Vous êtes sur la page 1sur 16

Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

www.elsevier.com/locate/soildyn

Modelling liquefaction-induced building damage in earthquake


loss estimation
Juliet F. Bird a,*, Julian J. Bommer b, Helen Crowley c, Rui Pinho c
a
Arup Geotechnics, The Arup Campus, Blythe Gate, Blythe Valley Park, Solihull, West Midlands, B90 8AE, UK
b
Department of Civil and Environmental Engineering, Imperial College London, SW7 2AZ, UK
c
European School of Advanced Studies in Reduction of Seismic Risk, c/o EUCENTRE, Via Ferrata 1, 27100 Pavia, Italy
Accepted 4 October 2005

Abstract
The assessment of building damage caused by liquefaction-induced ground deformations requires the definition of building capacity and
vulnerability as a function of the demand, as well as damage scales to describe the state of the damaged building. This paper presents a framework
for resolving these issues within the context of earthquake loss estimations, where large variations in building stock and ground conditions must be
considered. The principal modes of building response to both uniform and differential ground movements are discussed and the uncertainties in
their evaluation are highlighted. A unified damage scale is proposed for use in both reconnaissance and assessment of all modes of building
damage, including ‘rigid body’ response of structures on stiff foundations to uniform or differential ground movements. The interaction of ground
shaking and liquefaction in the context of induced structural damage is also briefly considered. The paper raises important aspects of earthquake
loss estimations in regions of liquefaction potential, which remain relatively poorly defined at present.
q 2005 Elsevier Ltd. All rights reserved.

Keywords: Liquefaction; Ground deformation; Building damage; Earthquake loss estimations

1. Introduction buildings and facilities, due to future earthquakes. The main


components of an earthquake loss model are shown in Fig. 1.
The potential for earthquake-induced liquefaction to cause The scale of an earthquake loss estimation may range from an
significant damage to buildings in earthquakes is well known. entire country to cities or even districts within cities. The larger
Destructive earthquakes in Kobe in 1995 and in Taiwan and the study area, the more likely it is that ground shaking-induced
Turkey in 1999 are prime examples of this, with numerous damage to buildings will dominate the overall losses, and that
reports of buildings settling, tilting, sliding, or otherwise any secondary hazards affecting a particular zone will be of less
suffering due to the failure of the underlying soils. The impact significance. For example, in the 1999 Kocaeli earthquake in
of liquefaction has been greatest on lifelines such as roads, Turkey, damaging hazards included surface fault rupture,
mainly due to the failure of bridges, and buried pipelines, as coastal subsidence, slope failures and liquefaction, but the
well as to port structures, the clearest case in point for the latter dominant cause of damage over the entire region was
being the damage caused by the Kobe earthquake. Although undoubtedly ground shaking [1]. An estimation of building
there are cases where spectacular damage to buildings has been damage for the entire Kocaeli region could, therefore,
caused by liquefaction, such as in Adapazari due to the 1999 reasonably concentrate on ground-shaking, but for more
Kocaeli earthquake in Turkey, liquefaction-induced damage to detailed studies within the region there is an increased
buildings in major earthquakes typically contributes to earth- likelihood that local features such as liquefiable soils will
quake losses only over relatively small areas [1]. influence the results, as was seen in the city of Adapazari. The
Earthquake loss estimation is a technique used to quantify framework for estimating liquefaction damage presented in this
potential losses in a given region or to a particular portfolio of
paper is primarily intended to be applied on a local scale.
Comprehensive earthquake loss estimations require inter-
* Corresponding author. Tel.: C44-121-213-3000. action between Earth scientists, engineers, public and private
E-mail address: Juliet.bird@arup.com (J.F. Bird). owners of facilities, lifeline operators, planners and financiers
0267-7261/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. and as such are truly multi-disciplinary. Two significant
doi:10.1016/j.soildyn.2005.10.002 features from an engineering perspective are the regional
16 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

than calculating their magnitude. Even in the field of


HAZARDS
Ground shaking performance-based seismic design, there is little guidance on
Ground Failure what are acceptable damage levels due to foundation
Fault Rupture
movements.
INVENTORY
The prediction of liquefaction demand comprises three
Buildings and Lifelines
stages. The first is to determine whether the soils are
DIRECT susceptible to liquefaction using mainly qualitative criteria
PHYSICAL
DAMAGE such as those presented by Youd and Perkins [4]. Given a
positive response to this, the likelihood of liquefaction being
triggered by the scenario earthquake is required (Section 2.1).
INDUCED PHYSICAL DAMAGE
Fire following earthquake
Finally, the demand, in terms of the expected ground
Release of hazardous materials deformations, must be defined (Section 2.2).
Flood following earthquake A preference has been noted in previous papers [5,6] for the
Debris
use of scenario based loss estimations over those based upon a
probabilistic seismic hazard assessment (PSHA). This pre-
ference, despite the additional computational expense it entails,
is reiterated herein, since to estimate both the liquefaction
LOSS potential and the resulting ground deformations, the earthquake
Social losses – casualties and shelter magnitude, its duration, and the source-to-site distance are
Direct economic impact
Indirect economic impact
fundamentally important variables, and therefore, a scenario-
based approach, which explicitly defines the magnitude and
Fig. 1. Components of an earthquake loss estimation (after FEMA [2]).
distances, is essential.

nature of loss estimations and the fact that the models deal with
2.1. Triggering of liquefaction
existing buildings and infrastructure, about which very little
may be known. Both these features present a considerable
Current standard practice for the assessment of liquefaction
contrast to site-specific design or even site-specific assessment.
potential is essentially deterministic, using one of a number of
Techniques used for site-specific design cannot be employed in
published relationships between the in situ density of the soil
this context, because the demands of time and data acquisition
and the cyclic shear stress induced by ground shaking (e.g. [7]).
would be excessive.
Application of deterministic methodologies goes no further
The objective of this paper is to obtain an assessment of the
than ascertaining that the factor of safety (FS) against
distribution amongst damage levels for a particular building
liquefaction is greater than or less than unity. For risk
stock as a consequence of an earthquake-induced liquefaction.
assessments, the probability associated with the result FS!1
These results can be simply translated into a direct economic
is also needed. Even where there is an apparently significant
loss once the total financial value of the building stock is
safety margin in the calculation, there may still be an
known. However, as discussed in detail in Section 3.2, the
associated, albeit low, probability of failure, which comes
definition of damage necessarily differs from that traditionally
from three principal areas of uncertainty: the ground motion
used in loss estimation models since a building can be rendered
level, the empirical procedures used to evaluate liquefaction
uninhabitable by the effects of liquefaction without suffering
resistance, and the natural variability and heterogeneity of the
any structural damage to the load-bearing system.
soil properties and stratigraphy. Probabilistic approaches
towards liquefaction assessment have been developed by,
2. Estimating liquefaction demand
amongst others, Cetin et al. [8] and Rodriguéz-Marek et al. [9],
although these are not so widely used as the deterministic
Methodologies for the assessment of liquefaction potential
methods.
and the resulting ground deformation have been the focus of
research for many years. Despite this, when it comes to
determining the impact these deformations will have on 2.2. Quantification of permanent ground deformations
existing structures, the published literature is generally
inadequate [3]. There are many reasons for this, not least that The evaluation of liquefaction-induced displacements
the preferred design approach is to mitigate the liquefaction beneath a building is far from straightforward. A number of
potential by ground improvement or site re-location. Generally, methodologies are available, all of which use different
a structural engineer will determine the deformation tolerances approaches and uncertain variables; none of these methods
for a structure, and the geotechnical engineer will ensure have overwhelming support in their favour compared to the
through design that these criteria are satisfied. Similarly, in the others. There are multiple complex mechanisms which can
estimation of expected ground deformations, most soil cause damaging ground deformations. Table 1 summarises
mechanics theory is orientated towards the avoidance of available methodologies for the assessment of expected
large-scale deformations associated with ground failure, rather liquefaction-induced ground deformations.
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 17

