Review of methods addressing freshwater use in life cycle inventory and impact assessment

The way human health is affected by freshwater use depends on the level of economic development and welfare (Boulay et al. 2011b; Bayart 2008). If this is sufficient, the lack of freshwater will be compensated by the development of backup technologies [such as desalination or the import of water-intensive goods as virtual water (Allan 1996)]. These compensation activities need to be assessed with a new inventory and can, in turn, lead to environmental impacts via other interventions involved in the compensation activities (e.g., climate change impacts caused by energy consumption for desalination). If the level of economic development is not sufficient to cover these costs, freshwater use will lead to water deprivation for primarily three functions which fulfill essential human needs depending on local conditions: domestic use (hygiene and ingestion), agriculture, and aquaculture/fisheries. Industrial functions of freshwater closed to human essential needs (e.g., house building and provision of pharmaceuticals) are not considered in this framework because they are more likely to consider compensation strategies rather than suffering from freshwater deprivation (Boulay et al. 2011b). Water quality degradation leads to water deprivation when it creates a loss of functionality for users who need water at a higher quality level than the released one. Users who are able to use freshwater at that or a lower quality level would not be deprived. The extent of water quality degradation depends on the amount and intensity of chemical, biological, and thermal pollution withdrawn and is related to the sanitation capacity. The withdrawn freshwater represents an adverse impact depriving users from a given amount of water at ambient water quality; the released freshwater (negative LCI flow) results in a burden reduction by making available the same amount of water for users capable to use water at that quality. Current models agree that the way human health is affected by water use depends on the level of economic development and welfare. They acknowledge that under given conditions, water use can lead to deprivation for essential human needs such as agriculture, fisheries, and domestic use and ultimately to malnutrition and spread of diseases. However, there is currently no sufficient information to determine whether freshwater use in a low-income water-stressed region would lower water availability for domestic users or rather only affect other users, e.g., agricultural, fisheries, or industries (Boulay et al. 2011b).

Malnutrition and spread of diseases are interconnected, i.e., malnutrition could, for example, make a person more vulnerable to the spread of diseases, and reciprocally, some enteric diseases could affect the ability to absorb nutrients and thus contribute to malnutrition. Freshwater use ultimately leads to an aggregated impact on human health, generally expressed in disability-adjusted life years (Motoshita et al. 2010a, b; Boulay et al. 2011b; Pfister et al. 2009).

Water use can also affect the ecosystem, for instance, by changes in the river, lake, or wetland flow quantity (e.g., due to surface water withdrawals); changes in the level of the groundwater table (e.g., due to groundwater withdrawal); changes in flow regimes (e.g., due to turbined water use); and loss of freshwater quality. Similarly to human health, degradation corresponds to the consumption of freshwater of a higher quality (with a higher ecological value or ecological functionality) and the release of freshwater of lower quality (with a lower ecological value, thus affecting all the ecological users needing a better water quality, but not the users able to deal with a lower quality).

It should be noted that the latter cause–effect chain is related to the deprivation of freshwater of a given quality and not to the aquatic ecotoxicity, aquatic eutrophication, and aquatic acidification impact of this degradation. The midpoint impacts related to freshwater deprivation, which depend on water scarcity and water quality, eventually lead to species diversity change in aquatic and terrestrial ecosystems. The extent of these changes depends on the ecological value of water in the considered ecosystem. Ultimate impacts on ecosystem quality are commonly expressed in potentially disappeared fraction (PDF) of species on a given surface or volume during a given time (PDF per square meter per year or PDF per cubic meter per year) (van Zelm et al. 2011; Hanafiah et al. 2011).

Milà i Canals et al. (2009) suggest that changes caused by production systems on the amount of rainwater available to other users (ecosystems) through changes in the fractions of rainwater that follow infiltration, evapotranspiration, and runoff should be included as impacts on ecosystem quality. This is closely linked to the impact of land occupation and transformation on green water availability through the variation of stock of water stored as soil moisture available for plant uptake (green water).

Consumption of all freshwater types as well as withdrawal and release of fossil groundwater can respectively lead to overuse of renewable water bodies or exhaustion of nonrenewable fossil groundwater. Overuse of renewable water bodies depends on the water renewability rate. These midpoint impacts affect water flows and funds and ultimately have an effect on the resources stock. This reduction of available water affects other cause–effect chains by increasing local water scarcity. Different approaches exist to characterize the impact on resources encompassing the abiotic depletion potential given in antimony equivalents (Sb-eq) (Milà i Canals et al. 2009) at the midpoint level, and the backup technology concept expressing the resource damage in megajoules (MJ) surplus energy- (Pfister et al. 2009) or exergy-based methods given in megajoules of exergy (MJex) (Boesch et al. 2007) at the endpoint level.

