1 Introduction

The third phase of the UNEP-SETAC Life Cycle Initiative (http://www.lifecycleinitiative.org/activities/phase-iii/) has launched a flagship project to provide global guidance and build consensus on environmental life cycle impact assessment (LCIA) indicators. This global process is starting with a limited number of life cycle impact category indicators developed within a consistent framework. At the May 2013 international workshop of the Life Cycle Initiative in Glasgow (Jolliet et al. 2014), water use was selected as one of the four initial topics due to its high relevance and the lack of established consensus in the recently developed methods. Indeed, assessment models on water use and related indicators have been shaping only over the last years, and the WULCA working group (Water Use in LCA) received the mandate to lead the harmonization and consensus building for water consumption impact assessment in LCA, tackling the challenge to build consensus on a rapidly evolving science. It should be noted that water quality indicators are excluded from this work as they are addressed in other groups.

The WULCA working group’s overall goal since 2007 focuses on providing practitioners, from industry, governments, NGOs, and academia, with a consensus-based and consistent framework to assess, compare, and disclose the environmental performance of products and operations regarding freshwater use. The main scientific deliverables so far have been a framework for water use impact assessment in LCA (Bayart et al. 2010), a review of existing methods for water use in LCA (Kounina et al. 2013), and two papers from a quantitative method comparison and application (Boulay et al. 2015a, b). Additionally, numerous disseminating activities, including scientific seminars and trainings ensured broad stakeholder involvement. The outcomes of these activities serve as building blocks for the development of a consensus-based method.

The WULCA group commenced the development of the consensus-based indicator(s) in January 2014. This activity is planned to last 2 years. Building on the framework developed by (Bayart et al. 2010; Kounina et al. 2013), this first step led to the identification of three indicators on which to focus (see Fig. 1): (1) the impact pathway leading to damages on human health, which is already modeled by different methods and ready for harmonization (Boulay et al. 2015a) and hence included in the work, and a consensus-based method defining this impact pathway is under development; (2) the ecosystem impact pathway, which includes several methodologies with possible complementary assessments (Kounina et al. 2013) and therefore requires development of a consistent framework and identification of a common midpoint indicator early in the impact pathway; and (3) a water stress or scarcity indicator, which answers the demand from industry and policy makers for a generic (not human- or ecosystem-oriented) and recommended midpoint indicator. This latter indicator is not explicitly located on the cause-effect relationship from the environmental intervention to endpoint damages. While being independent from other impact pathways and not directly linked to any endpoint damages, it encompasses potential deprivation of both ecosystem and human, hence aiming to represent potential impacts more comprehensively than any other available LCA-oriented method assessing the “water scarcity footprint” (ISO 2014).

Fig. 1
figure 1

Framework of impact pathways for water use in LCA. The three indicators in blue are the focus of this project. The impact pathways printed in faded colors are not mature enough to be included. Only the stress-based generic midpoint was discussed in the workshops

A series of three expert workshops, including non-LCA experts from hydrology, eco-hydrology, and water supply science, was organized specifically on the topic of this generic midpoint indicator. They were held in Zurich on 10th September, in San Francisco on 5th October, and in Tsukuba on 27th October 2014. In total, 49 experts attended. The specific objectives of the workshops were twofold. First, it was to present the identified options of the stress-based indicator narrowed down by the active members of WULCA during the first 8 months of the project and to receive comments on the relevance, usefulness, acceptability, and focus of the selected indicator. Second, the workshop covered different challenges in the modeling of the indicator and presented the experts with background information and specific questions. This paper summarizes the discussions and outcome of these workshops. Where no agreement was reached, the working group of active members is considering all inputs received and continues the work.

2 Scope and format of the indicator

The following three questions and associated indicators were previously identified by the active members in order to better map the potential metrics and their purposes. These were used to clarify the scope and meaning of the indicator.

  1. I.

    Questions: “To which extent are humans using the available water in this region?” and “What is the potential of affecting water availability for human uses?”

    This question can be answered with the traditional indicator type based on withdrawal-to-availability (WTA) ratios or consumption-to-availability (CTA) ratios, as described in Boulay et al. (2015a).

  2. II.

    Question: “What is the potential of depriving another user from water, with no specification of who the user is (i.e. humans or ecosystems)?”

    In this case, depriving refers to using water that is being used by any user, leading to active adaption and/or suffering of consequences. This question can be answered by an indicator that includes human and ecosystems water demand with respect to availability, which is defined as demand-to-availability (DTA). The definition of demands is the critical point in characterizing human and ecosystem needs.

  3. III.

    Questions: “How much water is available in this region?”, “How dry is the region?”, and “How critical is the water use assessed in comparison to the water available in this area?”

    These questions can be typically answered by an aridity index which quantifies the absolute availability of water in a region, typically relating available water with area or potential evaporation.

Experts in Zurich were confronted with these options and the recommendation from the active member group that question II was best suited for the generic indicator being developed by this working group. The experts in Zurich agreed that the indicator to be developed should answer question II above and possibly include aspects of question III. Discussions arose on which indicator is the best to answer that question. Most agree that DTA is a good option, but two concerns were identified: (1) the special case of arid areas (question III) and (2) the definition and inclusion of ecosystem water demand. These are further described below.