Table 1
Selected methodologies for determining liquefaction-induced permanent ground deformations

Reference Description Variable input Comments


parametersa
I. Lateral ground deformation
Zhang and Zhou Empirical prediction of liquefaction-induced lateral
[10] spread
Youd et al. [11] Equations obtained through multi-linear regression of M, R, T15, (N1)60, Shown to be accurate within a factor of 2
empirical data FC, D50,
Bardet et al. [12] Multi-linear regression of empirical data M, R, T15, slope (S) Same dataset as Youd et al. [11] but in a probabilistic
and free face ratio framework
(W)
Youd and Perkins Equations for Liquefaction Severity Index (LSI), defined M, R Maximum value. Single topographic/geologic environ-
[13] as maximum lateral deformation in inches, with 0! ment only
LSI!100
Rauch and Martin Empirical prediction of liquefaction lateral spread (i). M, R, PGA, Average displacements 3 equations, for different levels of
[14] duration input data; site, regional and geotechnical
(ii). M, R, PGA,
duration, Lslide,
Stop, Hface (if free
face present)
(iii). M, R, PGA,
duration, Lslide,
Stop, Hface, ZFSmin,
Zliq
II. Settlement
Tokimatsu and Charts relating in situ density measurements to earth- (N1)60, CSR, FC Shown to yield correct results within a factor of 2-3 [18]
Seed [15] quake induced shear strain. Laboratory and empirical
data
Ishihara and Charts relating FL to in situ density and volumetric strain FL, N1, FC
Yoshimine [16]
III. Combined lateral and vertical movement
Shamoto et al. [17] Charts for volumetric and shear strains T/s 0 , Na, FC Uses soil mechanics theory as well as empirical data.
a
T15, thickness of layer with (N1)60!15; D50, median particle size diameter; CSR, cyclic shear stress ratio; Hface, Height of free face; FC, fines content; FL, factor
of safety against liquefaction; Lslide, maximum horizontal length of lateral spread; M, earthquake magnitude; N1, SPT N value corrected for overburden; Na, adjusted
SPT N value (Shamoto et al., 1998); (N1)60, SPT N value adjusted for energy and overburden; PGA, peak ground acceleration; R, source-to-site distance; Stop,
average slope across the surface of the lateral spread; T15, thickness of soil with SPT N!15; Vs, shear wave velocity of soil; ZFSmin, depth to minimum factor of
safety against liquefaction; Zliq, depth to top of liquefied layer

Most commonly used assessment methods rely heavily on generally accepted to be accurate only to within a factor of 2 or
empirical data. One reason for this is the difficulty in obtaining 3 [18,11] and their predictive capacity tends to be worst for
measured in situ soil properties for input to soil mechanics small-to-moderate (0.3–0.75 m) deformations [21].
theory, and the variability associated with such properties [18]. Selection of the calculation methodology is best made on a
Another reason is that geotechnical analysis techniques do not case-by-case basis; with categorical recommendations there is
generally allow for large strain deformations of soils [19]. Even a danger that regional features, which may affect their validity
where full constitutive soil models in a finite element in some way, are overlooked. For example, in the HAZUS
framework are used, many of these cannot go beyond the methodology [22], the simplified approach for estimating
initiation of liquefaction, at which point the procedure will lateral spread combines the Liquefaction Severity Index (LSI)
typically break down and fail to converge (e.g. [20]). relationships derived by Youd and Perkins [13] with the
Furthermore, numerical models have been found to be attenuation relationships of Sadigh et al. [23] to derive a
particularly sensitive to small variations in input parameters relationship for LSI in terms of peak ground acceleration
[21], an undesirable feature in such an uncertain field as loss (PGA) only. Notwithstanding the simplifications involved, or
estimation. Youd et al. [11] state that ‘no physical theory’ the fact that significant scatter in both empirical relationships is
exists to confirm their empirically-based relationships. ignored, neither of these relationships are applicable outside of
For the purpose of loss estimations, empirical processes, the western USA, and the former is only valid for the specific
based upon easily obtainable parameters, or at least parameters environment of ‘wide active flood plains or other areas of
that can be estimated using a reasonable level of judgement, are gently-sloping late Holocene fluvial deposits’.
preferable to complex numerical approaches [3]. The For the prediction of lateral movements, the EPOLLS
uncertainty associated with either approach should always be (Empirical Prediction of Liquefaction Lateral Spreading)
considered. The empirical methods shown in Table 1 are methodology [14], has some advantageous features, in that it
18 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

provides three different levels of empirical relationship, Furthermore, it should be noted that these uncertainties are
depending on the amount of data available. An additional over and above the significant epistemic uncertainties
advantage is that the formulae estimate the average rather than associated with the primary part of a loss estimation; building
the maximum displacement, which is more appropriate to the damage due to ground shaking, all of which have been shown
requirements of a loss estimation study. The authors also to have a significant impact upon the estimated distributions of
present a study of the statistical variation of horizontal damage [e.g. 27,28].
movements within a lateral spread [24].
For estimation of vertical movements, both the methods of 2.4. Differential movements
Tokimatsu and Seed [15] and Ishihara and Yoshimine [16]
have been found to be accurate to within a factor of 2 or 3. Both Of particular concern in terms of assessing building damage
methodologies deal with settlements caused by the change in are the expected differential settlements or differential lateral
volumetric strain as pore water pressures dissipate after movements. Ishihara and Yoshimine [16] state that ‘because of
liquefaction has taken place. This is only one potential cause nonhomogeneous conditions in the soil deposits, the settle-
of vertical foundation movements, others include loss of ments seldom occur uniformly even in small localised areas
ground due to sand boiling, the vertical component of lateral and differential settlements become the major cause of the
deformation of a volume of soil, or ‘punching’ failure of damage to lifelines or other facilities’. An example of large
foundations due to reduced bearing capacity [21]. All of these differential settlements is those that occur beneath buildings
hazards are potentially damaging to foundations; unfortunately located on the boundary between liquefied and non-liquefied
there are no reliable procedures for their estimation, which is soils. Bardet and Kapuskar [29] observed that the worst
currently a major shortcoming in this field. damage in the Marina district after the 1989 Loma Prieta
earthquake was on such a boundary. Differential ground
settlements will occur due to heterogeneity in soil stiffness and
2.3. Uncertainties associated with liquefaction demand stratigraphy both laterally and with depth.
Estimating differential ground movements on a regional
The uncertainty associated with estimating the expected scale has an even greater uncertainty than the estimation of
permanent ground deformation arises from the following areas: uniform or average movements. This is principally because of
the lack of sufficient geotechnical data; a borehole at each
† The level of earthquake hazard in terms of the uncertainty
corner of a building would allow a reasonable estimation of the
associated with the ground-motion estimation.
variability in the settlements, even though retaining the other
† The likelihood of liquefaction triggering, based upon
uncertainties listed in the previous sub-section. In the absence
empirical data with associated scatter (e.g. Youd et al.
of this degree of detailed geotechnical data, for convenience,
[7]).
the variability may be modelled as random (i.e. aleatory),
† The choice of appropriate methodologies and input
assuming a statistical distribution such as normal or lognormal,
parameters for the evaluation of liquefaction probability,
with a mean and variance. Thus the distribution of differential
liquefaction-induced permanent ground deformation, and
settlements over the footprint of a building, in terms of a
building response. As well as the epistemic uncertainty in
percentage of the absolute settlements, can be obtained.
selecting a methodology, there will be uncertainties due to
Differential lateral movements will also be caused by the
scatter, and, due to the limitations of the selected approach. variability of in situ ground conditions, but these will also have
† The difficulty in accurately defining the thickness and a less random component related to geometric variability in
extent of different stratigraphic layers, due to natural terms of increasing distance from a free face or from the toe of
heterogeneity of soils (e.g. Power et al. [25]) and a slope. Rauch and Martin [24] found the variability of lateral
insufficiency of available geotechnical data. displacements across a slope to be reasonably represented by a
† Measurement or equipment biases of in situ data. Standard gamma distribution.
penetration test (SPT) measurements have been cited as The flow chart in Fig. 2 illustrates the input data and
having coefficients of variation from 25 to 50% [20,26]. decisions required to identify modes of liquefaction-induced
† The necessary simplification of soil properties and ground deformation beneath buildings in the context of a loss
stratigraphy, since even in the unlikely event that estimation. The relative likelihood of each mode must be
comprehensive geotechnical data were available, it would evaluated and the associated magnitude of ground deformation
not be feasible on a large scale to incorporate this without quantified. The complexities and uncertainties in this process
some approximation. Ground conditions and building types are significant, and Fig. 2 summarises these without dictating
that are expected to respond in an approximately similar the choice of methodology and the level of input data, which
manner to liquefaction must be grouped into a manageable must be decided on an individual basis.
number of classifications. In reality each soil profile and
building will respond differently, and understanding and 3. Building response to liquefaction
thereby quantifying the effects of this grouping is one of the
most significant requirements of regional damage Structural damage to buildings during ground shaking is
estimations. mainly due to inter-storey drift, caused by inertial forces
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 19

