Vous êtes sur la page 1sur 17

Int J Life Cycle Assess (2016) 21:17991815

DOI 10.1007/s11367-016-1127-4

WATER USE IN LCA

Critical analysis of life cycle impact assessment methods


addressing consequences of freshwater use on ecosystems
and recommendations for future method development
Montserrat Nez 1 & Christian R. Bouchard 2 & Ccile Bulle 3 &
Anne-Marie Boulay 4 & Manuele Margni 4

Received: 7 October 2015 / Accepted: 25 April 2016 / Published online: 24 May 2016
# Springer-Verlag Berlin Heidelberg 2016

Abstract be overcome to further method harmonisation. This is precise-


Purpose Anthropic water uses can affect aquatic and terres- ly the aim of this study.
trial ecosystems through various pathways. To address these Methods Existing methods were analysed against a scheme. It
impacts in life cycle assessment, an array of impact assess- consists of four issues (1) covered impact pathway, (2) structure
ment methods can be applied. The currently well-known re- of characterisation model and factor, (3) fate factor modelling,
view of methods carried out by the UNEP/SETAC Life Cycle and (4) effect factor modelling, further specified based on ten
Initiatives WULCA working group (Kounina et al. Int J Life criteria. Among these, one criterion evaluates the legitimacy of
Cycle Assess 18(3):707721, 2013) recommends that practi- using steady-state fate factors to model pulse-type water uses
tioners Bsimultaneously apply all indicators to evaluate dam- based on the theorem presented by Heijungs (Environ Sci
age on ecosystem quality and to cautiously sum up the score Pollut Res Int 2:217224, 1995). New terminology is proposed
into a single metric^. This call for caution is attributed to the for a proper description of the criteria. All of the criteria eval-
fact that methods reviewed cover different ecosystem targets. uate scientific and technical aspects. The analysis approach
Their characterisation factors and units also vary. However, involves a qualitative description of each model.
the review lacks a detailed analysis of compatibilities and Results and discussion Important findings of the analysis in-
coherence between methods that identifies inconsistencies to clude the following: (1) for several methods, the environmen-
tal intervention proposed and its connection to the impact
assessment phase are debatable; (2) the location of the mid-
Responsible editor: Matthias Finkbeiner point stressor (i.e. the indicator of change in the environment
Electronic supplementary material The online version of this article due to the environmental intervention) along the causality
(doi:10.1007/s11367-016-1127-4) contains supplementary material, chain causes problems of compatibility among fate factors of
which is available to authorized users.
different models; (3) it is appropriate to use the steady-state
solution to find the new system condition after pulse-type
* Montserrat Nez
montse.nunez-pineda@irstea.fr
water use. None of the models have justified this fundamental
choice before. Recommendations for future method develop-
ment respectively, involve the following: (1) avoid the use of
1
Irstea, UMR ITAP, ELSA Research Group & ELSA-PACT inventory information in characterisation models and jointly
Industrial Chair for Environmental and Social Sustainability develop inventory data and characterisation factors to ensure
Assessment, 361 Rue Jean Franois Breton,
34196 Montpellier, France the applicability of impact assessment methods; (2) use the
2 more environmentally relevant midpoint stressor (the one along
Dpartement de gnie civil et de gnie des eaux, Universit Laval,
1065 Avenue De la Mdecine, Qubec, QC G1V 0A6, Canada the causality chain closer to ecosystem damage; and (3) justify
3 the use of the steady-state solution for fate factor modelling.
CIRAIG, UQAM, cole des sciences de la gestion, 315 St.Catherine
Street East, office R-1020, Montreal, QC H2X 3X2, Canada Conclusions This study identifies sources of inconsistency in
4 the indicator structures analysed and provides recommenda-
Mathematical and Industrial Engineering Department, CIRAIG,
Polytechnique Montreal, C.P. 6079, succ. Centre-Ville, tions that will foster harmonisation. The current mismatch
Montral, QC H3C 3A7, Canada between methods leads us to not recommend aggregating
1800 Int J Life Cycle Assess (2016) 21:17991815

indicators into one single metric until a common framework 2013) and (b) dam operation (Humbert and Maendly
underpins existing and future water use impact assessment 2008).1 A question that naturally arises from such a vari-
methods for ecosystem quality. ety of models is as follows: how complementary and com-
patible are these models?
Keywords Characterisation models . Ecosystems . Answering this question requires a detailed analysis of these
Freshwater use . Life cycle assessment . Life cycle impact models that goes beyond the initial generic attempt made by
assessment Kounina et al. (2013). In this study, the authors recommended
that researchers Bsimultaneously apply all reviewed indicators
to evaluate damage on ecosystem quality^, namely those pre-
1 Introduction sented by Pfister et al. (2009), Hanafiah et al. (2011), van Zelm
et al. (2011) and Humbert and Maendly (2008). The authors
Human freshwater consumption and quality degradation suggested summing the score into a single metric while
can have several consequences (e.g., decrease in soil interpreting results cautiously as the methods cover different
moisture, water temperature increase, soil and endpoints. Despite this recommendation, these and other more
freshwater salinisation, and reduced freshwater availabili- recent methods for assessing ecosystem quality damage have
ty in water bodies or land flooding). These changes in had very limited applications. To improve understanding of
environmental conditions can threaten aquatic and terres- available methods and to foster the development of new
trial ecosystems. Furthermore, climate change will modify harmonised approaches, this study aims at providing a thor-
the distribution and seasonal availability of water around ough critical analysis of impact assessment methods that ad-
the world (IPCC 2007). Patterns of unsustainable human dress consequences of freshwater use on ecosystems. Our study
freshwater use and consumption combined with effects of complements the comparison of water scarcity indices and hu-
climate change may result in widespread damage to the man health damage indicators carried out by Boulay et al.
provisioning of ecosystem services and biodiversity (2015), as such type of comparison is missing for the ecosys-
(MEA 2005). It is therefore essential to develop model tem quality area of protection. The cientific quality, comple-
capabilities to assess the environmental impacts of fresh- mentarity and compatibility of the models are identified and
water use on ecosystem quality. discussed. Beyond the critical analysis of existing models, ter-
Life cycle assessment (LCA) is a standardised technique minology and guidelines are proposed for the consistent inte-
for quantifying the potential environmental impacts of a prod- gration of characterisation models. In addition, a unified ap-
uct through its full cycle (ISO 2006a; ISO 2006b). In LCA, the proach for the development of fate factors is proposed. To reach
term freshwater use is a generic term that groups all types of our objectives, a scheme of analysis is proposed and applied to
human freshwater resources use. Consumptive use (or water models for water use impact assessment on ecosystem quality
consumption) is a specific type of water use whereby water is published by 2014. The following factors are out of the scope
removed from but not returned to the same watershed. of the study: methods that address water use-related impacts on
Degradative use (or water degradation) involves withdrawal the areas of protection human health and resource depletion,
and release in the same watershed after water quality has been models assessing water degradation in traditional emission-
altered (Bayart et al. 2010). oriented impact categories as well as consequences of the use
Until a few years ago, water use was considered in of nonconventional sources of water (e.g. desalinated water).
LCA only as an inventory flow. Impacts of water use on
ecosystems were only covered by emission-related impact
categories (namely freshwater ecotoxicity, freshwater
acidification, ionising radiation, and freshwater eutrophi- 2 Materials and methods
cation). Lately, several authors have proposed characteri-
sation models that assess impacts on ecosystem quality in 2.1 Scheme used to analyse methods
the freshwater use impact category. Most of these models
are endpoint oriented (i.e. indicators are close to areas of The scheme we propose may be used in any life cycle
protection) and assess consequences of water consump- impact assessment (LCIA) model irrespective of the impact
tion (Pfister et al. 2009; Hanafiah et al. 2011; van Zelm category. The selection of the criteria is inspired by the eval-
et al. 2011; Verones et al. 2012, 2013a, b; Tendall et al. uation approach of the UNEP/SETAC LC-Initiative Working
2014). A few models assess impacts due to freshwater Group (Margni et al. 2008) used in the International Reference
degradation (Verones et al. 2010; Amores et al. 2013; 1
Following Kounina et al. (2013), we included this method in our anal-
Pfister and Suh 2015; Quinteiro et al. 2015). Other
ysis although it is not yet published in a peer-review journal due to its
models account for (a) aquatic biodiversity alteration as innovative nature and accessible documentation for analysing underlying
a result of exotic fish species introduction (Hanafiah et al. modelling choices
Int J Life Cycle Assess (2016) 21:17991815 1801

