Climate change impacts on global agricultural water deficit


  • Xiao Zhang,

    1. Ven Te Chow Hydrosystems Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA
    Search for more papers by this author
  • Ximing Cai

    Corresponding author
    • Ven Te Chow Hydrosystems Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, USA
    Search for more papers by this author

Corresponding author: Ximing Cai, Ven Te Chow Hydrosystems Laboratory, Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA. (


[1] This paper assesses the change in crop water deficits (the difference between crop evapotranspiration and precipitation that is effective for crop growth) of 26 crops (including rainfed and irrigated) under current (1961–1990) and projected climates (2070–2099). We found that despite the universally rising mean temperature, crop water deficits are likely to decline slightly at the global scale, although changes vary by region. While the increasing precipitation and changing intra-annual precipitation distribution in many areas can lead to more effective rainfall for crop growth, the declining diurnal temperature range will play a key role in offsetting the warming effect at the global scale. Regionally, Africa and China are likely to benefit from lower water requirements, but the impacts on other regions, including Europe, India, South America, and the United States, are subject to the land-use types (rainfed or irrigated) and the uncertainty involved in the assessment approaches.

1 Introduction

[2] Water requirements and availability are critical factors that determine the extent of climate change impacts on agriculture [FAO, 2011]. Some aspects of climate change, such as longer growing seasons and warmer temperatures, may bring benefits for agriculture. However, climate change can also have adverse impacts on food production by increased water deficits (WD) resulting from growing water requirements (crop evapotranspiration) and reduced water availability from effective precipitation [Mimi and Jamous, 2010]. Thus, it is of key importance to investigate how climate change could affect agricultural water use, especially irrigation requirements, given that irrigated land produces about 40% of the global harvest [FAO, 2011]. Although a few studies have examined the issue [Döll, 2002; Fischer et al., 2007], their results differ greatly because of data limitations and oversimplified assumptions. Döll categorized all crops into two groups (rice and non-rice) and adopted a uniform growth period length of 150 days [Döll, 2002], while Fischer et al. assumed that future irrigation requirements were linearly proportional to the irrigated area [Fischer et al., 2007]. Those assumptions neglect many crop details necessary for crop water-use assessment. In this study, we provide a biophysically-based assessment of the climate change effect by using more comprehensive crop information and explicitly considering climate change projection uncertainty. We address the following key questions: how much will current irrigation requirements (i.e., WD for irrigated crops) and WD for rainfed crops be mitigated or aggravated as a result of climate change, and which regions will see these changes in WD? More broadly, what will the global situation of agricultural water be under climate change? By addressing these questions, we provide a global picture of agriculture water use together with its corresponding likelihood of occurrence under the various climate change scenarios. Through this analysis, implications for regional agricultural development considering water and land conditions will be discussed.

2 Methods

[3] The questions listed above are answered through a spatially explicit assessment of WD in both irrigated and rainfed land under the reference climate condition (1961–1990) and the projected scenarios (2070–2099), respectively. A range of potential changes is provided by incorporating various climate scenarios. Different general circulation models (GCMs) provide different (and even conflicting) projections for one region [Laurent and Cai, 2007]. So two data ensemble approaches, simple average method (SAM) and root mean square error minimization method (RMS), are adopted given the assumption that ensemble of the GCMs provides more reliable climate prediction than any single GCM [Murphy et al., 2004; Weigel et al., 2010]. In addition, two representative CO2 emission scenarios (A1B and B1) are used to represent a range of emission levels: A1B projects greater rates of greenhouse gas (GHG) emissions than B1, assuming CO2-equivalent GHG concentrations of 850 parts per million by volume (ppmv), compared to 600 ppmv under B1 [IPCC, 2007]. The combination of GCM ensemble approaches and CO2 emissions ends with four future scenarios: A1B-SAM, B1-SAM, A1B-RMS, and B1-RMS. Crop-wise water requirements of 26 crops are assessed, employing the most up-to-date crop data [Portmann et al., 2010]. Pasture lands are not included in the estimate. Diurnal temperature range (DTR), an index of climate change [Braganza et al., 2004], is used to explain changes in agricultural water requirements. In-depth regional analyses are performed for Africa, China, Europe, India, South America, and the United States, covering the world's key food producers. The details of the data sets and procedures on the WD assessment, including the validation of the assessment results under the reference scenarios, are provided in the Supporting Information.

