, the damping ratios are successfully evaluated for
15 events for Sakamoto, an arch dam. the results are summarized in Fig. 6 as the
relation with Ac and H ∗ with the similar estimation of the damping ratios by Ueda et
481
Q.106 - R.29
al. [4]. Dependency on Ac and H ∗ is clearly identified in these data in Fig. 6. These
relations are examined according to Eq. [5] and illustrated as a regression line in the
figures. Ueda et al. [4] suggested that an energy dissipation was additionally caused
in the arch dam due to the behavior of slide or disjunction of transverse joints during
earthquakes. These behaviors (referred to as non-elastic behavior in this paper)
could be restricted by closely contacted transverse joints of which degree depend
on Ac and H ∗, and additionally ambient temperature. These situations agree with the
characteristics of the damping of the arch dam examined here as shown in Fig. 6. It
is presumed that the scatter of the damping ratio around the regression line could
be due to the non-elastic behavior. A numerical simulation taking the joint behavior
into consideration may estimate quantitatively the effect of non-elastic behavior on
damping ratios of arch dams. However, authors consider it is a future issue.
Fig. 6
Damping ratio of an arch dam.
Taux d’amortissement d’un barrage-voûte.

4 After Ueda et al. [4] and its regression 4 D’après Ueda et al. [4] et sa régression
5 Two parameter regression under 5 Régression à deux paramètres sous la
the averaged water depth prodondeur moyenne du réservoir
6 Two parameter regression under 6 Régression à deux paramètres à vide
empty condition
7 Non-definitional water depth, H ∗ 7 Profondeur du réservoir non-dimensionnelle, H*
8 Two parameter regression under the 8 Régression à deux paramètres
averaged acceleration (26 cm/s2 ) avec l’accélération moyenne (26 cm/s2 )
Based on the regression of the damping ratios to Ac and H ∗ shown in Fig.6 it is

assumed that the growths of approximately 1.3% and 2% damping are expected to
the increment of 100 cm/s2 in Ac and a full impoundment of the reservoir, respectively.
In addition, it seems that the value of 1.5% is a lower limit in wide range of Ac and
482
Q.106 - R.29
H ∗. It is identified as a damping due to the material reason caused by the elastic

behavior of the dam. However, certain scatters around the regression is found in
the estimated damping shown in Fig. 6 as well as in Fig. 5. These situations are
discussed in Chapter 4.
3.3. DAMPING OF A ROCKFILL DAM
By the similar examination, the damping ratios are successfully evaluated

for 15 events out of 27 earthquakes for Ohuchi, a rockfill dam. The results are
summarized in Fig. 7 as the relation with Ac and H ∗. The damping ratios estimated
are ranging from 2.7% to 10% which are wider distribution than ones for concrete
dams. These reveal little tendency on H ∗ and non-linear dependency on Ac , even
that the range of H ∗ is limited in above 0.8. Referring to well-known characteristics
of rockfill dams which relate to shear strain of the dam, the damping characteristics
of a rockfill dam is examined in terms of the shear strain of the dam. The shear strain
is estimated by Eq. [6] as a representative one in the whole dam body. The results
are shown in Fig. 8 with the dominant frequency of the dam [11].
Ac 1
γm= × (6)
(2πfp )2 Hm
Where, γm : maximum shear strain in the whole dam body, Ac: maximum acceleration
fp : dominant frequency Hm : dam height
Fig. 7
Damping ratio of a rockfill dam.
Taux d’amortissement d’un barrage en enrochement.

4 Non-dimensional water depth, H ∗ 4 Profondeur du réservoir non-dimensionnelle, H*
483
Q.106 - R.29
Fig. 8
Dynamic characteristics of a rockfill dam.
Caractéristiques dynamiques d’un barrage en enrochement.

2 Dominant frequency (Hz) 2 Fréquence dominante (Hz)
3 Assumed max. strain 3 Contrainte max. supposée
(∗) Lower figure is extracted from the reference [11] and modified.
The dominant frequencies of the dam clearly decline in the shear strain (γm )
above 1 × 10−4 , suggesting that the shear modulus of the dam is degraded. The
damping ratios shows similar dependency that ones of 3% to 6 at lower strain area
increase 10% above 1 × 10−4 of γm . The maximum damping ratio of 10% is iden-
tified when Ac is 176 cm/s2 of which earthquake shows the largest amplification in
earthquakes examined. It is considered that these characteristics reasonably reflect
the material properties of the rockfill dam.
4. DISCUSSION ON THE DAMPING CHARACTERISTICS OF DAMS
The dam behavior during earthquakes is an integration of the dam response

itself to the earthquake and the interaction among the dam system. Therefore, the
damping estimated based on monitored earthquake records involves the character-
istics of all elements relating to the dam behavior during earthquakes. Hereinafter,
the damping is categorized into internal damping and external dissipation and
discussed in terms of the dam types of concrete dams, and rockfill dams.
484
Q.106 - R.29
4.1. DAMPING CHARACTERISTICS OF CONCRETE DAMS
The damping of concrete dams obviously depends on the water depth, indi-
cating that it involves the dissipation due to the interaction between the dam and the
reservoir. One of arch dams depends in addition on the accretion amplitude of the
response, indicating the energy dissipation in the dam due to the non-elastic behav-
ior. Preparing a regression which take such dependencies into consideration, the
hypothetical damping could be estimated without the influence of the water depth
and acceleration amplitude. Furthermore, the lower value than damping ratios esti-
mated corresponds to the situation where the interaction by the foundation and/or
the reservoir is little effective. These are assumed where only the internal damping
works. Based on the former values of approximately 3% and 1.5% and the latter
of approximately 2.7% and 1.8% for the concrete gravity dam and the arch dam
examined, respectively. Therefore, the ratios of the internal damping for the con-
crete gravity dam is evaluated as 3% and for the arch dam is as 1.5% and energy
dissipation due to non-elastic behavior.
The internal damping is only the damping parameter required for the dynamic
analysis of dams where the interaction among the dam system is adequately consid-
ered as the calculation scheme. It is one of the reasons why an adequate evaluation
of internal damping is essential for the dynamic simulation of dams. The reproduc-
tion analysis of the behavior during earthquakes is frequently conducted for such
evaluation of dams. Such analysis for the arch dam examined shows that the damp-
ing ratio of 2% for the internal damping is adequate for the earthquake with the
maximum acceleration of 102 cm/s2 . It moderately agrees with the consideration
above which provides the internal damping of 1.5% and energy dissipation of 1.3%
due to the non-elastic behavior against the acceleration of 102 cm/s2 .
The dissipation due to the interaction between a dam and a reservoir is eval-
uated in the maximum 2% and 1% for the concrete gravity dam and the arch dam
examined, respectively based on the regressions for both. It seems in Fig. 5 and Fig.
6 that the damping ratios obtained here involve another damping, which is assumed
to be the dissipation due to the interaction between a dam and a foundation. Based
on the regression and the value of the internal damping, these are estimated at
2.5% as a maximum, which are almost equivalent to other damping. However, the
damping dissipating to the foundation has not been adequately quantified in this
study. Authors think that the further study is necessary on this issue.
Regarding the similar examination of the damping of arch dams, Ueda et al.
[4] examined damping ratios for arch dams through the in-situ vibration test, in which
107.5 m high (I dam) and 130 m high (J dam) arch dams were excited at maximum
5 cm/s2 by the vibratory machine on the dam crest. The reproduction analysis found
that the internal damping of 1% and 1.5% were adequate for the I dam and the J dam,
respectively. The dissipation to the reservoir and the foundation were evaluated as
maximum 1% and 1.5% for the I dam by the half power method using the response
485
Q.106 - R.29
function of the dam. It is considered that these results are consistent to the author’s
study, considering that vibration of the dams are localized due to the limitation of the
capability of vibratory machine used.
4.2. DAMPING CHARACTERISTICS OF A ROCKFILL DAM
No indication of the interaction between a dam and a reservoir is found in

the damping characteristics of the rockfill dam examined. It is interpreted that the
infiltration of the water into the dam body absorb the interaction.
Regarding the interaction between a dam and a foundation in a rockfill dam,

its degree depends on the impedance ratio of these. The smaller impedance, i.e. the
smaller foundation modulus makes higher interaction. Meanwhile, rockfill dams have
high adaptability to wide range of the foundation characteristics. These situations
result that the characteristics of the interaction between a dam and a foundation of
a rockfill dam involve high variety and should be examined for each dam separately.
This issue will be leave for the further study. It is assumed that the damping evaluated
of the rockfill dam involves both an internal damping and the dissipation to foundation
in this paper.
A non-linear dependency on an acceleration amplitude is the outstanding char-

acteristics of the damping of the rockfill dam. It is interpreted that the non-linear
properties of dam material depending on shear strain of the dam are emerged.
Generally, such non-linearity is gradually obvious in the shear strain range above
1 × 10−4 . The damping ratios estimated show similar characteristics clearly in the
examined rockfill dam. The damping ratio of 10% at the shear strain of 1 × 10−4 might
be overestimated, while such damping is expected at the shear strain of 2.5 × 10−3
in the normal rockfill material. It is caused by the underestimation of the shear strain
using Eq. [6]. The local behavior of the dam crest, where the DE/TFM method treat
should be amplified more than the averaged amplification in the dam body. The
damping ratios of approximately 3% are estimated corresponding to slight shear
strain region. Such results coincide with the damping characteristics of the normal
rockfill material.
5. CONCLUSIONS
This paper investigates the typical damping characteristics during earthquakes

of 100 m high dams based on the monitored earthquake behavior using the DE/TFM
method, which is newly invented by authors. The study focused on the internal
damping and dissipation due to interaction among dam, foundation and reservoir
of a concrete gravity dam, an arch dam and a rockfill dam. The conclusions are
summarized below extracted from the study. The essential feature of the DE/TFM
486
Q.106 - R.29
method is a simplicity to evaluate the damping, thus preventing the arbitrary pro-
cessing and not requiring skilled background. The method uses only the monitored
earthquake records at the dam, which are processed by the method of Fourie trans-
formation and simple arithmetic operations in the frequency domain. The method
is particularly effective to treat a lot of accumulated records aiming the identifica-
tion of damping characteristics of existing dams. The internal damping of concrete
dams that is essential parameter for the dynamic analysis is examined based on the
damping estimation by the DE/TFM method. These are identified as 3% and 1.5%
respectively for the concrete gravity dam and the arch dam examined. An additional
damping proportional to the acceleration degree is necessary for the arch dam to
incorporate the non-elastic behavior such as displacement of transverse joints. The
interactions among a dam, a foundation and a reservoir cause the energy dissipa-
tion from the dam externally, which are evaluated as the damping ratios for concrete
dams examined. The damping ratios due to the interaction with the reservoir are
identified proportional to the water depth. Ones due to the interaction with the foun-
dation are expected as similar degree as other damping. A non-linear dependency
on an acceleration amplitude is the outstanding characteristics of the damping of the
rockfill dam. It is interpreted that the non-linear properties of dam material depend-
ing on shear strain of the dam have emerged. The infiltration of the water into the
dam body absorbs the interaction with the reservoir.
REFERENCES
[1] OHMACHI T., KATAOKA S. Evaluation of dynamic interaction effects of 2-

D dam-foundation-reservoir systems. Journal of JSCE, No.519/I-32, pp.199–
209, 1995. (in Japanese).
[2] MATSUMOTO N., YASUDA N., OKUBO M. Dynamic modulus, damping ratio
and Poisson’s ratio of course material. Civil engineering journal, No.28/7,
pp.382–387, 1986. (in Japanese).
[3] SATO N., SODA H., OTA K. Evaluation and application of dynamic character-
istics and properties based on monitored earthquake records in existing dams.
Engineering for dams, No.321'Cpp.40-47'C2013. (in Japanese).
[4] UEDA M., SHIOJIRI H., YOKOI M, TSUNEKAWA K. Damping ratios of exist-
ing arch dams estimated earthquake records etc. Proc. of 24th Workshop of
earthquake engineering of JSCE, pp.825–828, 1997. (in Japanese).
[5] Timoshenko, S. Vibration Problems in Engineering. D. Van Nostrand Com-

pany, Vol. 32,1929.
[6] JAPAN COMMISSION ON LARGE DAMS Technical committee report, Updat-

ing of large earthquake records in Japan. Large dams, No. 231, 2015. (in
Japanese).
487
Q.106 - R.29
[7] WATANABE N., ARIGA Y., CAO Z. Evaluation of dynamic stability of concrete
gravity dam by three dimensional dynamic analyses taking non-linearity into
account. Journal of JSCE, No.696/I-58, pp.99–110, 2002. (in Japanese).
[8] OHMACHI T., TAHARA T. Nonlinear earthquake response characteristics of

a central clay rockfill dam. Soils and Foundations, Vol.51, No.2, pp.227–238,
2011.
[9] CAO Z., KASHIWAYANANGI M., YOSHIDA M., ASAKA H. Evaluation method
of damping ratio using earthquake records and its application in dam engi-
neering. Proc. of 17th World Conference on Earthquake Engineering, Sendai,
Japan, 2020.
[10] CAO Z., KASHIWAYANANGI M., YOSHIDA M., ASAKA H. A high-accuracy

evaluation method of transfer function considering the interferences between
vibrating directions and its application in dam engineering, Electric Power Civil
Engineering, No.390 pp.13–21, 2017. (in Japanese).
[11] KASHIWAYANAGI M., ONISHI H., CAO Z., YODA M., HAYAKAWA S. Utilizing
statistics of the dam behavior during earthquakes for dam safety management,
Proc. of 15th symposium on rock engineering in Japan, 2021. (in Japanese).
488
Q.106 - R.30
-------
-------
CHARACTERISTIC EVALUATION AND COUNTERMEASURE PLANNING OF

RESERVOIR SEDIMENTATION UTILIZING THE LONG TERM SURVEY DATA IN
HYDROPOWER DAM (∗)
Hirofumi OKUMURA
Dam Resilience Office, CIVIL & Architectural Engineering Dept.
ELECTRIC POWER DEVELOPMENT COMPANY
Tetsuya SUMI
Professor, Disaster Prevention Research Institute
KYOTO UNIVERSITY
JAPAN
SUMMARY
The sedimentation management of hydropower dam reservoir is one of the

major problems for the sustainable operation of the plants in Japan. To evaluate the
sedimentation characteristics and plan a sedimentation management, utilizing the
results of annual sedimentation survey data that has been conducted for several
decades and has survey error is studied. It is found that, averaging the survey
data is effective to eliminate the error and grasp the relationship between annual
maximum dam inflow and annual sedimentation volume. Utilizing the relationship
and numerical analysis, draw-down operation with spillway remodeling is useful to
sustain the function of the hydropower reservoir.
∗ Évaluation des caractéristiques et planification des mesures de la sédimentation des réser-
voirs à l’aide des données de surveillance à long terme sur les barrages hydroélectriques
DOI: 10.1201/9781003211808-125 489

Q.106 - R.30
RÉSUMÉ
La gestion des sédiments des réservoirs des barrages hydroélectriques est

l’un des problèmes majeurs pour l’exploitation durable des centrales au Japon. Afin
d’évaluer les caractéristiques des sédiments et de planifier une gestion des sédi-
ments, on étudie les résultats de l’enquête annuelle sur les sédiments qui a été
menée pendant plusieurs décennies et qui a révélé des erreurs d’enquête. Il s’avère
que la moyenne des données de l’étude est efficace pour éliminer l’erreur et saisir
la relation entre le débit maximal annuel du barrage et le volume annuel de sédi-
ments. En utilisant la relation et l’analyse numérique, l’opération de rabattement
avec la re modélisation du déversoir est utile pour maintenir la fonction du réservoir
hydroélectrique.
1. INTRODUCTION
The sedimentation management of hydropower dam reservoir is one of the

major problems for the sustainable operation of the plants. In Japan, due to the
steep topography and plenty of and seasonal rainfall, the average of reservoir
sedimentation rate in Japan has been increasing to approximately 10%.
The purpose of this paper is to evaluate the sedimentation characteristics and

plan a sedimentation management utilizing the results of annual sedimentation sur-
vey data that has been conducted for several decades at hydropower dam reservoirs.
Although the results of sedimentation survey data presumed to be related to water
discharge and dam operation, it is difficult to confirm the relationship between the
two due to survey error. Therefore, it is needed to eliminate the effect of survey
error by averaging the measured data, which are annual maximum dam inflow and
the annual sedimentation volume. And the change of the relationship by draw-down
operation as sedimentation management is studied.
2. SEDIMENTATION SURVEY METHOD AND ERROR
2.1. SURVEY METHOD
Water utilization dams, including hydropower dams with a dam height of 15m
or more, are required by river administrator to conduct sedimentation survey once
a year and report the results of survey in order to evaluate the influences of sed-
imentation on the storage capacity of the reservoir and on surrounding facilities.
490
Q.106 - R.30
As shown in Fig. 1, the survey method is a cross-sectional survey at 400m intervals

in the reservoir and at 200m intervals upstream from the reservoir end. The results
of the cross-sectional survey are used to determine the amount of sedimentation in
the reservoir by the average cross-section method.
Fig. 1
Sedimentation survey method.
Méthode d’étude de la sedimentation.
1 Dam 1 Barrage
2 Sedimentation 2 Sédimentation
3 Survey ship 3 Navire de surveillance
4 Acoustic wave 4 Onde acoustique
2.2. SURVEY ERROR
Fig. 2 shows the relationship between the annual maximum dam inflow and
the annual sedimentation volume in the Sa dam reservoir based on result of sed-
imentation survey for 65 years from 1956 to 2019, which is owned by ELECTRIC
POWER DEVELOPMENT COMPANY (J-POWER). The figure shows a tendency of
increasing sedimentation volume, however there are several sedimentation decreas-
ing years. Sa dam reservoir is 326,848,000 m3 in capacity and 156m in dam height, it
is difficult to occur that sedimentation on the reservoir bed resurfaces into water and
flows down through the spillway or the intake of the plant. Therefore, the decrease
is considered to be caused by measurement error. And the annual sedimentation
volume varies by several times as the minimum one of the same annual maximum
dam inflow, it is also considered to be caused by the error.
491
Q.106 - R.30
Fig. 2
Relationship between the annual maximum dam inflow and the annual
sedimentation volume in Sa dam reservoir (original capacity 326,848,000 m3 ).
Rapport entre le débit maximal annuel du barrage et le volume annuel de
sédiments dans le réservoir du barrage Sa (capacité initiale de 326,848,000 m3 ).
3. ERROR ELIMINATION
3.1. AVERAGING
The relationship between the annual maximum dam inflow and the annual
sedimentation volume is derived from the results of sedimentation survey in order to
understand the sedimentation situation and apply it to the planning of sedimentation
management. The survey error will be eliminated by averaging the annual sedimen-
tation volume for the same level of annual maximum dam inflow. The nth largest
annual maximum dam inflow and annual sedimentation volume are calculated by
averaging of k data from n to n + k, as shown in Eq. [1] and Eq. [2].
the annual maximum dam inflown
1
i=n+k
= (the annual maximum dam inflow in ith year) (1)
k
i=n
the annual sediment volumen
1
i=n+k
= (the annual sediment volume in ith year) (2)
k
i=n
492
Q.106 - R.30
3.2. RESULT AND STUDY
The averaging method which treats sedimentation survey data with Eq. [1]
and Eq. [2] will be examined for eight dam reservoirs owned by J-POWER that
have been affected by sedimentation problems, shown in Table 1. The types of
regulating reservoirs shown in the table are classified according to differences in
power generating operations and sedimentation characteristics in the regulating
reservoir, and the classification method is shown in Table 2.
Table 1
Dam reservoirs for examination of survey data averaging method
Dam reservoir Dam Original reservoir Sedimentation Survey data
(types) height capacity as of 2019 period
Volume Rate
Fn Dam (regulating, river-type) 25 m 10,900,000 m3 1,681,000 m3 15.4% 1977–2019
Ko Dam (regulating, river-type) 34 m 9,700,000 m3 1,256,000 m3 12.9% 1966–2019
Ak Dam (regulating, inter.-type) 89 m 34,703,000 m3 11,655,000 m3 33.6% 1958–2019
Hi Dam (regulating, inter.-type) 38 m 4,240,000 m3 2,387,000 m3 56.3% 1960–2019
Ho Dam (regulating, lake-type) 32 m 493,000 m3 462,000 m3 93.7% 1969–2019
Ft Dam (regulating, lake-type) 76 m 43,000,000 m3 14,577,000 m3 33.9% 1961–2019
Sa Dam (storage reservoir) 156 m 326,848,000 m3 132,752,000 m3 40.6% 1956-2019
Ka Dam (storage reservoir) 101 m 130,000,000 m3 22,444,000 m3 17.3% 1960–2019
Table 2
Classification of hydropower dam reservoir for effective sedimentation management
At these dam sites, multi-year survey data were averaged using Eq. [1] and
Eq. [2], and approximate curves were obtained for each data, assuming a quadratic
relationship between the annual maximum dam inflow and the annual sedimentation
volume. The reason why a quadratic relationship is assumed, is that the trend of the
493
Q.106 - R.30
curves has both increasing and decreasing, and the distribution of the plot resembles
the shape of a linear or quadratic function. The coefficient of the relationship is
summarized in Fig. 3 to 5.
Fig. 3
Coefficient of determination between maximum annual dam inflow and annual
sedimentation volume (k = 1, 3, 5, 10).
Coefficient de détermination entre le débit annuel maximum du barrage et le
volume annuel de sédiments (k = 1,3,5,10).
Fig. 3 shows that the coefficient of determination between annual maxi-

mum dam inflow and annual sedimentation volume is higher in storage reservoirs
and lake-type regulating reservoirs than river-type and intermediate-type regulat-
ing reservoirs. Because storage reservoirs and lake-type regulating reservoirs are
generally large in capacity and most of the incoming sediment is deposited in the
reservoirs. In addition, the higher the number of data to be averaged, the higher the
coefficient.
494
Q.106 - R.30
Fig. 4
Coefficient of determination between annual maximum dam inflow and annual
sedimentation volume on each number of averaged data.
Coefficient de détermination entre le débit maximal annuel du barrage et le volume
annuel de sédiments sur chaque nombre de données moyennées.
Fig. 5
Coefficient of determination between annual maximum dam inflow and annual
sedimentation volume on each original reservoir capacity.
Coefficient de détermination entre le débit maximal annuel du barrage et le volume
annuel de sédiments sur la capacité initiale de chaque réservoir.
495
Q.106 - R.30
Fig. 4 shows that the coefficient of determination is about the marginal high
value when the number of data to be averaged is 5 or more. However, where the
measurement error is large and the coefficient of determination is relatively small,
the coefficient of determination tends to increase as the number of target data is
increased.
Fig. 5 shows the relationship between the original storage capacity which
indicates the scale of dam reservoir, and the coefficient of determination when the
average number of data items is 5. For the storage reservoirs and lake type regulating
reservoirs, the coefficient of determination to the power of 0.5 is about 0.8 or higher
for all locations, regardless of the size of the reservoirs. On the other hands, for the
river-type and intermediate-types regulating reservoirs, the coefficient to the power
of 0.5 tends to be smaller as the reservoir size increases. Where sediment deposition
and scouring occurs on the reservoir bed, survey in the smaller reservoir size is
considered to be more accurate, and this may be due to the fact that single-beam
cross-sectional surveying is not good in deep reservoir.
4. EVALUATION OF CHARACTERISTICS OF RESERVOIR SEDIMENTATION
The plots before and after the survey data averaging k = 5 of 8 dam reservoirs
in Table 1 are shown in Fig. 6 to 7, Fig.9 to 10. It can be seen in the figures that the
range of plot variations becomes smaller by averaging. The approximate quadratic
curves of before and after the averaging are on the figures, however they does not
change significantly.
Fig. 6
Relationship between annual maximum dam inflow and annual sedimentation
volume of river-type regulating reservoir (left: Fn dam, right: Ko dam).
Relation entre le débit maximal annuel du barrage et le volume annuel de
sédiments du réservoir régulateur de type fluvial (à gauche : barrage Fn, à droite :
barrage Ko).
496
Q.106 - R.30
Fig. 7
volume of intermediate-type regulating reservoir (left: Ak dam, right: Hi dam).
sédiments du réservoir régulateur de type fluvial (à gauche : barrage Ak, à droite :
barrage Hi).
4.1. RIVER-TYPE AND INTERMEDIATE-TYPE REGULATING RESERVOIR
For the river-type and intermediate-type regulating reservoirs, the approxi-

mate quadratic curves are convex upward shown in Figs. 6 to 7, indicating that the
sediment on the reservoir bed would resurface and flow down the dam through the
spillway or the intake of the plant.
In the river-type regulating reservoir, the annual sediment volume becomes

less than zero. Ko dam reservoir has the zero point at 7,000 m3 /s to 8,000 m3 /s
of the annual maximum dam inflow, and when dam inflow is 7,000 m3 /s to 8,000
m3 /s, the water surface profile of the reservoir is the same as river condition from
the dam site to upstream, as shown in Fig. 8. It means that if the dam water level is
Fig. 8
Water surface profile of Ko reservoir.
Profil de la surface de l’eau du réservoir de Ko.
497
Q.106 - R.30
low enough to make the reservoir river, sediment incoming the reservoir would pass
through the reservoir and the dam.
4.2. LAKE-TYPE REGULATING RESERVOIR AND STORAGE RESERVOIR
For lake-type regulating reservoirs and storage reservoirs, the approximate

quadratic curves are both convex upward and downward shown in Fig. 9 to 10, and
Fig. 9
volume of Lake-type regulating reservoir (left: Ho dam, right: Ft dam).
sédiments du réservoir régulateur de type fluvial (à gauche : barrage Ho, à droite :
barrage Ft).
Fig. 10
volume of storage reservoir (left: Sa dam, right: Ka dam).
sédiments du réservoir régulateur de type fluvial (à gauche : barrage Sa, à droite :
barrage Ka).
498
Q.106 - R.30
the curves don’t have zero point crossing vertical axis. This indicates that the most
of sedimentation in the reservoir would stay on the bed even in flood and it is difficult
to make sediment incoming the reservoir pass through the reservoir and the dam.
4.3. RELATIONSHIP BETWEEN SEDIMENTATION AND DAM INFLOW
Based on these trends, the relationships between the annual maximum dam
inflow and the annual sedimentation volume of each reservoir types are classified,
shown in Fig. 11.
Fig. 11
Relationship between the annual maximum dam flow and the annual
sedimentation volume of each reservoir types.
Relation entre le débit annuel maximum du barrage et le volume annuel de
sédiments de chaque type de réservoir.
5. EFFECT ON RESERVOIR SEDIMENTATION OF DRAW-DOWN