The inventory section contains both inventory databases and inventory methods. The ecoinvent database (Frischknecht et al. 2004; Ecoinvent 2007) and GaBi database (PE 2011) are widely used databases and contain elementary flows for freshwater withdrawal and turbined water. The WFN database (Water Footprint Network 2011) assesses the inventory consumptive and degradative flows of crops and derived crop products, farm animals, and animal products; biofuels; national consumption and production; as well as trade in crop and animal and industrial products according to the WFN method (Water Footprint Network 2011). Pfister et al.’s database (Pfister et al. 2011) assesses the freshwater consumption for the production of 160 crops. An additional source of data for consumptive and evapotranspirative use can be found for five crops and three livestock products (Hanasaki et al. 2010). The Quantis water database (Quantis 2011) is a database of water uses based on ecoinvent 2.2 developed in the aim of providing industrial stakeholders with datasets required to apply all existing impact assessment methods.

Inventory methods generally suggest concepts for a systematic classification of freshwater elementary flows according to their type (surface water, groundwater, precipitation water stored as soil moisture, whether intake water quality is considered, etc.) without providing respective data. Inventory methods also describe technical water flows such as cooling water and irrigation water. The reviewed inventory methods differ widely in their objective and level of detail. Some focus on defining water categories to allow quality to be considered (Vince 2007; Bayart 2008; Boulay et al. 2011a), and others, on providing inventory tools for organizations (Hoekstra et al. 2011; WBCSD 2010), integrating the effects of direct water use and of land occupation and transformation on water availability in a comprehensive methodology (Milà i Canals et al. 2009), or providing detailed hydrological modeling and classification of freshwater use data in specific sectors (e.g., Australian red meat sector) (Peters et al. 2010). Boulay et al. (2011a) was built on Vince’s (2007) and Bayart’s (2008) methods.

Midpoint impact assessment methods give an indicator either common to all areas of protection or specific to a defined area of protection. Methods covering all area of protections giving a single index related to water scarcity include the Swiss ecological scarcity (Frischknecht et al. 2006; Pfister et al. 2009; Ridoutt and Pfister 2010b), Water Impact Index of Veolia, Boulay et al. (2011b) methods, and Water Footprint impact indices (Hoekstra et al. 2011). Area of protection-specific midpoint indicators describe the impact pathway leading to a decrease in freshwater availability for contemporary human users (Bayart 2008), as well as changes in freshwater availability for ecosystems leading to freshwater ecosystem impacts (Milà i Canals et al. 2009) and changes in groundwater availability causing freshwater depletion (Milà i Canals et al. 2009). Milà i Canals et al. (2009) suggest to use different types of water indices (Smakhtin et al. 2004; Falkenmark et al. 1989; Raskin et al. 1997) to assess freshwater ecosystem impacts. Falkenmark et al.’s (1989) index focuses on human use by evaluating the fraction of the total annual runoff available for human use. Raskin et al. (1997) use a water use per resource refined by Smakhtin et al. (2004) by subtracting environmental freshwater requirements from the available resources to derive a water index focused on freshwater resources available for human use.

The overall “blue-green-gray water” footprint concept of Hoekstra et al. (2011) was generally classified as an inventory metric, given that precipitation water stored as soil moisture evapotranspirated by plants (“green water footprint”) and consumptive use of surface and groundwater (“blue water footprint”) represent physical metrics and are not further characterized. However, the gray water footprint can also be evaluated as a midpoint approach as gray water footprint denotes degradative freshwater use by characterizing the chemical pollution in water similar to “the critical dilution volumes approach”,¹ i.e., an equivalent amount of water needed to dilute an emission below an acceptable threshold. This method thus juxtaposes measurable inventory results of “blue” and green water footprint with a theoretical volume of “gray water” which corresponds to a characterized inventory results. Using the term gray water also creates the problem of having two competing definitions of this term circulating in the water industry² (Henriques and Louis 2011).

Endpoint impact assessment methods provide specific indicators for potential damages on the areas of protection of human health (Boulay et al. 2011b; Motoshita et al. 2010b, a; Pfister et al. 2009), ecosystem quality (Hanafiah et al. 2011; Pfister et al. 2009; van Zelm et al. 2011), and resources (Pfister et al. 2009; Boesch et al. 2007).