3 Results

[4] WDs on irrigated lands (i.e., irrigation requirements) have a high likelihood of decreasing in the world under all four scenarios, although by differing magnitudes (Table 1a). Regionally, Africa, China, and South America are predicted to have reduced irrigation requirements under all four scenarios; Europe, India, and the United States, however, are more scenario dependent. India is predicted to have an increase in irrigation requirement with A1B-SAM, B1-SAM, while Europe and the United States have higher predicted irrigation requirements under A1B-RMS only. Unlike irrigation requirements, the global WD changes for rainfed crops are more sensitive to the scenarios and depend on the GCM uncertainty assessment approaches (Table 1b). In general, RMS projects diminished water deficits for rainfed crops, while SAM predicts the opposite at the global scale. Regionally, Africa and China are still likely to benefit from lower water deficits, while Europe and the United States may anticipate rising deficits for rainfed crops. The changes in India and South America are subject to the ensemble approaches: RMS predicts decreases, while SAM predicts increases.

Table 1. Regional and Global Changes of Water Deficit for (a) Irrigation Crops and (b) Rainfed Crops under Reference Scenario (1961–1990) and Four Projected Scenarios (2070–2099)
 AfricaChinaEuropeIndiaSouth AmericaUnited StatesGlobal
(a) Irrigation crops
Reference86 173 48 341 44 121 1289 
(b) Rainfed crops
Reference492 139 385 305 265 422 2871 

[5] The change in WD for both irrigated crops and rainfed crops varies largely by region. Comparing the annual irrigation requirement changes for the aggregated crops (up to 26 crops in the various regions) between the projected climate (2070–2099) and the reference climate (1961–1990) (A1B-SAM scenario) (Figure 1a) indicates that the western U.S., southern Africa, and northern Australia may expect substantial decreases in irrigation requirements, whereas the southeastern U.S., northeastern South America, and northwestern India may have significant increases. Eastern Europe and southern China are likely to have minor increases, while Western Europe and northern China are likely to experience slight decreases. Comparing the annual WD for rainfed crops of the two time periods (Figure 1b) indicates western Sub-Saharan Africa may have to face a more severe water situation, with greater deficits, whereas water concerns in other parts of Africa may be reduced to some extent as a result of smaller water deficits. Water conditions in the western U.S. and northern China may also be improved, but the eastern U.S. and southern China are likely to have a slightly greater water deficit. Southern India, the west coast of South America, and Western Europe may have a smaller water gap in rainfed areas, while northern India, Eastern Europe, the northern Amazon, and northeast South America are likely to experience varying levels of water deficit increases.

Figure 1.

(a) Irrigation requirement changes for irrigated areas and (b) water deficit changes for rainfed areas of 26 crops between 2070–2099 and 1961–1990 under A1B-SAM scenario.

