Agricultural Systems 125 (2014) 33–41

Contents lists available at ScienceDirect

Agricultural Systems

journal homepage: www.elsevier .com/locate /agsy

Short Communication

WaterFootprint on AgroClimate: A dynamic, web-based toolfor comparing agricultural systems

0308-521X/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.agsy.2013.11.006

⇑ Corresponding author. Tel.: +1 352 392 1864x220.E-mail address: ddourte@ufl.edu (D.R. Dourte).

Daniel R. Dourte ⇑, Clyde W. Fraisse, Oxana UryasevUniversity of Florida, Agricultural and Biological Engineering, 1741 Museum Road, PO Box 110570, Gainesville, FL 32611, United States

a r t i c l e i n f o

Article history:Received 4 April 2013Received in revised form 6 September 2013Accepted 24 November 2013Available online 28 December 2013

Keywords:Water footprintSustainabilityManagementEfficiencyWater stress

a b s t r a c t

We introduce a dynamic web-resource for estimating seasonal water footprints of agricultural produc-tion in the U.S. This tool provides a system-specific water footprint accounting that responds to changesin location, time, soil, and management. Modifications to an existing crop growth model were made inorder to separate consumptive use of green and blue water; that is water from rainfall and water froma groundwater or surface water resource, respectively. This separation is an important distinction ofwater footprinting that allows for more direct assessments of impacts on water resources. The tool alsoprovides a local water stress index, based on regional water use and available supplies, and it displaystime series and cumulative rainfall during the period of crop production.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

A water footprint is a comprehensive measure of freshwaterconsumption that connects water use to a certain place, manage-ment system, time, and type of water resource. It is distinct fromthe common measure of water use, water withdrawals, because awater footprint only accounts for consumptive water use, whichis water that becomes unavailable for reuse locally in the shortterm due to evaporation, incorporation into products, or a substan-tial quality decline. It has been used to evaluate the impacts of spe-cific consumption and production practices on freshwater qualityand sustainability. A water footprint can be calculated for a kilo-gram of wheat (Mekonnen and Hoekstra, 2010a), a jar of pastasauce (Ridoutt and Pfister, 2010), a pizza (Aldaya and Hoekstra,2010), a kilogram of milk (Zonderland-Thomassen and Ledgard,2012), a barrel of oil (Dominguez-Faus et al., 2009), a pair of jeans(Chapagain et al., 2006), or for any kind of product, land area, con-sumer, business, or nation by following accounting practices thathave been largely standardized by the Water Footprint Network(Hoekstra et al., 2011).

The calculation of a water footprint of a product can include thetotal amount of freshwater consumed along the supply chain of aproduct. An aspect of water footprinting that makes it especiallyvaluable for sustainability and impact assessments is theseparation of freshwater consumption into: (1) green water use,

which is consumption from rainfall; (2) blue water use, which isconsumption from groundwater or surface water; and (3) greywater use, which can be described as the dilution water requiredto reduce contaminant concentrations to acceptable values. Itshould be noted that the common definition of grey water aswastewater that might be reused is redefined in the water foot-print literature as a theoretical volume of water required to reducecontaminant loads to a specified level. This distinction amonggreen, blue, and grey water footprints recognizes that theconsumptive use of rainfall, groundwater or surface water, andthe water quality impacts all have different economic costs andecological impacts.

Agriculture is responsible for 92% of the 9087 Gm3 annual waterfootprint of humanity (Hoekstra and Mekonnen, 2012), and it isestimated (Mekonnen and Hoekstra, 2010) that 78% of the globalagricultural water footprint is green (from rainfall), 12% is blue(irrigation from groundwater or surface water), and 10% is grey(freshwater contamination). Green, blue, and grey water foot-prints have been calculated for 126 crops and more than 200crop-derived products at a global scale; crop-specific water foot-prints have been reported at national and state-level scales usingweather data for 1996–2005 (Mekonnen and Hoekstra, 2010b).This extensive dataset provides excellent baseline estimates ofagricultural water footprints. However, management and climateare not stable with time, and tools to estimate the water footprintsof crop production for particular management systems and timescan be valuable for stakeholders interested in water resourcesustainability.

34 D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41

The WaterFootprint tool on AgroClimate (http://agroclimate.org/tools/Water-Footprint/), introduced here, is a user-friendlyweb-resource capable of simulating the system-specific greenand blue water footprints for crop production in the United States.The dynamic nature of the tool allows for the comparisons of waterfootprints for varying management practices in agriculture for dif-ferent seasons and regions in the U.S. The estimates from the toolcan be used in the context of evaluating freshwater availability,water stress impacts, the effective use of rainfall, and the signifi-cance of irrigation. Presently, there is much discussion andresearch concerning adaptation of agricultural systems to a chang-ing climate, but there are few metrics that can compare the resil-ience of different systems. A water footprint can serve as both ameasure of resource-use efficiency and also as a tool for evaluatingsustainability of freshwater systems. The objectives of this paperare to (1) introduce the AgroClimate WaterFootprint tool, (2) eval-uate performance of the tool, and (3) demonstrate potential appli-cations of the tool to show water footprint impacts from climatevariability, management, and production region.

