Water-scarcity footprints and water productivities indicate unsustainable wheat production in China

Water shortage is a critical constraint limiting China ’ s capacity for food security. To provide evidence supporting environmentally sustainable water use in food production, this study compared irrigation water productivities (IWPs) and water-scarcity footprints (WSFs) for China ’ s wheat production at high spatial resolution. Contrary to previous water productivity studies assessing crop yield over total water consumption, it was found that IWPs in China ’ s water-scarce northern regions were much lower than those in water-rich southern regions. The WSFs further demonstrated the larger environmental impacts resulting from irrigation in water-scarce northern regions. Hotspot regions, having IWPs in the lowest tercile (<5.2kgm − 3 ) and WSFs in the highest tercile (>0.058m 3 H 2 O e kg − 1 ), were mainly located in the Huang-Huai-Hai and northwestern regions and accounted for 34% of the cropping area but 61% of irrigation water use. Historically, the south was also an important contributor of China ’ s wheat production, but progressive shifts toward highly resource-e ﬃ cient cropping in the Huang-Huai-Hai region has occurred. The paradox is that gains in total crop water e ﬃ ciency have led to increased irrigation demand and water scarcity. Today, croplands suitable for wheat production lie fallow in some southern regions in the winter. A national reassessment of this situation is urgently needed.


Introduction
Feeding a growing population while minimizing global environmental impacts are twin challenges faced by the global food system (Davis et al., 2017;Foley et al., 2011;Scherer et al., 2018). As the world's most populous and largest food-consuming country, China's food security has been an issue of broad concern for a long time (Brown, 1997;Dalin et al., 2015;Heilig et al., 2009). Water shortage is recognized as the most critical constraint that limits China's capacity for food security (Du et al., 2014;Huang and Li, 2010;Khan et al., 2009). Although China's water resources are large in absolute terms, the average water resources per capita are less than one-third of the global average (Kang et al., 2017). Moreover, increasing demand, worsening water pollution, spatially and seasonally uneven water distributions, and climate change have aggravated the water shortage and deteriorated China's aquatic ecosystem (Cai et al., 2017;Xiong et al., 2009;Zuo et al., 2018).
The agricultural sector is responsible for the most water use in China, accounting for 63% (NBSC, 2011-2016NBSC, 2016NBSC, 2011. To address the water for food dilemma in China, improving crop water productivity (WP), especially irrigation water productivity (IWP), has been an important measure for ensuring China's water and food security. IWP refers to the total crop yield divided by the total amount of irrigation water used for crops, i.e. it is high if crop yields are high and irrigation is low. China's national grain IWP has increased from 0.8 kg m −3 in the early 1990s to 1.6 kg m −3 in 2013 (Kang et al., 2017). Currently, improving WP and IWP remains the chief concern for China's agricultural water management (Du et al., 2015; https://doi.org/10.1016/j.agwat.2019.105744 Received 4 April 2019; Received in revised form 6 August 2019; Accepted 7 August 2019 Geng et al., 2019;Kang et al., 2017). However, the WP (or IWP) concept has limited meaning when comparing the values from locations with different water scarcity backgrounds. This is because producing a crop with a high WP (or IWP) in a water-scarce region can cause more serious environmental impacts than producing a crop with a low WP in a water-rich region, as described by a life cycle assessment (LCA) based water-scarcity footprint (WSF) indicator (Ridoutt and Pfister, 2010a). Thus, the WP concept as well as other volumetric water-use indicators such as virtual water (VW), which for crops refers to the actual evapotranspiration over the crop yield (Chapagain and Hoekstra, 2008), or the VW-derived water footprint (WF vw ) (Aldaya et al., 2012), have the potential to misinform water management decisions (Ridoutt and Huang, 2012;Ridoutt and Pfister, 2010a).
It is reported that most global freshwater withdrawals currently occur in watersheds with extreme water stress (Ridoutt and Pfister, 2010b). That said, the urgent need to reduce the pressure that humanity exerts on the freshwater system does not arise from an absolute shortage of freshwater in the world but rather from the current pattern of freshwater use. Similar situations exist in the current state of crop production in China. For example, it is well-known that the extremely water-scarce North China Plain, with a water availability of less than 150 m 3 per capita per year, produces more than half of the national wheat and one-third of maize (CMWR, 2000-2015CMWR, 2015CMWR, 2000NBSC, 2011-2016NBSC, 2016NBSC, 2011. The growing demand for irrigation water in this region has resulted in an overdraft of groundwater and therefore falling groundwater tables and increased land degradation. A number of studies have focused on this issue and mitigation strategies have been identified, such as improving the WP (or IWP) of crop production from a technological perspective and restricting the amount of water extraction from a policy perspective (Dalin et al., 2015;Kang et al., 2017;Yang et al., 2003). However, as mentioned above, these strategies focus on the local conservation of water rather than considering environmental impacts at the national scale, failing to provide useful information for supporting environmentally sustainable water use in China's food production.
To present a picture of national water use associated with wheat production and identify relevant hotspots, this study compared both the IWPs and WSFs of wheat production across China. A geospatial simulation tool (GeoSim) was applied to manage the FAO AquaCrop model for spatial simulations of wheat yield and irrigation water consumption from 2010 to 2015 (2015 being the most recent year for which most of the data were available). Hotspot regions with low IWPs but high WSFs were identified with a spatial resolution of 5 arc-minutes and were also presented at the scale of China's agro-ecological zones (AEZs). The overall goal was to provide scientific evidence that will enable policies to set national agricultural water use priorities across regions by considering the environmental implications of meeting food security.