Life Cycle Data System (ILCD) handbook (JRC 2010; The more impact pathways are described, the more complete
Hauschild et al. 2013), by the requirements from ISO 14040 an assessment of potential damages of water use on ecosystem
and ISO 14044 standard (ISO 2006a; ISO 2006b) and by quality will be, but also the more potential problems related to
review works (Delbaere et al. 2009; Curran et al. 2011; inconsistencies between methods in the case where different
Kounina et al. 2013). The scheme consists only of scientific methods describe different pathways separately by using dif-
and technical issues (i.e. there are no stakeholder acceptance ferent structures of modelling and different indices.
criteria as in the ILCD handbook (JRC 2010) and in Kounina
et al. (2013)). It explores four issues that are further specified 2.1.2 Issue 2structures of the model and characterisation
based on ten criteria, as per Table 1. Some of our criteria have factor
not yet been applied to this typology of study. They were
selected because they refer to core of the scientific validity Criterion 2.1 The characterisation factor is expressed as a
of the methods in terms of their compliance with underlying fate factor multiplied by an effect factor. Influence of the
modelled physical processes (e.g. criterion 3.1) or to the spatial resolution on the characterisation factor structure
methods adhesion to fundamental requirements of LCA This criterion evaluates whether the model under analysis al-
(e.g. criterion 1.1). We do not examine aspects related to the lows considering both fate and effect coherently and separately.
ease of applicability, stakeholder acceptance and uncertainty Like the modelling structure for LCIAs of other impact path-
management as these issues are not indicative of the correct- ways (Margni et al. 2008; JRC 2010), to analyse this criterion
ness of the LCIA model itself. The critical analysis performed and to compare CFs of existing models, we adopt a character-
involves a qualitative description of each model. Based on the isation model structure that reflects the propagation of an envi-
proposed criteria, a scoring system can be further developed to ronmental intervention into environment (fate), followed by the
assess a models compliance with a specific criterion. This can alteration of an ecosystem (effect), as this structure suits the
be done only with a broad acceptance of such criteria. The examination of cause-effect chains typically used in LCA.
four issues and ten criteria are described further in the Please note that we use the widely accepted structure of CF
following. (Eqs. (1) to (4)) for the sake of method comparison in this
article. If this structure should be consensus-based recommend-
2.1.1 Issue 1covered impact pathway ed practice is out of the scope of this study.
An endpoint-oriented CF determines the additional or mar-
Criterion 1.1 The inventory consists of environmental in- ginal damage on ecosystem quality due to an incremental
terventions that are appropriately connected to the impact environmental intervention (water use) resulting from the
assessment model According to ISO 14044 (ISO 2006a; ISO product of a fate (FF) and effect (EF) factor:
2006b), the life cycle inventory (LCI) consists of elementary
flows entering and leaving the system under study and the Damages
C Fr;s F Fr;s  E Fs 1
LCIA step evaluates the effects of the elementary flows on Environmental interventionr
the environment. In this work, the term Benvironmental
intervention^ is preferred to the term Belementary flow^ as where the subscript r denotes the compartment where the envi-
the former is a broader term (Guine et al. 2001). As defined ronmental intervention occurs (receiving compartment) and
by Udo de Haes et al. (2002), an environmental intervention is where s denotes the compartment where the midpoint stressor
the amount of inert or living matter (including an area of land (term defined in criterion 2.2) is modelled. A compartment is
or a water stream) or energy (including noise and radiation) any water body or part of a water body (e.g. whole river or river
that is extracted, used or released into an environmental com- section). A given LCIA method may involve calculating one
partment. This criterion focuses on describing the appropriate- single CF for the entire modelled water compartment or many
ness of defining targeted environmental interventions for each CFs for smaller scale sections of a given compartment. In the
model analysed and whether a model effectively transmits latter case, CFs are usually aggregated as a final CF for a single
environmental interventions of the LCI to the characterisation compartment. Ways in which this is addressed affect CFs struc-
factor (CF) of the LCIA step. ture, and this can in turn affect degrees of compatibility between
methods. We include an analysis of this aspect in this criterion.
Criterion 1.2 Covered impact pathway and complemen- We do not discuss optimal CF spatial resolution level (i.e. if it
tarity with pre existing pathways is shown This criterion is should be sub-basin, basin, country or any other).
analysed by drawing the impact pathway described by each
method on a global cause-effect chain plot (Fig. S1 in ESM). Criterion 2.2 The midpoint stressor (MPS) is selected cor-
The plot is generated based on impact pathways covered by rectly We define a midpoint stressor as a change in an envi-
methods analysed in this work and is complemented with ronmental condition due to an environmental intervention that
qualitative information from a review of specialised literature. potentially places stress on ecosystems. The MPS can occur
1802 Int J Life Cycle Assess (2016) 21:17991815

Table 1 Critical analysis scheme


of impact assessment methods Issue Criteria
addressing freshwater use impacts
on ecosystems: identified issues 1. Covered impact pathway 1.1 The inventory consists of environmental interventions that are
and evaluation criteria appropriately connected to the impact assessment model
1.2 Covered impact pathway and complementarity with
pre existing pathways is shown
2. Structure of characterisation 2.1 The characterisation factor is expressed as a fate factor multiplied
model and factor by an effect factor. Influence of the spatial resolution on the
characterisation factor structure.
2.2 The midpoint stressor is selected correctly
3. Fate factor modelling 3.1 The fate factor equation is in accordance with underlying
physical laws that govern the modelled environmental intervention
3.2 The fate factor considers the current state of the environment
4. Effect factor modelling 4.1 The effect factor definition is coherent with the characterisation
factor and fate factor
4.2 The taxonomic and geographic coverage is as complete as possible
and the damage assessment is spatially resolved
4.3 Sub-steps of the effect mechanism are explicitly addressed in the effect
modelling
4.4 The effect factor considers the current state of the environment

anywhere along the cause-effect chain; it may be equal to the the FF is necessary to reflect uneven distributions and renew-
FF multiplied by the environmental intervention (as per Eq. ability levels among water sources (Yano et al. 2015).
(2))which would allow a consistency if applied coherently The FF definition of Eq. 2 is based on existing fate models
to all different impact pathwaysor not (e.g. if the character- for pollutant release (Heijungs 1995; Pennington et al. 2004).
isation factor structure is not split into a fate factor and an Z t
effect factor). Several appropriate MPSs may be selected MPSs dt
along the same impact pathway, but an appropriate MPS must FFr;s 0
2
be situated along the cause-effect chain. Note that a midpoint Environmental interventionr
stressor is not the same as a midpoint indicator. The term
stressor would have been used without any qualifier instead. Equation (2) recognises the time integrative nature of the
However, the term stressor has been applied in the LCI step environmental change due to an intervention. t may be set
(Udo de Haes et al. 2002). Adding Bmidpoint^ as a descriptor to either defined time frames, which allows calculating dy-
allows identifying the intermediary point in a causality chain namic solutions (e.g. t = 100 years for GWP100 in the climate
where a stressor is modelled. This criterion involves analysing change impact category), or a time-independent solution is
when the method under evaluation clearly identifies the mid- calculated (e.g. t in the toxicity impact categories) to
point stressor and whether the MPS is located along the cause- account for the final state of the environment when steady
effect chain, proving that there is a relationship between the state is reached.
environmental intervention, the MPS and the ecosystem attri- To solve Eq. (2), Heijungs (1995) has demonstrated the
bute to be protected. We also include in this criterion informa- mathematical relationship that may exist between the steady-
tion on the temporal resolution of the MPS. state solution of a continuous-type pollutant release and the
transient solution of a pulse-type pollutant release (Heijungs
1995). In practice, this means that LCIA may use much sim-
2.1.3 Issue 3fate factor modelling pler steady-state models rather than sophisticated time-
dependent unsteady-state models. However, before using
Criterion 3.1 The fate factor equation is in accordance steady-state models and data for water use FF modelling,
with underlying physical laws that govern the modelled Heijungs theorem should be checked. To determine the ap-
environmental intervention A fate model should address plicability of Heijungs (1995) theorem for characterising wa-
the propagation of a pulse-type environmental intervention into ter use-related impacts, this criterion evaluates the match be-
different environmental compartments, i.e. it should address a tween FFs of methods from the literature analysed in this
transient propagation problem that lasts for some time within study and FFs we derived following the abovementioned the-
the stressed compartment (Eq. (2)). For water use impact as- orem. We derived FFs for simple cases of the following envi-
sessment, the differentiation between water compartments in ronmental interventions: groundwater pumping, surface water
Int J Life Cycle Assess (2016) 21:17991815 1803

