The water footprint of agricultural products in European river basins

This work quantifies the agricultural water footprint (WF) of production (WFprod, agr) and consumption (WFcons, agr) and the resulting net virtual water import (netVWi, agr) of 365 European river basins for a reference period (REF, 1996–2005) and two diet scenarios (a healthy diet based upon food-based dietary guidelines (HEALTHY) and a vegetarian (VEG) diet). In addition to total (tot) amounts, a differentiation is also made between the green (gn), blue (bl) and grey (gy) components. River basins where the REF WFcons, agr, tot exceeds the WFprod, agr, tot (resulting in positive netVWi, agr, tot values), are found along the London–Milan axis. These include the Thames, Scheldt, Meuse, Seine, Rhine and Po basins. River basins where the WFprod, agr, tot exceeds the WFcons, agr, tot are found in Western France, the Iberian Peninsula and the Baltic region. These include the Loire, Ebro and Nemunas basins. Under the HEALTHY diet scenario, the WFcons, agr, tot of most river basins decreases (max −32%), although it was found to increase in some basins in northern and eastern Europe. This results in 22 river basins, including the Danube, shifting from being net VW importers to being net VW exporters. A reduction (max −46%) in WFcons, agr, tot is observed for all but one river basin under the VEG diet scenario. In total, 50 river basins shift from being net VW importers to being net exporters, including the Danube, Seine, Rhone and Elbe basins. Similar observations are made when only the gn + bl and gn components are assessed. When analysing only the bl component, a different river basin pattern is observed.


Introduction
The water footprint (WF, table 1) concept has been brought into water management science in order to show the importance of consumption patterns and the global dimensions in good water governance (Galli et al 2012, Vanham and Bidoglio 2013a. It is an indicator of direct and indirect water use. An assessment of the WF of all nations has recently been conducted by . It is important to distinguish between the WF of production (WF prod ) and the WF of consumption (WF cons ) of a geographical region. The first refers to the total use of domestic water resources within the region (for producing goods and services for either domestic consumption or for export). The second refers to the use of domestic and foreign water resources behind all goods and services that are consumed domestically. A balance between the two is reached by virtual water (VW) flows (imports (VW i ) and exports (VW e )), and more particularly by the netVW i (VW i minus VW e ), which equals WF cons minus WF prod . WFs consist of blue, green and grey components. Following the definition of Rockström et al (2009), green water is the soil water held in the unsaturated zone, formed by precipitation and available to plants, while blue water refers to liquid water in rivers, lakes, wetlands and aquifers. Irrigated agriculture receives blue water (from irrigation) as well as green water (from precipitation), while rainfed agriculture only receives green water. The green WF is thus the rainwater consumed by crops. The grey WF is an indicator of the degree of water pollution (Hoekstra et al 2011).
This paper conducts the following analyses for the river basins (partly) located in the EU and remaining Balkan countries (basin size larger than 1000 km 2 ) (figure 1): the reference (REF) WF of production for agricultural products (WF prod, agr ), the reference WF of consumption for agricultural products (WF cons, agr ) and the resulting reference net VW import for agricultural products (netVW i, agr ). Additionally the WF cons, agr for two diet scenarios is analysed: a healthy diet (HEALTHY) and a vegetarian diet (VEG). The resulting change in netVW i, agr is also assessed for all river basins. All these analyses are carried out for the total WF (WF prod, agr, tot , WF cons, agr, tot , netVW i, agr, tot ), the green WF (WF prod, agr, gn , WF cons, agr, gn , netVW i, agr, gn ), the blue WF (WF prod, agr, bl , WF cons, agr, bl , netVW i, agr, bl ) and the green plus blue WF (WF prod, agr, gn + bl , WF cons, agr, gn + bl , netVW i, agr, gn + bl ).
To date, the WF prod, agr has been assessed for some selected large river basins (Aldaya and Llamas 2009, Dumont et al 2013, Zeng et al 2012, Zhao et al 2010, but not covering the whole EU like in this study. The WF cons, agr has only been assessed in the literature on a national basis but not for river basins. The netVW i, agr, tot has been analysed in Vanham (2013a) for EU river basins. However, the author acknowledges that for policy options there is a need to compute the different components (green, blue and grey), which was not done in that study. The effect of diets on the WF cons has been carried out for the city of Milan (Vanham and Bidoglio 2014), China (Liu and Savenije 2008), Austria (Vanham 2013b), the EU as a whole (Vanham et al 2013b) and for four EU zones (Vanham et al 2013a). In the latter study, the importance of a regional analysis is stressed. The effect of different diets on the WF cons, agr (and the resulting change in netVW i, agr ) of a river basin has not been carried out to date. River basins, not administrative borders, are the key geographical entities for water management. The main instrument for the implementation of the European Water Framework Directive is the River Basin Management Plan (EC 2012). This study presents a first comprehensive agricultural WF accounting analysis of 365 European river basins, including the effect of various diet scenarios.
Environ. Res. Lett. 9 (2014) 064007 D Vanham and G Bidoglio Figure 1 River basins (size larger than 1000 km 2 ) that are located (partly) in the EU and remaining Balkan countries for which the assessment was carried out. The names of major basins are depicted. Indeed, in the framework of the global water-foodenergy-ecosystem nexus, the analysis conducted in this paper gives essential information.

