Water Footprint – Assessing Impacts of Water Use Along Product Life Cycles

Home , 2013 WDF World Cup, Water footprint

Water Footprint – Assessing Impacts of Water Use along Product Life Cycles

vorgelegt von M.Sc. / Dipl.-Ing. (FH) Markus Berger

Fakultät III Prozesswissenschaften der Technischen Universität Berlin zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften ‒ Dr.-Ing. ‒ genehmigte Dissertation

Promotionsausschuss: Vorsitzender: Prof. Dr. Martin Jekel Gutachter: Prof. Dr. Matthias Finkbeiner Prof. Dr. Stefanie Hellweg

Tag der wissenschaftlichen Aussprache: 16. Dezember 2013

Berlin 2014 D 83

Acknowledgement

“This dissertation would not have been possible without the support of many people.” 7,820 Google hits show that this is a default sentence in the acknowledgement section of many dissertations – and now I know why… There are a lot of people whom I would like to thank for their support and assistance throughout the last years – and especially during the last weeks.

First of all, I wish to thank my supervisors Prof. Dr. Stefanie Hellweg and Prof. Dr. Matthias Finkbeiner for their valuable feedback and the help they provided. A very special thank goes to you Matthias, for supporting me with your knowledge, experience, and Long Island Ice Tea with hardly any coke…

Furthermore, I would like to thank Annekatrin Lehmann for reading the first draft of this doctoral thesis and the detailed feedback she provided. Thanks a lot also to all co-authors of the journal publications upon which this dissertation is based. I am very grateful for the time, efforts, and feedback you dedicated to this work. In particular the data provided by Ruud van der Ent and Stefanie Eisner have been very important for the innovations presented here. As important as these datasets has been the help of Korbinian Brochnow who taught me how to process them in the GIS calculations. Moreover, the support provided by our project partners, especially by Dr. Jens Warsen and Dr. Stephan Krinke from Volkswagen, are highly acknowledged as they enabled the first water footprint study of industrial products. A big thank you goes to my colleague Vanessa Bach with whom I had the pleasure to collaborate in all the water footprint projects during the last years ranging from cows to seawater desalination plants. Furthermore, I wish to thank all colleagues from the Chair of Sustainable Engineering for supporting me with feedback and ideas as well as with coffee and chocolate.

A very special thank is extended to my mum, Ursula Berger, who spent her holidays proofreading this entire dissertation, with the exception of the acknowledgement – which is likely to be full of mistakes…

Finally, the biggest thank you goes to my family: Inga and Emil. Three lines in an acknowledgement section can hardly express the great support you provided and the sacrifices you made. Sorry for spending too many evenings in front of a computer rather than with you on a playground and thank you for giving me the strength to finish this work. Mops!

iii Abstract

Abstract

Freshwater scarcity is a relevant problem for more than 1 billion people around the globe. Therefore, the analysis of water consumption along the supply chain of products is of increasing relevance in current sustainability discussions. This thesis aims at enhancing the concept of water footprinting by reviewing and applying various water footprint approaches, identifying methodological challenges, and developing a novel water footprint method.

In a comprehensive literature review more than 30 water footprint methods, tools, and databases have been identified and discussed. The scopes of water footprint approaches differ regarding the types of water use accounted for, the distinction of watercourses, the inclusion of quality aspects, and the consideration of temporal and regional aspects such as water scarcity and sensitivity of population or ecosystems. As the most advanced methods require the highest resolution inventory data, the trade-off between precision and applicability needs to be addressed in future database and method developments.

As most of the water accounting and impact assessment methods have hardly been applied in practice, in this work a selection of methods has been tested in various case studies. Representing the first water footprint study of complex industrial products, water consumption and resulting impacts have been analyzed along the life cycles of Volkswagen’s car models Polo, Golf, and Passat. Based on inventory databases freshwater consumption throughout the cars’ life cycles has been allocated to material groups and assigned to countries according to import mix shares or location of production sites. By means of these regionalized water inventories, consequences for human health, ecosystems, and resources have been determined by using recently developed impact assessment methods. Water consumption along the life cycles of the three cars ranges from 52 – 83 m3/car. More than 95% of the water is consumed in the production phase, mainly resulting from producing iron, steel, precious metals, and polymers. Results show that water consumption occurs in 43 countries worldwide and that only 10% is consumed directly at Volkswagen’s production sites. Impacts on health tend to be dominated by water consumption in South Africa and Mozambique, resulting from the production of precious metals and aluminum. Consequences for ecosystems and resources are mainly caused by water consumption of material production in Europe.

Based on the review and case studies, methodological challenges in water footprinting have been identified and potential solutions have been presented. A key challenge is the current definition of water consumption according to which evaporated water is regarded as lost for the originating drainage basin per se. Continental evaporation recycling rates of up to 100% within short time and length scales show that this definition is not valid and needs to be revised. Also the inclusion of land iv Abstract use effects on the hydrological balance is questionable as land transformation often leads to higher water availability due to locally increased runoff. Unless potentially negative consequences, like flooding or water logging, and adverse effects on the global water cycle are considered, water credits from land transformation seem unjustified. Most impact assessment methods use ratios of annual withdrawal or consumption to availability to denote regional water scarcity. As these ratios are influenced by two metrics – withdrawal and availability – arid regions can appear uncritical if only small fractions of the little renewable supplies are used. Besides neglecting sensitivities to additional water uses, such indicators consider neither ground nor surface water stocks which can buffer water shortages temporally.

In addition to methodological challenges it has been discussed whether the water footprint should be a volumetric or impact oriented index. Authors favoring volumetric indicators claim that global freshwater appropriation is more important than regional impacts, easier to determine, and less error-prone than putting complex ecological interaction into mathematical models. However, as 1 m³ of water consumption in Mexico does not compare to 1 m³ of water consumed in Canada, water footprints need to consider regional impacts in addition to volumes. As shown in an example, volumetric water footprints can be misleading without additional interpretation because numerically smaller footprints can cause higher impacts.

Tackling the shortcomings of existing water footprint methods, the water accounting and vulnerability evaluation (WAVE) model has been developed. On the accounting level, atmospheric evaporation recycling within drainage basins is considered for the first time, which can reduce water consumption volumes by up to 33%. Rather than predicting impacts, WAVE analyzes the vulnerability of basins to freshwater depletion. Based on local blue water scarcity, the water depletion index (WDI) denotes the risk that water consumption can lead to depletion of freshwater resources. Water scarcity is determined by relating annual water consumption to availability in more than 11,000 basins. Additionally, WDI accounts for the presence of lakes and aquifers which have been neglected in water scarcity assessments so far. By setting WDI to the highest value in (semi-)arid basins, absolute freshwater shortage is taken into account in addition to relative scarcity. This avoids mathematical artefacts of previous indicators which turn zero in deserts if consumption is zero. As illustrated in a case study of biofuels, WAVE can help to interpret volumetric water footprint figures and, thus, promotes a sustainable use of global freshwater resources.

Keywords: water footprint, water use, water consumption, life cycle assessment, evaporation recycling, vulnerability

v List of abbreviations

List of abbreviations

A - Availability of freshwater, here: runoff plus upstream inflow

AFGWS - Adjustment factor for groundwater stocks

Alake - Surface area of lakes

AoP - Area of protection

AP - Acidification potential

Awetland - Surface area of wetlands

BIER - Basin internal evaporation recycling

BIER100 - Basin internal evaporation recycling within 100km

BIERhydrol-eff - Hydrologically effective basin internal evaporation recycling

C - Water consumption

CTA - Consumption-to-availability

DALY - Disability adjusted life years

deff - Effective depth of lakes and wetlands

E - Share of withdrawal consumed due to evapo(transpi)ration

EIA - Environmental impact assessment

EP - Eutrophication potential

ER - Evapo(transpi)ration recycling

FW - Freshwater

GW - Groundwater

GWP - Global warming potential

vi List of abbreviations

GWS - Groundwater stocks

ISO - International Organization for Standardization

JRC-IES - Joint Research Centre - Institute for Environment and Sustainability

LCA - Life cycle assessment

LCI - Life cycle inventory analysis

LCIA - Life cycle impact assessment

Max-s - Maximum scarcity

Min-s - Minimum scarcity

ODP - Ozone depletion potential

P - Precipitation

PGM - Platin group metals

POCP - Photochemical ozone creation potential

Q90 - Statistical low flow (runoff exceeded with a probability of 90%)

R - Long-term average runoff

SETAC - Society of Environmental Toxicology and Chemistry

SWS - Surface water stocks

T - Availability time horizon of surface water stocks

TDI - Turbo direct injection

UNEP - United Nations Environment Program

USEtox - UNEP-SETAC toxicity model

V - Vapor created in chemical reactions

vii List of abbreviations

Verband der Automobilindustrie e.V. (German Association of the VDA - Automotive Industry)

Vdam - Volume of dams and reservoirs

VR - Synthetically created vapor recycling

WaterGAP - Water – a Global Assessment and Prognosis

WAVE - Water accounting and vulnerability evaluation

WBCSD - World Business Council for Sustainable Development

WC - Water consumption

WCeff - Effective water consumption

WDI - Water depletion index

WF - Water footprint

WFN - Water Footprint Network

WHYMAP - World-wide Hydrogeological Mapping and Assessment Programme

WSI - Water stress index

WTA - Withdrawal to availability

WULCA - Water use in life cycle assessment (working group)

WW - Waste water x - Length of basin

α Runoff fraction

λ - Average local length scale of evaporation recycling

viii Table of content

Table of content

Acknowledgement ...... iii

Abstract ...... iv

List of abbreviations ...... vi

Table of content ...... ix

1 Introduction ...... 1

2 Review of water footprint methods ...... 4

2.1 Overview of water footprint methods, databases, and tools ...... 4

2.2 Comparison of water footprint methods, databases, and tools ...... 6

3 Application and comparison of water footprint methods in industrial case studies ...... 9

3.1 Water footprint of a car manufacturer’s production site (Daimler) ...... 9

3.2 Water in the copper production chain (EuroCopper) ...... 9

3.3 Water footprint of seawater desalination plants (Siemens) ...... 10

3.4 Water footprint of a flow regulator (Neoperl) ...... 10

3.5 Water Footprint of passenger cars (Volkswagen) ...... 11

3.5.1 Background ...... 11

3.5.2 Methodology ...... 11

3.5.2.1 Determination of water consumption ...... 11

3.5.2.2 Top-down regionalization of water inventories ...... 12

3.5.2.3 Sensitivity analysis ...... 14

3.5.2.4 Impact assessment ...... 15

3.5.3 Results and discussion ...... 16

3.5.3.1 Inventory ...... 16

3.5.3.2 Impact assessment ...... 19

3.5.3.3 Comparison between cars ...... 20

3.5.3.4 Sensitivity analysis ...... 21

ix Table of content

3.5.3.5 Comparison to other environmental interventions ...... 22

4 Challenges and potential solutions in water footprinting ...... 25

4.1 Inventory challenges ...... 25

4.1.1 Definition of freshwater consumption ...... 25

4.1.2 Aggregation of different types of water consumption ...... 27

4.1.3 Consideration of green water consumption ...... 27

4.2 Impact assessment challenges ...... 28

4.2.1 Challenges of midpoint scarcity indicators ...... 28

4.2.1.1 Absolute vs. relative freshwater scarcity ...... 29

4.2.1.2 Sensitivity to additional water withdrawal ...... 29

4.2.1.3 Withdrawal vs. consumption based scarcity indicators ...... 30

4.2.1.4 Annual vs. monthly water scarcity ...... 30

4.2.1.5 Determination of water availability ...... 31

4.2.2 Consideration of water quality ...... 31

4.2.3 Dealing with environmental credits ...... 32

4.2.4 Challenges of endpoint impact assessment methods ...... 32

4.3 Volumetric or impact oriented water footprints? ...... 33

5 The water accounting and vulnerability evaluation model – WAVE ...... 38

5.1 Water accounting model ...... 38

5.2 Vulnerability evaluation model ...... 41

5.3 Application of WAVE in a case study on biofuels ...... 45

5.4 Discussion of the WAVE model ...... 47

5.4.1 Scope of WAVE ...... 48

5.4.2 Water accounting model ...... 48

5.4.3 Vulnerability evaluation model ...... 50

5.4.4 The quality corrected risk of freshwater depletion ...... 52

5.4.5 Uncertainties in WAVE and sensitivity analysis ...... 52 x Table of content

5.4.6 Application of WAVE ...... 57

6 Conclusion ...... 58

6.1 Summary of main findings ...... 58

6.2 Remaining challenges in water footprinting ...... 61

7 Outlook ...... 64

7.1 Methodological trends ...... 64

7.2 The international standard on water footprinting – ISO/FDIS 14046 ...... 64

7.3 Water footprint – cure or tranquilizer? ...... 65

7.3.1 Actions to be taken based on water footprint results ...... 66

7.3.2 The water footprint – a means of mitigating global water stress? ...... 67

8 References ...... 69

Glossary ...... 79

List of Figures ...... 85

List of Tables ...... 88

Appendix...... 89

xi 1 Introduction

1 Introduction

1,400,000,000 km³ – that’s the total amount of water available on our planet. It covers two thirds of the Earth’s surface. However, only 3% of this volume is freshwater, of which 69% is locked up in glaciers and polar ice caps (Gleick 1996). The remaining 13 million km³ of usable freshwater sustain life on our planet but are distributed very unevenly around the globe. While some regions in Columbia, Indonesia, or New Zealand abound in water (> 3,000 mm annual precipitation), other places, such as the Atacama desert, the Sahel zone, or Saudi Arabia are extremely dry with less than 100 mm precipitation per year.

During the past century water use was growing twice as fast as the world’s population (UN and FAO 2007). The main reason for this was agricultural irrigation which is responsible for about 85% of global water consumption (Shiklomanov 2003) and has led to an increased water scarcity in many regions around the globe. Today, 1.2 billion people live in such water scarce regions and another 1.6 billion people suffer from economic water shortage. This means they don’t have access to safe drinking water due to missing opportunities to withdraw, purify, or transport water from aquifers and rivers (UN and FAO 2007). As a consequence of climate change, population growth, and changing consumption patterns in emerging nations, water scarcity is expected to increase significantly in many parts of the world (Alcamo and Henrichs 2002).

Taking into account these alarming figures, the analysis of water use along the supply chains of goods and products seems urgently necessary. In the nineties of the last century, the Virtual Water concept was developed which accounts for the consumption of ground and surface water (blue water), the evapo(transpi)ration of rainwater (green water), and the pollution of freshwater (gray water) (Allan 1998). A decade later, the Water Footprint was introduced as a tool which expresses the virtual water content of products, organizations, people, and nations (Hoekstra and Hung 2002) in a spatially and temporally explicit way. By revealing surprisingly high volumes, like 140 liters per cup of coffee (Chapagain and Hoekstra 2007) or 2,700 liters per cotton T-shirt (Chapagain et al. 2006), consumers have been made aware of the amounts of water consumed or polluted during the production of daily goods. Recently, water footprinting has also become a relevant issue in life cycle assessment (LCA) (ISO 14044 2006) due to its increased application in the renewable raw materials and energy sectors. In contrast to the volumetric virtual water and water footprint approaches, LCA aims at assessing regional impacts in addition to the volumes of water used along a product’s life cycle.

This different interpretation of the water footprint as a volumetric or impact oriented indicator has led to strong dispute in the scientific community. Some scholars highlight the need of additional

1 1 Introduction interpretation as 1 m³ of rainwater consumption in Brazil does not compare to 1 m³ of groundwater consumed in Egypt (Pfister and Hellweg 2009; Ridoutt and Huang 2012). In contrast, other authors argue that global freshwater appropriation is more important, as impacts are hard to predict and water is a global resource subject to virtual trade via products (Hoekstra et al. 2009; Hoekstra and Mekonnen 2012). However, as problems occur due to a regional and not a global shortage of water (Ridoutt and Huang 2012), consensus is increasing that the water footprint should measure impacts in addition to volumes as defined in the upcoming ISO standard on water footprinting (ISO/FDIS 14046 2014).

The overall target of this thesis is to enhance the water footprint concept from a methodological and practical point of view. Therefore, the following objectives are pursued:

1 Review of existing water footprint approaches 2 Application and comparison of different water footprint methods in industrial case studies 3 Identification of methodological challenges in water footprinting 4 Development of a state-of-the-art water footprint method for the accounting and assessment of water use

Each of the four objectives has been addressed in at least one peer reviewed scientific journal publication. Additionally, key findings have been presented at international conferences and in books. Industrial case studies are mostly documented in confidential project reports.

Figure 1 provides an overview of the publications which address the four objectives and sketches the structure of this thesis. The four key journal publications upon which this thesis is based are highlighted in an orange frame in Figure 1 and can be found in the appendix. In the following, the key publications are summarized and put into context in a separate chapter before a common conclusion and outlook is presented.

In order to enable a consistent terminology throughout this thesis, the terms water use and water consumption, which are often used synonymously, need to be defined (see also chapter Glossary). Adopting the terminology proposed by Owens (2001), water use describes the total withdrawal of freshwater which can be differentiated into consumptive, degradative, and borrowing water use. Consumptive water use (or water consumption) denotes the fraction of total water use which is not returned to the originating drainage basin due to evapo(transpi)ration, product integration, or discharge into other basins and the sea. Degradative water use is the part of withdrawal returned to the basin after quality degradation (e.g. waste water discharge). In contrast, borrowing water use expresses withdrawal and discharge with low or no quality degradation (e.g. cooling water).

2 1 Introduction

Figure 1 Objectives and structure of this thesis and corresponding publications. Key publications representing the basis for this work are marked in orange frames and are provided in the appendix. Other journal, conference, book, or project report publications are marked in black frames and are not provided in this dissertation.

3 2 Review of water footprint methods

2 Review of water footprint methods

A comprehensive literature review has been conducted in order to identify and assess existing water footprint methods. This chapter summarizes the findings presented in the review paper of Berger and Finkbeiner (2010) which currently is the top 1 cited paper of the journal Sustainability. Reprinted (adapted) with permission from MDPI AG, http://www.mdpi.com/2071-1050/2/4/919. By additionally considering water footprint methods which have been developed since its publication in 2010, a thorough update is provided in this work. After an overview of available water footprint methods is given, the individual accounting and impact assessment schemes are compared and discussed.

2.1 Overview of water footprint methods, databases, and tools

By means of a comprehensive literature review a broad set of methods enabling the accounting and impact assessment of water use has been identified. They can be categorized as stand-alone and LCA based methods (Figure 2). Moreover, databases and tools which facilitate water footprinting have been included in the review.

Figure 2 Water footprint methods, databases, and tools identified and classified in the literature review

4 2 Review of water footprint methods

Stand-alone methods like Virtual Water (Allan 1998) and the Water Footprint as defined by the WFN (Hoekstra et al. 2011) enable the analysis of water use throughout products’ or organizations’ supply chains. Results are usually presented on a volumetric level and potential regional consequences are – if at all – discussed on a qualitative level.

Next to stand-alone methods, 20 methods have been developed in an LCA context. LCI schemes developed by Vince (2007), Bayart et al. (2010), and Boulay et al. (2011a) propose a detailed accounting of water use which considers volumetric, geographical, watercourse, and quality information in order to satisfy inventory requirements of modern impact assessment methods. The accounting scheme of Berger et al. (2014) additionally considers effects of atmospheric moisture recycling within basins (for details see chapter 5.1). On the midpoint level, methods assess the consequences of water use or consumption in the middle of the cause–effect chain on human health (Bayart et al. 2009; Boulay et al. 2011b), ecosystems (Mila i Canals et al. 2008), resources (Hauschild and Wenzel 1998; Bösch et al. 2007; Mila i Canals et al. 2008; Berger et al. 2014) or unspecified for all areas of protection (Brent 2004; Pfister et al. 2009; Veolia 2011). Endpoint methods assess potential damages resulting from water use or consumption at the end of the cause–effect chain on human health (Motoshita et al. 2008; Pfister et al. 2009; Motoshita et al. 2011; Boulay et al. 2011b), ecosystems (Humbert and Maendly 2008; Pfister et al. 2009; Hanafiah et al. 2011) and resources (Pfister et al. 2009). Additionally, there are some rather specific endpoint methods which are not shown in Figure 2 and assess impacts of water use on biodiversity in the Netherlands (van Zelm et al. 2011), in wetlands of international importance (Verones et al. 2013a; Verones et al. 2013b), or in coastal wetlands due to salinity increase (Amores et al. 2013). Impact assessment methods which focus on the pollution of freshwater only and do not consider withdrawal, such as eutrophication (Guinee et al. 2002), are not included in this review.

An early summary of the water footprint approaches available along with a discussion of individual strengths and weaknesses has been published in the review paper of Berger and Finkbeiner (2010). Many of the methods developed after the publication of this review paper are summarized in the work of Kounina and colleagues (2013).

In order to reach consensus on methodological questions, an international standard on water footprinting is currently being developed (ISO/FDIS 14046 2014). This document defines the procedure for conducting a water footprint study and sets methodological requirements to be fulfilled in the inventory and impact assessment phases. An outlook on this standard is given in section 7.2.

In addition to water footprint methods as such, several databases have been identified which provide water use and consumption data for various products and materials. Databases can be divided into

5 2 Review of water footprint methods typical life cycle inventory (LCI) databases like GaBi (PE International 2013) and Ecoinvent (Ecoinvent centre 2013), sector and country specific databases (FAO 2013; Pfister et al. 2011a; Pfister et al. 2011b; Ono et al. 2012) and distinct water footprint databases like the Quantis Water Database (Quantis 2013) or the WaterStat database (WFN 2013c).

Moreover, several tools like the Global Water Tool (WBCSD 2013), the Local Water Tool (GEMI 2013a), the WF Assessment Tool (WFN 2013a), Collecting the Drops (GEMI 2013c), Connecting the Drops (GEMI 2013b), the Corporate Water Gauge (CSO 2013), and the Water Risk Filter (WWF 2013) have been identified which facilitate the accounting of a company’s (direct) water use and assess environmental, operational, legal, and reputational risks.

2.2 Comparison of water footprint methods, databases, and tools

In order to enable a comparison of the water footprint methods, databases, and tools identified in the (updated) literature review, a set of criteria has been developed for assessing their scope and applicability. These criteria comprise:

 The type of water analyzed  The type of usage considered  The inventory data required/provided  Areas of protection (AoP) addressed  Availability of characterization factors  ISO 14044 compliance regarding comparative assertions disclosed to the public

As shown in Table 1, most methods focus on consumptive blue water use. Green and gray water is mainly considered by stand-alone methods in order to address rainwater evapotranspiration of agricultural products and degradative freshwater use respectively. The analysis revealed that the inventory requirements of water footprint methods differ significantly. In general, scientifically advanced methods show a need for higher resolution inventory data. In addition to volumes and regional information, methods like Veolia (2011) or Boulay et al. (2011b) require information on watercourses and water qualities. In some cases, even temporal information is required to acknowledge varying water scarcity throughout the year (Hoekstra et al. 2012; Pfister and Baumann 2012). Even though this increased level of precision is appreciated from a scientific point of view, such inventory requirements are hard to fulfill – especially if complex background systems are involved. Hence, the trade-off between “precision” and “applicability” needs to be addressed in future studies and in the new international standard. Considerable differences have been detected concerning the availability of characterization factors. While some methods comprise characterization models but no factors (e.g. Hauschild and Wenzel 1998), other methods provide

6 2 Review of water footprint methods characterization factors on both drainage basin and country levels (e.g. Pfister et al. 2009; Berger et al. 2014). As some of the impact assessment models contain a weighting step (e.g. Frischknecht et al. 2009), they cannot be used in water footprint studies comprising comparative assertions disclosed to the public.

The level of detail provided in LCI databases differs significantly. While LCI databases like GaBi and Ecoinvent only provide information on the volumes and watercourses used, additional regional, quality, and even temporal information can be found in distinct water footprint databases. A similar variation concerning inventory requirements has been identified in the water footprint tools.

A detailed follow-up characterization of methods has been accomplished by the water use in LCA (WULCA) working group of the UNEP-SETAC Life Cycle Initiative (Kounina et al. 2013). This work is based on the review scheme of the International Reference Life Cycle Data System (JRC-IES 2011) and was co-authored by the author of this dissertation.

The review presented in this chapter, which updates the publication of Berger and Finkbeiner (2010), clearly shows that there is not only one “water footprint”. Next to stand-alone methods, databases, and tools, most methods have been developed in an LCA context. Impact assessment models range from rather simple scarcity indicators up to comprehensive endpoint models which describe complex cause-effect chains. In addition to differences concerning the addressed areas of protection and the availability of characterization factors, water footprint methods differ significantly regarding their inventory data requirements. While for some impact assess models the volume und the regional information are sufficient to enable applicability, other methods require additional quality or even temporal information. In order to support this theoretical comparison and to test applicability, some of the water footprint methods have been applied in industrial case studies, whose results are presented in the following chapter.

7 2 Review of water footprint methods

Table 1 Scope and characteristics of the water footprint methods, databases, and tools identified in the literature

review (updated from Berger and Finkbeiner (2010))

yes

yes***

ISO14044 ISO14044

no, weighting no,

compliance **

factors

Availability of

characterization

for basins/countries for

for 214 river basins 214 river for

for countries for

for basins/countries for

to be be calculatedto

for countries for

for basins/countries for

to be be calculatedto

for basins/countries for

for 7 countries for

for basins/countries for

for main basinsmain for

fixed exergy content

for South Africa South for

to be be calculatedto

for the main basins main the for

fied

unspeci-

Re-

sources

Eco-

systems

impact assessment

Areas of protection addressed in

health

Human

Quality

discharge

Quality

withdrawal

course

Water-

Geo-

graphy

Inventory data required/provided

Volumes

Water

consumption

x (gray)

Water use

Type of usage considered

gray

blue

* barrage water only, ** for comparative assertions comparative **disclosed only, for *** public, * the unlesswater barrage to aggregated 99 result is eco-indicator used

Water analyzed

green

Method

Water RiskWater Filter

Corporate Water Gauge Water Corporate

Connecting the Drops the Connecting

Collecting Drops the

WF AssessmentWF Tool

Local Tool Water

Global Water Tool Global Water

Ono et al.Ono (2012)

Pfister Pfister et al. (2011b)

Pfister Pfister et al. (2011a)

FAOSTAT

WaterStat

Quantis

GaBi

Ecoinvent

Boulay Boulay et al. (2011b)

Hanafiah et al. (2011)

Motoshita et al. Motoshita (2011)

Pfister Pfister et al. (2009)

Humbert & Maendly (2008) & Maendly Humbert

Motoshita et al. Motoshita (2008)

Berger et Berger al. (2014)

Veolia (2011)

Boulay Boulay et al. (2011b)

Pfister Pfister et al. (2009)

Bayart Bayart et al. (2009)

Frischknecht et al. Frischknecht (2009)

Mila i Canals et al. (2008)

Bösch et al. (2007)

Brent et al. Brent (2004)

Hauschild (1998) Wenzel and

Berger et Berger al. (2014)

Boulay Boulay et al. (2011a)

Bayart Bayart et al. (2010)

Vince (2007)

Water Footprint (WFN) Footprint Water Virtual Water Virtual

LCIA (endpoint) LCIA LCIA (midpoint) LCIA LCI

alone

Tools Databases Stand-

Life cycle assessment methods assessment cycle Life

8 3 Application and comparison of water footprint methods in industrial case studies

3 Application and comparison of water footprint methods in industrial case studies

The majorities of water footprint case studies available today remain on the volumetric level and have been conducted for agricultural products (WFN 2013b). Moreover, most of the water footprint methods identified and discussed in chapter 1 are recent developments and have hardly been applied in practice. Therefore, several of the newly developed impact assessment methods have been applied in industrial case studies in order to:

 Test the inventory data availability and applicability of the new impact assessment methods  Analyze the plausibility and robustness of the results obtained  Compare the results of impact assessment methods to those of volumetric methods  Compare the results of impact assessment methods among each other

As most of the case studies underlie confidentiality agreements with the commissioning companies, only a short description of these studies is presented here. However, the study accomplished with Volkswagen has been published (Berger et al. 2012) (appendix, paper II) and is described in more detail in section 3.5.

3.1 Water footprint of a car manufacturer’s production site (Daimler)

In this study a water balance of Daimler’s production site in Sindelfingen has been created in order to account for the freshwater withdrawal, the water consumption, and the wastewater discharge of the site. Based on this water balance, risks and weaknesses of the water circulation system as well as water losses and saving potentials were identified. Direct environmental impacts resulting from water consumption were assessed using a set of recently developed impact assessment methods. Additionally, indirect impacts resulting from energy and chemical usage required to run the water purification, the circulation system, and the waste water treatment were analyzed. Subsequently, advanced water reuse options were analyzed in terms of their benefits (less water consumption and energy demand for withdrawal and purification) and environmental burdens (more energy for cycling systems, fungicides, etc.). Finally, different scenarios for rainwater harvesting were analyzed considering both direct impacts of water consumption/saving and indirect consequences resulting from infrastructure.

3.2 Water in the copper production chain (EuroCopper)

In this study water use and consumption along the supply chains of primary and secondary copper sheets and tubes have been analyzed. Inventory data derived from literature and LCI databases was

9 3 Application and comparison of water footprint methods in industrial case studies compared to company data collected at copper mines, primary and secondary smelters, as well as sheets and tube producers. Based on this investigation, the water profile of European semi-finished copper products was determined considering import mixes of copper concentrate as well as primary and secondary production shares. Furthermore, different impact assessment methods were applied to estimate potential consequences of water consumption on human health, ecosystems, and resources in different regions along the copper production chain.

3.3 Water footprint of seawater desalination plants (Siemens)

In this study the water footprint of a Siemens seawater desalination plant has been determined. This is particularly interesting because on the one hand, seawater desalination plants produce fresh water, on the other hand, freshwater is also consumed during the production and energy generation required for their operation. After identifying appropriate databases as well as inventory and impact assessment methods, the “net water production” of a seawater desalination plant was determined. Subsequently, environmental benefits and burdens resulting from freshwater production and consumption were analyzed and compared using different impact assessment methods. In addition to various desalination technologies, also influences of the salinity, the energy source used, and the location of operation were examined. Finally, the environmental impacts of seawater desalination plants were compared to freshwater transportation by means of pipelines.

3.4 Water footprint of a flow regulator (Neoperl)

Products of the NEOPERL GmbH, like flow regulators and aerators, help to reduce freshwater use and consumption in households and industry. Moreover, energy demand and greenhouse gas emissions can be reduced due to the avoided withdrawal, purification, and treatment of water. On the other hand, the production and recycling of flow regulators also causes water consumption and other environmental interventions. Therefore, this study aimed at analyzing water consumption and water savings along the life cycle of a NEOPERL flow regulator and estimating the resulting regional impacts (or benefits) on human health, ecosystems, and resources. Moreover, the primary energy demand and the global warming potential were analyzed considering the environmental burden during production and recycling as well as credits obtained in the use phase. Finally, it was analyzed how regional impacts of water consumption would change if the flow regulator was produced by competitors in China assuming similar production processes but higher regional water scarcity.

10 3 Application and comparison of water footprint methods in industrial case studies

3.5 Water Footprint of passenger cars (Volkswagen)

This chapter summarizes the water footprint case study conducted for Volkswagen. The corresponding publication (Berger et al. 2012) has won Environmental Science and Technology’s 2012 Award Second Runner-Up Best Article – Environmental Technology (Hotze 2013). Reprinted (adapted) with permission from Berger, M., J. Warsen, S. Krinke, V. Bach, and M. Finkbeiner. 2012. Water Footprint of European Cars: Potential Impacts of Water Consumption along Automobile Life Cycles. Environmental Science and Technology 46(7): 4091-4099. Copyright 2012 American Chemical Society. http://pubs.acs.org/doi/abs/10.1021/es2040043

3.5.1 Background

Volkswagen has been analyzing the environmental effects of its cars and components by means of LCA for many years (Volkswagen AG 2010a). However, water consumption has not been considered yet due to a lack of awareness, inventory data, and impact assessment models. Therefore, this study has analyzed freshwater consumption and resulting impacts along the life cycles of the three Volkswagen car models Polo 1.2 turbocharged direct injection (TDI), Golf 1.6 TDI, and Passat 2.0 TDI (model year 2010). Further objectives comprised the discussion and comparison of impact assessment results and methods, the identification of significant life cycle stages and processes, and the analysis of how sensitive results are to the regional context. In order to estimate the relevance of water for the automotive industry, the potential damages resulting from water consumption have been compared to damages caused by other environmental interventions, like resource use and emissions. As existing water footprint studies focus on agricultural products, this analysis represents the first published water footprint case study conducted for complex industrial products.

3.5.2 Methodology

3.5.2.1 Determination of water consumption

Water consumption was analyzed in the production, use, and end of life phase (Figure 3). Data was derived from the GaBi LCI models which were established by means of the slim LCI interface system (Koffler et al. 2008) for the environmental commendations of the Polo, Golf, and Passat (Volkswagen AG 2010c, 2010b).

In order to apply impact assessment methods, which evaluate the consequences of the water consumption determined from the LCI models, the basic volume is not enough. Regionalized water inventories, which state the location where water consumption occurs, are needed to consider regional water scarcity conditions, the vulnerability of ecosystems, or socio-economic parameters affecting the sensitivity to water scarcity induced health damages. Such geographically explicit water inventories are determined in a top-down approach as described in the following chapter.

11 3 Application and comparison of water footprint methods in industrial case studies

Production

mining of raw materials (e.g. iron ore, bauxite, crude oil)

material production (e.g. steel, aluminium, polymers)

fabrication of components (e.g. hot stamping of trunk lid, die-casting of crankcase, injection moulding of glove box)

assembly (e.g. joining, painting)

Freshwater Use Wastewater

withdrawal production of crude oil discharge

in country in country

a, b, c… production of diesel in refinery a, b, c…

End-of-life

end-of-life vehicle

draining removal of of fluids battery, catalyst, spare parts

shredding of car body

treatment, material recycling, energy recovery of shredder residues

disposal of wastes

system boundaries Figure 3 Life cycle of the cars along which water consumption has been analyzed

3.5.2.2 Top-down regionalization of water inventories

In order to regionalize a spatially unspecific water inventory, the car’s total water consumption is divided into the shares consumed in the life cycle stages production, use, and end of life. For further specification, the water consumed in the production phase is assigned to manufacturing steps and to 15 material groups (specified in the German material classification in motor vehicle construction standard (VDA 231-106 1997)). Now the water consumption caused by the manufacturing steps and material groups is allocated to specific countries, based on production mixes, location of suppliers, production sites, etc. This procedure is illustrated in the following example of polymer components, which has been chosen due to its complexity.

12 3 Application and comparison of water footprint methods in industrial case studies

As it can be seen in Figure 4, the water consumption of the polymer production steps crude oil production, refinery, polymerization, and component fabrication has been determined based on data derived from the GaBi database (PE International 2013). As no datasets are available for the refinery and polymerization as such, the water consumption figures have been determined by means of a differential method. In order to estimate the water consumption of a refinery, water consumption figures of crude oil production have been subtracted from the average water consumption of monomers available in the GaBi database. In a similar way, the average water consumption of monomers has been subtracted from the weighted average water consumption of polymers used in a car in order to estimate the water consumption of the polymerization phase. The water consumption of the component fabrication has been determined based on average water consumption data of polymer extrusion and injection molding available in the GaBi database.

Figure 4 Top-down regionalization of water consumed in the production of polymer components showing the production steps and their fractions of water consumption, the basis used for the regionalization, and the water allocation shares of countries

After determining the water consumption shares of the individual production steps of the polymer components, the actual top-down regionalization can be accomplished. Hence, 1.9% of the water consumption resulting from the crude oil production has been allocated to twelve countries contributing to the European crude oil production mix (PE International 2013). The fraction of water consumed in the refinery (2.5%) has been assigned to countries based on the location of European refineries (Ecoinvent centre 2013). The share of water consumed in the polymerization step (42.6%) has been broken down to the countries hosting most of the European polymerization plants. The

13 3 Application and comparison of water footprint methods in industrial case studies fraction of water consumption resulting from the component fabrication (53.1%) is assumed to occur in Volkswagen’s production plants (Pamplona, Wolfsburg, Emden) und at suppliers located mainly in Germany and Belgium.

Based on these estimations, the total water consumption of the material group polymers has been allocated to 26 countries in which the production of polymer components or of their preliminary products takes place.

Regionalized water inventories of other materials, such as metals, can be determined more simply in theory as fewer production stages, e.g. mining and refinery, and locations are involved. However, in practice water consumption data is often only available for the aggregated metal datasets. Therefore, regionalization has been accomplished based on the import mix shares of the metal only.

Several assumptions have been necessary to create geographically explicit water inventories according to the top-down procedure described above. First, it has been assumed that materials are purchased according to average import mix shares, as Volkswagen’s specific supply situation could not be analyzed in detail. Second, by assigning the water consumption of material groups to countries based on import mix shares, it is presumed that the water intensity for producing a certain material is equal in all countries. Third, it is assumed that material production is accomplished exclusively in countries contributing to the material import mix. This neglects the fact that minor volumes of water consumed during the generation of electricity or the production of auxiliary products might have been consumed in other countries than those included in the mix.

3.5.2.3 Sensitivity analysis

Uncertainties caused by this top-down regionalization could be avoided if geographically differentiated water flows were available in the LCA databases, as it is already common practice for fossil energy carriers to consider different calorific values and qualities. In order to analyze the consequences resulting from these assumptions, a sensitivity analysis has been conducted. By means of a minimum and maximum scarcity (min-s/max-s) scenario, the respective water consumption of the 15 material groups have been assigned to the countries in the corresponding import mixes which show the lowest/highest physical water scarcity. Physical water scarcity has been determined based on the withdrawal to availability ratio (WTA) which relates annual regional withdrawal to the renewable water resources. As it might be too optimistic or too pessimistic to assume that all materials are derived exclusively from the countries of the lowest or the highest water scarcity, the scenarios should be regarded as boundaries for more realistic options.

14 3 Application and comparison of water footprint methods in industrial case studies

3.5.2.4 Impact assessment

In order to allow for a comprehensive impact assessment the following methods, representing different levels of sophistication and assessing different consequences, have been selected:

 The impact assessment method of Frischknecht et al. (2009) (ecological scarcity method), which assesses water consumption based on physical water scarcity, measured in eco- points/m³  The impact assessment method of Motoshita et al. (2011)1, which evaluates damage to human health caused by infectious diseases resulting from polluted water uptake as a consequence of domestic water scarcity, expressed in disability adjusted life years (DALY)/m³  The method of (Pfister et al. 2009), comprising five characterization/weighting models: o Freshwater deprivation, which assesses freshwater consumption based on physical water scarcity (m³ deprived/m³ consumed) o Damage to human health, which addresses health impacts resulting from malnutrition as a consequence of agricultural water shortage, quantified in DALY/m³. o Damage to ecosystem quality, which evaluates ecological consequences resulting from decreased biodiversity due to water shortage, measured in potentially disappeared fraction of species (PDF) m² year/m³ o Damage to resources, which assesses depletion of freshwater resources as a

consequence of water uses exceeding renewability rates, expressed in MJsurplus

energy/m³ o Overall damage, which aggregates impacts determined in the three previous characterization models to a single-score result, quantified in points/m³ Even though a broad set of recent impact assessment methods has been chosen, it should be noted that the selection of methods shown in Table 1 has been restricted by inventory data requirements. Since only volumetric and geographical data has been available, impact assessment methods which require additional quality (e.g. Boulay et al. 2011b) or watercourse (e.g. Mila i Canals et al. 2008) information could not be applied in this case study. As a second criteria, only those methods could be applied which provide characterization factors on a global level.

1 Since Motoshita et al. (2011) determined damage factors only for domestic water consumption, these factors have been multiplied by a country specific ratio of domestic to total water use (FAO 2013) to allow for an assessment of general water consumption. Therefore, the method is termed Motoshita et al. (2011)* in the following.

15 3 Application and comparison of water footprint methods in industrial case studies

3.5.3 Results and discussion

In the following chapters results are presented and discussed on the inventory and impact assessment levels. Moreover, a comparison between the three cars is presented, the sensitivity of results to different regionalization scenarios is analyzed, and impacts from water consumption are compared to impacts resulting from other environmental interventions.

3.5.3.1 Inventory

The water consumption along the life cycles of the three cars amounts to 51.7 m³ (Polo 1.2 TDI), 62.4 m³ (Golf 1.6 TDI), and 82.9 m³ (Passat 2.0 TDI). These figures are lower than previously reported data of 400 m³ virtual water consumption per car (Die Zeit 2009). However, both results have been derived from different methods. While our results have been determined based on LCA, which analyzes a distinct product system, the 400 m³ have been derived from economic input-output tables. Based on financial and environmental statistics the water consumption per US$ of industrial product is calculated. Hence, the previous study represents a rather rough estimate of an average industrial product and is based on economic, not physical data. This difference in the modeling approach does not allow for a detailed discussion of differences.

Determining the water consumption of the main life cycle stages, revealed that about 95% of the water is consumed in the production phase of all three cars (Figure 5a). This is in contrast with most other environmental interventions evaluated in Volkswagen’s LCA studies, like eutrophication or global warming. Apart from the impact category ozone depletion, these environmental interventions are usually dominated by the car’s use phase (Volkswagen AG 2010b). Hence, it can be seen that different processes than fossil fuel consumption are relevant from a water perspective. Yet, it should be remembered that these results were obtained assuming the use of fossil diesel, which – according to our data – has a low water consumption of 0.005 l/MJ compared to biodiesel consuming 217 – 335 l(blue water)/MJ on global average depending on the crop used (Gerbens-Leenes et al. 2009).

A significance analysis was accomplished to identify the contributions of individual materials and manufacturing steps to the impact assessment and water consumption results of the production phase for the Golf. As shown in Figure 5b, steel and iron materials as well as polymers contribute equally strong to most impact categories and water consumption (70 – 80%). In contrast, the contribution of light metals (aluminum and magnesium alloys) to total water consumption is lower and the share of special metals (gold, silver, and platinum group metals (PGM)) is higher than in conventional impact categories. The fact that less than 1 kg of precious metals is responsible for more than 20% of the overall water consumption throughout a Golf’s life cycle highlights the large material specific water consumption of these materials.

16 3 Application and comparison of water footprint methods in industrial case studies

b) Figure 5 Relative contributions of life cycle stages (a) and material groups to production impacts (b) in the impact categories eutrophication (EP), ozone layer depletion (ODP), photochemical ozone creation (POCP), global warming (GWP), acidification (AP), and in the water consumption inventory (WC) for the Golf 1.6 TDI.

The regionalized water inventories shown in Figure 6 reveal that water consumption occurs in 43 countries worldwide. Less than 10% are consumed directly at the production sites in Pamplona, Wolfsburg, and Emden resulting mainly from painting and evaporation of cooling water. Hence, more than 90% of the water consumption along the cars’ life cycles is caused by the material and energy production in the background system.

17 3 Application and comparison of water footprint methods in industrial case studies

c) Figure 6 Global water consumption throughout the life cycles of: a) the Polo 1.2 TDI, b) the Golf 1.6 TDI, and c) the Passat 2.0 TDI.

18 3 Application and comparison of water footprint methods in industrial case studies

The procedure to establish regionalized water inventories presented in chapter 3.5.2.2 can be used as an approximation to enable the application of impact assessment models as long as LCI databases do not provide geographically explicit inventory flows. This work refrains from publishing fixed allocation keys, as the assignment of water consumption to specific countries took Volkswagen’s specific supply situation into account. It should be noted that this top-down regionalization has been refined in further case studies conducted for Siemens (chapter 3.3) and Neoperl (chapter 3.4). Moreover, a more general regionalization procedure – independent from the specific supply situation of a company – shall be developed in a current research proposal (chapter 6.2).

3.5.3.2 Impact assessment

Based on the regionalized water inventories, the impact assessment models described in section 3.5.2 were applied in order to evaluate consequences resulting from water consumption in different countries. Figure 7 shows the results obtained by means of the water inventory and impact assessment methods normalized to the Polo for the default, min-s, and max-s scenarios.

Since absolute results for the Polo differ among the scenarios, Figure 7 only allows for comparing the cars within one impact category and scenario. For instance, in the method of Frischknecht et al. (2009) it can be seen that in the min-s scenario the Passat causes less impacts than the Polo. However, it cannot necessarily be concluded that the Passat’s impacts in the min-s scenario are lower than in the default scenario as both results are normalized to the impacts caused by the Polo which differ in each scenario. Comparisons of results obtained per car and impact category in different scenarios are shown separately in Figure 8.

Figure 7 Relative comparison of results on the inventory and impact assessment levels for the default scenario (bars), the min-s scenario (circles), and the max-s scenario (diamonds).

19 3 Application and comparison of water footprint methods in industrial case studies

The results obtained in the method of Frischknecht and colleagues (2009) and the impact category freshwater deprivation from Pfister et al. (2009) depend on two factors: the volume of water consumed and the physical water scarcity at the place of consumption. While the ecological scarcity method uses the WTA ratio as a weighting factor directly, freshwater deprivation uses a water stress index (WSI) as a characterization factor which is based on WTA, but additionally considers seasonal variation of water availability (Pfister et al. 2009). Despite different proportions, both methods are dominated by the water consumption in similar countries mainly Germany (due to high volumes) as well as Spain, Belgium, and South Africa (due to high scarcity).

While the method of Pfister and colleagues (2009) assessing health damages from malnutrition considers physical water scarcity and socio-economic aspects, the method of Motoshita et al. (2011)* measuring health damages from infectious diseases considers only socio-economic aspects. As physical water scarcity is high and the level of development is rather low, human health impacts measured according to Pfister and colleagues (2009) are dominated by the water consumed in South Africa resulting from the PGM production. Damages determined in the method of Motoshita et al. (2011)* are mainly caused by relatively low amounts of water (78 – 191 l) consumed in the aluminum production in Mozambique. In contrast, due to high sanitation standards and a high degree of development, the water consumption in countries like Spain or Australia does not cause damages to human health, despite high physical water scarcity in these countries.