pumping, and heat release. In a first attempt, FFs were derived species and taxa. Biodiversity also varies across the planet.
for single compartment systems. Therefore, EFs for water use impact assessment shall be spa-
tially explicit to cover regional specificities of local biodiver-
Criterion 3.2 The fate factor considers the current state of sity, thus acknowledging that this spatial variability of biodi-
the environment One variable that is always included in the versity is rarely considered in current LCIA models due to a
FF equations we derived (see criterion 3.1 above) is the back- lack of region specific effect data. At the same time, effect
ground level (e.g. C0, V0). This criterion assesses whether the factors should ideally be globally applicable to allow the as-
method under analysis considers the initial value of the MPS sessment of products with life cycles distributed over the
(i.e. background level) in the FF. As the background level can world. This criterion focuses on the ways in which each meth-
vary from one geographical location to another, it must be od accounts for species of different taxa and for regional spec-
referred to regional conditions whenever the adequate infor- ificities in terms of effects.
mation on the initial state is available.
Criterion 4.3 Sub-steps of the effect mechanism are explic-
2.1.4 Issue 4effect factor modelling itly addressed in the effect modelling The EF can be explic-
itly divided into several sub-factors and be expressed as a
Criterion 4.1 The effect factor definition is coherent with combination of exposure, incidence and damage modelling
the characterisation factor and fate factor As the CF is steps as in toxic emission models (Pennington et al. 2006)
considered as a product of an FF and EF, mathematical for- (Eq. (4)). For the purposes of this critical analysis, we adopt
mulations of these three factors should be consistent. For ex- this structure because it is well adapted to modelling the
ample, the numerator of the FF and its units should be the cause-effect chain from the FF to the latest environmental
same as the denominator of the EF. Therefore, to be consistent impacts on ecosystems. The criterion analyses whether the
with the adopted CF model structure (criterion 2.1) and FF studied methods explicitly adopt such mechanistic structure,
definition (criterion 3.1), the effect factor should relate the even if factors are set to a value of 1 (e.g. exposure factor =1).
time-integrated MPS to the endpoint indicator measuring the Damage Exposure Incidence Damage
damage on the ecosystem quality area of protection. Based on = (4)
Stressor Stressor Exposure Incidence
Eqs. (1) and (2), and like existing effect models for pollutant
release, this translates into the following equation for EF:
Effect Exposure Incidence Damage
Damages factor factor factor factor
E Fs Z t 3
Midpoint stressors dt
0
The exposure factor determines to what extent the target
This criterion focuses on how this equation applies to the considered (e.g. affected living species) is exposed to a change
analysed methods and on how the connection of FF and EF in in environmental conditions.
terms of form and metric are carried out for those methods. The incidence factor determines how a change in expo-
Ecosystem quality damage (the numerator in Eq. (3)) may be sure levels affects the quality of the ecosystem (e.g., in terms
modelled using different approaches (see criterion 4.4). of potentially affected fraction of species (PAF)). Different
approaches can be adapted from other impact categories
Criterion 4.2 The taxonomic and geographic coverage is (e.g., the species-area relationship used in the land use cat-
as complete as possible and the damage assessment is spa- egory or exposure level-response curves used in emission-
tially resolved In LCA, damages to ecosystem quality are related categories). Exposure-response curves that relate ef-
usually measured using biodiversity indicators. Biodiversity fects from a midpoint stressor on several species may be
can be compartmentalised into a hierarchy of levels. At each established based on species sensitivity distribution func-
level, indicators may be defined. Curran et al. (2011) differ- tions (SSD) that relate a species community response to
entiated between four broad biodiversity levels (genetic, spe- the exposure level used to determine the proportion of spe-
cies, community and ecosystems) and three types of indicators cies affected of the incidence factor. The incidence factor
(composition, structure and function). Overall, existing LCIA relies on the same type of threshold for all exposure-
methods assume that species adequately represent the quality response curves.
of ecosystems. Individual species attributes or community in- The damage factor converts the potentially affected frac-
dices are chosen as a representative proxy of total biodiversity. tion of species (PAF) into a potentially disappeared fraction of
The larger the number of species of different taxa included in species (PDF) or into other metrics expressing biodiversity
indices, the more faithfully they can capture the real biodiver- damage in absolute terms (e.g. species year). To transform
sity loss as biodiversity is not evenly distributed among incidence in to damage, it is important to specify whether
1804 Int J Life Cycle Assess (2016) 21:17991815

SSD curves correspond to acute, sub-chronic or chronic ex- basin), receiving and stressed environmental
posure. Different relationships between affected species and compartments.
species loss have been proposed (Snell and Serra 2000; 3. Mathematical equation and unit of the fate factor at the
Goedkoop and Spriensma 2001; Jolliet et al. 2003): (1) native spatial scale.
PDF = PAF; (2) PDF = PAF 0.1; (3) PDF = PAF 0.5. Thus 4. Mathematical equation and unit of the effect factor at the
far, none of them have garnered broad scientific consensus. native spatial scale.
5. Mathematical equation and unit of the characterisation
Criteria 4.4 The effect factor takes considers the current factor at the native spatial scale.
state of the environment Ecosystem quality damage (the 6. Mathematical equation and unit of the characterisation
numerator of Eq. (3)) is sensitive to local conditions in real factor at the aggregated spatial scale.
life. LCIA models may apply three EF modelling approaches:
(1) marginal approach, for very small changes in the known For methods with the same native and aggregated spatial
local background, (2) average approach, calculated based on scales, information is reported under the aggregated spatial
the distance between the known local background and a pre- scale tag.
ferred state (typically set to a 0 effect state), and (3) linear
approach, for unknown background local level information
(Huijbregts et al. 2011; Hauschild and Huijbregts 2015). In
3 Results and discussion
LCA, local context information needed to model actual eco-
system responses in a product life cycle basis (i.e. from cradle
Characterisation models are qualitatively analysed below ac-
to grave) is not generally available. Hence, it is consensually
cording to the scheme introduced in Sect. 2. At the end of each
accepted in current practice that damage is not a function of
criterion description, we provide recommendations that can
time or MPS level but is linear to the environmental stressor
guide future developments.
(Hauschild and Huijbregts 2015). This linear approach has
been applied to all impact categories with varying degrees of
sophistication. For instance, toxicity (Rosenbaum 2015) and 3.1 Scheme used to analyse methods
particulate matter (Humbert 2015) models apply linear, no-
threshold curves, whereas acidification (Roy et al. 2012) and 3.1.1 Issue 1covered impact pathway
eutrophication (Struijs et al. 2011) consider a linearised re-
sponse based on the buffer capacity of the environment. Criterion 1.1 The inventory consists of environmental in-
Irrespective of the approach applied, once the EF curve is terventions that are appropriately connected to the impact
modelled, the impact will be proportional, i.e. linear, to the assessment model Most of the analysed methods have an LCI
environmental intervention set in the LCI (Hauschild and that consists of an environmental intervention and where the
Huijbregts 2015). This criterion evaluates whether the back- proposed characterisation factors can be directly applied to
ground level is a variable included in the EF equation and the assess the targeted environmental intervention. The
model approach applied. environmental intervention modelled by most LCIA
methods is a volume of consumed surface water or
2.2 Summary tables of existing models groundwater. However, for several methods, the
environmental intervention proposed and its connection to
We present individual tables (see Electronic Supplementary the impact assessment phase are debatable. Hanafiah et al.
Material (ESM)) as summaries of the characterisation models (2013) consider the mass of goods shipped (kg transported/
analysed. These tables are designed to facilitate an understand- year) as the environmental intervention. This is actually a
ing of the model structures and model comparison. Each table human activity that may lead to an environmental intervention
addresses the following topics: (i.e. an output to the environment) such as a biological alter-
ation involving the introduction of exotic species in a given
1. Covered impact pathway showing the start and end points watershed. The latter is in fact the midpoint stressor (the nu-
of the impact pathway, the fate and effect factor or the merator, see Eq. (2)) of the FF reported by Hanafiah et al.
directly calculated characterisation factor, steps of the (2013). The lack of proper inventory data and of inventory
cause-effect chain that are cited as explanations/ models that relate goods transportation to an actual environ-
justifications in model documentation but are not specif- mental intervention may explain this integration of an inven-
ically modelled in equations. tory and a fate models into a single relationship in the charac-
2. Structure of the characterisation model, including the terisation factor through what we call an adaptability factor.
identification of the native spatial scale (e.g. river section) We define an adaptability factor as a parameter containing
and the spatial scale of the aggregated CFs (e.g. river inventory information in the impact assessment phase.
Int J Life Cycle Assess (2016) 21:17991815 1805