OPERATION
5.1. EFFECT ON RESERVOIR SEDIMENTATION
The effect on reservoir sedimentation of draw-down operation during flood

of the Hi dam is studied. A one-dimensional numerical analysis of the riverbed
499
Q.106 - R.30
fluctuation is conducted under the conditions shown in Table 3. The results are shown
in Fig. 12, and it can be seen that it is effective to execute draw-down operation in
flood. And it is possible to make no sedimentation by remodeling the spillway crest
5.5 m lower than the current level.
The relationship between the annual maximum dam inflow and the annual
sedimentation volume is changed by draw-down operation, shown in Fig. 13. The
Table 3
Numerical analysis condition of the riverbed fluctuation at Hi dam
Item Contents
Dam and Reservoir Operation start: 1960, effective water depth: 3m
Dam height: 38 m, Bank length:124 m, CA = 233 km2
Capacity: 4,240,000 m3 (sedimentation rate: 56.3%)
Analysis term 1988–2009 (22 years)
Normal Operation Operation in effective water depth
3.0 m (EL.147∼EL.144 m) according to the record
Draw-down Draw-down Operation during flood when dam
Cases Operation 1 inflow is more than 500 m3 /s, and
(Current spillway) draw-down to water level that is spillway
crest EL.137 m + water depth
Draw-down Draw-down Operation during flood when dam
Operation 2 inflow is more than 500 m3 /s, and draw-down
(Remodeled to water level that is remodeled spillway
spillway) crest EL.131.5 m + water depth
Fig. 12
Numerical analysis result of the riverbed fluctuation at Hi dam.
Résultat de l’analyse numérique de la fluctuation du lit de la rivière au barrage
de Hi.
500
Q.106 - R.30
Fig. 13
Relationship between the annual maximum dam inflow and the annual
sedimentation volume in numerical analysis result at Hi dam (from 1988 to 2009).
sédiments dans le résultat de l’analyse numérique au barrage de Hi
(de 1988 à 2009).
reservoir water level during flood is the lower, the area of the plot variations is located
in the lower level. Where annual maximum dam inflow is under 500 m3 /s that is
criteria of the draw-down operation, the areas of the plot variations is almost same
in all cases. In the case of the Draw-down operation 2, annual maximum dam inflow
is the larger, the less is annual sedimentation volume like river-type. It means that
remodeling the spillway with draw-down operation could make the intermediate-type
reservoir river-type reservoir from the view point of sedimentation management.
The calculation result isn’t affected by survey error, so the range of calculation
plot variations is not so wide as result of the sedimentation survey.
5.2. SEDIMENTATION MANAGEMENT
Sedimentation management is now one of the most important problems to

keep sustainability of hydropower plant, but it was not so important as now at the time
of construction. Therefore, the operation rules were determined mainly for power
generation, and the spillway was designed for maximum flood volume assumed, not
considering sedimentation.
501
Q.106 - R.30
The result of averaging survey data and numerical analysis show the effect
of draw-down operation and spillway remodeling on reservoir sedimentation man-
agement. Indeed, in the Mimikawa river, draw-down operation project with spillway
remodeling in hydropower dams has been executed since the big flood disaster in
2005. The project is also expected for contribution to the environment in Mimikawa
river and coastal area. This fact indicates the possibility of new sedimentation
management in hydropower dam.
6. CONCLUSIONS
The conclusions of this study are as follows. In this paper, elimination of error
is discussed, also survey method with less error should be generalized.
1. Although the results of sedimentation survey include errors, they have the
relationship with the annual maximum dam inflow, and the relationship can be
confirmed by averaging for each of the same dam inflows.
2. The confirmed relationships can be classified to the sedimentation character-
istic types of hydropower dam reservoirs, as arranged by the authors.
3. According to the results of the numerical analysis of riverbed fluctuation, it is
possible to confirm a change in the relationship between the annual maximum
dam inflow and the annual sedimentation volume when the water level of the
dam is lowered during flood. This indicates that lowering the water level during
flood can suppress the sedimentation.
REFERENCES
[1] MINISTRY OF LAND, INFRASTRUCTURE, TRANSPORT AND TOURIAM.

Guideline of sedimentation condition survey, 2001. (in Japanese).
[2] OKUMURA H., SUMI T., Reservoir sedimentation management in hydropower

plant regarding flood risk and loss of power generation, International Sympo-
sium on Dams For A Changing World ICOLD 2012, Kyoto, Japan, 2012
[3] OKUMURA H., SUMI T., Sedimentation management by draw down operation
in hydropower regulating reservoirs considering properties of sedimentation
and facility condition, Journal of water engineering vol.55, 2010. (in Japanese).
[4] SUMI T., YOSHIMURA T., ASAZAKI K., KAKU M., KASHIWAI J., SATO T.:
Retrofitting and change in operation of cascade dams to facilitate sediment
sluicing in the Mimikawa River Basin, The 25th Congress of ICOLD, Stavager,
Q.99-R.45, 2015.
502
Q.106 - R.31
-------
-------
DIAGNOSTIC DES TÉMOINS D’AUSCULTATION DES TIRANTS

PRÉCONTRAINTS DU BARRAGE EL KANSERA (∗)
Abdelilah BOUKAIDI LAGHZAOUI

Directeur du Pôle Aménagements Hydrauliques et Hydroagricoles, Alkhibra
Siham BELHACHMI
Ingénieur Sénior, Alkhibra
Mariam MAHDAOUI
Chef de Service Auscultation des Ouvrages Hydrauliques,
Direction des Aménagements Hydrauliques
MAROC
RÉSUMÉ
Le barrage El Kansera réalisé entre 1926 et 1934 est le plus ancien barrage
réservoir du Maroc, situé sur l’Oued Beht, affluent de l’oued Sebou, à environ 20
km au Sud de la ville de Sidi Slimane.
Cet ouvrage en béton de conception mixte, poids et contreforts, est fondé sur
des terrains marno-calcaires.
En 1968, le barrage a été surélevé de 6 mètres, faisant passer le volume

de la retenue de 197 à 297 Mm3 . Ceci permettait de compenser la diminution du
volume régularisé résultant de l’envasement mais encore, d’accroître la superficie
irriguée de 3700 ha, tout en améliorant le laminage des crues et en faisant passer
le productible électrique annuel à 33 millions de kWh.
Outre les adaptations nécessaires des ouvrages annexes, la surélévation du

barrage a nécessité l’exécution de 77 tirants précontraints de 240 t de tension initiale
∗ Diagnosis of El kansera’s dam monitoring witnesses of pre-stressed tensions
DOI: 10.1201/9781003211808-126 503

Q.106 - R.31
qui ont été mis en place dans des forages d’un diamètre de 130 mm totalisant 4300
m de longueur.
Les tirants précontraints du type VSL sont ancrés de 15 à 20 m dans le rocher

de fondation et ont ainsi permis d’assurer la précontrainte totale de l’ouvrage : les
efforts dus à la surélévation du barrage sont entièrement repris par l’effort de pré-
contrainte. La longueur des ancrages en rocher a été définie à la suite de deux
essais d’arrachement et fixée à 5 m.
Quatre tirants témoins de contrôle munis de vérin conique à leur tête ont été
mis en place permettant de suivre ultérieurement le comportement de ceux-ci, dont
deux tirants sont implantés sur l’aile gauche du barrage et deux sont placés sur le
déversoir de l’évacuateur de crues.
Depuis 1969, un contrôle périodique du comportement des tirants précon-

traints témoins est réalisé moyennant des essais Lift-off qui consistent à mesurer
la pression appliquée sur le vérin, nécessaire pour le soulèvement de la tête du
tirant de 12 mm correspondant la fin de course du vérin et la comparer à la pression
initiale de la mesure effectuée dans les mêmes conditions après la mise en tension
initiale du tirant. La relaxation limite fixée par le constructeur est définie à 12%.
Les taux de relaxation mesurés sur les tirants témoins depuis l’année 1970
sont dans l’ensemble inférieurs à la valeur limite de 12%. Toutefois, des problèmes
d’accès aux tirants sur le déversoir sont rencontrés. Ce qui entrave la réalisation des
mesures de relaxation de ces tirants. Des solutions d’automatisation du système de
mesure pour effectuer des lectures à distance sont préconisées.
SUMMARY
The El Kansera dam built between 1926 and 1934 is the oldest reservoir dam
in Morocco, located on the Oued Beht, a tributary of the Oued Sebou, about 20 km
south of the city of Sidi Slimane.
This concrete structure of mixed design, gravity and buttresses, is based on

marl-limestone soils.
In 1968, the dam was raised by 6 meters, increasing the volume of the reser-
voir from 197 to 297 Mm3. This compensated for the decrease in the regulated
volume resulting from silting, but also increased the irrigated area by 3700 ha, while
improving the rolling of the floods and increasing the annual electricity production to
33 million kWh.
In addition to the necessary adaptations of the ancillary works, the raising of

the dam required the execution of 77 prestressed tie-rods of 240 t initial tension
504
Q.106 - R.31
which were installed in boreholes with a diameter of 130 mm totaling 4300 m in

length.
The VSL type prestressed tie rods are anchored 15 to 20 m in the foundation
rock and have thus ensured the total prestressing of the structure: the forces due to
the raising of the dam are entirely taken up by the prestressing force. The length of
the rock anchors was defined following two pull-out tests and set at 5 m.
Four control tie rods with conical cylinders at their heads were installed to
monitor the behaviour of the rock anchors. Two of these tie rods are located on the
left wing of the dam and two are placed on the weir of the spillway.
Since 1969, a periodic control of the behaviour of the prestressed tie rods has
been carried out by means of Lift-off tests which consist in measuring the pressure
applied to the jack, necessary for the lifting of the head of the rod by 12 mm corre-
sponding to the end of the stroke of the jack and comparing it to the initial pressure
of the measurement carried out under the same conditions after the initial tensioning
of the rod. The relaxation limit set by the manufacturer is defined at 12%.
The relaxation rates measured on the control rods since 1970 are on the
whole lower than the 12% limit value. However, problems of access to the tie rods
on the spillway are encountered, which hinders the realization of the relaxation
measurements of these tie rods. Solutions for automating the measuring system to
take remote readings are recommended.
1. INTRODUCTION
Le barrage El Kansera réalisé entre 1926 et 1934 est le plus ancien barrage
réservoir du Maroc, situé sur l’Oued Beht, affluent de l’oued Sebou, à environ 20
km au Sud de la ville de Sidi Slimane.
Cet ouvrage en béton de conception mixte, poids et contreforts, est fondé sur
des terrains marno-calcaires.
En 1968, le barrage a été surélevé de 6 mètres, faisant passer le volume de

la retenue de 197 à 297 Mm3 . La surélévation du barrage a nécessité l’exécution
de 77 tirants précontraints de 240 t de tension initiale qui ont été mis en place dans
des forages d’un diamètre de 130 mm totalisant 4300 m de longueur.
Les tirants précontraints du type VSL sont ancrés de 15 à 20 m dans le rocher

de fondation et ont ainsi permis d’assurer la précontrainte totale de l’ouvrage.
Quatre tirants témoins de contrôle munis de vérin conique à leur tête ont été
mis en place permettant de suivre ultérieurement le comportement de ceux-ci, dont
505
Q.106 - R.31
deux tirants sont implantés sur l’aile gauche du barrage et deux sont placés sur le
déversoir de l’évacuateur de crues.
Depuis 1969, un diagnostic et contrôle périodique du comportement des

tirants précontraints témoins est réalisé moyennant des essais Lift-off. La relaxation
limite fixée par le constructeur est définie à 12%.
2. PRÉSENTATION DU BARRAGE
Le barrage El Kansera est un ouvrage en béton de conception mixte, poids et

contreforts, de hauteur initiale maximale de 63 m et de longueur en crête de 170 m.
La majeure partie (aile droite, déversoir et partiellement l’aile gauche) est

de type gravité alors que la zone des conduites forcées est à contreforts. Cette
conception mixte est due à la géologie. En effet, c’est un des premiers barrages
construit en terrain marno-calcaire et la qualité du rocher rencontré n’a pas permis
à l’époque d’exécuter un barrage poids sur toute la longueur.
L’aménagement était destiné d’une part à l’irrigation d’un périmètre d’environ

25 000 hectares près de Sidi Slimane et la protection de celui-ci contre les crues
de l’oued Beht et, d’autre part, la production d’énergie électrique de 15.000 kWh
annuels grâce à la réalisation d’une usine hydro-électrique de pied.
En 1968, cet ouvrage fut surélevé de 6 mètres, faisant passer le volume de

la retenue de 227 à 297 Mm3 . Ceci permettait non seulement de compenser la
diminution du volume régularisé résultant de l’envasement mais encore, d’accroître
la superficie irriguée de 3700 ha tout en améliorant le laminage des crues et en
faisant passer le productible électrique annuel à 33 millions de kWh.
La surélévation du barrage a nécessité le haussement du déversoir de 6m, les

profils types des ailes de 5m, et l’exécution de 77 tirants ancrés sur 15 à 20 m dans
la fondation et dans le rocher sur une profondeur de 5 m dont la force de mise en
tension était de 240 tonnes. 48 d’entre eux ont nécessité un traitement par injection
de la zone d’ancrage.
Ce choix prévalait sur une surélévation traditionnelle par élargissement de la

section du barrage sur toute sa hauteur, puisqu’elle a permis d’éviter la modification
de la largeur de base du barrage, de garantir la stabilité contre le renversement, et
d’éviter l’apparition de contraintes de traction sur le parement amont du barrage.
Les tirants précontraints ont ainsi permis d’assurer la précontrainte totale de

l’ouvrage : les efforts dûs à la surélévation du barrage sont entièrement balancés
par l’effort de précontrainte.
506
Q.106 - R.31
Fig. 1
Coupe transversale du barrage El Kansera surélevé.
Transverse section of the raised El Kansera Dam.
1- Barrage existant 1- Existing dam

2- Surélévation 2- Raising
3- Tirants 240 tonnes 3- 240 tons tie-rods
4- Piézomètres existants 4- Existing piezometers
Fig. 2
Vue du barrage El Kansera surélevé.
View of the raised El Kansera dam.
507
Q.106 - R.31
3. DONNÉES DESCRIPTIVES DES TIRANTS DE PRÉCONTRAINTE
Les tirants précontraints mis en place sont identiques du type VSL et

présentent les caractéristiques suivantes :
- Profondeurs d’ancrage : 5 m ;
- Force de mise en tension : 240 tonnes ;
- Section totale : 1953 mm2 ;
- Constitué de 21 torons de 1/2 ;
- Allongement de rupture 4% ;
- Tension de rupture : 180 kg/mm2 , (1800 MPa) ;
- Limite élastique : 160 kg/mm2 , (1600 MPa) ;
- Relaxation à 120h : 4% ;
- Relaxation à 1000h : 8,5%.
Un inconvénient inhérent au recours à la précontrainte a néanmoins dû être

appréhendé : il s’agit du phénomène de relaxation qui se caractérise par une diminu-
tion des contraintes dans le temps dans une pièce soumise à une déformation
donnée.
Aussi, quatre tirants de contrôle ont été prévus afin de suivre l’évolution de la
force de précontrainte dans l’ouvrage. Il s’agit du:
- Tirant ndeg 17 dans le plot ndeg 8, sur l’aile gauche du barrage ;

- Tirant ndeg 24 dans le plot ndeg 6, tirant de référence dans la partie contrefort
du barrage ;
- Tirant ndeg 40 dans le plot du déversoir ndeg 4 ;
- Tirant ndeg 58 dans le plot ndeg 2 du déversoir.
A la différence des autres, les quatre tirants de contrôle n’ont pas été injectés
en leur partie supérieure. Pour protéger les câbles contre la corrosion, un produit
mou a été mis en place à l’intérieur de la gaine de protection du tirant avant la mise
en tension de celui-ci.
Chacune des têtes d’ancrage de ces tirants a été équipée d’un vérin torique
spécialement conçu. Mis en place entre la tête et la plaque d’appui, ce vérin permet
de mesurer en tout temps la tension du câble. Les caractéristiques de ces vérins
sont présentées ci-dessous :
- Référence : Vérin VSL 12868, Fabrication ndeg 8903, Ndeg série : 68319
- Section du piston : 508.9 cm2
- Course maximale du vérin : 12 mm
- Pression maximale de l’essai : 550 Kg/ cm2 .
508
Q.106 - R.31
Fig. 3
Implantation des têtes des tirants de contrôle dans le corps du barrage.
Implantation of the control rod heads in the body of the dam.
(a) Repères géodésiques (a) Geodesic landmarks
(b) Drains (b) Drain
(c) Piézomètres (c) Piezometer
(d) Pendules (d) Pendulum
(e) Tirants de contrôle (e) Control tie-rods
4. DIAGNOSTIC DES TIRANTS DE PRECONTRAINTES
Depuis 1969, un diagnostic des têtes des tirants précontraints témoins est
effectué de manière périodique ainsi que des essais de pesage ou lift off pour le
suivi du comportement de ces tirants.
Les dernières interventions de diagnostic et de pesage des tirants précon-

traints ont été réalisés au mois de juin 2020 par le bureau d’études Alkhibra pour le
compte de la Direction des Aménagements Hydrauliques (Maroc).
Les mesures de tension des tirants précontraints sont faites à l’aide d’une
pompe manuelle. Un système de mesure à distance n’a pas été prévu dans le
projet de surélévation du barrage.
Les tirants témoins sont bien marqués et facilement repérables tandis que les
têtes des autres 73 tirants ne sont ni accessibles ni repérables au niveau de la crête
du barrage.
509
Q.106 - R.31
Outre l’inspection des têtes de tirants, les opérations de diagnostic compren-

nent également l’inspection des manomètres de pompe, ceux des vérins ; les
manomètres étalons et des raccords hydrauliques aux vérins.
Il est constaté que l’accès pour effectuer l’inspection et les mesures reste
pénible pour les deux tirants situés au niveau du seuil déversant.
La tête de l’un des tirants situés au niveau du déversoir n˚40 (plot ndeg 4),
présente des signes prononcés de corrosion, ce qui se manifeste par la détérioration
des embouts soudés sur la tête du vérin servant au raccordement de la pompe et
du manomètre : Capuchon fileté de protection complètement rouillé et détérioré.
La corrosion constatée au niveau de ce tirant témoigne de fuites d’étanchéités

dans la niche abritant la tête du tirant.
Fig. 4
Vue des tirants témoins sur le déversoir.
View of the control tie-rods on the spillway.
5. ESSAIS DE PESAGE ET INTERPRÉTATION DES MESURES
5.1. CONDUITE DES ESSAIS DE PESAGE
L’essai de pesage se fait moyennant une pompe manuelle munie d’un

manomètre de mesure de pression. La pompe est raccordée sur l’une des deux
sorties réservées sur le vérin. Sur l’autre sortie, un deuxième manomètre est rac-
cordé au vérin. Il consiste en la mise en pression du vérin de pesage jusqu’à arrêt
510
Q.106 - R.31
de la pompe (arrêt de l’aiguille de son manomètre). Ce qui indique l’atteinte par la

tête du vérin d’un déplacement maximal défini par le constructeur à 12 mm.
Lors des visites de diagnostic des tirants témoins de contrôle de la pré-

contrainte du barrage, il a été procédé au pesage de l’ensemble des quatre
tirants.
Les mesures se sont déroulées avec succès pour trois tirants, alors que le
quatrième tirant ndeg 40 au niveau du plot 4 du déversoir a affiché un comporte-
ment anormal (pas de montée de pression). Les embouts soudés sur la tête du
vérin servant au raccordement de la pompe et du manomètre sont en effet bouchés
(capuchon fileté de protection complètement rouillé et détérioré). Il importait donc
d’enlever le doute sur le circuit de l’huile puis de procéder à la vérification du
fonctionnement du vérin.
Après les opérations de débouchage, de nettoyage et de purge, l’essai de

pesage a été reconduit avec succès.
Fig. 5
Conduite de l’essai de pesage du tirant précontraint témoin par pompe manuelle.
Conducting the weighing test by hand pump.
5.2. INTERPRÉTATION DES MESURES DE PRÉCONTRAINTE
L’interprétation des mesures de contrôles des tirants nécessite une compara-

ison entre les tensions de contrôle mesurées périodiquement, avec les valeurs
mesurées lors de la mise en tension (tension de clavage, "Iift off"), en tenant
511
Q.106 - R.31
Table 1
Niveaux de relaxation après pesage du 23/06/2020 (*)
TIRANT N◦ 58 TIRANT N◦ 40 TIRANT N◦ 24 TIRANT N◦ 17
PLOT 2 PLOT 4 PLOT 6 PLOT 8
Pression initiale de service (bar) 460 465 460 467
Pression lue pompe (bar) 415 415 430 425
Pression lue vérin (bar) 410 400 415 410
Pression réelle: moyenne (bar) 412.5 407.5 422.5 417.5
Pertes (bar) 47.5 57.5 37.5 49.5
Pertes (%) 10.33% 12.37% 8.15% 10.60%
(*) Relaxation levels after weighing on 23/06/2020.
compte des phénomènes influençant le comportement des tirants et la précision

des résultats (frottement, relaxation, etc.).
Pour le cas du barrage El Kansera, la relaxation limite fixée par le constructeur

est définie à 12%.
Les tests de pesage effectuées à l’occasion du contrôle des tirants témoins

de la précontrainte ont montré des valeurs de pressions qui varient de 400 à 415
bars au niveau du manomètre du vérin, alors que celui de la pompe a affiché des
valeurs supérieures avec des écarts de 5 à 15 bars.
L’évaluation du niveau de relaxation actuel atteint au niveau des points de

mesures lors des essais de pesage est synthétisée dans le tableau suivant :
Une analyse statistique de l’ensemble des mesures disponibles depuis 1974

à 2020 a révélé que les pertes de tension moyennes varient de 9.6 à 11 %, toutefois,
elles dépassent parfois la limite admissible de 12% fixé par le fournisseur. On con-
state, en outre, que l’évolution de la pression dans les tirants de contrôle n’a pas de
liaison claire tant avec le niveau de la retenue qu’avec les variations saisonnières.
6. CONCLUSION
L’évolution des mesures de tension des tirants précontraints témoins du bar-

rage El Kensera et par conséquent la relaxation dans ces derniers restent dans un
domaine acceptable et témoigne d’un état convenable des tirants.
La contribution des tirants précontraints à la stabilité du barrage n’a pas faibli

depuis leur installation comme en témoigne les mesures de pesage effectuées.
Le nombre de tirants de contrôle installé au niveau du barrage est très réduit

et doit être maintenu.
512
Q.106 - R.31
Compte tenu des problèmes rencontrés d’accès aux tirants sur le déversoir,
la réalisation des mesures périodiques de leur relaxation se trouve entravée. Dans
le cas de retenue pleine, des risques liés aux déversements sont à craindre. C’est
dans les situations exceptionnelles (avènement de crues, déversement important)
que l’information sur l’état du barrage devient très utile et ne doit pas s’arrêter.
Des solutions d’automatisation du système de mesure pour effectuer des

lectures à distance ont été recommandées.
L’outillage nécessaire à la réalisation des mesures de contrôle doit être péri-

odiquement entretenu (pompe, étalonnage des manomètres, utilisation d’huile
adéquate…) ainsi que la protection contre la corrosion des bouchons des embouts
et équipements en tête des tirants.
513
Q.106 - R.32
-------
-------
CHARACTERISTICS OF MICROSEISMIC b-VALUE ASSOCIATED WITH

LARGE DEFORMATION OF ROCK MASS IN HIGN GEOSTRESS
UNDERGROUND POWERHOUSE CAVERNS (∗)
Peiwei XIAO
State Key Laboratory of Hydraulics and Mountain River Engineering, College of
Water Resource and Hydropower, SICHUAN UNIVERSITY, Chengdu
CHINA GUODIAN JINSHAJIANG XULONG HYDROPOWER DEVELOPMENT
CO., LTD, Chengdu
Biao LI
School of Geoscience and Technology, SOUTHWEST PETROLEUM
UNIVERSITY, Chengdu
Xingguo YANG
State Key Laboratory of Hydraulics and Mountain River Engineering,
College of Water Resource and Hydropower, SICHUAN UNIVERSITY, Chengdu
Nuwen XU
State Key Laboratory of Hydraulics and Mountain River Engineering,
College of Water Resource and Hydropower, SICHUAN UNIVERSITY, Chengdu
CHINA
SUMMARY
The large deformation of rock mass in underground powerhouse caverns

has been a severe hazard of the hydropower engineering in Southwest China.
During the development of rock mass large deformation, a series of fractures
is generated, which can be monitored using the microseismic (MS) monitoring
technique. An MS monitoring system was established in Houziyan high geostress
underground powerhouse caverns, Southwest China. The MS clusters associated
∗ Caractéristiques de la valeur microsismique b associée aux déformations du rocher
fortement sollicité des cavernes des centrales électriques souterraines
514 DOI: 10.1201/9781003211808-127

Q.106 - R.32
with large deformation of rock mass were delineated and the MS b-values were
comparatively analysed. In addition, the evolution processes of the MS b-values
during the development of rock mass large deformation were researched. Results
showed that the MS b-value associated with large deformation of rock mass was
less than 1.0 in underground powerhouse caverns. Prior to the large deformation of
rock mass in the underground powerhouse caverns, the MS b-values in both high
stress underground powerhouse caverns decreased, which indicated that the pro-
portion of MS events with large magnitude increased. The results are meaningful to
understand the fracturing characteristics of MS sources associated with rock mass
large deformation and provide significant references for early warning of rock mass
large deformation in underground powerhouse caverns.
RÉSUMÉ
La grande déformation de la roche environnante des cavernes souterraines

des centrales électriques est devenue un grave danger pour les projets hydroélec-
triques du sud-ouest de la Chine. Dans le processus de déformation importante de
la masse rocheuse, une série de fissures sera générée, qui peuvent être surveil-
lées par l’utilisation de la technologie de surveillance microsismique. Un système
de surveillance MS a été mis en place dans les cavernes de la centrale souterraine
à forte sollicitation du sol à Houziyan, dans le sud-ouest de la Chine. Les agrégats
de spectre de masse liés à la grande déformation de la masse rocheuse sont délim-
ités, et la valeur b du spectre de masse est comparée et analysée. Le processus
d’évolution de la valeur MS-b lors du développement de grandes déformations de la
masse rocheuse est étudié. Les résultats montrent que la valeur MS-b des cavernes
souterraines des centrales électriques lorsque la roche environnante est déformée
est inférieure à 1,0. Avant la déformation massive des cavernes souterraines de la
centrale, les valeurs MS-b des deux cavernes souterraines de la centrale à haute
contrainte ont diminué, indiquant que la proportion d’événements de SP à haute
résistance a augmenté. Les résultats de la recherche sont d’une grande importance
pour comprendre les caractéristiques de fracture des sources de spectromètres de
masse liées aux grandes déformations des masses rocheuses, et fournissent une
référence importante pour l’alerte précoce des grandes déformations des cavernes
souterraines des centrales électriques.
1. INTRODUCTION
Large deformation of rock mass induced by the excavation unloading of large

underground powerhouse caverns will threaten the safety of personnel and equip-
ment, and also likely result in the deformation value of surrounding rocks exceeding
the operation request, which has been a significant problem during the construction
515
Q.106 - R.32
of the hydropower projects in Southwest China. Thus, it is of great significance for

underground surrounding rocks to identify the potential deformation zone, under-
stand the deformation mechanisms and conduct a real-time warning. During last a
few years, the microseismic (MS) monitoring technique has been applied to establish
the relationship between excavation-induced rock fractures (MS events) and large
deformation in underground powerhouse caverns [1]. Abundant rock fracturing infor-
mation such as magnitude, energy and apparent stress of the MS events can be
obtained using the MS monitoring technique. The magnitude-frequency relationship
of MS events reflects the evolution of fracturing scales in the failure process of rock
mass. Thus, it is significant to study the magnitude-frequency relationship of MS
events for rock mass large deformation.
The magnitude-frequency characteristics were firstly studied in seismology.