2. Methods

2.1. Web user interface

WaterFootprint on AgroClimate.org was designed to quicklyand easily calculate the green and blue water footprints of anagricultural system for historical and recent seasons for a varietyof production systems. Grey water is not accounted for as thisrequires much more detailed management data, and does notrepresent actual freshwater consumption, as it is more of an indexfor water quality impacts. The tool is hosted and maintained byAgroClimate.org (Fraisse et al., 2006; Fraisse, 2012), a web-basedclimate information and decision-support system that includesseasonal climate outlooks, expected impacts of managementoptions for different crops and climate scenarios, and a wide vari-ety of interactive tools that can help producers monitor currentconditions and plan for the season ahead.

User inputs in WaterFootprint were streamlined in an effort tobalance management system detail with the time required to runthe tool. Presently, input data include (1) location, (2) crop, (3)planting and harvest dates, (4) yield as input or simulated, (5) soiltexture, root zone depth, organic matter, (6) tillage, (7) irrigationmanagement, and (8) fertilizer application. Location can be selectedby clicking on a map or by inputting the zip code nearest to the pro-duction system. Crop choice is made by drop-down menu from fivegroups: Cereals, Pasture and forage, Fiber, Legumes, Fruits and veg-etables, Sugar and stimulants. Currently, selection can be madefrom among 71 annual crops, but the addition of selected perennialspecies is under development. Dates of planting and harvest are in-put or selected from a calendar. Yield can be simulated by the toolor input; it is recommended that yield be input as this reducesuncertainty in the water footprint calculation, and it is expectedthat the yield for a system could be known if the other managementdata are known. Soil is described based on the HC27 generic/proto-typical soil profiles that have been used for global crop modelingapplications (Nelson et al., 2009). These simplified soil descriptionsgive three choices for texture (sandy, silty, clayey), three choices ofroot zone depth (60, 120, and 180 cm), and three choices of organicmatter content (1.4%, 1.0%, and 0.4% in the top layer of soil). Tillageoptions are organized by groups: conservation, subsoil, conven-tional, weed-control. Tillage implement and direction (straight-row or contour) are used in the tool to adjust runoff Curve Number(SCS-CN; USDA, 1954). Three options are available to describeirrigation management: rainfed, auto-irrigation, and manual irriga-tion. For the auto-irrigation option, the model initiates an irrigation

operation, filling the soil profile to field capacity, when the waterstress exceeds a user-specified threshold. For manual irrigation, aseasonal total irrigation depth and average frequency of irrigationare specified by the user. Fertilizer forms are available for thepossibility of adding functionality in the tool to account for waterquality impacts; currently fertilizer application is automated basedon plant nitrogen stress and a specified nitrogen stress factorthreshold.

2.2. Water balance and crop growth simulation

Simulations of crop growth and water balances in the Water-Footprint tool make use of the EPIC crop growth model (Environ-mental Policy Integrated Climate; Williams et al., 1989) withinthe framework of the SWAT hydrology model (Soil and WaterAssessment Tool; Arnold et al., 1998). EPIC uses a single, genericgrowth model to simulate all crops. Each crop is described in themodel using approximately 30 plant-specific parameters. Animplementation of EPIC was used for the AgroClimate WaterFoot-print tool because it has been shown to be useful for broad, spatialassessments of water footprinting (Liu et al., 2007). SWAT is a land-scape-scale hydrology model having physically-based processdescriptions; it was used in the WaterFootprint tool because it con-tains an updated, streamlined implementation of EPIC for plantgrowth and it has physically-based, well-established routines forwater balance modeling (Arnold et al., 1998; Gassman et al.,2007; Krysanvoa and Arnold, 2008) that provide some flexibilitybeyond what was available by using EPIC outside of the SWATstructure.

Calculating blue and green water footprints requires that totalET be separated into ET of rainfall (ETgreen) and ET of irrigation fromsurface or groundwater (ETblue). The WaterFootprint tool followsthe guidelines of Hoekstra et al. (2011) to make this separation.ETgreen is calculated as the minimum of effective rainfall (Peff)and ETtotal; Peff = P � DP � RO, where P is rainfall, DP is deep perco-lation, and RO is surface runoff. In irrigated systems, ETblue is thedifference between ETtotal and ETgreen: ETblue = ETtotal � ETgreen.These ET depths are converted to volumes (liters or gallons) anddivided by yields (kg/ha or lb/ac) to calculate the blue and greenwater footprints.