Humbert and Maendly (2008) use the amount of water pragmatic solution for managing deficient inventory data.
turbined in a dam as the environmental intervention. To be When enhanced inventory datasets become available, the
consistent with their inventory, the CF relates potential impact adaptability factor should be expressed in inventories to
on species (in PDF) per area of the dam to the m3 of turbined generate coherent methods that always require the same
water. This is determined by dividing the impact associated types of standardised elementary flows. Overall, a joint
with the dam by the quantity of water turbined in each dam. development of inventory and characterisation models
This comes with the underlying assumption that the impact of should improve the coherence between these two phases
a dam is proportional to the amount of turbined water, which is of any LCA method.
not necessarily true as other key variables, such as changes in
downstream river flow variability, may influence damage on
Criterion 1.2 Covered impact pathway and complementarity
aquatic biodiversity. To capture all potential impacts of dams
with pre existing pathways is shown Figure 1 shows that the
on ecosystems, we recommend assessing not only dam oper-
models generally complement one another in the target impact
ation impacts, as the current method does, but also dam con-
pathways, although they do not exhaustively cover all possi-
struction impacts. Appropriate environmental interventions
ble pathways.
should accordingly be reported separately in the inventory.
Models for multiple compartment systems (e.g. impacts
For dam construction, we propose using the increase (flooded
on ecosystem caused by modifications of surface water
area) or decrease (dry out area) in the surface of a water
flow due to groundwater consumption) are still to be de-
body, i.e. change in land use (m2), as the inventory flow
veloped. Interconnections between surface and ground
as this is, for instance, the case for the ecoinvent database
water bodies are only unambiguously addressed in
(Weidema et al. 2013). For dam operation, the inventory
Verones et al. (2012, 2013a) and Amores et al. (2013) in
flow may well be the m3 of water turbined over a time
the case of wetlands. Van Zelm et al. (2011) also
period (m3 year), recognising that dam operation involves
modelled the environmental intervention and the effect
a change in river flow variability, but the amount of water
of water withdrawal on different compartments, but unlike
turbined may also be insufficient information alone for esti-
the abovementioned methods, FF and EF equations do not
mating corresponding impacts on ecosystem. Other data,
explicitly show established connections between the
such as mitigation practices (e.g. fish scales and derivation),
modelled compartments (i.e. relation between aquifer wa-
stream regulation practices (e.g. maintaining the river levels
ter extraction (FF) and effect on soil-water dependent ter-
during the reproduction period or not) may be important to
restrial plants (EF) is not made explicit in the equations
report on in the inventory. Further work on Humbert and
but is included in the characterisation model applied).
Maendlys (2008) impact assessment model is needed for
So far, research efforts have mainly focused on model-
the establishment of relevant connections with the quantity
ling surface and groundwater consumption pathways.
of water turbined as reported in the LCI (which may itself
Changes in water balance and water filtration caused by
need to be specified further).
land occupation and transformation are addressed by sev-
In Verones et al. (2010), the environmental intervention is
eral methods, although with partial pathways at either the
the product of a surplus temperature released and a cooling
LCI level (Nez et al. 2013), or midpoint (Mil i Canals
flow rate (C m3 water discharge/day), which is not an ele-
et al. 2009; Saad et al. 2013) and endpoint (Cao et al.
mentary flow currently reported in life cycle inventories (on-
2015) LCIA levels.
ly the amount of cooling water used, with no information on
Connections between water consumption and water qual-
the temperature increase for this water, is reported). This
ity degradation have thus far been limited to salinity increase
elementary flow relates well to their FF. However, the envi-
(Amores et al. 2013). No methods report on the negative
ronmental intervention metric should not depend on the
effect of lowering river water availability on the dilution of
ways in which heat is released into water (e.g. it should
toxic and nontoxic emissions. Furthermore, many environ-
not depend on the water cooling flow rate to make the en-
mental stressors that affect water quality are not yet consid-
vironmental intervention more general). It should be an
ered in LCA (e.g. bacterial contamination and water colour
amount and not a flow rate, as is the case for other inventory
change).
elements. It should also be expressed as a common heat unit
(e.g. the joule). Hence, we propose using heat release into
water, expressed in joules, as the environmental intervention Recommendations The diversity of hydrological patterns
for thermal pollution. and of the interconnections between water bodies makes it
difficult to comprehensively cover all impact pathways.
Recommendations The use of the adaptability factor Therefore, it would be useful to determine the relative
adopted by some impact assessment methods bridges the importance of different pathways in overall mechanisms
current gap between the LCI and LCIA and serves as a as was done by, for example, Weidema (2001) for the land
1806 Int J Life Cycle Assess (2016) 21:17991815

Fig. 1 Impact pathways currently covered by existing ecosystem quality analysis carried out: Mil i Canals et al. (2009), Nez et al. (2013), Saad
methods. Each method is shown in a different colour. The following et al. (2013), Pfister and Suh (2015), and Quinteiro et al. (2015)
methods are shown in this figure but are out of the scope of the critical

use impact category. This would help prioritise modelling tables) and later aggregated, based on different weighted
efforts. averages, native factors in the final CF spatial scale (e.g.
water basin, see Baggregated spatial scale^ information).
3.1.2 Issue 2structure of the characterisation model The following of the models analysed use this tiered CF
and factor structure: Humbert and Maendly (2008), Pfister et al.
(2009), Verones et al. (2010), van Zelm et al. (2011),
Criterion 2.1 The characterisation factor is expressed as a Hanafiah et al. (2013), and Tendall et al. (2014). Note
fate factor multiplied by an effect factor. Influence of the that not all these above-listed methods explicitly show
spatial resolution on the characterisation factor structure during method documentation such native fates, effects
An analysis of this criterion reveals that all of the char- or characterisation factors. For example, Pfister et al.
acterisation models distinguish between fate and effect (2009) first derived CFs on a 0.5 grid before aggregat-
sub-models with the exception of methods developed ing these damage factors by watershed and country,
by Pfister et al. (2009) and Humbert and Maendly which are the final spatial resolutions provided in the
(2008). Mathematical equations of the models analysed article. The rest of the methods directly calculated CFs
are presented in the summary tables (Sect. 3.2). As it can at the final spatial scale of regionalisation.
be seen in these tables, CFs are calculated following very
different approaches. Some methods have a tiered CF Recommendations The appropriate native resolution scale
structure whereby the authors first modelled the CF in should be chosen adequately, as within this scale all the com-
a part of the target water compartment (e.g. a grid cell, partment is homogeneously affected by the environmental in-
see Bnative spatial scale^ information in the summary tervention. The ways in which aggregation and weighting of
Int J Life Cycle Assess (2016) 21:17991815 1807