Other approaches exist to estimate impact on resources that attempt to account for the emergy flows put into place by natural processes to make available a given resource at a given state (Zhang et al. 2010; Rugani et al. 2011) but are not evaluated in this review because they are not specific to the characteristics of freshwater resource. Emergy is defined as the measure of both the work of nature and that of humans in generating products and services, i.e., a record of previously used-up available energy that is a property of the smaller amount of available energy in a transformed product (Odum 1996).

Water indices are originally non-LCA-based indicators that express a measure of human and environmental water needs or of the fraction of resource available to meet these needs. Water indices can be used as characterization factors for midpoint (Raskin et al. 1997; Smakhtin et al. 2004; Falkenmark et al. 1989) and endpoint (Sullivan et al. 2003; Döll 2009) impact assessment methods when applied to freshwater consumptive or degradative use. Such indices can be considered as human use oriented (Gleick 1996; Falkenmark et al. 1989; Ohlsson 2000; Seckler et al. 1998; Sullivan et al. 2003; Döll 2009), ecosystem use oriented (Smakhtin et al. 2004), or cover all three areas of protection (Alcamo et al. 2007; Raskin et al. 1997; Pfister et al. 2009; Frischknecht et al. 2006; Hoekstra et al. 2011; Boulay et al. 2011b). In this work, the terminology “water scarcity index” is related solely to withdrawal-to-availability ratio (Smakhtin et al. 2004; Alcamo et al. 2007; Raskin et al. 1997; Seckler et al. 1998; Pfister et al. 2009; Frischknecht et al. 2006; Bayart et al. submitted) or consumption-to-availability ratio (Boulay et al. 2011b; Hoekstra et al. 2011). Water scarcity indices can be based solely on a measure of water scarcity or include, additionally, a measure of water quality (Boulay et al. 2011b). The details of the implementation of water indices in a LCA context, i.e., the water type to be considered in the inventory phase, needs to be specified in order to make water indices applicable in a method.

Uncertainties are generally large in life cycle impact assessment, especially on the endpoint level, and are yet generally not quantified in most of methods. Only a few authors, i.e., Pfister and Hellweg (2011), reported uncertainties for human health and WSI indicators on watershed and country level.

The impact pathways covered by current methods regarding human health include the lack of freshwater for hygiene and ingestion resulting in the spread of communicable diseases (Motoshita et al. 2010b; Boulay et al. 2011b), water deprivation for irrigation causing in malnutrition (Pfister et al. 2011; Motoshita et al. 2010a; Boulay et al. 2011b), and water deprivation for freshwater aquaculture and fisheries resulting in loss of productivity and food supply (Boulay et al. 2011b). Indirect impact of freshwater use, i.e., impact on human health and conflict creation, is not covered by existing methods. The cause–effect chain modeling is based on hydrological and socioeconomical data (Pfister et al. 2009; Boulay et al. 2011b; Motoshita et al. 2010b, a). Some of them consider the water scarcity index used at the midpoint level (Pfister et al. 2009; Boulay et al. 2011b). The level of economic development is considered in studied methods through parameters such as Human Development Index (Pfister et al. 2009), house connection to water supply (Motoshita et al. 2010b), or adaptation capacity based on gross national income (Boulay et al. 2011b). All methods consider the reduction of human health impacts in case the level of economic development is sufficient to cover compensation mechanism costs, but none of them include the impact of the development and functioning of compensation mechanisms. Not expanding the system boundary is a common approach in attributional LCA. Some of the cause–effect chain relationships have been calculated based on empirical data, e.g., malnutrition rate and human development index (Pfister et al. 2009), water scarcity, and accessibility to safe water (Motoshita et al. 2010b). Other cause–effect chains rely on the multiplication of key parameters (Boulay et al. 2011b). Both approaches are relevant but need to be further characterized by a measure of uncertainty to assess the deviation of potential impacts estimation. Endpoint indicators are generally regionalized on a country (Pfister et al. 2009; Motoshita et al. 2010b; Boulay et al. 2011b) or watershed level (Pfister et al. 2009; Boulay et al. 2011b).