[6] A statistical analysis further illustrates the results stated above. The Wetness Index (WI), defined as the ratio of effective precipitation to crop evapotranspiration, is calculated for all crops in both rainfed and irrigated areas. Changes in the WI reflect the joint effect of changes in temperature and precipitation. The exceedance probabilities are plotted and compared by the reference scenario and the various projection scenarios for both rainfed and irrigated crops, respectively. The aggregated exceedance curve for 26 crops is then generated by utilizing crop area as the weight and presented for different regions in Figure 2. The results indicate Africa and China are likely to have improved water conditions for both irrigated and rainfed lands as the exceedance probabilities of all future climate scenarios are larger than those of the current climate scenario over nearly all WI levels. However, in extremely dry areas (with WI lower than 0.2) in Africa, the improvements for irrigated crops are minor. Impacts in Europe are minor, and negative effects are possible in dry areas. For rainfed crops, slightly advantageous conditions are expected for very wet areas, but opposite results show for very dry areas; for irrigated crops, slightly advantageous conditions are expected with relatively wet areas with high likelihood (the exceedance probability is approximately 70%). Influences in India, South America, and the United States are scenario dependent, and the ensemble approach plays a greater role in shaping the curves. For rainfed crops in India, compared to the current climate scenario, all future scenarios project slightly better conditions for relatively wet areas (WI greater than 0.6); for dry areas, the SAM scenarios generate larger water deficit for rainfed crops, while other scenarios still project better water conditions. For irrigated crops in India, improvements are projected for wet lands but no significant change for dry areas. In South America, the RMS scenarios project advantageous results for both irrigated and rainfed areas. However, comparing the SAM scenarios to the current climate scenario, WI increases in very wet and very dry areas but decreases in the areas with a large range of intermediate WI, which shows a high likelihood of larger WDs. In the United States, the beneficial influence of climate changes are likely to occur with high confidence for dry areas (WI less than 0.4) for both rainfed and irrigated crops, but climate change will likely be detrimental for areas with medium WI. The adverse change may significantly affect agricultural production in the U.S. and South America since major high-yield crops are planted in the medium WI zones, such as maize in the Midwest of the U.S. and sugarcane in southeast of Brazil.

Figure 2.

Wetness Index (Peff/ETc) exceedance curve area aggregated of 26 crops on rainfed (upper) and irrigated (lower) area in (a) Africa, (b) China, (c) Europe, (d) India, (e) South America, and (f) the United States under the reference scenario and four future scenarios.

[7] Crop-wise analysis of WD is also performed to investigate how different crops are likely to be impacted. Six main crops with high WD are selected and examined, as shown in Figure S2. Wheat and maize do not exhibit much alteration in irrigation requirements, while rice and fodder grass may have diminished requirements with relatively low uncertainty. Changes in sugar cane irrigation requirements are comparatively neutral but subject to high uncertainty. Cotton is more likely to have higher requirements, but the change in magnitude is scenario sensitive. In terms of their water deficits, wheat, maize, and fodder grass exhibit wider ranges of change than in the case of irrigation requirements, while barley and other perennials have more narrow ranges. Rainfed maize and soybean are likely to be negatively affected by climate change with higher water deficits. In contrast, wheat and other perennials may benefit from climate change with lower WDs, although the likelihood of this occurring is different for each. The differences in crops' responses mainly result from their various growth periods, different water requirements and availability, and the differing spatial distributions of their growth.

4 Discussion

[8] An issue with the GCM model variability is the fact that we use only six GCMs due to the data requirement for the assessment and limited availability from the Intergovernmental Panel on Climate Change (IPCC) global data sets: only these six GCMs can provide all the required inputs, especially the daily maximum and daily minimum temperature. However, these six GCMs may not represent well the complete range of GCM predictions for some regions. To address this issue, we compare the predictions of precipitation from the six GCMs with those from 13 other GCMs (Figure S5). The mean differences in the outputs for the six regions are within [−0.03, 0.25] mm/day (Table S2). By country or regional aggregation, China and the U.S. have slightly lower precipitation predictions from the six GCMs, which supports the general finding of declining WD in these countries; other regions have slightly higher values from the six GCMs, which supports the results of increasing WD for rainfed crops in India, South America, and Europe or implies that the decreases in WD may occur with smaller magnitudes or even opposite direction. Regions such as the eastern U.S. and northern India tend to have relatively lower precipitation from the six GCMs than the 13 GCMs, suggesting that the projected WD increases in these regions can occur with smaller magnitude. In contrast, the ensemble precipitation prediction from the six GCMs is slightly higher in the western U.S., eastern Brazil, Australia, and vast areas of Sub-Saharan Africa, except the region around DR Congo. Thus, the predicted WD decreases need additional examination.