General
The methodology to compute WF prod, agr and WF cons, agr (and resulting netVW i, agr ) river basin values based upon national values (as obtained from Hoekstra and Mekonnen (2012a)) and a WF prod, agr geodataset for crops (WF prod, agr-crops ) is presented in figure 2. To assess the effect of the two diets (HEALTHY and VEG) on the national WF cons, agr , the approach of Vanham et al (2013a) is applied. The most detailed geographical level for such an analysis is the national level, due to data restrictions. In this paper such an analysis is conducted for each of the 40 nations separately-based upon regional Food-Based Dietary Guidelines (FBDG) for HEALTHY as presented in table A.1-and results then transposed to river basin level. To date, such a detailed national assessment with respect to these 40 nations has only been conducted for Austria (Vanham 2013b). For the VEG, all meat of the healthy diet is substituted by an increase in the intake of products from the group pulses, nuts and oilcrops with equal caloric value and protein content.

River basins
The catchment database for continental Europe (CCM2), developed by Vogt et al (2007) (based on the digital elevation model SRTM-Shuttle Radar Topography Mission-of 90 m resolution), was used to identify the river basins (figure 1). Selected basins have to fulfil two conditions: (1) they are fully or partly located in the EU28 and remaining Balkan countries; (2) they have a surface area larger than 1000 km 2 .

Accounting framework
We follow the Global Water Footprint Standard developed by the Water Footprint Network (Hoekstra et al 2011). National data on the green, blue and grey WF cons, agr for each nation are obtained from  and Mekonnen and Hoekstra (2011b). The WF cons of agricultural products is calculated with the bottom-up approach, based upon direct underlying national data on consumption from FAO food balance sheets (FBS) FAO (2014). Three geodatasets (GIS-rasters) for the green, blue and grey WF prod, agr for crops (WF prod, agr-crops )(Mekonnen and Hoekstra 2011a, Mekonnen and Hoekstra 2010) with a 5 arc minute spatial resolution were obtained from the Water Footprint Network. These geodatasets are calculated based upon the crop growing areas (on a 5 by 5 arc minute grid cell resolution) from Monfreda et al (2008) and Portmann et al (2010). These geodatasets also comprise the crops which are used as feed. National blue (service water) and green (grazing) WF prod, agr data for livestock (WF prod, agr-liv ) were obtained from Mekonnen and Hoekstra (2012) and Mekonnen and Hoekstra (2011b). The period for which the analyses were made is 1996-2005. This period is identified as REF within this study.