Historical and Is soil susceptible No


Stop
geological data to liquefaction?

Yes
Earthquake
characteristics and
geotechnical data Regional P(L) < X *
probability of Stop
liquefaction
Choice of calculation
methodology >X*
P(L) –

Level ground? Geotechnical and Sloping face or free


topographical data face?
Yes Yes

Mean and variance Mean and variance


of expected Choice of calculation of expected lateral
settlement methodology spreading
(regional) (regional)

Flexible
foundations
only

P(uniform Mean and variance Mean and variance


settlement) of expected of expected
differential differential lateral
settlement (local) movement (local)

Fig. 2. Information required for a complete assessment of building vulnerability to liquefaction-induced ground deformations. *X is a predefined value below which
the probability of liquefaction is considered negligible.

induced by the excitation of the base and by the passage of building response to earthquake-induced ground failure, the
waves ascending the building. The inter-storey drift induces foundation system will also be of importance. For a portfolio of
distress in the structural elements, leading to loss of buildings, knowledge of the foundations will be more uncertain
functionality and then to load bearing capacity. Structural than that of the structural system, which in itself requires some
response to vibratory motion lends itself to analytical significant assumptions, short of carrying out a building-by-
modelling of the elastic and inelastic deformations, permitting building survey. Even from visual surveys, foundation types
the development of mechanically-based methods for evaluat- cannot be easily ascertained, and a greater depth of
ing building vulnerability to strong ground-motion. One investigation is required. There are, however, some assump-
example of this is the capacity spectrum method embodied in tions that can reasonably be made: for example that poorly
the HAZUS methodology [22,30]. An alternative approach has designed buildings are likely to have poorly designed
been developed by Crowley et al. [31]. The evaluation of foundations, and well-designed modern buildings are more
building vulnerability to liquefaction-induced ground defor- likely to have considered liquefaction susceptibility in the
mations is less straightforward, because of the variety of ways foundation design and to have stiff or deep foundations, or some
in which buildings can respond, which creates the need to use form of ground improvement. Such assumptions necessarily
more than one approach. The different modes of response of add an additional level of uncertainty to an inventory, which
buildings to ground failure also creates the need for broader must be considered when interpreting the results.
definitions of damage, since these cannot be based only on Given that this methodology is likely to be applicable to
deformations in structural members. relatively small study areas, where it is known that there is a
The first step in developing any method for evaluating prevalence of liquefiable soils, it is a reasonable premise that
building vulnerability is a classification system for buildings, enough information on the local building stock will be
based on the concept of buildings in any category having available to make appropriate judgments regarding the
essentially the same mode of response and failure, and distribution of foundation types. For example, in Adapazari,
comparable levels of susceptibility to the earthquake hazard investigations after 1999 revealed that a very high percentage
under consideration. Crowley et al. [31], in common with most of the beskat1 buildings were founded on reinforced concrete
loss estimation methodologies, define building vulnerability in
terms of the structural system and the member properties 1
Literally ‘five-storied’, but used to describe 4–6 storey, reinforced concrete
(dimensions and deformation characteristics). For assessing frame apartment blocks.
20 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

(a) (b)

∆v ∆v

Fig. 3. Schematic illustration of frame buildings with stiff shallow foundations subjected to liquefaction-induced ground deformations: (a) Uniform vertical
settlements, and (b) differential vertical settlements leading to rigid-body rotation of structure and foundation.

mat foundations of 1–1.5 m thickness [32]. For a study area the because the damage more closely resembles damage caused by
size of Adapazari it is feasible to gather such information ground shaking, comprising cracking and yielding of structural
through dialogue with local engineers and builders, whereas for members, it is not always immediately identified as being due
a larger region, such as the model compiled for the whole of to ground failure. However, this explanation seems improbable
Turkey [5], such depth of investigation is clearly unrealistic. considering the building shown in Fig. 5 for example, where
the ground deformation is obvious.
3.1. Rigid-body and differential movements Fig. 6 summarises the modes of response to ground
deformations for any given structural and foundation classifi-
There are a number of deformation modes that buildings cation. Where ground deformation causes ‘rigid body’ building
may experience when subject to liquefaction-induced ground response, the important consideration for determining the
deformations [33]. These modes can be divided into two broad damage state is the acceptability of the system performance.
categories: rigid-body movements, whereby the structure The damage state cannot be easily classified in terms of the
moves without significant internal deformation, and differential deformations of structural members, therefore there is no
movements. The type of response will depend primarily on the apparent analytical solution in this case, and an empirical
foundation type: for shallow foundations, the distinction will solution is required. Where there is sufficient flexibility in the
be whether these are rigid (Fig. 3) or flexible (Fig. 4). foundations for the walls or columns to move independently
A review of field reports from recent earthquakes [1] and thus differentially (as depicted in Fig. 4), the response is
revealed a much larger number of occurrences of rigid body referred to as ‘structural damage’, because the damage occurs
damage rather than structural damage to buildings on flexible in the structural elements and is related to the deformation
(i.e. unrestrained) foundations. Hundreds of buildings were induced in them. Analytical solutions are possible in this case;
reported to have settled or tilted in the earthquakes in Turkey a procedure for estimating expected damage levels in the
and Taiwan in 1999 and the Philippines in 1990, amongst many framework of an earthquake loss estimation is presented by
others. A notable example of structural damage, caused by Bird et al. [33]. These solutions are compatible with the
differential foundation movement, is the failure of the low-rise displacement-based earthquake loss assessment methodology
wooden structures of the Marine Laboratory at Moss Landing presented by Crowley et al. [31].
following the Loma Prieta earthquake in 1989 [34]. Other cases
have been reported in the 1999 Kocaeli and Chi–Chi 3.2. Building damage states and losses
earthquakes (e.g. Fig. 5) but overall there are relatively few
such reports. Notwithstanding the scarcity of case histories, this Where building response to ground failure comprises
failure mode clearly represents a genuine hazard to buildings. structural damage, as described above, damage states can be
Marino [35] notes that the most common earthquake-induced classified using the same schemes used for structural damage
foundation damage is separation of foundation elements, caused by ground shaking, e.g. Table 2. Such schemes are an
resulting from the irregularity of foundation settlement due to important component of loss estimations, in allowing analysis
lateral spreading, inducing tension, bending or tilting in and presentation of results, comparison with field observations,
shallow foundations. and providing users with an understanding of the physical
The smaller number of reports of structural damage due to condition of different classes of buildings, and consequences in
differential foundation movements may be because in many terms of repair and occupancy.
cases even fairly flexible foundations have sufficient relative The absence of rigid body rotation failures related to ground
stiffness, compared to soft underlying soils, to cause structures deformations, such as tilting, settlement or sliding, from most
to behave as rigid bodies. Alternatively, it is possible that, existing damage scales has been discussed previously [e.g. 1,
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 21

