High-resolution water footprints of major crops in China from cities to grids

Abstract

Rapid population and economic growth have led to an increased demand for water and food, thereby exacerbating the water scarcity crisis. Therefore, an objective assessment of water usage in crop production is crucial for ensuring national food security and alleviating water scarcity. However, the city-scale crop production water footprint (CWF) in China remains incomplete, while grid-scale CWF data are plagued by the limitations of coarse crop statistics. To fill this knowledge gap, we propose a novel methodology for developing a high-resolution inventory of CWFs. Based on this methodology, we quantified the city-scale water footprint (WF) of three major crops (rice, wheat, and maize) in China and allocated the CWFs of individual cities to 3 km × 3 km grids through a top-down downscaling approach to create a high-resolution CWF inventory. The results show that the average annual CWF of the three crops from 2000 to 2020 was in the order of rice (2.50 × 10¹² m³), maize (1.55 × 10¹² m³) , and wheat (1.21 × 10¹² m³). During the study period, the dependence on green water resources for crop production in China increased, especially for maize, which showed a relative increase of 106.76% in green water demand, in addition to optimal water use efficiency, with the dual advantages of combining high yield and low irrigation dependence. At a grid scale of 3 km × 3 km, the CWFs of the three crops followed the same order as at the city scale. Furthermore, the green water footprint (GWF) of each of the three crops increased at different rates during the study period, with maize showing a particularly significant increase of 59.26%. Meanwhile, the blue water footprint (BWF) per unit area for rice and wheat increased, while the BWF for maize decreased by 0.94%. This finding implies that maize cultivation is more efficient in utilizing rainwater resources, thereby reducing dependence on blue water. The inventory established in this study can assist in optimizing crop production in various regions of China, thereby mitigating the effects of water scarcity and facilitating sustainable agricultural development.

Introduction

Rapid socio-economic and population growth has led to a sharp increase in water withdrawal and consumption, resulting in a global water scarcity crisis^1–3. In addition, the unpredictable impacts of climate change pose increasing challenges to water sustainability^4,5. Global per capita freshwater availability has declined by more than 20% compared to the 20th century, while China’s per capita water availability is only a quarter of the global average^6,7. Furthermore, China’s water resources face three interrelated challenges: unequal temporal distribution, uneven spatial distribution, and deteriorating water quality^8,9. These factors impede the development and utilization of water resources, thereby undermining China’s sustainable socio-economic development.

Agriculture, the most significant sector in China, reached an impressive 695 million tons of crop production in 2023, responsible for 24.6% of global crop production. At the same time, agriculture accounts for 62% of the country’s total water consumption, making it China’s most important water sector¹⁰. With increasing pressure on water resources and ever-growing demand for food, the sustainable development of agriculture and food security in China represents the most critical existential challenge of the 21st century. It is therefore essential to conduct a comprehensive assessment of the actual water consumption associated with crop production to ensure food security and nutrition and to alleviate water scarcity^11,12.

As early as the 1990s, Allan and Session¹³ first introduced the concept of ‘virtual water’, defined as the amount of water used in the production of goods and services. Subsequently Hoekstra¹⁴ improved and expanded on this concept. He proposed the concept of ‘water footprint (WF)’, which is defined as the amount of virtual water contained in products produced or consumed by an individual, a household, a sector, or an entire nation. It serves as a measure of the human water resource consumption. To provide a more comprehensive and detailed indicator for assessing the amount of water consumption in the production of important goods, Mekonnen and Hoekstra² further developed the concept of the WF. The improved WF not only describes water consumption in detail, but also reveals water consumption and pollution during production or consumption, and comprehensively assesses freshwater appropriation. It is therefore regarded as a comprehensive indicator for studying the water situation from a consumption perspective^15,16. The concept has since been widely used in agriculture-related research, among which the crop production water footprint (CWF) is a method to assess crop water use and productivity by measuring the green water footprint (GWF), blue water footprint (BWF) and grey water footprint (GrWF) consumed during crop production¹⁷. The GWF is the amount of rainwater or water in the soil consumed, the BWF is the amount of water extracted from surface water or groundwater, and the GrWF is the amount of water needed to dilute pollutants^14,18,19. The CWF provides a more accurate quantification of water demand during crop production than traditional agricultural water use calculations and therefore provides more insightful recommendations for agricultural water management^20,21.