Ecosystem damage denotes the loss of biodiversity and is influenced by water scarcity and the regional sensitivity of vascular plants (Pfister et al. 2009). Again, the water consumption in South Africa dominates the impact assessment result with 56% (Golf) to 67% (Passat). Damages caused by the depletion of resources only occur in countries where water withdrawal exceeds the renewability rate (WTA>1). As this is not the case in Central Europe, where most of the water is consumed, large shares of water consumption do not contribute to resource damage. This impact category is dominated by the water consumption in Spain and Ukraine which contribute 55% (Passat) to 67% (Polo) to the overall result depending on the car.

3.5.3.3 Comparison between cars

By showing the results of the water inventory and impact assessment methods normalized to the Polo, Figure 7 allows for a comparison of the three cars. In the default scenario it can be seen that the increased water consumption of the Golf and Passat are reflected by the ecological scarcity method and the model of Motoshita et al. (2011)*, showing that these methods lead to similar conclusions as the inventory in this scenario. Yet, in the categories developed by Pfister et al. (2009), the impacts of the Polo and Golf are regarded as rather similar despite different water consumption. This can be explained by two facts. First, similar water consumption is weighted higher at the Polo’s

20 3 Application and comparison of water footprint methods in industrial case studies production site in Spain than at the Golf’s production site in Germany. This compensates the advantages of the lower water consumptions in the material production resulting from the reduced weight of the Polo in comparison to the Golf. Second, some impact categories, especially the one developed by Pfister and colleagues (2009) measuring damages to human health, are dominated by the water consumption of the PGM production in South Africa. As the PGM contents of the Polo and Golf are comparable, results of these impact categories are similar, too. Since the Passat contains more PGM than the Polo and Golf, the same reasoning can explain the higher impacts in the human health categories. In contrast, the water consumption in South Africa does hardly affect damage to resources since WTA is below 1 in most drainage basins, which, according to Pfister et al. (2009), means that no depletion of water resources occurs. For that reason the Passat scores only slightly worse in this impact category due to the larger water consumption of the larger material production.

3.5.3.4 Sensitivity analysis

Since the regionalization of water inventories contains several assumptions and impact assessment results strongly depend on regional aspects, a sensitivity analysis was accomplished. Therefore, the respective water consumption of the 15 material groups was fully assigned to the countries in the corresponding import mixes which show the lowest/highest physical water scarcity (min/max-s scenario).

Figure 8 Relative differences in impact assessment results between max/min-s scenarios and the default scenario

Figure 8 displays the differences between the min/max-s scenarios and the default scenario in relation to the default scenario. Hence, a difference of +100% means that impacts have doubled compared to the default scenario. A result of 0% indicates no changes and a result of -100% means that impacts are zero in this scenario. While results shown in Figure 7 could only be compared among

21 3 Application and comparison of water footprint methods in industrial case studies cars but not between scenarios, results presented in Figure 8 can only be compared among scenarios but not between cars. Hence, in the method of Frischknecht et al. (2009) it can be concluded that impact of the Polo have doubled in the max-s compared to the default scenario. However, it cannot necessarily be said that impacts of the Golf are higher than impacts of the Polo in the max-s scenario as both results are normalized to their differing default scenario results.

As shown in Figure 5b, more than 70% of water consumption derives from the production of steel and iron materials as well as polymers, which occurs mainly in Central and Northern Europe (except the iron ore and crude oil production). Hence, regional shifting of large water consumption between the scenarios takes place within Europe. Only the shift of water consumption deriving from the PGM production between South Africa and Russia in the max-s and min-s scenarios causes significant changes in the regionalized inventories outside Europe. As water scarcity, vulnerability of ecosystems, and socio-economic parameters are rather similar throughout Central and Northern Europe, variations in the impact assessment results were mainly caused by the shifting of water consumption between South Africa and Russia.

As it can be seen in Figure 8, variations from the default scenario range from -100 to +150 % in all impact assessment models with the exception of Motoshita et al. (2011)*, where the min-s scenario leads to higher impacts than the max-s scenario. This can be explained by the fact that this method is only sensitive to socio-economic parameters and doesn’t account for physical water scarcity which is used as a parameter to define min-s and max-s. Hence, instead of using a default min/max-s scenario, individual minimum and maximum scenarios could have been determined for each impact assessment method based on its characterization factors.

3.5.3.5 Comparison to other environmental interventions

An advantage of the ecological scarcity method is that it enables a comparison of water consumption related impacts to impacts resulting from other environmental interventions like the consumption of fossil and mineral resources or emissions to air, water, and soil. In a similar way, the method of Pfister et al. (2009) allows for a comparison to other environmental damages determined by means of the eco-indicator 99 method (Goedkoop and Spriensma 2001). Also impacts on human health determined according to Motoshita et al. (2011)* can be compared to health damages calculated by eco-indicator 99. In that way, the contribution of water consumption to total impacts caused along automotive life cycles and, thus, the relevance of water consumption to the environmental performance of the automotive industry can be estimated. As shown in Figure 9 for the production of the Golf, water consumption affects the total results to an extent of 0 – 7% depending on the impact category. While similar results were identified for the Polo slightly higher percentages ranging from 0 – 13% were obtained for the Passat.

22 3 Application and comparison of water footprint methods in industrial case studies

Figure 9 Relative contribution of water consumption in the production of a Golf 1.6 TDI to the total impacts according to the ecological scarcity method and the impact assessment models of eco-indicator 99 (hierarchist approach) and Motoshita et al. (2011)* and Pfister et al. (2009)

For all cars it has been shown that water consumption mainly affects damages to ecosystems rather than damages to human health and resources. This can be explained by the fact that most of the water consumption occurs in Europe, where the level of development avoids water induced health damages. Moreover, water use in Europe usually does not exceed the renewability rate, which prevents the depletion of freshwater resources. The rather low contribution of water consumption to the overall damage can be explained by the application of normalization and subsequent weighting of the three damage categories (human health 40%, ecosystems 40%, and resources 20%) in the eco- indicator 99 methodology. It should be noted that these figures only reflect the production phase of cars. When considering the whole life cycle, the share of water related damages decreases even more as other environmental interventions increase significantly while water consumption and related impacts stay rather constant.

Furthermore, a comparison of damages to human health and ecosystem quality resulting from water consumption and freshwater pollution in the production of the Golf was accomplished. Impacts caused by water pollution result from the emission of carcinogenic substances in the category human health and from the release of acidifying, eutrophying, and eco-toxic substances in the category ecosystem quality (Goedkoop and Spriensma 2001). As shown in Figure 10, damages from water consumption are smaller than damages resulting from emissions into freshwater in both damage categories.

23 3 Application and comparison of water footprint methods in industrial case studies

Figure 10 Relative comparison of non-water related damages to damages resulting from water pollution and consumption in the impact categories human health and ecosystems in the production of a Golf 1.6 TDI according to the methods eco-indicator 99 (Goedkoop and Spriensma 2001) and (Pfister et al. 2009)

Up to now, water footprints have been determined for agricultural products mainly and usually remain on the volumetric level (WFN 2013b). The case studies presented in this chapter prove that it is possible to accomplish water footprint studies in industrial product systems as well. However, as no geographically explicit water inventories are available in today’s LCI databases, an elaborate top- down regionalization is needed to enable the application of impact assessment methods. It should be noted that it is only possible to apply those impact assessment methods for which volumetric and regional information is sufficient. Methods which additionally require quality or even temporal information are not applicable considering today’s inventory data availability. Further methodological challenges identified in the literature review (chapter 2) and in the case studies are presented in the following chapter.

24 4 Challenges and potential solutions in water footprinting

4 Challenges and potential solutions in water footprinting

After reviewing a broad set of water footprint methods (chapter 2) and applying them in different industrial case studies (chapter 3), practical and methodological challenges have been identified and discussed in Berger and Finkbeiner (2013) (appendix, paper III). This chapter summarizes and updates the challenges and potential solutions identified by Berger and Finkbeiner (2013) on the inventory and impact assessment level to inspire future methodological developments. Finally, as it is an ongoing debate whether water footprints should be volumetric or impact oriented indicators, the pros and cons of both approaches are discussed. Reprinted (adapted) with permission from WILEY, © 2012 by Yale University, http://onlinelibrary.wiley.com/doi/10.1111/j.1530- 9290.2012.00495.x/abstract.

4.1 Inventory challenges

4.1.1 Definition of freshwater consumption

Freshwater consumption is defined as water “lost” for a drainage basin by either evapo(trans)piration, product integration, or discharge into sea water or another basin (Bayart et al. 2010). However, there is no scientific rationale why water withdrawal in one basin and release in another one should not be seen as consumption in the former and credit in the latter. Also, the assumption that evapo(transpi)rated water is lost for drainage basins is solely based on the fact that 70% of all precipitation occurs over the oceans (New et al. 2001). When the fate of evaporation is analyzed, it turns out that on global average 57% of terrestrial evaporation returns as precipitation over land (van der Ent et al. 2010). Thus, water consumption in one drainage basin may lead to precipitation in the same or another basin.

Figure 11 Average continental evaporation recycling ratio (van der Ent et al. 2010) – reproduced by permission of American Geophysical Union

25 4 Challenges and potential solutions in water footprinting

As it can be seen in Figure 11, continental evaporation recycling is highly site-dependent and ranges from below 10% in Patagonia and New Zealand to more than 90% in tropical regions and in the Himalayas. According to a follow-up study conducted by van der Ent and Savenije (2011), length and time scales of this atmospheric moisture recycling vary between 500 km and 3 days in the Congo basin up to more than 7,000 km and more than 30 days in deserts. Hence, it depends very much on the location and size of the drainage basin whether evapo(transpi)rated water is recycled in the same or in another basin or precipitates over oceans.

In order to incorporate these findings in water inventories, the internally recycled fraction of water evaporated in e.g. the Rhine basin and the shares that precipitate in other basins, like Elbe or Danube, and over sea need to be determined (Figure 12). While the inverse case, i.e. the evaporative sources of precipitation in a particular drainage basin, has been analyzed already (Dirmeyer and Brubaker 2007), more research is needed to determine the fate of evaporation in a geographically explicit manner. Taking into account atmospheric moisture recycling, current water inventories will change drastically as evapo(transpi)rated water is no longer considered as consumed per se. There might be even water credits resulting from the precipitation of water vapor which was synthetically created in the combustion of fossil fuels. Hence, cars might have positive carbon but negative water footprint results, i.e. they produce water, if they are run in regions with high continental evaporation recycling.

Figure 12 Fractions of water evaporated in the Rhine basin which are recycled internally and precipitate over other drainage basins and sea

26 4 Challenges and potential solutions in water footprinting

4.1.2 Aggregation of different types of water consumption

As one of the most prominent volumetric methods, the water footprint according to Hoekstra et al. (2011) comprises the consumption of ground and surface water (blue water), the consumption of soil moisture due to evapotranspiration (green water), and the pollution of freshwater due to waste water discharges (gray water).

The simple aggregation of these three types of water consumption implies equal weighting and the possibility of compensating e.g. blue by green water consumption, which is not always possible. Green water is relevant for local ecosystems and agriculture and can reduce the demand for blue irrigation water. However, aggregating blue and green water to single numbers, hides the advantages of rainfed compared to irrigated agriculture. For example, producing 1 t of wheat consumes only 1,123 m³/t in Egypt (= 216 m³green + 907 m³blue) but 1,363 m³/t in Canada (= 1,358 m³green + 5 m³blue) (Mekonnen and Hoekstra 2010). Hence, aggregated numbers conceal the environmental meaning.

4.1.3 Consideration of green water consumption

Concerning the green water consumption of e.g. agricultural plants, it should be noted that natural vegetation also causes evapotranspiration – which can even be higher than the evapotranspiration of agricultural plants (Núñez et al. 2013). For that reason, some authors suggest determining the “net green water footprint”, i.e. the difference in evapotranspiration between agricultural and natural land (SABMiller and WWF 2009). However, the relevance of net or total green water consumption is controversial.

On the one hand, soil moisture is available for local plants only and cannot be used by surrounding eco-systems or withdrawn for human needs. On the other hand, green water is the most important resource for growing crops in global food production. It should be managed well in order to ensure an efficient use of soil moisture (shift from unproductive evaporation to productive transpiration). Since there are many countries, like Algeria or Morocco, which suffer from blue water scarcity but have enough green water to grow crops, a combined consideration is needed (Rockström et al. 2009).

A further relevant question that should be addressed is how the green water footprint affects blue water availability. According to Ridoutt and Pfister (2010), this issue is closely related to land use as land transformation can lead to an altered surface runoff and groundwater recharge. Yet, the consideration of altered blue water availability from land use changes as proposed in Mila i Canals et al. (2008) or in the new international standard (ISO/FDIS 14046 2014) would often result in negative blue water footprints. This can be explained by the fact that groundwater recharge and surface

27 4 Challenges and potential solutions in water footprinting runoff often increase when natural vegetation is transformed into agricultural land, which has led to globally increased river discharges of 7% (Rost et al. 2008). However, this implies that an organization could compensate its blue water footprint by land use changes such as deforestation. It ignores the fact that natural vegetation is an important foundation for the global water cycle and land use changes affect this cycle. According to Rost and colleagues (2008), the increase in runoff resulting from land transformation equals the decrease in evapotranspiration (net green water footprint), which will lead to decreased precipitation in other places. Hence, blue water credits resulting from land use changes should only be given if the consequent decrease in precipitation in other drainage basins is taken into account as well. Specific losses of precipitation in drainage basins, resulting from land transformation and decreased evapotranspiration in another basin, could be determined by means of geographically explicit evaporation recycling ratios as mentioned above (Figure 12). Also, potentially on-site negative impacts resulting from increased runoff such as flooding, increased soil salinity, or water logging should be considered when crediting local blue water increase.

The above discussion of challenges and potential solutions identified on the inventory level is now completed by a discussion of methodological challenges on the impact assessment level comprising midpoint scarcity indicators, the consideration of water quality, environmental credits, and endpoint assessment models.

4.2 Impact assessment challenges

4.2.1 Challenges of midpoint scarcity indicators

Regional freshwater scarcity is the key indicator to assess various consequences of water consumption. Therefore, most midpoint methods use a ratio of annual water withdrawal to availability (Equation 1) to express regional freshwater scarcity. The freshwater availability of a drainage basin (A) expresses the annually renewable freshwater volumes within the basin which can be quantified by means of blue water runoff (plus upstream inflows if the basin is divided into subcatchments). Different terms like water utilization level (Falkenmark 1989), use-to-resource ratio (Raskin et al. 1996), or withdrawal-to-availability ratio (WTA) (Alcamo et al. 2007) can be found for this ratio shown in Equation 1.

Equation 1

This section discusses challenges of scarcity indicators concerning absolute and relative water scarcity, sensitivity to additional water withdrawal, withdrawal vs. consumption based scarcity indicators, and the proper determination of water availability.

28 4 Challenges and potential solutions in water footprinting

4.2.1.1 Absolute vs. relative freshwater scarcity

Freshwater scarcity indicators like the WTA ratio are influenced by two metrics – withdrawal and availability. Consequently, they denote a relative but not an absolute water scarcity. As shown by means of the water stress index (WSI), which was developed by Pfister and colleagues (2009) based on WTA, this can lead to deceptive effects (Figure 13). Countries like Belgium or cities like London, which abound in water, are regarded as critical just because large shares of the renewable supply are used (but not consumed). In contrast, countries in the arid Sahel zone, like Chad, are considered as uncritical because only minor fractions of the small renewable water supplies are used due to a lack of industry and low population density.

In order to overcome this shortcoming, it may be reasonable to consider absolute freshwater scarcity in addition to relative scarcity. For instance, the scarcity indicator could be set to the highest level per se in drainage basins which receive precipitation below e.g. 200 mm/a or are located in hyperarid regions, in which precipitation is less than 5% of potential evapotranspiration (UNEP 1997).

Regions with low absolute

but high relative water

scarcity

Regions with high

absolute but low relative

water scarcity

4.2.1.2 Sensitivity to additional water withdrawal

Closely linked to the discussion of absolute vs. relative water scarcity is the sensitivity of a drainage basin to water scarcity resulting from additional water withdrawal. An assumed annual water use of 1 m³/year and an availability of 10 m³/year would lead to the same result (WTA = 0.1) as an annual water use of 1,000 m³/year and a renewable supply of 10,000 m³/year. However, adding 1 m³ of

29 4 Challenges and potential solutions in water footprinting water use would double the scarcity ratio in the first case (WTA = 0.2) but leave it rather unchanged in the second (WTA = 0.1001). Hence, drainage basins with small renewable water supplies are more sensitive to additional water uses than basins with large renewability rates. It has to be noted that an additional water use of 1 m³/year in the first example is contradictory to the assumption of marginal changes, not influencing the overall environmental situation, which is common practice in life cycle impact assessment (Guinee et al. 2002).

4.2.1.3 Withdrawal vs. consumption based scarcity indicators It is often discussed whether physical freshwater scarcity should be measured by WTA ratios, as applied in e.g. Pfister et al. (2009), or by means of consumption to availability (CTA) indicators (Equation 2), used in recent method developments of Boulay et al. (2011b) and Hoekstra et al. (2011).

Equation 2

On the one hand, WTA tends to overestimate physical water scarcity as it also comprises borrowing and degradative water uses, such as cooling water, which are returned to the basin from which they were withdrawn. On the other hand, there can be water scarcity resulting from competing uses and from freshwater pollution, too. These problems are implicitly covered by WTA as withdrawal includes both consumptive and degradative forms of water use. However, considering freshwater degradation by means of scarcity indexes in addition to water quality indices and emission based impact categories in LCA can lead to an overestimation of impacts as described in section 4.2.2. Even though competing uses of freshwater may lead to scarcity in densely populated areas in some cases, CTA generally seems to express physical water scarcity in a more meaningful way. Moreover, it ensures consistency between the inventory and the impact assessment levels as water consumption is assessed by a characterization factor based on consumption as well.

4.2.1.4 Annual vs. monthly water scarcity

Freshwater scarcity indicators are usually determined based on annual averages of water withdrawal/consumption and availability. However, the hydrological situation in a basin might vary throughout a year (Savenije 2000). For this reason, some authors introduce monthly scarcity factors (Hoekstra et al. 2012) which are especially relevant for agricultural products grown during particular seasons. However, such approaches require temporally explicit inventory data, which is difficult to obtain – especially if complex background systems are involved. As an alternative, consumption weighted annual averages of monthly scarcity factors can be used (Pfister and Baumann 2012). Yet, this does not overcome the key methodological challenge of a monthly scarcity assessment: the consideration of inter-monthly storage capacities which can buffer water scarce periods throughout

30 4 Challenges and potential solutions in water footprinting the year (Pfister and Baumann 2012). Moreover, the temporal resolution of water scarcity assessments also determines the required spatial resolution. Large basins can have flow times of several months from spring to mouth, which makes a monthly assessment difficult.

4.2.1.5 Determination of water availability

Another challenge for water scarcity indicators is the proper determination of water availability. In their current states, WTA or similar indicators only consider the annual renewability rate, i.e. groundwater recharge and surface runoff. However, ground and surface water stocks, such as aquifers or lakes, are neglected. From a resources perspective this is justified as a water use that exceeds the renewability rate leads to a depletion of water reservoirs. This can cause drastic consequences of up to 50 m declined groundwater tables in the High Plains Aquifer (USA) or North China Plains Aquifer (UNEP 2012). However, when consequences on human health and ecosystems are assessed, it is important to consider ground and surface water stocks as they can compensate for an overuse of renewable supplies. Hence, impacts are less severe if reservoirs are available which can buffer water shortage for a certain time. While data on groundwater recharge and surface water runoff is available on high regional resolution (e.g. Alcamo et al. 2003; Döll and Fiedler 2008), geographically explicit figures of total ground and surface water stocks are hard to obtain, which is a clear need for further research.

4.2.2 Consideration of water quality

A further point of concern is the consideration of water quality degradation by means of the gray water concept (e.g. Hoekstra et al. 2011), the use of withdrawal based scarcity ratios (e.g. Pfister et al. 2009), or consideration of quality indicators (e.g. Boulay et al. 2011b). Especially if the water footprint analysis is accomplished in an LCA context, the pollution of water is often covered by other impact categories such as eutrophication, acidification, and eco or human toxicity (Guinee et al. 2002). If quality degradation is considered in traditional impact categories and in water footprint indicators, an overestimation can occur if mutually exclusive impact pathways are considered. For instance, an emission of cadmium to freshwater is considered in the impact category human toxicity by modeling the exposure route of cadmium to humans via fish. Yet, cadmium emissions are also considered in the water footprint method of Boulay and colleagues (2011b), which estimates health damages of malnutrition as a consequence of quality degradation affecting agriculture and fisheries. However, there can be either the negative effect of cadmium uptake caused by eating fish or the negative effect of malnutrition by not having enough fish to eat.

However, the consideration of water quality degradation and emission oriented impact categories does not necessarily lead to an overestimation in all cases. Emission based impact categories neglect

31 4 Challenges and potential solutions in water footprinting the quality of the water withdrawn, the fact that water pollution can increase water scarcity in arid regions, or they take into account different, yet complementary impact pathways. For instance, the human toxicity potential considers the uptake route via potable water (Rosenbaum et al. 2008) and the characterization model of Boulay et al. (2011b) considers malnutrition occurring if water quality degradation precludes irrigation in agriculture. As a lack of drinking water is not considered in Boulay et al. (2011b) and since the exposure to pollutants via agricultural irrigation is not included in the human toxicity potential, these impact pathways are not exclusive but complementary. If impact oriented water footprint methods like Veolia (2011) are applied, which considers quality and generically assesses consequences on human health and ecosystems, it is hard to say whether impacts are overestimated or not as no distinct impact pathways are described. Hence, potential overestimation should be discussed in the interpretation of water footprint studies.

4.2.3 Dealing with environmental credits

Further need for improvement was detected in the question of environmental credits. If, for instance, water is withdrawn from a fossil aquifer and after its use discharged into surface water courses, this water is made available to other users. Thus, depletion of fossil water resources, e.g. in copper mining in Chile, might be beneficial for ecosystems and human needs. Consensus has to be found on how to deal with this trade-off between two areas of protection.

4.2.4 Challenges of endpoint impact assessment methods

Endpoint impact assessment models assess damages resulting from water consumption at the end of the cause-effect chain concerning the areas of protection human health, ecosystems, and resources. Therefore, existing endpoint methods like Pfister et al. (2009) or Boulay et al. (2011b) model particular impact pathways, such as water consumption leads to less water available for irrigation, leading to less productive agriculture, leading to health impacts due to malnutrition. Even though the meaning of impact assessment results obtained from endpoint models is higher than rather abstract midpoint results, uncertainty increases when modeling complex cause-effect chains. Various assumptions are necessary. Regressions used to describe impact pathways are often of low statistical significance. This shows that the link between water scarcity and actual impacts is not straightforward and depends on many more parameters. Thus, the main challenge of endpoint impact assessment methods is to increase understanding of the relationship between water consumption, water scarcity, and damages occurring at the area of protections. Moreover, agreement on the impact pathways to be considered and the modeling procedures is needed as currently different impact assessment methods can lead to different results.

32 4 Challenges and potential solutions in water footprinting

4.3 Volumetric or impact oriented water footprints?

Volumetric water footprint methods like Virtual Water (Allan 1998) and the Water Footprint according to Hoekstra and Hung (2002) provide information about the global freshwater appropriation of products along their life cycles. However, as pure volumetric figures do not allow for assessing the consequences of water consumption, authors like Pfister and Hellweg (2009) and Ridoutt and Huang (2012) conclude that volumetric water footprints are incomplete and can be misleading.

In contrast, scholars favoring the volumetric approach state that global freshwater appropriation is more relevant than regional impacts (Hoekstra et al. 2011). Moreover, Hoekstra and colleagues regard impact oriented water footprints without physical interpretation as “completely meaningless”. In their opinion, regional environmental conditions influencing impacts are poorly described by current impact assessment schemes, which leads to “questionable weighting choices” (Hoekstra et al. (2009) in reply to Pfister and Hellweg (2009)). Furthermore, Mekonnen and Hoekstra (2012) consider volumetric water footprints as consistent with carbon footprints (Finkbeiner 2009). Misinterpreting the meaning of midpoint characterization models, these authors argue that “the carbon footprint is a measure of the amount of greenhouse gases emitted to the environment from human activities and does not describe environmental impacts” (Mekonnen and Hoekstra 2012, p. 187).

It should also be noted that impact oriented water footprint methods provide characterization factors, which are derived from characterization models describing the environmental mechanisms, rather than weighting factors, which include value judgments. Actually the gray water concept used in volumetric water footprint methods, developed by those authors, depends on legal thresholds and, thus, represents a subjective distance-to-target weighting rather than a scientific characterization. In addition, there are many pollutants for which legal limits are lacking. As a consequence, they are simply neglected in such an approach.

Despite their strong criticism, Hoekstra and colleagues (2011) also provide a method to determine “water footprint impact indices”, which equals an impact based approach. However, these impact indices have been developed to enable an additional interpretation – the actual water footprint will stay on the volumetric level.

So, should the water footprint really be a volumetric indicator based on the arguments that environmental mechanisms are hard to model and that required inventory information is difficult to collect? This point of discussion is illustrated in the following conceptual example, in which volumetric and impact oriented water footprints are compared. In order to highlight the analogy, we further assessed greenhouse gas (GHG) emissions on an inventory and impact assessment level.

33 4 Challenges and potential solutions in water footprinting

Figure 14 Conceptual example of water consumption and greenhouse gas emissions

When the two products shown in Figure 14 are assessed on the inventory level by means of volumetric water footprints, product A (1 m³ water consumption) scores better than product B (2 m³ water consumption). However, taking into account regional water scarcity and degree of development, water consumption of 1 m³ in Iran is likely to cause more sever impacts than 2 m³ in Germany.

In order to analyze these potential consequences arising from this water consumption, the following impact assessment methods have been applied:

 Frischknecht et al. (2009) – ecological scarcity method  Motoshita et al. (2008) – human health impacts from malnutrition  Boulay et al. (2011b)2 – water stress index and human health impacts  Motoshita et al. (2011)3 – human health impacts from infectious diseases  Pfister et al. (2009) – freshwater deprivation, human health impacts from malnutrition, ecosystem impacts, resource impacts

As it can be seen in Figure 15, the two methods relying only on physical water scarcity (Frischknecht et al. (2009) and freshwater deprivation included in Pfister et al. (2009)) show similar results since both are based on WTA. In contrast, the water stress index developed by Boulay and colleagues (2011b) indicates a relatively higher relevance of water consumption in Germany because it was determined for a certain water quality and relies on a ratio of water consumption to the statistical low flow (Q90). Q90 expresses the runoff which will be exceeded with a probability of 90% in a long term perspective (ISWS 2013).

2 As the method of Boulay et al. (2011b) requires quality information, the consumption of good quality surface water has been assumed in both cases. 3 Since Motoshita et al. (2011) determined damage factors only for domestic water consumption, these factors have been multiplied by a country specific ratio of domestic to total water use (FAO 2013) to allow for an assessment of general water consumption. Therefore, the method is termed Motoshita et al. (2011)* in the following.

34 4 Challenges and potential solutions in water footprinting

Figure 15 Relative results of volumetric and impact based water footprints of the theoretical example comparing two products causing 1 m³ of water consumption in Iran and Germany respectively

Also damages to human health are evaluated differently. In the methods of Motoshita and colleagues, impacts resulting from water consumption of 2 m³ in Germany are considered more relevant (34 %) in the malnutrition category (Motoshita et al. 2008) than in the infectious diseases model (21 %) (Motoshita et al. 2011*). In contrast, Pfister and colleagues (2009), who also determine damages from undernourishment, consider health impacts in Germany as nil arguing that in countries with a human development index larger than 0.88 malnutrition is not an issue. These different results in two methods assessing malnutrition can be explained by the fact that the method of Motoshita et al. (2008) additionally considers indirect effects. Accordingly, agricultural water scarcity in Germany leads to an import of food from other countries causing impacts outside the country. Without the internalization of such indirect effects, the method would also consider damages resulting from water consumption of 2 m³ in Germany (7.6 E-14 DALY) negligibly low compared to 1 m³ in Iran (1.7 E-9 DALY). The comprehensive characterization model of Boulay et al. (2011b), which assesses damages on human health due to malnutrition and infectious diseases, leads to a result of zero health impacts in Germany considering the high adaptation capacity to water stress.

While wealth reduces human health impacts in the method of Pfister et al. (2009) and Boulay and colleagues (2011b), it cannot avoid ecosystem damages which are relevant in Germany as well. Even though water use is relatively high in an industrialized and densely populated country like Germany (WTA = 0.21 (FAO 2013)), no damage to resources is detected as withdrawal does not exceed the renewability rate (WTA < 1), which avoids the depletion of water stocks.

35 4 Challenges and potential solutions in water footprinting

In contrast to the volumetric water footprint, all characterization models applied show that it is more relevant to consume 1 m³ of water in Iran than 2 m³ in Germany. Hence, without additional qualitative interpretation as proposed (but hardly applied) by Hoekstra and colleagues (2011), volumetric water footprints can be misleading as numerically smaller footprints can cause more severe impacts. Therefore, water footprints should be impact based indicators in the same way as carbon footprints are.

In fact, a volumetric water footprint would equal a mass based carbon footprint, according to which product A would also be preferable as only 1 kg instead of 2 kg GHG are emitted. Obviously, no one would argue for such an approach as the individual contribution of methane and carbon dioxide to global warming needs to be taken into account. This characterization would result in an opposite ranking as product A causes 21 and product B 2 kg CO2-equivalents (IPCC 2007). Consequently, in the same way methane and carbon dioxide emissions cannot be compared on a kilogram level, water consumption in Iran and Germany cannot be compared on a cubic meter level (also shown by Pfister and Hellweg (2009)).

A main difference between carbon and water footprinting is the agreement on characterization models, based on which characterization factors for greenhouse gases and regional water consumption are determined. While the global warming potential is currently established as an internationally agreed characterization model for the carbon footprint (despite inherent weaknesses of the model), there are several characterization models available for water consumption. Even though different models can lead to different results, this cannot be regarded as an inconsistency. It is clear that different methods, which describe different impact pathways of water consumption to human health, ecosystems, and resources, lead to diverse implications as they are influenced by various parameters. This also reflects the complexity of reality as e.g. water consumption in water scarce but developed countries may lead to the depletion of water resources while impacts on human health can be nil. Even within one area of protection, like human health, results can vary as the sensitivity of a population to malnutrition or infectious diseases might differ. Hence, as long as no comprehensive impact oriented water footprint method, which addresses all impact pathways in a consistent manner, is available, several characterization methods should be applied to analyze various implications. Uncertainties in the environmental fate and in the characterization models are a challenge and have to be addressed in carbon and water footprinting. However, they do not justify accomplishing inventory related footprints which can be misleading due to the lack of environmental interpretation.

36 4 Challenges and potential solutions in water footprinting

In this chapter a broad set of methodological challenges and potential solutions has been identified. On the inventory level, the current definition of water consumption neglects atmospheric evaporation recycling effects which is expected to change water consumption figures significantly. On the impact assessment level, challenges have been found especially in midpoint scarcity indexes. This includes for instance, the use of withdrawal versus consumption based indicators, the consideration of ground and surface water stocks in the determination of freshwater availability, and the consideration of absolute in addition to relative freshwater scarcity. Based on the challenges presented along with the potential solutions, a new water footprint method is introduced in the following chapter.

37 5 The water accounting and vulnerability evaluation model – WAVE

5 The water accounting and vulnerability evaluation model – WAVE

In order to tackle some of the methodological shortcomings discussed in the previous section, a new inventory and impact assessment method – the water accounting and vulnerability evaluation (WAVE) model is developed. On the inventory level, a new water accounting approach considers atmospheric evaporation recycling effects and, therefore, allows for a determination of more realistic water consumption figures. In order to translate volumes into potential impacts, the vulnerability of a basin to freshwater resource depletion is evaluated by means of a new blue water scarcity indicator. Hence, WAVE will help to interpret volumetric virtual water studies and can be used as an inventory and characterization model in LCA and water footprinting (Boulay et al. 2013). The methodological development and testing of WAVE are described in the following chapter based on Berger et al. (2014) (appendix, paper IV). Reprinted with permission from Berger, M., R. van der Ent, S. Eisner, V. Bach, and M. Finkbeiner. 2014. Water accounting and vulnerability evaluation (WAVE) – considering atmospheric evaporation recycling and the risk of freshwater depletion in water footprinting. Environmental Science and Technology in press, DOI 10.1021/es404994t. Copyright 2014 American Chemical Society. http://pubs.acs.org/doi/abs/10.1021/es404994t

5.1 Water accounting model

A new inventory method for the accounting of water use is introduced. In addition to freshwater withdrawals (FW) and wastewater discharges (WW) considered in existing inventory schemes (Berger and Finkbeiner 2010), the new water accounting model explicitly considers the share of withdrawal which is consumed due to evapo(transpi)ration (E). Moreover, vapor created synthetically in chemical reactions, e.g. by burning fossil fuels, is regarded in an explicit way (V) (Figure 16).

Figure 16 Water inventory flows along the life cycle of a product considered in WAVE

38 5 The water accounting and vulnerability evaluation model – WAVE

In order to consider the effects of atmospheric evaporation recycling (van der Ent and Savenije

2011), the effective water consumption (WCeff), which represents the sum of effective water consumptions in each basin n (WCeff,n), is introduced.

∑ Equation 3

WCeff,n is determined by subtracting total waste water discharges (WW), evapo(transpi)ration recycling (ER) and synthetically created vapor recycling (VR) from freshwater withdrawals (FW) occurring within basin n.

∑( ) Equation 4

As shown in Equation 5 and Equation 6, the volumes of evapo(transpi)ration and synthetically created vapor recycled within a basin are determined by multiplying the volumes of evapo(transpi)ration (E) and synthetically created vapor (V) with the basin internal evaporation recycling ratio (BIER) and the runoff fraction (α).

Equation 5

Equation 6

According to van der Ent and Savenije (2011), BIER is estimated based on the length of the basin in the direction of the main moisture flux (x) and the basin’s average length scale of the evaporation recycling process (λ).

( )

Equation 7

The length scale of the evaporation recycling process (λ) has been calculated based on an atmospheric water accounting model considering evaporation, precipitation, winds, and humidity (van der Ent et al. 2010; van der Ent and Savenije 2011). For simplification each basin is assumed to be a square with x representing the side length determined via the square root of the basin’s surface area.

BIER has been determined for more than 11,000 basins on a global level (Figure 17) and varies from 0% in the Sahel zone to 38% in the Congo basin. Thus, significant shares of the water consumed in a product system due to evapo(transpi)ration can be returned to the originating drainage basin via precipitation. In the same way, water vapor created in chemical reactions can be returned to the basin of origin to a noticeable extent. In drainage basins located in industrialized countries like USA, Japan, or within the EU BIER ranges from 1 to 10 %. Since the share of evaporation recycling increases with distance, large drainage basins show higher BIER values than small basins when λ is constant.

39 5 The water accounting and vulnerability evaluation model – WAVE

Figure 17 Basin internal evaporation recycling (BIER) ratios denoting the fraction of evaporated water returning to the originating basin via precipitation

It should be noted that only a fraction of the evaporation recycling, which is returned to the originating basin via precipitation, will be available as ground or surface water. Since WAVE focuses on blue water only, the runoff fraction (α) is implemented. Based on data derived from the hydrological model WaterGAP2 (Döll et al. 2003; Flörke et al. 2013), α is determined by relating a drainage basin’s long-term average runoff (R), i.e. groundwater recharge and surface runoff, to the total precipitation (P) within the basin (Equation 8). As shown in Figure 18, α is highest (>80%) in basins located in Alaska and the Himalayas and in the Amazon basin.

Equation 8

Figure 18 Runoff fraction (α) which denotes the ratio of the long-term average runoff (blue water) and precipitation within a drainage basin.

40 5 The water accounting and vulnerability evaluation model – WAVE

The resulting hydrologically effective basin internal evaporation recycling (BIERhydrol-eff), which is obtained by multiplying BIER with α, is shown in Figure 19. Since α is comparably low in Central Africa, large BIER ratios determined in e.g. the Congo basin (38%) are reduced when considering the hydrologically effective fraction (BIERhydrol-eff = 11%). Even though BIERhydrol-eff is below 5% in most of the world’s drainage basins, it reduces blue water consumption significantly in basins in the Himalayas, Alaska, south-east Asia, and the North of South America (10-33%).

Figure 19 Hydrologically effective basin internal evaporation recycling (BIERhydrol-eff) ratios denoting the fractions of evaporated water returning to the originating basin as blue water

5.2 Vulnerability evaluation model

In order to assess consequences resulting from water consumption, many impact assessment models developed in LCA try to describe impacts on the areas of protection resources, ecosystems, and human health (Berger and Finkbeiner 2010). As mentioned in section 4.2.4, some authors model concrete cause-effect chains, like water consumption leads to less water available for irrigation, leading to less productive agriculture, leading to health impacts due to malnutrition (Pfister et al. 2009; Boulay et al. 2011b). However, such endpoint models rely on various assumptions. Moreover, regressions used to describe impact pathways are often of low statistical significance. This shows that the relation between water consumption and impacts – especially on human health and ecosystems – is not straightforward and depends on multiple variables.

Therefore, this work focuses on freshwater resources only and evaluates the regional vulnerability of drainage basins to blue water depletion. This vulnerability approach distinguishes the WAVE model from conventional characterization models describing consequences on resources, such as Pfister et al. (2009). It is not intended to “predict” impacts on freshwater resources but to denote the risk that water consumption in a certain region will lead to freshwater depletion.

41 5 The water accounting and vulnerability evaluation model – WAVE

This risk of freshwater depletion (RFD) can be determined by multiplying the effective water consumption in each basin with its corresponding water depletion index (WDI). WDI denotes the vulnerability of drainage basins to freshwater depletion based on physical blue water scarcity. It can be used to interpret volumetric water footprints on a qualitative level or as characterization factors for impact assessment in water footprinting and LCA.

∑( ) Equation 9

Tackling the shortcomings related to WTA, WDI is based on the consumption-to-availability (CTA) ratio, which relates annual water consumption to annual availability (A). Data for C and A is available in WaterGAP2 for more than 11,000 basins on a global level. As shown in Equation 10, CTA is modified in two steps.

Equation 10

First, annually usable surface water stocks (SWS) are added to A in order to consider lakes, wetlands, and dams in the scarcity index. While storage volumes of dams (Vdam) are available directly (Lehner et al. 2011), volumes of lakes and wetlands are determined by multiplying their surface areas

(Alake/wetland) per basin (Lehner and Döll 2004) with an effective depth (deff = 5 m for lakes and 2 m for wetlands). In order to combine the volumes of dams, lakes, and wetlands (km³) with the flows C and A (km³/a), an annually usable fraction of 1% of the total volumes is used in the determination of SWS (Equation 11). This means that ground and surface water stocks can be used for at least 100 years – even if no renewability occurs.

∑ ( ( )) Equation 11

The effective depths of lakes and wetlands are derived from WaterGAP2 (Döll et al. 2003; Flörke et al. 2013) which represents the only data source for this kind of information on a global level. Just as the time horizon of 100 years, they can be regarded as conservative estimates which are exceeded in many basins. Such methodological choices are in line with the vulnerability approach applied in this work, which aims at assessing the risk that freshwater depletion can occur rather than predicting impacts. In order to evaluate the influence of deff and the time horizon on the final result, a comprehensive sensitivity analysis is accomplished (see section 5.4.5).

In contrast to SWS, volumes of groundwater stocks (GWS) are not available on a global level.

Therefore, an adjustment factor (AFGWS) is introduced, which reduces the scarcity ratio based on the availability of groundwater. Using data provided by WHYMAP (Richts et al. 2011), AFGWS is defined based on geological structure and annual recharge as shown in Table 2. The scarcity reduction rates

42 5 The water accounting and vulnerability evaluation model – WAVE have been derived from discussions with the developers of the WHYMAP (Struckmeier and Richts 2013). In line with the vulnerability approach, moderate reduction rates are selected. Their influence on the final result has been analyzed in a sensitivity analysis presented in section 5.4.5. Fossil groundwater stocks are excluded from this analysis as they cannot be quantified on a global level and it is not sure that they can be accessed in every part of the world.

Table 2 Adjustment factor for ground water stocks (AFGWS) reducing water scarcity based on geological structure and annual recharge

Annual Scarcity

Geological structure recharge (mm) reduction AFGWS

> 300 10.0% 0.900 Major ground water basin 100-300 7.5% 0.925

> 300 5.0% 0.950 Complex hydrogeological structure 100-300 2.5% 0.975

Others - 0.0% 1.000

The factor WDI aims at assessing a basin’s vulnerability to freshwater depletion based on CTA* as shown in Equation 12. It can be understood as an equivalent volume of depleted water resulting from a volume of water consumption.

Equation 12 ( )

Similar to existing scarcity indexes (Pfister et al. 2009; Boulay et al. 2011b), the logistic function plotted in Figure 20 leads to non-linear transformation of physical water scarcity into vulnerability to freshwater depletion. This is important as in the upper and lower ranges of CTA* doubled scarcity does not necessarily lead to doubled vulnerability to depletion. WDI turns 1 above a CTA* of 0.25 (Figure 20), which is regarded as the threshold of extreme water stress (Richter et al. 2011).

43 5 The water accounting and vulnerability evaluation model – WAVE

* Figure 20 Logistic function determining WDI based on CTA ; S-curve leads to larger spreading of WDI in medium scarcity ranges 0.05

Since CTA* expresses a ratio of consumption to availability, the resulting WDI takes into account relative freshwater scarcity only. In order to consider absolute freshwater shortage as well, WDI is set to 1 per se in semi-arid and arid basins (UNEP 1997) shown in Figure 21. This setting is relevant as freshwater resources are highly vulnerable to depletion in (semi-)arid regions regardless of the relative scarcity.

Figure 21 Basins classified as dry subhumid, semi-arid, and arid according to UNEP (1997)

As shown in Figure 22, WDI is at the highest level in many drainage basins located in Central Asia, India, Saudi Arabia, Australia, Northern and Southern Africa, Mexico, the south-west of the USA, and the Andes. In contrast, little or no freshwater resource depletion is caused by water consumption in most basins located in Russia, Canada, Northern Europe, or around the equator.

44 5 The water accounting and vulnerability evaluation model – WAVE

Figure 22 Factors WDI expressing vulnerability of basins to freshwater resource depletion [m³depleted/m³consumed]

In order to ensure the applicability and validity of the water accounting model (section 5.1) and vulnerability evaluation model (this section), BIER, BIERhydrol-eff and WDI factors have been tested in a case study on biofuels, whose results are presented in the following section.

5.3 Application of WAVE in a case study on biofuels

The methodology developed above is tested by means of an existing water footprint study of bioethanol produced from sugar cane in five producing countries (Mekonnen and Hoekstra 2011). Table 3 shows the blue water consumption (evapotranspiration of blue irrigation water) required to produce 1 GJ of bioethanol from sugarcane in Columbia, Mexico, Thailand, Australia, and Zambia. By means of country specific factors for BIERhydrol-eff and WDI, the effective water consumption (WCeff) and the risk of freshwater depletion (RFD) are determined (Table 3). In order to compare the results obtained by means of the WAVE model to those obtained by other impact assessment methods, potential impacts resulting from WCeff are additionally evaluated by means of the models developed by Pfister et al. (2009) and Frischknecht et al. (2009)

It should be noted that a consideration on the country level has its limitations as plants may be grown in particular basin whose hydrogeological situation may differ from the country average. For example sugar cane from Australia is mainly produced in the coastal areas(Griggs 2007) showing less severe water stress than the country average. However, as Mekonnen and Hoekstra (2011) provides data on the country and state level and since we provide BIER and WDI on the level of countries and drainage basins, the country level is the lowest common denominator to be used in this study. Moreover, the testing of the WAVE model and the comparison to other impact assessment methods is regarded more important than the absolute result in this case.

45 5 The water accounting and vulnerability evaluation model – WAVE

Table 3 Analysis of water consumption and effective water consumption required to produce 1 GJ of bioethanol and the resulting risk of freshwater depletion (RFD, WAVE), freshwater deprivation (Pfister et al. 2009), and ecological scarcity (Frischknecht et al. 2009)

Water inventory WAVE Pfister et al. (2009) Frischknecht et al. (2009)

Country WDI Water stress index Freshwater Eco-factor Ecological WC BIER WCeff RFD [m³depleted/ [m³deprived/ deprivation [eco-points/ scarcity [m³] hydrol-eff [m³] [m³depleted] m³consumed] m³consumed] [m³deprived] m³consumed] [eco-points]

Colombia 3.594 10.0% 3.234 0.017 0.053 0.037 0.121 0.000 0.001

Mexico 14.233 0.7% 14.133 0.776 10.971 0.756 10.685 0.280 3.957

Thailand 18.289 4.8% 17.412 0.053 0.916 0.534 9.298 0.440 7.661

Australia 22.695 0.2% 22.649 0.905 20.505 0.402 9.105 0.023 0.521

Zambia 38.904 4.1% 37.401 0.010 0.374 0.012 0.434 0.003 0.101

Figure 23 shows the water consumption figures on a relative scale and presents the reductions in

WCeff compared to WC resulting from the consideration of BIERhydrol-eff. Moreover, potential impacts determined by means of WAVE as well as by the models of Pfister et al. (2009) and Frischknecht et al. (2009) are shown normalized to the highest result in each category (for absolute results see Table 3).