(a)

Rotational failure
of columns

∆v

(b)

Hinges form in
beams

∆v

Fig. 4. Schematic illustration of frame buildings with flexible shallow foundations to liquefaction-induced ground deformations: (a) Differential vertical settlements
due to soil variability, and (b) differential settlement at the centre of a multi-bay frame.

36]. Even where there is no, or minor, damage to the structural damage state of a rotated or settled building is best described
elements of such buildings, they must be considered to be using empirical solutions, classifying the damage level in terms
damaged, where damage may be defined as ‘a change in the of the functionality and reparability of the building. Damage
condition of the structure that adversely affects its future can be considered according to three criteria: aesthetic, where
structural performance’ [37]. In extreme cases, such as the the movements are perceptible to the owners or inhabitants but
buildings shown in Fig. 7, the damage state is complete, since do not impact the functionality; serviceability, or in the worst
the buildings are only suitable for demolition. In this context, case, stability.
damage can only be defined in terms of the consequences for Damage to a building that has settled uniformly may
the building use and the safety of the contents and inhabitants. include: damage to the ground floor slab (Fig. 8), sand ejecta
However, the actual cost of the kind of failure illustrated in filling the ground floor, disruption to services, damage due to
Fig. 7 could potentially be even higher than that due to dragging down of adjacent attached buildings and service-
complete structural collapse, since in addition to the ability issues related to change of levels of doors and entrances.
replacement cost of the building there will be the costs Additionally there may be aesthetic issues due to level
associated with the demolition of the tilted or settled building, differences between a settled building and neighbouring
and excavation of the existing foundations. buildings, or sidewalks. The damage state will be related to
Figure 8 illustrates another possible failure whereby the extent of settlement and ease of repair. Unacceptable
settlement causes damage to the floor slab; this may not be settlement or tilt may require structural remedial works in the
damage to the main load-carrying system, but it will represent a form of jacking-up and underpinning, or by grout injection. In
weakening of the structure due to loss of diaphragm action. At the worst cases, repair costs may be excessive and the buildings
the same time, the disruption to the function of the building will be demolished.
could be very severe. For consistency with the structural damage levels described
As mentioned previously, there is no analytical approach for in Table 2, both rotational and settlement limits must also be
the estimation of damage levels of buildings that have rotated related to the repair cost ratio, although determining the cost of
as a rigid-body, since there is no structural demand (apart from foundation repairs is significantly more complex than structural
possible PKD effects at larger rotations). Therefore the repairs. Issues that complicate the estimation of likely repair
22 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

damage. The apartment building shown in Fig. 7(b) tilted by


approximately 5 degrees and was subsequently demolished,
whereas Hwang et al. [38] report buildings that tilted by up to 9
degrees in the Chi-Chi earthquake and were repaired using oil
jacks to re-level them.
Proposed limits for grouping the rigid body response of
buildings to ground failure into similar ranges that coincide
with structural damage definitions are presented in Table 3.
This scale enables comparison of damage in different regions
and different earthquakes as well as providing improved
descriptors of predicted damage distributions. Sources of
information for determining acceptable levels include reports
from past earthquakes, technical papers related to earthquake
damage levels, and published tolerances for non-earthquake
related foundation deformations [39].
These limits have a high associated uncertainty due to the
lack of available data regarding repair methods and costs for
settled and rotated buildings. There are clearly many factors
influencing rigid body damage levels that are both difficult to
define and to quantify. Therefore the body of field data needs to
be significantly expanded, and supplemented with local
knowledge and judgement; without this it is not possible to
quantify the uncertainties associated with these limits.
Fig. 5. Liquefaction induced damage of a technical school building after the The limits presented in Table 3 for stiff foundations
1999 Chi–Chi earthquake. (Courtesy: Marshall Lew, Los Angeles Tall subjected to ground deformations are compared in Fig. 9
Buildings Structural Design Council).
with vulnerability curves derived for reinforced concrete frame
buildings on flexible foundations [33]. The former are
costs of buildings that have undergone rigid body deformation deterministic, in that they ignore the variability in building
include the strength of the foundations to survive the planned dimensions associated with the grouping of many buildings
remedial action, which may not be easily determined, and into a single classification. The latter include variability in the
regional variabilities in construction practices and tolerance to geometric and material properties, but not in the demand. An

BUILDING FOUNDATION MODE OF GROUND SOLUTION FOR


CLASSIFICATION CLASSIFICATION DEFORMATION DAMAGE ESTIMATION
Uniform vertical

Differential vertical
‘Structural damage’.
Flexible foundations Analytical solutions
Differential lateral
possible
Combined lateral and vertical

Uniform vertical

Stiff shallow foundations Differential vertical ‘Rigid body’ response.


RC Frame Empirical solutions
Buildings OR
floating piles Differential lateral required.

Combined lateral and vertical

Uniform vertical Significant damage is


unlikely and can be
Differential vertical ignored on a regional
Piles to firm scale
stratum
Differential lateral
Outside the scope of this
Combined lateral and vertical paper

Fig. 6. Modes of liquefaction-induced damage within one building classification.


J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 23

Table 2
Structural damage state descriptions for RC frame buildings (after Crowley et al. [31])

Structural damage band Description


None to slight Linear elastic response, flexural or shear type hairline cracks (!1.0 mm) in some members, no yielding in any critical
section
Moderate Member flexural strengths achieved, limited ductility developed, crack widths reach 1.0 mm, initiation of concrete
spalling
Extensive Significant repair required to building, wide flexural or shear cracks, buckling of longitudinal reinforcement may occur
Complete Repair of building not feasible either physically or economically, demolition after earthquake required, could be due to
shear failure of vertical elements or excessive displacement