native factors into the final factor are performed (e.g. area- relationship between the environmental intervention and the
weighted, water consumption-weighted) are also an important midpoint stressor (i.e. environmental intervention = MPS, then
factor that should be unified for the sake of methods FF = 1). Both soil moisture and groundwater levels are appro-
compatibility. priate MPS, as both fall along the target impact pathway.
Another issue related to MPS selection relates to the tem-
Criterion 2.2 The midpoint stressor is selected correctly poral resolution. Environmental conditions vary throughout
All of the methods adequately describe the environmental the year. Arthington et al. (2006) emphasise the fact that river
mechanism, identify the MPS that affects the environment flow rates significantly fluctuate over time and that these fluc-
and propagate the impact of the MPS on the related endpoint tuations have important effects on ecosystems. The authors
indicator. For example, an increase in river water temperature conclude that a yearly average flow rate does not properly
(MPS) may affect fish biodiversity (Verones et al. 2010). capture this reality. Hanafiah et al. (2011), for example, only
Similarly, a decrease in water discharge into the river (MPS) consider the change in annual average river discharge flow,
can also affect fish populations (Hanafiah et al. 2011). whereas Amores et al. (2013) and Verones et al. (2010) ac-
Humbert and Maendly (2008) and Pfister et al. (2009) do not count for seasonal variations in their fate models.
explicitly identify the MPS in their equations as these methods
do not divide the characterisation model into FF and Recommendations Selecting the same MPS and units should
EF components. For Humbert and Maendly (2008), assumptions facilitate further harmonisation and comparison processes of
on how the MPS (change in water flowing through the dam) impact pathways that lead to the same MPS and of methods
affects fish species cannot be validated as these authors base their that address the same pathway. When multiple MPSs can be
model on empirical values and on a technological model, thus selected, we recommend selecting the more environmentally
disabling the quantification of the environmental mechanism that relevant indicator, which is the one positioned along the
relates the cause to the consequence. Pfister et al. (2009) did not causality chain which is closer to ecosystem damage.
model fate and thus assume that 1 m3 of water pumped from Regarding temporal resolutions of the MPS, future
rivers and aquifers deprives plants of 1 m3 of soil-water (MPS). developments should investigate the relevance of including
Definitions of the MPS and its position along the environmental condition variability as done by, e.g. Amores
causality chain may cause problems of compatibility et al. (2013) and Verones et al. (2010). As application of
among the FFs of different models. For example, methods in an LCA context may then quickly become data
Verones et al. (2012, 2013a) modelled changes in water intensive, the parsimony principle should not be forgotten. In
flowing into wetlands due to water consumption. some cases, it may be worth modelling this in detail, while in
However, FF meanings and metrics of both models cannot many othercases it may be sufficient to consider such variabil-
be directly compared: Verones et al. (2012) used the vol- ity through uncertainty estimates, depending on the degree of
ume of water infiltrating into the wetland as the MPS precision required for a given goal and scope.
(m3), whereas Verones et al. (2013a) used changes in wet-
land surface area (m2). The use of wetland volume reduc- 3.1.3 Issue 3fate factor modelling
tion (Verones et al. 2012) is comparable to the use of FFs
of other water-related impact pathways (Hanafiah et al. Criterion 3.1 The fate factor equation is in accordance
2011). However, assessing wetland area loss is compara- with underlying physical laws that govern the modelled
ble to land use-related impacts and may for instance be environmental intervention We show below the mathemat-
made compatible with the model addressing effects of ical derivation of the fate factor for a model that evaluates
dam construction on aquatic biodiversity. Although the effects of groundwater pumping. Similar FF derivations for
close relationship between land use and water use impacts surface water extraction and heat release to water are
has been largely discussed in LCA (Mil i Canals and de shown in the Supporting Information (Sect. S2). To com-
Baan 2015) and some suggestions have been made to pare these derivations with those of previous studies
harmonise land and water use impacts (Pfister et al. (Heijungs 1995), an FF for the pollutant release case is also
2011; Verones et al. 2015), so far, no agreement has been derived (Sect. S2). For each FF, derivation involves the
reached to render the assessment of both types of impacts following modelling steps:
fully compatible.
The same problem of MPS compatibility applies to the two 1. Steady-state modelling of the initial equilibrium state
methods that address potential impacts of water consumption 2. Transient modelling of the evolution toward a new equi-
on terrestrial vascular plant species. Van Zelm et al. (2011) librium state after the start of a continuous environmental
identified the change in groundwater volume as the MPS, intervention
while Pfister et al. (2009) used the change in soil moisture as 3. Transient modelling of the return to the initial equilibrium
a cause of plant damage. The latter authors assumed a linear state after a pulse environmental intervention
1808 Int J Life Cycle Assess (2016) 21:17991815

4. MPS selection and FF derivation dV


GRRGPRGDR 11
dt
In each case, the first three steps are schematically illustrat-
ed by a figure.
Where
Fate factor modelling for groundwater pumping
GDR k H  h 12
1. Steady-state modelling of the initial equilibrium state As V = V0 at t = 0, integrating Eq. (11) produces
(Fig. 2a)
A groundwater body is modelled very simply as a res- kH
hhstat h0 hstat e A t 13
ervoir with continuous and constant recharge (GRR) and
discharge (GDR). Initially, groundwater is pumped at a
constant rate (GPR0). The discharge is assumed to follow
Darcys law: kH
VV stat V 0 V stat e A t 14
GDR0 k H  h0 5
where hstat and Vstat are the new steady state, or equi-
where GDR0 is the initial volumetric discharge rate librium, height and volume toward which the system
(m3/s), h0 is the initial hydrostatic head (m), and kH will tend:
is a global hydraulic resistance (m2/s) that is supposed
GRRGPR0 GPR
to be constant. hstat 15
The initial ground water volume, V0 (m3) is kH

V 0 A  AG0 A  h0 h* 6
 
GRRGPR0 GPR
V stat A h* 16
kH
where h* is the elevation difference between the bot-
tom of the groundwater body and the free surface of 3. Transient modelling of the return to the initial equilibrium
the discharge water body (m). state after a pulse environmental intervention (Fig. 2c)
Assuming that water density variations are negligi- At t = 0, a given volume of water (Vcons) is instantaneous-
ble, a steady-state mass balance on the water body ly withdrawn from the ground water body. This tends to
becomes a volumetric balance: decrease the groundwater volume. A transient volumetric
balance gives
0 GRRGPR0 GDR0 7
dV
GRRGPR0 GDR 17
dt
Then
GRRGPR0
h0 8
kH dV kH
V 0 V 18
dt A

  As V = V0 at t = 0, integrating Eq. (18) produces


GRR GPR0
V0 A h* 9
kH kH
V V 0 V cons e A t 19
2. Transient modelling of the evolution toward a new equilib-
rium state after the start of a continuous environmental inter-
vention (Fig. 2b) 4. MPS selection and FF derivation
At t = 0, a continuous and constant additional pumping of As in van Zelm et al. (2011), the MPS is selected as the
water suddenly starts (GPR). groundwater volume deficit (V0 V).
Then Then, the time-integrated MPS for the pulse intervention,
, is
GPR GPR0 GPR 10
Z Z kH A
This tends to decrease the groundwater body volume (V). A V 0 V dt V cons e A t dt V cons 20
transient volumetric balance on the groundwater body gives: 0 0 kH
Int J Life Cycle Assess (2016) 21:17991815 1809

Groundwater
Fig. 2 Groundwater pumping recharge rate, V0
fate modelling a) Inial state GRR

Groundwater
pumping rate,
GPR0
h0
AG0
Groundwater
h*

Discharge
Groundwater water body
kH
discharge rate,
GDR0

b) Connuous water pumping c) Pulse water pumping


Groundwater Groundwater
recharge rate, recharge rate, Vcons
V
GRR GRR

Groundwater Groundwater
pumping rate, pumping rate,
GPR0 + GPR GPR0
AG0 AG

A
Discharge Discharge
Groundwater water body Groundwater water body
kH kH
discharge rate, discharge rate,
GDR GDR
Groundwater volume

Groundwater volume
V0 V0

Vcons
Vstat

t=0 Time t=0 Time

Then: in the MPS to the change in the environmental


intervention.
A The FF we derived for a simplified case of groundwater
FF 21
V cons k H pumping rate increase (see Eqs. (5) to (22) above) is the same
as the FF proposed by van Zelm et al. (2011), who use a
According to Eqs. (16) and (19), comprehensive hydrogeological model to compute steady-
state groundwater levels after a given change in groundwater
V 0 V stat pumping. It is, however, in disagreement with the FF of 1
FF 22 implicitly considered by Pfister et al. (2009) for water
GPR
pumping from aquifers.
Hanafiah et al. (2011) and Tendall et al. (2014) assumed
that 1 m3 per year of water consumed through human activi-
The continuous-type groundwater pumping case can then ties leads to a 1 m 3 reduction of available water
be used to calculate the pulse-oriented FF. downstream per year, i.e. FF = 1. The FF selected is consistent
Our FF derivations proved that there is a linear rela- with the FF that we derived for the change in water discharge
tionship between the steady-state environmental stressor rate due to water use (Sect. S2 of the ESM, Eqs. (S46)(S47)).
value and the rate of MPS change. Steady-state models However, this simplified approach does not account for any
and data may thus be used to find new system conditions links between different water compartments (e.g. surface and
after a pulse-type change. When comparing Eq. (2) and ground water).
our derived FFs to existing FFs, we found that none of the For thermal pollution, the FF derived in this study
models explicitly use Eq. (2) reflecting the finite nature of (Sect. S2, Eq. (S61)) is different from the FF proposed in
environmental intervention change and the integrative na- Verones et al. (2010). These authors use the same MPS (i.e.
ture of MPS change modelled in LCA. Furthermore, none numerator) but a different environmental intervention (i.e. de-
of the models justify the use of steady-state conditions nominator) than us. While for them the elementary flow is a
and data to model the fate factor of a pulse intervention. thermal discharge (in C m3 water discharge/day), in this
They rather define the FF as the ratio between the change study it is an amount of heat release (in J, see Sect. 3.1.1).
1810 Int J Life Cycle Assess (2016) 21:17991815

