Improvements in crop water productivity increase water sustainability and food security—a global analysis

Crop production and water consumption data for irrigated and rainfed crops were provided by Siebert and Döll [23], who computed daily water balances and corresponding crop water requirements using the FAO Penman–Monteith method of reference evapotranspiration [29] modified by crop coefficients and considering crop areas and crop calendars [30]. For irrigated crops, actual evapotranspiration (AET) was assumed to be equal to potential evapotranspiration (PET), with irrigation water compensating for any rainwater deficit; supply-side limitations to irrigation were thus not incorporated into the model. In addition, our analysis of irrigated crops evaluates beneficial irrigation only, as data on use for other irrigation-related activities (salt flushing, soil preparation in rice paddies) are not globally available [23]. The ratio between AET and PET was used to separate and downscale yields between irrigated and rainfed crops. Yield data were originally developed by Monfreda et al [17] by combining census data for agricultural production with estimates of cropland extent derived from satellite imagery. The data have a spatial resolution of five arc minutes and report 7-year average yield and 5-year average water consumption centred on the year 2000.

To evaluate climate and non-climate effects on crop water productivity, we divided cropland into 20 climatic zones (SI figure 1 available at stacks.iop.org/ERL/8/024030/mmedia) defined by the ratio of mean annual precipitation (P) to mean annual potential evapotranspiration (PET) [31]. We attributed the difference in maximum water productivity across climatic zones to the effects of climate. In contrast, variation in water productivity within each climatic zone was attributed to non-climate effects. For statistical power, we considered only crops with an extent of >50 000 ha (irrigated) or >100 000 ha (rainfed) in a given climatic zone. Further detail on methods is provided in the supplementary information (SI, available at stacks.iop.org/ERL/8/024030/mmedia).

We calculated the Generalized inequality index (Gini) [32] to quantitatively compare inequalities in water productivity among crops and across regions. Cumulative water consumption and cumulative kilocalorie production for each crop at 5 arc minute resolution were ordered by water productivity; the 'curves' in figure 3(b) are compilations of these points.

For each crop, we calculated the production-weighted 20th percentile value for each climatic zone where P < PET. In each climatic zone, we identified all cropland where water productivity for each crop was less than its 20th percentile value. We then calculated the calories that could be added or water saved were the water productivity of any crop below its 20th percentile value increased to that value (SI table 1 available at stacks.iop.org/ERL/8/024030/mmedia). Because improvements for each crop are measured against its own 20th percentile value in the same climatic zone, improvements are scaled to the range of variability for each crop.

Water consumption by crops varies substantially across the globe (SI figure 2 available at stacks.iop.org/ERL/8/024030/mmedia), reflecting differences in cropping density, crop choice, soil characteristics, irrigation availability, and agricultural management as well as climatic drivers of evapotranspiration. The distribution of crop water consumption among climatic zones (figure 1) illustrates the crops likely to contribute substantially where water resources are scarce. In arid climates (P/PET < 0.2), rainwater consumption is dominated by millet and sorghum while irrigation water is mainly used for wheat and rice. In semi-arid and dry-sub-humid climates (0.2 < P/PET < 0.7), rainwater consumption is dominated by wheat, rice, and maize while irrigation water use is dominated by wheat, rice, sugarcane, and maize. For the 16 crops studied, 52% of rainwater consumption (by irrigated and rainfed crops) and 82% of irrigation water consumption occur in regions where precipitation is potentially limiting (P < PET).

Zoom In Zoom Out Reset image size

We found a large range in crop water productivity. For example, in large parts of Sub-Saharan Africa maize water productivity was below 1 kcal l⁻¹ (0.3 kg m⁻³) while the highest water productivities of maize (>6 kcal l⁻¹/1.7 kg m⁻³) were computed for the US corn belt and Northern China (figure 2). Field-based studies (such as those surveyed in [21, 22, 33]) document a similar range of water productivities; our analysis extends their insights to a broader set of climates, locations, and crops.

Zoom In Zoom Out Reset image size

Variation in water productivity causes water consumption in certain areas to be disproportionately large. On the least water productive rainfed cropland in precipitation-limited regions, ∼46% of water was consumed in the production of just 20% of maize calories. Irrigated crops consume both rain and irrigation water—for those low productivity irrigated croplands in precipitation-limited regions where irrigation water provides at least 50% of total water, ∼44% of irrigation water consumption by maize went to 20% of food production.

Crop water productivity is not evenly distributed (figure 3). For example, for irrigated wheat in precipitation-limited areas, the USA has a national-level water productivity (2.9 kcal l⁻¹/0.9 kg m⁻³) that is close to the world average (2.8 kcal l⁻¹/0.9 kg m⁻³), but water productivity is unevenly distributed (Gini of 0.32) compared to, for example, China, where irrigated wheat water productivity is uniformly high (4.3 kcal l⁻¹/1.3 kg m⁻³) and evenly distributed (Gini of 0.09).

Zoom In Zoom Out Reset image size

Variation in crop water productivity is not solely a function of climate. While each crop we analysed has its highest water productivity in specific climatic zones, water productivity varied within each climatic zone (SI figures 3 and 4 available at stacks.iop.org/ERL/8/024030/mmedia). For most crops, we see as much variability in water productivity resulting from non-climate drivers as from climate (figure 4).