Figure 23 Relative presentation of blue water consumption required to produce 1 GJ of bioethanol from sugar cane,

reduction of water consumption due to consideration of BIERhydrol-eff, and potential impacts determined by means of WAVE and by the impact assessment methods of Pfister et al. (2009) and Frischknecht et al. (2009)

46 5 The water accounting and vulnerability evaluation model – WAVE

On the inventory level, the consideration of the hydrologically effective evaporation recycling by means of BIERhydrol-eff leads to a reduction of water consumption between 0% in Australia and 10% in Columbia. If the total basin internal evaporation recycling (BIER) is taken into account, the recycled fractions of evapotranspirated irrigation water will increase significantly (up to 24% in Zambia).

The assessment of potential impacts resulting from irrigation water consumption in the countries considered leads to different conclusions than a volumetric analysis. All three assessment methods come to the result that the largest water consumption in Zambia (39m³/GJ) actually causes the lowest impacts, as no water scarcity is detected in this country on an annual basis. However, this also shows the limitations of an annual assessment method in a seasonal product system. As discussed in detail in section 4.2.1.4, water consumption and water scarcity can vary throughout the year – especially in countries with dry and wet seasons (Savenije 2000).

The relatively high water consumption of bioethanol produced in Thailand (18 m³/GJ) is evaluated differently in the assessment models. While the risk of freshwater depletion (RFD) is considered low in the WAVE model, significant impacts are expected in the methods of Pfister et al. (2009) and Frischknecht et al. (2009). The reason for this can be found in the underlying methodologies. While a consumption-to-availability ratio is considered in WAVE, the two other impact assessment models are based on a withdrawal-to-availability ratio according to which Thailand is much more water scarce. The three assessment methods also lead to different conclusions regarding bioethanol produced from sugarcane in Australia. While the required irrigation water consumption of 23 m³/GJ causes the highest risk of freshwater depletion in WAVE, it is considered less relevant than production in Mexico and Thailand in the other impact assessment models. In the method of Frischknecht et al. (2009), suggested by the European Union for the product environmental footprint (European Union 2013), water consumption in Australia (23 eco-points/m³) is even regarded far less relevant than in countries abounding in water like Germany (910 eco-points/m³). These unexpected results can be explained by the fact that the methods of Frischknecht et al. (2009) and Pfister et al. (2009) consider relative freshwater scarcity only. Yet, even though only a comparably small fraction of Australia’s water availability is used, the country suffers from absolute freshwater shortage. This highlights the relevance of considering absolute scarcity by means of aridity in the WAVE model.

5.4 Discussion of the WAVE model

In the following section, the WAVE model is discussed concerning scope, the water accounting and vulnerability evaluation methods, inclusion of quality aspects, uncertainties, and application.

47 5 The water accounting and vulnerability evaluation model – WAVE

5.4.1 Scope of WAVE

The WAVE model developed in this work focuses on blue water consumption which occurs mainly due to evapo(transpi)ration or product integration of ground and surface water. While the vulnerability evaluation model is restricted to assess blue water consumption only, BIER and

BIERhyrdol-eff can also be used to assess the basin internal recycling of plant evapotranspiration (green water consumption).

As the WAVE model does not consider water quality degradation in its default version, two possibilities of including water quality aspects are presented. When the method described here is applied in an LCA study, freshwater pollution is assessed by means of impact categories like eutrophication, acidification, or human and eco-toxicity. However, similar to the gray water footprint, this approach does not consider the water quality of freshwater inputs. Therefore, a possibility of determining the quality corrected effective water consumption (WCq,eff,n) which can be used as the basis for calculating the quality corrected risk of freshwater depletion (QRFD), is presented in section 5.4.4.

In the WAVE model, evaporation recycling and freshwater scarcity are determined based on annual averages. As mentioned in the case study, this is a limitation as climatic conditions influencing evaporation recycling as well as the hydrological situation in a basin might vary throughout a year. Especially the combination of these effects can be relevant in semi-arid drainage basins, as BIER may be high in the rainy season when water is abundant but low in the dry season when water scarcity is of concern. However, considering data uncertainty and open methodological questions regarding inter-monthly storage capacities explained in detail in section 4.2.1.4, this work refrains from providing monthly or weighted annual scarcity factors.

5.4.2 Water accounting model

The accounting approach presented in this work considers the basin internal recycling of the share of water withdrawal consumed due to evapo(transpi)ration. Even though this complies with the definition of water consumption (Bayart et al. 2010), it may appear arguable whether the reduction of actual water consumption is reasonable in large basins, like the Danube, where evaporation recycling can occur after hundreds of kilometers. In our opinion such an approach is justified for three reasons. First, a drainage basin delineation of WaterGAP2 is used which divides the world’s 34 largest drainage basins into sub-catchments. This avoids extremely long evaporation recycling distances that would otherwise occur in e.g. the Congo basin. Second, in several countries, like the USA or Australia, water is transported from withdrawal to use through pipelines over long distances. The fraction of withdrawal returned to the river also reduces the water consumption volume in such cases. Hence, atmospheric water transport should be treated in the same way as anthropogenic

48 5 The water accounting and vulnerability evaluation model – WAVE transport. Third, in many basins the downwind transport of evaporated water is in fact an upstream transport from a river perspective as e.g. in the Amazon basin (van der Ent et al. 2010). This means that recycled evaporation is actually returning water, which is available for a possible second consumption.

Nevertheless, additional BIER100 ratios are determined as a sensitivity check by restricting the evaporation recycling distances (x) to 100 km (Equation 7). As shown in Figure 24, BIER100 is significantly lower than BIER (Figure 17) with a maximum of 19% determined in a Colombian drainage basin. Apart from Australia, Northern Africa, Saudi Arabia, and large regions in Central Asia where BIER100 is below 1%, it ranges from 1 – 5% in most of the world’s basins.

BIER100 <1% 1%- 5%- 10%- >15%

Figure 24 Basin internal evaporation recycling (BIER) ratios, determined for a basin side length of 100 km

It should be noted that the model simplification that basins are of quadratic shape leads to an under or overestimation of BIER and BIER100 depending on the actual shape and the prevailing wind directions (Figure 25).

Wind

x x x

a) b) c)

Figure 25 Relation between BIER and wind direction/basin shape leading to: a) BIER as modeled in Figure 17; b) lower BIER as modeled due to shorter evaporation recycling distance x (Equation 6); b) higher BIER as modeled due to longer evaporation recycling distance x (Equation 6) 49 5 The water accounting and vulnerability evaluation model – WAVE

The consideration of basin internal evaporation recycling leads to an interesting effect in agricultural product systems: Even the evapotranspiration of green water by agricultural plants can cause blue water benefits. The reason is that parts of this green water evapotranspiration will be recycled within the drainage basin (BIER), of which parts will be hydrologically effective (α). Moreover, the accounting model developed explicitly considers the emission of water vapor created in chemical reactions and its partial return due to atmospheric moisture recycling effects (Figure 16). Consequently, the combustion of fossil fuels may lead to negative effective water consumption if the synthetically created vapor recycling is higher than the difference between freshwater withdrawals, wastewater discharges, and evaporation recycling (Equation 2).

So far, WAVE has considered the evapo(transpi)ration recycling within drainage basins only leading to a global average BIER of 1%. However, the average continental evapotranspiration returning as continental precipitation amounts to 57% (van der Ent et al. 2010). Thus, the examination of basin internal evaporation recycling (BIER) effects should be extended to a fate of evaporation analysis (Berger and Finkbeiner 2013) which considers the fractions of evapo(transpi)ration returning to other basins as well.

5.4.3 Vulnerability evaluation model

In this work the vulnerability of a drainage basin to freshwater depletion is evaluated. Based on physical water scarcity, the water depletion index (WDI) denotes the risk that water consumption leads to freshwater depletion.

In contrast to most other water scarcity indicators used as impact factors in water footprinting, WDI is based on a consumption instead of withdrawal-to-availability ratio. Even though withdrawal implicitly accounts for quality degradation as well, a consumption based indicator expresses water shortage more realistically as large shares of cooling water, which are returned with low quality degradation due to temperature increase, are excluded.

Moreover, for the first time ground and surface water stocks are included in a water scarcity indicator. As shown in Figure 26, the consideration of aquifers, lakes and wetlands leads to a scarcity reduction of up to 10% in many basins around the globe; especially in Canada, Central Africa, Central Europe, South America, and Russia. Even higher reductions of more than 80% are achieved in small basins in Alaska and the Himalayas. Thus, the consideration of ground and surface water stocks leads to a further scarcity reduction in regions which are under low water stress anyway. Even though the logistic function diminishes this reduction effect in the final WDI result, the difference between water scarce regions, like Saudi Arabia, and regions abounding in water, such as Canada, is increased. This higher precision in water scarcity assessment is especially relevant when comparing water consumption in different regions.

50 5 The water accounting and vulnerability evaluation model – WAVE

Figure 26 Relative changes in CTA scarcity ratio due to the consideration of ground and surface water stocks

By setting WDI to the highest value (1.00) in arid and semi-arid basins, WAVE considers absolute freshwater shortage in addition to relative scarcity. This helps to avoid the mathematical artifact that dry regions are regarded uncritical if consumption is close to zero. Figure 27 presents the influence of this setting on the final WDI result. Without this consideration of absolute freshwater shortage, most of the arid and semi-arid basins (Figure 21) would have significantly lower WDI results than shown in Figure 22. Especially, impacts from water consumption in the Sahel zone or in Australia would be close to zero – as it is the case in existing impact assessment methods like Frischknecht et al. (2009).

Figure 27 Changes in WDI due to the consideration of absolute freshwater shortage in relation to relative scarcity

So far, WAVE has only assessed the vulnerability of drainage basins to freshwater depletion from a blue water resource perspective. In future research also the vulnerability to human health and ecosystem impacts should be analyzed. By means of sensitivity factors the risk that water consumption in water scarce regions can lead to impacts could be analyzed. However, especially

51 5 The water accounting and vulnerability evaluation model – WAVE when assessing the vulnerability to health impacts, the consumption and availability of green water need to be considered in addition to blue water. This combined approach is needed as both types of water are equally important for food production and there are many countries which suffer from blue water scarcity but have enough green water to grow crops (Rockström et al. 2009).

5.4.4 The quality corrected risk of freshwater depletion

In order to consider quality aspects of the water flows entering and leaving the product system (Figure 16), the implementation of water quality indicators (Q) is proposed according to the methodology of the Water Impact Index (Veolia 2011).

A quality corrected effective water consumption (WCq,eff,n) can serve as the basis for calculating the quality corrected risk of freshwater depletion (QRFD).

∑( ) Equation 13

WCq,eff,n considers the quality of each freshwater input (FW), waste water output (WW), evaporation recycling flow (ER), and synthetically created vapor recycling (VR) in each drainage basin n.

∑( ) Equation 14

As suggested by Veolia (2011), water quality indicators can be determined by relating a target concentration of a pollutant (Ctarget,P) to its actual concentration in the water flow (Cactual,P). When several pollutants are of concern, Q is determined based on the pollutant leading to the highest target concentration exceedance.

Equation 15 ( )

Target concentrations can be determined based on legal thresholds, actual concentrations would have to be measured for each water flow entering or leaving the product system. Hence, a detailed water quality consideration which overcomes the limitations of simplified approaches like LCA or gray water causes high efforts and data demands.

5.4.5 Uncertainties in WAVE and sensitivity analysis

Uncertainties in the accounting and in the vulnerability evaluation model of WAVE are difficult to assess as predicted evaporation recycling rates and the risk of freshwater depletion can hardly be calibrated with reality.

As mentioned before, methodological choices are made from a conservative point of view. Hence, rather low effective depths of lakes and wetlands, a long availability time horizon of surface water stocks, and low scarcity reduction rates considering groundwater stocks have been selected. In line

52 5 The water accounting and vulnerability evaluation model – WAVE with the vulnerability approach applied in this work, this is meant to avoid too high scarcity reductions due to the inclusion of ground and surface water stocks. Nevertheless, the influence of these methodological choices can be analyzed by means of sensitivity analyses. As shown in Table 4, eight sensitivity scenarios are considered. First, the values of the effective depths, time horizons, and scarcity reductions are doubled and halved in a separate scenario at a time. Second, a combination of the above-mentioned settings leading to a minimum and a maximum scarcity reduction is analyzed.

Table 4 Scenarios examined in the sensitivity analysis including description and parameter settings

Sensitivity scenario Description Parameter setting

d 0.5x Bisection of effective depths for lakes and deff, lakes = 2.5 m / deff, wetlands = 1 m wetlands

Doubling of effective depths of lakes and d 2x deff, lakes = 10 m / deff, wetlands = 4 m wetlands

Bisection of availability time horizons of surface T 0.5x T = 50 years water stocks

Doubling of availability time horizons of surface T 2x T = 200 years water stocks

Bisection of scarcity reduction rates accounting Bisection of reduction rates shown GW 0.5x for the presence of groundwater stocks in Table 2

Doubling of scarcity reduction rates accounting Doubling of reduction rates shown GW 2x for the presence of groundwater stocks in Table 2

Combination of parameter settings leading to combined min d 2x / T 0.5x / GW 2x minimum scarcity

Combination of parameter settings leading to combined max d 0.5x / T 2x / GW 0.5x minimum scarcity

As it can be seen in Figure 28a and Figure 29b respectively, a bisection of the effective depths or a doubling of the availability time horizon leads to an increase in WDI of 1-10% in some basins located mainly in the USA, Canada, and Russia. Vice versa, a doubling of the effective depths and a bisection of the time horizon causes a scarcity reduction in the same order of magnitude (Figure 28b and Figure 29a). Hence, the settings of the effective depths of lakes and wetlands as well as the choice of the availability time horizon influence WDI only in those basins in which the annually usable fraction of surface water stocks is relevant compared to runoff.

53 5 The water accounting and vulnerability evaluation model – WAVE

Figure 28 Relative increase of default WDI results when (a) the effective depths of lakes/wetlands are halved (2.5 m lakes, 1 m wetlands); (b) the effective depths of lakes/wetlands are doubled (10 m lakes, 4 m wetlands)

b) Figure 29 Relative decrease of default WDI results when (a) the availability time horizon of surface water stocks is halved (50 years); (b) the availability time horizon of surface water stocks is doubled (200 years)

54 5 The water accounting and vulnerability evaluation model – WAVE

When the scarcity reduction rates which consider groundwater stocks are halved or doubled, WDI increases by 1-10% (Figure 30a) or decreases by 1-20% (Figure 30b) in many drainage basins in Europe and in few basins in south-east Asia, Brazil, and the USA.

Figure 30 Relative increase of default WDI results when (a) the scarcity reduction rates accounting for the presence of groundwater are halved compared to the default settings shown in Table 2; (b) the scarcity reduction rates accounting for the presence of groundwater are doubled compared to the default settings shown in Table 2

Since the parameter settings of groundwater stocks influence different drainage basins than those of surface water stocks, the combined minimum and maximum sensitivity scenarios lead to altered WDI results in many basins around the globe (Figure 31). Especially the combined doubling of effective depths and bisection of availability time horizons leads to an amplification of differences between WDI results obtained in these scenarios and the default WDI.

55 5 The water accounting and vulnerability evaluation model – WAVE

Figure 31 Relative decrease of default WDI results when (a) the parameter settings leading to minimum scarcity are combined (double effective depths, half availability time horizon, double scarcity reduction rates, as shown in Table 4); (b) the parameter settings leading to maximum scarcity are combined (half effective depths, double availability time horizon, half scarcity reduction rates, as shown in Table 4)

As it can be seen in Figure 28 to Figure 31, the parameter variation does not influence the WDI results in many (semi-)arid drainage basins as they are set to the highest value (1.00) per se due to the consideration of absolute freshwater shortage. The basis on which absolute water shortage is defined (degree of aridity) influences the number of basins affected significantly (Figure 21). Changes in WDI resulting from the consideration of absolute freshwater scarcity in semi-arid and arid basins are shown in Figure 27.

As mentioned in section 5.4.6, the default values of BIER, BIERhydrol-eff, and WDI are provided on a country level in addition to the level of drainage basins (Appendix). Obviously, uncertainties can be relevant when determining country averages – especially in countries with regions of different water scarcity like the United States or China. Therefore, the minimum and maximum factors of basins within the country are provided in addition to the average factors (Appendix). They can be used to evaluate whether a difference between alternatives is significant or within the uncertainty range.

It should be noted that there are many sources of uncertainty in water footprinting. Starting from inventory databases, significant differences can be found in the water consumption figures of materials and processes (Ecoinvent centre 2013; PE International 2013). Further uncertainties are added in the top-down regionalization required to determine geographically explicit water

56 5 The water accounting and vulnerability evaluation model – WAVE inventories of complex industrial product systems (section 3.5.2.2), which are a prerequisite for impact assessment. Thus, providing quantitative uncertainty figures in the WAVE model alone might pretend a level of precision which does not exist in practice. Therefore, it is recommended to additionally discuss potential uncertainties on a qualitative level by considering the methodological limitations addressed in the discussion.

5.4.6 Application of WAVE

BIER, BIERhydrol-eff, and WDI are determined on the level of drainage basins, as they reflect hydrologic conditions best. All factors are made available by means of a layer file which can be implemented into Google Earth (Dataset S1 in Berger et al. (2014)). Since inventory information is usually not available on such a detailed geographic resolution, all factors are provided on the country level as well (Dataset S2, Appendix). For the determination of country specific BIER, BIERhydrol-eff, and WDI consumption weighted averages are used. This adds higher weight to those basin fractions within a country which contribute a higher share to the country’s total consumption. Uncertainties related to the creation of country averages are discussed in section 5.4.5 and presented in Dataset S2 (Appendix).

BIER, BIERhydrol-eff, and WDI are determined on the level of drainage basins, as they reflect hydrologic conditions best. Since inventory information is often not available on such a detailed geographic resolution, all factors are provided on the country level as well. In order to promote the applicability of the WAVE model, BIER, BIERhydrol-eff, and WDI are made available free of charge on both drainage basin and country levels in a Google Earth layer and spreadsheet, respectively: http://www.see.tu- berlin.de/wave/parameter/en/. For the determination of country specific factors, which are presented in the appendix of this dissertation, consumption weighted averages are used. This adds higher weight to those basin fractions within a country which contribute a higher share to the country’s total consumption. Uncertainties related to the creation of country averages are discussed in section 5.4.5 and quantified in the spreadsheet/appendix.

57 6 Conclusion

6 Conclusion

In this chapter the main conclusions drawn from the review, the case studies, the identification of challenges, and the development of the WAVE model are summarized. Moreover, the remaining challenges which should be addressed in future research are presented.

6.1 Summary of main findings

Water scarcity affects more than 1 billion people and is likely to be of increasing global concern due to climate change, population growth, and changing consumption patterns in developing countries. Therefore, the analysis of water use along product life cycles is relevant in order to identify hotspots in the supply chain, show potential for improvement, and allow for comparisons between alternatives. However, when analyzing the impacts of water use the basic volume is not sufficient, as 1 m³ of groundwater consumption in Saudi Arabia does not compare to 1 m³ of rainwater consumed in Germany. Additional information, such as regional scarcity or sensitivity of the population and ecosystems is needed in order to translate volumes into impacts.

This thesis aims at enhancing the applicability and methodological background of water footprinting by conducting a review, accomplishing industrial case studies, identifying methodological challenges, and developing a new water footprint method.

In a comprehensive literature review more than 30 stand-alone methods, LCA based inventory and impact assessment methods as well as databases and tools have been identified. Water footprint methods range from simple volumetric measures to advanced characterization models which describe complex cause-effect chains. The analysis has shown that most methods focus on consumptive blue water use, while green and gray water use are almost exclusively considered in the water footprint of Hoekstra and colleagues (2002) and the global water tool (WBCSD 2013). Moreover, it is a general trend that the scientifically most advanced methods show higher inventory data requirements. In addition to volume and location, quality information (Boulay et al. 2011b), information on watercourses (Mila i Canals et al. 2008), and even temporal information (Pfister and Baumann 2012) can be required. As this data demand is hard to satisfy – especially when complex background systems are involved – future developments should focus on both more detailed inventory databases and applicable impact assessment methods.

The application of a broad set of water footprint methods in various industrial case studies has confirmed the trade-off between precision and applicability in current method developments. In industrial product systems it is challenging to obtain reliable volumetric figures along the supply chain. Geographical, quality, or even temporal data is not available in today’s LCI databases and can

58 6 Conclusion only be collected with great additional effort. In order to provide at least geographical information, which is a prerequisite for all impact assessment methods, a top-down approach has been developed (section 3.5.2.2). Based on import mix shares, the location of production sites and suppliers, etc. the material specific water consumption is allocated to the respective countries. As illustrated by means of the water footprint study conducted for Volkswagen, such regionalized inventories include many countries because supply chains originate from mining and refining activities of crude oil and metals around the globe.

In the particular case of Volkswagen, water consumption along the life cycles of the Polo, Golf, and Passat occurred in 43 countries amounting to 51.7 m³, 62.4 m³, and 82.9 m³ respectively. Unlike other environmental interventions like global warming or acidification which are dominated by the use phase, the cars’ production causes more than 95% of the total water consumption. Within the production only 10% of the water consumption is caused at Volkswagen’s production sites. Most of the water is consumed in the production of iron, steel, precious metals, and polymers. The application of a broad set of impact assessment methods revealed the limitations of a purely volumetric analysis. While the Golf causes higher water consumption than the Polo, impact assessment results obtained by the method of Pfister and colleagues (2009) are similar in both cars as the Polo is produced in Spain and the Golf is assembled in Germany. Even though the applied top- down regionalization contains several assumptions, a sensitivity analysis revealed relatively stable impact assessment results, as most production processes take place in regions with similar water scarcity in Central Europe. The fact that impact categories evaluating damages to human health are dominated by the water consumption of a few liters in developing countries, highlights the need of impact assessment in water footprinting.

Based on the review and application of various water footprint methods, methodological challenges in water footprinting have been identified and it has been discussed whether the water footprint should be a volumetric or impact oriented indicator. Shortcomings have been detected in the current definition of freshwater consumption. The assumption that water is generally “lost” for a basin due to evapo(transpi)ration is refuted by recent studies showing relevant continental evaporation recycling rates within short time and length scales (van der Ent and Savenije 2011). Moreover, the aggregation of green, blue, and gray water in some volumetric approaches (Allan 1998; Hoekstra and Hung 2002) is questionable as it implies equal relevance and mutual substitutability, for which a scientific rationale is lacking. Especially the importance of green water consumption is controversial, as soil moisture is available for local plants only and cannot be used by surrounding ecosystems or for human withdrawal. However, it is of great importance when considering water scarcity in global food production as there are many countries which suffer from blue water scarcity but not necessarily from green water shortage (Rockström et al. 2009). When the effects of land use changes

59 6 Conclusion on blue water availability are considered, care has to be taken, as land transformation from natural to arable land may lead to increased blue water availability due to altered surface runoff and groundwater recharge patterns. However, water credits from land use changes do not seem justified without considering the potentially adverse effects of increased runoff and without taking into account the relevance of natural vegetation to the global water cycle. On the impact assessment level, most methods rely on ratios of annual water withdrawal/consumption to renewability rate (WTA/CTA) to consider relative regional freshwater scarcity. As absolute freshwater shortage is neglected, arid regions can be regarded as uncritical if only a small fraction of the little renewable supply is used/consumed. Moreover, WTA and CTA do not take into account ground and surface water stocks, which can buffer temporal overexploitation, and do not express the sensitivity of drainage basins to additional withdrawal/consumption. Consequences of water consumption depend on regional scarcity, the type of watercourse used, water quality, the time of withdrawal, as well as the sensitivity of ecosystems and population. For this reason, volumetric water footprints can be misleading without substantial additional interpretation as numerically smaller footprints can cause larger impacts.

Aiming to enhance the analysis of freshwater consumption along products’ life cycles, the water accounting and vulnerability evaluation (WAVE) model has been introduced. On the accounting level, the atmospheric evaporation recycling within drainage basins is considered by means of the basin internal evaporation recycling (BIER). Results show that, depending on the climatic conditions and the size of the basin, up to 38% of evapo(transpi)rated water can return to the originating basin via precipitation. When taking into account the runoff fraction, i.e. the hydrologically effective share of precipitation, 10-33% of evapo(transpi)ration return as blue water in basins in the Himalayas, Alaska, south-east Asia, and the North of South America. In order to allow for an assessment of potential consequences resulting from water consumption, the vulnerability of drainage basins to blue water depletion is evaluated. This vulnerability approach distinguishes the WAVE model from conventional characterization models describing consequences on resources. Rather than “predicting” impacts on freshwater resources, WAVE denotes the risk that water consumption in a certain region will lead to freshwater depletion. This risk of freshwater depletion (RFD) can be determined by multiplying the effective water consumption in each basin with its corresponding water depletion index (WDI). WDI denotes the vulnerability of drainage basins to freshwater depletion based on physical blue water scarcity. In contrast to previous works, water scarcity is measured based on a consumption rather than a withdrawal-to-availability ratio and accounts for the presence of ground and surface water stocks. In order to consider absolute freshwater shortage in addition to relative scarcity, WDI is automatically set to the highest value in (semi-)arid basins. WDI is at the highest level in many drainage basins located in Central Asia, India, Saudi Arabia, Australia, Northern and Southern Africa,

60 6 Conclusion

Mexico, the south-west of the USA, and the Andes. In contrast, little or no freshwater resource depletion is caused by water consumption in most basins located in Russia, Canada, Northern Europe, or around the equator. As illustrated in a case study of biofuels, the analysis of the risk of freshwater depletion can lead to different conclusions than purely volumetric water footprint figures and overcomes shortcomings of existing indicators. Thus, by adding environmental meaning to freshwater consumption volumes, WAVE can support scientists in other disciplines and decision makers in mitigating global water stress.

6.2 Remaining challenges in water footprinting

The water accounting and vulnerability evaluation (WAVE) model presented in the previous chapter has overcome several methodological challenges described in chapter 4. For instance, WAVE considers evaporation recycling, ground and surface water stocks, as well as the vulnerability of arid basins. Nevertheless, there remain methodological and practical challenges which have to be addressed by future research.

The consideration of evaporation recycling in the WAVE model is a promising approach to determine more realistic water consumption figures. As mentioned in section 5.4.2, the current model should be refined in order to consider the real shape of the basin and the prevailing wind directions. In the long term it would be desirable to extend the basin internal evaporation recycling (BIER) analysis to a fate of evaporation analysis. In addition to analyzing the fraction of evaporation recycled within a drainage basin, the share of evaporation returned to other basins and the sea could be determined in such an approach. Hence water consumption in one drainage basin might cause benefits in other basins. A comprehensive fate of evaporation analysis would allow for assessing consequences of land use change on the global water cycle. A reduced evapotranspiration, resulting from e.g. the transformation of forests to agricultural land, will lead to decreased precipitation in other regions which could be quantified by such an approach. In a current cooperation between TU Delft and the Potsdam Institute for Climate Impact Research, such fate of evaporation analysis is being conducted for the Amazon basin (Zemp et al. 2013). Results will show whether a global consideration will be possible in the future.

Even though the accounting of green water is controversial as explained in section 4.1, the relation between green and blue water should be analyzed in more detail. Currently, green and blue water are regarded as isolated types of water. However, in reality there can be a connection between them as green water can turn into blue water and vice versa. Hence, the consumption of green water might influence the availability of blue water. A deeper understanding of this relationship will help to analyze the consequences of land use changes, which affect both green and blue water consumption and availability.

61 6 Conclusion

As highlighted in the case studies (chapter 3), the application of impact assessment methods is challenging since no geographical information is available in current LCI databases. In an attempt to incorporate the WAVE model into the GaBi software, PE International is currently trying to provide such regionalized water consumption figures by analyzing the supply chains modeled in their master database. Geographically explicit inventory databases are the most promising option in the long term. In the meantime a top-down regionalization based on import mix shares, locations of suppliers, etc. can be an alternative to enable the application of impact assessment tools. For this reason, the regionalization model developed in this work (section 3.5.2.2), which represents the supply situation for Volkswagen, should be generalized to be applicable for other studies as well.

A further methodological challenge is the inclusion of water quality aspects in both inventory and impact assessment approaches. Even though a potential solution is presented in this work (section 5.4.4), resulting high inventory data demands are hard to satisfy. This could be reduced by the implementation of quality categories (Boulay et al. 2011a). Yet, care has to be taken to avoid potential overestimation when assessing water use in LCA studies in which impact categories like eutrophication or human and eco toxicity are included. As mentioned in section 4.2.2, mutually exclusive impact pathways have to be avoided. For instance, cadmium polluted water could either cause toxic effects by eating fish or malnutrition by not having enough fish to eat but not both.

Finally, impact assessment methods – especially those modeling endpoint effects – share the challenge that the relations between water consumption in water scarce regions and damages to human health and ecosystems are rather uncertain. As shown in a regression analysis (Figure 32), adverse effects of malnutrition or infectious diseases can be excluded in developed countries (HDI above 0.9). However, below a HDI of 0.8, health damages can occur in countries with low and high risk to freshwater depletion in equal measure. This shows that further basic research is needed to understand the relationship between water consumption, water scarcity, and impacts. In particular when assessing the vulnerability to health impacts, the consumption and availability of green water need to be considered in addition to blue water. This combined approach is needed as both types of water are equally important for food production and there are many countries, like Algeria or Morocco, which suffer from blue water scarcity but have enough green water to grow crops (Rockström et al. 2009).

62 6 Conclusion

Caption: WDI ≤ 0.1 0.1 < WDI ≤ 0.5 0.5 < WDI ≤ 1.0

Figure 32 Regression between Human Development Index (HDI) and human health damages resulting from malnutrition (left) and diarrhoeal diseases (right) measured in disability adjusted life years (DALY). Results are presented for three ranges of the water depletion index (WDI)

The methodological and practical shortcomings described above are revisited in a current research proposal which will be submitted to the Deutsche Forschungsgesellschaft in the near future. It is hoped that these challenges can be tackled in cooperation with other institutes in a three-year research project.

63 7 Outlook

7 Outlook

After presenting the conclusions drawn from this thesis and the remaining challenges, this chapter provides an outlook on methodological trends, the international standardization process, and the future relevance of the water footprint.

7.1 Methodological trends

As revealed by the review of a broad set of water footprint methods (Berger and Finkbeiner 2010, chapter 2), the scientific advancement of impact assessment methods often leads to inventory requirements which cannot be satisfied by today’s inventory databases. For instance, recent impact assessment developments by e.g. Boulay et al. (2011b), Veolia (2011), or Pfister and Baumann (2012) require geographical, water quality, and even temporal information. The rather limited application of these approaches compared to methods with relatively low inventory requirements such as Pfister et al. (2009) highlights the trade-off between scientific precision and applicability. Even though the WAVE model demands only volumetric and geographical information, it is expected that the trend towards data demanding impact assessment methods is ongoing. Therefore, it is urgently necessary that LCI databases make such data available in the long term.

Since about 20 different inventory and impact assessment methods have been developed for LCA, which model various cause-effect chains and address different areas of protection, the WULCA group of the UNEP-SETAC Life Cycle Initiative started to develop a consensus model. Similar to USEtox in the toxicity impact modeling (Rosenbaum et al. 2008), a group of international researchers aims at developing a model which addresses all relevant impact pathways of water use and provides a harmonized and consistent impact assessment.

7.2 The international standard on water footprinting – ISO/FDIS 14046

In addition to the consensus impact assessment model mentioned above, the international community is finalizing an international standard on water footprinting (ISO/FDIS 14046 2014). Aiming at “providing transparency, consistency, and credibility for assessing water footprint and reporting water footprint results of products, processes or organizations” the standard includes “principles, requirements and guidelines” on water footprinting. After defining a consistent terminology and describing underlying principles, the methodological framework is presented and guidance on reporting and critical review is provided.

64 7 Outlook

In line with the LCA structure (ISO 14044 2006), the framework of a water footprint analysis comprises the goal and scope definition, the water footprint inventory analysis, the water footprint impact assessment, and the interpretation of results.

The standard clearly states that the water footprint assessment is an impact based measure. Contrary to the definition of Hoekstra and colleagues (2011), a water footprint inventory can be reported but shall not be termed “water footprint”. It is stated that a water footprint assessment can be used as both a stand-alone analysis and part of an LCA containing additional environmental information. The water footprint assessment should be a comprehensive analysis comprising water availability and water pollution aspects. If only single aspects of such a comprehensive analysis are taken into account, this should be reflected in the name of the study. For example, a “water availability footprint” considers only the volume of water consumed and the resulting impacts. In contrast, a “water eutrophication footprint” assesses the impacts of eutrophication caused by water pollution but neglects the consumed volume.

Rather than proposing a specific inventory and impact assessment method, the standard defines criteria which have to be fulfilled in an ISO compliant water footprint study. For instance, elementary flows should include information concerning quantities, type of watercourse, water quality, types of water use, geographical location, time, and emissions. In impact assessment, water availability footprints should be determined by means of characterization models assessing “the contribution of the product, process or organization to pressure on water availability”. In a similar way, water footprints addressing water degradation should be determined by characterization models describing “the contribution of the product, process or organization to impacts related to water degradation”. The preferred water footprint profile contains several impact categories measuring water availability and degradation footprints.

According to this definition, the WAVE model developed in this work can be used as a characterization model for water availability footprints. In order to obtain a comprehensive water footprint profile it could be extended by quality parameters (Q) enabling the determination of a quality corrected effective water consumption as described in section 5.4.4. Alternatively, conventional impact categories, like eutrophication or human and eco-toxicity (Guinee et al. 2002), could be evaluated in addition to WAVE.

7.3 Water footprint – cure or tranquilizer?

As illustrated in this dissertation, the water footprint has developed considerably from a simple volumetric measure to an advanced impact assessment tool which is applicable even in complex industrial case studies. Hence, water footprint results are of increasing robustness and can support

65 7 Outlook stakeholders in industry and politics when analyzing technical or political options. However, there remain questions which are discussed in the following chapters:

1) What actions should be taken based on water footprint results?

and

2) Can the water footprint help to mitigate global water stress?

7.3.1 Actions to be taken based on water footprint results

Considering the water footprint results of case studies accomplished in this dissertation and by other researchers, one might conclude that e.g. cotton textiles should be avoided completely (Chapagain et al. 2006), no wheat should be imported from Morocco (Pfister et al. 2011b), and biofuels produced in water scarce countries (Mekonnen and Hoekstra 2011) should be banned.

Even though this could be preferable from a pure water perspective, there are other aspects which should be considered, too. As shown in Figure 33, water consumption is only one aspect among other environmental interventions, such as global warming or human toxicity. In some cases, e.g. the comparison of biofuels to fossil fuels (Berger et al. 2012), the water and carbon footprints can lead to opposing preferences. Such trade-offs between indicators become even more relevant when including the economic and social dimensions in life cycle sustainability assessment studies (Finkbeiner et al. 2010). Thus, methodological developments like WAVE can help to increase precision and reliability when assessing impacts of water consumption but should not be used as a sole basis for decision-making.

Figure 33 Water consumption as one indicator among others in the context of sustainability

66 7 Outlook

As learned from the case studies described in chapter 3, companies are often not primarily interested in absolute water footprint figures but want to understand where relevant water consumption occurs within the supply chain of their products in terms of volume and impacts. Based on this information, companies can analyze the hotspots in greater detail as identified potential impacts do not necessarily mean that there are real damages.

Such a more detailed analysis can be facilitated by e.g. the water stewardship concept (EWS 2013) which analyzes the water consumption of an organization in greater detail. Taking into account the specific local situation, water stewardship approaches evaluate the environmental, operational, legal, and reputational risks associated with an organization’s water consumption. Supporting an “out of the fence approach”, water stewardship identifies opportunities and solutions in cooperation with the public, authorities, and other water users within the basin. In addition to reduction and recycling options, solutions to reduce impacts of an organization’s water use can also include offsetting measures, such as rainwater collection or drinking water purification projects within the basin.

Similar to the differences between environmental impact assessment (EIA) and life cycle assessment (LCA), water stewardship analyzes water consumption and the resulting consequences in greater detail – but can hardly be applied throughout a complex supply chain. Therefore, it can be a promising symbiosis to identify potential hotspots by means of the water footprint and analyze real risks and opportunities by means of water stewardship projects.

7.3.2 The water footprint – a means of mitigating global water stress?

The question whether environmental assessment methods like LCA (ISO 14044 2006), the carbon footprint (Finkbeiner 2009), or the water footprint (ISO/FDIS 14046 2014) can really help to reduce environmental impacts is difficult to answer. The water footprint can support decision-makers in analyzing potential impacts resulting from water consumption throughout the life cycle of products. This information can be used to identify hotspots, reveal potential for improvement, set reduction targets, and compare alternatives. Especially in combination with water stewardship activities, actions can be taken which reduce impacts on freshwater resources, ecosystems, and human health.

Therefore, the key question is whether this potential is really utilized. So far, water footprints have mainly been determined to inform stakeholders about the volumes and resulting impacts of water consumption occurring along the supply chain of products or business activities. Even though this awareness raising is relevant as such, it is crucial for the water footprint to take the next step: from information to actions that will reduce negative consequences of water consumption. In order to achieve this goal, the water footprint needs to become a management instrument. Similar to LCA, which is successfully implemented as a research and development tool in some companies

67 7 Outlook

(Finkbeiner et al. 2001), water footprint results can become one aspect in the complex decision- making process. In addition to numerous conventional parameters like costs, design, or quality, impacts resulting from a product’s water consumption will then need to be considered in every decision.

It is hoped that the practical applications and methodological enhancements presented in this work will support this transition of the water footprint from an awareness raising to a decision tool which will help to mitigate global water stress.

68 8 References

8 References

Alcamo, J. and T. Henrichs. 2002. Critical regions: A model-based estimation of world water resources sensitive to global changes. Aquatic Sciences 64: 352-362.

Alcamo, J., M. Flörke, and M. Märker. 2007. Future long-term changes in global water resources driven by socio-economic and climatic changes. Hydrological Sciences Journal 52(2): 247-275.

Alcamo, J., P. Doll, T. Heinrichs, F. Kaspar, B. Lehner, T. Rosch, and S. Siebert. 2003. Development and testing of the WaterGAP 2 global model of water use and availability. Hydrological Sciences Journal 48(3): 317-337.

Allan, J. A. 1998. Virtual water: a strategic resource, global solutions to regional deficits. Ground Water 36(4): 545-546.

Amores, M. J., F. Verones, C. Raptis, R. Juraske, S. Pfister, F. Stoessel, A. Ant n, F. Castells, and S. Hellweg. 2013. Biodiversity Impacts from Salinity Increase in a Coastal Wetland. Environmental Science and Technology 47(12): 6384−6392.

Bayart, J. B., C. Bulle, M. Margni, F. Vince, L. Deschenes, and E. Aoustin. 2009. Operational characterisation method and factors for a new midpoint impact category: freshwater deprivation for human uses. In Proceedings of the SETAC Europe: 19th Annual Meeting May 31 - June 04, 2009, Gothenborg, Sweden.

Bayart, J. B., C. Bulle, A. Koehler, M. Margni, S. Pfister, F. Vince, and L. Deschenes. 2010. A framework for assessing off-stream freshwater use in LCA. International Journal of Life Cycle Assessment 15(5): 439-453.

Berger, M. and M. Finkbeiner. 2010. Water footprinting - how to address water use in life cycle assessment? Sustainability 2(4): 919-944.

Berger, M. and M. Finkbeiner. 2013. Methodological challenges in volumetric and impact oriented water footprints. Journal of Industrial Ecology 17(1): 79-89.

Berger, M., J. Warsen, S. Krinke, V. Bach, and M. Finkbeiner. 2012. Water Footprint of European Cars: Potential Impacts of Water Consumption along Automobile Life Cycles. Environmental Science and Technology 46(7): 4091-4099.

Berger, M., R. van der Ent, S. Eisner, V. Bach, and M. Finkbeiner. 2014. Water accounting and vulnerability evaluation (WAVE) – considering atmospheric evaporation recycling and the risk of freshwater depletion in water footprinting. Environmental Science and Technology in press, DOI 10.1021/es404994t.

69 8 References

Bösch, M. E., S. Hellweg, M. A. J. Huijbregts, and R. Frischknecht. 2007. Applying Cumulative Exergy Demand (CExD) Indicators to the ecoinvent Database. International Journal of Life Cycle Assessment 12: 181-190.

Boulay, A.-M., A. Y. Hoekstra, and S. Vionnet. 2013. Complementarities of Water-Focused Life Cycle Assessment and Water Footprint Assessment. Environmental Science & Technology 47(21): 11926–11927.

Boulay, A.-M., C. Bouchard, C. Bulle, L. Deschenes, and M. Margni. 2011a. Categorizing water for LCA inventory. International Journal of Life Cycle Assessment 16(7): 639-651.

Boulay, A.-M., C. Bulle, J.-B. Bayart, L. Deschenes, and M. Margni. 2011b. Regional Characterization of Freshwater Use in LCA: Modelling Direct Impacts on Human Health. Environmental Science and Technology 45(20): 8948–8957.

Brent, A. C. 2004. A life cycle impact assessment procedure with resource groups as area of protection. International Journal of Life Cycle Assessment 9: 172-179.

Chapagain, A. K. and A. Y. Hoekstra. 2007. The water footprint of coffee and tea consumption in the Netherlands. Ecological Economics 1(64): 109-118.

Chapagain, A. K., A. Y. Hoekstra, H. H. G. Savenije, and R. Gautam. 2006. The water footprint of cotton consumption: An assessment of the impact of worldwide consumption of cotton products on the water resources in the cotton producing countries. Ecological Economics 60(1): 186-203.

CSO. 2013. The Corporate Water Gauge. Available at http://www.sustainableorganizations.org. Center for Sustainable Organizations. Accessed July 30, 2013.

Die Zeit. 2009. Wie viel Wasser verbrauchen wir? Available at http://www.zeit.de/online/2009/25/infografik-wasser. Accessed August 09, 2013.

Dirmeyer, P. A. and K. L. Brubaker. 2007. Characterization of the Global Hydrologic Cycle from a Back- Trajectory Analysis of Atmospheric Water Vapor. Journal of Hydrometeorology 8: 20-37.

Döll, P. and K. Fiedler. 2008. Global-scale modeling of groundwater recharge. Hydrology and Earth System Sciences 12(3): 863–885.

Döll, P., F. Kaspar, and B. Lehner. 2003. A Global Hydrological Model for Deriving Water Availability Indicators: Model Tuning and Validation. Journal of Hydrology 270(1-2): 105-134.

Ecoinvent centre. 2013. Ecoinvent 3 LCI databse. Available at http://www.ecoinvent.ch/. Accessed July 30, 2013.

70 8 References

European Union. 2013. Commission Recommendation of 9 April 2013 on the use of common methods to measure and communicate the life cycle environmental performance of products and organisations. Official Journal of the European Union L124, 56.

EWS. 2013. European Water Stewardship. Available at http://www.ewp.eu/activities/ews/Global. Accessed September 18, 2013.

Falkenmark, M. 1989. The massive water scarcity now threatening Africa: Why isn't it being addressed? Ambio 18(2): 112-118.

FAO. 2013. FAOSTAT. Available at http://faostat.fao.org/. Food and Agricultural Organization of the United Nations. Accessed July 30, 2013.

Finkbeiner, M. 2009. Carbon footprinting—opportunities and threats. Internation Journal of Life Cycle assessment 14(2): 91-94.

Finkbeiner, M., K. Ruhland, H. Cetiner, and B. Stark. 2001. Life Cycle Engineering and Design for Environment of the Mercedes-Benz C-Class. In 2001 Environmental Sustainability Conference & Exhibition. November 12-14, 2001, Graz, Austria

Finkbeiner, M., E. M. Schau, A. Lehmann, and M. Traverso. 2010. Towards Life Cycle Sustainability Assessment. Sustainability 2(10): 3309-33223322.

Flörke, M., E. Kynast, I. Bärlund, S. Eisner, F. Wimmer, and J. Alcamo. 2013. Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study. Global Environmental Change 23(1): 144-156.

Frischknecht, R., R. Steiner, and N. Jungbluth. 2009. The Ecological Scarcity Method - Eco-Factors 2006 - A method for impact assessment in LCA. Bern, Swizerland: Federal Office for the Environment.

GEMI. 2013a. The Local Water Tool. Available at http://www.gemi.org/localwatertool/. Global Environmental Management Initiative. Accessed September 18, 2013.

GEMI. 2013b. Connecting the Drops. Available at http://www.gemi.org/water/overview.htm. Global Environmental Management Initiative. Accessed September 18, 2013.

GEMI. 2013c. Collecting the Drops. Available at http://www.gemi.org/waterplanner/. Global Environmental Management Initiative. Accessed September 18, 2013.

Gerbens-Leenes, W., A. Y. Hoekstra, and T. H. Van der Meer. 2009. The water footprint of bioenergy. Proceedings of the National academy of Sciences of the United States of America 106(25): 10219-10223

71 8 References

Gleick, P. H. 1996. Water resources. In Encyclopedia of Climate and Weather, edited by S. H. Schneider. New York: Oxford University Press.

Goedkoop, M. and R. Spriensma. 2001. The Eco-indicator 99 - A damage oriented method for Life Cycle Impact Assessment, Methodology Report. Amersfoort, The Netherlands: Product Ecology Consultants (PRe).

Google Inc. 2011. Google Earth. Available at http://earth.google.com/. Accessed January 07, 2011.

Griggs, P. D. 2007. Deforestation and sugar cane growing in Eastern Australia. Environment and history 13(3): 255-283.