Since 1944, Gutenberg and Richter [2] had studied the magnitude-frequency char-
acteristics of the seismic events for more than twenty years in the California-Nevada
region and proposed the well-known G-R law. They defined the b-value to describe
the relationship between the seismic magnitude and frequency. Wyss et al [3] pointed
out that along fault zones the low b-values seemed to correspond with asperities
while high b-values were related to creeping sections of faults. Moreover, the rela-
tionships were revealed between b-values and other parameters such as faulting
style [4] and earthquake depth [5]. As for engineering microseismicity, Legge and
Spottiswoode [6] evaluated spatial characteristics of b-values of mining-induced MS
events and found that low b-values were potentially induced by microseismic events
distributed in two dimensions, while high b-values might be resulted from three-
dimensional failures. Liu et al [7] have revealed that prior to large seismic events
in mines (such as rockbursts, roof collapses, etc.), the b-values showed a gradual
tendency to decrease.
Therefore, it can be seen that the b-value can provide important information
on the potential hazards. In the present study, a typical underground engineering
at high geostress level in Southwest China was selected as the object. Through
analysing the b-values associated with rock mass large deformation, the b-value in
high geostess underground caverns was investigated. Moreover, the b-value varia-
tions were revealed during rock mass large deformation, which contributes to laying
the foundation for the early warning of large deformation hazards in underground
powerhouse caverns.
2. MS MONITORING DATA
2.1. ENGINEERING BACKGROUND
The Houziyan hydropower project was constructed on the Dadu River in

Sichuan Province, China. The underground powerhouse caverns were designed in
516
Q.106 - R.32
the right bank mountain, with a total installed capacity of 1700 MW. The underground
powerhouse caverns mainly include the main powerhouse, the main transformer
chamber, and the tailrace surge chamber. The caverns were excavated by layer and
their excavation sizes are 219.5 m × 29.2 m × 68.7 m (length × width × height),
139 m × 18.80 m × 25.2 m (length × width × height), and 140.5m × 23.5 m ×
75.0 m (length × width × height), respectively. The underground powerhouse is gen-
erally at a depth of 280-510 m horizontally and 400-660 m vertically. The geo-stress
measurement results show that the maximum principal geo-stress in the under-
ground powerhouse caverns is 21.46-36.43 MPa and the rock mass strength-
to-stress ratio is only 2-4, which is a typical high geo-stress example of large
hydropower projects [8]. During the excavation of the caverns, weak rockbursts
and rock failures occurred at times.
2.2. MS MONITORING SYSTEM
An MS monitoring system was established to monitor the evolution process of

microfractures during the large deformation of rock mass of the underground power-
house caverns. Sixteen acceleration sensors were installed in the layer drainage gal-
leries and the upstream sidewall of the main transformer chamber, as shown in Fig. 1.
Fig. 1
Spatial layout of sensors of Houziyan underground powerhouse caverns.
517
Q.106 - R.32
2.3. CALCULATION OF THE B-VALUE
The wellknown G-R law proposed by Gutenberg and Richter [3] well reflects
the relation between seismic magnitudes and frequency, expressed as:
lgN = a − bmM (1)
where N is the event count whose magnitude is no less than mM , a and b are
constants for the given statistical data and can be acquired by the lgN − mM
diagram. The parameter b, also called b-value, represents the proportion of high
and low magnitude of events. Large b-values usually mean low-magnitude events
are dominant while small b-values indicate high-magnitude events occurred more
frequently.
3. B-VALUES ASSOCIATED WITH ROCK MASS LARGE DEFORMATION
3.1. CALCULATING RESULTS OF B-VALUES
The MS monitoring system had been in operation since April 12, 2013, when
the fouth bench of the main powerhouse was to be excavated. Myriads of MS
events induced by excavation unloading were acquired in the Houziyan underground
caverns. During the MS monitoring period, the downstream sidewall of the main
powerhouse was the main deformation region. Four rock mass large deformation
problems associated with MS clusters are shown in Fig.2. It can be seen from the
absolute displacement process diagram of multipoint extensometers that the rock
mass large deformation had obvious step increments and the increasing values of
the displacements of rock mass excavation surfaces were respectively 9.89 mm,
6.35 mm, 7.90 mm, and 8.54 mm. A few days before and during the large defor-
mation of rock mass, the MS events accumulated around the deformation regions
(Fig.2). Thus, it can be verified that rock mass large deformation has a close relation
with MS activities.
The MS b-values calculated in Fig.3a and b correspond to the MS clusters in

Fig.2a and b. Using Equation (1), we can obtain statistics for mM and N . Linear fitting
is then implemented with mM as the x-coordinate and lgN as the y-coordinate. The
slope of this straight line is the b-value. The b-values associated with four rock mass
large deformation problems in Houziyan underground powerhouse are respectively
0.7821 and 0.8201.
518
Q.106 - R.32
Fig. 2
Rock mass large deformation in MS clustering region of Houziyan underground
powerhouse caverns. (a) MS cluster recorded from April 16, to May 20, 2013 and
the absolute displacement process diagram of multipoint extensometer M4CF 3-6,
(b) MS cluster recorded from July 4 to July 28, 2013 and the absolute
displacement process diagram of multipoint extensometer M4CF 3-5.
Fig. 3
b-values of MS clusters associated with rock mass large deformation of in
Houziyan underground powerhouse caverns. (a) and (b) are the calculating results
respectively corresponding to (a) and (b) in Fig. 2.
519
Q.106 - R.32
3.2. TEMPORAL VARIATIONS OF B-VALUES DURING THE DEVELOPMENT OF ROCK

MASS LARGE DEFORMATION
The temporal variations of b-values during the development of rock mass large
deformationis shown in Fig.4. During the development of rock mass deformation, the
b-values decreased prior to the rock mass large deformation. Take the rock mass
large deformation occurring from May 2 to May 14, 2013 for example, as shown
in Fig.4. The b-value decreased from 0.8282 (ten days before the rock mass large
deformation), to 0.7852 (five days before), and to 0.7312 (one day before). During
the large deformation, the b-value was 0.7792. Before the large deformation, the
intensity of rock fracturing increased, resulting in the increasing proportion of high-
magnitude MS events. Thus, the decrease of b-value can be used as a significant
index for early warning of rock mass large deformation in underground powerhouse
caverns.
Fig. 4
Evolution of b-values during the development of rock mass large deformation in the
Houziyan underground caverns.
4. CONCLUSIONS
The MS monitoring technique was applied in the high geostress underground

powerhouse caverns at different stress levels. The MS clusters associated with rock
mass large deformation were delineated and the b-values of these MS clusters were
comparatively analysed. The b-value was less than 1.0 in high stress underground
caverns, which indicated that high-magnitude MS events occurred frequently in high
stress underground caverns.
During the development of the large deformation of the rock mass, the
evolution processes of MS b-values were investigated. Prior to rock mass large
deformation, the b-values gradually decreased, which could be regarded as an indi-
cator of potential large deformation of rock mass. Consequently, the evolution of
520
Q.106 - R.32
the b-value can provide significant references for early warning of rock mass large
deformation in underground powerhouse caverns.
ACKNOWLEDGMENTS
The authors are grateful for the financial support from the National Key R&D
Program of China (No. 2017YFC1501100), the National Natural Science Foundation
of China (Nos. 51809221 and 51679158).
REFERENCES
[1] DAI F., LI B., XU N. W., FAN Y. L., ZHANG C. Q. Deformation forecasting
and stability analysis of large-scale underground powerhouse caverns from
microseismic monitoring. International Journal of Rock Mechanics and Mining
Sciences, vol. 86, pp. 269–281, 2016.
[2] GUTENBERG B., RICHTER C. F. Frequency of earthquakes in California.

Bulletin of the Seismological Society of America, vol. 34, no. 4, pp. 185–188,
1944.
[3] WYSS M., SHIMAZAKI K., WIEMER S. Mapping active magma chambers by
b-values beneath the off-Ito volcano, Japan. Journal of Geophysical Research,
vol. 102, pp. 20413–20422, 1997.
[4] SCHORLEMMER D., WIEMER S., WYSS M. Variations in earthquake-size

distribution across different stress regimes. Nature, vol. 437, pp. 539–542,
2005.
[5] WIEMER S., MYSS M. Mapping the frequency-magnitude distribution in

asperities: An improved technique to calculate recurrence times. Journal of
Geophysical Research, vol. 102, pp. 15115–15128, 1997.
[6] LEGGE N. B., SPOTTISWOODE S. M. Fracturing and microseismicity ahead

of a deep gold mine stope in the pre-remnant and remnant stages of mining.
in 6th ISRM Congress, Montreal, Canada, pp. 2071–1078, 1987.
[7] LIU J. P., FENG X. T., LI Y. H., XU S. D., SHENG Y. Studies on temporal and
spatial variation of microseismic activities in a deep metal mine. International
Journal of Rock Mechanics and Mining Sciences, vol. 60, pp. 171–179, 2013.
[8] CHENG L. J., LI Z. G., WANG J. S., WANG D. K., XIA X. Design report on
reinforcement measures for surrounding rock mass in underground group cav-
erns of Houziyan hydropower station along Dadu River. HydroChina Chengdu
Engineering Corporation, Chengdu, China, pp. 1–163, 2014.
521
Q.106 - R.33
-------
-------
FIELD TEST OF SOFT SOIL PRELOADING IN THE ESTUARINE AREA OF THE

PEARL RIVER (∗)
Shuaidong YANG & Zhihuai HUANG

PEARL RIVER WATER RESOURCES RESEARCH INSTITUTE, Guangzhou
Xiaoliang WANG
South China Sea Institute of Planning and Environmental Research, SOA,
Guangzhou
Mi ZHOU
State Key Laboratory of Subtropical Building Science,
South China Institute of Geotechnical Engineering, SOUTH CHINA UNIVERSITY
OF TECHNOLOGY, Guangzhou
CHINA
SUMMARY
It is always a challenge for engineers to predict the settlement of soft sub-

grade under the action of heap load in estuarine area. This paper studies the
post-construction settlement of the soft soil foundation in the estuary area by the field
test of the pre-loading in the Pearl River Estuary area, reveals the settlement and
deformation characteristics of the pre-loading on the soft soil in the estuary area
through the field test, and establishes its post-construction settlement prediction
model, which provides guidance for similar projects in the later stage.
∗ Essai in situ de préchargement du sol dans la région de l’estuaire de la rivière des Perles.
522 DOI: 10.1201/9781003211808-128

Q.106 - R.33
RÉSUMÉ
Les propriétés de tassement des fondations meubles sous l’effet des charges
constituent un défi pour les ingénieurs, et la prévision précise du tassement après
la construction peut aider à orienter les pratiques d’ingénierie. Dans le cas de
l’estuaire de la rivière des Perles, le présent rapport étudie les problèmes de sédi-
mentation dans la zone estuarienne après la construction en étudiant les essais
de précompression et préchargement. Des essais in situ ont permis de mettre en
évidence les caractéristiques de déformation après tassement du sol meuble. Sur
la base de cette caractéristique, le modèle de prédiction de tassement après la
construction est établi pour fournir des orientations pour les projets ultérieurs du
même type.
1. INTRODUCTION
The sand drain preloading method proposed by the American engineer Moran
in 1925 [3], was the earliest form of surcharge preloading treatment technology. The
installation of sand drains (e.g. packed sand drain or normal sand drain) in the soft
soil foundation will enhance the permeability and accordingly, the permeability is
significantly enhanced. The water trapped in any soil formation could easily flow
out and the consolidation process occurs. With the wide application of sand drain
method in engineering, many scholars at home and abroad [2, 4–10] have carried
out a series of studies on this method.
Nowadays, more and more projects use plastic drainage boards as verti-
cal drainage channels in preloading. Compared with traditional sand drain, their
performance and quality are superior. In terms of construction, light construction
equipment can be used on soft soil foundations and the advantages are obvious, e.g.
the efficiency is higher, the construction is more convenient, the drainage capacity
is stronger, and the project cost is lower, natural foundation soil suffers less distur-
bance. Based on these superior properties, the plastic drainage boards has been
widely used. At present, their good reinforcement has been utilized in the soft soil
foundation treatment project of reclamation [1, 11–13].
This paper focus on the reinforcement of the complex soil foundations in the
estuary of the Pearl River by preloading method. Based on the field monitoring data,
the effects of the reinforcement on soft soil subgrade were analyzed according to the
settlement, horizontal displacement, and dissipation of the excessive pore pressure.
The deformation mechanisms were discussed and the corresponding prediction
method based on the in-situ monitoring data on the specific soil formation of the
estuary of the Pearl River was proposed.
523
Q.106 - R.33
2. BACKGROUND
This work relies on the real water system monitoring project in Nansha,
Guangzhou. The soil type of the foundation is a typical estuary soft soil, and its
specific parameters are as follows:
Table 1
Soil Parameters
Soil type Silt Silt clay Medium-Fine sand Granite Silt clay
Bulk density γ (kN/m3 ) 16.0 17.0 19 19.5 19.1
Saturated density γsat (kN/m3 ) 17.8 19.0 20 21.5 20.8
Hydraulic conductivity Kh (m/d) 1.1E-4 8.6E-5 5.2 1.5E-7 6.0E-5
Young’s modulus E (kPa) 2100 2100 6000 20000 1500
Poisson ratio ν 0.39 0.35 0.3 0.3 0.35
Cohesion c (kPa) 7.8 11.4 0 45 16
Friction angle ϕ (◦ ) 7.3 12.8 28 12 21.8
The plain soil is adopted as preloading. The elevation of the stacked soil is
8.5m. The clear height is about 4m, and the slope is 1:2.5. A geotextile is installed
between the stacked soil and the soft soil foundation. A600mm thick sand cushion
plays the role of water-conducting layer. The stacked soil is installed layer by layer
and the loading rate is adjusted according to the real situation in order to avoid the
damage caused by the unreasonable loading rate.
The plastic drainage boards are laid in an average depth of 20m, with an
interval of 1.2m, in a regular triangular distribution (as shown in Fig. 1a). After
the construction of the drainage board, the stacked load is added, as shown in
Fig.1b.
Fig. 1
In-situ photos.
524
Q.106 - R.33
Fig. 2
Schematic of preloading scheme.
3.1. SETTLEMENT OF THE GROUND
The log of the settlement for a typical section started on December 2, 2014,
and the preload ended on February 11, 2015. Until August 11, 2015, the preloading
equipment has safely run for 180 days, which satisfied the design requirement.
According to the monitoring data (Fig. 3), the settlement occurred with the start of the
stacking. When the stacking finished, the settlement was about 700mm. The moni-
toring continued until the settlement was stable. The maximum average settlement
was1083.1mm and the maximum average rate of change is 0.2mm/d. According to
Fig. 3, about 75% of the settlement occurred during the stacking stage. After the
stabilization of the stack, the settlement was relatively close at three different posi-
tions, which prove that 1/3 of the trapezoidal stacking section has almost the same
settlement. This typical settlement owing to the soil deformation of the preloaded
foundation in the estuary area belongs to the overall settlement, i.e., the settlement in
the middle is larger than that of the settlement on both ends, and the corresponding
failure mode is overall instability.
Fig. 3
Settlement of the typical section 0+700.
525
Q.106 - R.33
3.2. PORE PRESSURE
Pore pressure monitoring is mainly used to control the filling rate. Fig. 4 shows
the pore water pressure diagram of a typical section. During filling, the pore pres-
sure increases with the increase of the filling thickness, and the pressure gradually
dissipates in the later stage. According to the pore pressure monitoring data, the
ratio of pore pressure increment to load increment during filling was always less than
0.6, which meets the design requirements. From the graph of pore water pressure,
it can be seen that in the initial stage of stacking, the pore pressure increases very
rapidly, and it gradually dissipates after reaching the maximum filling height. The
evolution law of pore water pressure is consistent with the development of the foun-
dation surface. That is, in the initial stage of stacking, the pore pressure increases
significantly. As the stacking ends, the pore pressure gradually dissipates.
Fig. 4
Evolution of the pore pressure.
3.3. LAYERWISE SETTLEMENT
It can be seen from the layer wise settlement diagram (Fig. 5) that as the
depth increases, the settlement decreases, which shows that the reinforcement
range of preloading is limited. With the spread of the overload stress, the greater
the depth of the soil, the weaker the effect of the reinforcement. Therefore, the
surface soil has the best reinforcement effect, which is consistent with the previous
conclusions made on pore pressure. According to the monitoring data of layered
settlement, the section with the smallest settlement is 0+450, and the settlement
value is 803.5mm. According to the comparison of the data of surface settlement
and layered settlement, the surface settlement of layered settlement is closer to the
surface settlement, which also verifies the reliability of the observation results.
526
Q.106 - R.33
Fig. 5
Evoluation of the layerwise settlement.
3.4. DEEP HORIZONTAL DISPLACEMENT
Fig. 6 shows the deep horizontal displacement of the typical section. The data
is obtained from the Inclinometer tube. With the occurrence of the surcharge, the
Fig. 6
Settlement at the deep depth.
527
Q.106 - R.33
horizontal displacement gradually increases, and the main horizontal displacement

is concentrated at 20m below the ground surface. After the filling, the horizontal
displacement further develops with the consolidation of the soil while the magni-
tude became smaller. The maximum horizontal displacement value is 47.5mm. The
horizontal displacement of the initial heap load changes with the increase of the
thickness of the heap, and the change of the horizontal displacement gradually
stabilizes after full load.
3.5. HORIZONTAL DISPLACEMENT OF THE SURFACE
The horizontal displacement of the foundation surface is shown in Fig. 7. The

trend of the horizontal displacement at the foundation surface is basically the same
as that of at deep depth. At the initial stage of filling, the horizontal displacement
of the surface develops rapidly. Then, the horizontal displacement also gradually
develops before the height of the mound reaches the maximum. As a result, the
horizontal displacement of the surface soil has further developed. The development
trend of the horizontal displacement of the surface can also reflect that the soft soil
load settlement trend in the estuary area mainly occurs during the filling period, and
then it gradually changes until it stabilizes.
Fig. 7
Evolution of the horizontal displacemen.
4. PREDICTION MODELS FOR THE POST-CONSTRUCTION SETTLEMENT
4.1. TRADITIONAL PREDICTION MODEL
It is important to predict and analyze the long-term settlement and deforma-

tion of soft soil foundation based on the monitoring data. Firstly, a large amount of
long-term monitoring data is analyzed. Secondly, a relatively simple calculation
528
Q.106 - R.33
model is used to fit the relevant parameters. With the calibrated model, the set-
tlement and deformation in the future can be predicted. In general, the prediction
models can be divided into two categories: an empirical regression analysis method
and a theoretical method. The empirical regression analysis method utilize a rela-
tively simple mathematical function for fitting calculation. Because of its intuitiveness
and simplicity, it is a simplified calculation method for the settlement in practical engi-
neering. There are many types of commonly used empirical regression analysis
models, mainly including hyperbolic model [14], exponential model [15], logarithmic
model [16], logistic curve method [17], etc. The theoretical method is established
on a certain theoretical basis. The corresponding simplified formula is firstly formu-
lated and the measured data are then used to obtain the relevant parameters. The
calibrated model can be used to predict the further settlement, such as the Asaoka
method[18] , gray system [19], Hoshino Law [20], etc.
After many attempts and comparisons, it was found that the hyperbolic method
is the most appropriate one for the prediction of the settlement in the Estuary area of
the Pearl River. The ingredients of the hyperbolic method is described in detail here.
This method is applied for the settlement of the soft soil foundation under fill
load. For projects with a generally high height, the actual loading process line can be
roughly divided into a step shape, and the load is divided into two levels: P1 and P2,
which are applied at t1 and t2 respectively. Assuming that the settlement caused by
each level of load can be described by a hyperbola and has the same consolidation
characteristics, After summing up the settlement deformation corresponding to each
level, the total settlement can be expressed as:

m
tk
St = Sk∞ (1)
a + tk
k=1
Where is m the number of the loads applied. tk corresponds to the time when the k th
load is added. a is a parameter reflecting the consolidation rate of soils.
As it is assumed that the consolidation characteristics of the soft soil foundation

remain unchanged, the parameter a, which is related to the properties of the soil,

won’t change with the load. If t ≥ tk , then tk = t − tk ; if t < tk (the k th load has not

been applied), take tk = 0; Sk∞ is the final settlement corresponds to the k th load,
which is assumed to be proportional to the load Pk as:
Sk∞ = βPk (2)
Subsisting Eq. [2] into Eq. [1] leads to:

m
tk
St = βPk (3)
k=1
a + tk
The main task of the regression analysis method of the hyperbolic model is to
use the monitoring data under different load levels to back-analyze the appropriate
parameters α and β. Then the formula (6-12) can be used to predict the settlement
of the landfill at any time.
529
Q.106 - R.33
The hyperbolic form of the post-construction settlement prediction formula can

be expressed as:
t − t0
S t = S0 + (4)
α + β (t − t0 )
In the formula: S0 is the initial settlement; St is the settlement at time t; t0 is the

starting time, generally taken as the end of the filling construction; α and β are the
coefficients obtained by the curve regression of the monitoring data.
Eq. [4] is expressed as

t − t0
= α + β (t − t0 ) (5)
St − S0

After some manipulation, the relation between t − t0 St − S0 and t − t0 is plot.
And the parameters α and β are as follows:
It can be seen from Fig. 8 and Table 2 that there is no significant difference
for the intercept α obtained by numerical fitting according to the Eq. [5], while the
Table 2
Fitted Parameters
Cases L/h1 S/L h2/h1 α β
1 6 0.1 10 0.109 0.00417
2 11 0.055 10 0.10549 0.00457
3 15.5 0.039 10 0.08491 0.00508
4 10 0.09 10 0.0864 0.00426
5 10 0.12 10 0.07809 0.00395
6 10 0.15 10 0.07415 0.00378
7 10 0.06 2.5 0.1121 0.00568
8 10 0.06 5 0.08751 0.0049
9 10 0.06 10 0.10672 0.00471
Fig. 8
Fitting results by the hyperbolic model.
530
Q.106 - R.33
slope β is different. Therefore, the intercept α is fixed as 0.1.The slope β is affected

by three parameters. According to the data in Table 2, after curve regression, the
fitting formula (R2 = 0.95) is obtained as shown in Eq. [6], and the fitting results are
shown in Fig. 9:
−0.0047 −0.2117
L S 0.65
β = 0.00242 exp (6)
h1 L h2 h1
In this equation, L is the depth where the plastic drainage board installed. h1
is the thickness of the hard layer. S is the interval of the plastic drain board. h2 is the
thickness of soft clay (silt) interlayer.
Fig. 9
Fitting results by the hyperbolic model.
4.2. NEW MODEL FOR POST-CONSTRUCTION SETTLEMENT PREDICTION
According to the results of the parameter analysis in the previous section, the
revised post-construction settlement prediction formula can be expressed as:
t−t0
St = S 0 + −0.0047 −0.2117
0.1+ 0.00242 hL S
exp h0.65h (t−t0 ) (7)
1 L 2 / 1
Where S0 is the initial settlement, which is generally taken as the settlement at the
end of stacking. t0 is the end of stacking.
531
Q.106 - R.33
Eq. [7] takes into account the influence of the complex soil layer in the area of
the Pearl River, and on the basis of Eq. [4], three parameters related to the thickness
of the soft clay (silt) interlayer, the depth
of the
drainage
plate and the interval of
the drainage plate are introduced. h2 h1 , L2 h1 , S L, the new post-construction
settlement calculation formula is obtained.
Compared with the original Eq. [4], for the complex soil foundation in the
area of the Pearl River, the proposed model applies not only to a single soil layer,
but also to multi-layer complex foundations. The calculation process no longer
requires
curve regression. By substituting the formation information parameter

h2 h1 and the relevant parameters of the plastic drainage plate L h1 , S L, the
prediction result could be directly obtained based on the initial settlement S0 and
the initial time t0 after the completion of the pile load obtained from the on-site
monitoring.
4.3. MODEL VALIDATION
We took the monitoring data of the section 0+600 to verify the applicabil-
ity and accuracy of the newly proposed model. The calculated results are as
follows:
Fig. 10
Comparison between monitoring data and predictions by the proposed model
(0+600).
It can be seen from Fig. 10 that the proposed model in this work can accu-
rately reflect the trend of the settlement, and accurately predict the settlement after
construction at any time.
532
Q.106 - R.33
5. CONCLUSIONS
In this paper, based on the existing calculation method for the settlement of
soft soil foundation after loading and the monitoring data, a hyperbolic model applies
to the complex foundation settlement characteristics of the Pearl River Estuary is
proposed:
(1) Based on the settlement data obtained in the previous article, considering
the influence of the thickness of the soft soil interlayer in the foundation, the inter-
val and depth of drainage board on post-construction settlement, the introduction
of three influencing factors is related. The existing model of hyperbolic method is
revised and a new formulation is proposed.
(2) The new calculation formula can be applied to multi-layer complex foun-
dations, without the necessitate of curve regression analysis. Based on the initial
settlement after the completion of the heap load, the relevant parameters can be
substituted to predict the post-construction at any moment.
REFERENCES
[1] Liu Hanlong, Zhao Minghua. Review of ground improvement technical and its
application in China[J]. CHINA CIVIL ENGINEERING JOURNAL, 2016, 49(1):
96–115.
[2] Ye Shulin, Ye Guanbao. Foundation treatment and underpinning technology