Numerous studies have calculated CWFs at various spatial scales, including global^22–24, national^25–27, basin^28–30, and sub-national administrative zone levels^31–33. At the city scale, Cao, et al.³⁴ calculated the CWF for 13 cities in Jiangsu Province from 1996 to 2015; Li, et al.³⁵ examined the intercity variations in the CWFs of nine crops grown in the Beijing-Tianjin-Hebei region of China from 2000 to 2013; and Wu, et al.³⁶ evaluated the CWFs of wheat and maize for city-level cities in the Northwest Arid Zone from 2011 to 2020; Cai, et al.³⁷ also assessed the agricultural WF in China from 2000 to 2017 and analyzed its spatial and temporal characteristics. However, CWF studies at the city scale have primarily focused on the trend of specific CWFs within particular areas or on the total CWFs in Chinese cities, neglecting a comprehensive analysis of the spatial and temporal variations in CWFs for individual crops at the city scale across mainland China. Furthermore, research on agricultural water use has shifted toward incorporating finer spatial resolution. This shift is driven by evidence from studies indicating that high resolution assessments of water scarcity enable policymakers to prioritize regions with critical water stress, which aligns with governmental initiatives promoting precision management of agricultural water resources³⁸. In pursuit of this objective, Wang, et al.³⁹ constructed an annual average CWF dataset for 21 crops at a spatial resolution of 5 acre across China from 2000 to 2018, while Wang and Shi⁴⁰ established an annual water use dataset of 15 crops at a spatial resolution of 1 km × 1 km across China from 1991 to 2019. Nonetheless, both studies suffer from the following problems. Firstly, the cropping data of all crops in these studies were calibrated using provincial statistical data, resulting in significant uncertainty when downscaling to the grid. Furthermore, the studies only considered the average annual CWF of the crops, omitting the capture of dynamic trends in water use for each crop over time.

Overall, while these studies provide valuable insights into agricultural water use, it is important to address the above limitations for enhancing the accuracy and relevance of water management strategies within the context of sustainable agricultural development. Here, we present the annual CWF (given the inaccessibility of some of the polluting data, our study only considers the GWF and BWF) of three major crops (rice, wheat, and maize) in China from 2000 to 2020 (specifically 2000, 2005, 2010, 2015, and 2020) at both the city scale and a fine-grained resolution of 3 km × 3 km. Initially, we employed the CROPWAT software to calculate the CWF of 356 cities in China, utilizing agricultural and meteorological data sourced from each city. Subsequently, we adopted a downscaling top-down methodology to construct a high-resolution CWF inventory, drawing inspiration from previous studies^41,42. Crop distribution data were selected as allocation factors to downscale the city-scale CWF to the grid scale. Lastly, we validated the uncertainty of the constructed high-resolution CWF inventory, thereby affirming the scientific rigor of our research methodology. The resulting crop water use data facilitate the accurate implementation of water-saving agricultural policies and technologies, thereby providing a scientific foundation for promoting sustainable agricultural development and formulating food security strategies.

Materials and methods

Overall framework

Figure 1 shows the overall framework of the study: (1) collecting and processing of data, including climate remote sensing data, literature data and crop planting distribution data, (2) calculating CWF for 356 Chinese cities from 2000 to 2020 based on climate data and literature data, (3) reclassifying the resolution of the three crop distribution maps to 3 km × 3 km and using them as allocation factors to downscale city-scale CWF to 3 km × 3 km, (4) verifying the uncertainty of creating CWF inventory.

Fig. 1.

The overall framework of the study.

Study area

China (3°31′00″−53°33′47″N, 73°29′59.79″−135°2′30″E) has a land area of approximately 9.6 × 10⁶ km² dominated mainly by temperate climatic conditions, with tropical climatic conditions prevailing in smaller areas. The study area covers mainland China, excluding Hong Kong, Macao, Taiwan, Sansha and some cities in Xinjiang. The area features a complex geographical environment, diverse crop planting structures, various planting patterns, and different planting habits (Fig. 2)^43–45. Rice, wheat, and maize are the major food crops in China. Production of these crops totaled 607 million tons, accounting for 90.7% of China’s total crop production⁴⁶. Therefore, these three crops were included in this study.