Guinee, J. B., M. Gorree, R. Heijungs, G. Huppes, R. Kleijn, A. de Koning, L. van Oers, A. W. Sleeswijk, S. Suh, H. A. Udo de Haes, H. de Bruijn, R. van Duin, and M. A. J. Huijbregts. 2002. Life cycle assessment - An operational guide to the ISO standards. Dordrecht, The Netherlands: Kulwer Academic Publishers.

Hanafiah, M. M., M. A. Xenopoulos, S. Pfister, R. S. E. W. Leuven, and M. A. J. Huijbregts. 2011. Characterization Factors for Water Consumption and Greenhouse Gas Emissions Based on Freshwater Fish Species Extinction. Environmental Science and Technology 45(12): 5272– 5278.

Hauschild, M. and H. Wenzel. 1998. Environmental Assessment of Products. Vol. 2. New York, NY, USA: Chapman & Hall, Thomson Science.

Hoekstra, A. Y. and P. Q. Hung. 2002. Virtual Water Trade: A Quantification of Virtual Water Flows between Nations in Relation to International Crop Trade. In Value of Water Research Report Series 11. Delft, The Netherlands: UNESCO-IHE.

Hoekstra, A. Y. and M. M. Mekonnen. 2012. Reply to Ridoutt and Huang: From water footprint assessment to policy. Proceedings of the National academy of Sciences of the United States of America 109(22): E1425.

Hoekstra, A. Y., W. Gerbens-Leenes, and T. H. van der Meer. 2009. Reply to Pfister and Hellweg: Water footprint accounting, impact assessment, and life-cycle assessment. Proceedings of the National academy of Sciences of the United States of America 106(40): 114.

Hoekstra, A. Y., A. K. Chapagain, M. M. Aldaya, and M. M. Mekonnen. 2011. The Water Footprint Assessment Manual - Setting the Global Standard. London, Washington, DC: earthscan.

Hoekstra, A. Y., M. M. Mekonnen, A. K. Chapagain, R. E. Mathews, and B. D. Richter. 2012. Global Monthly Water Scarcity: Blue Water Footprints versus Blue Water Availability. PLoS One 7(2): e32688.

72 8 References

Hotze, M. 2013. Recognizing Environmental Science & Technology’s Contributors: Top Papers 2012. Environmental Science and Technology 47(10): 4961-4962.

Humbert, S. and R. Maendly. 2008. Characterization factors for damage to aquatic biodiversity caused by water use especially from dams used for hydropower. In 35 LCA forum Zurich - Water in LCA. June 5, 2008, Zurich, Switzerland.

IPCC. 2007. Climate Change 2007 - The Fourth Assessment Report. United Nations, Intergovernmental Panel on Climate Change.

ISO 14044. 2006. Environmental management - Life cycle assessment - Requirements and guidelines (ISO 14044:2006), edited by International Organisation for Standardisation. Geneva, Switzerland.

ISO/FDIS 14046. 2014. Water footprint - principles, requirements and guidance edited by International Organization for Standardization. Geneva, Switzerland.

ISWS. 2013. Illinois Streamflow Assessment Model. Available at http://www.sws.uiuc.edu/data/ilsam/definitions.asp. Illinois State Water Survey. Accessed October 22, 2013.

JRC-IES. 2011. ILCD handbook general guide for life cycle assessment—recommendations for life cycle impact assessment in the European context—based on existing environmental impact assessment models and factors. Luxembourg: Joint Research Centre - Institute for Environment and Sustainability.

Koffler, C., S. Krinke, L. Schebek, and J. Buchgeister. 2008. Volkswagen slimLCI - a procedure for stream-lined inventory modelling within Life Cycle Assessment (LCA) of vehicles. International Journal of Vehicel Design 46(2): 172-188 (special issue).

Kounina, A., M. Margni, J.-B. Bayart, A.-M. Boulay, M. Berger, C. Bulle, R. Frischknecht, A. Koehler, L. M. i. Canals, M. Motoshita, M. Núñez, G. Peters, S. Pfister, B. Ridoutt, R. v. Zelm, F. Verones, and S. Humbert. 2013. Review of methods addressing freshwater use in life cycle inventory and impact assessment. International Journal of Life Cycle Assessment 18(3): 707-721.

Lehner, B. and P. Döll. 2004. Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology 296(1-4): 1-22.

Lehner, B., C. R. Liermann, C. Revenga, C. Vörösmarty, B. Fekete, P. Crouzet, P. Döll, M. Endejan, K. Frenken, J. Magome, C. Nilsson, J. C. Robertson, R. Rödel, N. Sindorf, and D. Wisser. 2011. High resolution mapping of the world’s reservoirs and dams for sustainable river flow management. Frontiers in Ecology and the Environment 9(9): 494–502.

73 8 References

Mekonnen, M. M. and A. Y. Hoekstra. 2010. A global and high-resolution assessment of the green, blue and grey water footprint of wheat. Hydrology and Earth System Sciences 14(7): 1259– 1276.

Mekonnen, M. M. and A. Y. Hoekstra. 2011. The green, blue and grey water footprint of crops and derived crop products. Hydrology and Earth System Sciences 15(5): 1577-1600.

Mekonnen, M. M. and A. Y. Hoekstra. 2012. The blue water footprint of electricity from hydropower. Hydrology and Earth System Sciences 16(1): 179-187.

Mila i Canals, L., J. Chenoweth, A. Chapagain, S. Orr, A. Anton, and R. Clift. 2008. Assessing freshwater use in LCA: Part I - inventory modelling and characterisation factors for the main impact pathways. International Journal of Life Cycle Assessment 14(1): 28-42.

Motoshita, M., N. Itsubo, and A. Inaba. 2008. Development of impact assessment method on health damages of undernourishment related to agricultural water scarcity. In Proceedings of EcoBalance 2008 - The 8th International Conference on EcoBalance. December 10-12, 2008, Tokyo, Japan.

Motoshita, M., N. Itsubo, and A. Inaba. 2011. Development of impact factors on damage to health by infectious diseases caused by domestic water scarcity. International Journal of Life Cycle Assessment 16(1): 65-73.

Murray, C. and A. Lopez. 1996. The Global Burden of Disease. Boston, MA, USA: Harvard School of Public Health (on behalf of the World Health Organization, World Bank).

New, M., M. Todd, M. Hulme, and P. Jones. 2001. Precipitation Measurements and Trends in the Twentieth Century. International Journal of Climatology 21(15): 1899-1922.

Núñez, M., S. Pfister, P. Roux, and A. Antón. 2013. Estimating Water Consumption of Potential Natural Vegetation on Global Dry Lands: Building an LCA Framework for Green Water Flows. Environmental Science and Technology 47(21): 12258−12265.

Ono, Y., Y. D. Kim, K. Horiguchi, and N. Itsubo. 2012. The development of water inventory database considering water consumption. In Proceedings of EcoBalance 2012 - The 10th International Conference on EcoBalance. November 21-23, 2012, Yokohama, Japan.

Owens, J. W. 2001. Water Resources in Life Cycle Impact Assessment - Considerations in Choosing Category Indicators. Journal of Industrial Ecology 5(2): 37-54.

PE International. 2013. GaBi LCI database. Available at http://www.gabi-software.com. Accessed March 8, 2013.

74 8 References

Pfister, S. and S. Hellweg. 2009. The water ‘‘shoesize’’ vs. footprint of bioenergy. Proceedings of the National academy of Sciences of the United States of America 106(35): E93-E94.

Pfister, S. and J. Baumann. 2012. Monthly characterization factors for water consumption and application to temporally explicit cereals inventory. In 8th International Conference on LCA in the Agri-Food Sector. Rennes, France, 2-4 October 2012.

Pfister, S., A. Koehler, and S. Hellweg. 2009. Assessing the environmental impacts of freshwater consumption in LCA. Environmental Science and Technology 43(11): 4098-4104.

Pfister, S., D. Saner, and A. Koehler. 2011a. The environmental relevance of freshwater consumption in global power production International Journal of Life Cycle Assessment 16(6): 580-591.

Pfister, S., P. Bayer, A. Koehler, and S. Hellweg. 2011b. Environmental Impacts of Water Use in Global Crop Production: Hotspots and Trade-Offs with Land Use. Environmental Science and Technology 45(13): 5761–5768.

Quantis. 2013. The Quantis Water Database. Available at http://www.quantis- intl.com/waterdatabase.php. Accessed July 30, 2013.

Raskin, P. D., E. Hansen, and R. M. Margolis. 1996. Water and sustainability: global patterns and long- range problems. Natural Resources Forum 20(1): 1-5.

Richter, B. D., M. M. Davis, C. Apse, and C. Konrad. 2011. Short communication: A presumptive standard for environmental flow protection. River Research and Applications DOI: 10.1002/rra.1511.

Richts, A., W. Struckmeier, and M. Zaepke. 2011. WHYMAP and the Groundwater Resources of the World 1:25,000,000. In Sustaining Groundwater Resources, edited by J. A. A. Jones. Heidelberg: Springer.

Ridoutt, B. G. and S. Pfister. 2010. A revised approach to water footprinting to make transparent the impacts of consumption and production on global freshwater scarcity. Global Environmental Change 20(1): 113-120.

Ridoutt, B. G. and J. Huang. 2012. Environmental relevance—the key to understanding water footprints. Proceedings of the National academy of Sciences of the United States of America 109(22): E1424.

Rockström, J., M. Falkenmark, L. Karlberg, H. Hoff, S. Rost, and D. Gerten. 2009. Future water availability for global food production: The potential of green water for increasing resilience to global change. Water Resources Research 45(7): W00A12.

75 8 References

Rosenbaum, R. K., T. M. Bachmann, L. S. Gold, M. A. J. Huijbregts, O. Jolliet, R. Juraske, A. Koehler, H. F. Larsen, M. MacLeod, M. Margni, T. E. M. J. Payet, M. Schuhmacher, D. v. d. Meent, and M. Z. Hauschild. 2008. USEtox—the UNEP-SETAC toxicity model: recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. International Journal of Life Cycle Assessment 13(7): 532-546.

Rost, S., D. Gerten, and U. Heyder. 2008. Human alterations of the terrestrial water cycle through land management. Advances in Geosciences 18(1): 43-50.

SABMiller and WWF. 2009. Water Footprinting: Identifying and Addressing Water Risks in the Value Chain Working and Godalming, United Kingdom: SABMiller Plc and World Wide Fund for Nature UK.

Savenije, H. H. G. 2000. Water scarcity indicators; the deception of the numbers. Physics and Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere 25(3): 199-204.

Shiklomanov, I. A. 2003. World Water Resources at the Beginning of the 21st Century. Cambridge: Cambridge University Press.

Struckmeier, W. and A. Richts. 2013. Personal communication (February 19, 2013), Federal Institute for Geosciences and Natural Resources.

UN and FAO. 2007. Coping with water scarcity - challenge of the twenty first century.

UNEP. 2012. A Glass Half Empty: Regions at Risk Due to Groundwater Depletion. United Nations Environment Programme - Global Environmental Alert Service. .

UNEP. 1997. World Atlas of Desertification. Edited by N. Middleton and D. Thomas. 2 ed. United Nations Environment Programme, London: Arnold. van der Ent, R. J. and H. H. G. Savenije. 2011. Length and time scales of atmospheric moisture recycling. Atmospheric Chemistry and Physics 11: 1853-1863. van der Ent, R. J., H. H. G. Savenije, S. Bettina, and S. C. Steele-Dunne. 2010. Origin and fate of atmospheric moisture over continents. Water Resources Research 46. van Zelm, R., A. M. Schipper, M. Rombouts, J. Snepvangers, and M. A. J. Huijbregts. 2011. Implementing Groundwater Extraction in Life Cycle Impact Assessment: Characterization Factors Based on Plant Species Richness for the Netherlands. Environmental Science and Technology(45): 629-635.

VDA 231-106. 1997. Material classification in motor vehicle construction; Structure and nomenclature (VDA 231-106). Frankfurt, Germany: Verband der Automobilindustrie e.V. (VDA) (German Association of the Automotive Industry).

76 8 References

Veolia. 2011. The Water Impact Index and the First Carbon-Water Analysis of a Major Metropolitan Water Cycle. Available at http://www.veoliawaterna.com/north-america- water/ressources/documents/1/10975,Water_Impact_Index-White_Paper.pdf. Accessed July 3, 2013.

Verones, F., S. Pfister, and S. Hellweg. 2013a. Quantifying area changes of internationally important wetlands due to water consumption in LCA. Environmental Science and Technology (accepted) DOI: 10.1021/es400266v.

Verones, F., D. Saner, S. Pfister, D. Baisero, C. Rondinini, and S. Hellweg. 2013b. Effects of consumptive water use on wetlands of international importance. Environmental Science and Technology (accepted) DOI: 10.1021/es403635j.

Vince, F. 2007. Proposition for the LCI framework (unpublished).

Volkswagen AG. 2010a. The Life Cycle of a Car - Environmental Commendations Document Progress. Available at http://en.volkswagen.com/en/company/responsibility/service/download/download.html. Group Research, Environment Affairs Product. Accessed August 08, 2013.

Volkswagen AG. 2010b. The Golf - Environmental Commendation. Available at http://en.volkswagen.com/en/company/responsibility/service/download/download.html. Group Research - Environment Affairs Product. Accessed August 08, 2013.

Volkswagen AG. 2010c. The Polo - Environmental Commendation. Available at http://en.volkswagen.com/en/company/responsibility/service/download/download.html. Group Research - Environment Affairs Product. Accessed August 08, 2013.

WBCSD. 2013. The WBCSD Gloabl Water Tool. Available at http://www.wbcsd.org/web/watertool.htm. World Busines Council for Sustainable Development. Accessed June 30, 2013.

WFN. 2013a. The Water Footprint Assessment Tool. Available at http://www.waterfootprint.org/tool/home/. Water Footprint Network. Accessed September 18, 2013.

WFN. 2013b. Product Water footprints. Available at http://www.waterfootprint.org/?page=files/productgallery. Water Footpint Network. Accessed September 19, 2013.

77 8 References

WFN. 2013c. The WaterStat database. Available at http://www.waterfootprint.org/?page=files/WaterStat. Water Footprint Network. Accessed July 30, 2013.

WWF. 2013. The Water Risk Filter. Available at http://waterriskfilter.panda.org/. World Wide Fund for Nature. Accessed September 18, 2013.

Zemp, D., R. Van der Ent, J. Heinke, C.-F. Schleussner, A. Rammig, J. Donges, H. M. J. Barbosa, and G. Sampaio. 2013. Cascading effects of deforestation in the Amazon on moisture recycling and forest resilience. In Proceedings of 11th INTECOL Congress, Ecology. August 18-23, 2013, London, United Kingdom.

78 Glossary

Glossary

In this chapter technical terms used throughout this dissertation are defined and explained. In some cases the internationally agreed terminology provided in the ISO standards on water footprint4 and life cycle assessment5 is cited

Safeguard subject to be protected, usually human health, Area of protection - ecosystem, and resources

Level of absolute water shortage; potential evapotranspiration 5-20 Arid - times higher than precipitation (UNEP 1997)

The annually renewable freshwater volumes within the basin which Availability - can be quantified by means of runoff (plus upstream inflows if the basin is divided into sub-catchments)

Basin internal evapo- Return of evaporated water by precipitation within the basin of - ration recycling (BIER) origin

Blue water - Ground and surface water stocks (Hoekstra et al. 2011)

Withdrawal and discharge with low or no quality degradation (e.g. Borrowing water use - cooling water)

Global warming potential of a product, which denotes its Carbon footprint - contribution to climate change, resulting from greenhouse gas emission along its life cycle

4 Direct citation from ISO/FDIS 14046. 2014. Water footprint - principles, requirements and guidance edited by International Organization for Standardization. Geneva, Switzerland.

5 Direct citation from ISO 14044. 2006. Environmental management - Life cycle assessment - Requirements and guidelines (ISO 14044:2006), International Organisation for Standardisation. Geneva, Switzerland.

79 Glossary

Factor derived from a characterization model which is applied to Characterization factor - convert an assigned life cycle inventory analysis result to the common unit of the category indicator5

Environmental claim regarding the superiority or equivalence of one Comparative assertion - product versus a competing product that performs the same function5

Consumption-to- Ratio of annual consumption to annual availability in a basin - availability (WTA) ratio expressing relative freshwater scarcity

Measure for the loss of human health resulting from years of lost life Disability adjusted life - expectancy and living with a disease(disability (Murray and Lopez years (DALY) 1996)

Area from which direct surface runoff from precipitation drains by Drainage basin/ basin - gravity into a stream or other water body4

Level of absolute water shortage; potential evapotranspiration 1.5-2 Dry-subhumid - times higher than precipitation (UNEP 1997)

Endpoint indicator - Indicator expressing an impact at the end of a cause-effect chain

Detailed analysis of local environmental impacts resulting from a Environmental impact - project or decision taking into account the specific regional situation assessment in order to inform stakeholders and decision makers

Table relating sector-specific environmental impacts to sector- Environmental input- - specific turnover. Used for the estimation of environmental impacts output table caused per monetary unit spent in a specific economic sector.

Evapo(transpi)ration - Evaporation and/or evapotranspiration

Evaporation recycling - Return of evaporated water by precipitation

80 Glossary

Evapotranspiration - Sum of evaporation and plant transpiration

Fate of evaporation Spatial analysis of the return of evaporation to the originating and - analysis other basins by precipitation (Berger and Finkbeiner 2013)

Groundwater body that has a negligible rate of natural recharge on Fossil groundwater - the human time-scale4

Freshwater - Water having a low concentration of dissolved solids4

Volume of water polluted due to waste water discharges or emission Gray water - of pollutants (Hoekstra et al. 2011)

Soil moisture and water on soil (muds) available for local Green water - evapotranspiration (Hoekstra et al. 2011)

Water which is being held in, and can be recovered from, an Groundwater - underground formation4

Hydrologically effective basin internal - Fraction of BIER returning as blue water evaporation recycling

(BIERhydrol-eff)

Class representing environmental issues of concern to which life Impact category - cycle inventory analysis results may be assigned5

Natural water bodies and dams which store water from rainy Inter monthly storage - seasons and release it in dry seasons and, thus, can buffer capacities differences in monthly water scarcity levels

Consecutive and interlinked stages of a product system, from raw Life cycle - material acquisition or generation from natural resources to final disposal5

81 Glossary

Compilation and evaluation of the inputs, outputs and the potential Life cycle assessment - environmental impacts of a product system throughout its life cycle5

Phase of life cycle assessment aimed at understanding and Life cycle impact evaluating the magnitude and significance of the potential - assessment environmental impacts for a product system throughout the life cycle of the product5

Phase of life cycle assessment involving the compilation and Life cycle inventory - quantification of inputs and outputs for a product throughout its life analysis cycle5

Midpoint indicator - Indicator expressing an impact in the middle of a cause-effect chain

Net green water The difference in evapotranspiration between agricultural and - footprint natural land (SABMiller and WWF 2009)

Collection of unit processes with elementary and product flows, Product system - performing one or more defined functions, and which models the life cycle of a product5

Regionalized/ Inventory providing spatially explicit information where water geographically explicit - use/consumption occurs water inventory

Fraction of precipitation leading to surface runoff and groundwater Runoff - recharge

Level of absolute water shortage; potential evapotranspiration 2-5 Semi-arid - times higher than precipitation (UNEP 1997)

Systematic procedures for estimating the effects of the choices Sensitivity analysis - made regarding methods and data on the outcome of a study5

82 Glossary

Statistical low flow Runoff which will be exceeded with a probability of 90% in a long - (Q90) term perspective (ISWS 2013)

Water in overland flow and storage, for example rivers and lakes, Surface water - excluding seawater4

Temporally explicit Inventory providing temporally explicit information in which month - water inventory water use/consumption occurs

Transpiration - Evaporation of water by plants

Water accounting and The method developed in this work enabling the accounting of vulnerability evaluation - water use and the analysis of the vulnerability of a basin to potential (WAVE) model impacts resulting from it

Water footprint study which considers water consumption and Water availability - water scarcity related impacts but neglects water quality footprint degradation and resulting impacts (ISO/FDIS 14046 2014)

Accumulation of water which has definite hydrological, Water body / water - hydrogeomorphological, physical, chemical and biological course characteristics in a given geographical area4

The share of total water use which is not returned to the originating Water consumption / - drainage basin due to evapo(transpi)ration, product integration, or consumptive water use discharge into other basins and the sea (based on Owens 2001)

Water degradation / The part of withdrawal returned to the basin after quality - degradative water use degradation (e.g. waste water discharge) (based on Owens 2001)

Indicator expressing the vulnerability to freshwater depletion in a Water depletion index basin determined based on blue water scarcity; WDI can be - (WDI) understood as an equivalent volume of potentially depleted water resulting from a volume of water consumption

83 Glossary

Water footprint study which assesses freshwater degradation due to Water eutrophication emissions of nitrates, phosphates, etc. leading to eutrophication - footprint impacts but neglects water consumption and water scarcity related impacts (ISO/FDIS 14046 2014)

Metric(s) that quantify(ies) the potential environmental impacts Water footprint - related to water4

Activities related to accomplishing water footprint studies and/or to Water footprinting - develop water footprint methods

Physical (e.g. thermal), chemical and biological characteristics of Water quality - water with respect to its suitability for an intended use by human or ecosystems4

Extent to which demand for water compares to the replenishment of Water scarcity - water in an area, e.g. a drainage basin4

Detailed analysis of an organization’s water consumption taking into account the specific local situation in order to evaluate the environmental, operational, legal, and reputational risks. In an “out Water stewardship - of the fence approach”, water stewardship identifies opportunities and solutions in cooperation with the public, authorities, and other water users within the basin. (EWS 2013)

Water stress - General pressure resulting from water scarcity

Water use - Total input of freshwater into a product system

Anthropogenic removal of water from any water body or from any Water withdrawal - drainage basin, either permanently or temporarily4

Withdrawal-to- Ratio of annual withdrawal to annual availability in a basin - availability (WTA) ratio expressing relative freshwater scarcity

84 List of Figures

List of Figures

Figure 2 Water footprint methods, databases, and tools identified and classified in the literature review ...... 4

Figure 3 Life cycle of the cars along which water consumption has been analyzed ...... 12

Figure 5 Relative contributions of life cycle stages (a) and material groups to production impacts (b) in the impact categories eutrophication (EP), ozone layer depletion (ODP), photochemical ozone creation (POCP), global warming (GWP), acidification (AP), and in the water consumption inventory (WC) for the Golf 1.6 TDI...... 17

Figure 6 Global water consumption throughout the life cycles of: a) the Polo 1.2 TDI, b) the Golf 1.6 TDI, and c) the Passat 2.0 TDI...... 18

Figure 7 Relative comparison of results on the inventory and impact assessment levels for the default scenario (bars), the min-s scenario (circles), and the max-s scenario (diamonds)...... 19

Figure 8 Relative differences in impact assessment results between max/min-s scenarios and the default scenario ...... 21

85 List of Figures

Figure 11 Average continental evaporation recycling ratio (van der Ent et al. 2010) – reproduced by permission of American Geophysical Union ...... 25

Figure 12 Fractions of water evaporated in the Rhine basin which are recycled internally and precipitate over other drainage basins and sea ...... 26

Figure 14 Conceptual example of water consumption and greenhouse gas emissions ...... 34

Figure 15 Relative results of volumetric and impact based water footprints of the theoretical example comparing two products causing 1 m³ of water consumption in Iran and Germany respectively ...... 35

Figure 16 Water inventory flows along the life cycle of a product considered in WAVE ...... 38

Figure 17 Basin internal evaporation recycling (BIER) ratios denoting the fraction of evaporated water returning to the originating basin via precipitation ...... 40

Figure 18 Runoff fraction (α) which denotes the ratio of the long-term average runoff (blue water) and precipitation within a drainage basin...... 40

Figure 19 Hydrologically effective basin internal evaporation recycling (BIERhydrol-eff) ratios denoting the fractions of evaporated water returning to the originating basin as blue water ...... 41

* Figure 20 Logistic function determining WDI based on CTA ; S-curve leads to larger spreading of WDI in medium scarcity ranges 0.05

Figure 21 Basins classified as dry subhumid, semi-arid, and arid according to UNEP (1997) ...... 44

Figure 22 Factors WDI expressing vulnerability of basins to freshwater resource depletion

[m³depleted/m³consumed] ...... 45

Figure 23 Relative presentation of blue water consumption required to produce 1 GJ of bioethanol

from sugar cane, reduction of water consumption due to consideration of BIERhydrol-eff, and potential impacts determined by means of WAVE and by the impact assessment methods of Pfister et al. (2009) and Frischknecht et al. (2009) ...... 46

Figure 24 Basin internal evaporation recycling (BIER) ratios, determined for a basin side length of 100 km ...... 49

86 List of Figures

Figure 26 Relative changes in CTA scarcity ratio due to the consideration of ground and surface water stocks ...... 51

Figure 27 Changes in WDI due to the consideration of absolute freshwater shortage in relation to relative scarcity ...... 51

Figure 29 Relative decrease of default WDI results when (a) the availability time horizon of surface water stocks is halved (50 years); (b) the availability time horizon of surface water stocks is doubled (200 years) ...... 54

Figure 33 Water consumption as one indicator among others in the context of sustainability ...... 66

87 List of Tables

List of Tables

Table 1 Scope and characteristics of the water footprint methods, databases, and tools identified in the literature review (updated from Berger and Finkbeiner (2010)) ...... 8

Table 2 Adjustment factor for ground water stocks (AFGWS) reducing water scarcity based on geological structure and annual recharge ...... 43

Table 4 Scenarios examined in the sensitivity analysis including description and parameter settings ...... 53

88 Appendix

Appendix

Paper I: Berger, M. and M. Finkbeiner. 2010. Water footprinting - how to address water use in life cycle assessment? Sustainability 2(4): 919-944.

Reprinted with permission from MDPI AG. http://www.mdpi.com/2071-1050/2/4/919

Paper II: Berger, M., J. Warsen, S. Krinke, V. Bach, and M. Finkbeiner. 2012. Water Footprint of European Cars: Potential Impacts of Water Consumption along Automobile Life Cycles. Environmental Science and Technology 46(7): 4091-4099.

Reprinted with permission from Berger, M., J. Warsen, S. Krinke, V. Bach, and M. Finkbeiner. 2012. Water Footprint of European Cars: Potential Impacts of Water Consumption along Automobile Life Cycles. Environmental Science and Technology 46(7): 4091-4099. Copyright 2012 American Chemical Society. http://pubs.acs.org/doi/abs/10.1021/es2040043

Paper III: Berger, M. and M. Finkbeiner. 2013. Methodological challenges in volumetric and impact oriented water footprints. Journal of Industrial Ecology 17(1): 79-89.

Paper IV: Berger, M., R. van der Ent, S. Eisner, V. Bach, and M. Finkbeiner. 2014. Water accounting and vulnerability evaluation (WAVE) – considering atmospheric evaporation recycling and the risk of freshwater depletion in water footprinting. Environmental Science and Technology, in press, DOI 10.1021/es404994t

Reprinted with permission from Berger, M., R. van der Ent, S. Eisner, V. Bach, and M. Finkbeiner. 2014. Water accounting and vulnerability evaluation (WAVE) – considering atmospheric evaporation recycling and the risk of freshwater depletion in water footprinting. Environmental Science and Technology in press, DOI 10.1021/es404994t. Copyright 2014 American Chemical Society. http://pubs.acs.org/doi/abs/10.1021/es404994t

Country factors for BIER, BIERhydrol-eff, and WDI

89 2010, , 919-944; doi:10.3390/su2040919 OPEN ACCESS sustainability ISSN 2071-1050 www.mdpi.com/journal/sustainability

Water Footprinting: How to Address Water Use in Life Cycle Assessment?

Markus Berger * and Matthias Finkbeiner

Department of Environmental Technology, Technische Universität Berlin, Office Z1, Strasse des 17, Juni 135, 10437 Berlin, Germany; E-Mail: [email protected]

* Author to whom correspondence should be addressed; E-Mail: [email protected]; Tel.: +49-30-314-25084; Fax: +49-30-314-78815.

! " # $ !

Abstract: As freshwater is a vital yet often scarce resource, the life cycle assessment community has put great efforts in method development to properly address water use. The International Organization for Standardization has recently even launched a project aiming at creating an international standard for ‘water footprinting’. This paper provides an overview of a broad range of methods developed to enable accounting and impact assessment of water use. The critical review revealed that methodological scopes differ regarding types of water use accounted for, inclusion of local water scarcity, as well as differentiation between watercourses and quality aspects. As the application of the most advanced methods requires high resolution inventory data, the trade-off between ‘precision’ and ‘applicability’ needs to be addressed in future studies and in the new international standard.

Keywords: water use; life cycle assessment; water footprint

1. Introduction

%% &'(

Freshwater is a precious resource on our planet. It is crucial to sustain life and cannot be replaced by any other substance. However, freshwater is scarce in some regions, countries, or even continents, 2010, 920 leading to manifold problems. With regard to human health, this can include for instance malnutrition due to lack of agricultural irrigation water. Such problems are relevant for about a third of the world’s population who are threatened by a lack of water to meet daily needs [1]. In terms of ecosystems, water scarcity can affect biodiversity, as sensitive species might not be able to cope with reduced freshwater availability. Hence, freshwater needs to be managed properly in order to achieve the United Nation’s millennium goals regarding human wellbeing and intact ecosystems [2]. Life cycle assessment (LCA) is a widely accepted and applied environmental management tool to measure the various environmental interventions caused by products from cradle to grave [3]. Yet, when assessing the environmental performance of a product by means of LCA, attention is usually drawn on the energy consumed along a product’s lifespan or on the emission of greenhouse gases and toxic substances. In contrast, the use of freshwater throughout a product’s life cycle is often neglected. This can be explained by the history of LCA, which was developed in industrial countries that usually do not suffer from water scarcity. Furthermore, LCA was traditionally used to assess industrial products, which require rather low amounts of water in their production. However, there are also specific methodological challenges that both the inventory and the impact assessment have to face for water use. Difficulties result from the fact that freshwater is not ‘consumed’, but rather circulates in global cycles. Furthermore, freshwater availability varies around the globe, different watercourses fulfill different ecological functions, and different water qualities enable different uses. Yet, when accomplishing LCA studies of agricultural products, biofuels, or renewable raw materials, water consumption can be substantial [4]. Hence, it needs to be considered, as otherwise problem shifting from, for instance, ‘global warming’ to ‘water scarcity’ can occur. Such a severe deficiency is not acceptable for a methodology that has been developed to support sustainable decision making and is even in conflict with the principle of ‘comprehensiveness’ required by International Organization for Standardization (ISO) in the ISO 14040 standard [5]. Even though this challenge has not been tackled for a long time, method development is making considerable progress today. Pushed from initiatives like the World Business Council on Sustainable Development (WBCSD) or the UNEP/SETAC Life Cycle Initiative, comprehensive methods to account for water use on both inventory and impact assessment level have been developed. Furthermore, in addition to the ‘carbon footprint’, which can be regarded as a single-impact LCA only addressing greenhouse gases [6], the ISO has recently started to establish an international standard to assess water use in LCA. Taking into account the recent efforts in method development, the standardization process and the increased public awareness, it may become true that “water is the new carbon” as claimed by a recent article in the British newspaper ‘The Independent’ [7]. Therefore, this paper aims at reviewing a broad range of scientific methods that have been developed up to now, which account for water use on both inventory and impact assessment levels. After presenting an overview of the methods, the individual advantages and shortcomings of each method are discussed. Based on the identified gaps, research recommendations for method improvement are derived.

2010, 921

% % ) (

The review of research articles dealing with assessment of water use in LCA or case studies revealed a lack of a consistent terminology. In order to provide consistent wording throughout this article, the terminology proposed by the UNEP/SETAC Life Cycle Initiative [8] has been adopted. In general, the total input of freshwater into a product system is referred to as ‘water use’. As parts of the water input is released from the product system as waste water, the remaining part which has become unavailable due to evaporation or product integration is referred to as ‘water consumption’. Moreover, the term ‘freshwater use’ is divided into the categories ‘in-stream freshwater use’ and ‘off-stream freshwater use’. While in-stream freshwater use describes an use of freshwater (e.g., for hydroelectric power or ship traffic), off-stream freshwater use comprises any use of freshwater that requires a prior removal of freshwater from the water body. Additionally, freshwater use can be divided into ‘freshwater degradative use’ and ‘freshwater consumptive use’. Freshwater degradative use is characterized by withdrawal and discharge of freshwater into the same watershed after quality alteration. In contrast, freshwater consumptive use occurs when used freshwater is not released into the same watershed from which it was withdrawn due to product integration, evaporation, or discharge into different watersheds. Based on these two sub-divisions, the following four types of freshwater use are divided:

In-stream freshwater degradative use, e.g., temperature increase of water retained in dams or reservoirs In-stream freshwater consumptive use, e.g., additional evaporation of water retained in dams or reservoirs Off-stream freshwater degradative use, e.g., increase of biochemical oxygen demand between water catchment and waste water treatment plant effluent Off-stream freshwater consumptive use, e.g., the fraction of irrigation water that is evaporated

It should be noted that this paper focuses on methods accounting for in- and off-stream freshwater consumptive uses, which assess the consequences of water that is ‘lost’ in a particular region. Methods assessing the consequences of degradative uses (freshwater pollution) leading to eutrophication, eco-toxicity, human-toxicity, % [9], are not reviewed in this work. Based on the concept introduced by Allan [10], several authors divide water into three categories: green, blue, and gray water. The green water consumption describes the evapotranspiration of rainwater during plant growth, which is especially relevant for agricultural products. Blue water consumption is the volume of ground and surface water that evaporates during production. Thus, it comprises the amount of water that is not returned into the environmental compartment from which it has been withdrawn initially. As the water that is returned to the environment (e.g., effluent of waste water treatment plants) can be of lower quality, the gray water describes the total amount of water that is polluted by that effluent. Hence, gray water equals the volume of water required to dilute the used water until it reaches commonly agreed quality standards. Besides the specification of different types of water use, it should be noted that the term ‘water footprint’ has two meanings. On the one hand it refers to the specific method introduced by 2010, 922

Hoekstra [11], which is described below. On the other hand, ‘water footprinting’ describes the activities of addressing water use in LCA in general. To avoid confusion the term ‘water footprint according to Hoekstra’ is used when the specific method is referenced. With regard to impact assessment, lots of methods use the withdrawal-to-availability (WTA) ratio for calculating characterization factors for water use and/or consumption. As shown in Equation 1, WTA is defined as the ratio of total annual freshwater withdrawal for human uses in a specific region (W) to the annually available renewable water supply in that region (A). Hence, WTA serves as an index for local water scarcity. + B , * * +) (1)

As this ratio as well as its components is named differently in different methods we ‘translate’ the method specific names into the terminology introduced here when describing the methods.

2. Methods for Accounting and Assessing Water Use in LCA

The application of the life cycle perspective to product water footprints leads to methods that reveal the entire amount of freshwater required to produce a product. This comprises the water use in the manufacturing process as well as water used in background processes such as the mining of raw materials, the production of materials and semi-finished products, or the generation of electricity. Furthermore, the water used during the product’s use, disposal, or recycling is taken into account. A broad range of currently developed methods assessing water use from a life cycle perspective were identified by literature research in cooperation with the working group on water assessment of the UNEP/SETAC Life Cycle Initiative. With the exception of the methods virtual water [10] and water footprint according to Hoekstra [11], most methods have been developed to support life cycle inventory (LCI) and life cycle impact assessment (LCIA) modeling within LCA. However, similar to carbon footprinting, the methods can also be extracted and used as ‘stand alone’ methods when focusing exclusively on water use. All methods are described in the following section starting from pure water inventories, midpoint- (middle of cause-effect-chain), up to damage oriented endpoint (end of cause-effect-chain) impact assessment schemes. It should be noted that this order does not reflect the scientific value of a method—it is just guiding the reader through the methodological development.

%% + ,

The simplest way to determine a water footprint is to use the water inventory of the product or organization under study. By subtracting the waste water effluents from the freshwater inputs the freshwater consumption due to evaporation, product integration, and leakages can be determined. In this way, the water footprints of production steps or organization units can be determined and aggregated to a complete organization or product water footprint. Water inventories can be established by means of LCA databases like ecoinvent [12] and GaBi [13], tools such as the WBCSD Global Water Tool [14], and according to frameworks proposed by e.g., Vince [15] or the UNEP/SETAC Life 2010, 923

Cycle Initiative [8]. Depending on the database, tool, or framework the information content of the inventory can differ considerably. LCA databases usually only classify the input and output fluxes according to the watercourses from which the water is withdrawn and to which it is released (ground-, surface-, seawater, %). In contrast, the WBCSD Global Water Tool [14] contains further information regarding the location of withdrawal and the respective scarcity in this area. Frameworks established by Vince [15] or the UNEP/SETAC Life Cycle Initiative [8] go further by differentiating the different qualities of water fluxes entering and leaving the product system.

% % - + + .! ( /'

The concept of virtual water [10] was the first attempt towards product water footprinting and was developed by Allan in the early 1960s [16]. The method accumulates all quantities of water that have been consumed along the production chain of a product. Hence, it comprises water used in the actual manufacturing processes as well as water used in background processes such as material or energy production. In contrast to the water inventory virtual water is divided into three categories: green, blue, and gray water as described in the section on Terminology. The water footprint according to Hoekstra [11] was introduced in 2002 and relies on the virtual water concept, but additionally includes spatial and temporal information [16]. Accordingly, the quantitative water footprint of a product is the same value as its virtual water content. Furthermore, water footprints were calculated for individuals, organizations, or nations by multiplying all products and materials consumed with their respective virtual water content and by adding the direct water consumption of the person, organization, or nation [16].

%0% 12,#

Within the environmental design of industrial products (EDIP) programme [17], a set of impact categories has been established to support LCIA in LCA studies. By means of the impact category EDIP resources the consumption of renewable and non-renewable resources can be assessed. With regard to water consumption this comprises the following steps. Initially, the volume of freshwater consumed along the product’s life cycle is normalized according to Equation 2: ! ( ! 3! ! !4 ( ! ! (2) 1990

As this figure does not contain any information concerning water scarcity, the normalized consumption is divided by the time span in which the resource will still be available. In order to determine the water availability time span, the total regional freshwater supply (renewable and non-renewable) is divided by the difference of annual regional consumption and annual regional regeneration of freshwater. Here it should be noted that the annual regional regeneration of freshwater equals the annually available renewable water supply (A) as defined in Chapter 1.2. ( ! supply . ( ! . ( ( (3) 2010, 924

Hence, water consumption in regions with high water scarcity scores higher than the same water consumption in regions where water is abundant. The result represents the share of the product’s water consumption to the per-capita water availability in the reference year 1990. Finally, the weighted water consumption can be aggregated with other resource consumptions leading to a single indicator, which enables a comprehensive resource consumption assessment.

%4% 56, + 6! 17(

Exergy can be regarded as the useable fraction of energy, which can be converted into work [18]. The cumulated exergy demand (CExD) [18] or its enhancement, the cumulative exergy extraction from the natural environment (CEENE) [19], were proposed as indicators for resource consumption in LCIA. The basic idea of both concepts is to multiply each resource input into a product system by its respective exergy content. Hence, CExD and CEENE represent the exergy taken from the natural environment by a technical product system or denote the ‘physical chemical price the natural environment pays for the withdrawal toward our industrial society’ [19]. With regard to water use the exergy demand is calculated in two different ways, depending on the type of water use. For in- and off-stream freshwater consumptive uses, the volume of water consumed is multiplied by its chemical exergy content of 50 MJ/m³ [18]. In contrast, the exergy demand of water used in hydroelectric power plants (in-stream degradative water use) is calculated based on the potential energy of the barrage water. The main advantage is that exergy contents can be determined for all types of resources including minerals, metals, water, biomass, renewable and non-renewable energy carriers, and land use. This overcomes the shortcomings of conventional resource LCIA indicators, which only account for certain resources. For instance, the cumulated energy demand (CED) [20] only covers energy carriers or the abiotic depletion potential [9] accounts exclusively for non-renewable abiotic resources. Consequently, exergy is a more comprehensive indicator accounting for all types of resource use and for the ‘quality’ of the resource consumption in terms of lost exergy. Water consumption can be assessed as a type of resource use and can be aggregated and compared with the consumption of other resources.

%$% 1 (

The ecological scarcity method [21] has been developed to support LCIA in LCA. The method provides eco-factors for a range of substances expressing their environmental impact. In LCIA elementary flows compiled in the LCI can simply be multiplied by their corresponding eco-factors. The results, which are expressed in eco-points, can then be aggregated to a single-score indicator expressing the overall environmental impact of the product analyzed. In general the eco-factors are calculated according to the following equation, which contains a characterization, normalization and weighting step:

2 F !int V 14 !int C . S 1 G W 9 4 4 D T 4 H X . D . T 63 E U tan (4) (! ) 8 3 +(( 2010, 925

The method also provides eco-factors for water use. In contrast to water consumption, which comprises the evaporated fraction only, water use denotes the total input of freshwater into the product system. When calculating eco-factors for water use, no characterization (conversion of LCI flow to the common unit of the impact category [22]) is performed, %%, water is not characterized according to quality or type of water source. Regarding normalization (impact category result in relation to a reference region [22]), the total annual freshwater withdrawal for human use in a specific region (W as described in Chapter 1.2) is assigned to 1 eco-point. In terms of weighting, the method uses a political distance-to-target weighting procedure in which the ratio of a current (F) to a critical flow (Fc) needs to be determined. In the context of weighting for water use assessment, the method incorporates the political target of preventing water stress. According to the OECD [23] water stress occurs if the water pressure, which equals WTA as described in Chapter 1.2, is larger than 20%. Hence, as long as no more than 20% of the annually available renewable water supply (A as defined in Chapter 1.2) is used by human activities, no harm for ecosystems is expected. Accordingly, the current flow equals the total annual freshwater withdrawal for human uses (W) in the region or country where the water use occurs and the critical flow is set to 20% of the annually available renewable water supply (A) of that region. The square of the weighting factor leads to an above average weighting if the critical flow is significantly exceeded. Thus, as shown in Equation 5 and Table 1, the weighting factor is dependent on the WTA and can range from 0.0625 to 56.3. Multiplying the result by the constant c (1012/a) leads to a more convenient dimension.

C S2 2 2 currentflow C total annual freshwater withdrawal for human uses (W) S 2 C 1 S W(( D T D T +) 4D T E criticalflow U E annuallyavailable renewable water supply(A) 4 20% U E 20% U (5)

Table 1. WTA ranges and resulting weighting factor assuming a critical flow of 20% of the renewable water supply [21].

WTA used for Weighting WTA calculation factor low <0.1 0.05 0.0625 moderate 0.1–<0.2 0.15 0.563 medium 0.2–<0.4 0.3 2.25 high 0.4–<0.6 0.5 6.25 very high 0.6–<1.0 0.7 16.0 extreme >1.0 1.5 56.3

%:% 56,

A site specific impact assessment method for South Africa has been introduced by Brent [24], which assesses the use and pollution of water-, air-, land-, and mined abiotic resources. In terms of water, the use of ground and surface water is simply aggregated without characterization in the sub-resource group ‘water quantity’. The pollution of water is denoted in the sub-resource group ‘water quality’ by normalizing the results for the impact categories eutrophication, acidification, human- and eco-toxicity based on ambient environmental quantity and quality objectives. 2010, 926

Subsequently, the results of the two sub-resource groups are combined within the main resource group ‘water’ by a distance-to-target weighting in which the two sub-resource groups are multiplied by a factor expressing the ratio of current ambient state to target ambient state. Both normalization and weighting is accomplished site specific for four South African regions to better reflect the site specific effects of water and land use impacts. Finally, the four resource impact indicators (RII) expressing the threats for the main resource groups water, air, land, and mined abiotic resources are combined to a single-score environmental performance resource impact indicator (EPRII). This aggregation is accomplished by means of a ranking procedure and a further weighting step that includes political value choices of the South African government and manufacturing industries.