Wheat yield and irrigation water consumption modeling
Wheat yields and irrigation water consumptions were modeled by the widely applied FAO AquaCrop model (version 6.1) (http://www. fao.org/aquacrop), which simulated attainable yield as a function of water consumption under rain-fed or irrigated conditions on a daily step Steduto et al., 2009). To facilitate the use of AquaCrop and other models for a high number of spatial simulations, a model-independent and open-source tool named GeoSim was previously developed as a plug-in for Quantum GIS (QGIS, https://www. qgis.org) and its application has been improved to efficiently work at China's national scale Thorp and Bronson, 2013). GeoSim automates batch simulations with AquaCrop for different locations by passing geospatial data from polygons in a base shapefile to the model input files, and then similarly passing model outputs back to the base shapefile (Thorp and Bronson, 2013).
Six types of input data were used for the simulations: 1) crop distribution; 2) climate data; 3) crop parameters; 4) soil parameters; 5) initial soil water conditions; and 6) management data. A shapefile (vector data) of national wheat distribution in 2015 was created from a raster dataset with a resolution of 5 arc-minutes (Appendix A., Fig. A.1). Daily climate data from 825 meteorological stations across China from 2010 to 2015 were obtained from the National Meteorological Information Center (NMIC, http://data.cma.cn). Crop parameters including sowing dates, sowing density, growing stages and harvest dates were also obtained from the NMIC. Additional crop data such as crop transpiration, yield formation, and soil water stresses were based on the conservative wheat parameters provided by the AquaCrop reference manual (http://www.fao.org/aquacrop) and studies which documented these parameters for both spring and winter wheat production modeling in China . Primary soil data applied to identify soil textures were obtained from the Harmonized World Soil Database (Wieder et al., 2014). The indicative values of several soil hydraulic parameters for each soil type were obtained from the AquaCrop reference manual. To avoid modeling failures caused by low soil water content that affects canopy development, the initial soil water contents were assumed to be at field capacity, which was the default value of AquaCrop. Due to the lack of detailed national irrigation management data, irrigation water consumption and wheat yields under irrigation conditions were modeled by applying the option of "Determination of Net Irrigation Requirement" in AquaCrop, which are calculated by adding a small amount of water to the soil profile each day when the root zone depletion exceeds a specified threshold. The threshold for the allowable root zone depletion was set as 50% of the readily available soil water, which was identified as a threshold for wheat irrigation management in China . The default rain-fed condition in AquaCrop was applied to simulate the wheat yield under rainfed conditions. Other management effects such as fertilizer application and ground surface cover were disregarded. Groundwater characteristics such as the depth and quality were not considered due to the lack of a detailed national dataset. All the input files of climate, crop, soil, initial condition and irrigation management were prepared according to the AquaCrop formats. Besides these files, the AquaCrop modeling requires day numbers to indicate the first and last days of cropping and simulation periods for each year: the calculation of these variables follows from our previous study .
After all the input files were prepared, GeoSim was used to pass only the names of the files and day numbers corresponding to each spatial unit in the base layer shapefile to AquaCrop's project file, which controls the AquaCrop simulations. By minimizing the spatial inputs and identifying unique response units-unique combinations of climate, crop, soil and day numbers, the efficiency of AquaCrop in conducting national-scale simulations was improved . As the model completed the simulation, the results for wheat yields and irrigation water consumptions under irrigated conditions and wheat yields under rain-fed conditions were post-processed by GeoSim and native QGIS tools. Finally, wheat yields and irrigation water consumptions from the base shapefile were converted back to raster datasets with a resolution of 5 arc-minutes. Further details on the post-processing is described in our previous study .