intersection of the vertical line with vulnerability curve for that pragmatic solution therefore seeks to deal with all the
limit state at a probability of failure, Pf, of 0.5 would suggest uncertainties, rather than eliminate (or ignore) them.
nominal agreement between the two. For rigid body rotations
there is a slightly lower tolerance to small movements, being
4.1. HAZUS
controlled by aesthetic issues rather than the qualitative
parameter of yield stress, but higher tolerances to larger
In terms of recognising the need to have a ‘ground-failure
movements, being less sensitive to increasing movements than
component’ and including all of the necessary steps to analyse
in the case of structural deformations.
damage due to ground failure, the HAZUS [22] loss-estimation
methodology has made a significant contribution to the field. In
4. Liquefaction in earthquake loss estimation models particular, the HAZUS methodology is advantageous to loss
modellers in being comprehensive, openly available, and fully
The subject of this paper must be put into context by documented. Recognising the low level of geotechnical and
reiterating that for buildings the liquefaction component of a building inventory data that most loss estimations have to use,
loss estimation model is, in the majority of cases, less HAZUS presents a simplified approach that seeks to resolve the
significant to the final results than the ground shaking contradictory demands of minimising the amount of additional
component (e.g. [1]). Nonetheless, the fact that liquefaction data required and producing realistic estimates of the resulting
is a secondary hazard in the context of earthquake-induced damage. It is the only published method for evaluating building
losses by no means diminishes its importance. There is a need damage due to ground failure that does so in a detailed manner,
for comprehensive methodologies capable of estimating taking account of the expected mode of failure and foundation
damage caused by all expected hazards to all elements of a type. Previous methodologies such as ATC-13 [40] simply
region’s infrastructure. applied a factor to the calculated ground shaking-induced
Any methodology to be added into a loss estimation should damage to estimate the additional damage due to ground
provide a pragmatic solution to the issues of spatial uncertainty failure.
and variability in ground conditions and exposed building Nonetheless, there are many additional uncertainties
stock, while still producing a realistic and meaningful generated through the simplifying assumptions made in
estimation of the expected damage distribution as a result of HAZUS, and these are poorly constrained and considered.
earthquake hazards. The particular requirements of a frame- Bird et al. [36] showed that despite the use of published
work for incorporating liquefaction into a loss model are: references and logical analysis to determine the expected
ground deformations, the procedure is based in a large part on
1. Consistency with other components of the model, in terms subjective judgement, and therefore has no particular advan-
of complexity, accuracy and treatment of uncertainties. tage over other empirically-based procedures. The principal
2. Appropriate input data, commensurate with available shortcomings of the HAZUS ground-failure component relate
resources for a regional study, where detailed in situ not to the assumptions and over-simplifications made in the
measurements are unlikely to be feasible in terms of cost or treatment of ground failure, discussed by Bird et al. [36], but
time. rather to what it fails to do, namely:
3. Compatibility of results with the estimation of damage
caused by ground shaking, i.e. using consistent damage † It does not distinguish between different shallow foundation
definitions as described in the previous section. systems, i.e. flexible or stiff, which will respond very
differently to ground failure
In terms of accuracy, the optimum approach would be to † The thickness of the liquefied layer is not evaluated; clearly
predict the probability of liquefaction, the expected ground this will influence the extent of the damage. Without in situ
deformation and the resulting damage to buildings using in situ data, to do so is almost impossible, which highlights the
geotechnical data and detailed building inventory data difficulty in attempting to incorporate liquefaction without
including foundation type. This would require specialist such data.
input, and above all, analysis to determine the building † It does not consider different building responses to uniform
response, and is therefore likely to be prohibitive. The or differential foundation deformations, and how these will
24 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

Ruptured ground
floor slab

∆v ∆v

Fig. 8. Potential building response due to uniform vertical settlements in level,


homogeneous ground.

relationships between uncertain input and predicted results,


are all ignored.

To estimate building damage due to a geotechnical


phenomenon, as HAZUS attempts to do, with the only input
data being a susceptibility rating between none and very high,
according to the ratings presented by Youd and Perkins [4], is
intuitively flawed. Expected earthquake losses due to liquefac-
tion cannot be estimated in the absence of any geotechnical
information, at least not in a quantitative manner, and not from
the limited information currently available in this field. The
authors believe that to do so represents an unrealistic
simplification, because of the complex and unquantified
uncertainties associated with the subsequent results.

4.2. An improved framework for ground-failure


induced losses

Relative to the uncertainties associated with the ground-


shaking demand and building response, those associated with
the occurrence and consequences of liquefaction are, if not
larger, certainly more complex, and a number of additional
variables, with large associated uncertainties, are necessary to
define them.
Analytical solutions have been developed for building
vulnerability to ground deformations, for buildings on flexible
footings, presented by Bird et al. [33]. These have the
advantage of being based upon straightforward structural
mechanics rather than empirical data, and are fully compatible
with the estimation of ground shaking-induced structural
damage presented by Crowley et al. [31], thus satisfying the
Fig. 7. Buildings damaged beyond repair due to rigid body rotation caused by three requirements listed at the beginning of Section 4.
foundation failure in the city of Adapazari, Turkey, after the 1999 Kocaeli
Furthermore, the probabilistic framework allows rigorous
earthquake: (a) rotation only, (b) combined settlement and rotation. Photos
courtesy of the Earthquake Engineering Field Investigation Team (EEFIT). treatment of the associated uncertainties.
Fig. 10 illustrates the principal stages in the evaluation of
building damage in a region with liquefaction potential.
vary with superstructure and foundation type is not Complete definition of each stage in Fig. 10 becomes complex.
considered. The illustration shown in the figure uses the analytical
† Natural variability of soil properties, epistemic uncertain- procedures presented by Bird et al. [33]. The uncertainties
ties related to the choice of calculation methodology, and within this procedure are described in Section 2.3.
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 25

Table 3
Suggested limit states for rigid body settlement and rotation due to earthquake-induced ground deformations beneath RC frame buildings [39]

Damage state Structural damage (see Table 2 for full descrip- Additional description (rigid body deformation) Settlement (D) Rotation (q)
tion) only only
Slight Hairline cracks only Repairs may be necessary for aesthetic reasons D%0.1 m q%0.68 1/100
Moderate Some cracks in load-bearing elements Repairable damage, serviceability and/or 0.1 m!D%0. 0.68!q%2.38 1/
functionality affected 3m 100–1/25
Extensive Wide cracks and buckling of longitudinal Uninhabitable, but repairable 0.3 m!D%1. 2.38!q%4.68 1/
reinforcement 0m 25–1/12.5
Complete Repair not feasible, shear failures or excessive Demolition cheaper than repair. Structural integ- R1.0 m qR4.68R1/12.5
displacement rity affected, possible instability

4.3. Combining liquefaction and ground-shaking hazards If the probability of liquefaction is negligible, then damage
will be due to ground shaking only, and if the occurrence of
The complex issue of combined ground shaking and liquefaction is certain (i.e. P(L)Z1) then damage will be due to
liquefaction has been touched upon in many previous liquefaction only.
publications. Post-earthquake field reports of tilted or settled
buildings that do not exhibit any signs of structural damage or 2. The two hazards do interact, i.e. the final damage state of
distress due to ground shaking, are common, suggesting a ‘base any building is a function of both the initial damage caused
isolation’ effect, whereby the softening of the ground due to by ground shaking and any subsequent damage caused by
liquefaction significantly increases the period of the ground liquefaction. In this case the final damage distribution is
shaking, such that for most typical structural periods the calculated as follows:
spectral accelerations are very low. Despite the theoretical and
observational support to this hypothesis, it has been clearly
PðDSÞ Z PðLÞPðDSjShaking plus LiquefactionÞ
demonstrated (e.g. [1,41]) that some strong shaking will occur
before the pore water pressures have built up sufficiently for C ð1KPðLÞÞPðDSjShaking onlyÞ (2)
liquefaction to initiate. A number of case histories of reports of
buildings damaged by both shaking and liquefaction are
presented by Bird and Bommer [1]. Furthermore, as Bray and In this scenario, the strong ground shaking, even though shorter
Stewart [42] reasonably observe, where a building has than the full duration, is of sufficient amplitude and duration to
collapsed, which many would if subjected to strong ground cause some initial damage to the building. When the building is
shaking followed by liquefaction, it is impossible to determine still in some damage state less than collapse, the subsequent
the extent of liquefaction-induced deformations. This remains deformation imposed on it due to liquefaction causes further
a complex and somewhat controversial issue. Presently, it can damage, either by increasing the deformation of the columns
be concluded that the ‘base-isolation’ effect of liquefaction (Fig. 4) or by causing the damaged building to settle and rotate
cannot be reliably predicted and cannot therefore be (Fig. 3). The final damage state is thus a function of the damage
guaranteed, either for design or assessment. Two potential
scenarios are described below: 1
0.9
1. The two hazards do not interact, therefore, any group of
0.8
buildings has a probability P(X) of being affected by
0.7
liquefaction, and a probability of 1KP(X) of being
0.6
damaged by ground shaking. The final damage distribution
Pf | ∆ FV