%"% 56, 56, ( + ; ( 6 6 ( < $=

This method attempts to differentiate between different types of water use in LCI and provides two midpoint impact categories for LCIA. In terms of LCI modeling, Mila i Canals and colleagues propose differentiating between inputs of green water (soil moisture), blue water (ground and surface water), fossil blue water (non-renewable ground water), and water use due to land use changes. Next to differentiating the input of freshwater into a product system, the use of water should be categorized into evaporative and non-evaporative use. Additionally, procedures for calculating different types of water consumption are provided. Furthermore, the method discusses the following impact pathways resulting from water use:

Water use leading to insufficient freshwater availability causing impacts on human health Fossil and aquifer groundwater use above renewability rate leading to reduced availability of freshwater as a resource for future generations—freshwater depletion (FD) Water use leading to insufficient freshwater availability causing effects on ecosystem quality—freshwater ecosystem impacts (FEI) Land use changes leading to changes in freshwater availability causing effects on ecosystem quality—freshwater ecosystem impacts (FEI)

While no method is provided to describe the impacts to human health, Mila i Canals and colleagues propose ways of quantifying the impacts of water use to freshwater depletion (FD) and freshwater ecosystem impacts (FEI) according to the impact pathways shown in Figure 1. 2010, 927

Figure 1. Inventory requirements and impact pathways resulting from different types of water use addressed by Mila i Canals and colleagues, based on [25].

life cycle inventory impact pathways areas of protection

input output

green water

soil moisture R a i blue water change in environ- n evaporative use mental river flow river/lake change in availa- ecosystem W bility for aquatic a aquifer ecosystems quality t evaporative use change in ground- e water table r land use change in evapo- change in return transpiration & runoff to ecosystem

fossil blue water non-evaporative use change in long- natural fossil water term availability resources evaporative use

The midpoint impact category FD assesses the reduced availability of the resource freshwater for future generations if the water use exceeds the renewability rate of the respective body of water. As surface watercourses such as rivers usually have a high renewability rate, it is assumed that only the consumption of water from aquifers (evaporative use) and fossil water (evaporative and non-evaporative use) can contribute to that impact category. In order to provide characterization factors (factors converting the LCI flow to the common unit of the impact category [22]) the method of Guinee and colleagues to determine the depletion of abiotic resources [9] is adapted to water use as shown in Equation 6: 1 2 2# 4 2 2 (6)

Thus, the abiotic depletion potential (ADP) of a water resource i serves as a characterization factor, which is dependent on the extraction rate of resource i (ER), the regeneration rate of resource i (RR), the ultimate reserve of the resource i (R), the ultimate reserves of the reference resource antimony

(RSb), and the deaccumulation rate of antimony (DRSb). As shown in Equation 5, strongly overexploited water resources (ER > RR) will result in higher characterization factors, whereas sustainably used resources (ER = RR) will result in a characterization factor of 0. Yet, underexploited water resources (ER < RR) would result in negative characterization factors. Following Mila i Canals and colleagues, such positive effects are excluded from the calculation as no water depletion occurs. The second midpoint impact category FEI aims to assess the ecological consequences of water use in a certain region. In contrast to FD, the consumption of fossil blue water is excluded as it fulfils 2010, 928 minimal ecological functions. Hence, only the evaporative use of blue water (surface water and aquifers) as well as water use due to land use changes are taken into account. In order to obtain concrete characterization factors the water stress indicator (WSI) developed by Smakhtin and colleagues [26] is suggested. +; +, (7) + 1+

As it can be seen from Equation 7, the WSI of region i denotes the ratio of total water use (WU) to the difference between renewable water reserves (WR) and the environmental water requirement (EWR) in that region. The basis for this indicator is the water use per resource indicator (WUPR) [27], which relates the total water use to the renewable water reserves in a region. In order to harmonize terminology, it should be noted that WU equals the total annual freshwater withdrawal for human uses (W), WR equals the annually available renewable water supply (A), and WUPR equals WTA as defined in Chapter 1.2. +; +;# (8) +

Hence, the WSI enhances the WUPR or WTA indicator by ‘reserving’ a certain amount of freshwater necessary to sustain the ecological functions in a particular region. Depending on the local water scarcity and the respective ecosystem demand, site specific characterization factors are obtained assessing the severity of additional human water use.

%>% 63 8 ,! 6( Q. 2! / ;R

Bayart and colleagues [28] introduced a freshwater accounting and impact assessment method that follows the requirements of the framework proposed by the UNEP/SETAC Life Cycle Initiative [8]. Accordingly, freshwater inputs into and outputs from the product system are categorized based on their quality (high/low) and resource type (surface/groundwater), which enables the quantification of losses and gains of different freshwater types on the inventory level. Subsequently, the authors proposed the new midpoint impact category ‘freshwater deprivation for human uses’ to assess the consequences of freshwater consumption regarding contemporary human uses. Depending on the type (i) of freshwater consumed, characterization factors (CF) are calculated according to Equation 9, which express the ‘m³ potable water equivalent unavailable for human uses’ per m³ of water consumed.

6. 4; 4A 41 6 (9)

While the regional freshwater scarcity (α) is calculated by means of WTA, the number of potential uses depending on freshwater type and quality is expressed by means of the functionality factor (U). The quality factor (Q) denotes the quality of the consumed freshwater based on the energy demand required to transform the water quality i into drinking water quality. Furthermore, the compensation ability (CA) to adapt to increased water scarcity based on socio-economic parameters is taken into account. Following this procedure the authors determined a set of characterization factors for different countries and different freshwater types. 2010, 929

%B% / / 2( ; ( +

Motoshita and colleagues [29] modeled the cause-effect-chain of agricultural water scarcity to undernourishment related health damages in two steps. First, the reduced availability of agricultural water, as a potential consequence of water consumption, will diminish agricultural productivity. This relationship is described on country scale in a prediction model which regards crop productivity per unit dietary energy as being proportional to agricultural water use. Subsequently, undernourishment related health damages resulting from decreased agricultural productivity are assessed in a regression model and expressed in units of ‘disability adjusted life years’ (DALY). DALY is the unit of a health indicator developed by the World Health Organization (WHO) that expresses the total amount of lost health due to premature death and disability resulting from illnesses and injuries [30]. Besides linking diminished agricultural productivity to health damages, the regression model analyses the effects of variables like average dietary energy consumption, medical treatment, and health expenditure by means of non-linear and multiple regression analysis. Depending on the local vulnerability, Motoshita and colleagues determined damage factors ranging from 10–9 to 10–7 DALY per m³ of water consumed.

%% / / 2( , 2 ( 2 + 6!

According to the WHO, 10% of the total worldwide diseases result from lacking access to clean drinking water, and lacking water for sanitation and hygiene [31]. For that reason, Motoshita and colleagues [32] analyzed the cause-effect-chain that links water consumption and the occurrence of infectious diseases. According to the authors, the consumption of freshwater in a particular region leads to shortage of safe drinking water, which results in drinking of unsafe water. Subsequently, the ingestion of unsafe drinking water leads to the intake of infectious sources, resulting in health damages caused by infectious diseases. By means of multiple regression analysis the authors modelled this cause-effect-chain taking into account variables such as house connection rate to water supply and sanitation, average dietary consumption, gini coefficient of dietary energy consumption, undernourished population, annual average temperature, and health expenditure per capita. As a result, Motoshita and colleagues determined country specific characterization factors expressing the health damage resulting from the consumption of freshwater. Depending on the local circumstances, the resulting health damage ranges from 6 × 10–13 to 2 × 10–4 DALY per m³ freshwater consumed.

%% 63 . ( 1 ( 2( C 17

Van Zelm and colleagues [33] are currently developing characterization factors expressing the contribution of groundwater extraction to ecosystem damage in The Netherlands. The characterization factors are calculated by means of a fate and an effect factor. The fate factor denotes the change in average spring groundwater level (dASG) as a consequence of the change in the extraction rate (dq). Information for determining fate factors is obtained from the National Hydrological Instrument of The Netherland (NHI). NHI describes the Dutch hydrological situation in a 250 250 m raster based on the MODFLOW model [34] and enables the calculation of 2010, 930 generic as well as soil type specific Dutch fate factors. The effect factor expresses the change in potentially not occurring fraction of species (dPNOF), which results from the change in average spring groundwater level (dASG). The effect factors are determined by means of multiple regression equations expressing the relation of changes in the groundwater level to the potential occurrence of plant species based on information of 690 plant species in the MOVE model [35]. F C V F#8D.V 6. 4 G W 4 G W B E C HX H X (10)

Finally, as shown in Equation 10, characterization factors (CF) are calculated by multiplying the area of each grid cell (A) by its corresponding fate and effect factors.

% % 2( E 1 6 + ; 2

Hydropower is generally regarded as an environmentally friendly source of electric energy. However, this view typically neglects the effects on ecosystems resulting from the damming of water, which are obviously very site dependent. Therefore, Maendly and Humbert [36] developed specific characterization factors for assessing effects on aquatic biodiversity resulting from this so-called in-stream freshwater use. By relating the fraction of disappeared species (PDF) on the original river surface area flooded in both the up-stream and down-stream zones (A) to the water throughput or electricity generated (Q), the characterization factor (CF) is obtained. Finally, the characterization factors of different sections (CFsection,i) are aggregated to the overall characterization factor (CF), which denoted the environmental damage expressed in the widely applied unit of potentially disappeared fraction (PDF m² a) [37] per m³ or kWh.

6. 6. 1 #2. 4 B sec, A B sec, , (11) (

%0% ,! . 6! ( # 6 ( <0>=

The method developed by Pfister and colleagues enables a comprehensive impact assessment of freshwater consumption on both midpoint and endpoint level. Referring to the virtual water terminology, the method only accounts for blue water consumption, %%, the consumption of ground and surface water. On midpoint level, a regional ‘water stress index’ (WSI) is introduced, which serves as a characterization factor for the proposed impact category ‘water deprivation’. It should be noted that the water stress index (WSI) introduced here must not be confused with the water stress indicator [26] (also WSI) that is suggested by Mila i Canals and colleagues [25] as characterization factor for the impact category freshwater ecosystem impacts (FEI). The WSI according to Pfister % relies on WTA as defined in Chapter 1.2 and has been calculated for more than 10,000 watersheds by means of the global WaterGAP2 model [39]. However, the regional hydrologic situation might vary throughout the year due to seasonal precipitation differences. This seasonal variation might cause additional water 2010, 931 stress if the wet seasons cannot fully compensate for the dry seasons due to lacking storage capacities of the individual water shed or additional evaporation of stored water. By introducing a variation factor (VF) such effects are taken into account and are included in the modified WTA ratio WTA*. In order to achieve continuous characterization factors between 0.01 and 1, the WSI is calculated according to the following logistic function.

+, 1 4+) C 1 S 1 6.4 *D 1T (12) E 0.01 U

All amounts of blue water consumption can then be multiplied by their specific regional WSI to obtain characterized results, which can be aggregated in the midpoint impact category water deprivation. Next to this midpoint indicator, the method also comprises three endpoint impact categories enabling damage assessment according to the eco-indicator 99 framework [37] in the areas of protection human health, ecosystem quality, and resources. In terms of ( the method refers to the impact pathway of malnutrition due to lack of irrigation water. In order to quantify the damage to human health resulting from malnutrition

(ΔHHmalnutr.) as a consequence of water consumption (WUconsumptive) in a particular region, the entire cause-effect-chain is modelled. Starting from the water stress index (WSI) and the percentage of agricultural water use to total water use (WU%agriculture), the water deprivation for agricultural purposes (WDF) is quantified. By means of an effect factor (EF), which incorporates the per-capita water requirement to prevent malnutrition (WRmalnutr.) and the human development factor (HDF), which is calculated based on the human development index, the annual number of malnourished people is calculated. Finally, the overall human health effects resulting from a certain number of malnourished people are quantified by means of a damage factor (DF) in the unit DALY based on statistical health data. // +, 4+; 4 /2. 4+ 1 4 2. 4+; malnutr.,i %( , malnutr.,i malnutr.,i malnutr.,i !, +2. 1. (13) 6. malnutr.,i

In order to assess ( E resulting from a certain freshwater consumption, the ecological cause-effect-chain needs to be modelled. It is assumed that withdrawals of blue water reduce the availability of green water, which is crucial for vegetation in many ecosystems. As proposed in the eco-indicator 99 framework [37], ecosystem damage is measured in potentially disappeared fraction of species (PDF), which are a measure for the vulnerability of vascular plant species biodiversity (VPBD). In order to assess vegetation damage that is related to water shortage, net primary production (NPP) has been chosen as a proxy for two reasons. First, a high correlation between NPP and VPBD has been revealed. Second, there are spatial data globally available assessing constraints to net primary production due to water shortage (NPPwat-lim) by means of indices ranging from 0 to 1 [40]. As shown in Equation 14, the damage to ecosystem quality (ΔEQ) is determined by multiplying NPPwat-lim by the ratio of water consumption (WUconsumptive) to precipitation (P). This ratio indicates the area-time equivalent necessary to recover the consumed (blue) water by annual precipitation. 2010, 932

+; 1A 6. 4+; 8## 4 cos!, 1A, !, lim, # (14) #2. 4

2( is the third area of protection assessed in the eco-indicator 99 framework [37]. It denotes the depletion of natural resources and is measured in units of surplus energy, which indicates the additional energy required to mine a resource of lower concentration after a resource extraction took place. In this context the concept has been modified and the surplus energy for replacing an amount of depleted freshwater by means of seawater desalination is determined according to the following equation. 1 4 . 4+; ! , ! (15)

The damage to resources (ΔR) resulting from water consumption (WUconsumptive) is calculated by multiplying the energy demand for desalination (Edesalination) by the fraction of water consumption contributing to freshwater depletion (Fdepletion). While Edesalination is fixed to a value of 11 MJ/m³,

Fdepletion is dependent on the withdrawal to availability (WTA) ratio.

I+) 1 L +) 1 . +) ! , J (16) L K0 +) 1

After determining the damage of freshwater consumption to human health, ecosystem quality, and resources, a normalization and weighting based on weighting factors from the eco-indicator 99 (hierarchist perspective) [37] can be accomplished to obtain a single-score indicator. This indicator denotes the overall damage caused by freshwater consumption and can be aggregated and compared to damage caused by other environmental interventions (e.g., emissions or waste) deriving from the product system under study.

3. Discussion

In the previous section, various methodological approaches for water footprinting have been described, which differ significantly regarding scope, information value, relevance, and data requirements. Starting from pure inventory methods, water footprinting has evolved in terms of differentiation between different types of water, different types of water use, as well as inclusion of quality and spatial information denoting local water scarcity conditions. Moreover, LCIA of freshwater use has been brought forward in terms of modeling effects on resources, ecosystems, and human health on both midpoint and endpoint levels. The simplest way to account for water use is the water inventory of an organization or product system. Despite the fact that the inventory methods rely on the same principle, %%, subtracting waste water effluents from freshwater inputs, even these methods differ with regard to differentiation of resource types, inclusion of spatial information, and water quality. The information content of databases like ecoinvent [12] and GaBi [13] is rather limited as neither geographical nor quality-related information are included up to now. Moreover, the correctness of the available data is 2010, 933 arguable as it is unclear whether all relevant water flows, especially those from the background system and cooling water run in circulation systems, are included or not. These doubts regarding the correctness of data are increased due to the fact that there are large differences (up to a factor of 10) between the water use and consumption data of materials determined from the two databases. Yet, the absence of better data sources and a commonly agreed impact assessment method make the GaBi [13] and ecoinvent [12] water inventories still the most widely applied method in addressing water use in LCA. Other concepts, such as the frameworks proposed by Vince [15], propose the inclusion of spatial and quality information in order to increase the relevance. These detailed inventories are the basis for impact assessment approaches discussed below. The virtual water and the water footprint according to Hoekstra [16] can be regarded as advanced water inventories as they take into account green and gray water, while conventional inventories only account for blue water use. Additionally, the water footprint according to Hoekstra [16] contains spatial information where water is withdrawn. Even though this information is not reflected in characterization factors, it enables denoting the fraction of water consumption occurring in water scarce areas. With regard to agricultural products, the consumption of green water, %%, the evapotranspiration of rain water, is of great importance [16] and, thus, its accounting overcomes a severe shortcoming. Gray water consumption, which denotes the volume of water necessary to dilute waste water until common quality standards are reached, can be regarded as a midpoint impact category for degradative water uses. However, as the method does not clearly define ‘common standards’ for water quality the concept should be regarded as rather vague. Depending on the thresholds for pollutants chosen as ‘common standards’, the amount of gray water will vary substantially. An advantage of the gray water concept is that it enables aggregating freshwater consumptive and degradative uses already on the inventory / midpoint level. However, the pollution of water is often covered by other impact categories such as eutrophication, acidification, eco toxicity, or human toxicity potentials [9]. Hence, one needs to pay attention to avoid double counting as waste water effluents should be regarded as either freshwater pollution or consumption but not as both. Moreover, pure inventory based water footprints can be meaningless or even misleading with regard to impact assessment, as relatively low water footprints in water scarce areas can be of more environmental relevance than large water footprints in areas where water is abundant. For that reason authors like Ridoutt and Pfister [41] point out the necessity of characterized water footprints. The authors propose a framework that accounts for blue and gray water consumption in the same way as the water footprint according to Hoekstra [16]. In contrast, green water consumption is not calculated as the total evapotranspiration of rain water but as the difference in blue water formation from green water as a consequence of land use change. Subsequently, this volumetric figure is multiplied by the regional water stress index (WSI) developed by Pfister and colleagues [38], which serves as a characterization factor and leads to a characterized water footprint measured in H2O equivalents. The authors argue that this would make water footprinting more consistent with the current practice of carbon footprinting, which also comprises the accounting of greenhouse gas emissions along with a characterization step leading to CO2 equivalents.[6] However, the developers of the water footprint according to Hoekstra [16] argue that changing the water footprints from volumetric measures to characterized indices might weaken its position in water resource management and that an aggregated index is not the intention of the method. The main argument against characterized water footprints is 2010, 934 that such footprints can also be misleading as long as the environmental impact routs are not sufficiently reflected in the impact assessment methods [42]. Next to the inventory related methods, there are also methods enabling the impact assessment of water consumption. It should however be noted that most of these methods are relatively new and have been published within the last two years. Thus, hardly any experience gained from their application in case studies is available, which only allows for a discussion on a theoretical level. The impact categories EDIP resources [17] or CExD [18] and CEENE [19] assess water consumption in the context of conventional resource consumption. The EDIP resources impact category accounts for the local depletion of freshwater and the global depletion of other resources. By enabling an aggregation of the results obtained from water consumption and the consumption of other raw materials the method enables a sound assessment of resource depletion. However, the method only addresses the depletion of resources and does not address any other effects related to water consumption. CExD and CEENE assess each resource input based on its respective exergy content. Even though the exergy concept enables aggregation of any type of resource use, it does not take into account the local scarcity of water as its exergy content is calculated based on its chemical composition or potential energy content. Thus, exergy can neither express the local depletion of water resources in a meaningful way nor account for any other consequences to human health or ecosystems related to water consumption. The ecological scarcity method [21] provides eco-factors for water use that comprise a normalization and a weighting based on WTA. The method can be adapted from Swiss conditions to the hydrological situation in any other country and, thus, enables a site specific assessment of water consumption as the local scarcity of water will determine the magnitude of the eco-factor. The authors recommend applying the eco-factors for freshwater use instead of consumption arguing that the weighting is based on a use-to-availability ratio (WTA) as well. On the one hand this enables consistency between inventory flows and weighting as both rely on water use. Furthermore, assessing water use instead of water consumption better reflects how water intense a product system really is. For example, a water consumption of 1 m³ can mean that 10 m³ of water are withdrawn and 9 m³ are released as waste water. However, 1 m³ of water consumption can also mean that 1,000 m³ of water are withdrawn and 999 m³ are released as waste water. On the other hand, only the fraction of water use that is consumed leads to water scarcity, as the remaining part is released after quality alteration and is covered by other eco-factors assessing the emission of pollutants. A clear advantage of the method is the fact that the ecological threat of water use can be aggregated and compared with other environmental impacts resulting from raw material extractions or emissions. However, one needs to be aware that the method contains a subjective weighting based on political value choices. Therefore, according to ISO 14044 the method cannot be applied in LCA studies, which are intended to be published and contain comparative assertions [22]. Another distance-to-target method has been developed by Brent [24] to promote site specific impact assessment in four South African regions. Similar to the ecological scarcity method [21] the method accounts for water use rather than consumption and the effects in terms of water can be aggregated and compared to other environmental impacts. Taking into account the subjective weighting based on political value choices and expert judgement, the method is also not applicable in LCA studies that 2010, 935 contain comparative assertions disclosed to the public [22]. Even though the method has been developed for South Africa, the procedures could be applied in other regional contexts as well. Mila i Canals and colleagues [25] developed a method that comprises a detailed accounting scheme for water use on LCI level and two impact categories to assess water use from a resource and ecosystem perspective. With regard to ecosystem impact the method assumes that only evaporative uses of surface and aquifer blue water have an effect on ecosystems. Other authors like Bayart % [8] suggest including non-evaporative uses as well, %%, the discharge of used water into another catchment area than the one where withdrawal occurred. Moreover, the water stress index [26] that is proposed as a characterization factor for freshwater ecosystem impact, is only available for the main river basins, which restricts the global applicability of the method. Additionally, the method is still lacking characterization factors describing the relevant impacts of freshwater deprivation on human health. Moreover, the separate accounting of green water consumption is recommended but no impacts on freshwater resources and ecosystems are considered. A clear advantage is the fact that this method along with the framework proposed by Ridoutt and Pfister [41] are so far the only ones which account for water losses due to changes in evapotranspiration and runoff as a consequence of land use changes. Unfortunately, this also leads to a trade-off between more detailed LCI information enabling sound LCIA and the associated data requirements which can hardly be satisfied, especially with regard to background processes. A midpoint impact category that assesses the consequences of freshwater consumption on human health, and hence overcomes a research gap of the method of Mila i Canals % [25], has been developed by Bayart and colleagues [28]. Both the inventory scheme and the characterization factors, that account for local scarcity, number of potential uses, water quality, and socio-economic adaptability, have been developed in line with the recommendations of the UNEP/SETAC Life Cycle Initiative [8]. However, as high quality water allows for more uses than low quality water, the parameters quality and functionality are interdependent and, thus, there is a danger of double counting. Next to the inventory and midpoint related methods, several endpoint oriented methods have been developed that enable damage assessment of water use to different areas of protection such as human health, ecosystem quality, and resources. Two endpoint oriented methods have been developed by Motoshita and colleagues [29, 32] that enable the quantification of damages to human health resulting from malnutrition and infectious diseases as a consequence of lacking agricultural and clean drinking water, respectively. However, especially the impact pathway linking water use to infectious diseases is controversial. Even though the method takes into account socio-economic parameters such as house connection rates to water supply and sanitation, infectious diseases are very often a consequence of poverty rather than of physical water scarcity. Even though they enable practitioners to assess a wide range of water uses, the damage oriented methods described above are still rather specific. They either focus on a particular country, a specific type of water use, a particular type of damage, or a specific impact pathway. Thus, practitioners would have to apply a whole set of methods when accomplishing case studies in which different types of water uses occur. Another problem lies in the fact that results obtained by different methods are often not comparable and are expressed in different units. 2010, 936

A more comprehensive LCIA method that aims at assessing environmental impacts of water consumption on both midpoint and endpoint level has been developed by Pfister and colleagues [38]. Yet, the limitation of only accounting for off-stream blue water consumption limits the applicability especially with regard to agricultural products where green water consumption is significant [16]. First, the midpoint impact category ‘water deprivation’ is introduced with water stress index (WSI), serving as characterization factor based on the withdrawal-to-availability (WTA) ratio. WTA is also used in the weighting of the ecological scarcity method [21], in the water stress indicator of the method of Mila i Canals % [25], and in the calculation of characterization factors in the method of Bayart and colleagues [28]. Even though it seems reasonable to develop characterization factors that express the ratio of total water use to renewable water reserves there are some problems connected with these types of indicators. For example a relatively dry country like Greece has a three times lower WTA (10%) than Germany (31%) even though the renewable water supply in Germany is much bigger [21]. This phenomenon can be explained by the higher water use in Germany. However, it illustrates the problem that obviously dry countries can have low characterization factors as long as the water use is low too, which is especially relevant for developing countries. Hence, the WTA indicator does not express the vulnerability of a region to an additional water withdrawal. Moreover, the WTA ratio only relates water use to renewable water reserves and neglects non-renewable water resources. Yet, especially if the water use exceeds the renewable water supply it is of great importance whether substantial fossil water resources are available or not as they can ‘buffer’ temporarily overexploit renewable watercourses. Besides the midpoint impact category the method introduced by Pfister and colleagues [38] comprises three endpoint impact categories assessing damage to human health, ecosystem quality, and resources in accordance to the eco-indicator 99 framework [37]. With regard to the calculation of damages to human health the authors only consider the impact pathway of malnutrition resulting from a lack of water for irrigation. Health damages that result from pure lack of drinking water are not taken into account as they result from extreme events like droughts or wars, which are not considered in LCA. Furthermore, damages to human health that result from the spread of diseases due to lacking hygiene are also neglected arguing that it is too difficult to assess such effects as they depend on local parameters. Even though this is correct, the same argument is true for health effects resulting from malnutrition. Moreover, Motoshita % [32] showed a way of determining damages to human health resulting from lack of hygiene and quantified these damages as even higher than those resulting from malnutrition. In terms of damages to resources Pfister % [38] determine the fraction of water consumption that contributes to water depletion. Subsequently, the energy required for producing the same volume from seawater desalination is determined in order to obtain a result in MJ surplus energy. However, surplus energy actually denotes the additional energy required to mine a resource (of lower concentration) after a resource extraction took place [37]. As such an energy demand can hardly be determined for renewable resources like water, the approach of Pfister and colleagues [38] to use the energy required for seawater desalination instead is understandable. However, it is more a ‘trick’ to obtain the unit required in the eco-indicator 99 concept [37] and the result can hardly be compared to or aggregated with other surplus energy demands resulting from the consumption of fossil or mineral resources. 2010, 937

Finally, the method enables aggregation of the three damage categories to one single-score eco-indicator. Similar to the ecological scarcity method [21] the single-score result enables aggregation of and comparison to other damages resulting from raw material consumptions or emissions of the product system under study. However, as the single-score aggregation contains a subjective weighting based on decisions of an expert panel, the aggregated eco-indicator 99 result cannot be applied in LCA studies that are intended to be published and contain comparative assertions [22]. In order to support the application of their method, Pfister and colleagues provide a layer [43] that can be added to the Google Earth software [44]. This tool enables an easy determination of site specific characterization factors for the midpoint and endpoint categories for thousands of water catchment areas around the world. Even though these recent efforts in terms of damage modeling complete the set of water use assessment methods, endpoint modeling is controversial in LCIA. On the one hand it enables quantifying the damage to areas of protection like human health, ecosystem quality, and resources, which is more meaningful than results of midpoint impact categories. On the other hand uncertainties increase the longer the modelled cause-effect-chain is, making the results less reliable. Yet, another advantage of endpoint modeling is the possibility of aggregating damages that result from water use and other environmental interferences like emissions or resource abstractions. Hence, it is possible to evaluate efforts aiming to save water from a more holistic perspective. In areas of no or little water scarcity it may be possible that measures to save water in industry, e.g., due to reusing and cycling of water, cause a higher environmental damage than the status quo. In such a case the damage resulting from the energy consumption of pumps or from the use of fungicides in cycling systems might be higher than the environmental damage avoided due to the decreased water consumption. In contrast, one and the same action, that is counterproductive in areas of no or little water scarcity, might be beneficial in water scarce areas where the avoided damages of water consumption would be significantly higher. Table 2 lists the different methods discussed in this paper and shows the different scopes regarding type of water and water use accounted for, as well as inclusion of spatial and quality information. Moreover, the areas of protection addressed by the method in impact assessment and the respective level in the cause-effect-chain are shown along with a statement concerning ISO 14044 [22] compliance in case of application for comparative assertions disclosed to the public. A further analysis of methods and indicators is currently accomplished by the UNEP/SETAC Life Cycle Initiative. The cooperative work of LCA experts from both academia and industry characterizes water scarcity indicators, water inventory schemes, and impact assessment methods by means of a detailed list of criteria. Next to performing a comprehensive criteria based comparison the working group aims at supporting LCA practitioners in choosing the best suitable method for a particular situation [45]. 2010, 938

Table 2. Scope of methods accounting and assessing water use in LCA. Method Type of water use Type of water Spatial Quality Impact assessment ISO 14044 [22] consumptive degradative green blue gray differen- differen- Area of Level in compliance of tiation tiation protection cause-effect- comparative assertions chain disclosed to the public Water inventories off-stream in-stream - x - x x - - x [8,12-15] [8,12-15] [12,13] [8,14,15] [8,15] Virtual water [10], water off-stream, off-stream x x x x - ecosystem midpoint (gray x footprint [11] in-stream (gray water) [11] (gray water) water) EDIP resources [17] off-stream - - x - x - resources midpoint x Exergy [18,19] off-stream in-stream - x - - - resources midpoint x (barrage water) Ecological scarcity off-stream - - x - x - resources midpoint - method [21] Brent [24] off-stream off-stream - x - x - ecosystem midpoint - Mila i Canals % [25] off-stream, - x x - x - resources & midpoint x in-stream ecosystem Bayart % [28] off-stream - - x - x x human health midpoint x Motoshita % [29] off-stream - - x - x - human health endpoint x (malnutrition) Motoshita . [32] off-stream - - x - x - human health endpoint x (infectious diseases) van Zelm % [33] off-stream - - x - x - ecosystem endpoint x (ground water) Maendly and Humbert - in-stream - x - x - ecosystem endpoint x [36] (barrage water) (barrage water) Pfister % [38] off-stream - - x - x - resources, midpoint, x ecosystem, endpoint (only midpoint and human health non-agregated endpoint results) 2010, 939

4. Recommendations for Improvement and Development

Comprehensive recommendations for the development of methods to account for water use in LCA have been provided by the UNEP/SETAC Life Cycle Initiative. After presenting the key proposals published in the initiative’s framework paper [8], recommendations for improvement and development are given based on the previous discussion.

4%% ;81# 1) 6 5 6 ,

The working group of this initiative focuses on off-stream freshwater consumptive use of blue water. Starting on the LCI level the framework suggests the provision of spatial information of water withdrawal and release to account for local scarcity conditions. Moreover, the inventory should distinguish the quality of water input and output fluxes (high or low) as well as the type of watercourse from which water is withdrawn and to which it is released (ground or surface water). With regard to LCIA the authors identified the following three elements of concern connected with water use:

Sufficiency of freshwater resource for contemporary human users Sufficiency of freshwater resource for existing ecosystems Sustainable freshwater resource basis for future generations and future uses of current generations

Based on the information compiled in the LCI, the three impact routes linking inventory data to the elements of concern should be modelled along the cause-effect-chain as described below. ,! ! ' ! % Here the authors [8] recommend differentiating between a compensation and a deficiency scenario depending on socio-economic parameters. In wealthy countries it is assumed that people do not have to suffer from deficiencies as they are able to compensate for water scarcity by e.g., seawater desalination. The environmental effects of such compensation measures can be assessed by means of conventional impact categories. In contrast, the reduced availability of freshwater in less developed countries forces humans to abstain from uses provided by the water. On the midpoint level the impact category ‘water deprivation for human uses’ expressed in ‘m³ of freshwater equivalent unavailable for humans’ is proposed. Characterization factors should take into account the regional freshwater scarcity, the number of functionalities provided by the freshwater, as well as the water quality. As stated earlier, these recommendations are put into practice in the method developed by Bayart and colleagues [28]. At the end of the cause-effect-chain, the area of protection human life comprising the endpoint categories human health and labour can be affected. ,! ! ' 7( F % On midpoint level the freshwater scarcity for ecosystems could be described in the category ‘water deprivation in ecosystems’ measured in ‘m³ of freshwater unavailable for ecosystems’. Appropriate characterization factors should account for the regional scarcity in an area as well as for the ecological value of the resource. With regard to the area of protection biotic environment, the endpoint categories ‘biotic productivity’ and ‘biodiversity’ will provide adequate indicators. 2010, 940

,! ! ' F ( % If water extraction exceeds the renewability rate, ‘water depletion’ expressed in ‘m³ of freshwater equivalent depleted’ accounts for the loss of water for future generations on the midpoint level. According to local degree of consumption and individual renewability rates, characterizations factors need to account for regional aspects. Due to high uncertainties in modeling future effects on human life and biotic environment, only the area of protection abiotic environment, comprising the endpoint category ‘abiotic natural resources’, is taken into account so far.

4% % ( C! 8

The comparison of the scopes of the methods shown in Table 2 reveals that most of the methods focus on a specific type of water use, which is off-stream freshwater consumptive use of blue water. Other water uses like in-stream and degradative uses are still underrepresented in the methodological development. Furthermore, the consumption of green water, which is especially relevant in terms of crop cultivation, is only accounted for in the methods virtual water [10], water footprint [11], and Mila i Canals % [25]. Yet, none of the three methods provides a characterization model for the assessment of negative effects resulting from this type of water use. However, as the consumption of green water can cause water deprivation for ecosystems and also may reduce the renewability of ground and surface water, this is a severe shortcoming, which should be addressed by future research efforts. Even though local water scarcity is taken into account by the latest method developments it is a general deficiency that most methods do not account for differences in terms of water quality of input and output fluxes yet. So far, only Bayart and colleagues [28] have accounted for this phenomenon, which is especially relevant when assessing effects of water use on human health. Moreover, most methods do not distinguish between water sources from which water is withdrawn and to which it is released. However, with regard to the areas of protection human health, ecosystem quality, and resources there is a difference whether water is withdrawn from a river, or lake, or aquifer. For example, the withdrawal of water from a fossil aquifer might not cause any effects on ecosystem quality. From an ecosystem perspective it might even be beneficial to withdraw fossil groundwater as it will become available to the ecosystem after its use. On the other hand the withdrawal of water from a lake with a high renewability rate might not cause resource depletion. The only methods accounting for this phenomenon are provided by Mila i Canals % [25] and Bayart and colleagues [28] who partly differentiate types of blue water use and their respective impact pathways (see Figure 1). However, this proposal also denotes the trade-off that all advanced water footprinting methods have to face—increased detail and sophistication of the methods with regard to inventory modeling and impact assessment lead to substantially increasing data requirements. Hence, the efforts to determine water footprints will increase as more and more information regarding type of water use, type of water, water quality and local scarcity are required. Especially when background processes such as the mining of raw materials, the production of semi-finished products, or the generation of electricity are taken into account, such data is costly to collect and currently not sufficiently available in public or commercial databases. As a consequence, method developers must address the trade-off between ‘scientific quality’ and ‘applicability’. From our perspective, there is currently no method or indicator that can be regarded as a broadly accepted standard—like for example the ‘global warming potential’ 2010, 941 for climate change. Simple methods like water inventories are fairly well applicable, but obviously lack relevance and information quality. More advanced methods like the impact category ‘freshwater deprivation for human uses’ [28] refer to the current state of knowledge but suffer from prohibitively high data demands. It will take substantial research efforts to develop a comprehensive and scientifically robust impact assessment method for water use in LCA—despite significant progress in recent years. In the short term, the most urgent task for the scientific community is to develop an intermediate approach between the inadequate inventory methods and the incomplete impact assessment approaches available today. The demand for better informed decision-making support on water use issues is obvious in both private and public organizations. The challenge is to satisfy this demand with a method that paves a consistent way towards tomorrow’s method refinements once the necessary data get available.

5. Conclusions

Freshwater is a vital yet often scarce resource sustaining live on our planet that needs to be managed properly to ensure human health and ecosystem quality. Hence, it is surprising that life cycle assessment—a tool to promote sustainable decision making—accounts for lots of environmental interventions, but so far often neglects water use. Having realized this shortcoming, which is especially relevant concerning agricultural products and biofuels, the life cycle assessment community has put great efforts in method development to properly address water use. On both inventory and impact assessment level, lots of accounting models have been developed. The International Organization for Standardization has recently even launched a project aiming at creating an international standard for water assessment in life cycle assessment. Taking into account the significant progress in method development, an overview of a broad range of methods developed to enable accounting and impact assessment of water use has been provided within this paper. Moreover, the individual methodological advantages as well as shortcomings have been discussed and resulting research gaps have been identified. The analysis revealed that the methodological scopes differ significantly regarding the types of water use accounted for, the inclusion of local water scarcity conditions, as well as the differentiation between watercourses and quality aspects. In conclusion there are promising methodological developments enabling sound accounting and impact assessment of water use in life cycle assessment. However, most methods focus on the assessment of off-stream consumptive use of blue water while other types of water use are underrepresented. Moreover, as different watercourses fulfill different functions, more detailed inventories and impact pathways need to be considered in water use assessment. Yet, the application of the most advanced methods requires high resolution inventory data, which can hardly be satisfied, especially with regard to background processes in the production chain. Hence, the trade-off between ‘precision’ and ‘applicability’ needs to be addressed in future studies and in the new international standard.

Acknowledgements

The authors would like to express sincere thank to Aliston Watson (Ministry of Agriculture and Forestry, New Zealand) and Vanessa Bach (Technische Universität Berlin) for their valuable 2010, 942 comments on this paper. Moreover, the feedback of Anna Kounine, Manuele Margni, and Sebastien Humbert, who are coordinating the water assessment working group of the UNEP/SETAC Life Cycle Initiative, was highly appreciated.

References

1. International Water Management Institute. + .G + 5 6! + ( ( ; Earthscan: London, UK, 2007. 2. United Nations World Water Assessment Programme. ) ; 8 + + 2 ! ! 0N+ 6(( + ; UNESCO: Paris, France, 2009. 3. Schnoor, J.L. LCA and Environmental Intelligence? 1% % ) % 2009, 40, 2997. 4. Dominguez-Faus, R.; Powers, S.E.; Burken, J.G.; Alvarez, P.J. The water footprint of biofuels: a drink or drive issue? 1% % ) % 2009, 40, 3005–3010. 5. 1 (N5 6 N#! .' I,D 44 :J; European Committee for Standardisation: Brussels, Belgium, 2006. 6. Finkbeiner, M. Carbon footprinting—opportunities and threats. ,% % 56 2009, 4, 91-94. 7. Rajan, A. Forget carbon: you should be checking your water footprint. ) ,!, 21 April 2008. 8. Bayart, J.B.; Bulle, C.; Koehler, A.; Margni, M.; Pfister, S.; Vince, F.; Deschenes, L. A framework for assessing off-stream freshwater use in LCA. ,% % 56 2009, (submitted). 9. Guinee, J.B.; de Bruijn, H.; van Duin, R.; Gorree, M.; Heijungs, R.; Huijbregts, M.A.J.; Huppes, G.; Kleijn, R.; de Koning, A.; van Oers, L.; Sleeswijk, A.W.; Suh, S.; Udo de Haes, H.A. 5 6 N D! C ,D ; Centre of Environmental Science, Leiden University (CML): Leiden, The Netherlands, 2001. 10. Allan, J.A. Virtual water: a strategic resource, global solutions to regional deficits. C + 1998, 36, 545–546. 11. Hoekstra, A.Y.; Hung, P.Q.; Virtual Water Trade: A Quantification of Virtual Water Flows between Nations in Relation to International Crop Trade. In - + ! ; UNESCO-IHE: Delft, The Netherlands, 2002. 12. 6 Homepage. Available online: http://www.ecoinvent.org/ (accessed on 12 March 2009). 13. #1 , Homepage. Available online: http://www.gabi-software.com (accessed on 8 July 2009). 14. +&62 Homepage. Available online: http://www.wbcsd.org/web/watertool.htm (accessed on 30 June 2009). 15. Vince, F. #! 56, .'; 2007, (unpublished). 16. + .! 8' /!(. Available online: http://www.waterfootprint.org (accessed on 23 June 2009). 17. Hauschild, M.; Wenzel, H. 1 #; Chapman & Hall, Thomson Science: New York, NY, USA, 1998; Volume 2. 18. Bösch, M.E.; Hellweg, S.; Huijbregts, M.A.J.; Frischknecht, R. Applying Cumulative Exergy Demand (CExD) Indicators to the ecoinvent Database. ,% % 56 2007, , 181–190. 2010, 943

19. Dewulf, J.; Bosch, M.E.; De Meester, B.; van der Vorst, G.; van Langenhove, H.; Hellweg, S.; Huijbregts, M.A.J. Cumulative Exergy Extraction from the Natural Environment (CEENE): A Comprehensive Life Cycle Impact Assessment Method for Resource Accounting. 1% % ) % 2007, 4, 8477–8483. 20. Verein Deutscher Ingenieure (VDI). 9 1(F&(G 2G &(; VDI 4600; Beuth Verlag GmbH: Berlin, Germany, 1997. 21. Frischknecht, R.; Steiner, R.; Jungbluth, N. ' 3 ' ( 9!!N'' :N +'( 3( ' 3; Öbu-Netzwerk für nachhaltiges Wirtschaften: Bern, Swizerland, 2009. 22. ISO. 1 (N5 6 NE C I,D 444 :J; European Commitee for Standardisation: Brussels, Belgium, 2006. 23. D162 9 1 ,; OECD Environment Directorate: Paris, France, 2004. 24. Brent, A.C. A life cycle impact assessment procedure with resource groups as area of protection. ,% % 56 2004, B, 172–179. 25. Mila i Canals, L.; Chenoweth, J.; Chapagain, A.; Orr, S.; Anton, A.; Clift, R. Assessing freshwater use in LCA: Part I—inventory modelling and characterisation factors for the main impact pathways. ,% % 56 2008, 4, 28–42. 26. Smakhtin, V.; Revenga, C.; Döll, P. A pilot global assessment of environmental water requirements and scarcity. + ,% 2004, B, 307–317. 27. Raskin, P.; Gliek, P.; Kirshen, P.; Pontius, G.; Strzepek, K. + . 5(F( # # ; Swedish Environment Institute/United Nations: Stockholm, Sweden, 1997. 28. Bayart, J.B.; Bulle, C.; Margni, M.; Vince, F.; Deschenes, L.; Aoustin, E. Operational characterisation method and factors for a new midpoint impact category: freshwater deprivation for human uses. In #( 1) 6 1! B (, Gothenborg, Sweden, 31 May–4 June 2009. 29. Motoshita, M.; Itsubo, N.; Inaba, A. Development of impact assessment method on health damages of undernourishment related to agricultural water scarcity. In #( 1( , 6 1& , Tokyo, Japan, 10–12 December 2008. 30. Murray, C.; Lopez, A. ) C & 2; Harvard School of Public Health (on behalf of the World Health Organization, World Bank): Boston, MA, USA, 1996. 31. Prüss-Üstün, A.; Bos, R.; Gore, F.; Bartram, J. +G & / ; World Health Organization: Geneva, Swizerland, 2008. 32. Motoshita, M.; Itsubo, N.; Inaba, A.; Aoustin, E. Development of damage assessment model for infectious diseases arising from domestic water consumption. In #( 1) 6 1! B (, Gothenburgborg, Sweden, 31 May–4 June 2009. 33. Van Zelm, R.; Rombouts, M.; Snepvangers, J.; Huijbregts, M.A.J.; Aoustin, E. Characterization factors for groundwater extraction based on plant species occurence in The Netherlands. In #( 1) 6 1! B (, Gothenborg, Sweden, 31 May–4 June 2009. 2010, 944

34. McDonald, M.G.; Harbaugh, A.W. A modular three-dimensionalfinite-difference groundwater flow model. In )E +F ,(; United States Geological Survey: Denver, CO, USA, 1988. 35. Bakkenes, M.; De Zwart, D.; Alkemade, J.R.M. D-1 -( 0% ( ; National Institute of Public Health and the Environment: Bilthoven, The Netherlands, 2002. 36. Maendly, R.; Humbert, S. Empirical characterization model and factors assessing aquatic biodiversity damages of hydropower water use. ,% % 56 2009, (submitted). 37. Goedkoop, M.; Spriensma, R. ) 1F BBN 2( D 5 6 ,! G ( !; Product Ecology Consultants (PRe): Amersfoort, The Netherlands, 2001. 38. Pfister, S.; Koehler, A.; Hellweg, S. Assessing the environmental impacts of freshwater consumption in LCA. 1% % ) % 2009, 40, 4098–4104. 39. Alcamo, J.; Doll, P.; Heinrichs, T.; Kaspar, F.; Lehner, B.; Rosch, T.; Siebert, S. Development and testing of the WaterGAP 2 global model of water use and availability. / (% % % 2003, 4>, 317–337. 40. Nemani, R.R.; Keeling, C.D.; Hashimoto, H.; Jolly, W.M.; Piper, S.C.; Tucker, C.J.; Myneni, R.B.; Running, S.W. Climate-driven increases in global terrestrial net primary production from 1982 to 1999. 2003, 0, 1560–1563. 41. Ridoutt, B.G.; Pfister, S. A revised approach to water footprinting to make transparent the impacts of consumption and production on global freshwater scarcity. C % 1% 6( 2010, , 113–120. 42. Hoekstra, A.Y.; Gerbens-Leenes, W.; van der Meer, T.H. Reply to Pfister and Hellweg: Water footprint accounting, impact assessment, and life-cycle assessment. #8 2009, :, 114. 43. #G % Homepage. Available online: http://www.ifu.ethz.ch/staff/stpfiste (accessed on 7 January 2010). 44. C( ,% Homepage. Available online: http://earth.google.com/ (accessed on 7 January 2010). 45. Kounina, A. Review of methods addressing water in life cycle assessment. In #( 5 6 ,O, Boston, MA, USA, 29 September–2 October 2009; pp 108–109.

© 2010 by the authors; licensee Molecular Diversity Preservation International, Basel, Switzerland. This article is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/). Article

pubs.acs.org/est

Water Footprint of European Cars: Potential Impacts of Water Consumption along Automobile Life Cycles † ‡ ‡ † † Markus Berger,*, Jens Warsen, Stephan Krinke, Vanessa Bach, and Matthias Finkbeiner † Technische Universitaẗ Berlin, Department of Environmental Technology, Chair of Sustainable Engineering, Office Z1, Strasse des 17. Juni 135, 10623 Berlin, Germany ‡ Volkswagen AG, Group Research, Environmental Affairs Product, P.O. Box 011/1774, 38436 Wolfsburg, Germany

*S Supporting Information

ABSTRACT: Due to global increase of freshwater scarcity, knowledge about water consumption in product life cycles is important. This study analyzes water consumption and the resulting impacts of Volkswagen’s car models Polo, Golf, and Passat and represents the first application of impact- oriented water footprint methods on complex industrial products. Freshwater consumption throughout the cars’ life cycles is allocated to material groups and assigned to countries according to import mix shares or location of production sites. Based on these regionalized water inventories, consequences for human health, ecosystems, and resources are determined by using recently developed impact assessment methods. Water consumption along the life cycles of the three cars ranges from 52 to 83 m3/car, of which more than 95% is consumed in the production phase, mainly resulting from producing iron, steel, precious metals, and polymers. Results show that water consumption takes place in 43 countries worldwide and that only 10% is consumed directly at Volkswagen’s production sites. Although impacts on health tend to be dominated by water consumption in South Africa and Mozambique, resulting from the production of precious metals and aluminum, consequences for ecosystems and resources are mainly caused by water consumption of material production in Europe.