[M]. China Architeclure Publishing & Media Co., ltd., 2005.
[3] Luo Jiajin. Study on Reinforcement Mechanics and Consolidation Settlement

Analysis of Solf Foundation Treatment with the Method of Prefabrieated Drain
with Preloding[D]. Central South University, 2013.
[4] Wang Yi-sun. Discounted Well-diameter Method to Take the Effect of Drain
Well Resistance into Consideration[J]. Chinese Journal of Geotechnical
Engineering, 1982, 4(3): 45–54.
[5] Xie Kanghe. Sand well foundation:consolidation theory.Numerical analysis

and optimization design[D]. Zhejiang University, 1987.
[6] Fang Yingguang. Analysis of Consolidation Deformation of Layered Saturated

Clay Withsand Drains[J]. CHINA CIVIL ENGINEERING JOURNAL, 1997,
30(5): 49–51.
[7] Indraratna B., Redana I. W. Plane-Strain Modeling of Smear Effects Asso-

ciated with Vertical Drains[J]. Journal of Geotechnical & Geoenvironmental
Engineering, 1997, 123(5): 474–478.
533
Q.106 - R.33
[8] ZHAO Wei-bing, AI Ying-bo, ZHANG Jing. Study on post-construction set-

tlement of thick soft roadbed improved by drainage-preloading[J]. HYDRO-
SCIENCE AND ENGINEERING, 2003(1): 28–33.
[9] Lou Rong. Study on the Simplified Methods for the Numerical Analysis of
Sand-Drained Ground[D]. Zhejiang University, 2006.
[10] Yoshikuni H., Nakanodo H. CONSOLIDATION OF SOILS BY VERTICAL

DRAIN WELLS WITH FINITE PERMEABILITY[J]. Soils & Foundations,
2008,14(2): 35–46.
[11] Gao Zhiyi, Miao Zhonghai. Nanning airport soft soil foundation vacuum
precompression construction[J]. China Harbour Engineering, 1992(1): 18–22.
[12] Yu Zhiqiang. Application of vacuum combined loading preloading method of

cipher-plastic drainage board in Tianjin Nanjiang Port Area[J]. China Harbour
Engineering, 1996(2): 19–24.
[13] Liu Zhibao. Application of vacuum preloading method to Reinforcing soft soil
foundation on hangzhou Highway[J]. SUBGRADE ENGINEERING, 1997(5):
40–44.
[14] Yin Zongze. Earthwork principle[M]. China Water Power Press, 2007.
[15] Shi Yisheng, Yao Chenghuan, Jiang cheng. Discussion on the Settlement
Calculation of Subgrade Under High Fill in Airports[J]. Soil Engineering and
Foundation, 2001, 15(3): 22–25.
[16] Sowers G. F. Compressibility of broken rock and the settlement of rockfills:

International Conference on Soil Mechanics and Foundation Engineering[C],
Montreal, Canada, 1965.
[17] Zhu Zhiduo, Zhou Lihong. Application of Logistic model in settlement prediction
during complete process of embankment construction[J]. Chinese Journal of
Geotechnical Engineering, 2009, 31(6): 965–969.
[18] Asaoka A. Observational procedure of settlement prediction[J]. Soils & Foun-

dations, 2008,18(4): 87–101.
[19] Zhang Hui, Gao De-li. Multilevel Comprehensive Evaluation Method is Based
on the Fuzzy Mathematics and Grey Theory[J]. MATHEMATICS IN PRACTICE
AND THEORY, 2008, 38(3): 1–6.
[20] TANG Zhaohail, LIU Xiao, LI Jufeng. Experiential Curve Nonlinear Regression
Model for Sediment Prediction of Soft Soil Foundation[J]. Geological Science
and Technology Information, 2005, 24(s1): 106–108.
534
Q.106 - R.34
-------
-------
LASERGRAMMÉTRIE ET VIDÉO-TACHÉOMÉTRIE AU SERVICE DE LA

SURVEILLANCE ET DE LA SÉCURITÉ DES BARRAGES (∗)
Franck SCHMIDT
Institut d’ingénierie du territoire, heig-vd
Jonathan FAURIEL
Surveillance et sécurité des barrages, ALPIQ SA
Jean-Claude KOLLY
Surveillance et sécurité des barrages, Groupe E
Reynald BERTHOD
Département barrages, Stucky SA
Vincent BARRAS
Directeur, 3sigmas SA
SWITZERLAND
RÉSUMÉ
La lasergrammétrie et la vidéo-tachéométrie sont deux technologies qui per-

mettent de figer l’ouvrage à un moment donné et de limiter l’impact sur l’exploitation
hydraulique. Grâce à la vidéo-tachéométrie il est possible de réduire le temps
d’acquisition d’un réseau géodésique de l’ordre de 50%. Ce gain de temps ne
réduit pas la précision, car les mesures sont réalisées à l’identique d’une campagne
géodésique traditionnelle et traitées avec les mêmes logiciels de compensation par
les moindres carrés.
Avec la lasergrammétrie, il est possible d’acquérir en quelques heures un

ouvrage, avec une importante densité de points. L’avantage, est la mise en évidence
∗ Lasergrammetry and video-tacheometry for surveillance and security of dams.
DOI: 10.1201/9781003211808-129 535

Q.106 - R.34
de déplacements sur tout l’ouvrage avec une précision de l’ordre de ±3mm. Cette
méthode présente toutes un inconvénient : la difficulté à mettre en évidence
les déplacements tangentiels. Des tests complémentaires sont en cours pour y
remédier.
Pour le génie civil, les cartes de colorisation des mouvements sur tout
l’ouvrage peuvent être utilisées comme calage d’un modèle à l’aide d’éléments finis
conçus et utilisés pour comprendre le comportement de l’ouvrage. D’autre part,
cela permet également de mettre en évidence des anomalies structurelles sur les
parties de l’ouvrage non instrumentées, et de mieux comprendre le comportement
de l’ouvrage.
SUMMARY
Lasergrammetry and video-tacheometry are two technologies that make it

possible to freeze the structure at a given moment and limit the impact on hydraulic
operations. With the video-tacheometry it is possible to reduce the acquisition time
of a geodetic network by around 50%. This time saving does not reduce accuracy,
as the measurements are carried out in the same way as a traditional geodetic
campaign and processed with the same least-squares compensation software.
With lasergrammetry, it is possible to acquire a work with a high density of

points in a few hours. The advantage is the demonstration of movements over the
entire work with a precision of ±3mm. The disadvantage of this method is that it
is difficult to detect tangential movements. Complementary tests are in progress to
remedy this.
For civil engineering, colour maps of the movements over the entire struc-
ture can be used as a basis for a model using finite elements designed and used
to understand the behaviour of the structure. On the other hand, this also makes
it possible to highlight structural anomalies on the non-instrumented parts of the
structure, and to better understand the behaviour of the structure.
1. INTRODUCTION
L’auscultation fait partie intégrante de la vie d’un ouvrage. Actuellement l’état

de l’art est réalisé par un réseau et une campagne de mesures géodésiques. Cepen-
dant les mesures géodésiques, classiques, comportent des points faibles. Depuis
plusieurs années de nouveaux capteurs d’acquisition existent, tels que les scanners
lasers 3D et la vidéo-tachéométrie.
Ce document a pour but de rappeler l’état de l’art de la technique de mesures

géodésiques, puis avec trois exemples de démontrer la valeur ajoutée des nouvelles
technologies dans la surveillance des ouvrages d’art.
536
Q.106 - R.34
2. ÉVOLUTION DES BESOINS POUR LA SURVEILLANCE
Le vieillissement des ouvrages en Suisse est un défi majeur pour l’ingénierie

des barrages. La surveillance et la sécurité des ouvrages d’accumulation évoluent
avec le temps et doivent régulièrement s’adapter à de nouveaux challenges. En
particulier, certaines pathologies du béton et leurs conséquences sur la sécurité
des ouvrages sont difficiles à caractériser. Les nouvelles technologies permettent
de mieux connaître le comportement de la structure, de la décrire avec précision
pour pouvoir trouver des solutions ingénieuses tout en respectant la sécurité des
biens et des personnes.
Dans ce contexte, les techniques de lasergrammétrie de vidéo-tachéométrie

peuvent être de précieux outils permettant de répondre à différentes questions
posées lors des analyses d’ouvrages. Elles peuvent, en particulier, apporter des
solutions afin de :
• Permettre des mesures plus rapides, ce qui limite les contraintes sur
l’exploitation de l’aménagement.
• Saisir, sous certaines conditions, les déplacements sur la totalité de la
superficie du barrage et de ses appuis.
• Saisir les déplacements particuliers de parties d’ouvrage et anticiper
un suivi.
• Améliorer les modèles de calcul permettant de comparer les déplace-
ments mesurés et calculés en améliorant la connaissance de l’ouvrage à
surveiller.
3. LA MESURE GÉODÉSIQUE CLASSIQUE
3.1. OBJECTIFS DES MESURES GÉODÉSIQUES [1]
La surveillance d’un ouvrage, par un réseau géodésique, a pour objectif de

déterminer, les coordonnées absolues de points de surveillance, dans un cadre de
référence supposé fixe tout au long de la vie de l’ouvrage. A l’aide de ces coor-
données, il est alors possible de calculer et de quantifier le déplacement absolu de
l’ouvrage. La précision de la mesure doit être telle que l’incertitude de mesure soit
fortement inférieure aux mouvements de l’ouvrage.
Les coordonnées des points de surveillance, matérialisés par des cibles

ou prismes, sont déterminées par mesures terrestres (angles et distances), niv-
ellement géométrique et/ou GNSS. Le réseau de mesures se compose aussi de
points de référence qui déterminent un cadre stable et pérenne. Les coordon-
nées calculées de chacun des points doivent être accompagnées d’indicateurs
statistiques de précision et de fiabilité comme les intervalles de confiance sur les
déplacements.
537
Q.106 - R.34
3.2. TRAITEMENT DES MESURES GÉODÉSIQUES
Les différentes techniques d’acquisition ne fournissent pas directement des

positions précises et fiables dans le cadre de référence de l’ouvrage. Les mesures
enregistrées durant la campagne de relevés doivent être traitées et combinées,
afin d’aboutir aux coordonnées des points de surveillance. Ces coordonnées sont
ensuite comparées avec les états précédents afin de quantifier les mouvements de
l’ouvrage.
Les méthodes de calcul modernes permettent de traiter les observations par

la méthode des moindres carrés. À chaque étape du traitement, des indicateurs
statistiques rigoureux sont calculés validant ou non les mesures. Il est alors possible
d’éliminer les fautes et d’obtenir des coordonnées optimales d’un point de vue statis-
tique. Les méthodes par moindres carrées fournissent également des indicateurs
de précision et de fiabilité sur chaque coordonnée.
3.2.1. Les logiciels de compensation
Actuellement sur le marché Suisse deux grandes familles de logiciels sont

disponibles :
• Compensation 2D + 1 : traitement séparé de la planimétrie et de l’altimétrie.

Ce sont les plus courants, par exemple le logiciel LTOP de Swisstopo [2].
• Compensation 3D : traitement combiné de la planimétrie et de l’altimétrie
dans un modèle de compensation par moindres carrés tridimensionnels, par
exemple le logiciel Trinet+ du Dr. Sébastien Guillaume [3].
L’avantage principal d’un logiciel de compensation basé sur un modèle 2D+1

est la simplicité du traitement des réseaux de mesure. La perte de rigueur par la
dissociation de la planimétrie et de l’altimétrie va impacter la précision et la fiabilité
des résultats.
Une compensation tridimensionnelle (3D) va permettre d’exploiter et de com-

penser de manière rigoureuse la géométrie du réseau géodésique. Toutes les
mesures seront ajustées en un bloc, dans le même système de coordonnées
cartésien. Cependant la compensation 3D est plus complexe, elle demande des con-
naissances sur le champ gravitationnel, pris en compte la déviation de la verticale,
et impose la mise en place d’un système de coordonnées topocentrique.
3.2.2. Recommandations
Toutes les mesures doivent être compensées d’une manière ou d’une autre
afin que les résultats puissent être validés. Dans le cas de la surveillance des bar-
rages avec un fort développement altimétrique, il est recommandé de traiter les
538
Q.106 - R.34
mesures dans un logiciel de compensation tridimensionnel, cette méthode relie les

observations brutes aux inconnues de coordonnée de façon directe et rigoureuse.
3.3. LIMITES DES MESURES GÉODÉSIQUES
3.3.1. Le temps d’acquisition
Dans le cas d’un ouvrage de grande envergure, l’acquisition des mesures

d’un réseau de surveillance doit se dérouler sur plusieurs jours. Afin d’obtenir des
mesures précises, il est indispensable que l’ouvrage soit dans le même état statique
durant toute la campagne de mesure. L’exploitant doit s’efforcer d’atteindre la cote
du lac souhaitée et de la maintenir. Cette exigence aura un impact non négligeable
sur la production électrique. Pour l’exploitant, la réduction du temps d’acquisition
est donc un élément déterminant.
D’autre part, les conditions thermiques changeantes vont provoquer des défor-
mations de l’ouvrage durant la période de mesures. En fonction de la température du
site, de l’orientation de la vallée et de l’exposition de l’ouvrage au soleil, un réchauf-
fement ou refroidissement différencié de ses faces et de ses appuis va entrainer
des déformations de l’ouvrage. Le même point aura bougé entre des mesures réal-
isées depuis des stations différentes. Cela aura pour conséquence d’engendrer des
imprécisions dans le calcul final et donc une perte de qualité sur les coordonnées.
3.3.2. Faible résolution spatiale
Dans la pratique actuelle, on matérialise des points de surveillance sur

l’ouvrage. Le nombre de ces points est généralement faible, donnant donc une
image très ponctuelle et simplifiée des déformations de l’ouvrage. S’il est toujours
possible de rajouter des points selon les informations recherchées, en le faisant on
augmente la durée des mesures, ce qui nuit à la précision des résultats.
4. DE NOUVELLES TECHNOLOGIES AU SERVICE DES INGÉNEIURS ET

EXPERTS DES BARRAGES
Depuis plus de 25 ans deux technologies bien particulières sont présentes sur
le marché des capteurs et proposées pour le suivi des ouvrages d’accumulation : la
vidéo-tachéométrie et les scanners lasers 3D terrestres (TLS). Quelle est la valeur
ajoutée et quelles sont les limites de ces nouveaux outils?
Le tableau de comparaison ci-dessous résume les forces et faiblesse des

différentes technologies:
539
Q.106 - R.34
Tableau 1
Comparaison des technologies de mesure
GÉODÉSIE CLASSIQUE VIDÉO-TACHÉOMÉTRIE SCANNER LASER
Vitesse d’acquisition – + ++
Géoréférencement ++ ++ –
Précision finale ++ ++ –
Densité de points – – ++
4.1. LA VIDÉO-TACHÉOMÉTRIE
La vidéo-tachéométrie permet d’acquérir très rapidement les mesures angu-

laires sur le terrain. Cette technologie va permettre de figer plus rapidement
l’ouvrage dans une position donnée, limitant l’impact sur l’exploitation et réduisant
les effets thermiques. Le gain de temps par rapport à une mesure traditionnelle est
d’environ 50%.
La vidéo-tachéométrie permet de “ court-circuiter ”, sur le terrain, le fastidieux

travail du pointé fin sur les cibles. Après l’orientation de l’instrument, la caméra
acquiert une image détaillée de chaque cible. La détermination précise des mesures
angulaires est réalisée en post-traitement, au bureau, avec l’appui d’algorithmes de
reconnaissance d’images. La suite des ajustements géodésiques reste inchangée
et les outils de la mesure géodésique classique sont utilisés.
4.1.1. Description de la technologie
La vidéo-tachéométrie est une combinaison de deux systèmes : un

tachéomètre, respectivement une station totale, et un capteur vidéo. Aujourd’hui,
la plupart des stations totales sont équipées d’une caméra coaxiale avec axe de
visée. La caméra coaxiale est judicieusement installée dans la lunette de l’appareil,
elle profite ainsi du grossissement de l’optique qui peut atteindre un facteur 30.
4.1.2. Transformer son tachéomètre en vidéo-tachéomètre
Au sein de l’ETH, le Dr. Sébastien Guillaume a développé un vidéo-

tachéomètre avec une caméra qui vient se fixer en lieu et place de l’oculaire [4-5-6].
Fig. 1
De gauche à droite avec caméra CCD fixée sur : TDA 5005, TS15, MS50 [5], puis
images issues du système QDaedalus cibles, prismes,
mesure automatique d’une cible [6].
540
Q.106 - R.34
Cette caméra est couplée à un ordinateur et au programme “ QDaedalus ” qui permet

de piloter et d’automatiser les mesures. Dans ce cas, la mise en place au niveau
de l’oculaire présente l’avantage d’être universel, mais l’inconvénient de ne plus
permettre un pointé à l’œil, si ces n’est à travers le programme et les images issues
de la caméra.
Durant les mesures, seules les images sont acquises comme décrit ci-dessus.
Les contraintes sur l’exploitation sont par conséquent réduites au maximum. Le
pointé fin sur la cible ou le prisme est réalisé en post-traitement au bureau, soit
manuellement ou de façon semi-automatique.
La précision du système est au moins équivalente à un pointé manuel. Les

divers travaux et projets de recherche [5] et [6] ont permis de déterminer la précision
sur des cibles et prismes, comme représentés à la figure 1 ; elle est de 0.2 mgon,
ce qui représente un écart transversal de 0.3 mm à 100m.
4.1.3. Vidéo-tachéométrie : conclusion
Cette technologie permet de gagner du temps. Il est possible de post-traiter

les mesures comme des mesures traditionnelles. En revanche la densité de points
mesurés reste faible. Comme les mesures sont automatiques, il serait envisageable
d’augmenter la densité des cibles ; on pourrait ainsi obtenir un plus grand nombre
de vecteurs de déplacement entre deux états de mesures avec la précision et la
fiabilité d’une géodésie traditionnelle.
4.2. LA LASERGRAMMÉTRIE (OU SCANNERS LASERS 3D)
La lasergrammétrie s’est développée rapidement pour devenir aujourd’hui

quasiment incontournable. L’instrument motorisé tourne sur lui-même dans un
champ de vision paramétré. Il mesure un angle horizontal et vertical, de même
que la distance jusqu’à l’objet, à une vitesse de l’ordre de 1’000’000 de points par
seconde. Cet instrument permet d’acquérir rapidement un nuage de points 3D très
dense d’un ouvrage. La résolution d’acquisition (par exemple, 1 point par centimètre
à 100 m), est indéniablement la valeur ajoutée de cette technologie.
Aussi séduisante soit-elle, la lasergrammétrie possède ses limites. La plus

importante est le calage et le géoréférencement des nuages de points entre dif-
férents états de mesure. L’appareil mesure la position 3D de points quelconques
et différents à chaque numérisation. Il en résulte que les logiciels de compensation
traditionnelle ne peuvent pas être utilisés.
A l’heure actuelle, les différents projets et tests qui ont été menés ont démontré
qu’il est possible de réaliser l’auscultation d’ouvrages par lasergrammétrie, mais
avec une précision inférieure à celle de la géodésie classique [7-8]. En l’état actuel
541
Q.106 - R.34
de l’art concernant le calage et le géoréférencement, il est possible d’atteindre une

précision inférieure à 3 mm.
4.2.1. L’interprétation des déplacements
Comme mentionné précédemment, le scanner laser ne permet pas la mesure

de points prédéfinis sur l’ouvrage. Cela signifie que les mesures de déplacement ne
se font pas point à point, mais entre une facette déterminée par des points dans l’état
de référence et la facette perpendiculairement la plus proche dans l’état déformé.
Certains déplacements ne sont donc pas détectables, comme le montre la figure 2
ci-dessous. Les déplacements tangentiels à la direction de mesure sont difficilement
visibles. Seule la composante radiale d’un déplacement, dans la direction générale
de mesure, peut être facilement détectée.
Toutefois, avec la lasergrammétrie il est possible de mettre en valeur les

déplacements de tout l’ouvrage et de son environnement. Le gain d’information est
considérable. La déformée de l’ouvrage dans son ensemble est mise en évidence,
ce qui peut, le cas échéant, révéler des comportements atypiques.
Fig. 2
Détection de mouvements.
4.2.2. Lasergrammétrie : conclusion
La lasergrammétrie est parfaitement adaptée à la détection de mouvements

d’une amplitude supérieure à 4-5 mm, à condition qu’ils soient principalement ori-
entés selon une direction radiale à l’ouvrage. Elle permet de saisir l’ouvrage à un
temps "t", dans des états de charge différents, afin de valider, par exemple, leur
comportement structurel. La lasergrammétrie apporte un complément d’information,
mais elle ne peut pas encore, ou difficilement, remplacer la géodésie classique. Le
développement de stations totales avec fonction de numérisation pourra peut-être,
dans un avenir proche, combiner avantageusement les deux techniques.
542
Q.106 - R.34
5. EXEMPLES D’UTILISATION
5.1. LE BARRAGE DE TOURTEMAGNE [9]
5.1.1. Situation, géométrie et particularité
Le premier exemple d’application est un barrage voute situé dans les Alpes
suisses, d’une hauteur de 33 m, particulièrement mince et sensible à la température.
Avec une longueur au couronnement de 145 m et une vallée resserrée en pied,
l’ouvrage est considéré comme étant d’une envergure plutôt petite.
5.1.2. Le réseau géodésique
Deux mesures géodésiques sont effectuées tous les cinq ans, chacune
demandant 2 jours de travail à une équipe de 3 personnes. Les points de surveil-
lance sont placés au niveau du couronnement et deux points supplémentaires sur
le parement aval à la clé. Les différents points sont déterminés à l’aide d’un réseau
trigonométrique comportant des mesures d’angle et de distance, de même qu’un
nivellement géométrique.
Fig. 3
Réseau géodésique de l’ouvrage (©Geosat SA [9]).
L’assemblage des mesures et le calcul des coordonnées des points de surveil-

lance sont réalisés avec le logiciel de compensation par les moindres carrés LTOP.
La précision finale obtenue sur les points de surveillance est de 0.4 mm, tant en
planimétrie qu’en altimétrie, à 95% de fiabilité.
543
Q.106 - R.34
5.1.3. Validation du comportement chaud-froid
Les mesures de déformation mensuelles de l’ouvrage sont déterminées par un

pendule situé en clé, depuis le couronnement jusqu’au pied aval. A cet endroit, les
mouvements indiquent une légère déformation irréversible vers l’aval. En parallèle,
les mesures géodésiques montrent un déplacement plastique vers l’amont.
Pour mieux comprendre le comportement de l’ouvrage, deux campagnes de

mesures par lasergrammétrie 3D sont effectuées, la première à l’état chaud (15◦ C),
la seconde à froid (–10◦ C).
Durant l’état estival, les mesures scanner laser sont couplées à la géodésie.
En hiver, la présence de neige ne permet pas de réaliser une mesure géodésique
complète. Les mesures scanner laser et quelques mesures géodésiques princi-
pales sont effectuées afin que le calage entre les deux états soit précis et fiable.
Les mesures par scanner laser terrestre ont été réalisées depuis les deux rives,
deux stations étant nécessaires pour acquérir la totalité de l’ouvrage. Le temps
d’acquisition est d’une demi-journée.
5.1.4. Mouvements observés par scanner laser
La précision d’assemblage des nuages de points mesurés à partir des deux

stations est inférieure à 1.5 mm pour les deux états. La surface de l’ouvrage est de
bonne qualité (béton peu poreux et sans détérioration). Dans ces bonnes conditions,
il est possible de détecter des mouvements significatifs supérieurs à 2 mm.
En lasergrammétrie, l’une des étapes délicates est la mise en valeur des

résultats. Les valeurs obtenues sont projetées sur des plans ou coupes. La figure
Fig. 4
Déplacement entre l’état chaud et froid (©Geosat SA).
544
Q.106 - R.34
ci-dessous montre les déplacements entre les états chaud et froid. Afin de faciliter la
visualisation, la coloration continue 3D a été projetée sur une élévation développée
de l’ouvrage rendant les données facilement interprétables.
Un plan de déplacement au niveau du couronnement met en relation les

déplacements observés par scanner laser et ceux observés par méthode tradi-
tionnelle. La courbe noire passe, pour la quasi-totalité des points, dans l’ellipse
de confiance des points de surveillance déterminés par la géodésie classique,
démontrant la bonne qualité de la lasergrammétrie. L’intervalle de confiance sur
les déplacements observés par scanner laser est de ± 2 mm.
Les mouvements surfaciques mesurés en tout point peuvent alors être utilisés
comme calage d’un modèle par éléments finis conçu et utilisé pour comprendre le
comportement particulier de l’ouvrage.
Des campagnes ultérieures de mesure avec cette technologie permettront de

confirmer et de suivre le comportement irréversible de l’ouvrage.
Fig. 5
Profils au niveau du couronnement (©Geosat SA).
5.2. LE BARRAGE DE SCHIFFENEN
5.2.1. Situation et géométrie
Le barrage de Schiffenen est implanté dans une large vallée, il obstrue le

passage à la Sarine. Son couronnement est long de 417 m pour une hauteur de
47 m et 7 m d’épaisseur au couronnement. Cet ouvrage est le plus grand et le plus
large présenté dans cet article.
545
Q.106 - R.34
Fig. 6
Plan du réseau géodésique (issu de l’article ref [8] ©Swisstopo).
5.2.2. Comparaison de trois méthodes de mesure de déformation sur les

barrages
Le site de Schiffenen a servi de laboratoire pour le projet de test de dif-

férentes méthodes d’acquisition pour la mesure de déformation des barrages, initié
par l’Office Fédéral de Topographie Swisstopo (Suisse). Ce projet de recherche
et développement [8] a pour but de comparer différentes techniques d’acquisition
et d’analyse des déplacements sans contact et de les confronter à la géodésie
traditionnelle afin de qualifier leur précision.
Les méthodes d’acquisition et d’analyse des déplacements testées dans ce

projet sont :
• Géodésie traditionnelle : Swisstopo (S. Condamin). Ces mesures font état

de valeur de référence.
• Lasergrammétrie : Haute Ecole d’Ingénieurieet de Gestion du Canton de
Vaud (CH) (V. Barras).
• Interférométrie Radar : Ecole polytechnique de Zürich (L. Schmid, A.
Wieser).
Ici nous parlerons uniquement de la partie lasergrammétrique acquise avec

le Riegl VZ2000i. Un article sur ce projet a été publié en 2020 [7].
546
Q.106 - R.34
5.2.3. États de mesure et déplacements attendus
Là également, la comparaison est faite entre un état froid et un état chaud.