The latter can be computed from the cooling flow rate and an EF that effectively expresses a change in an impact on eco-
from surplus temperature release, as specified in Verones systems due to a change in a midpoint stressor (Eq. (3)).
et al. (2010), but it can also be directly determined from heat However, three methods have used a different midpoint stressor
evacuated from an industrial process. Our approach can be in fate and effect factors. Such a correction allows obtaining the
applied to different heat transfer processes (e.g. convection, desirable metric of the resulting indicator. These methods are as
radiation and conduction). We suggest following the latter follows: (1) Verones et al.s (2010) method, where the midpoint
method as it involves collecting more straightforward and stressor in the FF is the change in river water temperature
precise information and it represents a more relevant environ- accumulated over the river volume (C m3) while in the EF, it
mental intervention for assessing potential impacts of thermal is the temperature increase (C); (2) van Zelm et al.s (2011)
emissions. method, which uses an FF that ends in the numerator with the
Regarding dam-related impacts (Humbert and Maendly lowering in groundwater volume (m3) whereas the EF links,
2008), and as suggested in criterion 1.1 above, dam construc- over two steps, the groundwater table head (m) to soil moisture
tion can be modelled as a land transformation process. Then, reduction (ordinal scale of Ellenberg values) and the soil mois-
the FF should be equal to 1, i.e. 1 m2 of area newly flooded or ture reduction to plant species loss; and (3) Amores et al.s
dried out leads to environmental condition changes accross (2013) method, which connects the salt mass increase (kg) in
this 1 m2 surface. The FF for dam operation still needs to be the FF to the salt concentration (kg/m3) in the EF. In all three
developed. cases, discontinuities in units are accounted for by using a
constant volume (Verones et al. 2010; Amores et al. 2013b)
Recommendations Future LCIA methods should justify or area (van Zelm et al. 2011). However, this is not explicitly
the use of the steady-state solution for modelling the fate shown in the equations. The three discontinuities in the char-
factor. The applicability of Heijungs (1995) theorem for acterisation factors are shown in the covered impact pathways
water uses affecting multiple compartments, as modelled listed in the summary tables (Sect. 3.2).
in Verones et al. (2012, 2013a), Amores et al. (2013) and
Tendall et al. (2014) must be proven to legitimate the use Recommendations Future methods should ensure good and
of steady-state models and data under such conditions. transparent connections of meanings and metrics between fac-
tors that compose the characterisation factor.
Criterion 3.2 The fate factor considers the current state of
the environment Most existing methods account for back- Criterion 4.2 The taxonomic and geographic coverage is
ground levels when defining the initial value of the MPS, as complete as possible and the damage assessment is spa-
which is the level of the stressor prior to the environmental tially resolved The taxonomic coverage varies consider-
intervention. This is the case for methods presented in the able between methods: half rely only on one taxon (e.g.
following: Humbert and Maendly (2008), Verones et al. fish for assessing damage of dams in Humbert and
(2010, 2012, 2013a), van Zelm et al. (2011), Amores et al. Maendly (2008) and the introduction of exotic species in
(2013), and Hanafiah et al. (2013). Only the endpoint indica- Hanafiah et al. (2013)), whereas the other half use at least
tor developed by Pfister et al. (2009) and models where the FF two taxa (e.g. plants and birds in Verones et al. (2012) for
is explicitly set to 1(Hanafiah et al. 2011; Tendall et al. 2014) assessing biodiversity damage in wetlands). Datasets of
do not. taxa other than vascular plants, fish and birds are often
unavailable, explaining why methods are typically very
Recommendations Background information on the midpoint limited in their biological scope. Four of the methods
stressor level is needed to rigorously demonstrate that steady- use an approach based on species sensitivity distribution
state models and data can be used for fate modelling (see the functions to calculate the proportion of species that is
FF equations in Table S1, ESM). Therefore, whenever the affected or absent (Verones et al. 2010; van Zelm et al.
required information is available, we recommend including 2011; Amores et al. 2013; Hanafiah et al. 2013). This
background information on the stressor level prior to the in- approach is taken from eco-toxicity methods and requires
tervention, as this would allow developing sophisticated ap- the inclusion of at least three phyla to reflect physiology
proaches involving nonlinear ecosystem responses. variability and to ensure the representation of a minimum
diversity of biological responses (Henderson et al. 2011).
3.1.4 Issue 4effect factor modelling Only one of these methods (Hanafiah et al. 2013) does not
fulfil this requirement. Table 2 summarises the modelling
Criterion 4.1 The effect factor definition is coherent with approach and geographic coverage in existing models to
the characterisation factor and fate factor All of the assess impacts on ecosystem quality.
analysed methods with a fate and effect factor (except for Indicators for constructing EFs mostly reflect changes in
Humbert and Maendly (2008) and Pfister et al. (2009)) have biodiversity composition at species or community level.
Int J Life Cycle Assess (2016) 21:17991815 1811

Table 2 Effect modelling approach and geographic coverage of the analysed methods covering impacts of water use on ecosystems

Model Modelling approach Indicator Taxonomic coverage Geographic coverage Indicator


of EF or CF
Component Attribute

Humbert and Empirical observation Community Composition Aquatic species (fish) Global (nonalpine dams PDF
Maendly (2008) of fraction of species vs alpine dams)
that had disappeared
before and after dam
construction
Pfister et al. (2009) Environmental modelling Ecosystem Function w-NPP used as a proxy for Global PDF
based on GIS: NPP the richness of terrestrial
limited by water vascular plants
availability (w-NPP)
Verones et al. SSD species to Species Composition 36 aquatic species of The Rhine River and its PDF
(2010) temperature temperate regions. Five Aare tributary to
phyla (Annelida, the Sea
Arthropoda, Cnidaria,
Mollusca, and Chordata)
Hanafiah et al. Species-discharge Community Composition Aquatic species (fishes) 214 watersheds located PDF
(2011) relationship at a latitude of 42
van Zelm et al. SSD based on the Species Composition 625 terrestrial vascular plants The Netherlands PNOF PDF
(2011) absence/presence of of Central Europe.
plant species as a Four phyla (Magnoliophyta,
function of Coniferophyta,
groundwater levels Pteridophyta,
and Lycopodiophyta)
Verones et al. Species-area relationship Community Composition Aquatic and terrestrial Santa Rosa wetland, PDF
(2012) species (plants and birds) Peru
Verones et al. Species-area relationship Community Composition Aquatic and terrestrial Global (10 % of SE
(2013b) and and species (waterbirds, wetlands covered.
ecosystem structure nonresidential birds, Data gaps due to a
amphibians, water- lack of data or an
dependent mammals, absence of wetlands)
and reptiles)
Amores et al. SSD species to salinity Species Composition 18 aquatic species. Five Wetland Albufera de PDF
(2013) phyla (Magnoliophyta, Adra, Spain
Tracheophyta, Arthropoda,
Spermatophyte, and
Chordata)
Hanafiah et al. SSD native species Species Composition Aquatic species (fish) in Rhine-Main-Danube PDF
(2013) variation to exotic European river basins river
species introduction One phylum: Chordata
Tendall et al. Species-discharge Community Composition Aquatic species (fish and Europe GSE
(2014) relationship and and macroinvertebrates)
ecosystem structure

GIS geographic information system, GSE global species extinction equivalents (weighted by vulnerability), NPP net primary productivity, PAF
potentially affected fraction of species, PDF potentially disappeared fraction of species, SE global species equivalents lost, SSD species sensitivity
distribution