  1. 1.

    Arid areas

    One of the main concerns expressed with respect to most existing indicators, and potentially a DTA-based indicator as well, is the special case of arid regions. An arid region was defined by UNEP for example (1997) as a region where the potential evapotranspiration is larger than five times the water received from precipitation. The concern with arid region is that indicator values in such areas (e.g., Morocco) can be lower than some other regions which have more water but more activity (e.g., Belgium). This outcome is explained by the fact that water use in arid areas is generally low and, even if availability is low as well, CTA, WTA, or DTA can be lower in some arid areas than in other more water-abundant regions. This would not be a problem in itself, if the result reflects the answer to the question that is asked regarding the potential to deprive other users; however, it does potentially bring disagreement for two reasons. First, some believe that arid areas should show significant water stress independently of the potential to cause damage to other users, with the reasoning that such areas should not appear to be more favorable to water use than a water abundant region, even if the latter shows more users competing, since the lower water usage is probably already influenced by the arid conditions. Second, while in LCA the assessment is assumed to be for a marginal environmental contribution which does not affect the background of the environmental state (assuming a ceteris paribus condition), a water use in an arid region may in fact not respect this hypothesis. Indeed, a supply chain situated in dry area might use a high percentage of the available water and hence have a high influence. For these reasons and for the fact that hydrological experts emphasized that data quality on water availability is much lower in these regions, the conclusion was that arid areas correspond to special cases that should be integrated with special attention in the model. Four approaches were discussed that can allow this:

    1. a.

      Combining information on arid areas into the DTA indicator, similarly to Berger et al. (2014) by including a conditional modeling aspect that applies for arid areas

    2. b.

      Using a different indicator format: 1 / (availability − human and ecosystem demand)

    3. c.

      Using a combination of two parts, one that assesses the extent to which the resource is being used (i.e., demand/availability, as an indicator of the “pressure”), and one that assesses the general state of aridity (i.e., area/availability as an indicator of “absolute availability”), and have them combined in a function (multiplication or other)

    4. d.

      Leaving arid areas out of the scope of the method because of too high uncertainty and questionable interpretation. However, this was found to be impractical for systematic LCA application

  2. 2.

    Ecosystem water demand

    The trade-off between uncertainty associated with including environmental water requirement (EWR) (as quantitative information on the relation between streamflow regimes and biodiversity and health of the river ecosystems is scarce), or lower relevance associated with excluding it, was discussed thoroughly. Although most participants accepted to include EWR, no complete agreement was found on this and some members/experts believe exclusion may be a better choice as CTA or WTA could be considered a “proxy” (albeit an unspecific one) for ecosystem impacts.

3 Modeling challenges

In addition to the general discussions on the goal of the indicator and different formats it could take, specific modeling questions were identified and discussed. The section below summarizes these discussions.

3.1 Defining and quantifying human and ecosystem water demand

Human and ecosystem “demand” or “needs” should be defined in the same way in order to minimize biases. When asked if demand or needs should differentiate essential versus less essential water, experts recommended to stay away from “essential” versus “useful” categorization of human water use as too much value judgment and uncertainty would be present. Assessment of actual water use is preferred.

For ecosystem water demand, experts presented different methodologies on aquatic ecosystem water requirement assessment. Recommendation was made to use EWR median and maximum of different methods or a range of 35 to 80 % of pristine (natural) conditions to account for the uncertainty and account for the temporal variability addressing low and high flow situations of actual (current) water requirements. Discussion occurred on the question whether the environment can always use all water that is there, such that environmental demand would be all available water. Choosing the current ecosystem demand would therefore carry the hypothesis that the current state is the state of reference. While no ideal state of reference could be identified, this one was chosen for its consistency with the current state of reference of human demand.

3.2 Green water/terrestrial ecosystems

Water demand of terrestrial ecosystems cannot, for consistency reasons, be included without including green water availability (i.e., precipitation water that has entered the soil), so this topic was addressed as one. The group of active members presented the following position to the group of experts, which was accepted in all three workshops:

Terrestrial ecosystems primarily consume green water from soil (green water) and, through capillary rise, also from groundwater (blue water) whose mechanisms of consumption are not yet well understood. It is unclear at this point how including green water demand and availability may change the meaning of the indicator as it is meant for assessment of blue water consumption (i.e., water extracted from rivers, reservoirs, lakes, and aquifers). While a separate green water scarcity indicator could eventually be developed to assess green water consumption (as attempted for natural and agricultural ecosystems; Rockström et al. 2009, Gerten et al. 2013), we believe it is too soon at this point in the WULCA context and not the main focus on the current mandate. Given the current LCIA models, the main impacts from blue water consumption on terrestrial ecosystems are described by the model of Van Zelm et al. (2011) for the Netherlands, which pertains to groundwater withdrawals only, and could be considered representative of selected areas in the world where the water table is shallow and well connected with soil water. Apart from the model of Van Zelm et al. (2011), we agree that there are probably other impact pathways linking blue water consumption to terrestrial ecosystem impacts, e.g., through reduced flooding. For example, impacts may occur as a result of surface water consumption which can deplete groundwater stocks and, as a corollary, affect terrestrial ecosystems, or groundwater consumption and depletion may affect groundwater dependent ecosystems far away from the location of consumption (e.g., Verones et al. 2013). However, these impact pathways are not well known and according to Gerten et al. (2013), terrestrial ecosystems mostly rely on green water which itself depends on precipitation, soil physics, climate, and land use, rather than human consumption of blue water. For these reasons, it was decided to exclude green water and hence terrestrial ecosystem water requirements from this indicator.