0.5 vertical lines show suggested


can therefore be estimated as follows: rigid body limits for a 4m wide
0.4 building rotating due to
differential settlement beneath
0.3 its base
PðDSÞ Z PðLÞPðDSjLiquefactionÞ
0.2
C ð1KPðLÞÞPðDSjShakingÞ (1) 0.1 LS1 LS2 LS3

0
0 0.2 0.4 0.6 0.8
where Differential Settlement of Foundations, ∆ (m)
FV
DS damage state (e.g. slight, moderate, extensive or
complete) Fig. 9. Comparison of rigid body rotational limits with vulnerability curves for
differential settlements by Bird et al. [33]; curves are derived for poor quality
P(L) Probability of liquefaction. This should be most onerous
RC frame buildings, with 4 m beam length, with flexible foundations. Vertical
combination of TL (thickness of liquefied layer) and P(L) lines are for 4 m wide buildings with stiff foundations. LS1, boundary between
with respect to the damage distribution, therefore, a slight and moderate damage; LS2, boundary between moderate and extensive
number of scenarios may need to be considered. damage; LS3, boundary between extensive and complete damage.
26 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

Fig. 10. A framework for estimating damage in a zone of liquefaction susceptibility. The principal components are in the column on the left-hand side of the figure,
and a simplified illustration is shown on the right. This example considers P(L)Z0.5, differential vertical movements only and RC frame buildings with flexible
foundations. S, slight; M, moderate; E, extensive; C, complete damage (see Table 2). F(PGDjUV) signifies the CDF of PGD given that the ground deformation is
uniform vertical etc.

state reached at the onset of liquefaction, and hence, how much with field observations such that the complexities of the hazard
capacity remains to resist the liquefaction demand. interaction are implicitly included. A more rigorous approach
One suggested approach for including combined damage for determining the overall damage for the second scenario
would be to calibrate building vulnerabilities to liquefaction above (i.e. interaction), when there is sufficient data to do so, is
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 27

briefly considered. However, it is recognised that there remain (a)


difficulties in determining when this will occur as opposed to LS1 LS2 LS3
F
the first scenario (i.e. no interaction), and that such assessments
may in many cases be beyond the scope of a loss estimation slight moderate extensive complete
model. Nonetheless, the procedure described below is a useful
illustration of the potential influence of liquefaction following
ground shaking.
A simplified behaviour model to determine the displace-
ment capacity at the end of ground shaking is illustrated in
Fig. 11. If the structure has not exceeded its yield capacity, i.e.
it has remained elastic, then the full displacement capacity is
available to resist liquefaction. If, however, either the yield
capacity or the second limit state, which defines the boundary
∆y ∆’ LS2 ∆ LS2 ∆ LS3 Displacement
between moderate and extensive damage, has been exceeded, Capacity
then it is nominally assumed that such buildings are, on Average available displacement
capacity of structure exceeding
average, half way between the two limits, and the remaining LS1 at end of ground shaking
capacity is calculated accordingly. The capacity is calculated
making the simplifying assumption that the unloading path has (b)
the same gradient as the initial loading path. If the structure LS1 LS2 LS3
F
has suffered complete damage due to ground shaking (i.e. it has
failed the third limit state), then it cannot be further damaged slight moderate extensive complete
by the occurrence of ground failure.
A further simplifying assumption has been made that all
damage due to ground shaking occurs in the first part of the
earthquake and the liquefaction-induced ground deformation
will occur towards the end of, or subsequent to the earthquake.
This ignores the potential for low-frequency ground shaking or
low-frequency ground oscillations after the onset of liquefac-
tion. A further assumption is made in Fig. 11 that the same
structural element is loaded by both ground-shaking and
ground deformation. This is likely to be the case for buildings ∆y ∆’ LS2 ∆ LS2 ∆ LS3 Displacement
Capacity
exhibiting column sway (soft-storey) behaviour under ground Average available displacement
shaking, followed by deformation beneath flexible foundations capacity of structure exceeding
LS1 at end of ground shaking
as shown in Fig. 4(a).
Based upon this model of ‘residual’ capacity at the end of Fig. 11. Schematic illustration of force–displacement paths for a structural
cyclic loading due to ground shaking, the schematic diagram element exceeding each limit state.: (a) to the end of the strong ground shaking
shown in Fig. 12 can be followed to obtain the final damage cycle (b) from the onset of liquefaction, showing the average available capacity
the element will have left to resist liquefaction-induced deformations.
state.

Building Classification Damage state at end of Potential damage states


strong shaking after liquefaction

Slight

Moderate
Slight
Extensive
Complete

Moderate
Moderate
RC Frame Extensive
Buildings. Pad
footings Complete

Extensive
Extensive
Complete

Complete Complete

Fig. 12. Cumulative structural damage due to ground shaking followed by liquefaction.
28 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

In order to implement the procedure shown in Fig. 12, an Bray and Stewart [42] did not suffer any ground failure.
important variable is the degree to which the occurrence of This value ignores the buildings which collapsed and
liquefaction will modify the demand compared to the case therefore made it impossible to determine the occurrence of
where no liquefaction occurs. It is hypothesised that in this case ground failure and could therefore be an over-estimate of
the peak acceleration will still occur, but the number of cycles the true value. This can be compared with a factor of 75%
of strong motion would be reduced. assumed in HAZUS to represent the area of ground that
would not liquefy even given a high susceptibility to
4.4. Application of the new framework liquefaction.
† The proportion of buildings affected by liquefaction where
A demonstration of the approach described in the preceding the settlements were uniform rather than differential. The
sections and summarised in Fig. 10 is briefly presented using presence of relatively stiff raft foundations in Adapazari
measurements and observations from central Adapazari in the would certainly have influenced this by increasing the
1999 Kocaeli earthquake. The data from Adapazari allow an likelihood of uniform settlements. A nominal factor of 10%
estimation of both the strength of ground shaking and the has been assigned for illustrative purposes.
magnitude of the liquefaction-induced permanent ground
deformation. This means that some of the uncertainty The damage distribution has been predicted as follows:
associated with the estimation of the demand is reduced. The
remaining uncertainty relates to the spatial variability across (i) Ground shaking damage to RC frame buildings, using
the study region, and the natural heterogeneity of the ground the methodology presented by Crowley et al. [6].
conditions. The example presented considers the semi- (ii) Liquefaction damage to RC frame buildings with
hypothetical case of reinforced-concrete frame buildings on flexible foundations due to differential settlements
flexible foundations subjected to the same level demand as the using the analytical solutions presented by Bird et al.
buildings in central Adapazari, which were in reality mainly on [33].
stiff raft foundations. (iii) Combined ground-shaking and liquefaction, assuming
The ground shaking demand in downtown Adapazari must ‘base-isolation’, i.e. the two hazards are mutually
be estimated using attenuation relationships, since there were exclusive, and 60% of the buildings are damaged by
no strong-motion instruments in the immediate area. However ground shaking only, the other 40% by liquefaction
there were sufficient strong-motion records from the near-fault only.
region to guide the selection of an appropriate attenuation (iv) Combined ground shaking and liquefaction with no
relationship and hence a degree of reliability in the calculated ‘base-isolation’ i.e. over the 40% of the area where
response spectra, as described by Bird et al. [36]. Thus the liquefaction occurs, buildings are damaged by ground
variability of demand is modelled by allowing a range of site shaking then liquefaction, as suggested in Fig. 12.
classifications from D to E, and a range of source-site distances
Since the demand is predetermined for the sake of this
of 6–8 km.
illustration, it is possible to inspect the hypothetical distri-
Almost no lateral spreading was observed in Adapazari,
therefore only settlement-induced damage is considered. The butions of damage due to both ground shaking and liquefaction
ground deformation demand has been estimated from the range in the same locality.
of settlements reported for individual buildings by Yoshida Fig. 13 shows that liquefaction alone is less damaging to the
et al. [43], assuming that these are directly analogous to the RC frame buildings than ground shaking alone. Combining the
differential settlements in the ground beneath the buildings. two hazards in the manner suggested in Fig. 12 would increase
The median and standard deviation of the differential 0.8
settlements have been calculated from the data, where the shaking differential settlement
Percentage of affected buildings