River basin WF and VW values
The methodology to compute river basin values based upon raster values (for the green, blue and grey WF prod, agr-crops ) and national values (for the green, blue and grey WF prod, agr-liv and WF cons, agr ) is presented in figure 2. The application of this methodology is presented in online supplementary figure A.1, available at stacks.iop.org/ERL/00/000000/mmedia for green water and in figure A.2 for blue water. To assess a green and blue WF prod, agr-liv raster, national WF prod, agr-liv data for grazing and livestock service water were extrapolated by means of the gridded livestock of the world (GLW) rasters (with a 1 km spatial resolution (year 2000)) for different livestock types FAO (2013). The group horses, donkeys and mules is not represented by a GLW raster. To spatially distribute this livestock type, national stock data FAO (2014) are interpolated by means of the cattle GLW raster. Raster geodata of the green, blue and grey WF cons, agr are obtained by multiplying national WF cons, agr values Mekonnen 2012a, Mekonnen andHoekstra 2011b) with the population raster of CIESIN (2005).
Environ. Res. Lett. 9 (2014) 064007 D Vanham and G Bidoglio Figure 2 Workflow of the methodology used in the paper.

Diets
In this study three diets-the current diet (REF, 1996(REF, -2005, a healthy diet (HEALTHY) based on regional FBDG ( In this paper, a diet scenario analysis and its effect on the WF cons, agr is conducted for each of the 40 nations separately based upon regional FBDG (figure 3). This assessment is much more detailed than in Vanham et al (2013a), where such an analysis was only done for four aggregated EU zones. The amounts of fish recommended by the respective FBDG are substituted by meat. The reason for this is that no WF data for fish have been published thus far. Like in Vanham et al (2013a), a VEG includes the intake of milk and milk products (cheese, butter, yoghurt, etc). All meat is substituted by the group pulses, nuts and oilcrops, by an increase in the intake of pulses and soybeans. National data on food consumption (period 1996-2005)-on which basis the WF cons is calculated -were taken from the FAO FBS (FAO 2014 Figure 4 shows that river basins where the WF cons, agr, tot exceeds the WF prod, agr, tot substantially (resulting in positive netVW i, agr, tot values), are found along the densely populated and heavily industrialized London-Milan axis. Major river basins include the Thames (WF prod, agr, tot = 130 363 m 3 km −2 , WF cons, agr, tot = 1025 948 m 3 km −2 , netVW i, agr, tot = 895 585 m 3 km −2 ), Scheldt (WF prod, agr, tot = 200 524 m 3 km −2 , WF cons, agr, tot = 704 998 m 3 km −2 , netVW i, agr, tot = 504 474 m 3 km −2 ), Rhine (WF prod, agr, tot = 109 720 m 3 km −2 , WF cons, agr, tot = 369 261 m 3 km −2 , netVW i, agr, tot = 259 541 m 3 km −2 ) and Po basins (WF prod, agr, tot = 219 630 m 3 km −2 , WF cons, agr, tot = 465 324 m 3 km −2 , netVW i, agr, tot = 245 694 m 3 km −2 ). Other such basins include the Tajo basin (which encompasses the city of Madrid) or small urban river basins like the Besòs river basin in which a large part of the city of Barcelona is located.

Net VW i, agr for the diet scenarios
The changes in WF cons, agr for the diet scenarios also lead to changes in the netVW i, agr of river basins. Figure 6 shows these changes for netVW i, agr, tot and netVW i, agr, gn + bl . For the netVW i, agr, tot , HEALTHY results in a shift from netVW i to netVW e for 22 of 365 river basins, amongst which the Danube basin (from 2243 m 3 km −2 to −9387 m 3 km −2 ) (table A.4). Two small basins shift from netVW e to netVW i . VEG results in a shift from netVW i to netVW e for 50 river basins,   Percentage reduction in WF cons, agr, tot (above) and WF cons, agr, gn + bl (below) within the river basins due to different diets (left HEALTHY; right VEG).