[9] The trend of decreasing WDs, especially irrigation requirements, might not be intuitive but is not surprising if we examine relevant historical global and regional data. Global pan evaporation has been decreasing for the last few decades [Peterson et al., 1995] despite the continuously rising global mean temperature. Although pan evaporation does not fully represent ground evapotranspiration, it provides some valuable insights regarding evaporation change patterns and its main influencing factors. Ohmura and Wild [2002] also illustrated that a warmer atmosphere did not necessarily indicate more evaporation. For example, the observed reference evapotranspiration in North China, Yangtze River catchment, and the Qinghai-Tibetan Plateau area shows a decreasing trend over the past 50 years [Song et al., 2010; Xu et al., 2006; Zhang et al., 2009] despite the increasing mean temperature.

[10] The decrease in evapotranspiration that has occurred even as the mean temperature has risen can be explained by the declining DTR, the difference between daily maximum temperature and daily minimum temperature. DTR is an index independent of variations in global mean temperature and has been identified as a suitable indicator of climate change [Braganza et al., 2004]. Peterson et al. found a strong correlation between the diminished pan evaporation and decreased DTR [Peterson et al., 1995]. Shen et al. [2010] also found a similar consistency in the arid region of China. The decline in DTR mainly results from the increased cloud coverage and/or aerosol concentration, which leads to a substantial decrease in global solar irradiance [Roderick and Farquhar, 2002]. The fact that decreased DTR results in reduced evapotranspiration can be explained by three factors: (1) the average vapor pressure deficit remains more or less constant despite the rising mean temperature since the decreased DTR offsets the warming effect, preventing it from facilitating any significant increase in evapotranspiration [Roderick and Farquhar, 2002]; (2) since the lower DTR is strongly correlated with increased cloud cover and/or aerosol concentration, the incoming solar radiance is reduced, and the lack of energy further decreases the evapotranspiration rate [Allen et al., 1998; Roderick and Farquhar, 2002]; (3) lower DTR may imply higher humidity, which also impedes evapotranspiration [Braganza et al., 2004]. IPCC's report (AR4) stated that the future warmer climate is likely to cause decrease in DTR [IPCC, 2007], and this has been demonstrated by other studies as well [Kaas and Frich, 1995]. The average DTR changes between a future scenario (e.g., A1B-SAM) and the baseline are displayed in Figure S3. Based on the GCM prediction and other studies [Easterling et al., 1997; Le, 2011], the DTR is likely to globally decline in the future except for in parts of the Middle East and the central part of China, and decreases in potential evapotranspiration can be anticipated due to its strong correlation with DTR. Figure S4 shows the difference of the reference ET (ET0) between the projection under A1B-SAM scenario and the reference scenario. The global average ET0 is projected to decline, which may indicate that decreased DTR outweighs the increased average temperature. Around the globe, the influence of decreased DTR dominates in the west of North America, Western Europe, and eastern Russia, in addition to the majority of Australia and sub-Saharan Africa. However, in some regions such as mid-eastern North America, Saharan Africa, and Mideast, ET0 is likely to rise as a result of the rising mean temperature.

[11] Regions with declining WD in the future may have greater opportunity for agricultural development than those with increasing WD. Meanwhile, adaptations in terms of crop patterns, cropping calendars, and land-use types (with or without irrigation) can help reduce irrigation requirements or rainfed crop water deficit in regions projected to be most affected by climate change. It is particularly interesting to explore insights for agricultural adaptations to climate change that can be gained by considering the coupling of changes in water requirement and land suitability. These adaptations can be examined by overlaying the map of projected land suitability for agriculture in the future (obtained from our previous work) [Zhang and Cai, 2011] onto the map of present rainfed and irrigated areas. Table 2 shows three land categories which provide implications for adaptations. Land in category 1 will not be able to achieve regular agricultural productivity without utilizing irrigation or other adaptation options; category 2 identifies areas where irrigation requirement is reduced to the extent which allows reasonable rainfed crop harvest; category 3 offers opportunities for new agricultural land development if needed in the future. The assumption with the categorization is that rainfed lands, either observed or projected, are suitable for cultivation with precipitation, while irrigated lands are marginally suitable for agriculture if without irrigation. Moreover, it is worth noting that some existing rainfed areas have marginal suitability in many arid and semi-arid areas, and such situation is likely to continue to the future, which is not necessarily a result from climate change.