La mesure hivernale est réalisée en février 2019 et l’estivale à la fin août 2019.
Le gradient de température de l’air entre les deux états est d’environ 10◦ C. Les
déplacements attendus au sommet de la voûte sont de l’ordre de 3 cm. Les résultats
de la lasergrammétrie sont confrontés à la géodésie afin de valider les déplacements
et la précision.
5.2.4. Mesure par scanner laser et calages des nuages de points
Dans le cadre de ce projet, la stratégie consiste à fixer les coordonnées plan-

imétriques et altimétriques du centre des scanners, puis d’orienter les nuages l’un
par rapport aux autres à l’aide des plans communs. La configuration du site per-
met un très fort recouvrement inter scan sur la voute de l’ouvrage. En vue de cette
stratégie la densité de points est définie à un point tous les 1.6cm à 100m, de
même que le scanner laser a été placé dans les embases déterminées par le réseau
géodésique. Cette technique de calage est possible uniquement à l’aide de logiciels
professionnels qui permettent de fixer des degrés de liberté pour chaque nuage de
points. Comme la station inertielle du VZ2000i n’est pas d’une finesse suffisante,
les 3 rotations (lacet, tangage et roulis) sont à déterminer.
Avec cette méthode il a été possible d’atteindre une précision de calage des
angles de ± 0.006 gon, pour chaque état. Cette méthode sans point de calage a
permis d’accélérer le travail sur le terrain, mais donne des résultats mitigés pour les
orientations des scans. Pour la suite du projet, des ajustements sur des données
tachéométriques complémentaires ont été effectués. Ce recalage pose un problème
notoire lorsque l’on ne dispose pas de ces compléments. Pour un objet de cette taille,
l’utilisation de quelques cibles communes entre stations de scan est actuellement
encore vivement conseillée.
5.2.5. Comparaison entre les états de mesure
Grâce aux modélisations 3D il est possible de limiter l’impact du bruit du

nuage de points (principe proche de la moyenne par moindres carrés appliquée à
la géodésie). La comparaison entre l’ouvrage à l’état hivernal et estival est reportée
sur une carte en coloration continue reproduite ci-dessous. Les déplacements, au
centre de l’ouvrage, atteignent 3 cm. Dans cet exemple une coupe verticale a été
générée au niveau d’un pendule. Il est ainsi possible de valider les observations
scanner laser avec celles faites sur le pendule. Les valeurs extraites de la laser-
grammétrie diffèrent au maximum de 2 mm par rapport aux déplacements mesurés
sur 3 niveaux.
547
Q.106 - R.34
Fig. 7
A gauche déplacements scanner laser, à droite coupe vertical au niveau d’un
pendule.
Pour réaliser une comparaison rigoureuse avec la géodésie, des cibles déter-
minées par intersections de directions aux 2 états sont projetées sur les surfaces
modélisées en hiver et en été. En comparant la variation de ces projections avec
les résultats traditionnels, il a été possible d’évaluer l’écart des déplacements 3D
sur 18 points entre les 2 techniques. En moyenne, suite aux traitements effectués
avec le Riegl, le projet a mis en évidence la statistique suivante :
Tableau 2
Écarts entre 18 cibles géodésiques et scanner laser
E [mm] N [mm] Z [mm] 3D [mm] rad [mm] tang [mm]
Ecart-type 3.8 3.7 1.3 5.2 5.2 0.6
Moyenne 1.3 –2.4 0.4 –3.3 –3.3 0.2
Mediane 1.0 –1.8 0.2 –2.7 –2.6 0.0
Minimum –5.6 –8.7 –1.1 –9.9 –9.3 –0.3
Maximum 9.1 0.9 3.4 5.5 5.6 1.9
5.3. LA FALAISE DE HAUTERIVE
5.3.1. Situation
Comme son nom l’indique, il ne sera pas question ici d’un ouvrage d’art mais
d’une falaise de molasse. La surveillance d’éléments naturels est également impor-
tante, car la chute de blocs ou des glissements de terrain peuvent endommager les
infrastructures.
548
Q.106 - R.34
La falaise de Hauterive est située à l’amont de la centrale hydroélectrique du

même nom, proche de Fribourg en Suisse. Haute d’une quarantaine de mètres elle
a présenté un danger pour la production hydroélectrique après un décollement de
près de 250 m3 de molasse en 2012. Actuellement, des décollements de petits
blocs sont régulièrement détectés.
5.3.2. Apport de la lasergrammétrie
La détection de mouvements par lasergrammétrie a montré tout son sens

dans ce cas. La surveillance surfacique a été couplée au suivi de cibles mesurées
par méthode tachéométrique. Le but, en 2014, était de comparer les 2 tech-
niques et de valider ce procédé exhaustif sur une surface naturelle avec de
la végétation variant suivant les saisons. Dans ce cas, les cibles placées
sur des blocs rocheux sains et nettoyés, n’ont pas montré de mouvements
significatifs.
5.3.3. Les mesures par scanner laser
Les mesures par scanner laser sont réalisées depuis le pied de la falaise.
Le secteur étudié s’étend sur environ 200 m de longueur. En tenant compte de
la végétation et des infrastructures en place, 4 stations de mesure sont néces-
saires afin d’acquérir tout le site. En parallèle aux mesures par scanner laser, des
images haute résolution sont acquises pour valider ou invalider les secteurs en
mouvement mis en évidence par la lasergrammétrie (la référence [10] détaille le
travail)
Fig. 8
Acquisition scanner laser du site.
549
Q.106 - R.34
5.3.4. Mise en évidence des blocs tombés
Après assemblage des nuages et un filtrage rigoureux des données, la détec-

tion des chutes de blocs d’une surface supérieure à 0.25 m2 (50cm x 50cm) est faite
en comparant la modélisation 3D de deux états de mesures. Au vu de la configuration
du site et de la forme de la falaise, le seuil de détection est fixé à 1 cm.
Comme on peut le voir ci-dessous, la zone claire représente la partie sta-

ble de la falaise. Les différentes taches sombres sont les déplacements observés
entre deux états de mesure. La technique a permis de détecter le décollement des
“ feuilles ” de molasse avant leur chute.
Fig. 9
Mise en évidence des chutes de blocs et découlements.
La figure ci-dessous montre la comparaison entre l’image de gauche avec les

zones en mouvement identifiées à partir des données laser et celle de droite qui
montre la chute effective des blocs. Ici l’apport de l’image permet de valider les
différences observées par scanner laser en montrant qu’elles sont bien réelles et
non pas liées à une évolution de la végétation.
Fig. 10
Relation entre image et détection de chute de bloc dans les données scanner laser.
550
Q.106 - R.34
Après une surveillance de 3 ans (2014 à 2017), la méthode a montré sa fiabilité

et son efficacité. Les craintes d’une poursuite de glissement de grands secteurs se
sont avérées infondées et la surveillance a été suspendue.
6. LE FUTUR
Les scanners lasers 3D et, dans une moindre mesure la vidéo-tachéométrie,

permettent d’envisager des améliorations de la surveillance des barrages dans cer-
tains cas, que nous présentons ci-dessous. Nous n’avons pas la prétention de
présenter l’ensemble des apports de ces deux technologies, car les ouvrages de
retenues sont très différents les uns des autres, par leur situation, leurs dimensions,
leur construction, leur surveillance.
6.1. SAISIR L’INFLUENCE DE PARTIES D’OUVRAGES
Les résultats obtenus par scanner laser sur le barrage de Schiffenen ont été
examinés avec attention par l’exploitant et ses ingénieurs chargés de la surveillance,
en particulier pour vérifier si des particularités constructives ou des fissures de
l’ouvrage influençaient ses déformations harmonieuses issues des calculs statiques.
Ces particularités constructives sont les suivantes:
• L’appui de la centrale sur le parement du barrage sur les blocs 7, 8 et 9

• La jonction de la cage d’ascenseur qui s’appuie contre le couronnement du
bloc 7
• Les deux tours de prises d’eau qui s’appuient contre le couronnement des
blocs 8 et 9 côté amont
• Les surépaisseurs de béton armé construites autour des trois évacuateurs de
crues
Fig. 11
Jonctions des tours de prise d’eau et de l’ascenseur avec le couronnement du
barrage.
551
Q.106 - R.34
Fig. 12
Coupe d’un évacuateur de crues avec la galerie.
• Les fissures horizontales de la galerie située sous l’évacuateur de surface côté

aval
Sur la base des résultats obtenus, il semble que les surépaisseurs des évac-
uateurs de crues agissent comme des raidisseurs et que l’appui de la centrale
limite les déplacements à la base des blocs 7, 8 et 9. Ces effets pourraient perme-
ttre d’améliorer la connaissance des déformations dans le modèle déterministe et
d’améliorer son calage.
6.2. SUIVRE LES DÉGÂTS SUR DES PAREMENTS OU DES APPUIS
Certains barrages ont été recouverts de béton projeté. C’est le cas du barrage
de la Maigrauge en Suisse, construit en 1870, qui a été recouvert à différentes épo-
ques. Cette gunite vieillit et se desquame. Les épaisseurs de gunite qui chutent sont
épaisses de quelques cm. Des mesures par scanner laser permettraient, comme
pour la falaise de Hauterive présentée au chap. 5.3.3, de suivre les dégâts et de
prévoir une réhabilitation.
Les appuis rocheux victimes de chutes de pierres ou de blocs pourraient

également être suivis de la même façon.
6.3. SUIVRE LES DÉPLACEMENTS DE PARTIES D’OUVRAGES PEU INSTRUMENTÉES
Il n’est pas toujours possible d’installer des pendules dans certains barrages, à
cause de leur épaisseur trop restreinte ou de leur courbure prononcée. Un scannage
régulier permettrait de saisir les mouvements de ces ouvrages particuliers.
552
Q.106 - R.34
6.4. SUIVRE LES DÉPLACEMENTS D’UN OUVRAGE EN CAS D’ÉVÈNEMENT

EXTRAORDINAIRE
À la suite d’un séisme ou de déformations extraordinaires, les exploitants et

les ingénieurs chargés de la surveillance d’ouvrages doivent pouvoir disposer de
résultats de mesures de déformations le plus rapidement possible. Que ce soit
par scanner laser ou par mesures vidéo tachéométriques tout gain de temps est
bienvenu.
Si l’on veut utiliser le laser scannage, il faut disposer d’un état de référence
lors de chaque saison au moins pour pouvoir comparer les résultats des deux
époques.
6.5. LES PROJETS DE RECHERCHES
Comme évoqué précédemment, les mesures par scanner laser ne permettent

pas d’identifier de manière claire et précise les déplacements tangentiels. Pour
combler ce manque, l’une des idées est de suivre en 3D le centre d’une empreinte-
relief d’un secteur. (ou si l’ouvrage à une structure trop lisse, de de matérialiser
des sphères). Ces secteurs peuvent être définis à postériori et leur nombre peut
évoluer en fonction des mouvements suspectés, car les traitements pour le suivi
se font sur le nuage 3D. Quelques tests succincts ont été réalisés sur les données
de Schiffenen et sur d’autres ouvrages avec un relief plus prononcé. Les résultats
sont encourageants, mais doivent être affinés sur d’autres barrages. La précision
de ces déterminations est moindre que celle d’une cible observée par géodésie.
Cependant, si l’ouvrage est sensible aux effets thermiques et/ou hydrostatiques,
les mesures par scanner laser permettraient de déterminer très rapidement tous
ces points. Si cela fonctionne de manière fiable, cela permettrait de combiner les
avantages de la mesure surfacique avec une analyse ponctuelle [11].
En parallèle, la vidéo-tachéométrie va être déployée sur des ouvrages tests

d’envergure. Cette technologie est très prometteuse car elle va permettre de
mesurer plus rapidement les réseaux géodésiques tout en assurant la même préci-
sion que des mesures traditionnelles. De plus, les pointés pouvant être automatisés,
l’impact humain du pointé manuel des cibles est minimisé, tout en assurant une
meilleure homogénéité entre les états de mesure [11].
7. CONCLUSION
Les deux technologies présentées ici apportent des réponses à la probléma-

tique actuelle des exploitants et spécialistes du génie civil : limiter les contraintes
d’exploitations et figer rapidement un ouvrage à un instant donné.
553
Q.106 - R.34
À l’aide de la lasergrammétrie, il est possible d’établir des cartes de colori-

sation continue des déplacements, permettant aux ingénieurs civils de mettre à
jour et valider les modèles par éléments finis. D’autre part, comme il ne faut que
très peu de temps pour acquérir un ouvrage, les contraintes d’exploitation sont
réduites au maximum. Pour l’instant, cette technologie est bien adaptée pour la
détection de mouvements saisonniers ou à des états de charge différents. Vu
qu’il est possible d’ausculter tout type d’objet sans matérialisation, elle est tout
particulièrement adaptée à la surveillance des falaises ou des massifs d’appui, afin
de valider ces divers secteurs rarement instrumentés, Cette technologie permet
d’assurer la protection des installations et personnes.
La vidéo-tachéométrie est une technologie très prometteuse, actuellement

opérationnelle sur de petits ouvrages ou dans le cadre de mesures industrielles.
Plusieurs projets sont en cours afin de la mettre en production sur des grands
ouvrages. Dans l’avenir il sera possible de mesurer un réseau géodésique avec
un gain de temps de 50%, tout en garantissant les mêmes précisions.
REFERENCES
[1] La géodésie pour la surveillance d’ouvrage des ouvrages d’accumulation

(2013), Comité Suisse des Barrages, sous-groupe géodésie, P-H Cattin, G.
Darbre, C-A Eder, J-C Kolly, M. Lutz, F. Walser, A. Wiget.
[2] Manuel d’utilisation et Description du programme LTOP, Mai 2003 Erich Gubler,
Swisstopo.
[3] Trinet+ Logiciel de Compensation de réseaux 3D Mode d’emploi, V6.1, Dr.

Sébastien Guillaume, Cyril Muller, Paul-Henri Cattin.
[4] QDaedalus: Augmentation of Total Stations by CCD Sensor for Automated

Contactless High-Precision Metrology, (2012) Sébastien GUILLAUME, Beat
BUERKI, Sylvain GRIFFET and Hélène MAINAUD DURAND, Switzerland.
[5] Contribution of the Image-Assisted Theodolite System QDaedalus to Geodetic

Static and Dynamic Deformation Monitoring (2016), S. Guillaume, J. Clerc, C.
Leyder, J. Ray, M. Kistler.
[6] Master Thesis: Automated Measurement of Passive Targets for Dam Monitor-
ing using an Image-Assisted Theodolite (2015), Jonas Clerc.
[7] Staumauerüberwachung – Vergleich dreier Technologien für epochenweise

Deformationsmessungen (2020) Andreas WIESER, Sebastian CONDAMIN,
Vincent BARRAS, Lorenz SCHMID und Jemil BUTT.
[8] Auscultation sans réflecteur (2017) Vincent Barras – Bertrand Cannelle –

Guillaume Chapotte – Revue XYZ 151, pages 43 à 50.
554
Q.106 - R.34
[9] Selon communication : Geosat SA, Christian Hagin, Route du manège 59b
CH-1950 Sion.
[10] Auscultation : complémentarité du laser et de l’image Revue Géomatique

Suisse 9/2018 Pages 264–267.
[11] Selon communication: Dr. Sébastien Guillaume and M.E. Franck Schmidt,
Institute of Territorial Engineering (INSIT), School of Management and Engi-
neering Vaud (HEIG-VD), University of Applied Sciences of Western Switzer-
land (HES-SO), 1 Route de Cheseaux, CH-1401 Yverdon-les-bains.
555
Q.106 - R.35
-------
-------
FIBER OPTIC DISTRIBUTED SENSING SYSTEM FOR MONITORING OF

TAILINGS STORAGE FACILITIES (∗)
Daniele INAUDI
Chief Technology Officer, SMARTEC SA
Riccardo BELLI
S&S Coordinator, Project Engineer, SMARTEC SA
Régis BLIN
Sales Engineer, SMARTEC SA
SWITZERLAND
SUMMARY
The types of measurement to be carried out as well as instrument location

must be selected according to the type and conditions of the tailings dam itself.
Tailing dams deformation and internal erosion are detected at an early stage with
distributed fiber optic sensing system. Fiber optic distributed sensing is increas-
ingly recommended by designers as a key monitoring solution for dam safety
management.
RÉSUMÉ
Les types de mesures à effectuer ainsi que l’emplacement des instruments

doivent être choisis en fonction du type et des conditions de la digue de retenue des
résidus elle-même. La déformation et l’érosion interne de ce type de barrage sont
détectées à un stade précoce grâce à un système de détection à fibres optiques
réparties. Cette technologie de surveillance est de plus en plus recommandée par
∗ Instrumentation d’un barrage de résidus à l’aide d’un système de mesures distribuées par
fibre optique
556 DOI: 10.1201/9781003211808-130

Q.106 - R.35
les concepteurs comme une solution de surveillance clé pour la gestion de la sécurité
des digues de retenue des résidus.
1. INTRODUCTION
A fiber optic distributed sensing system concentrates on the monitoring of

parameters useful to determine the short- and long-term behavior of the structure
in response to the external actions, including, e.g. seismic activities as well as long-
term phenomena such as seepage, tunneling, sinkholes, and other geo-structural
movements.
Monitoring of large retaining structures with such a system provides direct

strain/deformation and temperature measurements on the broad interior area,
enabling early detection, localization and sizing of subsurface damages. Measure-
ments are provided every meter along the entire sensing cable, this is particularly
important in such cases where the location of defects cannot be predicted, and
localized geotechnical instrumentation is therefore useless. Fiber optic distributed
sensing cables are relatively easy to deploy, are generally available for retrofit instal-
lation inside the dam structure using trenches on the downstream face or laid during
construction below the geomembrane liner within the bedding sand layer. Mapping
the cable route enables exact correlation between a linear coordinate on the cable
length and a position in the dam. Cable installation is envisaged to be carried out
during further raising of the embankment; additional cable sections are plugged to
the previously installed cables and measured with the existing measurement sys-
tem, requiring only an update of the mapping. The measurement system is placed
remotely from the sensing area since a section of sensing cable can be used to
link to the sensing area at the storage facility several kilometers away without any
performance degradation. The system is available for on-line, 24/7 real-time oper-
ational data management. Such fiber optic distributed sensing systems allow the
storage facilities to be monitored with long, accurate, continuous and precise data
for geotechnical diagnostic purposes.
This contribution will introduce the principles, benefits, and limitations of this
technology, presenting a relevant application example of use in tailings dams.
2. DETECTION WITH FIBER OPTIC SENSING
2.1. BENEFITS
A distributed sensing system installed in a dam or levee is able to detect the

following failure: structural movement, overtopping, under levee seepage, through
levee seepage, piping (internal erosion), external erosion, differential settlement and
landslides. The list of direct and indirect monitored parameters is presented Table 1.
557
Q.106 - R.35
Table 1
Monitored parameters
PRIMARY MONITORED ADDITIONAL MONITORED INDIRECTLY DETERMINED
PARAMETERS PARAMATERS PARAMETERS
Distributed average strain Seepage/leakage Dam rupture monitoring
Distributed temperature detection Dam settlement
Dam load re-distribution
Dam vertical displacement
Dam horizontal displacement
Sinkholes detection
Unlike “discrete monitoring systems” there are no coverage “gaps” between

sensors, providing seamless coverage. The system collects data day and night, 24
hours a day. Using current internet technologies, effective warning systems can be
integrated into current management methods. Sensor cables can be retrofitted on
existing levees and dams by means of trenching the cable into the face of the struc-
ture, on either the upstream or downstream side. Often trenching is only necessary
to a depth of protection for the cable itself.
A levee or dam is in many ways like a chain, only as strong as its weakest link (or
lowest point). An individual levee can be many miles long and can meander through
urban, rural, remote, inaccessible and sometimes dangerous areas. Vegetation,
sheer size and scale, can limit levee access and mask indicators from even the
most well trained inspectors. Areas of structural weakness, even when relatively
obvious, are difficult to identify. The DiTeSt system can not only detect structural
failures, but also detect very slight differential settlements that can be considered
as early warning of future issues as illustrated on sketch Fig. 1.
Fig. 1
Detection principle
2.2. WORKING PRINCIPLE
The most developed technologies of distributed fiber optic sensors are based
on Raman and Brillouin scattering. Both systems make use of a non-linear
558
Q.106 - R.35
interaction between the light and the silica material of which a standard optical fiber
is made. If light at a known wavelength is launched into a fiber, a very small amount
of it is scattered back at every point along the fiber. The scattered light contains
components at wavelengths that are different from the original signal. These shifted
components contain information on the local properties of the fiber, in particular its
strain and temperature. After an appropriate analysis of those signals, the system
can therefore provide a calibrated strain or temperature reading at every meter along
the sensing cable, as depicted in Fig. 2.
Typically a temperature resolution in the order of 0.1˚ C, a strain resolution of

20 µm/m and a spatial resolution of 1 m over a measurement range up to 50 km are
obtained with commercially available interrogators.
Fig. 2
Fiber Optic Distributed Sensing System.
2.3. TYPICAL LAYOUT
The location of the cable will depend on many different factors that must be
carefully considered by the engineer. Location of the sensor cable is critical for
detecting both strain and seepage temperature change, see example Fig. 3.
With regard to strain, the sensing cable must be placed where changes in
the shape of the structure will physically occur, directly transferring strain to the
cable. The same is true for seepage. The sensing cable must be placed in a location
where water is moving close to the cable, creating trends in temperature directly to
the cable.
559
Q.106 - R.35
Fig. 3
Layout of sensing cable for temperature monitoring.
3. FIBER OPTIC DISTRIBUTED SENSING SYSTEM
3.1. SENSING CABLE
Since the Brillouin frequency shift depends on both the local strain and tem-
perature of the fiber, the sensor setup will determine the actual sensitivity of the
system.
For measuring temperatures, it is sufficient to use a cable designed to shield

the optical fibers from an elongation of the cable. The fiber will therefore remain
in its unstrained state and the frequency shifts can be unambiguously assigned to
temperature variations. Measuring distributed strains requires a specially designed
sensor; a mechanical coupling between the sensor and the host structure along the
whole length of the fiber has to be guaranteed. Table 2 below introduces different
cable designs to measure strain and temperature in different applications.
Table 2
Examples of Distributed Sensing Cables
560
Q.106 - R.35
Traditional fiber optic cable design aims to the best possible protection of the
fiber itself from any external influence. In particular, it is necessary to shield the
optical fiber from external humidity, side pressures, crushing and longitudinal strain
applied to the cable. These designs have proven very effective in guaranteeing
the longevity of optical fibers used for communication and can be used as sensing
elements for monitoring temperatures in the -40˚ C to +80˚ C range, in conjunction
with Brillouin or Raman monitoring systems.
For Brillouin scattering systems, it is important to guarantee that the optical

fiber does not experience any strain that could be misinterpreted as a temperature
change due to the cross-sensitivity between strain and temperature. On the other
hand, the strain sensitivity of Brillouin scattering prompts to the use of such systems
for distributed strain sensing, a goal contradicting all experience from telecommuni-
cation cable design where the exact opposite is required. When sensing distributed
strain it is necessary to simultaneously measure temperature to separate the two
components. This is usually obtained by installing a strain and a temperature sens-
ing cables in parallel. Some cable design combines the two functions into a single
packaging as the cable installation depicted Fig. 4, and make use of distributed
Raman sensing for temperature monitoring.
Fig. 4
Distributed strain sensing cable for movements and landslides detection in Perez
Caldera II Project, Chile.
3.2. MONITORING CABINET
The measurement cabinet includes DiTeSt BOTDA distributed strain sens-

ing systems combined with DiTemp DTS temperature sensing interrogator as the
cabinet Fig. 5. The DiView data management system combines information from
561
Q.106 - R.35
both systems to analyze strain and temperature variations and displays the results
in a single user interface. Alerts can be set based on different conditions for
strain and temperature variations and enable easy localization of the triggering
events.
Fig. 5
Monitoring cabinet with DiTeSt BOTDA (bottom) and DiTemp DTS (top).
3.3. GRAPHICAL USER INTERFACE
The main functions of data management are aimed to measure distributed

sensors automatically and process the large amount of data automatically to detect
and localize the undesired events. The operator can view in real time the sensor’s
measurement history in graphical form. The system is also able to trigger alerts
(message, mail and phone call) and show warnings on the display. Warnings are
generated for different types of events, including: strain, temperature, leakages
and cracks. The system is able to combine measurements from different sens-
ing cables, to obtain complex results, such as temperature compensated strains
displayed on a user-friendly view as on Fig. 6. Another data analysis module is
dedicated to the detection of cracks from distributed strain data. All information
related to a sensor are stored in a single database structure. All data can be eas-
ily accessed from third-party software for further analysis or integration in SCADA
systems.
562
Q.106 - R.35
Fig. 6
Graphical User Interface (GUI) for distributed monitoring.
4. CASE STUDY – MAREESBURG TAILINGS DAM
4.1. CHALLENGES
The Mareesburg tailings dam is part of Der Brochen Project Platinum Mine
located in South Africa in the Limpopo province. Mareesburg tailings storage facility
is an expansion to the current mine infrastructure to allow storage of tailings material
pumped via penstock pipelines from the Mototolo Operation Concentrator where the
ore is processed. Mareesburg is a hillside tailings storage facility of 115 meters height
and a maximum footprint area of 133 ha capable to store 65 million m3 of tailings
material.
4.2. IMPLEMENTATION OF DISTRIBUTED FIBER OPTIC MONITORING SYSTEM
In order to safeguard the tailings dam and maintain integrity and safety at the
mining site, 8 km of Hydro & Geo sensing cable for strain and temperature monitoring
were laid during dam construction below the geo membrane liner within the bedding
563
Q.106 - R.35
layer. The Hydro & Geo cable does simultaneously both deformation and seepage
detection allowing strains monitoring at the liner/foundation interface below the dam
and seepage monitoring through any liner defects. Measurements are carried out
automatically and fed into the data acquisition software able to show the results
remotely and to trigger alerts (relays, SCADA, SMS, mail or phone call) in case of
critical events.
REFERENCES
[1] INAUDI, D., Distributed Fiber Optic Sensors for Dams and Levee Deformation
Monitoring, GEG New Orleans, 2015.
[2] INAUDI D. Permanent Safety Assessment of Dams, Levees, Reservoirs,