General
This paper shows that there are substantial differences in the amounts and characterizations of the current (REF) WF prod, agr , WF cons, agr and resulting netVW i, agr in European river basins. The diet scenarios show substantial shifts in the WF cons, agr , with resulting shifts in net VW i, agr amounts. From the perspective of FBDG, the WF cons, agr for both the HEALTHY and VEG scenarios can be regarded as being sustainable. From the water use perspective, the WF cons, agr can be regarded as being more sustainable under the VEG than under the HEALTHY scenario. Not assessed in this paper is the preferred consumption of local and seasonal food, Environ. Res. Lett. 9 (2014) 064007 D Vanham and G Bidoglio Figure 6 Shift in netVW i, agr, tot /netVW e, agr, tot (above) and netVW i, agr, gn + bl /netVW e, agr, gn + bl (below) within the river basins due to different diets (left HEALTHY; right VEG).
which can have some additional effect on the quantity and composition (green, blue, grey) of the WF cons, agr . As a next step, the sustainability of the current river basin WF prod, agr should be assessed, with the relevant blue, green and grey WF sustainability indicators (Vanham andBidoglio 2013a, Hoekstra et al 2011). Already today several of the analysed river basins experience (blue) water stress . The maximum sustainable WF prod, agr should be addressed per river basin. This can mean that the reference WF prod, agr of some river basins will have to be reduced while that of others can still be increased. Different pathways to achieving such an outcome are available. Sustainable agricultural intensification is identified as the way forward by different authors (Foley et al 2011, Godfray et al 2010, Tilman et al 2011. Yield gaps need to be closed on underperforming lands, while also reducing the environmental impacts of agriculture. If such a maximum sustainable WF prod, agr as well as the WF cons, agr of a HEALTHY and/or VEG diet were to be implemented, it is anticipated that a sustainable situation from a water resources point of view is reached. Additionally other resources/indicators such as land use and GHG emissions need to be assessed. Only by evaluating different indicators, integrated policy options can be defined.

Uncertainty in data and methodology
Different assumptions and simplifications were made in this study. Due to data availability restrictions, the results of this assessment are not absolute and must be regarded as best estimates based upon direct underlying data on production (for the WF prod , agr ) and consumption (for the WF cons , agr ). Both can be calculated by means of the top-down or bottomup approach (Hoekstra et al 2011). In this Letter, both are calculated with the bottom-up which is based upon direct underlying data on production and consumption-FAOSTAT data (FAO 2014)-and is less sensitive to trade data than the top-down approach . As described in Vanham and Bidoglio (2013b), the balance WF prod + VW i = WF cons + VW e within the geographic WF accounting scheme as calculated for the EU with the bottomup approach does not hold 100%. For the EU28, for agricultural products, these values are 552 km 3 (WF prod, agr ), 360 km 3 (VW i, agr ), 759 km 3 (WF cons, agr ) and 95 km 3 (VW e, agr ) (Vanham and Bidoglio, 2013b). The balance therefore shifts between 855 and 912 km 3 (range of about 6%). Theoretically this balance should close. This is however not the case due to practical complexities with data (availability of and inconsistencies in the underlying databases). As such the results of this assessment need to be regarded as best estimates.
The assumptions related to the methodology used to compute river basin values (figure 2, figure A.2 and figure A.3) were discussed in detail in Vanham (2013a). An important assumption is the fact that the WF cons, agr raster is obtained by spatially disaggregating national WF cons, agr data by means of a population raster dataset. This assumes that the consumption pattern is homogenous within each country. This is in reality not the case and thus a simplification. To account for this heterogeneity, regional statistics within a country could be used. Such detailed consumption data are however currently lacking (Hoff et al 2014, Vanham andBidoglio 2014).
The grey WF methodology needs to be further standardized (Vanham and Bidoglio 2013a), therefore WF prod, agr, tot and WF cons, agr, tot results were additionally shown without the grey component (WF prod, agr, gn + bl and WF cons, agr, gn + bl ).
Methodology assumptions and data availability for computing the WF cons, agr of the different diet scenarios were discussed in detail in Vanham et al (2013a). For the diets of the different zones e.g., average values were chosen from selected FBDG (table A.1), although recommendations for specific product groups often indicate a range of intake. Correction factors to compute intake values from consumption data are based upon a list of publications but are not available on a zonal/regional level (Vanham et al 2013a). An important issue is also that, although regional FBDG include fish, WF values for fish (and seafood) have not yet been published. In our analyses, these recommended amounts were substituted by meat. The WF cons , agr for the HEALTHY and VEG diets thus include the protein and energy intake of fish (substituted by meat), but the WF cons , agr calculated under the REF diet scenario does not represent the current intake of fish (and seafood) at all. Figure A.6 shows the REF and recommended intake values of meat (including offals) and fish (including seafood) for the 40 countries. Indeed, substantial fish (including seafood) amounts are part of the REF diet in all zones. By not incorporating these values, the current WF cons , agr is in fact underestimated. Especially for different countries in the FBDG zone NORTH (Finland, Lithuania, Norway, Sweden), the recommended intake of meat including fish (49.3 kg per cap per year) is exceeded when including, but not reached when excluding, the latter. This explains higher HEALTHY WF cons , agr values as compared to (underestimated) REF WF cons , agr values. Figure A.7 shows that during the past decades the intake of meat (including offals) and fish (including seafood) has evolved to some extent. As compared to REF (1996REF ( -2005, countries with very high intakes (France, Spain) have decreased their intake whereas many with low intake values show a steady increase.