Methods addressing ecosystem quality cover different parts of the cause–effect chains relevant to ecosystem services and biodiversity. The cause–effect chains that current methods cover regarding damages to ecosystem quality are the decrease of terrestrial biodiversity due to freshwater consumption (Pfister et al. 2009), decrease of aquatic biodiversity due to turbined water use, disappearance of terrestrial plant species due to groundwater withdrawal and related lowering of the water table (van Zelm et al. 2011), and the effects of freshwater consumption on freshwater fish species (Hanafiah et al. 2011). These endpoint methods do not use water scarcity indices as elements of the modeling equations. Rather, they are applied to different water types and uses and should be used complementarily. Most methods consider the ecological value of freshwater resources through an empirical observation of decreased biodiversity or of other proxy data such as net primary production (Pfister et al. 2009; van Zelm et al. 2011) and from a mechanistic perspective, e.g., by relating fish species richness to river discharge (Hanafiah et al. 2011).

Some cause–effect chains, e.g., the impact due to water deprivation related to water quality degradation on aquatic ecosystems, still need to be covered by additional methods. Endpoint methods addressing ecosystem quality have different levels of spatial differentiation: no differentiation, generic or for a specific region (van Zelm et al. 2011), archetype (e.g., alpine and non-alpine dams) country (Pfister et al. 2009), or watershed (Pfister et al. 2009; Hanafiah et al. 2011). This variability of the differentiation level reflects the diversity of the parameters considered in the cause–effect chain.

Methods addressing the area of protection resources quantify the impact on future freshwater availability through a backup technology approach to evaluate the impact of freshwater consumption above their renewability rate (Pfister et al. 2009) or through the exergy content of the freshwater resource (Boesch et al. 2007). In contrast to the Pfister et al.’s method (2009), Boesch et al.’s (2007) method is not specific to water resources and does not consider water scarcity.

None of the evaluated endpoint methods cover the cause–effect chain comprehensively; the pathway addressing impact due to fossil groundwater depletion is poorly known and is not covered by available methods. Furthermore, estimation of impact of consumption over the renewability rates lacks differentiation between different water types, and change in green water availability is not covered. Pfister et al.’s (2009) method is a spatially differentiated method on a watershed and a country level, whereas Boesch et al.’s (2007) method is not differentiated.

From a business and industry perspective, data availability on freshwater use as well as harmonized reporting formats are limiting factors for establishing meaningful water footprints of products, processes, and organizations (Koehler 2008). A balanced approach between LCIA methods and business data requirements is therefore needed to make characterization methods broadly applicable and meaningful. In order to link up with emerging LCI and LCIA methods, inventory databases should preserve the maximum freedom to provide necessary flows for application of different impact methods. The following recommendations for inventory database developments were drawn based on existing LCI and LCIA methods and are evaluated as necessary:

The following additional optional guidelines could be integrated:

General recommendations for inventory methods are the following:

In order to ease their applicability, LCIA shall, in general, show robust examples linking the inventory of freshwater types with all needed calculation steps to apply characterization factors and aggregate results for obtaining related mid- or endpoint indicators.

The water consumption or withdrawal to availability ratio has been recognized as a representative proxy for scarcity, in comparison to other indices, e.g., water use per capita, which reflects rather a socioeconomic situation. Midpoint methods addressing water scarcity shall (1) include water storage capacity in the modeling of total water availability within a geographical unit, (2) be quantitatively compared to evaluate the trade-off between easiness of application and cause–effect chain coverage and related uncertainty between indicators based solely on water scarcity (Pfister et al. 2009; Ridoutt and Pfister 2010b; Frischknecht et al. 2006; Milà i Canals et al. 2009) and more comprehensive midpoint indicators (Boulay et al. 2011b), (3) provide further empirical evidence of the link between water scarcity, water deprivation, and impact on different areas of protection to evaluate the relevance of mid- versus endpoint indicators. In an LCA perspective, water scarcity indicator does not refer to any potential impact. This does not necessarily mean that an endpoint is ultimately affected. Water stress index is, for example, involved in Pfister et al.’s and Boulay et al.’s endpoint models for human health, but human health is not affected if the economic development level is sufficient. Clear evidence of the link between water scarcity, water deprivation, and impact on different areas of protection would be needed to evaluate the relevance of mid- versus endpoint indicators.

Next steps towards a consistent framework for application of endpoint methods are as follows.

For all mid- and endpoint methods, uncertainties of input data as well as model uncertainty still need to be evaluated and documented. Mid- and endpoint methods covering human health and ecosystem quality impact shall provide characterization factors with monthly differentiation to reflect variability related to meteorological conditions and associated ecosystem changes.