Waterways with Fiber Optic Distributed Sensing, ICOLD Vienna, 2018.
564
Q.106 - R.36
-------
-------
MULTI-OBJECTIVE CALIBRATION OF DAM BEHAVIOUR ANALYSIS MODEL

FOR GRAVITY DAMS: APPLICATION TO ROBIEI DAM (∗)
Marius BÜHLMANN
Axpo Power AG, Dr. sc. ETH ZURICH (Formerly VAW, ETH ZURICH)
David F. VETSCH
Senior Researcher, Dr. Sc. Eth, Laboratory of Hydraulics, Hydrology and
Glaciology (VAW), ETH ZURICH
Georges R. DARBRE
Dam Safety Consultant and former Swiss Commissioner for Dam Safety, Ph.D.
Riccardo RADOGNA
Officine Idroelettriche della Maggia SA (Ofima), M. Sc.
Robert M. BOES
Director, Prof. Dr. sc. techn. ETH, Laboratory of Hydraulics, Hydrology and
Glaciology (VAW), ETH ZURICH
SWITZERLAND
SUMMARY
Dams can be affected by long-term processes such as alkali aggregate reac-

tion, concrete ageing and irreversible rock mass deformation, potentially resulting
in serious damage. Therefore, an early detection of deficiencies by use of a proper
monitoring system in combination with a dam behaviour analysis model is essen-
tial. A dam behaviour analysis model links these measured displacements with the
environmental conditions, i.e. above all water level and temperature distribution
in the dam body. Usually, displacements and environmental conditions are linked
∗ Calibration multi-objectif du modèle d’analyse du comportement des barrages-poids
DOI: 10.1201/9781003211808-131 565

Q.106 - R.36
empirically by statistical methods. Dam behaviour models are commonly set up and
calibrated to fit the displacement at one measurement level only.
The goal of the present work is to use a new approach that allows for multi-
objective calibration at several levels. The basis of this approach is a deterministic
beam model that can be used to calculate the displacement at each level. The
calibration is performed by using a Bayesian likelihood function, with which prior
information about the material parameters can be considered. The Markov Chain
Monte Carlo Method (MCMC) is used as optimization algorithm. This does not only
provide the most probable parameter set but also their distribution and correlations.
The newly developed approach has been applied to the Robiei hollow gravity
dam in Switzerland. The results are shown for the calibration period 1999 - 2006.
Except for the bottom level, a shift in the displacements in late 2008 can clearly be
identified. The size of this shift was found to be of the same order of magnitude at
all levels above. This leads to the conclusion that a damage on the measurement
equipment occurred at the bottom level. Furthermore, the displacements at crest
level show a slight trend that is not present at all levels. The example of this case
study shows the potential of using a deterministic beam model in combination with
multi-objective calibration. Only one set of physically meaningful parameters is cal-
ibrated for one block of a gravity dam. This leads to simple models that allow for
comprehensible inference by engineering judgement.
RÉSUMÉ
Les barrages peuvent être affectés par des processus à long terme tels que la
réaction alcali-granulat, le vieillissement du béton et la déformation irréversible des
masses rocheuses, qui peuvent conduire à de graves dommages. Il est donc essen-
tiel de détecter rapidement des défaillances en utilisant un système de surveillance
approprié en combinaison avec un modèle d’analyse du comportement des bar-
rages. Un modèle d’analyse du comportement des barrages relie les déplacements
mesurés aux conditions environnementales, avant tout au niveau d’eau et à la répar-
tition de la température dans le corps du barrage. Habituellement, les déplacements
et les conditions environnementales sont liés empiriquement par des méthodes
statistiques. Les modèles de comportement des barrages sont généralement établis
et calibrés pour s’adapter au déplacement à un seul niveau donné.
L’objectif du présent travail est d’utiliser une nouvelle approche qui permet
un calibrage multi-objectifs à plusieurs niveaux. La base de cette approche est un
modèle de poutre déterministe qui peut être utilisé pour calculer le déplacement
à chaque niveau. La calibration est effectuée en utilisant une fonction de vraisem-
blance bayésienne avec laquelle les informations préalables sur les paramètres
des matériaux peuvent être prises en compte. La méthode de Monte Carlo par
566
Q.106 - R.36
Chaîne de Markov (MCMC) est utilisée comme algorithme d’optimisation. Elle four-
nit non seulement l’ensemble des paramètres les plus probables, mais aussi leur
distribution et leurs corrélations.
Cette nouvelle approche a été appliquée au barrage-poids évidé de Robiei en

Suisse. Les résultats sont présentés pour la période d’étalonnage 1999 - 2006.
À l’exception du niveau inférieur, on peut clairement identifier un décalage des
déplacements à la fin de 2008. L’ampleur de ce décalage s’est avérée être du
même ordre de grandeur à tous les niveaux supérieurs. Cela permet de conclure
qu’un dommage sur l’équipement de mesures s’est produit au niveau inférieur. En
outre, les déplacements au niveau de la crête montrent une légère tendance qui
n’est pas présente aux autres niveaux. L’exemple de cette étude de cas montre le
potentiel de l’utilisation d’un modèle de poutre déterministe en combinaison avec un
calibrage multi-objectifs. Un seul ensemble de paramètres physiquement significat-
ifs est calibré pour un bloc d’un barrage-poids. Cela conduit à des modèles simples
qui permettent une inférence compréhensible par jugement d’ingénieur.
1. INTRODUCTION
Dams can be affected by long-term processes such as alkali aggregate reac-

tion, concrete ageing and irreversible rock mass deformation, from which they can
suffer serious damage. Therefore, an early detection of deficiencies by use of a
proper monitoring system in combination with a dam behaviour analysis model is
essential. In concrete dams, the displacements are commonly measured by pen-
dulums. A dam behaviour analysis model links measured displacements with the
environmental conditions, i.e. above all the water level and the temperature distribu-
tion in the dam structure. A common way to relate displacements and environmental
conditions are statistical methods [1]. The shape functions of the relations are
described by a model equation that has to be selected in advance. For the latter,
statistical model equations in combination with multiple linear regression (MLR) anal-
ysis are used. In recent past, there is also a trend to hybrid models. Hybrid models
are statistically optimised deterministic models (also called adjusted deterministic
models), where shape functions are evaluated based on the result of the deter-
ministic model [1]. Common deterministic models are finite element (FE) models or
simplified beam models of the dam structure.
Dam behaviour models are commonly set up and calibrated to fit the displace-
ment at one level li . If several levels are analysed, different sets of parameters are
estimated. However, deterministic models can be used to calculate the displace-
ment at several points of the dam, e.g. at different measurement levels of a block.
This allows for a multi-objective calibration of the material parameters, where the
measured displacement is simultaneously matched with the model output at several
levels. Unfortunately, common MLR analysis is not suitable for such a calibration.
567
Q.106 - R.36
There are several nonlinear calibration methods to perform multi-objective calibra-

tion. One method is the Markov Chain Monte Carlo method (MCMC) that is based on
Bayesian inference. The advantage of a Bayesian approach is that prior knowledge
about the parameters can be brought into the model. Since the model parameters
are physically-based, engineering knowledge or results from material tests can be
used. However, for MCMC a sampling framework is used that is computationally
intensive. Thus, efficient models with short computation time (e.g. 2-5 s for one
simulation) are preferred.
For this study, a simple structural beam model with elastic foundation was
set up to estimate the displacements of gravity dams. It was successfully cali-
brated with the MCMC method. The model and the MCMC method are described
in Chapter 2. In Chapter 3, the model is applied to the Robiei dam in the scope of
a case study.
2. METHODS
2.1. BEAM MODEL FOR GRAVITY DAMS
2.1.1. Model description
A typical vertical section of a gravity dam and acting loads that are considered
in the model are shown in Fig. 1a). The loads are the water pressure pw (z), the mean
temperature Tm (z) and the temperature difference Td (z). These temperatures have
to be estimated in a thermal pre-processing as presented in [2] or [3].
The displacement of a gravity dam can be modelled with a simple structural

beam model. Fig. 1b) and c) show the static system used for the investigations in
this study. For the calculation of the deformation due to the water pressure pw (z) and
the temperature difference Td (z), the beam is modelled with a symmetrical vertical
axis and a variable thickness tc (z) (Fig. 1b). The elastic foundation is modelled with
two springs and their constants kM and kV , respectively. These constants depend
on the material properties of the rock foundation and can be estimated by the Vogt’s
method described in [8]. Additionally, Young’s modulus Ec , Poisson’s ratio ν c and
thermal expansion coefficient α T must be provided. A model with vertical axis (e.g.
[4]) is simpler to implement but about 10–15 % weaker than a more complex model
with an inclined axis [3]. This can be compensated by using a Young’s modulus Ec
that is slightly larger.
The static system in Fig. 1b cannot be used to calculate the deformation due
to loads stemming from deviations from the mean temperature Tm . Furthermore,
568
Q.106 - R.36
the use of a model with inclined axis and the displacement in horizontal direction
estimated by the horizontal component of the beam extension would lead to an
underestimation of the displacement. Thus, a model similar to the one described
in [6] was used (Fig. 1c). In this model, the gravity dam is considered to consist
of two beams, one upstream (index us) and one inclined downstream (index ds),
that build a bipod. The geometry of the bipod is defined by the y-coordinate of the
upstream dam face yusf (z). By the definition of tc (z) and yusf (z), the y-coordinate
of the downstream face ydsf (z) is implicitly defined. With these coordinates, the
inclinations α usf (z) and α dsf (z) can be calculated. Since this static system is statically
determinate, the stiffness and therefore the thickness of the bipod is not relevant.
Further, the geometry is defined by the abutment level za and the crest level zcr .
Fig. 1
a) Vertical section of a gravity dam with acting loads; b) beam model for the
calculation of the displacement due to the water level and the temperature
difference Td ; c) bipod beam model to calculate the displacement due to a change
in mean temperature Tm .
a) Section verticale d’un barrage-poids avec charges; b) modèle de poutre pour le
calcul du déplacement dû à la pression hydrostatique et à la différence de
température Td ; c) modèle de bipied pour le calcul du déplacement dû au
changement de température moyenne Tm .
2.1.2. Equations and solution procedure
Due to the static determinacy, the deformation can be calculated directly by

applying the work theorem [5]. In doing so, a unit load F in the direction and on the
level zli of the deformation of interest is applied (see Fig. 1b). The deformation due
to water load for a water depth h can be calculated by
za Vy,i Vy,j Mx,i Mx,j RV ,i RV ,j RM ,i RM ,j
δh = + + + dz (1)
zli G Av,y E Ix kV kM
569
Q.106 - R.36
where Vy,i and Mx,i are the stress resultants due to the unit load F, Vy,j and Mx,j due
to the water level h. Ry,i and RM ,i are the reactions due to the unit load, and RV ,jj and
RM ,j due to h. The same static system can be used to calculate the deformation due
to a linear temperature difference Td (z) by the work theorem:
Zli Td (z)
δTd = αT Mx,i dz (2)
Za tC (z)
The displacement δ Td depends linearly on the thermal expansion coefficient

α T . The displacement due to a change in the mean temperature Tm (z) can be
obtained by
zcr
α T zF Tm (z)
δTm = Mx,i dz (3)
zcr − za za tan (αusf ) − tan (αdsf )
The above described equations are estimated by numerical integration. For

the integration of Eq. (1) to (3), a spatial discretisation with element lengths dz has
to be chosen. To be on the safe side, an element length of dz = 2 m was considered
for the applications [3]. The numerical integration is done with a midpoint Riemann
sum with constant element length dz. Since the temperature information is not mea-
sured continuously over the height but only at certain levels, it must be interpolated.
This is done by a linear interpolation between the different levels and a constant
extrapolation as described in [7].
Influence functions can be used for the calculation of the deformation. This
means that the deformation for a certain level li is calculated for unit shear force
Vy,j = 1, unit bending moment Mx,j = 1 and unit reaction forces Ry,j = 1 and
RM ,j = 1. Furthermore, the material properties can be set to 1 for the integration.
Consequently, the integration has to be performed only once for different load cases
and different material properties. After integration, the deformations of the individ-
ual components (shear, bending, elastic foundation) can be summed up according
to superposition. This makes the calculation of the deformation very efficient and
convenient for nonlinear optimisations.
2.2. MARKOV CHAIN MONTE CARLO CALIBRATION
The Markov Chain Monte Carlo (MCMC) method is used for nonlinear cali-
bration of the beam model. In this procedure, the material properties are optimised
such that the deformation on several level fits best.
570
Q.106 - R.36
2.2.1. Bayes theorem
Consider the deterministic model described in the previous section with its
material properties Ec and ν c and an elastic supporting with spring constants kM
and kV . From material tests and engineering knowledge, the mean value and the
standard deviation of the material properties and the spring constants are known.
With this knowledge, the deformation can be calculated. However, if there is further
knowledge, e.g. measurements of the displacement for certain loading conditions,
the material properties and the spring constants can be adjusted. The simplest
way would be to adjust material constants by a least squares procedure. How-
ever, if there are collinearities, this will lead to physically meaningless results. This
problem can be overcome by the use of prior knowledge in terms of Bayesian
inference.
Bayes’ rule (Eq. (4)) says that the prior distribution π (|y) of the parameters
(the material properties and the spring constants in the example of the beam model)
given the measured data y is proportional to the likelihood L(y|) of observing the
data y given the parameters multiplied by the prior distribution of the parameters
π() [9].
π(|y) ∝ L(y|)π() (4)
Eq. (4) can be examined for arbitrary parameter sets . Ideally, every possible
combination of parameters would be used but this would be very computationally
intensive. A common way to approximate the posterior distribution is to draw samples
from it by the MCMC method [9].
2.2.2. Algorithm
Markov chains define the path of the walk through the parameter space from
which the samples are drawn. By definition, they follow an autoregressive process
of first-order (AR(1)). There are different algorithms how the path of the Markov
chains can be described, the simplest one is a common random walk [10]. In this
study, the algorithm of Braak [11] with parallel Markov chains is used. This algorithm
considers the differential evolution algorithm of Storn and Price [12] (DE-algorithm)
for the update. The advantages of this algorithm are the simple implementation and
the convergence and therefore the good efficiency even if some parameters are
correlated [11]. The pseudo-code of the DE-MCMC algorithm is given in Fig. 2 (for
details see [11]).
The parameter γ de scales the step size to get accurate acceptance rates. It
can be chosen to γ de = 2.38p /2p, where p is the number of parameters in the model
[11]. The so-called temperature Tmh has an influence on the convergence as well. It
can be set as a variable or fixed, here a fixed value is used.
571
Q.106 - R.36
Fig. 2
DE-MCMC-algorithm according to Braak [11] and Storn & Price [12].
Algorithme DE-MCMC selon Braak [11] et Storn & Price [12].
2.2.3. Adequacy checking
The resulting samples shall be checked for convergence. Further, the auto-
correlation and the acceptance rates should be checked. The Markov chains start
from the random sampled initial values. Until a stationary ergodic distribution is
reached, several sampling steps are needed. This period is called burn-in and
572
Q.106 - R.36
must be removed before the distribution is analysed [13]. The length of the burn-in
period can be determined by the potential scale reduction factor PSRF of Brooks
and Gelman [14] that should be less than 1.1. The Markov chain samples of
the posterior distribution are autocorrelated by definition. To obtain independent
samples, the resulting Markov chains have to be thinned out. The lag that has
to be applied can be determined by the use of an ACF-plot. The acceptance
rate defines how much of the sampled parameter sets q are accepted by the
Metropolis-Hastings algorithm. An acceptance rate of around 0.25 shall be targeted
for multi-dimensional parameter spaces [15]. A high acceptance rate is an indicator
for poor convergence and there might be some regions that are not explored by the
chains. On the contrary, a low acceptance rate indicates that the chains move fast
through the parameter space and therefore the borders of the parameter space are
reached frequently [13]. Convergence is checked using the potential scale reduction
factor [14].
Fig. 3
DE-algorithm of Storn and Price [12]; a) step to next parameters j is defined by
the difference between the parameters of the two randomly chosen chains c1 and
c2 ; b) if the resulting parameters lie outside the range [ll , ul ], the vector is
mirrored at the limit (adapted from [11]).
Algorithme DE de Storn et Price [12]; a) le pas vers les paramètres j suivants est
défini par la différence entre les paramètres des deux chaînes choisies au hasard
c1 et c2 ; b) si les paramètres résultants se situent en dehors de la limite [ll ,
ul ], le vecteur est reflété à la limite (adapté de [11]).
2.2.4. Results
The results of the MCMC analysis are probability density functions of the sam-
pled parameters from the physically-based model. For further analysis, the most
probable parameters, i.e. the set of optimal parameters can be extracted.
573
Q.106 - R.36
3. CASE STUDY
3.1. ROBIEI GRAVITY DAM
The Robiei hollow gravity dam was built in the 1960s and features a central
block with a height of 68 m. The downstream face has an average slope of 1:0.75 and
the north-east oriented upstream face is vertical. To save concrete, small hollows
were placed inside the dam. Since they are small, the dam is treated as a common
gravity dam in this study. The reservoir is used as a regulating reservoir for pumped
storage operation. A pendulum and an inverted pendulum are used to measure
the displacement M on five levels. In addition, 18 concrete temperatures and three
water temperatures are measured on three different levels (Fig. 4). The readings
were taken manually once a month until 2006, since then they are recorded bi-
weekly.
Fig. 4
Robiei hollow gravity dam. Cross section with positioning
of the measurement equipment.
Barrage-poids à joints évidés de Robiei. Coupe transversale avec position de
l’équipement de mesure.
574
Q.106 - R.36
3.2. MODEL CALIBRATION
Before multi-objective calibration is done, the pre-processing of the data was

performed. These steps contained outlier analysis and heat conduction analysis to
compute mean temperatures Tm and temperature differences Td .
The sampled parameters are the Young’s moduli of concrete Ec and rock Er and
the thermal expansion coefficient of concrete α T . To consider independent spring
constants, two Young’s moduli Er are sampled, one for kV and one for kM . From the
sampled Young’s modulus of rock Er , the spring constants are estimated with the
Vogt’s method. The ratio b/a of an equivalent foundation area was estimated to be 10.
The Poisson’s ratios of concrete and rock are assumed to be constant with ν = 0.2.
The prior distribution of the concrete properties was assumed to be Gaussian with
Ec = 30 ± 10 GPa, αT = (1 ± 0.5) ∗ 10−5 1/K. Since knowledge about the stiffness of
the rock foundation is limited, a wide uniform distribution with Er = 30 ± 30 GPa is
chosen.
16 parallel chains with 10’000 iteration steps are chosen for the MCMC algo-
rithm. In addition, a temperature of TMH = 0.02 was used, which leads to reasonable
acceptance rates. To perform multi-objective-calibration and adapt the model to fit
on several measurement levels li (given as indices), the following likelihood function
with ei as the model residuals is used.
L(y|) = L(e1941 |)L(e1928 |)L(e1910 |)L(e1889 |)L(e1878 |) (5)
The prior information is considered by
π() = π(Ec )π(αT )π(Er,kV )π (Er,kM ) (6)
These equations are optimised by the MCMC algorithm described above.
3.3. RESULTS
3.3.1. Convergence
In Fig. 5, the PSRF is given for the different parameters. Convergence

(PSRF < 1.1) is reached very fast. Since there are enough iteration steps, the burn-
in period was set to iteration steps 1 - 2’500 which means that 7’500 samples are
left for further analysis. The acceptance rates of the individual chains are between
0.27 and 0.29, which is close to the target value of 0.25 [15].
575
Q.106 - R.36
Fig. 5
PSRF for Young’s moduli of a) concrete Ec ; b) rock due to shear force Er,kV ; c) rock
due to bending moment Er,kM ; and of d) thermal expansion coefficient αT .
PSRF pour les modules de Young de a) béton Ec ; b) roche dû à la force de
cisaillement Er,kV ; c) roche dû au moment de flexion Er,kM ; et à d) coefficient de
dilatation thermique α T .
3.3.2. Correlation structure
The correlation structure (autocorrelation) of the sampling decreases within

40 lags. Thus, only every 40th sampling parameter set is used to obtain indepen-
dent samples. This results in 187 independent parameter sets that can be used to
determine the joint probability distribution.
3.3.3. Parameters
In Fig. 6, the prior distributions (blue dashed lines) and the posterior
distributions (red solid lines) are given for the different parameters.
576
Q.106 - R.36
Fig. 6
Result of the MCMC calibration. The dashed blue lines correspond to the prior
distribution and the solid red lines to the posterior distribution of the parameters .
The blue points show the correlation between them. In addition, the correlation
coefficient is given in the upper right part.
Résultat de la calibration du MCMC. Les lignes bleues pointillées correspondent à
la distribution antérieure et les lignes rouges continues à la distribution postérieure
des parameters . Les points bleus montrent la corrélation entre eux. En outre, le
coefficient de corrélation est indiqué en haut à droite.
Although the prior distributions were chosen very broad, especially for the
Young’s modulus of rock and the thermal expansion coefficient, the parameters
converge well. The posterior distribution of Ec is smaller than those of Er,j . This
means that there is more uncertainty in the elastic abutment than in the concrete
properties. In addition, a slightly negative correlation of r = 0.38 between Ec and
Er,kM was detected. Based on the MCMC samples, the joint probability function can
be determined, leading to Ec = 30.2 GPa, Er,kV = 48.8 GPa, Er,kM = 34.0 GPa. and
α T = 0.91 · 105 1/K.
577
Q.106 - R.36
3.3.4. Behaviour analysis
Based on the resulting parameters, the displacement due to the effect of

the water level and the temperature is calculated for each level. Subsequently,
the adjusted behaviour indicator Madj = M − Prev is estimated. The concept of the
adjusted behaviour indicator is to highlight irreversible changes of a behaviour indi-
cator by subtraction of the modelled reversible variables Prev (displacement due to
water level and temperature) from the measurements of the behaviour indicator M
[16, 17]. The adjusted behaviour indicators Madj,i are shown in Fig. 7 for all levels. A
slight drift at crest level has occurred. In addition, a shift in the data in 2008 on the
levels equal to and above 1889 m a.s.l. is recognisable.
Fig. 7
Adjusted behaviour indicators resulting from the output of the multi-objective
calibrated beam model (black). In addition, the results of the MLR analysis to
search behaviour changes are shown (red). On all levels except the bottom, a shift
of around 0.5 mm is detected in summer 2008.
Indicateurs de comportement ajustés résultant de la sortie du modèle de poutre
calibré multi-objectifs (noir). De plus, les résultats de l’analyse MLR pour
rechercher les changements de comportement sont indiqués (rouge). Sur tous les
niveaux sauf la fondation, un décalage d’environ 0,5 mm est détecté à l’été 2008.
578
Q.106 - R.36
Then, MLR analysis was used to quantify these behaviour changes. A shift
of around 0.5 mm in 2008 was detected. This shift was not obvious and not visu-
ally recognisable in the raw pendulum readings at crest level. The size of the shift
corresponds to the estimated shifts at levels 1889 to 1928 m a.s.l. (mid level), which
are also around 0.5 mm. Furthermore, there is no shift at level 1878 m a.s.l (bottom
level) in 2008. This leads to the conclusion that there might have been a damage of
the measurement equipment of the direct pendulum at level 1878 m a.s.l. in 2008.
The operator confirmed that there were maintenance works in the pendulum shaft
at that time.
4. DISCUSSION
The case study exemplarily shows the potential of using a beam model in
combination with multi-objective calibration based on Bayesian inference. Only one
set of physically meaningful parameters is calibrated for one block of a gravity dam.
This leads to simple models that allow for comprehensible inference by engineer-
ing judgement. Moreover, a potentially abnormal behaviour can be detected and
correctly located by simultaneous analyses of the displacement on different levels.
Compared to standard MLR models, more effort for the set up and cal-
ibration is needed. On other hand, no variable selection procedure, e.g. the
choice of thermometer readings, is necessary. Thus, multicollinearity, which causes
instabilities in statistical models, is not present in these models.
5. SUMMARY AND CONCLUSIONS
For behaviour analysis of gravity dams, beam models allow for multi-objective
calibration, where the measured displacements are simultaneously matched with
the model output at several levels. The calibration can be automated by methods
based on Bayesian inference, where prior information about the material parameters
can be considered. Therefore, the Markov Chain Monte Carlo method was used as
optimisation algorithm. Since only one set of physically meaningful parameters is
calibrated for one block of the gravity dam, simple models that allow for comprehen-
sible inference by engineering judgement result. Moreover, due to the simultaneous
analyses of the displacement on different levels, a potentially abnormal behaviour
can be detected and correctly located.
ACKNOWLEDGEMENTS
The authors would like to thank the Swiss Federal Office of Energy SFOE for
the financial support of this project (S1/500834-01).
579
Q.106 - R.36
REFERENCES
[1] Swiss Committee on Dams (2003). Methods of analysis for the prediction and
the verification of dam behaviour. Wasser Energie Luft, 95(3/4): 73–110.
[2] Léger, P.; Leclerc, M. (2007). Hydrostatic, Temperature, Time-Displacement

Model for Concrete Dams. Journal of Engineering Mechanics (ASCE), 133(3):
267–277. http: //dx.doi.org/10.1061/(ASCE)0733-9399(2007)133:3(267).
[3] Bühlmann (2018). Statistical Methods for Dam Behaviour Analysis. VAW-
Mitteilung 242, Versuchsanstalt fu¨ r Wasserbau, Hydrologie und Glaziologie
(VAW), ETH Zu¨ rich.
[4] Rescher, O. J. (1965). Talsperrenstatik — Berechnung und Bemessung von

Gewichtsstau- mauern (‘Statics of dams — calculation of gravity dams’).
Springer-Verlag, Berlin, Germany. http://www.springer.com/de/
book/9783540033899 [in German].
[5] Marti, P. (2013). Theory of Structures — Fundamentals, Framed Structures,

Plates and Shells. ISBN 978-3-433-02991-6. Ernst & Sohn, Berlin, Germany.
[6] Kolly, J.-C.; Joos, B. (1995). D’un model simple de determination des deforma-
tions d’un barrage-poids sous l’influence de la temperature (‘A simple model
for the determination of the temperature influence of a gravity dam’). Proc.
Research and Development in the Fiels of Dams. Crans-Montana, Switzerland
[in French].
[7] Schnitter, N. (1969). Die sofortige U¨ berpru¨ fung von Staumauerdef-

ormationen (‘The immediate control of dam displacements’). Schweizerische
Bauzeitung, 87(49): 961–963. http://dx.doi.org/10.5169/
seals-70824 [in German].
[8] USBR (1938). Boulder Canyon project — final reports — part V — technical
investigation — Bulletin 1 — trial load method of analyzing arch dams. United
States, Department of the Interiour, Bureau of Reclamation (USBR), Denver,
USA.
[9] van Ravenzwaaij, D.; Cassey, P.; Brown, S. D. (2016). A simple introduction to
Markov Chain Monte-Carlo sampling. Psychonomic Bulletin & Review: 1–12.
1531-5320. http://dx.doi.org/10.3758/s13423-016-1015-8.
[10] Hastie, T.; Tibshirani, R.; Friedman, J. (2009). The Elements of Statistical
Learning — Data Mining, Inference, and Prediction. Springer, New York, USA.
http://www. springer.com/de/book/9780387848570.
[11] Braak, C. J. F. T. (2006). A Markov Chain Monte Carlo version of the ge- netic
algorithm Differential Evolution: easy Bayesian computing for real parameter
spaces. Statistics and Computing, 16(3): 239–249. http://dx.doi.org/10.1007/
s11222-006-8769-1.
580
Q.106 - R.36
[12] Storn, Rainer; Price, Kenneth (1997). Differential Evolution — A Simple and
Efficient Heuristic for global Optimization over Continuous Spaces. Journal of
Global Optimization, 11(4): 341–359. http://dx.doi.org/10.1023/
A:1008202821328.
[13] Robert, C. P.; Casella, G. (2010). Introducing Monte Carlo Methods with R.
Springer, New York, USA. http://www.springer.com/de/book
/9781441915757.
[14] Brooks, S. P.; Gelman, A. (1998). General Methods for Monitoring Con-
vergence of Iterative Simulations. Journal of Computational and Graphical
Statistics, 7(4): 434–455. http://dx.doi.org/10.1080/10618600.
1998.10474787.
[15] Gelman, A.; Roberts, G. O.; Gilks, W. R. (1996). Efficient Metropolis Jumping
Rules. Proc. Bayesian Statistics 5: Proceedings of the Fifth Valencia Interna-
tional Meeting, J. M. Bernardo; J. O. Berger; A. P. Dawid; A. F. M. Smith (eds.).
Oxford University Press, Oxford, UK: 599–607. ISBN 9780198523567.
[16] Penot, I.; Daumas, B.; Fabre, J.-P. (2005). Monitoring behaviour. International
Water Power & Dam Construction, 57(12): 24–27.
[17] Amberg, F. (2009). Interpretative models for concrete dam displacements.