Only two methods (Verones et al. 2013b; Tendall et al. purposes. The use of the same number and type of taxa to
2014) assess changes in communitiesy and ecosystems, be- assess effects on species should enhance consistency and
ing more relevant to evaluations of ecosystem quality comparability between methods. For the sake of complete-
damage. ness, future methods may continue investigating ways to con-
All of the methods are spatially resolved. However, geo- sider ecosystem and genetic levels of biodiversity.
graphic coverage remains incomplete for most methods.
Criterion 4.3 Sub-steps of the effect mechanism are explic-
Recommendations Providing global coverage for methods itly addressed in the effect modelling None of the models
that are only available for specific geographic areas, water- explicitly decompose the EF into one or several sub-factors.
sheds, or countries is an evident requirement for LCA Furthermore, none of the models explicitly address exposure,
1812 Int J Life Cycle Assess (2016) 21:17991815

which is implicitly assumed to be equal to a value of 1. This latter provided four linearised ecosystem responses in re-
choice should have been justified. Incidence and damage lation to groundwater levels. Largely due to a lack of data,
modelling approaches used differ from one study to another. species/communities or background levels considered in
Incidence factors used in Verones et al. (2010) and other methods are site generic, and effects are modelled
Amores et al. (2013) were derived from SSD curves for following a linear approach. For example, Hanafiah et al.
sets of exposure-response relationships for several spe- (2013) derived a generic EF for Europe and Verones et al.
cies of various taxa. Verones et al. (2010) based their (2010) derived EFs applicable to temperate regions.
PAF on one indicator: median lethal temperatures Amores et al. (2013) calculated a linear, background level
(LT50 for aquatic species) of a population. Amores independent effect at a 50 % hazardous salt concentration.
et al. (2013) based their PAF computations on the effec- These authors could have developed a more sophisticated
tive concentration affecting 50 % of the population nonlinear approach for the specific lagoon that they ex-
(EC50 for aquatic species) for several different effect amined from the high quality site-specific information that
indicators (death, growth, height, etc.) due to an increase they used.
in wetland salt concentrations. SSDs should be generated
using data for the same type and level of effect. Recommendations We recommend including site-specific
However, as data on one type of biological endpoint information on the state of the environment before the
(i.e., mobility, proliferation, growth, death, etc.) affecting midpoint stressor changes the environmental conditions
multiple taxa are often limited, in LCA, it is accepted of the affected water compartment. This would allow the
practice to use SSD functions for multiple biological development of sophisticated approaches with nonlinear
endpoints at the same time. ecosystem responses.
The severity of damage is addressed through two ap-
proaches, and neither of these are explicitly expressed in the 3.2 Summary tables of existing models
CF equation:
Table 3 shows a summary table for one of the assessed
1. The SSD-based PAF of the incidence factor is equiv- methods, Amores et al.s (201) method as it includes many
alent to the number of species lost (i.e. 1PAF = 1PDF). of the elements discussed in this study. Tables on the other
This is justified in Verones et al. (2010), as PAF is methods analysed in this work are collated in Sect. S3 of the
based on mortality data, while it is a simplification in ESM.
Amores et al. (2013) to compare salinity damages with
other commonly used ecosystem quality impact catego-
ries (e.g. eutrophication, ecotoxicity). 4 Conclusions
2. The inventory or MPS (e.g. reduction in water volume,
increase in exotic species) proportionally influences the end- We found that the models analysed have the potential to com-
point indicator PDF m3 year without previously accounting plement one another. However, we also identified sources of
for the exposed and affected species (Humbert and Maendly incompatibility and inconsistency and scientific drawbacks.
2008; Pfister et al. 2009). We therefore do not recommend aggregating existing indica-
tors into one single metric because thus far methods do not
apply the same guidelines that
Recommendations Subdivisions of the EF in several
sub-factors will make intermediate comparisons and the Correctly adopt basic principles of life cycle assessment
integration of methods easier to conduct. Further research on standards (e.g. sound definitions of environmental inter-
the conversion from affected species to disappeared species is ventions in the inventory step and their correct connec-
recommended to enhance the compatibility of endpoint tions to the impact assessment phase). As discussed, a
methods. joint development of inventory datasets and characterisa-
tion models is required to ensure coherence between these
Criterion 4.4 The effect factor considers the current state two life cycle assessment steps
of the environment Most models consider the state of the Standardise terminology, modelling and metrics of the
ecosystems prior to the environmental intervention by causality chain sections that are common to different
modelling the EF via a marginal approach. This is the characterisation models to facilitate cross comparisons
case for methods based on species-area (Verones et al. between results of intermediate sub-factors and final
2012; Verones et al. 2013b) and species-water flow rela- impact scores
tionships (Hanafiah et al. 2011; Tendall et al. 2014) and Include modelling of water compartments that have been
for the method presented by Van Zelm et al. (2011). The missing up to now (e.g. atmosphere) as well as links
Int J Life Cycle Assess (2016) 21:17991815 1813

Table 3 Summary table of Amores et al.s method according to the aspects listed in Sect. 2.2
LCIA method (Amores et al. 2013)
Covered impact
pathway1

Potential
Increase in affected
groundwater wetland
consumption aquatic
species

Subfactor 2

Structure of the Native spatial scale: None


EI
characterization Aggregated spatial scale: wetland
model2 lagoon
R S Receiving compartment, R:
groundwater
Stressed compartments, S: wetland
lagoon
D

Groundwater Wetland
Fate factor at Marginal change in the salt content of the wetland due to a marginal change in groundwater
aggregated consumption by agriculture:
spatial scale msalt, wetland kgsalt, wetland
FF= [=] 3
Qconsumed mconsumed
yr

Effect factor at Marginal change in the potentially affected fraction (PAF) of species due to a marginal change in
aggregated the salinity of the wetland :
spatial scale PAFwetland PAFwetland
EF= [=]
Csalt, wetland kgsalt /m3
wetland

Characterization Marginal change in PAF of species due to a marginal change in groundwater consumption in
factor at agriculture. The conversion 1 PAF = 1 PDF is applied:
aggregated
spatial scale PAFwetland (PAFm3 yr) wetland (PDFm3 yr) wetland
CF=FFEF= [=] =
Qconsumed m3consumed m3consumed

Source: Amores et al. (2013)


a
The diagonally hatched box indicates the impact pathway starting point, and the dotted box indicates the impact pathway final point. Dashed border
boxes indicate steps of the cause-effect chain that are cited as explanations/justifications in the model documentation but are not specifically modelled in
equations
b
The symbols for the modelled system structure schemes are as follows: EI, environmental intervention; R, receiving compartment, i.e. compartment
where the environmental intervention takes place; S, stressed compartment, i.e. compartment where the midpoint stressor is modelled; D, damage on
ecosystems; , propagation of an environmental intervention from one compartment to another
1814 Int J Life Cycle Assess (2016) 21:17991815