The WaterFootprint tool uses the water stress index (WSI;Pfister et al., 2009) to provide information on the local, relativescarcity of freshwater resources. WSI at the location of the produc-tion system being simulated is retrieved from a WSI database forthe U.S. WSI is a logistic function of WTA, the ratio of total annualfreshwater withdrawals to annual hydrological availability. WTA iscalculated from the WaterGAP2 global model that describes wateruse and annual availability, based on 1961–1990 climate period)for over 10,000 watersheds (Alcamo et al., 2003). WSI ranges from0.01 (low water stress) to 1 (high water stress), and a value of 0.5represents a threshold of severe water stress (Vorosmarty et al.,2000). Blue water equivalents or stress-weighted water footprints(Ridoutt and Pfister, 2010) can be calculated from WaterFootprinttool outputs by multiplying WSI and blue water footprint. Thestress-weighted water footprint allows for regional water footprintcomparisons that directly accounts for water scarcity.

2.3. Weather data

Daily weather data are required for the simulations of cropgrowth and hydrology. The most computationally expensiveservice of the WaterFootprint tool is the retrieval and formattingof daily weather data from a network of 5953 active U.S. weatherstations that are part of the Global Historical Climatology Network– Daily (GHCN-Daily; Menne et al., 2012). The WaterFootprint toolselects the closest station with adequate data (complete records of

D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41 35

maximum temperature, minimum temperature, and precipitation)based on the location of the production system being evaluated.Solar radiation, wind speed, and relative humidity are generatedusing the WXGEN weather generator implemented in SWAT(Sharpley and Williams, 1990), which is included in SWAT. Gener-ated radiation, wind speed, and relative humidity are based on adetailed climate database of 168 climate parameters for 1041 loca-tions in the contiguous U.S.; as with observed data, the locationused for the climate database is based on proximity to the produc-tion system. The WaterFootprint tool is currently functional in theUnited States, but it could be readily adapted for use in other re-gions by changing the routines for weather data retrieval.

2.4. WaterFootprint tool evaluation

Data from studies that reported measured evapotranspiration(ET) and crop yields were used for evaluation of the performanceof the WaterFootprint tool. The requirements for evaluation datawere: (1) measurements of ET (using lysimeters or EddyCovariance) and (2) measurements of crop yield and managementdata for locations in the U.S. These requirements resulted in rela-tively few experiments being available for evaluation. Consistentwith our own recommendation in using the WaterFootprint tool,reported yields were used as input in the WaterFootprint tool. Rootmean squared errors (RMSE), percent bias (PBIAS), and graphicalcomparisons were used to evaluate WaterFootprint toolperformance.

2.5. Climate and management impacts on water footprints

As a demonstration of an example application of the WaterFoot-print tool, the effects of seasonal climate variability on winterwheat water footprints were examined for two counties in Georgiaand in Alabama. The El Niño Southern Oscillation (ENSO) is theleading driver of seasonal and year-to-year climate variability inthe southeastern U.S. Impacts of ENSO on water footprints ofwinter wheat in these two counties were examined graphicallyby grouping simulated water footprints based on ENSO phase(ENSO year classification was based on the Oceanic Niño Index,the 3-month running mean of SST anomalies in the Niño 3.4 region– 5�N–5�S, 120�–170�W CPC 2012).

The two selected locations, Seminole county, GA and Houstoncounty AL are both top wheat producing counties in their states.For this demonstration, simulated water footprints in Alabamaand Georgia were for rainfed and irrigated systems, respectively.On average, about 6% of winter wheat acreage is irrigated inAlabama and about 40% is irrigated in Georgia (USDA, 2010). In

Fig. 1. (A) Observed and simulated water footprints of irrigated cotton (Howell et al., 20et al., 1996), and irrigated corn and soybeans in Nebraska (Suyker and Verma, 2009).

Georgia, the auto-irrigation option was selected in the WaterFoot-print tool with a threshold of 0.75. County-average yield data fromUSDA NASS QuickStats (NASS 2013; 1965–2007 in AL and 1969–2008 in GA) were used as input to the tool. Average planting andharvest dates of Extension production guides were used: planting20 November and harvesting 25 May in Houston county, AL;planting 15 November and harvesting 20 May in Seminole county,GA. Location was input by zip code based on farm locations havinghighest winter wheat subsidies for 1995–2011 (EWG, 2012):Gordon, AL – 36343 and Donalsonville, GA – 39845. Soils weredescribed in the tool as silty, medium root-zone depth, and med-ium soil carbon content based on soil descriptions from the statevariety trial reports at locations nearest to Gordon, AL and Donal-sonville, GA. Tillage was described as conventional, disk plow,straight row.