■ INTRODUCTION Therefore, the aim of this study is to analyze freshwater “Yet Another Footprint to Worry About: Water” was a headline consumption along the life cycles of three Volkswagen car in The Wall Street Journal1 with regard to the foundation of the models on both inventory and impact assessment levels. First, Water Footprint Network2 in 2008, which published regionalized water inventories are determined, showing surprisingly high figures of 70 L virtual water consumption3 country-specific water consumption figures, for the Polo 1.2 per apple or 2700 L per cotton T-shirt. Starting from such turbocharged direct injection (TDI), Golf 1.6 TDI, and Passat volumetric tools, which simply aggregate consumptions of 2.0 TDI (model year 2010). Based on these inventories, seven ground and surface water (blue water), soil moisture (green 4 impact assessment methods, which represent different levels of water ), and volumes of polluted freshwater (gray water), sophistication and model different impact pathways, are substantial methodological developments were undertaken applied. Further objectives comprise the discussion and recently. Modern impact-oriented water footprinting methods, comparison of impact assessment results, the identification of which were reviewed in a previous work,5 characterize water consumption based on parameters such as local scarcity or significant life cycle stages and processes, and the analysis of sensitivity of population and ecosystems and model complex sensitivity of results to altered regionalization scenarios. Finally, impact pathways. However, these methods were hardly tested the potential damages resulting from water consumption are or applied in complex industrial product systems. So far, most compared to damages caused by other environmental water footprint studies published focus on agricultural products interferences, such as resource use and emissions, in order to such as food,6,7 natural fibers,8,9 or bioenergy10 and biofuels.11 estimate the relevance of water for the automotive industry. Volkswagen has been analyzing the environmental effects of its cars and components by means of life cycle assessment Received: November 9, 2011 12,13 14 (LCA) for many years. However, due to lack of data and Revised: January 28, 2012 appropriate impact assessment models, the consumption of Accepted: March 5, 2012 freshwater has not yet been considered. Published: March 5, 2012

© 2012 American Chemical Society 4091 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article ■ METHODOLOGY The production phase comprised the mining of raw materials such as iron ore, the production of materials such as steel, the Water Inventory. Figure 1 shows the system boundaries in processing of components like trunk lids, and final assembly of which water consumption was analyzed along the life cycles of the car. In the use phase, crude oil production and the refinery the three cars, comprising the production, use, and end-of-life of the diesel required to run the cars for 150 000 km in the phases. It should be noted that water consumption denotes New European Driving Cycle (NEDC)16 were included. only the fraction of total water use that is not returned to the Accordingly, diesel production of 5700 L (Polo), 6750 L same river basin from which it was withdrawn due to (Golf), and 8550 L (Passat) was considered. Washing of cars evaporation, product integration, or discharge into other was not included in this analysis as it highly depends on watersheds and seawater.15 Transports between process steps, individual use, personal attitude, and the technique applied, for the generation of electricity, and the production of auxiliary which no reliable water consumption (evaporation) figures are materials were within the scope of this study but not shown available. Moreover, the washing of cars was not included in ’ explicity in Figure 1. Volkswagen s existing LCA studies upon which this study is based. Because we intended to compare impacts resulting from water consumption to damages caused by other environmental interferences, consistent system boundaries had to be ensured. Finally, it was expected that the contribution of car washing to water evaporation is rather small, but this assumption should be validated in future studies. The end-of-life phase was modeled in accordance with Volkswagen’s SiCon recycling process.17 It comprises the draining of fluids, the removal of batteries, catalysts, and spare parts, the shredding of the remaining body, the treatment and recycling of shredding residues, and the disposal of wastes. In contrast to conventional recycling approaches, this process allows for the recycling of nonmetallic shredder residues as well and enables recycling rates of 95% by weight. Environmental credits gained from the use of secondary materials in subsequent product systems were not considered. The process of modeling a life cycle inventory (LCI) of a whole car is very complex due to the fact that it involves registering thousands of components, together with any related upstream supply chains and production processes. Therefore, Volkswagen developed the slimLCI interface system,18 which enables a consistent data collection and automated modeling of the LCI in the GaBi LCA software.19 A description of the slimLCI procedure along with representative extracts of the LCI model created in the LCA software can be found in the Supporting Information. The LCI data used in this study were originally determined for the environmental commendations of − the Polo, Golf, and Passat,20 22 which are reviewed by independent experts according to ISO 14040/44 (2006).12,13 To apply impact assessment methods evaluating the consequences of the water consumption determined from the LCI models, the basic volume is not enough. Regionalized water inventories, which state the location where water consumption occurs, are needed to consider regional water scarcity conditions, the vulnerability of ecosystems, or socio- economic parameters affecting the sensitivity to water scarcity induced health damages.5 Such geographically explicit water inventories were determined in a top-down approach. First, the car’s total water consumption was divided into the shares consumed by the life cycle stages production, use, and end-of- life. For further specification, the water consumed in the production phase is assigned to manufacturing steps and to 15 material groups (specified in the German material classification in motor vehicle construction standard23). Finally, the water consumption caused by the manufacturing steps and material groups was allocated to specific countries based on production mixes, location of suppliers, production sites, etc. For instance, water consumption in the painting, final assembly, and recycling of the Polo, Golf, and Passat was Figure 1. Life cycle of the cars along which water consumption was ’ analyzed. assigned to Volkswagen s production sites and recycling operators in Germany and Spain. Water consumed by the

4092 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article material group aluminum was allocated to countries contribu- In general, country-specific characterization factors were used ting to the European aluminum import mix proportionally to in this study. As they reflect hydraulic conditions in countries their import shares.24 For polymers, water consumption was with inhomogeneous water scarcity, like Spain or the United further divided into water consumed in oil extraction, refinery, States, more realistically, it would be preferable to use polymerization, and component fabrications based on generic watershed-specific factors throughout this study. However, data available in the GaBi database.19 These water consumption this data is not available and cannot be generated in our top- shares were assigned to countries based on the European down approach described above. Moreover, depending on the import mixes of crude oil,19 the location of refineries and car, between 45 and 60% of water consumption occurs in polymerization plants,25 and on the location of plastic countries with similar water scarcity throughout the territory component productions conducted at suppliers and Volkswa- such as Germany, Sweden, or Russia. Yet, in order to specify gen production sites. the impacts of water consumed at Volkswagens production sites Impact Assessment. Several impact assessment methods in Pamplona, Spain (Polo), Wolfsburg, Germany (Golf), and are available to evaluate consequences for human health, Emden, Germany (Passat) watershed-specific characterization ecosystems, and resources resulting from water consumption. factors were applied. In the ecological scarcity method, In addition to spatially explicit water inventories, some methods characterization factors for the catchment areas Ebro require information on the type of watercourse,26 water (Pamplona, Spain), Weser (Wolfsburg, Germany), and Ems qualities,27 or time of use28 as different watercourses fulfill (Emden, Germany) were determined using data from the different ecological functions, different qualities enable different WaterGAP 2 model.33 For the five characterization models uses, and scarcity can vary throughout the year. However, such provided by Pfister and colleagues,9 watershed-specific factors high inventory requirements are hard to satisfy as the desired were derived from a Google Earth34 layer provided by the information is not available in today’s LCI databases.19,25 For authors.35 Only for the method of Motoshita et al. (2011)* no that reason we were restricted to apply methods for which site-specific characterization factors, which are influenced by regionalized water inventories are sufficient. parameters such as house connection rate to water supply and To allow for a comprehensive impact assessment the sanitation, were determined as they are very low and similar following methods, representing different levels of sophisti- throughout Europe. cation and assessing different consequences, have been Uncertainties and Sensitivity Analysis. Uncertainties selected: The ecological scarcity method,29 which assesses result from the LCI modeling, LCI databases, assumptions to water consumption based on physical water scarcity, measured establish the regionalized water inventories, and impact in eco-points/m3. The impact assessment method of Motoshita assessment models. Inevitable uncertainties resulting from the et al. (2011),30 which evaluates damage to human health caused LCI modeling of complex industrial products are discussed in by infectious diseases resulting from polluted water uptake as a Koffler et al. (2008)18 but cannot be quantified in the scope of consequence of domestic water scarcity, expressed in disability this study. Although efforts have been made to provide high- adjusted life years (DALY)/m3. The method of Pfister et al. quality water consumption data, limitations occur due to the (2009),9 comprising 5 characterization models: Freshwater lack of data in current LCI databases. For instance, the deprivation, which assesses freshwater consumption based on ecoinvent database25 reports only freshwater withdrawals but physical water scarcity (dimensionless); Damage to human no wastewater discharges in its data sets and, thus, allows only health, which addresses health impacts resulting from for the determination of water use but not consumption. Data malnutrition as a consequence of agricultural water shortage, sets from the GaBi database19 used in this study contain quantified in DALY/m3; Damage to ecosystem quality, which consistent water in- and output figures;36 however, often only evaluates ecological consequences resulting from decreased for the foreground system and for energy production but not biodiversity due to water shortage, measured in potentially for processes in the background system such as mining. As disappeared fraction of species (PDF) m2 year/m3; Damage to water consumption in electricity production provides a major resources, which assesses depletion of freshwater resources as a share of total water consumption in industrial processes,37 the consequence of water uses exceeding renewability rates, data sets can be used for water footprint calculations but may 3 expressed in MJsurplus energy/m ; and Overall damage, which underrepresent the real water consumption. These uncertain- aggregates impacts determined in the three previous character- ties, which result from partly lacking water consumption figures ization models to a single-score result, quantified in points/m3. in the background system of LCI data sets, can hardly be It should be noted that the ecological scarcity method uses in quantified without detailed insight into aggregated data sets general an average, policy-based approach while Motoshita et available in the GaBi database. Also methodological un- al.30 and Pfister et al.9 follow a marginal impact approach.31 The certainties of the impact assessment methods cannot be characterization factors which were used to calculate the assessed statistically, but they are addressed by applying seven impacts resulting from water consumption in different different characterization models and comparing the results countries are provided in the Supporting Information of the obtained. methods of Motoshita et al.30 and Pfister et al.9 Since There are few uncertainties from the determination of Motoshita et al.30 determined damage factors only for domestic regionalized water inventories for the manufacturing steps in water consumption, these factors were multiplied by a country- the foreground system. Yet, several assumptions were necessary specific ratio of domestic to total water use32 in order to allow to geographically differentiate water consumed in the back- for an assessment of general water consumption. Therefore, the ground system: It was assumed that materials are purchased method is termed Motoshita et al. (2011)* in the following. according to average import mix sharesspecific information For the ecological scarcity method only few characterization was not included. By assigning the water consumption of factors for OECD countries are provided. Factors for non- material groups to countries based on import mix shares, it is OECD countries were calculated using hydrological data from presumed that the water intensity for producing a certain AQUASTAT.32 material is equal in all countries. The assumption that material

4093 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article production is accomplished exclusively in countries contributing to the import mix neglects the fact that minor volumes of water consumed during the production of auxiliary products might have been consumed in countries other than those included in the mix. Because regionalized water inventories are a prerequisite for all impact assessment methods, the uncertainties mentioned above cannot be avoided, however, they can be quantified by means of sensitivity analysis. As several different possibilities concerning the geographical differentiation are feasible, we decided to set up a minimum and a maximum scarcity scenario. In the minimum scarcity (min-s) scenario the individual water consumption of the 15 material groups was assigned to the countries in the corresponding import and production mixes Figure 2. Relative contributions of life cycle stages to total results in which show the lowest physical water scarcity. In contrast, in the impact categories eutrophication (EP), ozone layer depletion the maximum scarcity (max-s) scenario the material group (ODP), photochemical ozone creation (POCP), global warming specific water consumption was fully allocated to the water (GWP), acidification (AP), and in the water consumption inventory scarcest country in the respective import and production mixes. (WC) for the Golf 1.6 TDI. Physical water scarcity was measured by means of the withdrawal-to-availability (WTA) ratio, which relates annual share of special metals (gold, silver, and platinum group metals freshwater use to the renewable water supply in a country. (PGM)) is higher than in conventional impact categories. Water consumption in the foreground system such as the However, these figures tend to overestimate the actual share of manufacturing and recycling of the Polo, Golf, and Passat special metals, as their supply has been modeled with 100% of remained assigned to Germany and Spain in both scenarios. As primary material. Yet, Volkswagen has been running an it might be too optimistic or too pessimistic to assume that all effective catalysts recycling program for years, which helps to materials were derived exclusively from the countries of the recover and recycle PGM in a closed loop system. Nevertheless, lowest or the highest water scarcity, the scenarios should be the fact that less than 1 kg of precious metals is responsible for regarded as boundaries between which realistic options are more than 20% of the overall water consumption throughout a possible. Golf’s life cycle highlights the large material specific water consumption of these materials. ■ RESULTS AND DISCUSSION After assigning the water consumption of materials and Water Inventory. The water consumption along the life production steps to countries based on import mixes, location cycles of the three cars amounts to 51.7 m3 (Polo 1.2 TDI), of production sites, etc., regionalized water inventories were 62.4 m3 (Golf 1.6 TDI), and 82.9 m3 (Passat 2.0 TDI). It established for the three cars. As shown in Figure 4, water should be noted that our figures are lower than previously consumption takes place in 43 countries worldwide. Less than reported data of 400 m3 virtual water consumption per car.38 10% is consumed directly at the production sites in Pamplona, This result was calculated based on economic input−output Wolfsburg, and Emden resulting mainly from painting and tables determining the water consumption per US$ of industrial evaporation of cooling water. Hence, more than 90% of the product.2 Hence, it only represents a rather rough estimate of water consumption along the cars’ life cycles is caused by the an average industrial product and is based on economic, not material and energy production in the background system. physical data. This fundamental difference in the modeling Detailed geographically explicit water inventories for the Polo, approach does not allow for a detailed discussion of differences. Golf, and Passat are available in the Supporting Information, Determining the water consumption of the main life cycle showing country-specific water consumption figures for stages, revealed that about 95% of the water is consumed in the production (separated by the assembly and production of 15 production phase of all three cars (Figure 2). This is in contrast material groups), use, and end-of-life. with most other environmental interferences evaluated in Impact Assessment. Based on the regionalized water Volkswagen’s LCA studies, like eutrophication or global inventories, the impact assessment models of the ecological warming,39 which are usually dominated by the car’s use scarcity method, Motoshita et al.30 and Pfister et al.9 were phase.21 Hence, it can be seen that different processes than applied in order to evaluate consequences resulting from water fossil fuel consumption are relevant from a water perspective. consumption in different countries. Figure 5 shows the results Yet, it should be remembered that these results were obtained obtained by means of the water inventory and impact assuming the use of fossil diesel, whichaccording to our assessment methods normalized to the Polo for the default, datahas a low water consumption of 0.005 L/MJ compared min-s, and max-s scenarios. Because absolute results for the to biodiesel consuming 217−335 L/MJ on global average Polo differ among the scenarios, Figure 5 only allows for depending on the crop used.10 comparing the cars within one impact category and scenario. A significance analysis was accomplished to identify the Comparisons of results obtained per car and impact category in contributions of individual materials and manufacturing steps to different scenarios are shown separately in Figure 6. the impact assessment and water consumption results of the The results of the ecological scarcity method and the impact production phase for the Golf. As shown in Figure 3, steel and category freshwater deprivation depend on two factors: the iron materials, as well as polymers, contribute equally strong to volume of water consumed and the physical water scarcity at most impact categories and water consumption (70−80%). In the place of consumption. Whereas the ecological scarcity contrast, the contribution of light metals (aluminum and method uses the WTA ratio as a weighting factor directly, magnesium alloys) to total water consumption is lower and the freshwater deprivation uses a water stress index (WSI) as a

4094 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article

Figure 3. Relative contributions of material groups to production impacts in the impact categories eutrophication (EP), ozone layer depletion (ODP), photochemical ozone creation (POCP), global warming (GWP), acidification (AP), and in the water consumption inventory (WC) for the Golf 1.6 TDI. characterization factor which is based on WTA, but additionally Comparison Among Cars. By showing the results of the considers seasonal variation of water availability.9 Despite water inventory and impact assessment methods normalized to different proportions, both methods are dominated by the the Polo, Figure 5 allows for a comparison of the three cars. In water consumption in similar countries mainly Germany (due the default scenario it can be seen that the increased water to high volumes), as well as Spain, Belgium, and South Africa consumption of the Golf and Passat are to a similar extent (due to high scarcity). Figure S3 in the Supporting Information reflected by the ecological scarcity method and the model of shows the water consumption in areas of different water stress Motoshita et al. (2011)*, showing that these methods lead to throughout the lifecycles of the three cars. similar conclusions as the inventory in this scenario. Yet, in the Whereas the method of Pfister and colleagues9 assessing categories developed by Pfister et al.,9 the impacts of the Polo health damages from malnutrition considers physical water and Golf are regarded as rather similar despite different water scarcity and socio-economic aspects, the method of Motoshita consumption. This can be explained by two facts. First, similar et al. (2011)* measuring health damages from infectious water consumption is weighted higher at the Polo’s production diseases considers only socio-economic aspects. As physical site in Spain than at the Golf’s production site in Germany. water scarcity is high and the level of development is rather low, This compensates the advantages of the lower water human health impacts measured according to Pfister et al.9 are consumptions in the material production resulting from the dominated by the water consumed in South Africa resulting reduced weight of the Polo in comparison to the Golf. Second, from the PGM production. Damages determined in the method some impact categories, especially the one developed by Pfister of Motoshita et al. (2011)* are mainly caused by relatively low and colleagues9 measuring damages to human health, are amounts of water (78−191 L) consumed in the aluminum dominated by the water consumption of the PGM production production in Mozambique. In contrast, due to high sanitation in South Africa. As the PGM contents of the Polo and Golf are standards and a high degree of development, the water comparable, results of these impact categories are similar, too. consumption in countries like Spain or Australia does not Since the Passat contains more PGM than the Polo and Golf, cause damages to human health, despite high physical water the same reasoning can explain the higher impacts in the scarcity in these countries. human health categories. In contrast, the water consumption in Ecosystem damage denotes the loss of biodiversity and is South Africa hardly affects damage to resources since WTA is influenced by water scarcity and the local sensitivity of vascular below 1 in most watersheds, which, according to Pfister et al.,9 plants.9 Again, the water consumption in South Africa means that no depletion of water resources occurs. For that dominates the impact assessment result with 56% (Golf) to reason the Passat scores only slightly worse in this impact 67% (Passat). Damages caused by the depletion of resources category due to the larger water consumption of the larger only occur in countries where water withdrawal exceeds the material production. renewability rate (WTA > 1). As this is not the case in Central Sensitivity Analysis. Beause the regionalization of water Europe, where most of the water is consumed, large shares of inventories contains several assumptions and impact assessment water consumption do not contribute to resource damage. This results strongly depend on local aspects, a sensitivity analysis impact category is dominated by the water consumption in was accomplished. The water consumption of the material Spain and Ukraine which contribute 55% (Passat) to 67% groups was assigned to the countries with the lowest and the (Polo) to the overall result depending on the car. In addition to highest physical water scarcity available in the material-specific this comparison of the results obtained by different impact import or production mixes. Figure 6 displays the differences assessment methods, a critical evaluation of the underlying between the max/min-s scenarios and the default scenario in characterization models can be found in Berger and relation to the default scenario. Hence, a difference of +100% Finkbeiner.40 means that impacts have doubled compared to the default

4095 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article

models with the exception of Motoshita et al. (2011)*, where the min-s scenario leads to higher impacts than the max-s scenario. This can be explained by the fact that this method is only sensitive to socio-economic parameters and does not account for physical water scarcity which was used as an indicator to define min-s and max-s. Comparison to Other Environmental Interferences. An advantage of the ecological scarcity method is that it enables a comparison of water consumption related impacts to impacts resulting from other environmental interventions like the consumption of fossil and mineral resources or emissions to air, water, and soil. In a similar way, the method of Pfister et al.9 allows for a comparison to other environmental damages determined by means of the eco-indicator 99 method.41 Also impacts on human health determined according to Motoshita et al. (2011)* can be compared to health damages calculated by eco-indicator 99. In that way the contribution of water consumption to total impacts caused along automotive life cycles and, thus, the relevance of water consumption to the automotive industry can be estimated. As shown in Figure 7 for the production of the Golf, water consumption affects the total results to an extent of 0−7% depending on the impact category. While similar results were identified for the Polo slightly higher percentages ranging from 0 to 13% were obtained for the Passat (Figure S4 in the Supporting Information). For all cars it was shown that water consumption mainly affects damages to ecosystems rather than damages to human health and resources. This can be explained by the fact that most of the water consumption occurs in Europe, where the level of development avoids water induced health damages. Moreover, water use in Europe usually does not exceed the renewability rate, which prevents the depletion of freshwater resources. The rather low contribution of water consumption to the overall damage can be explained by the application of normalization and subsequent weighting of the three damage categories (human health 40%, ecosystems 40%, resources 20%) in the eco-indicator 99 methodology. It should be noted that these figures only reflect the production phase of cars. When considering the whole life cycle, the share of water related damages decreases even more as other environmental interferences increase significantly while water consumption and related impacts stay rather constant. Furthermore, a comparison of damages to human health and Figure 4. Global water consumption throughout the life cycles of (a) ecosystem quality resulting from water consumption and the Polo 1.2 TDI, (b) the Golf 1.6 TDI, and (c) the Passat 2.0 TDI. freshwater pollution in the production of the Golf was accomplished. Impacts caused by water pollution result from scenario. A result of 0% indicates no changes and a result of the emission of carcinogenic substances in the category human −100% means that impacts are zero in this scenario. health and from the release of acidifying, eutrophying, and eco- More than 70% of water consumption derives from the toxic substances in the category ecosystem quality.41 As shown production of steel and iron materials as well as polymers, in Figure S5 in the Supporting Information, damages from which occurs mainly in Central and Northern Europe (except water consumption are smaller than damages resulting from the iron ore and crude oil production). Hence, local shifting of emissions into freshwater in both damage categories. large water consumption between the scenarios takes place Recommendations. Even though water consumption within Europe. Only the shift of water consumption deriving tends to be of minor relevance for European passenger cars from the PGM production between South Africa and Russia in run with petrol-based fuels, it can cause significant impacts in the max-s and min-s scenarios causes significant changes in the agricultural products such as food,6,7 natural fibers,8,9 or regionalized inventories outside Europe. As water scarcity, biofuels.10,11 However, conclusions drawn from water footprint vulnerability of ecosystems, and socio-economic parameters are studies that rely on current LCI databases have to be handled rather similar throughout Central and Northern Europe, with care. First, data sets either only contain water use figures variations in the impact assessment results were mainly caused (ecoinvent) or tend to underestimate water consumption due by the shifting of water consumption between South Africa and to the partly ignorance of water consumed in background Russia. As it can be seen in Figure 6, variations from the default processes (GaBi). Second, as water flows are not geographically scenario range from −100 to +150% in all impact assessment differentiated, uncertainties resulting from the top-down

4096 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article

Figure 5. Relative comparison of results on the inventory and impact assessment levels for the Polo, Golf, and Passat normalized to the results obtained for the Polo in the default scenario (bars), the min-s scenario (circles), and the max-s scenario (diamonds).

regionalization of water inventories are added. Such uncertainties could be avoided if spatially differentiated water flows were available in the LCA databases, as it is already common practice for fossil energy carriers to consider different calorific values.19 The general procedure to establish regionalized water inventories presented in this work can be used as an approximation, enabling the application of impact assessment models in complex product systems. We refrained from publishing fixed allocation keys, as the assignment of water consumption to specific countries took Volkswagen’s specific supply situation into account. Hence, it should be decided for each case study if for a certain material the global production mix, a national import mix, or the location of a specific supplier should be used. In terms of method development we currently observe a Figure 6. Relative differences in impact assessment results between trend toward sophisticated end-point impact assessment max/min-s scenarios and the default scenario in relation to results methods.5 Even though these methodological developments obtained in the default scenario. are appreciated from a scientific point of view, such methods often require even more inventory data. In addition to the spatial differentiation of water flows, the type of watercourse used,26 quality data,27 or even temporal information28 need to be known. However, such information is hardly available and costly to collect, especially if complex background systems are involved. Some authors developed water categories42 and provide default characterization factors27 in order to reduce inventory demands. Yet, often not even these requirements can be fulfilled by today’s inventory databases. Therefore, efforts should be put into the development of both more detailed inventory data sets and robust and applicable impact assessment methods, in order to promote the important assessment of water consumption and its consequences in LCA and other disciplines.

Figure 7. Relative contribution of water consumption in the ■ ASSOCIATED CONTENT production of a Golf 1.6 TDI to the total impacts according to the *S Supporting Information ecological scarcity method and the impact assessment models of eco- indicator 99 (hierarchist approach) and Motoshita et al. (2011)* and Additional results, figures, and country specific water Pfister et al. (2009).9 consumption of all life cycle stages. This material is available free of charge via the Internet at http://pubs.acs.org.

4097 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article ■ AUTHOR INFORMATION (16) European Union. Directive 98/69/EC of the European Parliament and of the Council of 13 October 1998 relating to measures to be taken Corresponding Author against air pollution by emissions from motor vehicles and amending *E-mail: [email protected]; phone: +49.(0)30.314- Council Directive 70/220/EEC.EuropianParliament:Brussels, 25084; fax: +49.(0)30.314-21720. Belgium, 1998; http://eur-lex.europa.eu/LexUriServ/LexUriServ. Notes do?uri=CONSLEG:1998L0069:19981228:EN:PDF. (17) Volkswagen AG. Life Cycle Assessment of End-of-Life Vehicle The authors declare no competing financial interest. Treatment; Group Research, Environment Affairs Product: Wolfsburg, Germany, 2005; http://www.volkswagen.com/vwcms/master_public/ ■ ACKNOWLEDGMENTS virtualmaster/en2/unternehmen/environmental_commendations. html. The research grant for this study was provided by the (18) Koffler, C.; Krinke, S.; Schebek, L.; Buchgeister, J. Volkswagen Volkswagen AG, Group Research. We thank the Center for slimLCI - a procedure for stream-lined inventory modelling within Life Environmental Systems Research (CESR) at the University of Cycle Assessment (LCA) of vehicles. Int. J. Veh. Des. 2008, 46 (2), Kassel for providing watershed specific WTA ratios for the 172−188 (special issue). basins Ebro, Weser, and Ems. Moreover, the provision of (19) PE International Website. http://www.gabi-software.com. updated characterization factors by Dr. Masaharu Motoshita (20) Volkswagen AG. The Polo - Environmental Commendation; from the National Institute of Advanced Industrial Science and Group Research - Environment Affairs Product: Wolfsburg, Germany, Technology (AIST) and fruitful discussions with Dr. Stephan 2010; http://www.volkswagen.com/vwcms/master_public/ Pfister from the Institute of Environmental Engineering at the virtualmaster/en2/unternehmen/environmental_commendations. Swiss Federal Institute of Technology Zurich are highly html. (21) Volkswagen AG. The Golf - Environmental Commendation; appreciated. Group Research - Environment Affairs Product: Wolfsburg, Germany, 2010; http://www.volkswagen.com/vwcms/master_public/ ■ REFERENCES virtualmaster/en2/unternehmen/environmental_commendations. (1) Alter, A. Yet Another 'Footprint; to Worry About: Water. The html. Wall Street Journal 17 February 2009. (22) Volkswagen AG. The Passat - Environmental Commendation; (2) Water Footprint Network Website; http://www.waterfootprint. Group Research - Environment Affairs Product: Wolfsburg, Germany, org. 2010; http://www.volkswagen.com/vwcms/master_public/ (3) Allan, J. A. Virtual water: A strategic resource, global solutions to virtualmaster/en2/unternehmen/environmental_commendations. regional deficits. Ground Water 1998, 36 (4), 545−546. html. (4) Falkenmark, M. The massive water scarcity now threatening (23) German Association of the Automotive Industry. Material Africa: Why isn’t it being addressed? Ambio 1989, 18 (2), 112−118. Classification in Motor Vehicle Construction; Structure and Nomenclature (5) Berger, M.; Finkbeiner, M. Water footprinting - How to address (VDA 231-106); Verband der Automobilindustrie e.V. (VDA) water use in life cycle assessment? Sustainability 2010, 2 (4), 919−944. (German Association of the Automotive Industry): Frankfurt, (6) Milài Canals, L.; Chapagain, A.; Orr, S.; Chenoweth, J.; Anton, Germany, 1997. A.; Clift, R. Assessing freshwater use impacts in LCA, part 2: Case (24) Leroy, C. Provision of LCI data in the European aluminium study of broccoli production in the UK and Spain. Int. J. Life Cycle industry - Methods and examples. Int. J. Life Cycle Assess. 2009, 14 − Assess. 2010, 15 (6), 598−607. (Suppl 1), 10 44. (7) Chapagain, A. K.; Hoekstra, A. Y. The blue, green and grey water (25) Ecoinvent Centre Website; http://www.ecoinvent.org/. footprint of rice from production and consumption perspectives. Ecol. (26) Mila i Canals, L.; Chenoweth, J.; Chapagain, A.; Orr, S.; Anton, Econ. 2011, 70 (4), 749−758. A.; Clift, R. Assessing freshwater use in LCA: Part I - inventory (8) Chapagain, A. K.; Hoekstra, A. Y.; Savenije, H. H. G.; Gautam, R. modelling and characterisation factors for the main impact pathways. The water footprint of cotton consumption: An assessment of the Int. J. Life Cycle Assess. 2008, 14 (1), 28−42. impact of worldwide consumption of cotton products on the water (27) Boulay, A.-M.; Bulle, C.; Bayart, J.-B.; Deschenes, L.; Margni, M. resources in the cotton producing countries. Ecol. Econ. 2006, 60 (1), Regional Characterization of Freshwater Use in LCA: Modelling 186−203. Direct Impacts on Human Health. Environ. Sci. Technol. 2011, 45 (20), (9) Pfister, S.; Koehler, A.; Hellweg, S. Assessing the environmental 8948−8957. impacts of freshwater consumption in LCA. Environ. Sci. Technol. (28) Hoekstra, A. Y.; Mekonnen, M. M. Global Water Scarcity: The 2009, 43 (11), 4098−4104. Monthly Blue Water Footprint Compared to Blue Water Availability for (10) Gerbens-Leenes, W.; Hoekstra, A. Y.; Van der Meer, T. H. The the World’s Major River Basins; Value of Water Research Report Series water footprint of bioenergy. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 53; UNESCO-IHE: Delft, The Netherlands, 2011. (25), 10219−10223. (29) Frischknecht, R.; Steiner, R.; Jungbluth, N. The Ecological (11) Gerbens-Leenes, W.; Hoekstra, A. Y. The water footprint of Scarcity Method - Eco-Factors 2006 - A Method for Impact Assessment in biofuel-based transport. Energy Environ. Sci. 2011, 4 (8), 2658−2668. LCA; Environmental studies no. 0906; Federal Office for the (12) ISO 14040. Environmental management - Life cycle assessment - Environment: Bern, Swizerland, 2009. Principles and framework (ISO 14040:2006); International Organ- (30) Motoshita, M.; Itsubo, N.; Inaba, A. Development of impact isation for Standardisation: Geneva, Switzerland, 2006. factors on damage to health by infectious diseases caused by domestic (13) ISO 14044. Environmental management - Life cycle assessment - water scarcity. Int. J. Life Cycle Assess. 2011, 16 (1), 65−73. Requirements and guidelines (ISO 14044:2006); International Organ- (31) Pennington, D. W.; Potting, J.; Finnveden, G.; Lindeijer, E.; isation for Standardisation: Geneva, Switzerland, 2006. Jolliet, O.; Rydberg, T.; Rebitzer, G. Life cycle assessment Part 2: (14) Volkswagen, A. G. The Life Cycle of a Car - Environmental Current impact assessment practice. Environ. Int. 2004, 30 (5), 721− Commendations Document Progress; Group Research, Environment 739. Affairs Product: Wolfsburg, Germany, 2010; http://www.volkswagen. (32) FAO Website; http://www.fao.org/nr/water/aquastat/main/ com/vwcms/master_public/virtualmaster/en2/unternehmen/ index.stm. environmental_commendations.html. (33) Alcamo, J.; Doll, P.; Heinrichs, T.; Kaspar, F.; Lehner, B.; Rosch, (15) Bayart, J. B.; Bulle, C.; Koehler, A.; Margni, M.; Pfister, S.; T.; Siebert, S. Development and testing of the WaterGAP 2 global Vince, F.; Deschenes, L. A framework for assessing off-stream model of water use and availability. Hydrol. Sci. J. 2003, 48 (3), 317− freshwater use in LCA. Int. J. Life Cycle Assess. 2010, 15 (5), 439−453. 337.

4098 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099 Environmental Science & Technology Article

(34) Google Inc. Website; http://earth.google.com/. (35) Swiss Federal Institute of Technology Zurich Website; http:// www.ifu.ethz.ch/staff/stpfiste. (36) PE International, Water footprint workshop - 27 October 2010. Stuttgart, Germany, 2010. (37) Pfister, S.; Saner, D.; Koehler, A. The environmental relevance of freshwater consumption in global power production. Int. J. Life Cycle Assess. 2011, 16 (6), 580−591. (38) Die Zeit Website; http://www.zeit.de/online/2009/25/ infografik-wasser. (39) Guinee, J. B., Ed. Life Cycle Assessment - An Operational Guide to the ISO Standards; Kulwer Academic Publishers: Dordrecht, The Netherlands, 2002. (40) Berger, M.; Finkbeiner, M. Methodological challenges in volumetric and impact oriented water footprints. J. Ind. Ecol. 2012, DOI: , DOI: 10.1111/j.1530-9290.2012.00495.x. (41) Goedkoop, M.; Spriensma, R. The Eco-indicator 99 - A damage oriented method for Life Cycle Impact Assessment, Methodology Report; Product Ecology Consultants (PRe): Amersfoort, The Netherlands, 2001. (42) Boulay, A.-M.; Bouchard, C.; Bulle, C.; Deschenes, L.; Margni, M. Categorizing water for LCA inventory. Int. J. Life Cycle Assess. 2011, 16 (7), 639−651.

4099 dx.doi.org/10.1021/es2040043 | Environ. Sci. Technol. 2012, 46, 4091−4099

RESEARCH AND ANALYSIS

Methodological Challenges in Volumetric and Impact-Oriented Water Footprints

Markus Berger and Matthias Finkbeiner

Keywords: Summary industrial ecology life cycle assessment (LCA) This work identifies shortcomings in water footprinting and discusses whether the water water consumption footprint should be a volumetric or impact-oriented index. A key challenge is the current water footprint definition of water consumption according to which evaporated water is regarded as water resources lost for the originating watershed per se. Continental evaporation recycling rates of up water withdrawal to 100% within short time and length scales show that this definition is not generally valid. Also, the inclusion of land use effects on the hydrological balance is questionable, as land transformation often leads to higher water availability due to locally increased runoff. Unless potentially negative consequences, such as flooding or waterlogging, and adverse effects on the global water cycle are considered, water credits from land transformation seem unjustified. Most impact assessment methods use ratios of annual withdrawal or consumption to renewability rates to denote local water scarcity. As these ratios are influenced by two metrics—withdrawal and availability—arid regions can be regarded as uncritical if only small fractions of the limited renewable supplies are used. Besides neglecting sensitivities to additional water uses, such indicators consider neither ground nor surface water stocks, which can buffer water shortages temporally. Authors favoring volumetric indicators claim that global freshwater appropriation is more important than local impacts, easier to determine, and less error prone than putting complex ecological interaction into mathematical models. As shown in an example, volumetric water footprints can be misleading without additional interpretation because numerically smaller footprints can cause higher impacts.

nations, water scarcity is expected to increase signiﬁcantly in Introduction many parts of the world (Alcamo and Henrichs 2002; UNEP Two-thirds of the Earth’s surface is covered with water, re- 2007). Consequently, sustainable water management is urgently sulting in a volume of 1.4 billion cubic meters (m3).1 However, needed and water footprinting is a tool promoting this shift by only 3% is freshwater, of which 69% is locked up in glaciers identifying the amounts of water used or consumed in orga- and polar icecaps (Gleick 1996). Hence only about 1% of the nizations, along product life cycles, or by people and nations global water resources is usable freshwater, which sustains life on (United Nations 2009). our planet. As this remaining fraction is unevenly distributed Having been neglected for many years due to a lack of both around the globe, water scarcity is a problem for more than awareness and appropriate methods for accounting and assessing 1.6 billion people worldwide (IWMI 2011). Taking into ac- water use and consumption, water footprinting is now a priority count population growth, demographic changes, industrial de- in current sustainability discussions. It has to be noted that there velopment, and changing consumption patterns in emerging is not only one water footprint method but different approaches

Address correspondence to: Markus Berger, Technische Universitat¨ Berlin, Department of Environmental Technology/Chair of Sustainable Engineering, Ofﬁce Z1, Strasse des 17. Juni 135, 10623 Berlin, Germany. Email: [email protected]

c 2012 by Yale University DOI: 10.1111/j.1530-9290.2012.00495.x

Volume 17, Number 1 www.wileyonlinelibrary.com/journal/jie Journal of Industrial Ecology 79 RESEARCH AND ANALYSIS to analyze the water use and consumption of organizations or methodological requirements for conducting a water footprint along product life cycles. study. Next to stand-alone methods, such as virtual water (Al- Generally the above described methods can be divided into lan 1998), water footprinting according to the Water Foot- volumetric and impact-oriented water footprints. While volu- print Network (WFN) (Hoekstra et al. 2011), the global water metric methods determine the global freshwater appropriation tool (WBCSD 2010), or the corporate water gauge (Center of products on an inventory level, impact-based water foot- for Sustainable Organizations 2011), many methods were de- prints aim at assessing the consequences resulting from wa- veloped in a life cycle assessment (LCA) context (ISO 2006a, ter consumption and require a characterization of individual 2006b). flows prior to aggregation. Depending on the method, charac- As shown in figure 1, LCA-related water footprint methods terization factors denote, for example, local freshwater scarcity range from simple water inventories to complex impact assess- (Frischknecht et al. 2009), water quality (Bayart et al. 2009), ment models. Water inventories (Bayart et al. 2010; Boulay the vulnerability of ecosystems (Pfister et al. 2009), or the sen- et al. 2011a; Ecoinvent 2011; PE International 2011; Quan- sitivity of the population to human health damages (Motoshita tis 2012) list and differentiate water input and output flows et al. 2011). according to criteria such as location, watercourse, or quality. Based on a comprehensive review (Berger and Finkbeiner On the midpoint level, methods assess the consequences of 2010), a criteria-based comparison (Kounina et al. 2011), water use or consumption in the middle of the cause–effect and an industrial application (Berger et al. 2012), this arti- chain on human health (Bayart et al. 2009; Boulay et al. 2011b; cle presents methodological challenges identified in volumetric Veolia 2011), ecosystems (Mila i Canals et al. 2008; Veolia and impact-oriented water footprint methods in order to inspire 2011), resources (Bosch¨ et al. 2007; Mila i Canals et al. 2008), or future methodological developments. Finally, as it is an ongo- generically for all areas of protection (Brent 2004; Frischknecht ing debate whether water footprints should be volumetric or et al. 2009; Hauschild and Wenzel 1998; Pfister et al. 2009). impact-oriented indicators (Hoekstra et al. 2009; Ridoutt and Endpoint methods assess potential damages resulting from wa- Pfister 2010), this article aims at discussing the pros and cons ter use or consumption at the end of the cause–effect chain of both approaches. on human health (Boulay et al. 2011b; Motoshita et al. 2008, 2011; Pfister et al. 2009), ecosystems (Maendly and Humbert 2011; Pfister et al. 2009; Van Zelm et al. 2011), and resources Methodological Challenges in Volumetric (Pfister et al. 2009). In order to reach consensus on method- Water Footprints ological questions, an International Organization for Standard- ization (ISO) standard on water footprinting is currently be- Freshwater consumption is defined as water “lost” for a catch- ing developed (ISO 2011), which defines the procedure and ment area by either evapo(trans)piration, product integration,

Figure 1 Overview of methods for inventorying and assessing water use and consumption.

80 Journal of Industrial Ecology RESEARCH AND ANALYSIS

Figure 2 Average continental evaporation recycling ratio (Van der Ent et al. 2010). Reproduced by permission of American Geophysical Union. or discharge into seawater or another river basin (Bayart et al. analyzed already (Dirmeyer and Brubaker 2007), more research 2010). However, due to a lack of data, some databases, such as is needed to determine the fate of evaporation in a spatially Ecoinvent (Ecoinvent 2011), only contain freshwater input but explicit manner. Taking into account atmospheric moisture no wastewater output flows and thus enable the determination recycling, current water inventories will change drastically as of water use but not water consumption. Other databases, such evapo(transpi)rated water is no longer considered as consumed as GaBi (PE International 2011), provide use and consumption per se. There might even be water credits resulting from the figures but only consider water used/consumed in the foreground precipitation of water vapor that was synthetically created in system and electricity production. Hence water used/consumed the combustion of fossil fuels. Hence cars might have positive in background processes, such as mining, grinding, and ore con- carbon but negative water footprints if they are run in regions centration, are not included. with high continental evaporation recycling. Despite data availability problems, there is no scientific As one of the most prominent volumetric methods, rationale why water withdrawal in one catchment and re- according to the Water Footprint Network (WFN; Hoekstra lease in another should not be seen as consumption in the et al. 2011) the water footprint comprises the consumption former and credit in the latter. Also, the assumption that of ground and surface water (blue water), the consumption of evapo(transpi)rated water is lost for watersheds is solely based soil moisture due to evapotranspiration (green water), and the on the fact that 70% of all precipitation takes place over the degree of freshwater pollution due to wastewater discharges oceans (New et al. 2001). However, when the fate of evapo- (gray water). The simple aggregation of these three types of ration is analyzed, it turns out that, on global average, 57% water consumption implies equal weighting and the possibility of terrestrial evaporation returns as precipitation over land of compensating (e.g., blue by green water consumption), (Van der Ent et al. 2010). Thus water consumption in one which is not possible in most cases. Green water is relevant for river basin may lead to precipitation in the same or another local ecosystems and agriculture and can reduce the amount catchment. of blue irrigation water. However, aggregating blue and green As can be seen in figure 2, continental evaporation recy- water to single numbers (e.g., 1,300 liters per kilogram [l/kg]3 of cling is highly site dependent, and ranges from less than 10% in southern Argentina and New Zealand to more than 90% in tropical regions and in the Himalayas. According to a study conducted by Van der Ent and Savenije (2011), average length and time scales of this atmospheric moisture recycling vary between 500 kilometers (km)2 and 3 days in the Congo basin to more than 7,000 km and more than 30 days in deserts. Hence it depends very much on the location and size of the catchment whether evapo(transpi)rated water is recycled in the same or in another basin or precipitates over oceans. In order to incorporate these findings in water inventories, the internally recycled fraction of water evaporated in, for example, the Rhine basin, Figure 3 Fractions of water evaporated in the Rhine catchment and the shares that precipitate in other catchments, such as (red shaded polygon) which are recycled internally and precipitate the Elbe or Danube, and over the sea need to be determined over other river basins (gray shaded polygons) and sea (Alcamo (figure 3). While the inverse case, that is, the evaporative et al. 2003), depending on local climatic conditions as described in sources of precipitation in a particular river basin, has been Van der Ent and colleagues (2010).