Irrigation water productivity calculation
Irrigation water productivity for wheat production in each grid cell (IWP grid , kg m −3 ) was defined as the total wheat production of each grid cell (P grid , kg) divided by the total irrigation water consumption of each grid cell (I grid , m 3 ): Huang, et al. Agricultural Water Management 224 (2019) where Y irri (t ha −1 ) is the wheat yield under irrigated conditions; Y rain (t ha −1 ) is the wheat yield under rain-fed conditions; A (ha) is the area of wheat; I cons (mm) is the amount of irrigation water consumption; F irri is the fraction of irrigated cropland compared to the total cropland in each grid cell. The areas of the irrigated croplands and the total croplands at a county level were obtained for the year 2015 from national statistics (NBSC, 2011-2016NBSC, 2016NBSC, 2011. This data was used to derive the F irri for each county, which was subsequently converted to a raster dataset (Fig. A.2). We assumed that in each county, the irrigation water use was equally distributed among the area equipped for irrigation. The IWPs of wheat production across China were presented in a raster map with a resolution of 5 arc-minutes. In addition, the gridbased IWPs were aggregated to average IWPs for each of China's agroecological zones (AEZs, Fig. 1) (Eq. (4)), which were defined based on climatic, soil and landform characteristics (Liu and Chen, 2005).

Water-scarcity footprint calculation
The water-scarcity footprint in each grid cell (WSF grid , m 3 H 2 Oe kg −1 ) was calculated by using a water scarcity index (WSI)-related to the ratio of water consumption to water availability-to express the environmental relevance of water use (Pfister et al., 2009): The gird-based WSIs were originated from a global dataset for the decade 2000-2009; then, the data were rescaled to the period of 2010-2015 by using the data of water availability, water use and precipitation for each of China's first-order basins (CMWR, 2000-2015CMWR, 2015CMWR, 2000Scherer and Pfister, 2016). The primary 30 arc-minute data were disaggregated by bilinear interpolation to 5 arc-minutes. The raster map of WSIs can be found in the Appendix A. (Fig. A.3).
The WSFs of wheat production across China were also presented in both a raster map with a resolution of 5 arc-minutes and a AEZ map. The WSFs for each of China's AEZs were calculated by the following equation:

Hotspots identification
To identify the hotspot regions, IWPs and WSFs at both grid and AEZ scales were classified as low, moderate and high by dividing them into terciles, i.e. if they were below the 33.3% quantile (here IWP < 5.2 kg m −3 ; WSF < 0.0065 m 3 H 2 Oe kg −1 ), between 33.3% and 66.7%, and above 66.7% (here IWP > 15.8 kg m −3 ; WSF > 0.058 m 3 H 2 Oe kg −1 ). Thus, there can be nine combinations of IWPs and WSFs. The combinations were mapped using a matrix for color coding, and the corresponding wheat areas identified by each color were calculated and presented as percentages of the total wheat area. The hotspot regions were defined as the regions with low IWPs but high WSFs, which means a lot of irrigation water is used for wheat production in water scarce regions. Trade-offs can occur if IWP is low (i.e. water is wasted) but WSF is low (i.e. environmental impacts are low because of water abundance), or if IWP is high (i.e. water is saved) but WSF is also high (i.e. environmental impacts are high because of water scarcity). Since there was considerable uncertainty in the choice of boundaries, which would be further increased by combining two indicators with different units, we judged this qualitative representation as adequate for the type of information being analyzed rather than including more precise quantitative details.

Wheat production and irrigation water consumption
The highest amounts of grid-based wheat production were mainly observed in the Huang-Huai-Hai AEZs (Fig. A.4a). The total production of these AEZs contributed to approximately 60% of the total national wheat production. Similarly, the highest amount of irrigation consumption for wheat production was also observed in the in the Huang-Huai-Hai AEZs (Fig. A.4b). Since the amount of precipitation is very low in most of the region during the winter wheat cropping season, irrigation is essential for wheat growth. It accounted for approximately 66% of the total national irrigation consumption for wheat production.