A second application of the WaterFootprint tool: to moredirectly demonstrate the impacts of differences in seasonal rainfall,the water footprints of irrigated and rainfed peanut production inMilton, Florida (Santa Rosa county) were calculated for abnormallydry and wet years, 2007 and 2008, respectively. Planting and har-vest dates used were 4 April and 28 September, respectively, inboth 2007 and 2008. Soils were described in the tool as silty,medium root-zone depth, and low soil carbon content based onsoil information from NRCS Web Soil Survey (Soil Survey Staff,2012). The auto-irrigation option was selected, with a thresholdof 0.75, for water footprint calculations of the irrigated system.County-average peanut yields for Santa Rosa county, a leadingpeanut producing area of Florida, were 3474 and 4555 kg/ha in2007 and 2008 (NASS 2013). Reported county yields aggregateboth irrigated and rainfed production and were used here tocompare to the averages of simulated yields across irrigated andrainfed systems.

A third application of the WaterFootprint tool was regionalwater footprints comparisons between Plains, Texas (Yoakumcounty, northwestern Texas) and Newton, Georgia (Baker county,southwestern Georgia). This was demonstrated by simulating irri-gated peanut production for summer seasons (June–July–August,JJA) that were near the 50th percentile and 90th percentile of driestsummers based on the 1981–2010 climate normals. The seasons ofnear-normal rainfall corresponded to the years of 2012 and 2005 inGeorgia and Texas, respectively, and the very dry seasons corre-sponded to the years of 1997 and 1990 in Georgia and Texas,respectively. Georgia is the leading peanut-producing state in thenation, and Texas is among the top five peanut-producing statesin the U.S. The counties of the farms used for this water footprintaccounting are top producing counties in their states. Irrigatedproduction was simulated to show blue water footprint impacts;

04); (B) observed and simulated water footprints of irrigated corn in Texas (Howell

Fig. 2. (A) Observed and simulated water footprints of rainfed corn in Virginia (Roygard et al., 2002) and soybean in Nebraska (Suyker and Verma, 2009); (B) observed andsimulated crop yields for all corn and soybean, both irrigated and rainfed, experiments used for water footprint evaluation.

Fig. 3. Water footprints, right vertical axis, of rainfed winter wheat in El Niño andLa Niña years; Houston county, Alabama. Percent difference in total waterfootprints for El Niño and La Niña years, left vertical axis, based on difference fromall-years average (1965–2007). Yield input from USDA-NASS county yield data.

36 D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41

nearly 100% and 70% of peanut acreage is irrigated in Texas andGeorgia, respectively (USDA, 2010).

3. Results

Evaluation of the WaterFootprint tool showed good perfor-mance in comparisons with measured water footprint values(Figs. 1 and 2). Root mean squared errors (RMSE) of water footprintprediction were 106 l/kg for irrigated systems (corn, and soybean;n = 15) and 189 l/kg for rainfed systems (corn and soybean; n = 8).PBIAS for irrigated and rainfed water footprint predictions were�13% and 0.1%, respectively. A separate evaluation of yield predic-tion from the WaterFootprint tool shows an RMSE for predictedyields of 1501 kg/ha and 2111 kg/ha for irrigated and rainfed corn,respectively. Average observed yields from the data used were11,457 kg/ha for irrigated corn (n = 9 Texas and Nebraska) and6749 kg/ha for rainfed corn (n = 6, Virginia). PBIAS for irrigatedand rainfed corn yield predictions from the WaterFootprint toolwere 12% and 11%, respectively. These data suggest that the

WaterFootprint tool is simulating both crop yield and ET withacceptable accuracy. In this initial assessment, the simplified man-agement inputs and variable distance to weather stations seem tonot be causing substantial reductions in performance of the tool.Simulated water footprints of the irrigated systems used for evalu-ation shows a tendency of the WaterFootprint tool to underesti-mate water footprints compared to observations (Fig. 1); thismight be explained by the positive bias in yield prediction havingthe effect of reducing simulated water footprints. In the rainfedsystems used for evaluation, there was no clear bias in water foot-print estimation for corn and soybean combined (Fig. 2), but therewas less skill in yield predictions in rainfed systems, as indicatedby the larger RMSE of predicted corn yields.