Zoom In Zoom Out Reset image size

For irrigated crops, we evaluated water productivity based on combined evapotranspiration of rain and irrigation water. Water productivity of only irrigation (blue) water is a function of both fraction of overall water originating as irrigation water and total crop water productivity. The fraction of irrigation water is strongly correlated with aridity (Pearson's r² is 0.96 for wheat, 0.85 for maize, and 0.99 for rice), while water productivity of irrigated crops is not (SI figure 4 available at stacks.iop.org/ERL/8/024030/mmedia). We therefore focus on overall water productivity of irrigated crops. Though the full impact of increasing water productivity of irrigated crops would not be gained in irrigation savings, it would result in a lower irrigation water requirement, particularly in more arid climates.

If water productivity on only the least productive, precipitation-limited croplands were increased, the impact would be substantial. For illustrative purposes, we assume all rainfed improvements go towards production and all irrigated improvements go to water savings. On rainfed cropland in precipitation-limited regions, increasing water productivity to the 20th percentile could increase production by ∼120 trillion kilocalories per year while holding water consumption constant (figure 5(a)). This amounts to a ∼30% increase over current production in these areas and would provide food for ∼110 million people annually (assuming an average food requirement of 3000 kcal d⁻¹ cap⁻¹ [34]). On irrigated croplands, increasing the lowest water productivity in precipitation-limited regions could save ∼77 km³ yr⁻¹ of water without reducing food production (figure 5(b)). This water savings amounts to more than one quarter of current water consumption on low water productivity irrigated cropland and would be enough to meet the annual domestic water demand of ∼1.4 billion people (assuming a demand of 150 l d⁻¹ cap⁻¹ [35]).

Zoom In Zoom Out Reset image size

To identify countries where water savings on irrigated cropland would be most beneficial, we analysed irrigation water consumption on cropland with the lowest 20% of water productivity in precipitation-limited climates (figure 6). India accounts for more than 40% of total irrigation water consumption in these areas, 67 km³ yr⁻¹, primarily in wheat and rice cultivation. Because India has so much cropland in these areas, however, irrigation intensity (consumption per hectare) is ∼520 mm yr⁻¹, close to the median value of irrigation intensity on all precipitation-limited, low water productivity cropland (∼500 mm yr⁻¹). By contrast, in similar climates in Uzbekistan, food crops consume only ∼3 km³ yr⁻¹. However, irrigation intensity (∼1400 mm yr⁻¹) is nearly three times greater. Because of low total water consumption, improving irrigation efficiency of food crops in Uzbekistan is unlikely to alleviate basin-wide water shortages, though on-farm impacts may be dramatic. In India, by contrast, small but widespread improvements in irrigation intensity could have a substantial impact. Assessments should not assume that high irrigation consumption is linked to high irrigation intensity.

Zoom In Zoom Out Reset image size

We sought to quantify the range of tradeoffs between crop water use and production to explore, first, the feasibility of interventions to increase food production and reduce water consumption by improving crop water productivity, and, second, to quantify the potential magnitude of this type of intervention. The extensive variability in crop water productivity within climatic zones indicates that there are many non-climate related drivers of low crop water productivity, and these can potentially be managed. Even when limitations are due to soil or slopes, management techniques such as nutrient addition and terracing or furrowing can improve yields [36]. Water productivity is a ratio, and in many cases low water productivity is likely a function of low yield, so interventions to increase water productivity may focus more on improving yield than on decreasing water consumption.

Specific management strategies to increase crop water productivity must be tailored to local contexts; studies in Nigeria, Sudan, and Niger, for example, have shown that nitrogen and phosphorus limitation restrict millet and sorghum yields [37]. Water productivity could be increased by increasing yields via improved soil nutrient conditions and reduced wind-driven erosion. Examples of other interventions that have demonstrated improvements in water productivity include: rainwater harvesting and local water storage, applying drip or deficit irrigation, adjusting planting dates, and modifying tillage practices to reduce evaporation [21, 25, 33, 38–40]. In many cases, techniques for increasing water productivity, especially when it is low, require little or no additional water [41].

Any improvement in crop water productivity will have a positive effect on either food production or water savings. We demonstrate the magnitude of increasing low water productivity to the 20th percentile because this is a plausible level of increase. Allocative efficiency demonstrates that it is easier to increase resource use efficiency when resources are used inefficiently; improving the very lowest performing systems therefore should take less effort than improvements in high-performing systems [24, 25]. Implications of crop water productivity improvements to other levels are discussed in the SI (available at stacks.iop.org/ERL/8/024030/mmedia).

By limiting our analysis to 16 crops, our study does not provide a complete picture of all global crop water use or food production. Further, due to data limitations, we are unable to assess water requirements to prevent salinization or for non-beneficial irrigation. Our analysis thus underestimates the volume of water required for irrigated agriculture, particularly in arid regions where irrigation water is often the primary source of water for irrigated crops. In addition, we assume all irrigation demand is met and that irrigation is not limited by water supply. These issues loom large in local and regional management, and we hope to address them at the global scale in the future. We also do not assess production losses due to spoilage, animal feed, or post-consumer waste. The end points of food production provide important context in decision-making about water allocation but are outside the scope of the current research.

We address limitations of methodology in the SI (available at stacks.iop.org/ERL/8/024030/mmedia) in detail.