Irrigation water productivities and water-scarcity footprints
IWPs vary largely across regions, in line with differences in wheat yield and irrigation intensity. Most of the rain-fed or idly irrigated (unnecessarily irrigated without achieving any yield improvements) wheat production areas were located in the water-rich southern region (Fig. 2a). Low irrigation water consumption in some southern and northeastern regions resulted in much higher IWPs (> 20.0 kg m −3 ) compared to wheat grown in some northern, northwestern and southwestern regions (< 2.5 kg m −3 ). According to the AEZ scale (Fig. 2b), the Northwest AEZs had much lower IWPs (< 2.5 kg m −3 ), while some Southwest, Sichuan Basin and Northeast AEZs had much higher IWPs (> 20.0 kg m −3 ). If idle irrigation was avoided, most AEZs in the southern regions could have higher IWPs, e.g., the IWP of the Southwest AEZ encoded as 9.2 may increase from 31.3 to 35.6 kg m −3 .
Larger environmental impacts resulted from irrigation in water- Fig. 1. China's first-order agro-ecological zones (AEZs). The number of the AEZs represents codes linked to their names (Liu and Chen, 2005). Each firstorder AEZ includes several sub-order AEZs.
scarce regions, such as the Huang-Huai-Hai and northwestern regions (Fig. 3a). The WSFs of wheat in these regions were higher than 0.10 m 3 H 2 Oe kg −1 . In contrast, the WSFs of wheat produced in some southern and northeastern regions were less than 0.01 m 3 H 2 Oe kg −1 . At the AEZ scale (Fig. 3b), the Northwest AEZs as well as some Huang-Huai-Hai AEZs had much higher WSFs (> 0.10 m 3 H 2 Oe kg −1 ), while most Northeast AEZs and most AEZs in southern China had much lower WSFs (< 0.02 m 3 H 2 Oe kg −1 ).

Hotspot regions
The hotspot regions were identified as the grid cells with IWPs in the lowest tercile (< 5.2 kg m −3 ) but with WSFs in the highest tercile (> 0.058 m 3 H 2 Oe kg −1 ) (marked red in Fig. 4a). The hotspot regions accounted for 34% of the total wheat cropping area (Fig. 4c), 32% of the national wheat production, and 61% of the national irrigation consumption for wheat production. Most hotspot regions were located in the Huang-Huai-Hai and northwestern regions. The most sustainable water use was found in the northeastern and southern regions, accounting for 11% of the national wheat cropping area (dark green in Fig. 4a). At the AEZ scale (Fig. 4b), the hotspot AEZs were identified as the Northwest and several Huang-Huai-Hai AEZs. Some AEZs in the Huang-Huai-Hai region and surroundings also had very high WSFs, but moderate IWPs (light coral in Fig. 4c).