Berger and Finkbeiner, Methodological Challenges in Volumetric and Impact-Oriented Water Footprints 81 RESEARCH AND ANALYSIS wheat) (WFN 2011) hides the advantages of rain fed compared In addition to methodological shortcomings identified by to irrigated agriculture. Hoekstra and Mekonnen (2011), this ratio is influenced by two Concerning the green water consumption of, for example, metrics: use and availability. As shown by means of the method agricultural plants, it should be noted that natural vegeta- of Pfister and colleagues (2009) in figure 4, this can lead to tion also causes evapotranspiration, which can be even higher deceptive effects. Countries such as Belgium, where water is than the evapotranspiration of agricultural plants (Nunez et al. abundant, are regarded as relatively critical just because large 2010). For that reason, some authors suggest determining the shares of the renewable supply are used (not consumed). In “net green water footprint,” that is, the difference in evapotran- contrast, countries in the arid Sahel zone, such as Sudan, are spiration between agricultural and natural land (SABMiller and considered as uncritical because only minor fractions of the WWF 2009). However, the relevance of net or total green wa- small renewable water supplies are used due to a lack of industry ter consumption is questionable, as soil moisture is available for and low population density. local plants only and cannot be used by surrounding ecosystems In order to overcome this shortcoming, it may be reason- or withdrawn for human needs. The actual question that should able to define criteria such that the scarcity indicator is set to be addressed is how the green water footprint affects blue water the highest level per se. For instance, watersheds could be con- availability. According to Ridoutt and Pfister (2010), this issue sidered as extremely water scarce if they receive precipitation is closely related to land use, as land transformation can lead below a threshold (e.g., 200 mm) or are located in hyperarid to an altered surface runoff and groundwater recharge. Yet the regions in which precipitation is less than 5% of potential evap- consideration of altered blue water availability from land use otranspiration (UNEP 1997). changes as proposed by Mila i Canals and colleagues (2008) It is generally discussed whether physical water scarcity or in the new ISO standard (ISO 2011) would often result should be measured by WTA ratios, as applied by, for example, in negative blue water footprints. This can be explained by the Pfister and colleagues (2009), or by means of consumption- fact that groundwater recharge and surface runoff often increase to-availability (CTA) indicators, used in the recent method when natural vegetation is transformed into agricultural land, developments of Boulay and colleagues (2011b) and Hoek- which has led to globally increased river discharges of 7% (Rost stra and colleagues (2011). On the one hand, WTA tends et al. 2008). However, this implies that an organization could to overestimate physical water scarcity, as it also comprises compensate its blue water footprint by land use changes such borrowing and degradative water uses, such as cooling wa- as deforestation. It ignores the fact that natural vegetation is ter, that are returned to the watershed from which they were an important foundation for the global water cycle and land withdrawn. On the other hand, there can be water scarcity use changes affect this cycle. According to Rost and colleagues resulting from competing uses and from freshwater pollution, (2008), the increase in runoff resulting from land transforma- which are implicitly covered by WTA. However, considering tion equals the decrease in evapotranspiration (net green water freshwater degradation by means of scarcity indexes in addi- footprint), which will lead to decreased precipitation in other tion to water quality indices and emission-based impact cat- places. Hence blue water credits resulting from land use changes egories in LCA can lead to an overestimation of impacts, as should only be given if the consequent decrease in precipitation described later. Even though competing uses of freshwater may in other watersheds is taken into account as well. Specific losses lead to scarcity in densely populated areas in some cases, CTA of precipitation in watersheds resulting from land transforma- generally seems to express physical water scarcity in a more tion and decreased evapotranspiration in another catchment meaningful way. Moreover, it ensures consistency between the could be determined by means of spatially explicit evaporation inventory and the impact assessment levels, as water con- recycling ratios, as mentioned above (figure 3). Also, potential sumption is assessed by a characterization factor also based on on-site negative impacts resulting from increased runoff, such consumption. as flooding, increased soil salinity, or waterlogging, should be Despite this discussion, neither CTA- nor WTA-based ap- considered when crediting local blue water increases. proaches consider the sensitivity of a watershed to water scarcity resulting from additional water use/consumption. An assumed annual water use of 1 cubic meter per year (m3/year) and a 3 Methodological Challenges In renewability rate of 10 m /year would lead to the same result = 3 Impact-Oriented Water Footprints (WTA 0.1) as an annual water use of 1,000 m /year and a renewable supply of 10,000 m3/year. However, adding 1 m3 Local freshwater scarcity is the key indicator to assess vari- of water use would double the scarcity ratio in the first case ous consequences of water consumption. Therefore most meth- (WTA = 0.2) but leave it rather unchanged in the second ods use a ratio of annual water use to renewability rate to (WTA = 0.1001). Hence catchments with small renewable express local freshwater scarcity, known in different terms as water supplies are more sensitive to additional water uses than water utilization level (Falkenmark 1989), use-to-resource ratio basins with large renewability rates. It should be noted that an (Raskin et al. 1996), or withdrawal-to-availability ratio (WTA) additional water use of 1 m3/year in the first example is contra- (Alcamo and Henrichs 2002): dictory to the assumption of marginal changes not influencing the overall environmental situation, which is common practice annual water use WTA = . (1) in life cycle impact assessment (LCIA; Guinee et al. 2002). renewabilityrate

82 Journal of Industrial Ecology RESEARCH AND ANALYSIS

Figure 4 Google Earth layer (Google Inc. 2011) showing water stress in global watersheds according to Pﬁster and colleagues (2009) ranging from very low (water stress index [WSI] < 0.1, dark blue shading) to extreme (WSI = 1, dark red shading).

Another challenge for water scarcity indicators is the proper emissions are also considered in the water footprint method of determination of water availability. In their current states, Boulay and colleagues (2011b), which estimates health dam- WTA or similar indicators only consider renewable ground- ages of malnutrition as a consequence of quality degradation water recharge and surface runoff but neglect ground and sur- affecting agriculture and fisheries. However, there can be either face water stocks. From a resources perspective this is justified the negative effect of cadmium uptake caused by eating fish or because a water use that exceeds the renewability rate leads the negative effect of malnutrition by not having enough fish to a depletion of water reservoirs. This can cause drastic conse- to eat. quences, such as a 50 meter (m) decline in groundwater tables in It should be noted that the consideration of water quality the High Plains Aquifer (USA) or North China Plains Aquifer degradation and emission-oriented impact categories does not (UNEP 2012). However, when consequences on human health necessarily lead to an overestimation. Emission-based impact and ecosystems are assessed, it is important to consider ground categories neglect the quality of the water withdrawn and the and surface water stocks, as they can compensate for an overuse fact that water pollution can increase water scarcity in arid re- of renewable supplies. Hence impacts are less severe if reservoirs gions, or they take into account different, yet complementary are available that can buffer water shortages for a certain time. impact pathways. For instance, the human toxicity potential While data on groundwater recharge and surface water runoff considers the uptake route via potable water and the charac- are available at high spatial resolution (e.g., Alcamo et al. 2003; terization model of Boulay and colleagues (2011b) considers Doll¨ and Fiedler 2008), spatially explicit figures of total ground malnutrition occurring if water quality degradation precludes and surface water stocks are hard to obtain, which is a clear irrigation in agriculture. As a lack of drinking water is not need for further research. considered by Boulay and colleagues (2011b), and since the A further point of concern is the consideration of water exposure to pollutants via agricultural irrigation is not included quality degradation by means of the gray water concept (e.g., in the human toxicity potential, these impact pathways do not Hoekstra et al. 2011), the use of withdrawal-based scarcity ratios exclude each other. If impact-oriented water footprint methods (e.g., Pfister et al. 2009), or consideration of quality indicators such as those of Veolia (2011) are applied, which considers qual- (e.g. Boulay et al. 2011b). Especially if the water footprint anal- ity and generically assesses consequences on human health and ysis is accomplished in an LCA context, the pollution of water ecosystems, it is hard to say whether impacts are overestimated is often covered by other impact categories such as eutrophica- or not, as no distinct impact pathways are described. Hence po- tion, acidification, and eco- or human toxicity (Guinee et al. tential overestimation should be discussed in the interpretation 2002). If quality degradation is considered in traditional impact of water footprint studies. categories and in water footprint indicators, an overestimation Further need for improvement was detected in the question can occur if mutually exclusive impact pathways are considered. of environmental credits. If, for instance, water is withdrawn For instance, an emission of cadmium to freshwater is consid- from a fossil aquifer and after its use is discharged into sur- ered in the impact category human toxicity by modeling the face water courses, this water is made available to other users. exposure route of cadmium to humans via fish. Yet cadmium Thus depletion of fossil water resources might be beneficial for

Berger and Finkbeiner, Methodological Challenges in Volumetric and Impact-Oriented Water Footprints 83 RESEARCH AND ANALYSIS ecosystems and human needs. Consensus has to be found on how to deal with this trade-off between two safeguard subjects. Another challenge was identified regarding the extent of inventory data required by different water footprint methods. In order to apply impact-oriented methods, the pure volume of water used or consumed is not enough. As shown in table 1, most methods require geographical information denoting where the water is used. Additionally, some methods can only be applied if the type of watercourse used, the water quality, or even the time of withdrawal is known. However, such information is rarely available and costly to collect, especially if complex background systems are involved (Berger Figure 5 Theoretical example of water consumption and et al. 2012). As the scientifically most advanced methods show greenhouse gas emissions. the highest inventory data requirements, there is a trade-off between precision and applicability in impact-oriented water footprints. This shows the need for both more detailed in- of midpoint characterization models, these authors argue that ventory datasets and practicable water footprint methods not “the carbon footprint is a measure of the amount of greenhouse requiring overly detailed inventory information. Therefore ef- gases (GHGs) emitted to the environment from human activi- forts to establish regionalized inventory datasets as realized in ties and does not describe environmental impacts” (Mekonnen the Quantis Water Database (Quantis 2012) are of great im- and Hoekstra 2012). portance. Also, the development of water categories (Boulay Despite their strong criticism, Hoekstra and colleagues et al. 2011a) denoting the type of watercourse and quality are a (2011) also provide a method to determine “water footprint promising development. They can be incorporated as elemen- impact indices,” which equals impact assessment. It should also tary flows in inventory datasets and fulfill the data requirements be noted that impact-oriented water footprint methods provide of advanced impact assessment methods (Boulay et al. 2011b). characterization factors, which are derived from characterization models describing the environmental mechanisms, rather than weighting factors, which include value judgment. Actu- Volumetric or Impact-Oriented Water ally the grey water concept used in volumetric water footprint Footprints? methods, developed by those authors, depends on legal thresh- With respect to the development of water accounting tools, olds and thus represents a subjective distance-to-target weight- methods such as virtual water (Allan 1998) or the water foot- ing rather than a scientific characterization. In addition, there print according to WFN (Hoekstra and Hung 2002) started by are many pollutants for which legal limits are lacking. As a aggregating the volumes of water consumed in the various life consequence, they are simply neglected in such an approach. cycle phases at different places worldwide. By publishing sur- So, should the water footprint really be a volumetric indi- prisingly high figures of, for example, 140 l per cup of coffee cator based on the arguments that environmental mechanisms or 2,400 l per hamburger (WFN 2011), the authors did pio- are hard to model and that required inventory information is neering work in making people aware of the large amount of difficult to collect? This point of discussion is illustrated in the water consumed in the production of daily goods. However, as following theoretical example in which volumetric and impact- discussed above, these numbers should be handled with care, oriented water footprints are compared. In order to highlight as they contain an aggregated consumption of green, blue, and the relevance, we further assessed GHG emissions on an inven- gray water, for which a scientific rational is lacking. As pure vol- tory and impact assessment level. umetric figures do not allow for assessing the consequences of When the two products shown in figure 5 are assessed on water consumption, authors such as Ridoutt and Pfister (2010) the inventory level by means of volumetric water footprints, conclude that volumetric water footprints are incomplete and product A scores better than product B. However, taking into can be misleading. account local water scarcity and degree of development, water In contrast, authors favoring the volumetric approach state consumption of 1 m3 in Iran may cause greater impacts than that global freshwater appropriation is more relevant than local 2m3 in Germany. impacts (Hoekstra et al. 2011). Moreover, Hoekstra and col- In fact, a volumetric water footprint would equal a mass- leagues regard impact-oriented water footprints without phys- based carbon footprint, according to which product A would ical interpretation as “completely meaningless.” In their opin- also be preferable, as only 1 kg instead of 2 kg GHG are emitted. ion, local environmental conditions influencing impacts are Obviously no one would argue for such an approach, as the in- poorly described by current impact assessment schemes, which dividual contribution of methane and carbon dioxide (CO2)to leads to “questionable weighting choices” (Hoekstra et al. 2009) global warming needs to be taken into account. This character- in reply to Pfister and Hellweg (2009). Furthermore, Mekonnen ization would result in an opposite ranking, as product A causes and Hoekstra (2012) consider volumetric water footprints as 25 kilograms carbon dioxide equivalents (kg CO2-eq) and prod- consistent with carbon footprints. Misinterpreting the meaning uctBcauses2kgCO2-eq (IPCC 2007). Consequently, in the

84 Journal of Industrial Ecology RESEARCH AND ANALYSIS x 2 2 x x 1 1 Net water consumption Location Water course Quality Time Inventory data requirements discharge Waste water (use) withdrawal Freshwater ¨ osch et al. (2007) x x Water footprint (WFN) x x x Global water toolCorporate water gaugeBoulay et al. (2011a) x x x x x x x x x x x x x PE International (2011)Quantis (2012) xMila i Canals et al.Bayart (2008) et al. (2009)Frischknecht et al. (2009)Pﬁster et al. x (2009)Boulay et al. (2011b)Veolia (2011) x xPﬁster et al. x (2009)Boulay et al. x (2011b)Maendly and Humbert (2011)Motoshita x et al. (2011)Van Zelm et al. (2011) x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x Endpoint Motoshita et al. (2008) x x Pollutant load needs to be known. 2 impactassessment B Brent (2004) x x x Life cycle Midpoint Hauschild and Wenzel (1998) x x Inventory data requirements of different water footprint methods Blue, green, grey water. 1 assessment inventory Ecoinvent (2011) x x Ta b l e 1 Method Stand-alone Virtual waterNotes: x x Life cycle Life cycle Bayart et al. (2010) x x x x x

Berger and Finkbeiner, Methodological Challenges in Volumetric and Impact-Oriented Water Footprints 85 RESEARCH AND ANALYSIS

Figure 6 Relative results of volumetric and impact-oriented water footprint methods for water consumption of 1 m3 Iran and 2 m3 in Germany normalized to the results obtained for 1 m3 of water consumed in Iran.

same way methane and CO2 emissions cannot be compared on with a human development index greater than 0.88, malnu- a kilogram level, water consumption in Iran and Germany can- trition is not an issue. These different results in two methods not be compared on a cubic meter level (also shown by Pfister assessing malnutrition can be explained by the fact that the and Hellweg [2009]). method of Motoshita and colleagues (2008) additionally con- In figure 6 it can be seen that impact-oriented water footprint siders indirect effects. Accordingly, agricultural water scarcity in methods lead to different results than volumetric inventories. Germany leads to the importation of food from other countries, Since Motoshita et al. (2011) determined damage factors only causing impacts outside the country. Without the internaliza- for domestic water consumption, these factors were multiplied tion of such indirect effects, the method would also consider by a country-specific ratio of domestic to total water use (FAO damages resulting from water consumption of 2 m3 in Germany 2011) to allow for an assessment of general water consumption. (7.6 E-14 disability-adjusted life year [DALY]) as negligibly low Therefore, the method is termed Motoshita et al. (2011)∗ in compared to 1 m3 in Iran (1.7 E-9 DALY). The comprehen- the following. In order to apply the impact assessment model of sive characterization model of Boulay and colleagues (2011b), Boulay et al. (2011b), the consumption of good quality surface which assesses damages of human health due to malnutrition water was assumed in both cases. and infectious diseases, leads to a result of zero health impacts As can be seen in figure 6, the two methods relying only on in Germany considering the high adaptation capacity to water physical water scarcity (ecological scarcity method and fresh- stress. water deprivation) show similar results since both are based on While wealth reduces human health impacts in the method WTA. In contrast, the water stress index developed by Boulay of Pfister and colleagues (2009) and Boulay and colleagues and colleagues (2011b) indicates a relatively higher relevance of (2011b), it cannot avoid ecosystem damages that are relevant water consumption in Germany because it was determined for a in Germany as well. Even though water use is relatively high certain water quality and relies on a ratio of water consumption in an industrialized and densely populated country such as Ger- to the statistical low flow (Q90).4 many (WTA = 0.21; FAO 2011), no damage to resources is Also, damages to human health are evaluated differently. detected, as withdrawal does not exceed the renewability rate In the methods of Motoshita and colleagues, impacts resulting (WTA < 1), which avoids the depletion of water stocks. In from water consumption of 2 m3 in Germany are considered contrast to the volumetric water footprint, all characterization more relevant (34%) in the malnutrition category (Motoshita models applied show that it is more relevant to consume 1 m3 et al. 2008) than in the infectious diseases model (21%) (Mo- of water in Iran than 2 m3 in Germany. Hence, without addi- toshita et al. 2011). In contrast, Pfister and colleagues (2009), tional qualitative interpretation as proposed by Hoekstra and who also determine damages from undernourishment, consider colleagues (2011), volumetric water footprints can be mislead- health impacts in Germany as nil, arguing that in countries ing, as numerically smaller footprints can cause more severe

86 Journal of Industrial Ecology RESEARCH AND ANALYSIS impacts. Therefore water footprints should be impact-based in- adverse effects of increased runoff and without taking into ac- dicators in the same way as carbon footprints. count the relevance of natural vegetation on the global water A main difference between carbon and water footprinting cycle. is the agreement on characterization models, based on which On the impact assessment level, most methods rely on ratios characterization factors for GHGs and regional water consump- of annual water withdrawal/consumption to renewability rate tion are determined. While the global warming potential is cur- (WTA/CTA) in considering local freshwater scarcity. As such rently established as an internationally agreed characterization ratios consist of two metrics, arid regions can be regarded as model for the carbon footprint (despite inherent weaknesses of uncritical if only small fractions of the limited renewable supply the model), there are several characterization models available are used/consumed. Moreover, WTA and CTA do not take for water consumption. Even though different models can lead into account ground and surface water stocks, which can buffer to different results, this cannot be regarded as an inconsistency. temporal overexploitation and do not express the sensitivity of It is clear that different models, which describe different impact watersheds to additional withdrawal/consumption. pathways of water consumption on human health, ecosystems, The consequences of water consumption depend on local and resources, lead to diverse implications as they are influ- scarcity, the type of watercourse used, water quality, the time enced by various parameters. This also reflects the complexity of withdrawal, as well as the sensitivity of ecosystems and pop- of reality as, for example, water consumption in water-scarce ulation. For this reason, volumetric water footprints can be but developed countries may lead to the depletion of water misleading without substantial additional interpretation, as nu- resources while impacts on human health can be nil. Even merically smaller footprints can cause larger impacts. Thus, in within one safeguard subject, such as human health, results order to support decision making, water footprints should be can vary, as the sensitivity of a population to malnutrition or impact-oriented indicators. infectious diseases might differ. Hence, as long as no comprehensive impact-oriented water footprint method that addresses Notes all impact pathways in a consistent manner is available, several characterization methods should be applied to analyze various 1. One cubic meter (m3,SI)= 103 liters (l) ≈ 264.2 gallons (gal). implications. Uncertainties in the environmental fate and in 2. One kilometer (km, SI) ≈ 0.621 miles (mi). = 3 ≈ the characterization models are a challenge and have to be ad- 3. One liter (l) 0.001 cubic meters (m ,SI) 0.264 gallons (gal). ≈ dressed in carbon and water footprinting. However, they do not One kilogram (kg, SI) 2.204 pounds (lb). 4. Q90 refers to the flow that is reached or exceeded with a 90% justify performing inventory-related footprints, which can be probablility. misleading due to the lack of environmental interpretation.

References Conclusions Alcamo, J., P. Doll, T. Heinrichs, F. Kaspar, B. Lehner, T. Rosch, and In order to promote future method developments, this ar- S. Siebert. 2003. Development and testing of the WaterGAP 2 ticle aimed at identifying methodological challenges in water global model of water use and availability. Hydrological Sciences footprinting and discussing whether the water footprint should Journal 48(3): 317–337. be a volumetric or impact-oriented indicator. Shortcomings Alcamo, J. and T. Henrichs. 2002. Critical regions: A model-based have been detected in the current definition of freshwater con- estimation of world water resources sensitive to global changes. sumption. The assumption that water is generally “lost” from a Aquatic Sciences 64: 352–362. watershed due to evapo(transpi)ration is refuted by recent stud- Allan, J. A. 1998. Virtual water: A strategic resource, global solutions to regional deficits. Ground Water 36: 545–546. ies showing relevant continental evaporation recycling rates Bayart, J. B., C. Bulle, A. Koehler, M. Margni, S. Pfister, F. Vince, and L. within short time and length scales (Van der Ent and Savenije Deschenes. 2010. A framework for assessing off-stream freshwater 2011). Moreover, the aggregation of green, blue, and gray wa- use in LCA. International Journal of Life Cycle Assessment 15(5): ter in some volumetric approaches (Allan 1998; Hoekstra and 439–453. Hung 2002) is questionable, as it implies equal relevance and Bayart, J. B., C. Bulle, M. Margni, F. Vince, L. Deschenes, and E. mutual substitutability, for which a scientific rational is lack- Aoustin. 2009. Operational characterisation method and factors ing. In particular, the importance of green water consumption for a new midpoint impact category: freshwater deprivation for is doubtful, as soil moisture is available for local plants only human uses. Paper presented at SETAC Europe, 19th Annual and cannot be used by surrounding ecosystems or for human Meeting, 31 May–4 June 2009, Gothenborg, Sweden. withdrawal. When effects of land use changes on blue water Berger, M. and M. Finkbeiner. 2010. Water footprinting—How to availability are considered, care has to be taken, as land trans- address water use in life cycle assessment? Sustainability 2(4): 919– 944. formation from natural to arable land may lead to increased Berger, M., J. Warsen, S. Krinke, V. Bach, and M. Finkbeiner. 2012. blue water availability due to altered surface runoff and ground- Water footprint of European cars: Potential impacts of water con- water recharge patterns. However, water credits from land use sumption along automobile life cycles. Environmental Science and changes seem unjustified without considering the potentially Technology. doi: 10.1021/es2040043

Berger and Finkbeiner, Methodological Challenges in Volumetric and Impact-Oriented Water Footprints 87 RESEARCH AND ANALYSIS

Bosch,¨ M. E., S. Hellweg, M. A. J. Huijbregts, and R. Frischknecht. entific and Cultural Organization Institute for Water Education 2007. Applying cumulative exergy demand (CExD) indicators to (UNESCO-IHE). the Ecoinvent database. International Journal of Life Cycle Assess- Hoekstra, A. Y. and M. M. Mekonnen. 2011. Global water scarcity: The ment 12: 181–190. monthly blue water footprint compared to blue water availabil- Boulay, A.-M., C. Bouchard, C. Bulle, L. Deschenes, and M. Margni. ity for the world’s major river basins. In Value of water. Research 2011a. Categorizing water for LCA inventory. International Journal Report Series 53. Delft, the Netherlands: United Nations Educa- of Life Cycle Assessment 16(7): 639–651. tional, Scientific and Cultural Organization Institute for Water Boulay, A.-M., C. Bulle, J.-B. Bayart, L. Deschenes, and M. Margni. Education (UNESCO-IHE). 2011b. Regional characterization of freshwater use in LCA: Mod- IPCC (Intergovernmental Panel on Climate Change). 2007. Climate elling direct impacts on human health. Environmental Science and change 2007 — The fourth assessment report. Geneva, Switzerland: Technology 45(20): 8948–8957. IPCC. Brent, A. C. 2004. A life cycle impact assessment procedure with ISO (International Organization for Standardization). 2006a. ISO resource groups as area of protection. International Journal of Life 14040. Environmental management — Life cycle assessment — Prin- Cycle Assessment 9: 172–179. ciples and framework. Geneva, Switzerland: ISO. Center for Sustainable Organizations. 2011. Corporate water gauge: A ISO (International Organization for Standardization). 2006b. ISO context-based solution for measuring the sustainability of organi- 14044. Environmental management — Life cycle assessment — Re- zational water use. www.sustainableorganizations.org/Corporate quirements and guidelines. Geneva, Switzerland: ISO. -Water-Gauge.pdf. Accessed 24 October 2011. ISO (International Organization for Standardization). 2011. ISO Dirmeyer, P. A. and K. L. Brubaker. 2007. Characterization of the global 14046 working draft 2. Water footprint — Principles, requirements hydrologic cycle from a back-trajectory analysis of atmospheric and guidance. Geneva, Switzerland: ISO. water vapor. Journal of Hydrometeorology 8: 20–37. IWMI (International Water Management Institute). 2011. Ecosystems Doll,¨ P. and K. Fiedler. 2008. Global-scale modeling of groundwater for water and food security. Nairobi, Kenya: United Nations Envi- recharge. Hydrology and Earth System Sciences 12: 863–885. ronment Programme. Ecoinvent. 2011. Welcome to the Ecoinvent Centre portal. Kounina, A., M. Margni, J.-B. Bayart, A.-M. Boulay, M. Berger, www.ecoinvent.org. Accessed 12 March 2011. C. Bulle, R. Frischknecht, A. Koehler, L. Mila-i-Canals, M. Falkenmark, M. 1989. The massive water scarcity now threatening Motoshita, M. Nuńez,˜ G. Peters, S. Pfister, B. Ridoutt, R. v. Zelm, Africa: Why isn’t it being addressed? Ambio 18(2): 112–118. F. Verones, and S. Humbert. 2011. Review of methods addressing FAO (United Nations Food and Agriculture Organization). 2011. freshwater availability in life cycle inventory and impact assess- Food and Agricultural Organization of the United Nations– ment. International Journal of Life Cycle Assessment. Submitted for Aquastat. www.fao.org/nr/water/aquastat/main/index.stm. Ac- publication. cessed 3 March 2011. Maendly, R. and S. Humbert. 2011. Empirical characterization model Frischknecht, R., R. Steiner, and N. Jungbluth. 2009. The ecological and factors assessing aquatic biodiversity damages of hydropower scarcity method — Eco-factors 2006 — A method for impact assess- water use. International Journal of Life Cycle Assessment. Submitted ment in LCA. Bern, Switzerland: Federal Office for the Environ- for publication. ment. Mekonnen, M. M. and A. Y. Hoekstra. 2012. The blue water foot- Gleick, P. H. 1996. Water resources. In Encyclopedia of climate and print of electricity from hydropower. Hydrology and Earth System weather,editedbyS.H.Schneider.NewYork,NY,USA:Oxford Sciences 16: 179–187. University Press. Mila i Canals, L., J. Chenoweth, A. Chapagain, S. Orr, A. Anton, Google Inc. 2011. Google Earth. http://earth.google.com. Accessed 7 and R. Clift. 2008. Assessing freshwater use in LCA: Part I — January 2011. Inventory modelling and characterisation factors for the main Guinee, J. B., M. Gorree, R. Heijungs, G. Huppes, R. Kleijn, A. impact pathways. International Journal of Life Cycle Assessment 14: deKoning,L.vanOers,A.W.Sleeswijk,S.Suh,H.A. 28–42. Udo de Haes, H. de Bruijn, R. van Duin, and M. A. J. Hui- Motoshita, M., N. Itsubo, and A. Inaba. 2008. Development of impact jbregts. 2002. Life cycle assessment — An operational guide to the assessment method on health damages of undernourishment re- ISO standards. Dordrecht, the Netherlands: Kluwer Academic lated to agricultural water scarcity. Paper presented at the Eighth Publishers. International Conference on EcoBalance, 10–12 December 2008, Hauschild, M. and H. Wenzel. 1998. Environmental assessment of prod- Tokyo, Japan. ucts, vol. 2. New York, NY, USA: Chapman & Hall, Thomson Motoshita, M., N. Itsubo, and A. Inaba. 2011. Development of impact Science. factors on damage to health by infectious diseases caused by do- Hoekstra, A. Y., A. K. Chapagain, M. M. Aldaya, and M. M. mestic water scarcity. International Journal of Life Cycle Assessment Mekonnen. 2011. The water footprint assessment manual — Set- 16(1): 65–73. ting the global standard. Washington, DC, USA: Earthscan. New, M., M. Todd, M. Hulme, and P. Jones. 2001. Precipitation mea- Hoekstra, A. Y., W. Gerbens-Leenes, and T. H. van der Meer. 2009. surements and trends in the twentieth century. International Jour- Reply to Pfister and Hellweg: Water footprint accounting, im- nal of Climatology 21: 1899–1922. pact assessment, and life-cycle assessment. Proceedings of the Na- Nunez, M., S. Pfister, A. Anton, P. Munoz, S. Hellweg, A. Koehler, tional Academy of Sciences of the United States of America 106: and J. Rieradevall. 2010. Using geographic information systems 114. and spatial data to assess water consumption of bioenergy. Paper Hoekstra, A. Y. and P. Q. Hung. 2002. Virtual water trade: A quan- presented at SETAC Europe, 20th Annual Meeting, 23–27 May tification of virtual water flows between nations in relation to 2010, Seville, Spain. international crop trade. In Value of water. Research Report Se- PE International. 2011. GaBi Software. www.gabi-software.com. Ac- ries 11. Delft, the Netherlands: United Nations Educational, Sci- cessed 8 March 2011.

88 Journal of Industrial Ecology RESEARCH AND ANALYSIS

Pfister, S. and S. Hellweg. 2009. The water “shoesize” vs. footprint of United Nations. 2009. The United Nations world water development re- bioenergy. Proceedings of the National Academy of Sciences of the port 3: Water in a changing world. World Water Assessment Pro- United States of America 106: E93–E94. gramme. Paris, France: United Nations Educational, Scientific Pfister, S., A. Koehler, and S. Hellweg. 2009. Assessing the environ- and Cultural Organization. mental impacts of freshwater consumption in LCA. Environmental Van der Ent, R. J. and H. H. G. Savenije. 2011. Length and time Science and Technology 43: 4098–4104. scales of atmospheric moisture recycling. Atmospheric Chemistry Quantis. 2012. Quantis water database. www.quantis-intl.com/ water- and Physics 11: 1853–1863. database.php. Accessed 10 February 2012. Van der Ent, R. J., H. H. G. Savenije, S. Bettina, and S. C. Steele- Raskin, P. D., E. Hansen, and R. M. Margolis. 1996. Water and sus- Dunne. 2010. Origin and fate of atmospheric moisture over conti- tainability: Global patterns and long-range problems. Natural Re- nents. Water Resources Research 46. doi:10.1029/2010WR009127 sources Forum 20(1): 1–5. Van Zelm, R., A. M. Schipper, M. Rombouts, J. Snepvangers, and M. Ridoutt, B. G. and S. Pfister. 2010. A revised approach to water foot- A. J. Huijbregts. 2011. Implementing groundwater extraction in printing to make transparent the impacts of consumption and life cycle impact assessment: Characterization factors based on production on global freshwater scarcity. Global Environmental plant species richness for the Netherlands. Environmental Science Change 1: 113–120. and Technology 45(2): 629–635. Rost, S., D. Gerten, and U. Heyder. 2008. Human alterations of the Veolia. 2011. The water impact index and the first carbon-water terrestrial water cycle through land management. Advances in analysis of a major metropolitan water cycle. www. veoliawa- Geosciences 18: 43–50. terna.com/north-america-water/ressources/documents/1/10975, SABMiller and WWF (World Wildlife Fund). 2009. Water footprinting: Water_Impact_Index-White_Paper.pdf. Accessed 5 September Identifying and addressing water risks in the value chain. Working and 2011. Godalming, UK: SABMiller Plc and WWF. WBCSD (World Business Council for Sustainable Develop- UNEP (United Nations Environment Programme). 1997. World atlas ment). 2010. Global water tool 2011. www.wbcsd.org/ of desertification, 2nd ed., edited by N. Middleton and D. Thomas. web/watertool.htm. Accessed 30 June 2010. London, UK: Arnold. WFN (Water Footprint Network). 2011. Water footprint. UNEP (United Nations Environment Programme). 2007. Global en- www.waterfootprint.org. Accessed 22 March 2011. vironmental outlook 4 — Environment for development. Nairobi, Kenya: UNEP. UNEP (United Nations Environment Programme). 2012. A About the Authors glass half empty: Regions at risk due to groundwater depletion. UNEP Global Environmental Alert Service. http://na.unep.net/ Markus Berger is a senior researcher and Matthias api/geas/articles/getArticleHtmlWithArticleIDScript.php?article Finkbeiner is full professor at the Chair of Sustainable Engi- _id=76. Accessed 10 February 2012. neering at Technische Universitaet, Berlin, Germany.

Berger and Finkbeiner, Methodological Challenges in Volumetric and Impact-Oriented Water Footprints 89

Article

pubs.acs.org/est

Water Accounting and Vulnerability Evaluation (WAVE): Considering Atmospheric Evaporation Recycling and the Risk of Freshwater Depletion in Water Footprinting † ‡ § † † Markus Berger,*, Ruud van der Ent, Stephanie Eisner, Vanessa Bach, and Matthias Finkbeiner † Technische Universitaẗ Berlin, Chair of Sustainable Engineering, Oﬃce Z1, Strasse des 17. Juni 135, 10623 Berlin, Germany ‡ Delft University of Technology, Department of Water Management, Postbus 5, 2600 AA Delft, The Netherlands § University of Kassel, Center for Environmental Systems Research, Wilhelmshöher Allee 47, 34109 Kassel, Germany

*S Supporting Information

ABSTRACT: Aiming to enhance the analysis of water consumption and resulting consequences along the supply chain of products, the water accounting and vulnerability evaluation (WAVE) model is introduced. On the accounting level, atmospheric evaporation recycling within drainage basins is considered for the ﬁrst time, which can reduce water consumption volumes by up to 32%. Rather than predicting impacts, WAVE analyzes the vulnerability of basins to freshwater depletion. Based on local blue water scarcity, the water depletion index (WDI) denotes the risk that water consumption can lead to depletion of freshwater resources. Water scarcity is determined by relating annual water consumption to availability in more than 11 000 basins. Additionally, WDI accounts for the presence of lakes and aquifers which have been neglected in water scarcity assessments so far. By setting WDI to the highest value in (semi)arid basins, absolute freshwater shortage is taken into account in addition to relative scarcity. This avoids mathematical artifacts of previous indicators which turn zero in deserts if consumption is zero. As illustrated in a case study of biofuels, WAVE can help to interpret volumetric water footprint ﬁgures and, thus, promotes a sustainable use of global freshwater resources.

■ INTRODUCTION After a comprehensive review13 and test of volumetric and 14 During the past century water use was growing twice as fast as impact oriented WF methods in an industrial case study, methodological challenges have been identified in both the world’s population, which has resulted in 1.2 billion people approaches.15 According to the current definition, water living in water scarce regions today.1 The analysis of water use consumption denotes the fraction of total water use (with- and associated impacts throughout the supply chains of 2 drawal) not returning to the originating drainage basin due to products, that is, the water footprint (WF), can serve as a product integration, discharge into seawater and other basins, relevant tool to mitigate water stress. 16 fi “ ” and evapo(transpi)ration. However, this de nition neglects Volumetric approaches, such as virtual water , consider the the fact that significant shares of evaporated water can return consumption of ground and surface water (blue water), the via precipitation within short time and length scales17 and, thus, evapo(transpi)ration of rainwater (green water), and the 3 should be regarded as water use but not consumption. To pollution of freshwater (gray water). By revealing surprisingly analyze the severity of a volume of water consumed, many ff 4 high volumes, like 140 L per cup of co ee or 2700 L per impact assessment methods3,18,19 use scarcity-based impact 5 cotton T-shirt, consumers have been made aware of the factors which rely on the withdrawal-to-availability (WTA) amounts of water consumed or polluted during the production ratio.20 However, WTA contains several shortcomings. First, of daily goods. Even the WF of nations and global virtual water withdrawals can include large shares of cooling water that are imports and exports have been analyzed based on WF estimates returned to the basin immediately and, therefore, lead to an of products, consumption patterns, and trade statistics.6,7 overestimation of water scarcity in industrial regions. Second, Despite the relevance of global freshwater appropriation ground and surface water stocks are neglected as “availability” figures, there is an ongoing debate whether the WF should be a includes runoff only. Yet, aquifers and lakes can buffer temporal volumetric indicator or an impact based measure like the water scarcity and, thus, are important when impacts of water carbon footprint.8 While some authors highlight the necessity consumption are to be assessed. Third, the WTA ratio turns of impact factors, as 1 m3 of water consumption in Canada does 3 9,10 not compare to 1 m of water consumed in Mexico, other Received: September 13, 2013 scientists argue that a product’s volumetric footprint is more Revised: February 21, 2014 important since freshwater is a global resource virtually traded Accepted: March 24, 2014 via products.11,12

Figure 1. Water inventory flows along the life cycle of a product considered in WAVE. zero when the numerator (withdrawal) is zero15 as it assesses are determined by multiplying the volumes of evapo(transpi)- relative scarcity only, but neglects absolute shortage of water. ration (Ei) and synthetically created vapor (Vi) with the basin Consequently, very dry regions, like the Sahel zone, can be internal evaporation recycling ratio (BIER) and the runoff regarded uncritical if withdrawal is close to zero due to the fraction (α). absence of population or industry. Tackling the above-mentioned challenges, the water ERBIERii=·E n ·α (3a) accounting and vulnerability evaluation (WAVE) model is introduced. It should be noted that there is no connection to a VRii=·V BIER n ·α (3b) different model describing the connection between “substances, 17 water and agrochemicals in the soil, crop and vadose According to van der Ent and Savenije, BIER is estimated environment”21 which is also termed WAVE. On the inventory based on the length of the basin in the direction of the main level, a new water accounting approach considers atmospheric moisture flux (x) and the average local length scale of the evaporation recycling effects and, therefore, allows for a evaporation recycling process (λ). determination of more realistic water consumption figures. In −−·x λλexp −x + order to translate volumes into potential impacts, the ()λ vulnerability of a basin to freshwater resource depletion is BIER = −x (4) evaluated by means of a new blue water scarcity indicator. Hence, WAVE will help to interpret volumetric virtual water The length scale of the evaporation recycling process (λ) has studies and can be used as an inventory and characterization been calculated based on an atmospheric water accounting 22 model in life cycle assessment (LCA) and water foot- model considering evaporation, precipitation, winds, and 23 printing. humidity.17,24 For simplification each basin is assumed to be a square with x representing the side length determined via the ■ WATER ACCOUNTING MODEL square root of the basin’s surface area. A new inventory method for the accounting of water use is BIER has been determined for more than 11 000 basins on a introduced. In addition to freshwater withdrawals and waste- global level (Figure 2) and varies from 0% in the Sahel zone to water discharges considered in existing inventory schemes,13 38% in the Congo basin. Thus, significant shares of the water the new water accounting model explicitly considers the share consumed in a product system due to evapo(transpi)ration can of withdrawal which is consumed due to evapo(transpi)ration. be returned to the originating drainage basin via precipitation. Moreover, vapor created synthetically in chemical reactions, e.g. In the same way, water vapor created in chemical reactions can by burning fossil fuels, is regarded in an explicit way (Figure 1). be returned to the basin of origin to a noticeable extent. Since In order to consider the effects of atmospheric evaporation the share of evaporation recycling increases with distance, large 17 ff recycling, the e ective water consumption (WCeff), which drainage basins show higher BIER values than small basins represents the sum of effective water consumptions in each when λ is constant. basin n (WCeff,n), is introduced. It should be noted that only a fraction of the evaporation recycling, which is returned to the originating basin via WCeff= ∑ WC eff,n (1) precipitation, will be available as ground or surface water. Since WAVE focuses on blue water only, the runoff fraction (α) WC is determined by subtracting total wastewater eff,n is implemented. Based on data derived from the hydrological discharges (WW ), evapo(transpi)ration recycling (ER ) and i i model WaterGAP2,25,26 α is determined by relating the long- synthetically created vapor recycling (VR ) from freshwater i term average runoff (R), that is, groundwater recharge and withdrawals (FW ) occurring within basin n. i surface runoff, to the total precipitation (P) within a drainage basin (eq 5). WC(FWWWERVR)eff,niiii=−−−∑ (2) R As shown in eqs 3a and 3b, the volumes of evapo(transpi)- α = ration and synthetically created vapor recycled within a basin P (5)

B dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article

Figure 2. Basin internal evaporation recycling (BIER) ratios denoting the fractions of evaporated water returning to the originating basins via precipitation.

As shown in Figure S1 in the Supporting Information (SI), α as characterization factors for impact assessment in water is highest (>80%) in basins located in Alaska and the Himalayas footprinting and LCA. and in the Amazon basin. The resulting hydrologically effective basin internal evaporation recycling (BIERhydrol‑eff), which is RFD=·∑ (WCeff,nn WDI ) (6) obtained by multiplying BIER with α, is shown in SI Figure S2. Since α is comparably low in Central Africa, large BIER ratios Tackling the shortcomings related to WTA, WDI is based on determined in, for example, the Congo basin (38%) are the consumption-to-availability (CTA) ratio,28 which relates reduced when considering the hydrologically effective fraction annual water consumption (C) to annual availability (A). The (BIERhydrol‑eff = 11%). Even though BIERhydrol‑eff is below 5% in freshwater availability of a drainage basin (A) expresses the most of the world’s drainage basins, it reduces blue water annually renewable freshwater volumes within the basin which consumption significantly in basins in the Himalayas, Alaska, can be quantified by means of runoff (plus upstream inflows if southeast Asia, and the North of South America (10−33%, SI the basin is divided into subcatchments). Data for C and A is Figure S2). available in WaterGAP2 for more than 11 000 basins on a global level. As shown in eq 7, CTA is modified in two steps. ■ VULNERABILITY EVALUATION MODEL C CTA*= ·AF In order to assess consequences resulting from water A + SWS GWS (7) consumption, many impact assessment models developed in LCA try to describe impacts on the areas of protection First, annually usable surface water stocks (SWS) are added resources, ecosystems, and human health.13 Some authors to A in order to consider lakes, wetlands, and dams in the model concrete cause-effect chains, like water consumption scarcity index. While storage volumes of dams (Vdam) are leads to less water available for irrigation, leading to less available directly,29 volumes of lakes and wetlands are productive agriculture, leading to health impacts due to determined by multiplying their surface areas (Alake/wetland) per 19,27 30 ff malnutrition. However, such end point models rely on basin with an e ective depth (deff = 5 m for lakes and 2 m for various assumptions. Moreover, regressions used to describe wetlands). impact pathways are often of low statistical significance. This In order to combine the volumes of dams, lakes, and shows that the relation between water consumption and wetlands (km3) with the flows C and A (km3/a), an annually impactsespecially on human health and ecosystemsis not usable fraction of 1% of the total volumes is used in the straightforward and depends on multiple variables. determination of SWS (eq 8). This means that ground and Therefore, this work focuses on freshwater resources only surface water stocks can be used for at least 100 years, even if and evaluates the regional vulnerability of drainage basins to no renewability occurs. blue water depletion. This vulnerability approach distinguishes ∑ ((VA+· d )) the WAVE model from conventional characterization models SWS = i dam,iii lake/wetland, eff, describing consequences on resources, such as ref 19. It is not 100 years (8) intended to “predict” impacts on freshwater resources but to denote the risk that water consumption in a certain region will The effective depths of lakes and wetlands are derived from lead to freshwater depletion. WaterGAP225,26 which represents the only data source for this This risk of freshwater depletion (RFD) can be determined kind of information on a global level. Just as the time horizon of by multiplying the effective water consumption in each basin 100 years, they can be regarded as conservative estimates which with its corresponding water depletion index (WDI). WDI are exceeded in many basins. Such methodological choices are denotes the vulnerability of drainage basins to freshwater in line with the vulnerability approach applied in this work, depletion based on physical blue water scarcity. It can be used which aims at assessing the risk that freshwater depletion can to interpret volumetric water footprints on a qualitative level or occur rather than predicting impacts. In order to evaluate the

C dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article fl fi * in uence of deff and the time horizon on the nal result, a Since CTA expresses a ratio of consumption to availability, comprehensive sensitivity analysis is accomplished in the SI. the resulting WDI takes into account relative freshwater scarcity In contrast to SWS, volumes of groundwater stocks (GWS) only. In order to consider absolute freshwater shortage as well, are not available on a global level. Therefore, an adjustment WDI is set to 1 per se in semiarid and arid basins34 shown in SI factor (AFGWS) is introduced, which reduces the scarcity ratio Figure S4. This setting is relevant as freshwater resources are based on the availability of groundwater. Using data provided highly vulnerable to depletion in (semi)arid regions regardless 31 fi by WHYMAP, AFGWS is de ned based on geological structure of the relative scarcity. and annual recharge as shown in Table 1. The scarcity As shown in Figure 3, WDI is at the highest level in many drainage basins located in Central Asia, India, Saudi Arabia, Table 1. Adjustment Factors for Ground Water Stocks Australia, Northern and Southern Africa, Mexico, the southwest (AFGWS) Reducing Water Scarcity Based on Geological of the U.S., and the Andes. In contrast, little or no freshwater Structure and Annual Recharge resource depletion is caused by water consumption in basins located in Russia, Canada, Northern Europe, or around the annual recharge scarcity equator. geological structure (mm) reduction AFGWS major groundwater basin >300 10.0% 0.900 ■ CASE STUDY 100−300 7.5% 0.925 complex hydrogeological >300 5.0% 0.950 The methodology developed above is tested by means of an structure 100−300 2.5% 0.975 existing WF study of bioethanol produced from sugar cane in fi 35 others 0.0% 1.000 ve producing countries. Table S1 in the SI shows the blue water consumption (evapotranspiration of blue irrigation water) required to produce 1 GJ of bioethanol from sugar reduction rates have been derived from discussions with the 32 cane in Columbia, Mexico, Thailand, Australia, and Zambia developers of the WHYMAP. In line with the vulnerability based on ref 32. By means of country specific factors for approach, moderate reduction rates are selected. Their ff fl fi BIERhydrol‑eff and WDI, the e ective water consumption (WCeff) in uence on the nal result has been analyzed in a sensitivity and the risk of freshwater depletion (RFD) are determined (SI analysis presented in the SI. Fossil groundwater stocks are fi Table S1). In order to compare the results obtained by means excluded from this analysis as they cannot be quanti ed on a of the WAVE model to those obtained by other impact global level and it is not sure that they can be accessed in every assessment methods, potential impacts resulting from WC are part of the world. eff ’ additionally evaluated by means of the models developed by The factor WDI aims at assessing a basin s vulnerability to fi 19 18 * P ster et al. (2009) and Frischknecht et al. (2009). freshwater depletion based on CTA as shown in eq 9. It can Figure 4 shows the water consumption figures on a relative be understood as an equivalent volume of depleted water scale and presents the reductions in WC compared to WC resulting from a volume of water consumption. Similar to eff resulting from the consideration of BIER ‑ . Moreover, existing scarcity indexes,19,27 the logistic function plotted in SI hydrol eff potential impacts determined by means of WAVE as well as by Figure S3 leads to nonlinear transformation of physical water the models of Pfister et al. (2009) and Frischknecht et al. scarcity into vulnerability to freshwater depletion. This is (2009) are shown normalized to the highest result in each important as in the upper and lower ranges of CTA* doubled category (for absolute results see SI Table S1). scarcity does not necessarily lead to doubled vulnerability to It should be noted that a consideration on the country level depletion. WDI turns 1 above a CTA* of 0.25, which is 33 has its limitations as plants may be grown in particular basin regarded as the threshold of extreme water stress. whose hydrogeological situation may differ from the country 1 average. For example sugar cane from Australia is mainly WDI = 36 11+−e−·40 CTA * 1 produced in the coastal areas showing less severe water stress ()0.01 (9) than the country average. However, as ref 35 provides data on

3 3 Figure 3. Factors WDI expressing the vulnerability of basins to freshwater resource depletion [m depleted/m consumed].

D dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article

Figure 4. Relative presentation of blue water consumption required to produce 1 GJ of bioethanol from sugar cane, reduction of water consumption fi due to consideration of BIERhydrol‑eff, and potential impacts determined by means of WAVE and by the impact assessment methods of P ster et al. (2009) and Frischknecht et al. (2009). the country and state level and since we provide BIER and The three assessment methods also lead to different WDI on the level of countries and river basins, the country level conclusions regarding bioethanol produced from sugar cane is the lowest common denominator to be used in this study. in Australia. While the required irrigation water consumption of Moreover, the testing of the WAVE model and the comparison 23 m3/GJ causes the highest risk of freshwater depletion in to other impact assessment methods is regarded more WAVE, it is considered less relevant than production in Mexico important than the absolute result in this case. and Thailand in the other impact assessment models. In the On the inventory level, the consideration of the hydro- method of Frischknecht et al. (2009), suggested by the logically effective evaporation recycling by means of European Union for the product environmental footprint,38 3 BIERhydrol‑eff leads to a reduction of water consumption between water consumption in Australia (23 eco-points/m ) is even 0% in Australia and 10% in Columbia. If the total basin internal regarded far less relevant than in countries abounding in water evaporation recycling (BIER) is taken into account, the like Germany (910 eco-points/m3). These unexpected results recycled fractions of evapotranspirated irrigation water will can be explained by the fact that the methods of Frischknecht increase significantly (up to 24% in Zambia). et al. (2009) and Pfister et al. (2009) consider relative The assessment of potential impacts resulting from irrigation freshwater scarcity only. Yet, even though only a comparably water consumption in the countries considered leads to small fraction of Australia’s water availability is used, the ff different conclusions than a volumetric analysis. All three country su ers from absolute freshwater shortage. This assessment methods come to the result that the largest water highlights the relevance of considering absolute scarcity by consumption in Zambia (39m3/GJ) actually causes the lowest means of aridity in the WAVE model. impacts, as no water scarcity is detected in this country on an annual basis. However, this also shows the limitations of an ■ DISCUSSION annual assessment method in a seasonal product system. As Scope. The WAVE model developed in this work focuses discussed in detail in the following chapter, water consumption on blue water consumption which occurs mainly due to and water scarcity can vary throughout the yearespecially in evapo(transpi)ration or product integration of ground and 37 countries with dry and wet seasons. surface water. While the vulnerability evaluation model is The relatively high water consumption of bioethanol restricted to assess blue water consumption only, BIER and 3 ff produced in Thailand (18 m /GJ) is evaluated di erently in BIERhyrdol‑eff can also be used to assess the basin internal the assessment models. While the risk of freshwater depletion recycling of plant evapotranspiration (green water consump- (RFD) is considered low in the WAVE model, significant tion). impacts are expected in the methods of Pfister et al. (2009) and As the WAVE model does not consider water quality Frischknecht et al. (2009). The reason for this can be found in degradation in its default version, two possibilities of including the underlying methodologies. While a consumption-to- water quality aspects are presented. When the method availability ratio is considered in WAVE, the two other impact described here is applied in an LCA study, freshwater pollution assessment models are based on a withdrawal-to-availability is assessed by means of impact categories like eutrophication, ratio according to which Thailand is much more water scarce. acidification, or human and eco-toxicity. However, similar to

E dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article the gray water footprint, this approach does not consider the simplification that basins are of quadratic shape leads to an water quality of freshwater inputs. Therefore, a possibility of under or overestimation of BIER and BIER100 depending on determining the quality corrected effective water consumption the actual shape and the prevailing wind directions (SI Figure (WCq,eff,n) which can be used as the basis for calculating the S6). quality corrected risk of freshwater depletion (QRFD), is The consideration of basin internal evaporation recycling presented in the SI. leads to an interesting effect when considering agricultural In the WAVE model, evaporation recycling and freshwater product systems: Even the evapotranspiration of green water by scarcity are determined based on annual averages. As agricultural plants can cause blue water benefits. The reason is mentioned in the Case Study, this is a limitation as climatic that parts of this green water evapotranspiration will be recycled conditions influencing evaporation recycling as well as the within the drainage basin (BIER), of which parts will be hydrological situation in a basin might vary throughout a year. hydrologically effective (α). Especially the combination of these effects can be relevant in Moreover, the accounting model developed explicitly semiarid drainage basins, as BIER may be high in the rainy considers the emission of water vapor created in chemical season when water is abundant but low in the dry season when reactions and its partial return due to atmospheric moisture water scarcity is of concern. For this reason, some authors recycling effects (Figure 1). Consequently, the combustion of introduce monthly scarcity factors39 which are especially fossil fuels may lead to negative effective water consumption if relevant for agricultural products grown during particular the synthetically created vapor recycling is higher than the seasons. However, such approaches require temporally explicit difference between freshwater withdrawals, wastewater dis- inventory data, which is difficult to obtain − especially if charges, and evaporation recycling (eq 2). complex background systems are involved. As an alternative, So far, WAVE has considered the evapo(transpi)ration consumption weighted annual averages of monthly scarcity recycling within drainage basins only leading to a global average factors can be used.40 Yet, this does not overcome the key BIER of 1%. However, the average continental evapotranspira- methodological challenge of a monthly scarcity assessment: the tion returning as continental precipitation amounts to 57%.24 consideration of intermonthly storage capacities which can Thus, the examination of basin internal evaporation recycling buffer water scarce periods throughout the year.40 Moreover, (BIER) effects should be extended to a fate of evaporation the temporal resolution of water scarcity assessments also analysis15 which considers the fractions of evapo(transpi)ration determines the required spatial resolution. Large basins can returning to other basins as well. have flow times of several months from spring to mouth, which Vulnerability Evaluation Model. In this work the makes a monthly assessment difficult. Therefore, this work vulnerability of a drainage basin to freshwater depletion is refrains from providing monthly or weighted annual BIER and evaluated. Based on physical water scarcity, the water depletion WDI factors but acknowledges them as an important subject to index (WDI) denotes the risk that water consumption leads to future research. freshwater depletion. Water Accounting Model. The accounting approach In contrast to most other water scarcity indicators used as presented in this work considers the basin internal recycling impact factors in water footprinting, WDI is based on a of the share of water withdrawal consumed due to evapo- consumption instead of withdrawal-to-availability ratio. Even (transpi)ration. Even though this complies with the definition though withdrawal implicitly accounts for quality degradation of water consumption,16 it may appear arguable whether the as well, a consumption based indicator expresses water shortage reduction of actual water consumption is reasonable in large more realistically as large shares of cooling water, which are basins, like the Danube, where evaporation recycling can occur returned with low quality degradation due to temperature after hundreds of kilometers. In our opinion such an approach increase, are excluded. is justified for three reasons. First, a river basin delineation of Moreover, for the first time ground and surface water stocks WaterGAP2 is used which divides the world’s 34 largest are included in a water scarcity indicator. As shown in SI Figure drainage basins into subcatchments. This avoids extremely long S7, the consideration of aquifers, lakes and wetlands leads to a evaporation recycling distances that would otherwise occur in, scarcity reduction of up to 10% in many basins around the for example, the Congo basin. Second, in several countries, like globe; especially in Canada, Central Africa, Central Europe, the U.S. or Australia, water is transported from withdrawal to South America, and Russia. Even higher reductions of more use through pipelines over long distances. The fraction of than 80% are achieved in small basins in Alaska and the withdrawal returned to the river also reduces the water Himalayas. Thus, the consideration of ground and surface water consumption volume in such cases. Hence, atmospheric water stocks leads to a further scarcity reduction in regions which are transport should be treated in the same way as anthropogenic under low water stress anyway. Even though the logistic transport. Third, in many basins the downwind transport of function diminishes this reduction effect in the final WDI result evaporated water is in fact an upstream transport from a river (SI Figure S3), the difference between water scarce regions, like perspective as, for example, in the Amazon basin.24 This means Saudi Arabia, and regions abounding in water, such as Canada, that recycled evaporation is actually returning water, which is is increased. This higher precision in water scarcity assessment available for a possible second consumption. is especially relevant when comparing water consumption in ff Nevertheless, additional BIER100 ratios are determined as a di erent regions. sensitivity check by restricting the evaporation recycling By setting WDI to the highest value (1.00) in arid and distances (x) to 100 km (eq 4). As shown in SI Figure S5, semiarid basins, WAVE considers absolute freshwater shortage fi BIER100 is signi cantly lower than BIER (Figure 1) with a in addition to relative scarcity. This helps to avoid the maximum of 19% determined in a Colombian drainage basin. mathematical artifact that dry regions are regarded uncritical if Apart from Australia, Northern Africa, Saudi Arabia, and large consumption is close to zero. SI Figure S8 presents the fl fi regions in Central Asia BIER100 ranges from 1 to 5% in most of in uence of this setting on the nal WDI result. Without this the world’s basins. Yet, it should be noted that the model consideration of absolute freshwater shortage, most of the arid

F dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article and semiarid basins (SI Figure S4) would have significantly (3) Hoekstra, A. Y.; Chapagain, A. K.; Aldaya, M. M.; Mekonnen, M. lower WDI results than those shown in Figure 3. Especially, M. The Water Footprint Assessment ManualSetting the Global impacts from water consumption in the Sahel zone or in Standard; Earthscan: London, Washington, DC, 2011. Australia would be close to zero, as it is the case in existing (4) Chapagain, A. K.; Hoekstra, A. Y. The water footprint of coffee impact assessment methods like Frischknecht et al. (2009). and tea consumption in the Netherlands. Ecol. Econ. 2007, 1 (64), 109−118. So far, WAVE has only assessed the vulnerability of drainage (5) Chapagain, A. K.; Hoekstra, A. Y.; Savenije, H. H. G.; Gautam, R. basins to freshwater depletion from a blue water resource The water footprint of cotton consumption: An assessment of the perspective. In future research also the vulnerability to human impact of worldwide consumption of cotton products on the water health and ecosystem impacts should be analyzed. By means of resources in the cotton producing countries. Ecol. Econ. 2006, 60 (1), sensitivity factors the risk that water consumption in water 186−203. scarce regions can lead to impacts could be analyzed. However, (6) Hoekstra, A. Y.; Mekonnen, M. M. The water footprint of especially when assessing the vulnerability to health impacts, humanity. Proc. Natl. Acad. Sci. U.S.A. 2012, 109 (9), 3232−3237. the consumption and availability of green water need to be (7) Suweis, S.; Rinaldo, A.; Maritana, A.; D’Odorico, P. Water- considered in combination with blue water. This combined controlled wealth of nations. Proc. Natl. Acad. Sci. U.S.A. 2012, 110 (11), 4230−4233. approach is needed as both types of water are equally important  for food production and there are many countries which suffer (8) Finkbeiner, M. Carbon footprinting Opportunities and threats. Int. J. LCA 2009, 14,91−94. from blue water scarcity but have enough green water to grow ‘‘ ’’ 41 (9) Pfister, S.; Hellweg, S. The water shoesize vs. footprint of crops. bioenergy. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 (35), E93−E94. Application. BIER, BIERhydrol-eff, and WDI are determined (10) Ridoutt, B. G.; Huang, J. Environmental relevanceThe key to on the level of drainage basins, as they reflect hydrologic understanding water footprints. Proc. Natl. Acad. Sci. U.S.A. 2012, 109 conditions best. Since inventory information is often not (22), E1424. available on such a detailed geographic resolution, all factors are (11) Hoekstra, A. Y.; Gerbens-Leenes, W.; van der Meer, T. H. Reply provided on the country level as well. In order to promote the to Pfister and Hellweg: Water footprint accounting, impact assess- applicability of the WAVE model, BIER, BIERhydrol-eff, and WDI ment, and life-cycle assessment. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 are made available free of charge on both drainage basin and (40), 114. country levels in a Google Earth layer and spreadsheet, (12) Hoekstra, A. Y.; Mekonnen, M. M. Reply to Ridoutt and Huang: respectively: http://www.see.tu-berlin.de/wave/parameter/en/. From water footprint assessment to policy. Proc. Natl. Acad. Sci. U.S.A. fi 2012, 109 (22), E1425. For the determination of country speci c factors consumption (13) Berger, M.; Finkbeiner, M. Water footprintingHow to address weighted averages are used. This adds higher weight to those water use in life cycle assessment? Sustainability 2010, 2 (4), 919−944. basin fractions within a country which contribute a higher share (14) Berger, M.; Warsen, J.; Krinke, S.; Bach, V.; Finkbeiner, M. to the country’s total consumption. Uncertainties related to the Water footprint of European cars: Potential impacts of water creation of country averages are discussed in the SI and consumption along automobile life cycles. Environ. Sci. Technol. quantified in the spreadsheet. 2012, 46 (7), 4091−4099. (15) Berger, M.; Finkbeiner, M. Methodological challenges in ■ ASSOCIATED CONTENT volumetric and impact oriented water footprints. J. Ind. Ecol. 2013, 17 (1), 79−89. *S Supporting Information (16) Bayart, J. B.; Bulle, C.; Koehler, A.; Margni, M.; Pfister, S.; Additional explanations, figures, and tables are available in the Vince, F.; Deschenes, L. A framework for assessing off-stream Supporting Information. This material is available free of charge freshwater use in LCA. Int. J. Life Cycle Assess. 2010, 15 (5), 439−453. via the Internet at http://pubs.acs.org. (17) van der Ent, R. J.; Savenije, H. H. G. Length and time scales of atmospheric moisture recycling. Atmos. Chem. Phys. 2011, 11, 1853− ■ AUTHOR INFORMATION 1863. (18) Frischknecht, R.; Steiner, R.; Jungbluth, N. The Ecological Corresponding Author Scarcity MethodEco-Factors 2006A Method for Impact Assessment in ffi *Phone: +49.(0)30.314-25084; fax: +49.(0)30.314-21720; e- LCA; Federal O ce for the Environment: Bern, Swizerland, 2009. mail: [email protected]. (19) Pfister, S.; Koehler, A.; Hellweg, S. Assessing the environmental impacts of freshwater consumption in LCA. Environ. Sci. Technol. Notes 2009, 43 (11), 4098−4104. The authors declare no competing financial interest. (20) Alcamo, J.; Flörke, M.; Marker,̈ M. Future long-term changes in global water resources driven by socio-economic and climatic changes. ■ ACKNOWLEDGMENTS Hydrol. Sci. J. 2007, 52 (2), 247−275. (21) Timmerman, A.; Feyen, J. The WAVE model and its We express sincere thanks to Stephan Pfister (Swiss Federal application; Simulation of the substances water and agrochemicals in Institute of Technology Zurich), Martina Flörke (Center for the soil, crop and vadose environment. Revista Corpoica 4 (1), 36-41. Environmental Systems Research, Kassel), as well as Andrea (22) ISO 14044. Environmental ManagementLife Cycle Assessment - Richts and Wilhelm Struckmeier (Federal Institute for Requirements and Guidelines (ISO 14044:2006); International Organ- Geosciences and Natural Resources, Hannover) for providing isation for Standardisation, Ed.; Geneva, Switzerland, 2006. datasets used in this work and many fruitful discussions. (23) Boulay, A.-M.; Hoekstra, A. Y.; Vionnet, S. Complementarities of Water-Focused Life Cycle Assessment and Water Footprint Assessment. Environ. Sci. Technol. 2013, 47 (21), 11926−11927. ■ REFERENCES (24) van der Ent, R. J.; Savenije, H. H. G.; Bettina, S.; Steele-Dunne, (1) United Nations Water and Food and Agricultural Organization S. C. Origin and fate of atmospheric moisture over continents. Water Coping with water scarcitychallenge of the twenty first century. Resour. Res. 2010, 46. (2) ISO FDIS 14046. Water FootprintPrinciples, Requirements and (25) Döll, P.; Kaspar, F.; Lehner, B. A global hydrological model for Guidance; International Organization for Standardization, Ed.; Geneva, deriving water availability indicators: Model tuning and validation. J. Switzerland, 2012. Hydrol. 2003, 270 (1−2), 105−134.

G dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX Environmental Science & Technology Article

(26) Flörke, M.; Kynast, E.; Barlund,̈ I.; Eisner, S.; Wimmer, F.; Alcamo, J. Domestic and industrial water uses of the past 60 years as a mirror of socio-economic development: A global simulation study. Global Environ. Change 2013, 23 (1), 144−156. (27) Boulay, A.-M.; Bulle, C.; Bayart, J.-B.; Deschenes, L.; Margni, M. Regional characterization of freshwater use in LCA: Modelling direct impacts on human health. Environ. Sci. Technol. 2011b, 45 (20), 8948− 8957. (28) Boulay, A.-M.; Bulle, C.; Bayart, J.-B.; Deschenes, L.; Margni, M. Regional characterization of freshwater use in LCA: Modelling direct impacts on human health. Environ. Sci. Technol. 2011b, 45 (20), 8948− 8957. (29) Lehner, B.; et al. High resolution mapping of the world’s reservoirs and dams for sustainable river flow management. Front. Ecol. Environ. 2011, 9 (9), 494−502. (30) Lehner, B.; Döll, P. Development and validation of a global database of lakes, reservoirs and wetlands. J. Hydrol. 2004, 296 (1−4), 1−22. (31) Richts, A.; Struckmeier, W.; Zaepke, M. WHYMAP and the groundwater resources of the world 1:25,000,000. In Sustaining Groundwater Resources; Jones, J. A. A., Ed.; Springer: Heidelberg, 2011; pp 159−173. (32) Struckmeier, W.; Richts, A. Personal communication (February 19, 2013). Federal Institute for Geosciences and Natural Resources, 2013 (33) Richter, B. D.; Davis, M. M.; Apse, C.; Konrad, C. Short communication: A presumptive standard for environmental flow protection. River Res. Appl. 2011, DOI: 10.1002/rra.1511. (34) United Nations Environment Programme. World Atlas of Desertification, 2 ed.; Arnold: London, 1997; p 192. (35) Mekonnen, M. M.; Hoekstra, A. Y. The green, blue and grey water footprint of crops and derived crop products. Hydrol. Earth Syst. Sci. 2011, 15 (5), 1577−1600. (36) Griggs, P. D. Deforestation and sugar cane growing in Eastern Australia. Environ. Hist. 2007, 13 (3), 255−283. (37) Savenije, H. H. G. Water scarcity indicators; the deception of the numbers. Phys. Chem. Earth, Part B 2000, 25 (3), 199−204. (38) European Union Commission Recommendation of 9 April 2013 on the use of common methods to measure and communicate the life cycle environmental performance of products and organisations. Off.J. Eur. Union 2013, L124, 56. (39) Hoekstra, A. Y.; Mekonnen, M. M.; Chapagain, A. K.; Mathews, R. E.; Richter, B. D. Global monthly water scarcity: Blue water footprints versus blue water availability. PLoS One 2012, 7 (2), e32688. (40) Pfister, S.; Baumann, J., Monthly characterization factors for water consumption and application to temporally explicit cereals inventory. In 8th International Conference on LCA in the Agri-Food Sector, Rennes, France, 2−4 October 2012, 2012. (41) Rockström, J.; Falkenmark, M.; Karlberg, L.; Hoff, H.; Rost, S.; Gerten, D. Future water availability for global food production: The potential of green water for increasing resilience to global change. Water Resour. Res. 2009, 45 (7), W00A12.

H dx.doi.org/10.1021/es404994t | Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Country factors for BIER, BIERhydrol-eff, and WDI

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin Afghanistan 5.3% 16.0% 1.0% 2.5% 8.3% 0.1% 0.95 1.00 0.58 Albania 2.4% 14.0% 1.0% 1.7% 4.9% 0.6% 0.27 0.98 0.01 Algeria 1.1% 7.0% 0.0% 0.3% 1.5% 0.0% 0.92 1.00 0.01 American Samoa 1.5% 2.0% 2.0% 0.8% 0.8% 0.8% 0.01 0.01 0.01 Andorra 3.3% 3.0% 3.0% 1.3% 1.5% 0.8% 0.20 0.66 0.08 Angola 16.0% 35.0% 0.0% 3.4% 7.0% 0.0% 0.30 1.00 0.01 Anguilla 0.7% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 Antigua & Barbuda 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Argentina 7.3% 19.0% 0.0% 1.1% 5.4% 0.0% 0.38 1.00 0.01 Armenia 7.7% 8.0% 2.0% 1.7% 1.7% 0.7% 0.95 1.00 0.18 Aruba 0.4% 0.0% 0.0% 0.0% 0.0% 0.0% 1.00 1.00 1.00 Australia 2.3% 6.0% 0.0% 0.2% 1.2% 0.0% 0.91 1.00 0.01 Austria 13.1% 14.0% 5.0% 4.7% 4.9% 1.5% 0.18 0.18 0.10 Azerbaijan 6.7% 8.0% 0.0% 1.5% 1.7% 0.0% 0.97 1.00 0.02 Bahrain 0.1% 0.0% 0.0% 0.0% 0.1% 0.0% 1.00 1.00 1.00 Bangladesh 8.3% 34.0% 1.0% 5.7% 32.8% 0.5% 0.19 0.95 0.01 Barbados 0.7% 1.0% 1.0% 0.2% 0.2% 0.2% 0.46 0.47 0.01 Belarus 7.9% 9.0% 3.0% 1.7% 3.2% 1.0% 0.11 0.13 0.01 Belgium 2.2% 6.0% 1.0% 1.0% 2.8% 0.3% 0.86 0.99 0.15 Belize 2.3% 5.0% 1.0% 0.8% 2.7% 0.5% 0.01 0.01 0.01 Benin 6.5% 8.0% 3.0% 1.4% 1.5% 1.0% 0.55 1.00 0.01 Bermuda 0.4% 0.0% 0.0% 0.1% 0.1% 0.1% 0.02 0.02 0.02 Bhutan 33.5% 34.0% 5.0% 32.8% 32.8% 1.9% 0.01 0.01 0.01 Bolivia 15.5% 22.0% 0.0% 3.5% 10.0% 0.0% 0.22 1.00 0.01 Bosnia & Herzegovina 13.6% 14.0% 1.0% 4.8% 4.9% 0.7% 0.18 0.18 0.01 Botswana 9.9% 24.0% 9.0% 1.0% 4.1% 0.4% 0.97 1.00 0.01 Brazil 10.1% 29.0% 0.0% 4.4% 17.1% 0.0% 0.08 1.00 0.01 British Indian Ocean Territory 1.4% 1.0% 1.0% 0.5% 0.5% 0.5% 0.02 0.02 0.02 British Virgin Is. 0.8% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Brunei 3.9% 6.0% 2.0% 2.2% 3.4% 1.6% 0.01 0.01 0.01 Bulgaria 6.1% 14.0% 1.0% 2.1% 4.9% 0.1% 0.56 1.00 0.01 Burkina Faso 7.5% 8.0% 6.0% 1.3% 1.5% 1.0% 0.52 1.00 0.01 Burundi 22.2% 25.0% 15.0% 3.5% 3.9% 2.3% 0.01 0.01 0.01 Cambodia 19.7% 23.0% 1.0% 8.8% 10.2% 0.6% 0.03 0.10 0.01 Cameroon 8.2% 23.0% 0.0% 2.8% 8.6% 0.1% 0.39 1.00 0.01 Canada 9.3% 20.0% 0.0% 3.8% 16.6% 0.0% 0.07 1.00 0.01 Cape Verde 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 1.00 1.00 1.00 Cayman Is. 0.7% 1.0% 1.0% 0.1% 0.1% 0.1% 0.22 0.22 0.22 Central African Republic 19.9% 27.0% 14.0% 4.7% 7.2% 2.7% 0.01 0.01 0.01 Chad 4.1% 16.0% 0.0% 1.0% 3.5% 0.0% 0.80 1.00 0.01 Chile 0.9% 22.0% 0.0% 0.6% 4.4% 0.0% 0.69 1.00 0.01

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin China 9.4% 34.0% 0.0% 4.1% 32.8% 0.0% 0.58 1.00 0.01 Christmas I. 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Cocos Is. 0.7% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 Colombia 20.2% 29.0% 0.0% 10.0% 21.5% 0.2% 0.02 1.00 0.01 Comoros 0.8% 1.0% 1.0% 0.4% 0.4% 0.3% 0.01 0.01 0.01 Congo 17.7% 27.0% 3.0% 5.7% 8.6% 1.3% 0.01 0.01 0.01 Congo, DRC 21.0% 38.0% 1.0% 5.7% 10.9% 0.3% 0.01 0.01 0.01 Cook Is. 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Costa Rica 3.5% 5.0% 1.0% 2.0% 3.5% 0.5% 0.01 0.02 0.01 Cote d'Ivoire 7.6% 9.0% 3.0% 1.7% 4.6% 0.9% 0.13 1.00 0.01 Croatia 13.1% 14.0% 1.0% 4.6% 4.9% 0.4% 0.17 0.18 0.01 Cuba 1.2% 2.0% 1.0% 0.3% 0.4% 0.0% 0.50 1.00 0.01 Cyprus 0.4% 1.0% 0.0% 0.2% 0.3% 0.2% 0.75 1.00 0.12 Czech Republic 8.4% 14.0% 4.0% 2.8% 4.9% 1.0% 0.14 0.18 0.03 Denmark 0.7% 1.0% 0.0% 0.4% 0.6% 0.2% 0.28 0.93 0.01 Djibouti 3.1% 5.0% 0.0% 0.9% 1.4% 0.0% 1.00 1.00 1.00 Dominica 0.6% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 Dominican Republic 1.8% 3.0% 1.0% 0.4% 0.7% 0.1% 0.30 1.00 0.01 Ecuador 13.7% 29.0% 0.0% 7.5% 16.3% 0.0% 0.19 1.00 0.01 Egypt 0.1% 1.0% 0.0% 0.0% 0.3% 0.0% 1.00 1.00 1.00 El Salvador 2.4% 4.0% 1.0% 1.3% 1.9% 0.7% 0.01 0.01 0.01 Equatorial Guinea 8.0% 23.0% 3.0% 5.2% 8.6% 2.4% 0.01 0.01 0.01 Eritrea 2.2% 4.0% 0.0% 0.5% 0.9% 0.0% 1.00 1.00 1.00 Estonia 1.8% 4.0% 0.0% 0.7% 1.7% 0.2% 0.02 0.02 0.01 Ethiopia 9.4% 18.0% 0.0% 2.5% 4.8% 0.0% 0.52 1.00 0.01 Falkland Is. 0.6% 1.0% 0.0% 0.1% 0.2% 0.1% 0.01 0.01 0.01 Faroe Is. 0.8% 1.0% 1.0% 0.6% 0.6% 0.6% 0.01 0.01 0.01 Fiji 1.1% 1.0% 0.0% 0.5% 0.6% 0.0% 0.01 0.01 0.01 Finland 3.1% 8.0% 1.0% 1.3% 3.3% 0.2% 0.20 0.77 0.01 France 3.2% 6.0% 0.0% 1.3% 4.3% 0.2% 0.22 1.00 0.01 French Guiana 6.7% 8.0% 1.0% 2.4% 2.7% 0.7% 0.01 0.01 0.01 French Polynesia 1.1% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 French Southern & Antarctic Lands 0.5% 1.0% 0.0% 0.2% 0.4% 0.2% 0.01 0.01 0.01 Gabon 16.0% 23.0% 2.0% 6.4% 8.6% 1.3% 0.01 0.01 0.01 Gaza Strip 0.3% 0.0% 0.0% 0.1% 0.1% 0.1% 1.00 1.00 1.00 Georgia 7.2% 8.0% 1.0% 2.0% 3.7% 1.0% 0.77 1.00 0.01 Germany 6.2% 14.0% 0.0% 2.5% 4.9% 0.2% 0.17 1.00 0.01 Ghana 7.1% 8.0% 2.0% 1.2% 1.8% 0.3% 0.01 0.02 0.01 Gibraltar 0.3% 0.0% 0.0% 0.2% 0.2% 0.2% 0.44 0.44 0.44 Glorioso Is. 0.7% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Greece 1.6% 4.0% 0.0% 0.9% 1.7% 0.2% 0.69 1.00 0.01 Greenland 0.7% 6.0% 0.0% 0.5% 2.7% 0.0% 0.01 0.01 0.01

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin Grenada 0.7% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 Guadeloupe 0.6% 1.0% 1.0% 0.1% 0.2% 0.1% 0.01 0.02 0.01 Guam 1.0% 1.0% 1.0% 0.7% 0.7% 0.7% 0.01 0.01 0.01 Guatemala 4.0% 7.0% 1.0% 2.0% 3.1% 0.5% 0.01 0.02 0.01 Guernsey 0.5% 1.0% 1.0% 0.3% 0.4% 0.3% 0.01 0.01 0.01 Guinea 5.2% 9.0% 1.0% 1.7% 5.4% 0.5% 0.65 1.00 0.01 Guinea-Bissau 2.6% 3.0% 1.0% 1.2% 1.9% 0.5% 0.01 0.01 0.01 Guyana 3.6% 24.0% 1.0% 1.5% 12.7% 0.2% 0.02 0.10 0.01 Haiti 1.5% 3.0% 1.0% 0.4% 0.7% 0.1% 0.16 0.95 0.01 Heard I. & McDonald Is. 0.8% 1.0% 1.0% 0.4% 0.4% 0.4% 0.01 0.01 0.01 Honduras 4.2% 5.0% 1.0% 1.9% 2.7% 0.4% 0.01 0.01 0.01 Hungary 13.8% 14.0% 14.0% 4.9% 4.9% 4.9% 0.18 0.18 0.18 Iceland 1.5% 2.0% 1.0% 1.2% 2.0% 0.3% 0.01 0.02 0.01 India 8.2% 34.0% 0.0% 3.8% 32.8% 0.0% 0.74 1.00 0.01 Indonesia 4.5% 21.0% 1.0% 2.3% 12.9% 0.0% 0.17 1.00 0.01 Iran 1.8% 8.0% 0.0% 0.9% 2.0% 0.0% 0.95 1.00 0.01 Iraq 3.1% 4.0% 0.0% 1.5% 2.0% 0.1% 0.93 1.00 0.87 Ireland 0.9% 2.0% 0.0% 0.6% 1.0% 0.3% 0.10 0.21 0.01 Isle of Man 0.6% 1.0% 1.0% 0.4% 0.4% 0.4% 0.02 0.02 0.01 Israel 0.4% 1.0% 0.0% 0.2% 0.3% 0.0% 0.97 1.00 0.65 Italy 3.8% 14.0% 0.0% 2.5% 4.9% 0.1% 0.64 1.00 0.01 Jamaica 0.8% 1.0% 1.0% 0.2% 0.4% 0.1% 0.39 1.00 0.01 Jan Mayen, Svalbard 1.9% 2.0% 0.0% 1.5% 1.6% 0.0% 0.01 0.01 0.01 Japan 1.2% 2.0% 0.0% 0.8% 2.4% 0.1% 0.48 1.00 0.01 Jarvis I., Baker I., Howland I., Johnston Atoll, Midway Is.; Wake I. 0.5% 1.0% 0.0% 0.1% 0.3% 0.1% 0.01 0.01 0.01 Jersey 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.41 0.41 0.41 Jordan 0.7% 3.0% 0.0% 0.3% 1.3% 0.0% 1.00 1.00 1.00 Juan De Nova I. 0.6% 1.0% 1.0% 0.2% 0.2% 0.2% 0.02 0.02 0.02 Kazakhstan 2.7% 16.0% 0.0% 0.7% 6.5% 0.0% 0.95 1.00 0.01 Kenya 6.7% 25.0% 1.0% 1.5% 4.8% 0.1% 0.71 1.00 0.01 Kiribati 0.3% 0.0% 0.0% 0.0% 0.0% 0.0% 0.01 0.01 0.01 Kuwait 1.0% 2.0% 0.0% 0.4% 0.6% 0.1% 1.00 1.00 1.00 Kyrgyzstan 4.4% 7.0% 2.0% 1.6% 4.0% 0.8% 1.00 1.00 0.01 Laos 19.9% 23.0% 2.0% 8.8% 10.2% 0.9% 0.02 0.05 0.01 Latvia 3.0% 4.0% 1.0% 1.2% 1.7% 0.3% 0.02 0.09 0.01 Lebanon 0.6% 1.0% 0.0% 0.4% 0.7% 0.2% 0.67 1.00 0.14 Lesotho 8.8% 9.0% 2.0% 0.8% 0.8% 0.6% 0.96 1.00 0.01 Liberia 6.4% 8.0% 3.0% 4.4% 5.4% 2.0% 0.01 0.01 0.01 Libya 0.4% 2.0% 0.0% 0.2% 0.8% 0.0% 1.00 1.00 1.00 Liechtenstein 6.1% 6.0% 6.0% 2.8% 2.8% 2.8% 0.15 0.15 0.15 Lithuania 3.5% 6.0% 1.0% 1.2% 1.7% 0.5% 0.02 0.03 0.01

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin Luxembourg 5.7% 6.0% 2.0% 2.6% 2.8% 0.9% 0.25 0.99 0.15 Macedonia 3.4% 4.0% 1.0% 1.7% 2.6% 0.6% 0.17 0.93 0.02 Madagascar 4.4% 7.0% 0.0% 2.0% 3.9% 0.0% 0.04 1.00 0.01 Malawi 20.4% 24.0% 2.0% 3.6% 4.3% 1.0% 0.01 0.01 0.01 Malaysia 3.9% 16.0% 1.0% 1.9% 8.0% 0.0% 0.03 0.25 0.01 Mali 5.0% 8.0% 0.0% 1.0% 1.5% 0.0% 0.99 1.00 0.01 Malta 0.4% 0.0% 0.0% 0.2% 0.2% 0.2% 1.00 1.00 1.00 Marshall Is. 1.3% 1.0% 1.0% 0.6% 0.6% 0.6% 0.01 0.01 0.01 Martinique 0.6% 1.0% 1.0% 0.3% 0.3% 0.2% 0.01 0.02 0.01 Mauritania 3.0% 4.0% 0.0% 0.6% 0.8% 0.0% 1.00 1.00 1.00 Mauritius 0.4% 0.0% 0.0% 0.1% 0.1% 0.1% 0.03 0.06 0.01 Mayotte 0.8% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Mexico 5.0% 10.0% 0.0% 0.7% 3.4% 0.0% 0.78 1.00 0.01 Micronesia 1.5% 2.0% 2.0% 0.9% 0.9% 0.9% 0.01 0.01 0.01 Moldova 7.8% 14.0% 1.0% 2.5% 4.9% 0.2% 0.45 0.83 0.06 Monaco 0.9% 1.0% 1.0% 0.6% 0.6% 0.6% 0.03 0.03 0.03 Mongolia 8.5% 17.0% 0.0% 2.7% 8.3% 0.0% 0.58 1.00 0.01 Montenegro 8.7% 14.0% 1.0% 3.7% 4.9% 0.7% 0.10 0.18 0.01 Montserrat 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Morocco 0.8% 2.0% 0.0% 0.3% 0.5% 0.0% 0.99 1.00 0.04 Mozambique 8.2% 24.0% 0.0% 1.2% 4.1% 0.0% 0.49 1.00 0.01 Myanmar 16.4% 34.0% 1.0% 12.3% 32.8% 0.0% 0.02 0.17 0.01 Namibia 4.8% 24.0% 0.0% 0.3% 4.1% 0.0% 1.00 1.00 0.01 Nauru 1.2% 1.0% 1.0% 0.4% 0.4% 0.4% 0.01 0.01 0.01 Nepal 14.8% 34.0% 15.0% 6.2% 32.8% 6.1% 0.95 0.95 0.01 Netherlands 2.3% 6.0% 1.0% 1.0% 2.8% 0.3% 0.74 1.00 0.01 Netherlands Antilles 0.6% 1.0% 1.0% 0.1% 0.1% 0.1% 1.00 1.00 1.00 New Caledonia 0.7% 1.0% 0.0% 0.2% 0.4% 0.1% 0.01 0.01 0.01 New Zealand 1.2% 3.0% 0.0% 0.7% 2.1% 0.1% 0.36 0.98 0.01 Nicaragua 3.1% 4.0% 1.0% 1.4% 2.0% 0.3% 0.01 0.02 0.01 Niger 3.6% 7.0% 0.0% 0.8% 1.5% 0.0% 1.00 1.00 1.00 Nigeria 7.0% 16.0% 0.0% 3.6% 7.6% 0.1% 0.28 1.00 0.01 Niue 0.8% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Norfolk I. 0.4% 0.0% 0.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 North Korea 1.2% 9.0% 0.0% 0.6% 2.2% 0.1% 0.50 1.00 0.02 Northern Mariana Is. 0.9% 1.0% 1.0% 0.7% 0.7% 0.7% 0.01 0.01 0.01 Norway 2.6% 6.0% 1.0% 2.0% 4.4% 0.4% 0.23 1.00 0.01 Oman 0.1% 0.0% 0.0% 0.1% 0.2% 0.0% 1.00 1.00 1.00 Pakistan 4.7% 16.0% 0.0% 1.8% 8.3% 0.0% 0.97 1.00 0.58 Palau 1.6% 2.0% 2.0% 0.8% 0.8% 0.8% 0.01 0.01 0.01 Panama 3.6% 24.0% 2.0% 2.1% 19.4% 1.0% 0.01 0.01 0.01 Papua New Guinea 6.1% 18.0% 1.0% 3.9% 12.0% 0.0% 0.01 0.01 0.01

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin Paraguay 16.6% 19.0% 4.0% 3.3% 5.4% 0.2% 0.01 0.01 0.01 Peru 7.8% 29.0% 0.0% 2.1% 16.3% 0.0% 0.73 1.00 0.01 Philippines 3.8% 8.0% 1.0% 2.2% 4.1% 0.4% 0.02 0.79 0.01 Poland 5.3% 14.0% 0.0% 1.4% 4.9% 0.2% 0.04 0.18 0.01 Portugal 1.9% 3.0% 0.0% 0.7% 1.0% 0.1% 0.55 1.00 0.01 Puerto Rico 0.8% 1.0% 1.0% 0.2% 0.3% 0.1% 0.08 1.00 0.01 Qatar 0.1% 0.0% 0.0% 0.0% 0.1% 0.0% 1.00 1.00 1.00 Reunion 0.5% 0.0% 0.0% 0.1% 0.2% 0.1% 0.15 0.32 0.01 Romania 13.4% 14.0% 0.0% 4.7% 4.9% 0.0% 0.21 1.00 0.01 Russia 6.8% 18.0% 0.0% 2.4% 10.7% 0.0% 0.08 1.00 0.01 Rwanda 23.6% 25.0% 15.0% 3.7% 3.9% 2.3% 0.01 0.01 0.01 Samoa 1.6% 2.0% 1.0% 0.7% 0.9% 0.7% 0.01 0.01 0.01 San Marino 2.1% 2.0% 1.0% 1.1% 1.1% 0.7% 0.24 0.82 0.24 Sao Tome & Principe 2.0% 2.0% 2.0% 1.1% 1.1% 1.1% 0.02 0.02 0.01 Saudi Arabia 1.7% 3.0% 0.0% 0.7% 1.3% 0.0% 1.00 1.00 1.00 Senegal 3.5% 4.0% 0.0% 0.7% 1.4% 0.0% 0.93 1.00 0.01 Serbia 12.7% 14.0% 2.0% 4.6% 4.9% 0.9% 0.18 0.93 0.02 Seychelles 1.1% 1.0% 1.0% 0.4% 0.4% 0.4% 0.01 0.01 0.01 Sierra Leone 5.7% 7.0% 3.0% 4.3% 4.8% 1.5% 0.01 1.00 0.01 Singapore 1.9% 2.0% 2.0% 0.7% 0.7% 0.7% 0.08 0.08 0.08 Slovakia 13.7% 14.0% 6.0% 4.8% 4.9% 1.6% 0.18 0.18 0.03 Slovenia 13.3% 14.0% 1.0% 4.7% 4.9% 0.4% 0.17 0.18 0.01 Solomon Is. 2.1% 2.0% 2.0% 1.0% 1.5% 0.0% 0.01 0.01 0.01 Somalia 2.4% 10.0% 0.0% 0.3% 1.4% 0.0% 1.00 1.00 1.00 South Africa 4.6% 9.0% 0.0% 0.5% 0.9% 0.0% 0.83 1.00 0.01 South Georgia & the South Sandwich Is. 1.0% 1.0% 1.0% 0.7% 0.8% 0.7% 0.01 0.01 0.01 South Korea 1.0% 2.0% 0.0% 0.5% 0.8% 0.1% 0.42 1.00 0.01 Spain 2.0% 3.0% 0.0% 0.6% 1.5% 0.0% 0.72 1.00 0.01 Sri Lanka 1.4% 2.0% 1.0% 0.6% 0.9% 0.1% 0.45 1.00 0.01 St. Helena 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 1.00 1.00 1.00 St. Kitts & Nevis 0.6% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 St. Lucia 0.6% 1.0% 1.0% 0.3% 0.3% 0.3% 0.02 0.02 0.01 St. Pierre & Miquelon 0.6% 1.0% 1.0% 0.4% 0.4% 0.3% 0.01 0.01 0.01 St. Vincent & the Grenadines 0.6% 1.0% 1.0% 0.2% 0.2% 0.2% 0.01 0.01 0.01 Sudan 2.9% 27.0% 0.0% 0.5% 7.2% 0.0% 0.89 1.00 0.01 Suriname 3.9% 15.0% 1.0% 1.5% 12.6% 0.4% 0.01 0.01 0.01 Swaziland 3.3% 3.0% 3.0% 0.7% 0.7% 0.5% 0.06 0.10 0.06 Sweden 2.0% 6.0% 0.0% 0.9% 4.4% 0.1% 0.15 0.98 0.01 Switzerland 6.2% 14.0% 4.0% 3.3% 4.9% 2.1% 0.36 0.78 0.04 Syria 2.0% 4.0% 0.0% 0.9% 2.0% 0.1% 0.94 1.00 0.65 Tajikistan 4.5% 16.0% 1.0% 2.1% 8.3% 0.3% 1.00 1.00 0.01 Tanzania 10.5% 25.0% 1.0% 1.8% 4.3% 0.1% 0.20 1.00 0.01

Basin internal evaporation Hydrologically effective basin Water depletion index recycling internal evaporation recycling Country BIER BIER hydrol- BIERhydrol- BIERhydrol- BIERmax BIERmin eff eff,max eff,min WDI WDImax WDImin Thailand 13.0% 23.0% 1.0% 4.8% 12.8% 0.2% 0.05 0.75 0.01 The Bahamas 0.6% 1.0% 1.0% 0.1% 0.4% 0.1% 0.02 1.00 0.01 The Gambia 2.6% 3.0% 1.0% 0.5% 0.5% 0.1% 0.95 1.00 0.01 Timor-Leste 1.1% 2.0% 1.0% 0.3% 0.5% 0.0% 0.02 0.21 0.01 Togo 5.0% 8.0% 3.0% 1.2% 1.5% 1.0% 0.01 0.01 0.01 Tokelau 1.5% 2.0% 2.0% 0.6% 0.6% 0.6% 0.01 0.01 0.01 Tonga 0.6% 1.0% 1.0% 0.1% 0.1% 0.1% 0.01 0.01 0.01 Trinidad & Tobago 1.0% 1.0% 1.0% 0.2% 0.3% 0.2% 0.01 0.02 0.01 Tunisia 0.8% 2.0% 0.0% 0.2% 0.8% 0.1% 0.94 1.00 0.03 Turkey 2.8% 8.0% 0.0% 1.1% 3.7% 0.2% 0.72 1.00 0.01 Turkmenistan 2.6% 5.0% 0.0% 1.0% 2.2% 0.0% 1.00 1.00 1.00 Turks & Caicos Is. 0.6% 1.0% 1.0% 0.1% 0.1% 0.1% 0.01 0.01 0.01 Tuvalu 1.2% 1.0% 1.0% 0.5% 0.6% 0.5% 0.01 0.01 0.01 Uganda 24.7% 38.0% 4.0% 4.0% 10.9% 0.5% 0.01 1.00 0.01 Ukraine 5.3% 14.0% 0.0% 1.2% 4.9% 0.1% 0.46 1.00 0.01 United Arab Emirates 0.1% 1.0% 0.0% 0.1% 0.4% 0.0% 1.00 1.00 1.00 United Kingdom 1.2% 2.0% 0.0% 0.6% 1.6% 0.0% 0.28 1.00 0.01 United States 4.3% 20.0% 0.0% 1.3% 16.6% 0.0% 0.55 1.00 0.01 Uruguay 5.8% 9.0% 0.0% 2.4% 2.8% 0.1% 0.01 0.16 0.01 Uzbekistan 2.4% 5.0% 0.0% 0.9% 2.2% 0.0% 1.00 1.00 1.00 Vanuatu 1.1% 1.0% 1.0% 0.4% 0.6% 0.0% 0.01 0.01 0.01 Vatican City 2.1% 2.0% 2.0% 1.1% 1.1% 1.1% 0.24 0.24 0.24 Venezuela 8.8% 25.0% 0.0% 4.3% 17.1% 0.0% 0.39 1.00 0.01 Vietnam 5.7% 23.0% 1.0% 2.7% 10.2% 0.0% 0.28 1.00 0.01 Virgin Is. 0.8% 1.0% 1.0% 0.3% 0.3% 0.3% 0.01 0.01 0.01 Wallis & Futuna 1.4% 1.0% 1.0% 0.6% 0.6% 0.6% 0.01 0.01 0.01 West Bank 0.4% 1.0% 0.0% 0.3% 0.3% 0.1% 1.00 1.00 1.00 Western Sahara 0.4% 1.0% 0.0% 0.2% 0.2% 0.0% 1.00 1.00 1.00 Yemen 1.4% 2.0% 0.0% 0.4% 0.8% 0.0% 1.00 1.00 1.00 Zambia 23.8% 35.0% 9.0% 4.1% 7.0% 2.3% 0.01 0.01 0.01 Zimbabwe 11.9% 24.0% 4.0% 1.8% 4.1% 0.9% 0.74 1.00 0.01