Table 2. Land Classifications Regarding Future Land Suitability for Agriculture and Current Land Use under A1B-SAM Scenarioa
CategoriesAfricaChinaEuropeIndiaSouth AmericaUnited StatesWorld
106 km2%106 km2%106 km2%106 km2%106 km2%106 km2%106 km2%
  1. aCategory 1: Land currently rainfed with declining productivity in the future without irrigation; category 2: Land currently irrigated but with less irrigation requirement in the future; and category 3: Land currently not cultivated but being suitable for cultivation in the future. The percentages are obtained by dividing the area by current rainfed area for categories 1 and 3 and by current irrigated area for category 2.

[12] Globally, around 10% of current rainfed area might be unsuitable for rainfed agriculture in the future, mostly due to either large WDs at present or increased WDs in the future due to climate change. India is the most affected region in terms of WDs for rainfed crops among the six shown in the table. Over half of the rainfed lands have sub-suitable productivity, and this situation is likely to exacerbate under climate change. Europe will have 25% of its current rainfed area that can be negatively affected by climate change. The effects in other four regions are comparatively small. Regarding water requirement on irrigated land, China is one of the regions that will benefit significantly, with nearly 90% of the irrigated area in the country expecting less irrigation water requirement; the United States, India, and Europe also expect reduction in irrigation requirement in considerable fraction of their current irrigated area. Although significant decrease of WD is estimated for irrigated crops in Africa, probably due to the current large WD, the decrease of WD will not improve the land to the condition for rainfed agriculture. This is why a relatively small increase of the land in category 2 is found for Africa. Furthermore, all regions will have considerable potential of developing new agricultural land with appropriate water availability; especially Africa will have the largest opportunity.

[13] The results indicate Africa may face a dilemma due to the altered distributions of arable land and water availability provoked by climate change. On the one hand, precipitation in many regions of Africa is likely to increase; on the other hand, the overall arable-land availability for rainfed crops may decline since climate change may cause unfavorable changes in soil temperature regime and soil moisture in most areas [Zhang and Cai, 2011]. Therefore, it is difficult to tell to what extent irrigation will be needed to maintain African land productivity. Eastern China has a high chance of benefiting from climate change because of its projected reduced irrigation requirements and retained land productivity: in fact, the irrigation demands in this region could be significantly reduced. In northwestern China, previously marginal land could become new agricultural land with regular productivity. The land classification map for Europe is mostly consistent with the water-change maps, showing that Western Europe may have lower irrigation demands with reasonable land suitability, while Eastern Europe is likely to be confronted with widening water gaps along with marginal suitability, making irrigation a necessary and important means of agricultural adaptation for the region. Irrigation is likely to continue to play an important role in India, particularly in the northwestern region. Although the southern part of India may have a wetter climate in the future, irrigation could still prove essential due to the prediction of a mix of various types of land. In South America, Venezuela, southern Brazil, and northeastern Argentina may see higher irrigation requirements to retain crop production there. By contrast, the western coast of South America is likely to have a wetter climate, but the land slope and elevation limits further agricultural expansion. The U.S. Midwest may encounter disadvantageous climatic changes that could increase water gaps between crop needs and precipitation. However, investment in irrigation may not be necessary, as the land will still be capable of achieving appropriate productivity for rainfed crops. Areas in the western U.S. may see decreased water deficits, but in spite of this, cropland in this region may retain below-suitable productivity due to its elevation and other limiting factors. Above all, adaptation and mitigation measures should be wisely planned and cautiously implemented by taking economic, social, and environmental factors into account. Moreover, irrigated areas are projected to increase in the future in order to produce more food to satisfy increasing food demand due to population growth and changes in diet [FAO, 2012]. The socioeconomic factors together with the impact of climate change on water requirements, especially for those regions with potentially increasing irrigation water requirements and the need of switching rainfed crops to irrigated due to climate factors, will face a great challenge to find sufficient water to match the new requirements, given the situation that many regions, especially arid and semi-arid regions, already suffer water shortage problems today. For those regions, more attention should be paid to potential agricultural changes so wise adaptation decisions can be made based on their water resources availability, financial capability (e.g., for the cost of increased irrigation), and environmental protection policies. It is worth noting that the land suitability is actually crop specific, which will affect crop water requirement.