median differential settlement was 8 cm, with a lognormal 0.7


standard deviation of 0.6 (i.e. the data required at Step 4 in 0.6
Fig. 10). These values are significantly lower that the
maximum settlements reported in Adapazari, which exceeded 0.5
1.0 m in areas, because they are differential values across the
0.4
footprint of a single building. For a 4 m wide building on rigid
foundations, 8 cm corresponds to a rotation of 1/50 or 0.3
approximately 18 which appears a reasonable estimate of the
0.2
median value based upon the observations of Bray and Stewart
[42]. 0.1
Additional information required to enable an estimation of
0
liquefaction-induced damage includes:
Slight Moderate Extensive Complete

† The proportion of buildings not affected by liquefaction. In Fig. 13. Comparison of damage distributions due to ground shaking only or
central Adapazari, at least 60% of the buildings surveyed by liquefaction-induced differential settlements only.
J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30 29

0.8 will respond as rigid bodies to deformations at foundation


shaking only level, and those on flexible (i.e. unrestrained) shallow
0.7
foundations, which will undergo structural deformations. As
Proportion of buildings

shaking and liquefaction


0.6 there is no obvious analytical solution to the former case,
0.5 shaking or liquefaction suggested limits have been proposed in Table 3, based largely
(base isolation)
0.4
upon empirical criteria. There remains an urgent need for
improved field data for the calibration of such scales. There is a
0.3 significant gap in current field survey data relating to post-
0.2 earthquake repair methods for damaged buildings. In parallel
0.1 with this is a need for improved analytical solutions, such as
those presented by Bird et al. [33], since even an improved
0
Slight Moderate Extensive Complete
empirical database will be small and sporadic in comparison to
those used to evaluate other insurance premiums. For piled
Fig. 14. Comparison of the two damage scenarios described in Section 4.3 to foundations many of the issues to consider are similar, i.e. the
the damage caused by ground shaking only. consequences of pile deformation on the supported building.
Additional complexities must also be introduced such as the
the overall level of damage as shown in Fig. 14. Only damage founding level of the piles compared to the thickness of the
caused by differential settlement to RC frame buildings with liquefied layer, and the lateral resistance capacity of the piles.
flexible foundations is considered in the liquefaction scenario, Methods for the identification of liquefaction susceptibility,
ignoring many of the complexities associated with different and the potential for the initiation of liquefaction under a given
foundation types and damage mechanisms. level of shaking, continue to be developed and refined, but
Fig. 14 shows the importance of the debate as to whether without the capability to determine what this will mean for the
‘base isolation’ will occur, since, the computed damage affected infrastructure, the usefulness of such methods is
distributions and their consequences are significantly different inevitably limited. The variability and uncertainty relating to
for the two scenarios in Section 4.3. An incorrect assumption of the demand remains a major issue for the evaluation of
no interaction between the two hazards could produce liquefaction-induced damage. Probabilistic procedures are
unconservative results. needed for the estimation of both average and differential
horizontal and vertical ground movements before a complete
methodology can be developed. This has been identified as one
5. Discussion
of the present challenges of geotechnical earthquake engineer-
ing (e.g. [44,21]).
The role of engineering judgement in the estimation of
This paper has shown that to incorporate liquefaction-
liquefaction-induced damage in earthquake loss models
induced damage into an earthquake loss model, following the
warrants a brief discussion. There is a tendency in earthquake
framework shown in Fig. 10, is a complicated procedure.
loss estimations to seek to standardise procedures such that
However, failure to consider each of these issues can produce
they can be applied to different regions by different users in
significant errors in the resulting damage estimations. Full
order to obtain comparable results. This feature is particularly
consideration as to the necessity of introducing liquefaction
desirable for insurance companies, and is the basis of the
into a loss model must be made [1]. Where it is deemed an
ground failure component of the HAZUS methodology [22],
essential component of the model, then the additional data and
which contains many simplifications, including ‘hidden’
simplifications based upon the developers’ judgement. In the computational requirements described in this paper should be
field of geotechnical engineering, the role of judgement is incorporated.
particularly significant, and, when extended to the field of
liquefaction engineering, it is essential to deal with uncertain- Acknowledgements
ties ranging from the choice of the most appropriate
methodology to the selection of input parameters. No frame- The work of the first and third authors was respectively
work could be developed that is universally applicable to all funded through a Marie Curie fellowship and the SAFERR
regions and ground conditions, thus it is imperative that Research Training Network from the European Commission.
decisions are made on the basis of individual studies. Financial assistance from the EPSRC and Arup is also
Probabilistic methods as presented by Crowley et al. [31] and gratefully acknowledged. The authors are grateful to the
Bird et al. [33], therefore, have a doubly important role to play, anonymous reviewer for constructive comments that improved
since as well as representing spatial variability they are the manuscript.
required to describe the uncertainties related to these
judgements. References
In the context of evaluating building damage caused by
liquefaction-induced ground deformations, separate solutions [1] Bird JF, Bommer JJ. Earthquake losses due to ground failure. Eng Geol
are required for buildings on stiff shallow foundations, which 2004;75(2):147–79.
30 J.F. Bird et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 15–30