In the first example application – to demonstrate climate vari-ability impacts – seasonal climate variability related to ENSOshowed clear impacts on winter wheat water footprints in the se-lected counties in Alabama and Georgia. For wheat in Houstoncounty, Alabama, a top wheat producing area in the state, waterfootprints were 21% greater (3053 l/kg) during La Niña years and12% lower (2236 l/kg) during El Niño years (Fig. 3) when comparedto the all-years average (1965–2007). In much of the southeasternU.S. during the winter and spring, El Niño typically brings coolerand wetter than average conditions while La Niña typically bringswarmer and drier than average conditions. Results of the Water-Footprint tool showed lower ET during El Niño years and higherET during La Niña years, suggesting that the higher temperaturesand solar radiation in La Niña winter/spring can result in more con-sumptive water use even while there may be less water availabledue to the lower-than-average rainfall. These results show thatthe water footprint impacts of ENSO are an amplified version ofthe more well-known signal of greater winter wheat yields in ElNiño and lower yields in La Niña for much of southern Alabamaand Georgia. The blue water footprint estimates of winter wheatproduction in Georgia show similar trends: blue water footprintsof this irrigated system were 31% greater (678 l/kg) during LaNiña years and 10% lower (465 l/kg) during El Niño years (Fig. 4).

As the second example application of the WaterFootprint tool,the management and climate impacts on water footprints of pea-nut production for dry and normal rainfall seasons are demon-strated in Fig. 5. For the area surrounding Milton, Florida, theaccumulated summer rainfall (JJA) for the region (Florida climatedivision: 1) in 2007 was in the 100th percentile, meaning it wasdrier than that of any JJA season in the 30-year period of climatenormals (1981–2010). JJA rainfall in 2008 was in the 34th percen-tile of JJA rainfall for the climate normals period, meaning it wasdrier than about a third of the JJA seasons in the climate normals

Fig. 4. Blue water footprints, right vertical axis, of winter wheat in El Niño and LaNiña years; Seminole county, Georgia. Percent difference in blue water footprintsfor El Niño and La Niña years, left vertical axis, based on difference from all-yearsaverage (1965–2007). Yield input from USDA-NASS county yield data. (Forinterpretation of the references to color in this figure legend, the reader is referredto the web version of this article.)

D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41 37

period. The water footprints of rainfed peanut production were1681 l/kg in 2007 and 1223 l/kg in 2008 (Milton, Florida; SantaRosa county). During the abnormally dry conditions of the 2007season, there was 37% more consumptive water use per unit of

Fig. 5. WaterFootprint tool results for rainfed and irrigated peanut production in Miltonblue water footprints and crop yields as labels in bar graphs, total and time-series rainfallfigure legend, the reader is referred to the web version of this article.)

rainfed peanut production in 2007 compared to the 2008 season,in which there was near normal rainfall. This illustrates thatalthough actual ET depths are reduced during dry years, the greaterreduction in yields can result in large water footprints in dry years(Fig. 5, Rainfed). Irrigated peanut production at the same locationalso shows a higher water footprint in the dry 2007 season. How-ever, having irrigation, which reduces the differences in both ETand yields between dry and normal years, resulted in only about8% more consumptive water use per unit of rainfed peanut produc-tion in 2007 compared to the 2008 (Fig. 5, Irrigated). In both 2007and 2008, the total water footprints of irrigated peanut productionare actually lower than that of the rainfed system; this results fromthe increased yields of irrigated production, and it highlights theimportance of separating blue and green water footprints. The bluewater footprints, those that can be used more directly for water re-source sustainability assessments, are only accounted for in irri-gated systems, and these are the water footprints that arguablyhave greater value for comparing water resource impacts acrosstime, region, or management system (Ridoutt and Pfister, 2010).Reported county yields, which for Florida combine irrigated andrainfed production, were 3474 and 4555 kg/ha in 2007 and 2008(NASS 2013, Santa Rosa county). The simulated yields from theWaterFootprint tool in Milton, FL were close to the county aver-ages. Yields averaged across irrigated and rainfed systems were2998 hg/ha and 4689 kg/ha in 2007 and 2008; this suggests theplant growth and yield simulations of the WaterFootprint toolwere performing well in these systems for these years.

The third example application of the tool examined the regionaldifferences in water footprints of irrigated peanut production inNewton, Georgia and Plains, Texas. Seasons in the 50th percentileand 90th percentile of driest summers were used for each location.Figs. 6 and 7 show lower total water footprints in the typically

, FL for dry and wet years, 2007 and 2008. Screenshots of outputs include green and, and regional water stress index. (For interpretation of the references to color in this

Fig. 6. WaterFootprint results for irrigated peanut production in Newton, Georgia in 2012 and Plains, Texas for 2005; these seasons rank close to the 50th percentile ofsummer rainfall based on 1981–2010 climate normals. Screenshots of outputs include green and blue water footprints in bar graph and table, yield, and total and time-seriesrainfall, and regional water-stress index. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

38 D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41

drier location in Texas due to less consumptive use of rainfall, butthe blue water footprints showed large differences between loca-tions. Assuming irrigation from groundwater resources, 54 l ofgroundwater were consumed per kilogram of peanut productionin Newton, Georgia, and 391 l of groundwater were consumedper kilogram of peanut production in Plains, Texas in the near-nor-mal rainfall seasons of 2012 and 2005 in Newton, Georgia andPlains, Texas, respectively (Fig. 6). In the dry seasons of 1997 (inNewton, Georgia) and 1990 (in Plains, Texas), it was estimated that422 l of groundwater were consumed per kilogram of peanut pro-duction in Newton, Georgia, and 585 l of groundwater were con-sumed per kilogram of peanut production in Plains, Texas(Fig. 7). These water footprint estimates in the dry years demon-strate how dry seasons or droughts can mask the typical regionalwater footprint differences between humid and dry regions. Thelarge WSI in the Texas location in this example suggest that thisis a region where there is substantial pressure on scarce freshwaterresources.