Conclusions
This study presents a comprehensive quantification and analysis of the WF prod, agr , the WF cons, agr and the resulting netVW i, agr for 365 EU river basins for a REF (1996REF ( -2005 and two diet scenarios (HEALTHY and VEG). The analysis differentiates between the different green, blue and grey WF components. Such a comprehensive analysis on the river basin scale is the first in its kind.
Substantial differences in amounts and characterizations of the river basins' REF WF prod, agr , WF cons, agr and resulting net VW i, agr are observed. The highly industrialized and densely populated river basins along the London-Milan axis show higher WF cons, agr than WF prod, agr amounts (for tot, gn + bl and gn), identifying them as net VW importer basins. These include major basins such as the Thames, Scheldt, Meuse, Seine, Rhine and Po basins. River basins where the WF prod, agr, tot exceeds WF cons, agr, tot amounts (net VW exporter basins)(for tot, gn + bl and gn) are found in western France (Loire, Garonne), the Iberian Peninsula (Ebro, Duero, Guadiana, Guadalquivir) and the greater Baltic region (Nemunas). These basins are sparsely populated with extensive agricultural areas. A different pattern is observed for the river basins when only the blue WF component is analysed. High WF prod, agr, bl values are concentrated in the Mediterranean region (Guadalquivir, Ebro) due to irrigated agriculture and to a lesser extent in the Benelux region (Scheldt) due to livestock service water. Blue net VW exporter basins are concentrated in the Mediterranean.
The different diet scenarios lead to substantial shifts in the WF cons, agr , with resulting shifts in net VW i, agr amounts. For the HEALTHY scenario, the WF cons, agr, tot of most river basins decreases with exceptions in northern and eastern Europe. High decreases are observed in e.g. the Loire (−30.5%) and Po (−31.7%) basins. As a result, 22 of the 365 river basins shift from being net VW importers to being net VW exporters. For the VEG scenario, a reduction is observed in 364 of 365 basins. High reductions are observed in e.g the Loire (−46.4%), Rhone (−44.6%), Scheldt (−44.6%) or Po (−44.3%) basins. This results in a shift of 50 river basins from being net VW importers to being net VW exporters. Similar observations are made for the WF cons, agr, gn + bl (and netV-W agr, gn + bl ) and for the WF cons, agr, gn (and netVW agr, gn ). With regard to the blue WF component, 15 basins (including the Loire) shift from being net VW importers to being net VW exporters under the HEALTHY scenario -and 33 basins (including the Loire and Tajo) under the VEG scenario.
Such reduced river basin WF cons, agr can contribute to sustainable water management both within the EU and beyond its borders. They could help to reduce the dependency of EU consumption on domestic and foreign water resources or even increase virtual water exports from the EU to other regions, thereby contributing to the mitigation of the growing water scarcity in other parts of the world (Vanham et al 2013b). As global land and water resources are finite, adaptations in both production and consumption need to be made.