Comparison with other research
Most previous studies focus on the water productivity (WP) of wheat production (Cao et al., 2015;Huang and Li, 2010;Liu et al., 2007aLiu et al., , 2007b, the virtual water (VW) (Siebert and Döll, 2010;Sun et al., 2013) or the VW-derived water footprint (WF vw ) (Cao et al., 2014;Mekonnen and Hoekstra, 2010;Zhuo et al., 2016). To our knowledge, IWPs and WSFs for wheat production have not yet been estimated with such a high spatial resolution at the national scale. This impeded conducting a detailed spatial comparison with other research. To ease the comparison, the grid-based IWPs and WSFs were aggregated to national and regional average values and compared with literature values after harmonizing the concepts (e.g., WP vs IWP) and units ( Table 1). The national and regional average IWPs in this study were higher than those derived from the literature (Cao et al., 2015;Huang and Li, 2010;Mekonnen and Hoekstra, 2010;Sun et al., 2013;Zhuo et al., 2016). An important reason is that most other studies used the wheat yield data from statistics or scaled the modeled yields to fit the statistics, but applied modeled water consumption values under full irrigation (Cao et al., 2015;Mekonnen and Hoekstra, 2010), which mixes optimal irrigation with suboptimal yields and may therefore overestimate the irrigation water consumption and result in lower IWPs. In contrast, by applying the option of "Determination of Net Irrigation Requirements" in AquaCrop without considering stress factors such as low fertility and high salinity, this study consistently assumes optimal conditions, but may overestimate the wheat yield and result in higher IWPs. The  (Liu and Chen, 2005). The integer part of the numbers indicates the code of first-order AEZs. For the names of the first-order AEZs, please refer to Fig. 1. Idle irrigation means the application of irrigation does not increase wheat yield compared with rain-fed wheat. Fig. 3. Water-scarcity footprints (WSFs, m 3 H 2 Oe kg −1 ) with a resolution of 5 arc-minutes (a) and (b) at the scale of agro-ecological zones (AEZs). White indicates no data or no wheat production. The number of the AEZs represents codes linked to their names (Liu and Chen, 2005). The integer part of the numbers indicates the code of first-order AEZs. For the names of the first-order AEZs, please refer to Fig.1.  Fig. 4. Hotspot regions with a resolution of 5 arc-minutes (a) and (b) at the scale of agro-ecological zones (AEZs). Hotspots depend on IWPs and WSFs. Their synergies and trade-offs are depicted by a color matrix (c). The pie chart displays the share of each combination of IWPs and WSFs (d). White indicates no data or no wheat production. The number of the AEZs represents codes linked to their names (Liu and Chen, 2005). The integer part of the numbers indicates the code of first-order AEZs. For the names of the first-order AEZs, please refer to Fig. 1. (For interpretation of the references to colour in the text, the reader is referred to the web version of this article.). average national wheat yield (the total production divided by the total cropping area) in this study was 6.5 t ha −1 , while the average national data from the statistics during 2010-2015 was 5.0 t ha −1 (NBSC, 2011-2016NBSC, 2016NBSC, 2011. Other reasons such as different studied time series and spatial resolutions may also result in higher IWPs in this study compared with previous studies. Higher wheat yields were also a reason for lower WSFs in this study compared with previous studies (Pfister and Bayer, 2014;Scherer and Pfister, 2016). In addition, this study calculated the national WSF as a production-weighted average based on gridded data for both water consumption and WSI, while Pfister and Bayer (2014) applied water consumption at the grid level but WSIs at the watershed level, and Scherer and Pfister (2016) focused their study on spatially explicit WSIs and only used national-scale crop production data to calculate WSFs for a few case studies. The different methods and different studied time series can also cause different results.
To identify the hotspot regions, the relative IWP and WSF values across China rather than the absolute values are more meaningful. Despite different absolute values among the studies, the spatial variation identified in IWPs and WSFs by this study resembles other studies (Huang and Li, 2010;Mekonnen and Hoekstra, 2010;Pfister and Bayer, 2014;Sun et al., 2013). The water-scarce northern regions such as the Huang, Huai, and Hai basins always had lower IWPs than the water-rich southern regions such as the Yangtze basin. For example, this study found that the IWP of wheat produced in the Hai basin was 3.5 kg m −3 , while that in the Yangtze basin was 14.4 kg m −3 . Similar results were obtained from Huang and Li (2010) who estimated an IWP of 1.8 kg m −3 for the Hai basin and an IWP of 19.4 kg m −3 for the Yangtze basin. Detailed regional WSF values were not available in previous studies, but the global map in Pfister and Bayer (2014) demonstrated a similar trend as this study, illustrating that the WSFs of the northern regions such as the Huang-Huai-Hai basins and northwest area were much higher than most southern regions.