[14] It should be noted that this study does not consider the impact of adaptation measures. In particular, it is assumed that cropping calendars and patterns will not change. First of all, the assumption makes such a global assessment feasible, since it is very difficult, if not impossible, to project specific crop calendar changes and farmers' adaptations regarding planting time and crop distribution around the world for 26 crops. Second, farmers' adaptations are likely to take advantage of the changed climate [IPCC, 2007], which would mitigate the negative effects. For example, studies show that changing crop phenology (e.g., advanced crop development, earlier sowing) can reduce WDs [e.g., Sacks and Kucharik, 2011; Siebert and Ewert, 2012]. Thus, assessments without considering adaptations provide a possible “worst-case scenario,” which requests cautious planning for appropriate adaptations to mitigate the impacts in the future. Moreover, warming climate may increase the potential of multicropping in some area [FAO, 2003], which improves water-use efficiency [Francis, 1987] while possibly demanding more water. Also, the effect of the increased CO2 concentration is not taken into account, which is expected to result in a more efficient photosynthesis and less crop water requirement as well.

[15] Finally, the global assessment is limited by data availability and resolution and modeling capability, which can be improved at the regional or local scale. For example, the different levels of preseason soil moisture can affect the crop water requirement, while it is considered as an average or normal condition; the paddy rice in some regions around the world needs a water layer to flood the field [Chapagain and Hoekstra, 2010], which is affected by rainfall during the crop season, while in this study it is assumed that the water layer requirement remains the same in the reference and projected scenarios. Nevertheless, such limitations should not significantly affect the result of the global assessment, and more regional and local studies are needed to further confirm the findings as well as provide more reliable information for local adaptations.

5 Conclusions

[16] Finally, despite the assumptions and uncertainties associated with the climate models and crop water simulation, this paper presents where and how agricultural water requirements (both irrigation requirements for irrigated crops and water deficits for rainfed crops) will likely be affected by climate change from a biophysical perspective. The predicted impacts display high heterogeneity spatially and vary with crop. The broad-scale analysis presented here aims to identify the regions where concerns may arise with relatively high likelihood: further studies at finer scales are needed for water resources planning and management. In particular, the predicted changes of WD for some countries or regions are sensitive to the emission and model scenarios. This uncertainty needs to be addressed through downscaling processes and/or the improvement of GCMs. It should also be noted that extreme weather events associated with climate change, such as heat waves and droughts, will affect irrigation at the local scale, and the extent to which these events will affect the regional, long-term estimation of water use needs additional investigation. Other factors, such as seasonal variability and monsoonal climates, may also affect the accuracy of the estimates and can be addressed with more refined data and more sophisticated climatic and agronomic models. Furthermore, studies investigating related issues, including agricultural water-use efficiency and drought-tolerant crops, would help provide a more comprehensive assessment of future agricultural water situation. The actual water requirement will finally be influenced by adaptation measures at the local level, most likely in a positive direction.


[17] The authors are grateful to Bo Teng for her assistance in data processing. This study is funded by the U.S. Department of Agriculture (USDA) through grant IND010576G1.