4. Discussion and conclusion

The WaterFootprint tool showed good performance in simulat-ing the water footprints of production systems in several locationsand seasons. Three example applications were illustrated, showingwater footprint impacts of: (1) climate variability, (2) irrigationmanagement, and (3) differences in production regions. The exam-ination of ENSO impacts on wheat water footprints in Alabama andGeorgia showed the temporal instability of water footprints in timeresulting from variability in climate that affected both yields andconsumptive water use. The example of peanut production in Mil-ton, Florida showed how a dry growing season can actually result

in larger water footprints when compared to seasons having morenormal rainfall due to the potentially large yield reductions in dryseasons. The effect on water footprints at the Florida location wasthat total water footprints were lower for irrigated production thanfor the rainfed system. This emphasized the importance of makingthe separation between green and blue water footprints, and it alsodemonstrated the interaction of the agronomic and hydrologic im-pacts of climate variability. Regional water footprint comparisonswere demonstrated by the water footprints of irrigated peanut pro-duction in Texas and Georgia. This showed that very different bluewater footprints can be expected between regions, even those withsimilar soils, management, and yields, due to seasonal rainfall dif-ferences, evaporative demands, and the corresponding impacts onirrigation requirements. Regional comparisons of crop productionwater footprints can be useful for evaluating water policies andirrigation restrictions or incentives.

The WaterFootprint tool on AgroClimate estimates the blue andgreen water footprints of specific agricultural production systems.The blue water footprints, and the regional water scarcity indica-tor, the water-stress index (WSI), to put them in context, can beused to compare the potential freshwater resource impacts of irri-gation between regions and management systems. Green wateruse, while not necessarily reducing the availability of freshwaterresources, is included in this tool so that all consumptive wateruse, either from surface or groundwater or rainfall, is accountedfor. Green water footprints can increase the information availablefor use in land use comparisons. For example, an agricultural sys-tem in some area means that the consumptive use of rainfall inthat same area cannot be used for another purpose, includinggrowing alternative crops or supporting the ecosystems and habi-tats of unmanaged lands. If the objective of water footprinting isestimating the impacts of agriculture on freshwater availability

Fig. 7. WaterFootprint results for irrigated peanut production in Newton, Georgia in 1997 and Plains, Texas for 1990; these seasons rank near the 90th percentile of lowestsummer rainfall based on 1981–2010 climate normals. Screenshots of the tool outputs include green and blue water footprints in bar graph and table, yield, and total andtime-series rainfall, and regional water-stress index. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41 39

for a specific location, then it is blue water footprints that shouldbe considered most carefully (Ridoutt and Pfister, 2010). Thereare numerous tools being developed and used to assess how agri-culture can adapt agriculture to a changing climate; by combiningthe impacts of management, location, and time on crop yield andwater use, water footprints can be used as one type of indicatorof the climate-resilience of a system.

The definition of a water footprint as consumptive water useper unit of production means that good cultural and water man-agement both can have the effect of reducing water footprints.Good pest management may not reduce consumptive water use,but it may reduce the water footprint of a system by increasingyields. Similarly, carefully-managed, efficient irrigation may notincrease crop yields, but the potentially lower consumptive wateruse would result in a reduced water footprint for the system. Forstakeholders among the agricultural supply chain, water footprint-ing can help in evaluating sustainability of irrigation waterresources or in comparing consumptive water use across manage-ment systems, regions, and time. Presently, most water footprint-ing in agriculture relies on the datasets of Mekonnen andHoekstra (2010b), but these data necessarily aggregate a varietyof production systems under averaged climate conditions, limitingtheir usefulness for evaluating water resource impacts of specificsystems under changing management and climate. The WaterFoot-print tool on AgroClimate advances water footprinting in agricul-ture by providing a free, web-based utility where interestedstakeholders can interactively develop water footprints estimatesof specific agricultural production systems in the U.S.

Acknowledgements

This tool was developed as part of an agricultural extensioninitiative: ‘‘Climate variability to climate change: Extension chal-lenges and opportunities in the Southeast USA,’’ and was supported

by Agriculture and Food Research Initiative Competitive Grant No.2011-67003-30347 from the United States Department of Agricul-ture (USDA) National Institute of Food and Agriculture (NIFA).