Proc. 23rd ICOLD Congress. International Commission on Large Dams
(ICOLD), Brasilia, Brasil. Q.91, R.209: 1–20.
581
Q.106 - R.37
-------
-------
NEW CONCEPT FOR CONCRETE SWELLING EVALUATIONS BASED ON

SURVEILLANCE AND MONITORING DATA (∗)
Russell Michael GUNN

Dam Safety Specialist,
SWISS FEDERAL OFFICE OF ENERGY (SFOE)
SWITZERLAND
SUMMARY
Concrete core samples are often extracted from dams with the primary objec-
tive of identifying (diagnostic phase) if the structure is undergoing a phenomenon of
concrete swelling due to alkali-aggregate/silicate reaction AAR/ASR, without nec-
essarily any major concern about the numerical methods and constitutive models
(calibration phase) that may be used to analyze and predict the effects of ASR (prog-
nostic phase) on dam safety and possible rehabilitation works. The main objective of
this document is to present a novel approach to assess firstly how, where and when
core samples are extracted from concrete dams based on in-situ humidity, thermal,
stress and other conditions and secondly, how to combine field observations derived
from surveillance and monitoring data with laboratory test results. The approach is
presented in terms of an easy to understand and follow flow chart that focuses on
surveillance and monitoring methods, material testing, data processing, and simple
numerical concepts that can be readily applied in standard software packages.
RÉSUMÉ
Les carottes de béton sont souvent extraites des barrages dans le but principal
d’identifier (phase de diagnostic) si la structure subit un phénomène de gonflement
∗ Nouveau concept pour l’évaluation du gonflement de béton basé sur les données de
surveillance et d’auscultations
582 DOI: 10.1201/9781003211808-132

Q.106 - R.37
du béton dû à une réaction alcali-granulat ou silicate AAR ou ASR, sans souci majeur
des méthodes numériques et lois constitutives (phase d’étalonnage) qui peuvent
être utilisées par la suite pour analyser et prédire les effets de l’ASR (phase pronos-
tique) sur la sécurité des barrages et les éventuels travaux de réhabilitation. L’objectif
principal de ce document est de présenter une nouvelle approche pour évaluer d’une
part comment, où et quand les carottes sont extraites des barrages en béton en
fonction de l’humidité in situ, de la chaleur, des contraintes et d’autres conditions
et d’autre part, comment combiner les observations in-situ dérivées de données de
surveillance et d’auscultation avec les résultats des essais en laboratoire. L’approche
est présentée sous la forme d’un organigramme facile à comprendre et à suivre ; il se
concentre sur les méthodes de surveillance et d’auscultation, les essais de matéri-
aux, le traitement des données et des concepts numériques simples qui peuvent
être facilement appliqués dans des progiciels standard.
1. INTRODUCTION
Between 1945 and 1970, Swiss hydropower and the construction of many
concrete dams went through the boom period and accounted for almost 90% of the
domestic energy production. By 1985, with the commissioning of five nuclear power
stations, this figure fell to 56%, still representing an important reliance on hydropower
as a source of renewable energy. Today, 69 gravity, 51 arch, 2 multiple-arch and 2
buttress concrete dams and 27 river weirs comprise the portfolio of hydropower
schemes in Switzerland. Four of these large dams are greater than 200 m and 25
are greater than 100 m. Concrete dams therefore represent an important source of
energy supply and revenue in Switzerland.
In Switzerland, mass concrete dams and their associated normally reinforced

appurtenant structures (spillways, water intakes, etc.) are ageing and some have
already been rehabilitated following the swelling effects of AAR (alkali-aggregate
reaction) [1]. These rehabilitation works and related studies were performed over a
period of more than 20 years running in parallel to the development of AAR research
projects that have led to a better understanding of AAR within the context of dam
engineering [2–5].
In the early stages of AAR identification many conflicting test results were
obtained about the potential or existence of AAR in dams. In one particular case
(a gravity dam), for which in-situ irreversible displacements and visual observations
intimated the presence of AAR, subsequent laboratory testing indicated that on
a petrographic level the aggregates were potentially reactive, however on a core
sample level (residual expansion tests) the mass concrete was far below AAR limits
and therefore potentially non-reactive. Shortly afterwards, for the same dam core
samples modified laboratory testing procedures and a critical reinterpretation of the
initial expansion phase due to water absorption indicated the contrary of previous
results and that the entire structure was comprised of potentially reactive materials.
583
Q.106 - R.37
Rehabilitation of the structure then consisted of releasing AAR induced stresses by

saw-cutting selected vertical joints in the dam body.
Experience gained from such projects highlights many important issues

related to the way that AAR is and can be realistically identified within Swiss dams.
This paper capitalizes on past studies that have been performed under partial or full
mandate of the SFOE such as the EPFL [2, 3], [4, 5] and EMPA [6, 7] by combining
these findings with concepts related to dam safety surveillance, the treatment and
evaluation of in-situ monitoring data and other concepts [8, 9], [10, 11]. Hence,
both research and dam safety works can be realistically structured, performed
and especially, validated. This “gluing” together of different industries proportion-
ally strengthens the consensus rather than settling for compromises that only lead
to more uncertainties.
2. AAR IN THE CONTEXT OF DAM ENGINEERING AND SAFETY
Many large Swiss concrete dams are built on gneiss, granite limestone and
slate foundations. Typically, the aggregates in dam concrete mixes use locally
sourced materials. The problematic minerals (mainly strained quartz) in some of
these aggregates may slowly (20 to 30 years post-construction) react to different
degrees with the alkaline pore solution (pH > 13) of concrete which in a moist envi-
ronment (> 80%) may lead to concrete swelling. AAR encompasses reactions with
silica Si O2 (ASR), silicates (ASSR, slower rates than ASR e.g. feldspar, mica, etc.)
and carbonates (ACR). Herein the focus is ASR.
2.1. GEOLOGICAL CONTEXT AND SUPPLEMENTARY CEMENTITIOUS MATERIALS
The geological context within which the dam is located, hence the reason why
herein we specifically refer to dams in Switzerland, needs to be well understood to
identify and understand the ASR phenomenon. ASR in highly reactive aggregates
such as Opal is identified by dark rims around the aggregate, whereas for granite
and gneiss, the ASR reaction process starts within the aggregate. Investigations on
mortar-bars, MBT [12] have shown that from 79 quarries distributed over Switzer-
land, approximately 85% of the rock sources can be classified as potentially reactive
(expansion > 0.11%) and that the slow reaction process starts within the aggregate
[6, 8]. This information already gives us some idea about what to expect from stan-
dard AAR testing [12, 13], [14, 15]. To avoid thermal cracking, mass concrete dams
typically require a two-staged process of thermal control: pre-and post-cooling [16].
The pre-cooling phase addresses the heat of hydration that can also be controlled
by supplementary cement materials, SCM such as pozzolan that also acts as a
countermeasure against the effects of ASR.
584
Q.106 - R.37
In conclusion, for a given concrete dam, it is vital to clearly identify both the
aggregate types and SCMs used in mass concrete mix designs, of which there may
be many within the dam body, so as to be able to identify the signs of ASR both in
terms of the spatial distribution and extent per pocket of ASR gel.
2.2. BEHAVIOR OF DAMS UNDERGOING EXPANSION DUE TO AAR
ICOLD [17] provides a useful insight into the swelling of concrete dams due to
AAR whereby it has been stated that “there are no potentially catastrophic or rapid
failure safety issues raised by AAR in dams”. However, the extent of the phenomenon
world-wide, the impact on the operation of hydraulic equipment governing the control
of reservoir storage volumes and in the general safety evaluation of the structure to
fulfil its design purpose, emphasizes the importance of understanding AAR within
the context of dam engineering and safety.
Generally [1], the expansion of concrete may be between 20 to 150

µm/m/year, whereby a 100 m high dam expands vertically by 2 to 15 mm/year
and a dam with a 250 m long crest length expands along its axis by 5 to 40
mm/year. In Switzerland [1], the mean expansion rate estimated in the verti-
cal direction varies between 1 and 30 µm/m/year and for a few dams for which
repair works have been executed, the total expansion reached 500 to 1000
µm/m. AAR expansions may be generally spread throughout the entire dam
or more other than not, localized to specific areas such as the crest, mid-
section or base and may generate compressive, tensile or mixed mode stresses
depending on the boundary conditions and hyperstatic characteristics of the
structure. Comparing AAR expansion values with those typically generated in
the dam due to self-weight and full hydrostatic pressure, it becomes apparent
that the magnitude of AAR irreversible displacements can be quite substantial.
Typically, vertical uplift or heave displacements and an upstream horizontal drift,
especially at the crest elevation, are associated with cracking in other areas of the
structure which may be classified as structurally stable or unstable [18]. Micro and
meso-scale cracking also leads to a reduction of tensile, compressive and elastic
modulus material properties in the order of 50 to 70%, 25 to 60% and 60 to 70%
respectively [1].
2.3. FACTORS INFLUENCING THE IDENTIFICATION OF AAR IN EXISTING DAMS
Typically, we start AAR investigations with the observed global behavior of the
structure based on visual inspection, dam instrumentation and/or geodetic readings
that show a progressive increase with time in irreversible deformations that could
be attributed to AAR and that cannot be attributed to other behavioral phenomena
585
Q.106 - R.37
such as plasticity and/or creep. Experience has shown that the identification and the
measure of extent of AAR in dams is highly dependent on the selection of extracted
material samples, whether it be from aggregate quarries used during the construc-
tion period, noting that quarries are difficult to access many years after construction
for technical and economic reasons, or crest, upstream, downstream faces of the
dam or within the structure such as cores extracted from control and access gal-
leries. The investigative process is often hampered by the fact that the visual signs of
AAR such as gel exudations, surface map cracks lined with dark rims and expansion
drifts are not always visible (low strains) [19] making it difficult to select specimens
from non-affected (reference specimens) and affected areas of the structure from
which core samples should be extracted and tested in the laboratory.
Clearly, all the actors in AAR studies such as geologists, material scientists,
engineers, analysts and dam owners/operators need to know the factors that influ-
ence and impact on how, where and to what extent AAR is identified in the structure
as given below.
Constituents of mass concrete:
• Aggregate types, noting also that mass concrete unlike structural concrete
may have maximum aggregate diameters of 120 mm and shapes that can be
angular (extracted from rock quarries) or rounded (extracted from rivers beds)
• Cement or cementitious material types that are typically selected to have a
low heat of hydration and the quantities per cubic meter ranging from 150 to
400 kg/m3 of concrete Fig. 1.
External boundary conditions:
• Temperature, humidity and multi-axial principal stress conditions in terms of

magnitude and direction [4].
Core samples:
• Size of extracted cores, how they are handled, prepared and tested. This
is very important for short and long-term AAR assessments [7] whereby the
recommended diameter and length of core samples are 100 mm and 200 mm
respectively.
In general, the characteristics and behavior of the structure as well as the

material properties at the micro, meso and macro scales are needed to benefit
from current state-of-the-art field and laboratory investigation methods as described
below.
586
Q.106 - R.37
Fig. 1
Schematic view of elements affecting the selection of AAR testing core samples.
Vue schématique des éléments affectant la sélection des carottes d’essais AAR.
3. AAR MATERIAL INVESTIGATIONS
Experience in Switzerland demonstrates that a clear strategy for AAR material

investigations is needed whereby clear goals at the very onset of the studies are
defined.
There is a real need to distinguish between those tests related to identification,

extent and prognosis of AAR to be able to define for example, rehabilitation AAR
counter measures.
3.1. AAR PREVENTION FOR NEW DAMS
Swiss Technical Note [12] unifies and updates existing AAR testing procedures
and methods mostly for new structures and formulates a three-level procedure (P1,
587
Q.106 - R.37
P2, & P3) to prevent AAR. Dams in Switzerland are mostly classified as P3 and
would therefore be subjected to all three levels of AAR testing.
3.2. AAR INVESTIGATIONS FOR EXISTING DAMS
The following strategy is proposed for the investigation of AAR in existing

dams:
• Phase 1: AAR stage tests, material source(s), e.g. quarries, and visual inspec-
tion and core sampling should cover all stages of the AAR process in the
short, medium and long term of the structures service life including infor-
mation needed for numerical modelling (constitutive material law calibration
tests). The stage or extent of AAR dictates the type of AAR tests that are to
be performed.
• Phase 2: AAR monitoring and data analysis. Using statistical, deterministic
(finite element modelling) and hybrid data analysis techniques to capture the
potential zones of AAR swelling (visible and non-visible) and hence, perform
analyses to assess the behavior and safety of the dam.
• Phase 3: AAR follow-up tests should be systematically performed every
5 years (expert review period for dams under Swiss direct surveillance guide-
lines) from the onset of AAR. Field, laboratory and desktop studies are
proposed herein to define the scope of testing.
The details of these tests have been presented in [20] and consist of
Phase 1 Material testing “Look-up tables” that split AAR testing into three distinct
parts:
• Classification of AAR tests by category, location (field, laboratory), phase

(design, build, service life), type (chemical, mechanical, numerical, visual,
crack index, optical and non-destructive), sample type (structure, core, pol-
ished slices and constituent materials such as aggregates), classification
(qualitative, quantitative, identification, severity, prognosis), reliability (well-
established, established, rarely used, research domain, test still in its infancy),
quality of information (excellent, good, satisfactory, poor, unsatisfactory) and
sample preparation (difficult, standard, easy or none).
• Influencing Factors: Specimen size, quantity of cement, additives, relative
humidity, temperature (◦ C), time (days) and cost (CHF).
• Results and Conclusions: Standard limits above which AAR is deemed to be
potentially deleterious or otherwise, laboratory and field observations with an
AAR classification index from levels 1 to 4 indicating low to high amounts of
588
Q.106 - R.37
AAR respectively and finally, a summary of the advantages and disadvantages

of the tests.
4. PROPOSED METHOD FOR DAM AAR IDENTIFICATION, EXTENT AND

PROGNOSIS EVALUATIONS
The proposed process for AAR evaluations is given in Fig. 2, highlighting:
• Where to select cores: To make the Engineer aware of the importance of

required strength zoning (cement content per m3 of concrete), at the dam
faces, core section and dam-foundation interface and of principal stress orien-
tations, the geographic orientation of the structure and ambient temperature
conditions, five potential AAR investigation zones are proposed Fig. 1, Fig. 2,
circled “concrete zones” (1) to (5).
• Zoning: Core drillings should ideally intersect several concrete zones and it
is recommended to discard the first 10 to 15 cm of core drilling that may
have suffered damage from frost, freezing and thawing. In addition, care
should be taken to avoid horizontal lift and vertical contraction joints since they
may have different material properties compared to the main dam concrete
zones.
• Surveillance and monitoring: Dams do not necessarily swell uniformly and
AAR may not always be visible to the eye. Hence, the application of statistical
and deterministic methods is proposed for four key regions within the structure
called “AAR Dam zones ↓” A to D. This information helps to select core loca-
tions, monitor the long-term behavior of the structure and calculate a fictitious
expansion/crack (micro) opening (CF ITN ) index that enables field and labo-
ratory measurements to be compared in both a qualitative and quantitative
manner.
4.1. PHASE 2: AAR MONITORING AND DATA ANALYSIS
The process of AAR identification starts with dam surveillance and moni-
toring data. AAR may be fully or partially visible to the eye (Direct Approach)
or non-visible (Indirect Approach). For the former, a Crack Index CITN (mm/m.
year) based on the LCPC method can be evaluated and core samples taken
accordingly [21]. For the indirect approach, four AAR zones (crest, crown base,
mid-height section and abutments) are prioritized and AAR displacements are
estimated using statistical and/or deterministic (finite element) methods. The
main objective at this phase of the studies is to compute a fictitious AAR crack
index CF ITN (mm/m. year) as a function of a given section height for the four
AAR zones.
589
Q.106 - R.37
Fig. 2
Schematic view of elements affecting the selection of AAR core samples [20].
Vue schématique des éléments affectant la sélection des carottes AAR.
590
Q.106 - R.37
4.1.1. Application of statistical data monitoring methods to AAR
Statistical methods uncouple the causes and effects and enable reversible &
irreversible displacements (drifts) such as those due to AAR to be estimated [11].
The procedure involves data regression analyses of the control (comparison) and
prediction data periods based on the selection of mathematical models for water level
fluctuations, temperatures (concrete, air, reservoir) and an exponential term(s) for
inelastic behavior (AAR, damage, etc.) [28]. The mathematical models for inelastic
behavior require a clear understanding of the potential phenomena and engineer-
ing judgement. The method and software originally developed by [28] and updated
by [11] uses the t-test (importance of variables), a variance inflation factor (to avoid
matrix ill-conditioning) and a user-friendly software tool that allows model variables
to be switched on and off. Statistical analyses can be applied over part or the entire
life-span of the structure and hence before and after the onset of AAR which nor-
mally develops some years after the first impounding of the reservoir [1]. Hence,
it seems possible to make a first estimation of the AAR displacements based on
statistical methods, for example using the software tool developed in Switzerland
called DamBASE [11] and/or deterministic methods as described in chapter 5.
Having defined/validated the location of AAR within the dam and computed
the fictitious AAR crack index CF ITN (mm/m. year) as a function of section height,
the process of AAR identification and testing may start in the laboratory and field for
selected core samples and zones of the dam.
4.1.2. Proposed AAR identification and extent procedure
The first round of laboratory tests which are mainly used for the prevention
of AAR in new structures, petrographic, MBT, CPT, etc. shall define if the AAR
limits have been exceeded and if so, as a first priority, optical microscopy should
be performed over the entire AAR investigated area(s) of the structure followed by
image microscopy (SEM, EDX) consisting of thin polished core slices impregnated
with uranyl acetate so that AAR becomes more visible under UV light. Scanning
Electron Microscopy, SEM and energy dispersive X-ray, EDX tests are performed to
identify AAR and the chemical composition of the sample respectively.
Optical microscopy on core samples defines a third measure of crack index,

crack sample index, CS ITN (mm/m. year) allowing the extent of AAR to be assessed
for different parts of the structure at a given time and principal stress direction within
the structure.
This method provides the Engineer with three indexes that define the extent
of AAR that have also been found to correlate with MBT and CPT tests thus clearly
facilitating the identification, extent and prognosis of AAR in dams [6].
591
Q.106 - R.37
4.2. PHASE 3: AAR FOLLOW-UP TESTS
Fig. 2 presents the long-term strategy for AAR investigations that is split into
two parts: Prognosis and Periodic Review (every 5 years) for the cases when AAR
limits are and are not exceeded respectively. The former implies residual expansion
tests on core samples [6] and numerical modelling and the latter, continued dam
surveillance and data monitoring.
In conclusion, following the step-by-step procedures given in Fig. 2 allows dam

owners, authorities and other institutions to arrive at a consensus on how to handle
AAR that is both effective and proportionally measured for the given dam.
5. DETERMINISTIC APPROACH TO AAR IDENTIFICATION,

EXTENT AND PROGNOSIS
The objectives of the approach presented herein are to identify areas or zones
of AAR in the structure (upstream and downstream faces) and fictitious AAR expan-
sion/crack displacements and openings, CF ITN for part of or the entire dam. This
information is extremely useful for the selection of core samples when AAR is not
visible to the naked eye, confirms visible AAR zones, provides a measure of AAR
extent and gives an indication of the long-term effects of AAR.
The proposed idea is essentially a derivative of the deterministic approach

described in reference [1]. This approach can also be used with statistical methods
such as DamBASE [11] and other closed form solutions for thermal loading [16].
Deterministic models have already been used in the evaluation of AAR. In their
complex form, constitutive material laws may be used to predict the effects of AAR at
the micro, meso and macro scales [3, 5, 23]. It is noted that the formulation, imple-
mentation, calibration against material tests and the ability to consistently model the
physical phenomenon in the medium to long term still remains a challenge [22].
5.1. FORMULATION OF THE DETERMINISTIC APPROACH TO AAR
Firstly we calibrate a finite element model of the structure (dam and foundation)
subjected to normal operation loads for a period of time without the effects of AAR
(say the first 5 to 10 years of AAR free operational service). Hence, the mechanical
material properties (moduli, Poisson’s ratio and coefficient of thermal expansion)
are defined. Time-dependent creep effects, normally predominate at the early life
cycle of the structure are implicitly included in the calibrated moduli values of rock
and concrete [25].
592
Q.106 - R.37
The finite element mesh is overlaid with a grid of intersecting lines consisting of
vertical lines where instrumentation is present such as pendulums, geodetic targets,
etc. and horizontal bands of unit temperature control sections (ranging from 1◦ C to
0◦ C) used for both external and internal fictitious AAR temperature loadings ideally
positioned where temperature device measurements exist. Hence, AAR swelling can
be quite simply simulated by an increase or decrease in temperature at intersecting
control points.
The FEA displacement response of the structure (dam and foundation) for
annual variable hydrostatic with a time step of weeks to months and unit temper-
atures defines influence lines for which control point coefficients can be stored in
a spreadsheet. Hence, the response of the structure at a given time for a given
hydrostatic and unit thermal loading can be obtained directly from the spreadsheet
database and no further FE analyses are required.
Using the spreadsheet, the known reservoir level and external ambient tem-
peratures at a given time provide control point (instrumentation reading points)
displacements.
The difference between the computed and measured values can be attributed
to AAR and temperatures fixed for AAR horizontal bands at different elevations until a
match is obtained between the computed and measured values using a spreadsheet
solver. The resulting AAR fictitious temperatures define AAR core sampling zones.
The principle of superposition for a linear system is used to sum the effects of
mechanical loads (hydrostatic pressure and thermal loading) and hence, the global
displacement at point M is given by:
u(M ) = U T (M ) + U M (M ) + U AAR (M ) (1)
Where is the displacement field of the dam due to the thermal loading and the
displacement field due to mechanical loading (i.e. hydrostatic pressure). The
mechanical displacement field is written as:

K
U M (M ) = M
αk uWL k
(M ) (2)
k=1
Where uWLM (M ) is the displacement at point M due to the hydrostatic pressure for a
k
water level of WLk . Activation coefficient αj is defined by αk = 1 if WLk is close enough
to the considered water level, otherwise zero. Hence, only one αj coefficient is equal
to 1, the others are 0. Different mechanical loadings are given by the K water levels
WLk (base to the crest elevation in steps of say 5 or 10 m) where K is the number
of hydrostatic loads. Elementary loadings are applied within the model to evaluate
and store the displacement fields at instrumentation locations (e.g. plumblines).
593
Q.106 - R.37
Now the thermal displacement field is written in accordance with the theory of
influence lines given by elementary thermal loads:

N
U T (M ) = Ti uiT (M ) + δTi viT (M ) (3)
i=1
Where: uiT (M ) is the displacement at point M due to the elementary thermal load
ref
case i defined by the temperature field fiT (M ) (see below); Ti − Ti is the constant
relative temperature and viT (M ) is the displacement at point M due to the elementary
temperature gradient δTi through the dam at level i. Finally, N is the number of
ref
elementary thermal load cases and Ti is the reference temperature at level i. The
constant and gradient temperature components define the temperature variation
across the thickness of the dam.
These terms can be computed from the thermocouples in the dam section or
from external water and air temperatures used in the STUCKY-DERRON method
[16] to provide equivalent constant and gradient temperatures at a given location
and time.
The elementary load cases have to verify the partition of unity, such as:

N
1= fiT (M ) (4)
i=1
The elementary temperature fields fiT (M ) are defined below noting that a linear
variation applies for intermediate values:
fiT (M at level j = i) = 1 and fiT (M at level j = i) = 0 (5)
To demonstrate the calculation method, the case for six temperature fields
is given in Fig. 3. The difference between computed and measured displace-
ments is attributed to AAR which is then simulated only by an increase in constant
temperature for each level i.
The resulting AAR zones can be concluded on the basis of fictitious temper-
ature zoning as shown in the Fig. 5, whereby both the intensity and location of
possible AAR reactions is now exposed or made visible.
The entire method is summarized in Fig. 4.
594
Q.106 - R.37
5.2. SCHEMATIC RESULTS OF THE DETERMINISTIC APPROACH APPLIED TO AAR