between them and other environmental compartments Heijungs R (1995) Harmonization of methods for impact assessment.
Environ Sci Pollut Res Int 2:217224
(e.g. soilatmosphere). This will improve the relevance
Henderson AD, Hauschild MZ, Van De Meent D et al (2011) USEtox fate
of the characterisation model with respect to the global and ecotoxicity factors for comparative assessment of toxic emis-
hydrological cycle sions in life cycle analysis: sensitivity to key chemical properties. Int
J Life Cycle Assess 16:701709
Huijbregts MAJ, Hellweg S, Hertwich E (2011) Do we need a paradigm shift
The development and adoption of these guidelines should
in life cycle impact assessment? Environ Sci Technol 45:38333834
help harmonise current and future water use impact assess- Humbert S (2015) Chapter 6: particulate matter formation. In Hauschild MZ,
ment methods for ecosystem quality. Huijbregts MAJ (ed) LCA CompendiumThe Complete World of
Life Cycle Assessment. Life Cycle Impact Assessment. Springer
Humbert S, Maendly R (2008) Characterization factors for damage to
Acknowledgments We thank Jean-Daniel Savard for his work, which
aquatic biodiversity caused by water use especially for dams used
inspired the schematic representation of the characterisation model. His
for hydropower
work constituted a first attempt at analysing several models that were re-
IPCC (2007) Climate Change 2007. The physical science basis.
analysed in the present study. We gratefully acknowledge the authors of
Contribution of Working Group I to the Fourth Assessment Report
the analysed methods for their valuable input. M. Nez acknowledges
of the Intergovernmental Panel on Climate Change. United
ANR, the Languedoc-Roussillon Region, ONEMA and its industrial part-
Kingdom and New York
ners (BRL, SCP, SUEZ, VINADEIS) for the financial support of the
ISO (2006a) ISO 14040. Environmental managementlife cycle assess-
Industrial Chair for Environmental and Social Sustainability
mentprinciples and framework
Assessment BELSA-PACT^ (grant no. 13-CHIN-0005-01).
ISO (2006b) ISO 14044. Environmental managementlife cycle assess-
mentrequirements and guidelines
Jolliet O, Margni M, Charles R et al (2003) Presenting a new method
References IMPACT 2002+: a new life cycle impact assessment methodology.
Int J Life Cycle Assess 8:324330
JRC (2010) ILCD Handbook: framework and requirements for LCIA
Amores MJ, Verones F, Raptis C et al (2013) Biodiversity impacts from models and indicators First edition
salinity increase in a coastal wetland. Environ Sci Technol 47: Kounina A, Margni M, Bayart J-B et al (2013) Review of methods ad-
63846392 dressing freshwater use in life cycle inventory and impact assess-
Arthington AH, Bunn SE, LeRoy PN, Naiman RJ (2006) The challenge ment. Int J Life Cycle Assess 18:707721
of providing environmental flow rules to sustain river ecosystems. Margni M, Gloria T, Bare J et al (2008) Guidance on how to move from
Ecol Appl 16:13111318 current practice to recommended practice in Life Cycle Impact
Bayart J-B, Bulle C, Deschnes L et al (2010) A framework for assessing Assessment
off-stream freshwater use in LCA. Int J Life Cycle Assess 15: MEA (2005) Ecosystems and human well-being. Synthesis
439453 Mil i Canals L, de Baan L (2015) Chapter 11: land use. In Hauschild M Z
Boulay A-M, Motoshita M, Pfister S et al (2015) Analysis of water use and Huijbregts MAJ (ed.) LCA CompendiumThe Complete World
impact assessment methods (part A): evaluation of modeling of Life Cycle Assessment. Life Cycle Impact Assessment. Springer
choices based on a quantitative comparison of scarcity and human Mil i Canals L, Chenoweth J, Chapagain A et al (2009) Assessing
health indicators. Int J Life Cycle Assess 20:139160 freshwater use impacts in LCA: part Iinventory modelling and
Cao V, Margni M, Favis BD, Deschnes L (2015) Aggregated indicator to characterisation factors for the main impact pathways. Int J Life
assess land use impacts in LCA based on the economic value of Cycle Assess 14:2842
ecosystem services. J Clean Prod 94:5666 Nez M, Pfister S, Roux P, Antn A (2013) Estimating water consump-
Curran M, de Baan L, De Schryver AM et al (2011) Toward meaningful tion of potential natural vegetation on global dry lands: building an
end points of biodiversity in life cycle assessment. Environ Sci LCA framework for green water flows. Environ Sci Technol 47:
Technol 45:7079 1225812265
Delbaere B, Serradilla AN, Snethlage M (2009) BioScore: a tool to assess Pennington DW, Payet J, Hauschild MH (2004) Aquatic ecotoxicological
the impacts of European Community policies on Europes indicators in life-cycle assessment. Environ Toxicol Chem 23:
biodiversity 17961807
Pennington DW, Margni M, Payet J, Jolliet O (2006) Risk and regulatory
Goedkoop M, Spriensma R (2001) The Eco-indicator 99. A damage
hazard-based toxicological effect indicators in life-cycle assessment
oriented method for Life Cycle Impact Assessment. Methodol Rep
(LCA). Hum Ecol Risk Assess An Int J 12:450475
Guine JB, Marieke G, Heijungs R et al (2001) Life cycle assessment. An Pfister S, Suh S (2015) Environmental impacts of thermal emissions to
operational guide to the ISO standards. Part 2A freshwater: spatially explicit fate and effect modeling for life cycle
Hanafiah MM, Xenopoulos MA, Pfister S et al (2011) Characterization assessment and water footprinting. Int J Life Cycle Assess 20:927936
factors for water consumption and greenhouse gas emissions based Pfister S, Koehler A, Hellweg S (2009) Assessing the environmental
on freshwater fish species extinction. Environ Sci Technol 45: impacts of freshwater consumption in LCA. Environ Sci Technol
52725278 43:4098104
Hanafiah MH, Leuven RSEW, Sommerwerk N et al (2013) Including the Pfister S, Bayer P, Koehler A, Hellweg S (2011) Environmental impacts
introduction of exotic species in life cycle impact assessment: the of water use in global crop production: hotspots and trade-offs with
case of inland shipping. Environ Sci Technol 47:1393413940 land use. Environ Sci Technol 45:57615768
Hauschild M, Huijbregts MAJ (2015) LCA CompendiumThe Quinteiro DAC, Arajo A et al (2015) Suspended solids in freshwater
Complete World of Life Cycle Assessment. Life Cycle Impact systems: characterisation model describing potential impacts on
Assess Springer. doi:10.1007/BF02978760 aquatic biota. Int J Life Cycle 20:12321242
Hauschild MZ, Goedkoop M, Guine J et al (2013) Identifying best Rosenbaum RK (2015) Chapter 8: ecotoxicity. In Hauschild M Z and
existing practice for characterization modeling in life cycle impact Huijbregts MAJ (ed.) LCA CompendiumThe Complete World
assessment. Int J Life Cycle Assess 18:683697 of Life Cycle Assessment. Life Cycle Impact Assessment. Springer
Int J Life Cycle Assess (2016) 21:17991815 1815

Roy P, Desche L, Margni M (2012) Life cycle impact assessment of Verones F, Hanafiah MM, Pfister S et al (2010) Characterization factors
terrestrial acidification: modeling spatially explicit soil sensitivity for thermal pollution in freshwater aquatic environments. Environ
at the global scale. Environ Sci Technol 46:82708278 Sci Technol 44:93649369
Saad R, Koellner T, Margni M (2013) Land use impacts on freshwater Verones F, Bartl K, Pfister S et al (2012) Modeling the local biodiversity
regulation, erosion regulation, and water purification : a spatial ap- impacts of agricultural water use: case study of a wetland in the
proach for a global scale level. Int J Life Cycle Assess 18: coastal arid area of Peru. Environ Sci Technol 46:49664974
12531264 Verones F, Pfister, Hellweg S (2013a) Quantifying area changes of inter-
Snell T, Serra M (2000) Using probability of extinction to evaluate the nationally important wetlands due to water consumption in LCA.
ecological significance of toxicant effects. Environ Toxicol Chem Environ Sci Technol 47:97999807
19:23572363 Verones F, Saner D, Pfister S et al (2013b) Effects of consumptive water
Struijs J, De Zwart D, Posthuma L et al (2011) Field sensitivity distribu- use on biodiversity in wetlands of international importance. Environ
tion of macroinvertebrates for phosphorus in inland waters. Integr Sci Technol 47:1224812257
Environ Assess Manag 7:280286 Verones F, Huijbregts MAJ, Chaudhary A et al (2015) Harmonizing the
Tendall DM, Hellweg S, Pfister S et al (2014) Impacts of river water assessment of biodiversity effects from land and water use within
consumption on aquatic biodiversity in life cycle assessmenta LCA. Environ Sci Technol 49:35843592
proposed method, and a case study for Europe. Environ Sci Weidema B (2001) Physical impacts of land use in product life cycle
Technol 48:32363244 assessment. Final report of the EURENVIRONLCAGAPS sub-
Udo de Haes H, Finnveden G, Goedkoop M et al (2002) Life-cycle project on land use
impact assessment: striving towards best practice. Society of Weidema B, Bauer C, Hischier R et al (2013) Overview and methodolo-
Environmental Toxicology and Chemistry (SETAC) gy. Data quality guideline for the ecoinvent database version 3.
van Zelm R, Schipper AM, Rombouts M et al (2011) Implementing Ecoinvent report 1 (v3). St. Gallen
groundwater extraction in life cycle impact assessment: characteri- Yano S, Hanasaki N, Itsubo N, Oki T (2015) Water scarcity footprints by
zation factors based on plant species richness for The Netherlands. considering the differences in water sources. Sustainability 7:
Environ Sci Technol 45:629635 97539772

Vous aimerez peut-être aussi