Appendix A

This appendix provides additional details about the processdescriptions used in the WaterFootprint tool of AgroClimate. Thefigure below summarizes the general structure and operation ofthe tool:

Simulating a crop water footprint using the WaterFootprint toolinvolves:

1. User input through the web-interface to describe the system,2. R program to collect input from the WUI, format all required

inputs, run EPIC-SWAT,3. SWAT simulation of growth and water balance processes,4. R program to read EPIC-SWAT output, calculate effective rain-

fall, ET from rainfall and ET from irrigation, do the unit conver-sions to provide water footprints (volume water/mass cropyield), write output for WUI,

5. WUI presents results to user.

The crop growth and hydrologic processes are described in de-tail in the SWAT documentation available at http://twri.tamu.edu/reports/2011/tr406.pdf. One of the key challenges in water foot-print accounting is separating green ET (consumptive water useof rainfall) and blue ET (consumptive water use of groundwateror surface water). Separate accounting for water use from rainfalland freshwater resources is also one of the biggest strengths ofwater footprinting. Making the separation requires a calculationof effective rainfall (Peff): the amount of rainfall that remains in acrop root zone and is available for evaporation and transpiration.

40 D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41

Green ET is then calculated, following the standards of the WaterFootprint Network (waterfootprint.org) as the minimum of ET total(simulated by SWAT) and Peff (calculated in the R routines fromdaily rainfall and daily SWAT outputs of runoff, RO, and deep per-colation, DP. RO is calculated within SWAT using the SCS CurveNumber approach (with Curve Number based on tillage implementand direction and crop type); deep percolation is also calculated inSWAT – whenever a soil layer’s water content exceeds field capac-ity, water drains from that layer as a function of saturated hydrau-lic conductivity, water content at field capacity and at saturation.The basic routines for the effective rainfall and green and blue ETcalculations are:

� Peff = P � RO � DP,� ETgreen = min(Peff, ETtotal),� ETblue = ETtotal � ETgreen; ETblue = 0 in rainfed systems.

Please contact the corresponding author if you would like to usethe code or if you would like to expand this dynamic water foot-print calculator to work in other regions of the world.

References

Alcamo, J., Doll, P., Henrichs, T., Kaspar, F., Lehner, B., Rösch, T., Siebert, S., 2003.Development and testing of the WaterGAP 2 global model of water use andavailability. Hydrol. Sci. J. 28 (3), 317–337.

Aldaya, M.M., Hoekstra, A.Y., 2010. The water needed for Italians to eat pasta andpizza. Agric. Syst. 103, 351–360.

Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large-area hydrologicmodeling and assessment: Part I. Model development. J. Am. Water Resour.Assoc. 34 (1), 73–89.

Chapagain, A.K., Hoekstra, A.Y., Savenije, H.H.G., Gautam, R., 2006. The waterfootprint of cotton consumption: an assessment of the impact of worldwideconsumption of cotton products on the water resources in the cotton producingcountries. Ecol. Econ. 60 (1), 186–203.

CPC (Climate Prediction Center, National Oceanic and Atmospheric Administration),2012. <http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml> (accessed 04.10.12).

Dominguez-Faus, R., Powers, S.E., Burken, J.G., Alvarez, P.J., 2009. The waterfootprint of biofuels: a drink or drive issue? Environ. Sci. Technol. 43 (9), 3005–3010.

EWG: Environmental Working Group, 2012. Houston County, AL: <http://farm.ewg.org/top_recips.php?fips=13253&progcode=wheat&regionname=SeminoleCounty>, Georgia and Seminole county, GA: <http://farm.ewg.org/top_recips.php?fips=01069&progcode=wheat&regionname=HoustonCounty,Alabama> (accessed10.09.12).

Fraisse, C.W., 2012. Agricultural Management Options for Climate Variability andChange: AgroClimate. University of Florida Institute of Food and AgriculturalSciences, Gainesville, FL, <http://edis.ifas.ufl.edu/ae491>.

Fraisse, C.W., Breuer, N., Zierden, D., Bellow, J.G., Paz, J.O., Cabrera, V., Garcia yGarcia, A., Ingram, K., Hatch, U., Hoogenboom, G., Jones, J.W., O’Brien, J., 2006.AgClimate: a climate forecast information system for agriculturalrisk management in the southeastern USA. Comput. Electron. Agric. 53 (1),13–27.

Gassman, P.W., Reyes, M.R., Green, C.H., Arnold, J.G., 2007. The soil and waterassessment tool: historical development, applications and future researchdirections. Trans. ASABE 50, 1211–1250.