[2] FEMA-249. Assessment for state-of-the-art earthquake loss estimation [23] Sadigh K, Egan JA, Youngs RR. Specification of ground motion for
methodologies. Federal Emergency Management Agency, Washington, seismic design of long period structures, 1986. Earthq Notes 1986;57(1):
DC, 1994 13 [Relationships are tabulated in Joyner and Boore (1988)].
[3] Juang CH, Yuan H, Li DK, Yang SH, Christopher RA. Estimating [24] Rauch AF, Martin JR. Predicting the maximum and distribution of
severity of liquefaction-induced damage near foundations. Soil Dyn displacements on liquefaction-induced lateral spreads. Fourth inter-
Earthq Eng 2005;25:403–11. national conference on recent advances in geotechnical earthquake
[4] Youd TL, Perkins DM. Mapping liquefaction-induced ground failure engineering and soil dynamics in honor of Professor W.D.Liam Finn, San
potential. ASCE J Geotech Eng Div 1978;104(GT4):433–46. Diego, California; 2001 [Paper No. 4.18].
[5] Bommer J, Spence R, Erdik M, Tabuchi S, Aydinoglu N, Booth E, et al. [25] Power MS, Perman RC, Wesling JR, Youngs RR, Shimamoto MK.
Development of an earthquake loss model for Turkish catastrophe Assessment of liquefaction potential in the San Jose, California, Urban
insurance. J Seismol 2002;6(2):431–46. Area. Fourth international conference on seismic zonation, vol. 2; 1991. p.
[6] Crowley H, Pinho R, Bommer JJ. A probabilistic displacement-based 677–84.
vulnerability assessment procedure for earthquake loss estimation. Bull [26] Baecher GB, Christian JT. Reliability and statistics in geotechnical
engineering. Chichester, England: Wiley; 2003.
Earthq Eng 2004;2(2):173–219.
[27] Crowley H, Bommer JJ, Pinho R, Bird JF. Epistemic uncertainty in loss
[7] Youd TL, Idriss IM, Andrus RD, Arango I, Castro G, Christian JT, et al.
models. Earthq Eng Struct Dyn; 2005; 34 (14): 1653–1685.
Liquefaction resistance of soils: summary report from the 1996 NCEER
[28] Al-Momani NM, Harrald JR. Sensitivity of earthquake loss estimation
and 1998 NCEER and NSF workshops on evaluation of liquefaction
model: how useful are the predictions. Int J Risk Assess Manage 2003;
resistance of soils. ASCE J Geotech Geoenviron Eng 2001;127(10):
4(1):1–19.
817–33.
[29] Bardet JP, Kapuskar M. An analysis of the sand boils due to liquefaction
[8] Cetin KO, Seed RB, Der Kiuregian A, Tokimatsu K, Harder LF, in the San Francisco Marina district during the 1989 Loma Prieta
Kayen RE, et al. Standard penetration test-based probabilistic and earthquake. J Geotech Eng ASCE 1993;119(3):543–62.
deterministic assessment of seismic soil liquefaction potential. J Geotech [30] Kircher CA, Reithermann RK, Whitman RV, Arnold CA. Estimation of
Geoenviron Eng 2004;130(12):1314–40. earthquake losses to buildings. Earthq Spectra 1997;13(4):643–61.
[9] Rodriguéz-Marek A, Luccioni LX, Cetin KO. Probabilistic assessment of [31] Crowley H, Pinho R, Bommer JJ. A probabilistic displacement-based
liquefaction over large areas. Fourth international conference on recent vulnerability assessment procedure for earthquake loss estimation. Bull
advances in geotechnical earthquake engineering and soil dynamics Earthq Eng 2004;2(2):173–219.
university of Missouri-Rolla 2002 [Paper No. 4.24]. [32] Sancio RB, Bray JD, Durgunoglu T, Onalp A. Performance of buildings
[10] Zhang J, Zhao JX. Empirical models for estimating liquefaction-induced over liquefiable ground in Adapazari, Turkey. 13th world conference on
lateral spread displacement. Soil Dyn Earthq Eng 2005;25:439–50. earthquake engineering, Vancouver, Canada 2004 [Paper No. 935].
[11] Youd TL, Hansen CM, Bartlett SF. Revised multilinear regression [33] Bird JF, Crowley H, Pinho R, Bommer JJ. Assessment of building
equations for prediction of lateral spread displacement. ASCE J Geotech response to liquefaction-induced differential ground deformation.
Geoenviron Eng 2002;28(12):1007–17. Engineering 2005; 38(4), in press.
[12] Bardet JP, Tobita T, Mace N, Hu J. Regional modeling of liquefaction- [34] Mejia LH, Hughes DK, Sun JI. Liquefaction at moss landing during the
induced ground deformation. Earthq Spectra 2002;18(1):19–46. 1989 Loma Prieta earthquake. 10th world conference on earthquake
[13] Youd TL, Perkins DM. Mapping of liquefaction severity index. ASCE engineering. Rotterdam: A.A.Balkema; 1992. p. 1435–40.
J Geotech Eng 1987;113(11):1374–92. [35] Marino GG. Earthquake damage inspection, evaluation and repair.
[14] Rauch AF, Martin JR. EPOLLS model for predicting average Lawyers & Judges Publishing Company, Inc., Tucson, AZ; 1997.
displacements on lateral spreads. ASCE J Geotech Geoenviron Eng [36] Bird JF, Bommer JJ, Bray JD, Sancio RB, Spence RJS. Comparing loss
2000;126(4):360–71. estimation with observed damage in a zone of ground failure: a study of
[15] Tokimatsu K, Seed HB. Evaluation of settlements in sands due to the 1999 Kocaeli earthquake in Turkey. Bull Earthq Eng 2004;2(3):
earthquake shaking. ASCE J Geotech Geoenviron Eng 1987;113(8): 329–60.
861–78. [37] Kehoe BE. Conflicting issues in post-earthquake damage evaluations
[16] Ishihara K, Yoshimine M. Evaluation of settlements in sand deposits Sixth US national conference on earthquake engineering, 1998. Oakland,
following liquefaction during earthquakes. Soils Found 1992;32(1): California: Earthquake Engineering Research Inst.; 1998.
[38] Hwang J-H, Yang C-W, Chen C-H. Investigations on soil liquefaction
173–88.
during the Chi-Chi earthquake. Soils Found 2003;107–23.
[17] Shamoto Y, Zhang J-M, Tokimatsu K. New charts for predicting large
[39] Bird JF. Earthquake loss estimations: modelling losses due to ground
residual post-liquefaction ground deformation. Soil Dyn Earthq Eng
failure. PhD Thesis. Imperial College London, University of London;
1998;17:427–38.
2005.
[18] Glaser SD. Estimation of surface displacements due to earthquake
[40] ATC. Earthquake damage evaluation data for California. ATC-13.
excitation of saturated sands. Earthq Spectra 1994;10(3):489–517.
Advanced Technology Council. Redwood City, California; 1985.
[19] Vick SG. Degrees of belief: subjective probability and engineering [41] Lopez FJ. Does liquefaction protect overlying structures from ground
judgement. Reston, VA: ASCE Press, American Society of Civil shaking? MSc dissertation Pavia. ROSE School, Italy; 2002.
Engineers; 2002. [42] Bray JD, Stewart JP. Damage patterns and foundation performance in
[20] Koutsourelakis S, Prévost JH, Deodatis G. Risk assessment of an Adapazari, in Kocaeli, Turkey earthquake of August 17, reconnaissance
interacting structure-soil system due to liquefaction. Earthq Eng Struct report. Earthq Spectra 2000;(Suppl. A):163–89.
Dyn 2002;31:851–79. [43] Yoshida N, Tokimatsu K, Yasuda S, Kokusho T, Okimura T.
[21] Seed RB, Cetin KO, Moss RES, Kammerer AM, Wu J, Pestana JM, et al. Geotechnical aspects of damage in Adapazari city during 1999 Kocaeli,
Recent advances in soil liquefaction engineering: a unified and consistent Turkey earthquake. Soils Found 2001;41(4):25–45.
framework. 26th annual ASCE Los Angeles geotechnical spring seminar, [44] Kramer SJ, Stewart JP. Geotechnical aspects of seismic hazards. In:
Long Beach, California. Keynote presentation 2003. Bozorgnia Y, Bertero VV, editors. Earthquake engineering: from
[22] FEMA. HAZUS-MH technical manual. Federal Emergency Management engineering seismology to performance-based engineering. FL, USA:
Agency, Washington, DC; 2003. CRC Press; 2004.

Vous aimerez peut-être aussi