3.3 Sources of data

Many methods exist to quantify water availability and human water use, and they can lead to non-negligible differences in water scarcity indicators (Boulay et al. 2015a). During the workshops, experts presented details on the global hydrological models and data sources: WaterGAP (Müller Schmied et al. 2014), LPJmL (Rost et al. 2008), Aqueduct 2.0 (Gassert et al. 2013), and H08 (Hanasaki et al. 2010). Differences in models originate namely from different hypothesis on evapotranspiration (and its individual components), the application and nature of calibration procedures, and different climatic data. It was noted that WaterGAP is the only water availability model of this suite that is calibrated to actual river discharge measurements and hence represents the current reality better, even if water balances are broken by this procedure, whereas the agricultural water use model of LPJmL is more comprehensive and may be more accurate in representing the coupled water-vegetation-soil dynamics. The option of using the mean of different models is discussed, but ultimately, no final decision was taken and the group of active members is considering all the received information before making their choice. A specific concern was raised regarding the desire to include infrastructure in water availability assessment (reservoirs, water transfers, etc) as well as to use datasets that will facilitate regular updates of the indicator.

For ecosystem water requirements, monthly models are presented and discussed by Pastor et al. (2013) and Hanasaki et al. (2008a, b).

3.4 Groundwater vs. surface water

Given adequate technology, most human users can in principle use surface or groundwater based on availability and infrastructure. Above-ground aquatic ecosystems (rivers, wetlands, deltas) rely on surface water, whereas terrestrial ecosystems may be affected by an overuse of groundwater as well as potentially surface water, but as mentioned above, this concept was decided as not sufficiently mature to be included. While in some cases it may be desirable to look at the stress for these resources separately, in general, this indicator aims to assess the overall pressure on the water resource. General agreement was found on not distinguishing surface and groundwater and providing only one generic indicator. Specifically, it was mentioned that providing separate indicators in addition to one aggregated indicator for both sources could lead to contradicting results. Consistency would be hard to reach as availability and reliability of inventory data regarding specific source of water is low, comparability would be lost and advice to the practitioner on which one to use and why would be arguable. Moreover, the relative importance of one source of water over another is often hard to justify and the connection between groundwater, surface water, and ecosystems is not modeled robustly enough at this point. For these reasons, the group agreed to focus on one indicator that does not distinguish different sources of water.

3.5 Temporal resolution

In LCA, water inventory data is most often available with no specific information on the time of the year the water was consumed. For this reason, it is clear that the provided indicator will be available as an annual metric. However, seasonality is a very important concept when it comes to water use and availability and constructing an indicator directly from annual data can result in significant differences as opposed to an aggregation of monthly indicators (Boulay et al. 2015b). Although it was recognized that monthly data on water use and availability are more uncertain than annual means, mean annual water availability would lead to an overestimation of actual water availability in semi-arid/arid areas and monsoonal areas as compared to others. The role of infrastructure in mitigating intra-annual water availability fluctuation is important, and although data availability may be a problem, it should be kept in mind for a monthly indicator. Other suggestions to account for temporal variations included using the statistical low-flow Q90 instead of total water availability or using an indicator which only represents the driest month. The group agreed to model monthly indicators to be used for those LCA practitioners who have access to temporal data related to water use and average the monthly values to obtain an annual one using a weighted average, to account for less-informed studies.

3.6 Geographical resolution

Similarly to the temporal resolution, the geographical resolution of the proposed indicator should match the resolution of available inventory data. At this time, only a small set of LCAs analyze water consumption site specifically; however, it is important that a methodology be developed which can allow differentiation at the sub-basin level. Again, recognizing the inconsistency in inventory data availability, these sub-basin level indicators could be aggregated to basin-level or country-level indicators. It was acknowledged that at the sub-basins scale, anthropogenic uses and flow disturbances should be considered, since otherwise, downstream sub-basins without consumption would show no stress even though the natural system is already disturbed by upstream consumption, as shown in Loubet et al. (2013). It was therefore recommended that the scale is chosen large enough that downstream/upstream effects are not so relevant anymore, but this was not further discussed. Consensus was found however on the aggregation of the indicator to the country level using consumption-based weighted averages, in order to represent the geographic probability distribution of the water use within the selected country.

4 Work process and next steps

The input, decisions, and points of discussion were compiled and brought back within the group of active members. The group is using the recommendations and works further on the harmonization of the points of disagreement. It is expected that a selection of indicators representing different ways to address the most important issues will be produced and tested in spring 2015. The analysis of the result should lead to a provisional recommendation by summer 2015.