Hoekstra, A.Y., Mekonnen, M.M., 2012. The water footprint of humanity. Proc. Natl.Acad. Sci. 109, 3232–3237.

Hoekstra, A.Y., Chapagain, A.K., Aldaya, M.M., Mekonnen, M.M., 2011. The WaterFootprint Assessment Manual: Setting the Global Standard. Earthscan, London.

Howell, T.A., Evett, S.R., Tolk, J.A., Schneider, A.D., Steiner, J.L., 1996.Evapotranspiration of corn—southern high plains. In: Cramp, C.R., Sadler, E.J.,Yoder, R.E. (Eds.), Proceedings of the International Conference inEvapotranspiration and Irrigation Scheduling, San Antonio Convention Center,San Antonio, Texas, November 3–6, 1996, pp. 158–165.

Howell, T.A., Evett, S.R., Tolk, J.A., Schneider, A.D., 2004. Evapotranspiration of full-,deficit-irrigated, and dryland cotton on the Northern Texas High Plains. J. Irrig.Drain. Eng. 130 (4), 277–285.

Krysanvoa, V., Arnold, J.G., 2008. Advances in ecohydrological modelling with SWAT– a review. Hydrol. Sci. 53, 939–947.

Liu, J., Williams, J.R., Zehnder, A.J.B., Yang, H., 2007. GEPIC – modelling wheat yieldand crop water productivity with high resolution on a global scale. Agric. Syst.94, 478–493.

Mekonnen, M.M., Hoekstra, A.Y., 2010. A global and high-resolution assessment ofthe green, blue and grey water footprint of wheat. Hydrol. Earth Syst. Sci. 14,1259–1276.

D.R. Dourte et al. / Agricultural Systems 125 (2014) 33–41 41

Mekonnen, M.M., Hoekstra, A.Y., 2010b. The green, blue and grey water footprint ofcrops and derived crop products, Value of Water Research Report Series No. 47,UNESCO-IHE, Delft, the Netherlands.

Menne, M.J., Durre, I., Vose, R.S., Gleason, B.E., Houston, T.G., 2012. An overview ofthe global historical climatology network-daily database. J. Atmos. OceanicTechnol. 29, 897–910.

NASS: National Agricultural Statistics Service, 2013. U.S. Department of Agriculture.Quick Stats – The National Agricultural Statistics Service Interactive. <http://quickstats.nass.usda.gov/> (accessed 03.29.13).

Nelson, G.C., Rosegrant, M.W., Koo, J., Robertson, R., Sulser, T., Zhu, T., Ringler, C.,2009. Climate Change: Impact on Agriculture and Costs of Adaptation.International Food Policy Research Institute, Washington, DC.

Pfister, S., Koehler, A., Hellweg, S., 2009. Assessing the environmental impactsof freshwater consumption in LCA. Environ. Sci. Technol. 43 (11), 4098–4104.

Ridoutt, B.G., Pfister, S., 2010. A revised approach to water footprinting to maketransparent the impacts of consumption and production on global freshwaterscarcity. Glob. Environ. Change 20, 113–120.

Roygard, J.K.F., Mark, M.A., Ray, K., 2002. No-till corn yields and water balance in themid-Atlantic coastal plain. Agron. J. 94, 612–622.

Sharpley A.N., Williams J.R., 1990. EPIC: Erosion/Productivity Impact Calculator 1.Model Documentation, Technical Bulletin 1768. U.S. Department of Agriculture.

Soil Survey Staff, 2012. Natural Resources Conservation Service, United StatesDepartment of Agriculture. Web Soil Survey. <http://websoilsurvey.nrcs.usda.gov/> (accessed 12.12.12).

Suyker, A.E., Verma, S.B., 2009. Evapotranspiration of irrigated and rainfed maize–soybean cropping systems. Agric. For. Meteorol. 149, 443–452.

USDA (United States Department of Agriculture), 1954. National EngineeringHandbook, sect. 4, Hydrology. U.S. Govt. Print. Office, Washington, DC.

USDA (United States Department of Agriculture), 2010. Farm and Ranch IrrigationSurvey (2008): 2007 Census of Agriculture, Volume 3, Special Studies, Part 1,AC-07-SS-1. U.S. Department of Agriculture, Washington, DC.

Vorosmarty, C.J., Green, P., Salisbury, J., Lammers, R.B., 2000. Global waterresources: vulnerability from climate change and population growth. Science289 (5477), 284–288.

Williams, J.R., Jones, C.A., Kiniry, J.R., Spanel, D.A., 1989. The EPIC crop growthmodel. Trans. ASAE 32, 497–511.

Zonderland-Thomassen, M.A., Ledgard, S.F., 2012. Water footprinting – acomparison of methods using New Zealand dairy farming as a case study.Agric. Syst. 110, 30–40.