INVESTIGATIONS
Applying the concepts presented above should lead to the results given in
Fig. 5. Clearly the schematic presentation of AAR zones in Fig. 5 indicates where
core samples should be taken and also provides the analyst with a good idea about
how AAR swelling displacements could be predicted by an equivalent temperature
distribution over the entire structure.
Such an equivalent temperature distribution over the height and width of the
dam (left to right banks) to simulate both non-uniform horizontal and vertical expan-
sions has also been presented within the context of AAR investigations of a dam in
Switzerland [1].
Fig. 3
Temperature field influence lines.
Lignes d’influence du champ de température.
Fig. 4
Influence line Method for AAR Core Sample Zoning.
Méthode de la ligne d’influence pour le zonage d’échantillons de carottes AAR.
595
Q.106 - R.37
Fig. 5
AAR Zoning for Core Sample Selections (colors indicate different levels of affected
and non-affected AAR regions).
Zonage AAR pour les sélections d’échantillons de base (les couleurs indiquent
différents niveaux de zones affectées at non-affectées par AAR).
Contrary to previous investigations, the approach described herein unifies the

notion of an equivalent temperature increase or decrease at specific regions to
obtain the swelling displacements of the structure with the physical phenomenon
rather than an artefact (temperature rise/lowering) to obtain a desired result (AAR
swelling displacements).
Hence, the AAR equivalent temperature loading also has a physical meaning
that the dam engineer can associate with the real behavior of the structure improving
the knowledge base decision process related to risk and safety.
The method above provides only an approximation of the effects of AAR on

the structure because differential displacements (difference between measured and
computed values) can also be attributed to other irreversible mechanisms such as
plasticity, creep and damage [26, 27].
However, with good field measurements, appropriate statistical/deterministic

data evaluations and engineering judgement, it should be possible to not only define
the potential areas of AAR, but also to get a much better insight into the behavioral
mechanisms of the swelling structure that of course shall be very important when
coming to the conception, design and analysis of rehabilitation works.
6. AAR PROGNOSIS TESTING
AAR prognosis evaluations may consist of 170-day residual expansion tests

performed on core samples with a recommended size of 100 mm diameter and
200 mm in length taken from both affected and non-affected AAR areas. These
596
Q.106 - R.37
tests are described in the reference [7] for which the results have been validated
& correlated against CPT and optical microscopy test results (expansion and crack
opening respectively). These findings are important since they establish several
important links related to AAR tests performed on concrete for new and existing
structures.
The residual expansion test overcomes the problem of using aggregate

sources (quarries from the construction period) that may be encountered for MBT
and CPT tests and moreover, it provides additional information related to the status
of irreversible AAR swelling (expansion at a given time, extent and rate).
The proposed method does not consider the initial water swelling/creep effects
at the initial phases of the test and therefore targets true AAR. To better understand
the results and conclusions of this work, the measured longitudinal expansion on
core samples the swelling phases and test results are schematically outlined in
Fig. 6.
Fig. 6
Different Phases of AAR Residual Expansion Prognosis Tests based on
Longitudinal Direction Expansion Measurements on Core Samples [7].
Différentes phases des tests de pronostic d’expansion résiduelle AAR basés sur
des mesures d’expansion en direction longitudinale sur des échantillons de
carottes [7].
7. CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER STUDIES
The identification, extent and prognosis of AAR remains a subject of great

importance in dam engineering and safety evaluations. This contribution presents
597
Q.106 - R.37
a methodical and detailed approach to AAR investigations for existing dams in

Switzerland that remains to be validated against case studies.
An overview of AAR within the context of Dam Engineering and Safety with spe-
cific emphasis on the geological properties of aggregates, the use of supplementary
cementitious materials, cement zoning within the dam, principal stress orientations,
environmental conditions (temperature and humidity), the behavior of dams under-
going expansion due to AAR with typical expansion values in mm per year and the
determining factors that influence the identification of AAR within existing dams is
given. This background information is deemed to be critical in AAR investigations
because it targets how, where and to what extent AAR is identified in the structure.
A detailed review of tests and investigations allowing the identification, extent

and prognosis of AAR to be evaluated for both new and existing dams is provided.
Three phases for the investigation of AAR are suggested. Phase 1, AAR Stage tests,
proposes “Look-up tables” for AAR Standard and Supplementary Tests that firstly
classify AAR tests in terms of scope, time, cost and quality of information and then
the essential factors that may influence test results and conclusions are given. This
provides the AAR investigative project team with a state-of-the-art overview of AAR
tests.
Phase 2 of AAR investigations look at the process by which AAR monitoring

and data analysis may be logically used. A strategic flow-chart approach to iden-
tification, extent and prognosis evaluations is given. This approach captures all of
the essential elements of AAR within the context of dam safety evaluations and pro-
poses that Phase 3 AAR follow-up tests are performed every 5 years in the event
that AAR has been initially detected in the structure. This proposal is deemed to be
essential to further validate the expansion in the structure and the prognosis tests
presented in chapter 6.
Finally, a new deterministic approach is proposed for the identification of AAR

within structures for which no visual signs of AAR, apart from expansion measured
by dam instrumentation, is apparent. This approach defines three expansion/crack
displacement and opening indexes that schematically illustrate AAR affected zones
of different magnitude and importance throughout the entire dam body. Knowledge
of these AAR zones with their respective extents allows test core samples to be
selectively extracted from the dam body. Moreover, the method defines fictitious
temperatures that allow correlation tests to be performed against laboratory resid-
ual expansion tests enabling the long-term effects of AAR and hence appropriate
rehabilitation works to be estimated.
DISCLAIMER
The work presented in this paper has been performed at SFOE, however the
statements and opinions herein are those of the author and do not necessarily reflect
a stance position of SFOE.
598
Q.106 - R.37
REFERENCES
[1] SWISSCOD (Swiss Committee on Dams), “Alkali-Aggregate Reaction (AAR)

in Swiss Dams”, Workshop, Switzerland, 13th May 2015.
[2] BEN HAHA M., “Mechanical Effects of Alkali Silica Reaction in concrete studied
by SEM-image analysis”, PhD Thesis N◦ 3516, École Polytechnique Fédérale
de Lausanne, May 2006.
[3] DUNANT C., “Experimental and Modelling Study of the Alkali-Silica–Reaction

in Concrete”, PhD Thesis N◦ 4510, École Polytechnique Fédérale de Lau-
sanne, November 2009.
[4] GIORLA A. B., “Modelling of Alkali-Silica Reaction under Multi-Axial Load”,

PhD Thesis N◦ 5982, École Polytechnique Fédérale de Lausanne, October
2013.
[5] CUBA RAMOS, A., I., “Multi-Scale Modeling of the Alkali-Silica Reaction in
Concrete”, PhD Thesis, École Polytechnique Fédérale de Lausanne, Spring
2017.
[6] LEEMANN A., GRIFFA M., “Diagnosis of alkali-aggregate reaction in dams”,

SFOE Publication N◦ 290934, Project N◦ 500863, Empa, Laboratory for
Concrete/Construction Chemistry, Überlandstrasse 129, CH-8600 Dübendorf.
(http://www.bfe.admin.ch/themen), Dec. 2013.
[7] MERZ, C., LEEMANN, A., “Assessment of the residual expansion potential of
concrete form structures damaged by AAR”, Cement and Concrete Research
52, pp. 182–189, 2013.
[8] SWISSCOD (Swiss Committee on Dams), “Concrete of Swiss Dams: Expe-

riences and Synthesis”, 2000, published in Wasser Energie Luft (WEL) 92.
Jahrgang, 2000, Heft 7/8, CH-5401, Baden, Switzerland.
[9] SWISSCOD, “Methods of analysis for the prediction and the verification of dam
behaviour”, April 2003, published in Wasser Energie Luft (WEL) Jahrgang,
2003, CH-5401, Baden, Switzerland.
[10] SWISSCOD, “Die Felsfundationen Schweizerischer Betonsperren, Übersicht

über die Fundationsverhältnisse”, Herbst 2006 (in German).
[11] GERBER, M., BÜHLMANN M. “DamBASE Version 1.0 May 2015”, Versuch-
sanstalt für Wasserbau Hydrologie und Glaziologie, VAW, 2015.
[12] SIA CAHIER TECHNIQUE 2042, “Prévention des désordres dus à la réaction
alcalis-granulats (RAG) dans les ouvrages en béton”, 2012, Correctif C1 2012
and Correctif C2 2015.
599
Q.106 - R.37
[13] SN 670 115, “Minéralogie et pétrographie qualitative et quantitative des

granulats”, 2015.
[14] SN 670 116, “Fillers – Minéralogie, pétrographie et minéraux argileux

gonflants”, 2015.
[15] SN EN 206-1, “Béton – Partie 1 : Spécification, performances, production et

conformité”, 2015.
[16] STUCKY A., DERRON M-H. “Problèmes thermiques posés par la construction
des barrages-réservoirs”, Science & Technique, Paul Feissly, libraire-éditeur,
Lausanne, 1957.
[17] ICOLD Bulletin 79, “Alkali-Aggregate Reaction in Concrete Dams”, 1991.
[18] AMBERG F., “Performance of Dams affected by Expanding Concrete”, Dams

and Reservoirs under Changing Challenges, edited by Robert M. Boes, CRC
Press 2011, Pages 115–122, Print ISBN: 978-0-415-68267-1, eBook ISBN:
978-0-203-80409-4
[19] GODART, B., de ROOIJ, M., WOOD, J. G. M., “Guide to Diagnosis and
Appraisal of AAR Damage to Concrete in Structures: Part 1 Diagnosis (AAR
6.1)”, April 2013.
[20] GUNN R. M., SCRIVENER K., A. LEEMANN A. "The identification, extent and
prognosis of alkali-aggregate reaction related to existing dams in Switzerland”,
Cfbr Chambéry, 13_15.06 2017.
[21] LCPC, “Détermination de l’indice de fissuration d’un parement de béton”,

Méthode d’essai LPC 47, 1997 (in French).
[22] SAOUMA V. E. “Numerical Modelling of AAR”, CRC Press, 2014.
[23] LARIVE C., “Apports combines de l’expérimentation et de la modélisation

à la compréhension de l’alcali-réaction et de ses effets mécaniques”, École
Nationale des Ponts et Chaussées Spécialité : Structures et Matériaux, June
1997 (in French).
[24] GRIMAL E., “Caractérisation des effets du gonflement provoqué par la réac-
tion alcali-silice sur le comportement mécanique d’une structure en béton”,
Université Paul Sabatier - Toulouse III, February 2007 (in French).
[25] GUNN, R. M., BOSSONEY, C. “Creep analysis of mass concrete dams”,

Hydropower and Dams Volume 3, Issue 6, 1996.
[26] GUNN, R. M., “Non-linear design and safety analysis of arch dams using
damage mechanics, Part 1: Formulation”, Hydropower and Dams, Volume 1,
Issue 1, 2001.
600
Q.106 - R.37
[27] GUNN, R. M., “Non-linear design and safety analysis of arch dams using
damage mechanics, Part 2: Applications”, Hydropower and Dams, Volume 1,
Issue 3, 2001.
[28] WEBER B., “Vorhersage des Verhaltens von Talsperren mit Hilfe des Soll-Ist-
Vergleichs – Statistischer Teil”, April 2002 (in German).
601
Q.106 - R.38
-------
-------
ADVANCED DETERMINISTIC MODELS TO ASSESS DAM DISPLACEMENTS

APPLICATION EXAMPLE AT 3 LARGE ARCH DAMS (∗)
Francesco AMBERG
Head of Special studies department,
LOMBARDI SA
SWITZERLAND
SUMMARY
The displacements of arch dams can vary significantly depending on water

level and temperature conditions, in particular in regions with relevant climate vari-
ations during the year. The regular assessment, if the current measured values
are correct and representative of a normal behavior is therefore not straightforward.
Suitable interpretative models are required to assist the engineer responsible of the
dam safety.
Interpretative models can be based on statistical analyses of the measures

available in the past or on structural analyses of the dam. The latter are commonly
referred as deterministic models and are by far more comprehensive than statistical
models. For dams is service since years, the establishment of deterministic mod-
els might represent an opportunity to verify certain aspects related with the dam
behavior and safety, e.g. orientation of pendulum readings, thermometers distribu-
tion within the dam body, basic elastic parameters of the dam and its foundation
and to assess if the dam behavior can be predicted by a structural model. Eventual
discrepancies need to be investigated and the actual reason identified.
The paper presents a reliable approach on how to establish accurate determin-

istic models, emphasizing in particular on the treatment of the demanding thermal
part, and shows the results obtained for three large arch dams in Switzerland:
Contra, Zeuzier and Emosson. The deterministic models of these three dams can
∗ Modèle déterministe pour analyser les déplacements des barrages – Exemple d’application
sur 3 grands barrages voûtes
602 DOI: 10.1201/9781003211808-133

Q.106 - R.38
predict the displacements so accurately, that it was possible to identify a visco-elastic

component in the behavior of the mentioned structures. The deterministic models
were then enhanced to include appropriately also this characteristic into the analysis.
The obtained interpretative models are very robust and can now be used for
long time. In fact, there are no reasons to update a reliable deterministic model, a
part if an actual modification in the response of the structure would occurs due for
example to a stiffness change. However, the identification of a change in stiffness
remains exactly one of the main objectives of dam monitoring.
RÉSUMÉ
Les déplacements d’un barrage voûte peuvent varier beaucoup en fonction

du niveau de la retenue et des conditions de température, en particulier dans les
régions avec des variation climatiques importantes durant l’année. La vérification
régulière, après chaque mesure, si les valeurs sont correctes et représentatives d’un
comportement normal, n’est donc pas toujours immédiate. Des modèles interpré-
tatifs sont donc disponibles pour assister les ingénieurs responsables de la sécurité
des barrages.
Les modèles interprétatifs peuvent se baser sur des analyses statistiques des
mesures disponibles dans le passé ou sur des analyses structurales du barrage.
Dans ces derniers cas on parle de modèles déterministes, qui sont plus complets
que les modèles statistiques. Pour les barrages en service depuis plusieurs années,
l’établissement de modèles déterministes peut représenter une bonne opportunité
pour vérifier certains aspects liés au comportement et à la sécurité du barrage,
comme l’orientation des mesures des pendules, la distribution des thermomètres
à l’intérieur du barrage, les paramètres élastiques du barrage et de ses fondations
et la confirmation si le comportement du barrage peut être reproduit par une anal-
yse structurale. Eventuelles différences doivent être étudiées afin d’en trouver les
causes.
Le rapport présente un possible approche pour établir des modèles détermin-

istes fiables, tout en mettant l’accent sur le traitement de la partie thermique, et
montre les résultats obtenus pour trois grands barrages voûte en Suisse : Contra,
Zeuzier et Emosson. Les modèles déterministes de ces trois barrages sont capables
de reproduire les déplacements si précisément, qu’il était possible d’identifier une
composante visco-élastique dans le comportement de ces trois structures. Les mod-
èles déterministes ont donc été améliorés pour aussi inclure de façon appropriée
cette particularité dans les modèles.
Les modèles interprétatifs finaux sont fiables et peuvent être utilisés pour
longtemps. En effet il n’y a pas de raison pour devoir mettre à jours un tel modèle
déterministe, sauf s’il y a eu une modification réelle du comportement du barrage,
par exemple à la suite d’une modification de sa rigidité. Dans tous les cas, c’est
bien l’identification de tels phénomènes l’un des objectifs de ces modèles et de
l’auscultation des barrages.
603
Q.106 - R.38
1. INTRODUCTION
Displacements of arch dams are subject to significant variations depending

on water level and temperature conditions. This is especially true in regions with
relevant seasonal fluctuations. Therefore, the regular assessment of whether the
measurements are correct and representative for a normal behaviour is not straight-
forward. Here, interpretative models present a very useful tool to analyse the dam
monitoring data as function of the operating conditions. Being integrated in a mon-
itoring software, interpretative models allow to detect immediately any abnormal
phenomenon in the behaviour of the foundation or the dam body itself.
Interpretative models can be based on statistical analyses of the measures

available or on structural analyses of the dam. The latter are commonly referred to
as deterministic models and are more comprehensive than statistical models, since
statistical models are not based on physically proven cause-effect relationships. On
the other hand, the quality of deterministic models, and thus the prediction reliability,
strongly depends on the correct interpretation of the whole physical process. It is
therefore crucial to adequately define the structural model, the representative mate-
rial parameters and the adequate loads. In particular, the definition of the proper
thermal loads on the structure from local temperature readings is relevant for the
accuracy of the prediction.
In the following, a reliable approach on how to establish accurate deterministic

models is presented, focussing on the pre-treatment of the thermal part and on the
adjustment procedure using the actual dam behaviour. Corresponding implemen-
tations are illustrated by showcasing three large arch dams in Switzerland: Contra,
Zeuzier and Emosson.
2. DETERMINISTIC MODEL
2.1. DEFINITIONS AND REFERENCE DISPLACEMENT
Interpretative models aim to describe the relationship between the operating

conditions and the behaviour of the structure. This relationship is mainly based
on the superposition principle, applicable for linear elastic structures, that allows a
simple summing up of the effects of the acting unit loads. Appropriate coefficients of
influence permit to quantify the effects of the different loads on the displacements.
The effect of hydrostatic pressure is usually [1] described by a polynomial function,
while the temperature effect is defined by a linear combination of the contribution of
various unit thermal loads.
The implementation of interpretative models allows to subtract the reversible

contributions of water level and temperatures from the measured displacements.
The obtained result, identified as reference deformation, corresponds to a theoretical
604
Q.106 - R.38
state of the dam at fixed operational conditions. It is pointed out that for deterministic
models it is possible to choose the corresponding reference load condition and thus
to define the constants of the various influence factors according to this choice.
δR (i) = δM (i) − δC (i) (1)
δR (i): reference displacement
δM (i): measured displacement
δC (i): total calculated displacement (reversible part)
The reference displacements should not be confused with the residual

between the measured and the calculated displacements (Equation (3)), even if
there are various common points between both. The main difference is that the
reference displacements allow to highlight the presence of eventual irreversible phe-
nomena of drifts, while the residual does actually not include any drift since the
irreversible part is generally already included in the displacement calculated by the
interpretative model.
2.2. CALIBRATION PROCEDURE
For existing dams it is useful, and recommended, to proceed with a calibra-

tion of the deterministic model on the base of the effectively observed behaviour.
This is an essential part of the safety assessment, since the actual versus theoret-
ical comparison allows to verify the validity of the model and of all its associated
assumptions. The calibration should however be performed according to certain cri-
teria, in order to maintain the physical background of the cause-effect relationship
established with the structural analysis. In practice the interference of the statistical
approach with the deterministic model should be limited to a minimum. The calibra-
tion should therefore concentrate on the main parameters of the structural analysis,
i.e. coefficient of thermal expansion and elastic modulus of concrete and founda-
tion rock mass. This principle can be expressed by the following equation giving the
reversible deformation calculated by a deterministic model.
δC (i) = CW,t · δW (i) + CT ,i · δT (i) (2)
CW,i : stiffness calibration factor
δW (i): calculated displacement due to water load
605
Q.106 - R.38
CT,i : calibration factor for the coefficient of thermal expansion
δT (i): calculated displacement due to thermal loads
The stiffness calibration factor applies only to the water load effect, since
the stiffness does not have any influence on displacements due to temperature
variations. The calibration factor CT ,i , on the other hand, applies only to thermal
displacements, since the effect of water pressure does not depend on the thermal
expansion coefficient.
The model optimization is done by an iterative process that modifies the

two factors exposed in Equation (2) in order to best reproduce the observed dam
behaviour. The effectiveness of this optimisation is evaluated based on the reduction
and minimization of the standard deviation of the residuals δ.
δ(i) = [δM (i) − δN (i)] − δN (i) (3)
δ(i): residual
δM (i): measured displacement
δN (i): non-reversible, time dependent displacement component
As it appears from Equation (3), the measured displacement is first corrected

by the non-reversible component, if relevant, before starting with the actual calibra-
tion. This non-reversible time dependent component can be defined following various
approaches, depending on the period considered during the service life and on the
expected phenomena causing the trend. A sound physical background based on
engineering judgement should not be neglected in the analysis of the non-reversible
component. For dams in operation for many years, it is for instance quite easy to
identify and describe possible trends exhibited by the actual behaviour over a long
period, with these tendencies being linear, accelerating or decreasing. The advan-
tage of performing this a priori correction of the non-reversible deformation from the
measured displacements is that the subsequent iterative calibration procedure can
effectively focus on the optimization of the reversible behaviour without the influence
of disturbing components.
After having obtained measured displacements characterized only by

reversible components, the optimization is done in several phases as follows:
• Phase 0: calculated displacements directly refer to the static calculation

(calibration factors CW,i and CT,i remain at 1) ;
• Phase 1: global elastic optimization – optimisation by considering single
factors CW and CT in order to minimise the sum of the standard deviations
of the residuals of all measuring points i;
606
Q.106 - R.38
• Phase 2: local elastic optimization – optimization of the calibration factors

individually and independently for each point.
The optimization is performed initially at the global level (phase 1), in which
case the final value of the two factors, after optimization, is valid for the entire model.
The optimization can also be performed locally (phase 2), with values of CW,i and
CT,i being searched independently for each reading point. This individual calibration
principally aims to verify the presence of trends in the stiffness calibration factor
CW,i indicating the necessity to adapt, in the structural model, the elastic modulus
of the rock mass, or of a certain foundation part, differently from the elastic modulus
of the concrete. The scope of phase 2 is therefore not to obtain the best actual-
theoretical comparison in all points, that probably can be easily obtained with a
statistical approach. Individual calibration factors should be adopted in the final
proposal of the calibrated determi

ICOLD - Twenty-Seventh International Congress On Large Dams - Vingt-Septième Congrès International Des Grands Barrages-CRC Press - Balkema (2022)

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

ICOLD - Twenty-Seventh International Congress On Large Dams - Vingt-Septième Congrès International Des Grands Barrages-CRC Press - Balkema (2022)

Transféré par

Droits d'auteur :

Formats disponibles

TWENTY-SEVENTH

27th May–3rd June 2022

Typeset by MPS Limited, Chennai, India

Published by: CRC Press/Balkema

ISBN: 978-1-032-07848-9 (Hbk)

TRANSACTIONS COMPTES RENDUS

GENERAL TABLE OF CONTENTS TABLE GÉNÉRALE DES MATIÈRES

GENERAL SYNTHESIS / SYNTHÈSE GÉNÉRALE

Questions discussed at preceding Questions discutées aux précédents

Wording of Question 104 Libellé de la Question 104

Wording of Question 105 Libellé de la Question 105

Wording of Question 106 Libellé de la Question 106

Wording of Question 107 Libellé de la Question 107

Table of Contents of Communications Table des Matières des

INNOVATION AWARDS / PRIX DE L’INNOVATION

TABLE DES MATIÈRES

First Congress. Stockholm (Sweden), Premier Congrès. Stockholm (Suède),

The 87th Annual Meeting of the La 87ème Réunion Annuelle de la

CONCRETE DAM DESIGN INNOVATION DANS LA

1. Innovations for arch dam analysis, 1. Innovations pour l’analyse, la

INCIDENTS AND ACCIDENTS INCIDENTS ET ACCIDENTS

DAMS AND CLIMATE CHANGE BARRAGES ET CHANGEMENT

1) Number of Papers and Communications by alphabetical order of countries/

Countries Nbr of Papers Nbr of Comm. Total of Papers & Comm.

Countries Q. 104 Q. 105 Q. 106 Q. 107 C. Total

b. Innovations for design, construction materials and placement methods,

c. Innovations for raising existing concrete dams.

d. Innovations for extremely high concrete dams.

e. Operational performance of concrete dams during the life cycle, including

INNOVATIONS DANS LA CONCEPTION ET LA PERFORMANCE DES

b. Innovations relatives à la conception, au choix des matériaux, aux méthodes

c. Innovations pour la surélévation des barrages en béton existants.

d. Innovations pour les barrages en béton de hauteur extrême.

e. Performance opérationnelle des barrages en béton pendant l’ensemble du

R. 1 ALI NOORZAD, ALIREZA DANESHYAR, PAYAM SOTOUDEH,

General Report / Rapport Général Q. 104

NONLINEAR RESPONSE OF FOUNDATION ROCK IN SEISMIC SIMULATION

Seismic simulation of concrete gravity dams considering nonlinear response

La simulation sismique des barrages-poids en béton, tenant compte de la

Earthquake analysis of concrete gravity dams is mostly performed under

Nonlinear behavior of concrete has drawn ample attention from numer-

In this paper, the novel damage-plastic method of Daneshyar and Ghaemian

2. CONTINUUM DAMAGE FORMULATION

Anisotropic damage-plastic model of Daneshyar and Ghaemian [19] is used

where Iijkl is the forth-order identity tensor as follows

and δij is the Kronecker delta, such that

thus, an isotropic description of damage is constructed. Degradation along different

Concrete behavior is susceptible to rate of loading, i.e. the fracture energy

and λ̄ and µ̄ are the Lame constants of virgin material.

[2] PEKAU OA, CHUHAN Z, LINGMIN F. Seismic fracture analysis of concrete

[4] EL-AIDI B, HALL JF. Non-linear earthquake response of concrete gravity

[5] BHATTACHARJEE SS, LEGER P. Seismic cracking and energy dissipation in

[6] GUANGLUN W, PEKAU OA, CHUHAN Z, SHAOMIN W. Seismic fracture

[7] HILLERBORG A, MODÉER M, PETERSSON PE. Analysis of crack forma-

[9] MIRZABOZORG H, GHAEMIAN M. Non-linear behavior of mass concrete in

[10] CERVERA M, OLIVER J, MANZOLI O. A rate-dependent isotropic damage

[12] BHATTACHARJEE SS, LÉGER P. Application of NLFM models to predict

[13] WEPF DH, FELTRIN G, BACHMANN H. Influence of time-domain dam–

[14] SKRIKERUD PE, BACHMANN H. Discrete crack modelling for dynamically

[16] CALAYIR Y, KARATON M. A continuum damage concrete model for earth-

[17] PAN J, ZHANG C, XU Y, JIN F. A comparative study of the different procedures

[18] ZHANG S, WANG G, YU X. Seismic cracking analysis of concrete gravity

[19] DANESHYAR A, GHAEMIAN M. Coupling microplane-based damage and