Implications of this study
Previous studies have reported that wheat production in China's water-scarce northern regions (e.g., Huang-Huai-Hai basins) had higher WPs than most southern regions (e.g., Yangtze basin) (Cao et al., 2015;Huang and Li, 2010;Liu et al., 2007a). This WP indicator, which integrates green (i.e. soil moisture) and blue water (i.e. groundwater or surface water) into a single assessment of water consumption, is confusing, as the consumption of green water is not equivalent to the consumption of blue water. By assessing the blue water-related IWPs, this study presents data contrary to what was previously thought; IWPs in China's water-scarce northern regions (e.g., the Huang-Huai-Hai and northwestern AEZs) were much lower than those in the water-rich southern regions (e.g., AEZs in the southwest and Sichuan Basin). This is because the irrigation consumption in the north was much higher than in the south. Thus, it is suggested that the WP results can mislead policy decisions for water management, as some research has argued that products should be traded from regions with high WPs to regions with low WPs (Chapagain and Hoekstra, 2008;Dalin et al., 2014). If more wheat would be sourced from China's water-scarce northern regions, it would require more irrigation and place a huge amount of pressure on the local water sources. However, although the IWP indicator can illustrate the unsustainable water use of China's wheat production to a certain degree, this concept can also be confusing because it still fails to consider variations in the environmental relevance of water from different locations. Several previous studies have illustrated that water use indicators that do not consider environmental relevance have the potential to misinform and motivate behaviors that potentially conflict with the goal of reducing pressure on freshwater systems Ridoutt and Huang, 2012;Ridoutt and Pfister, 2010a). This study, again, illustrated that high IWPs can also occur in water-stressed regions, resulting in very high environmental impacts (mango color in Fig. 4a).
By applying both IWP and WSF indicators, this study demonstrated that the current water use for wheat production in the water-scarce northern regions, where the IWPs were low but the WSFs were high, is highly unsustainable. The hotspot regions accounted for 34% of the total wheat cropping area. Considering all the regions with high WSFs, the contribution of hotspot regions was as high as 54% (marked red, light coral and mango colors in Fig. 4a). An important conclusion to be drawn from this study is that the pressure that wheat production puts on freshwater systems arises from the current patterns of water consumption, which often occurs in highly water-stressed regions. Wheat has a very wide suitable growing zone in China (IIASA/FAO, 2012). However, this study found that more than 60% of the total national wheat production was concentrated in the water-scarce Huang-Huai-Hai region, indicating that extensive wheat production in this region prohibited water sustainability. Davis et al. (2017) identified that optimizing the global distribution of major crops can reduce the current consumptive use of blue water by 12% while increasing the production, and this was substantial compared with solutions such as improvements in crop WP and the minimization of food waste. In the case of China's wheat production, the priority for reducing this pressure can be given to optimizing the current water use patterns by redistributing the national wheat cropping. Historically, southern China was also an important contributor of wheat production. However, the wheat production in some southern regions including the AEZs in Jiangnan, Sichuan, the southwest and the south has decreased by 44% from 1980 to 2015 (NBSC, 1981-2016NBSC, 2016NBSC, 1981. Due to numerous socioeconomic factors, especially the shifting of the labor force from rural to urban and suburban areas, more and more agricultural land with high quality lie fallow each winter in south China. A land area of 2.1 × 10 7 ha, which is suitable for winter wheat production, lie fallow in the middle-lower reaches of the Yangtze basin, accounting for 46% of the arable land in the region (Bao et al., 2014;Zhai et al., 2012). Thus, China's wheat production is experiencing a paradox-wheat is mainly produced in severely water-scarce regions while the suitable farmland in the waterrich regions lie fallow. To solve this problem, a national wheat cropping adjustment is necessary and urgent.
To overcome the mismatch of water availability and water consumption between the south and the north, China has been developing approximately 20 major water transfer projects, including the world's largest-the South-North Water Transfer Project (Liu et al., 2013). Furthermore, the Chinese government made a decision in 2011 to invest approximately 630 billion USD into water conservation from 2012 to 2020 and planned to establish a pricing mechanism to encourage saving water within a decade (Huang and Yang, 2017). On cultivated land, priority has been given towards improving land productivity through developing "high-standard farmlands" that are highly droughtand flood-resistant (Huang and Yang, 2017). All these strategies can alleviate the regional water stress to a certain degree. However, this study revealed that water consumption in some places has a greater potential for environmental harm than in others. It is the consumption patterns rather than the overall water consumption that deserves more attention in policy development. Fortunately, in 2016, the Chinese government announced a pilot plan called the Winter Fallow Policy in the Heilonggang region located in the Huang-Huai-Hai region, where the groundwater level has declined rapidly in recent years due to irrigation. By providing a certain number of subsidies called Ecological Compensations, farmers in the Heilonggang region were encouraged to abandon traditional wheat production. However, the Heilonggang region only accounted for a small part of the water-scarce region (less than 2% of the hotspot regions identified by this study). Thus, the Winter Fallow Policy has little impact on the regional water use. In addition, political decisions on land and water use lead to changes in the food supply system, which can have important environmental consequences extending beyond the local region . It is therefore essential that strategies are aimed at sustainable water use to better integrate food security, the socioeconomic situation and the environment.

Conclusions
This study has highlighted the importance of putting regional freshwater issues into a national context. The water indicator results obtained in this study lead to several strategic implications for China's wheat production. First, efforts to address environmental impacts benefit from being guided by WSFs rather than other volumetric-based indicators. High WSF values highlight the need for more urgent actions. Second, the national adjustment of cropping has a high potential to alleviate regional water stress. Opportunities to increase wheat production in southern regions can be explored. Third, regional decisions could avoid unintended negative consequences by integrating national water, food and socioeconomic considerations. It is critical that policies affecting land and water use consider the wider implications of meeting national food demands. Moving beyond these strategic recommendations, further research with a narrower scope is recommended to assess the specific managerial options. This case study of China's wheat production is likely to be representative of the challenges faced by many of the world's countries, where pressures on land and water resources are high and a sustainable means of increasing food supply must be found.