Fig. 2.

Planting structure of China’s three staple crops and geographical zoning map. (Source: author, generated using ArcGIS 10.8; https://desktop.arcgis.com/zh-cn/desktop/index.html).

Methodology

There are two paradigms in the CWF accounting system: total water footprint (TWF) and water use efficiency (WUE), which are complementary in terms of policy functions⁴⁷. The former characterizes the spatial pattern of regional water stress in terms of absolute water consumption, which serves the goal of “total control” (e.g., cross-basin water allocation); the latter analyses the efficiency of agricultural production through the water-yield ratio, which supports the “efficiency optimization” decision-making (e.g., promotion of water-saving technologies). Based on the urgent need to control the red line of total water use in China, this study mainly chooses TWF as the core indicator.

Quantifying CWF at city scale

This study utilizes the CROPWAT 8.0 (https://cropwat.informer.com/8.0/) model, developed by the Food and Agriculture Organization of the United Nations (FAO), to quantify the GWF and BWF of crops separately. Where GWF is calculated mainly based on the effective precipitation and evapotranspiration demand of the crop during its reproductive period, BWF is the amount of irrigation water consumed in the form of evapotranspiration during the biological period of the crop.

Referring to the WF calculation method proposed by Hoekstra, et al.⁴⁸, the GWF and BWF for the production are calculated using Eqs. (1)-(7) as follows.

1

2

3

4

5

6

7

where Y is the yield of the calculated crop (t); GWD stands for green water dependency and usually refers to the degree of dependence on green water (%); ET_green (mm), ET_blue (mm) denote the fraction of crop evapotranspiration from effective rainfall and irrigation water, respectively; H is the planted area of the calculated crop (ha); coefficient 10 converts water depth (mm) to water volume per unit land area (m³ ha^-1); P_e is the effective precipitation (mm); ET_c is crop evapotranspiration.

The ET_c is calculated using Eqs. (8) and (9) as follows.

8

where, K_c is the crop coefficient and ET₀ (m³ day⁻¹) is the reference crop evapotranspiration according to the Penman–Monteith formula recommended by FAO⁴⁹.

9

where R_n is net crop surface radiation (MJ m⁻² day⁻¹); G is soil heat flux (MJ m⁻² day⁻¹); T is average daily air temperature at 2 m height (℃); u₂ is wind speed at 2 m height (m s⁻¹); e_s is saturated water vapor pressure (kPa); e_a is actual water vapor pressure (kPa); is slope of saturated water vapor pressure curve (kPa ℃⁻¹); γ is dry and wet table constants (kPa ℃⁻¹).

We used the SCS (Soil Conservation Service) method proposed by the United States Department of Agriculture (USDA) in the CROPWAT model to calculate the effective precipitation⁵⁰. The formula is provided in Eqs. (10) as follows.

10

where P_a is the monthly average precipitation (mm).

Based on the above algorithm, we obtained the city-scale CWF in China from 2000 to 2020.

Creating high resolution CWF inventory through downscaling

A major challenge in building a high-resolution CWF inventory is how to allocate the city-scale CWF, which is calculated using CROPWAT, to a more detailed spatial scale. In this study, we employed a top-down downscaling methodology to assign CWFs of cities to grid cells that intersect with the city, utilizing a high-resolution allocation factor map. The allocation factor ratio, a value ranging from greater than 0 and less than or equal to 1, specifies the proportion of CWF in an area (a city in this study) that is allocated to a particular model grid cell (a 9 km² square in this study). Since the area of a given city encompassed multiple grid cells, allocation factors are needed to indicate the fraction of WF assigned to each individual grid cell. These allocation factors are created based on a Geographic Information System (GIS) shapefile, which includes high-resolution spatial distribution maps for rice, wheat, and maize.

The original resolution of the three crop distribution maps sourced from the online database is 1 km × 1 km, which need to be resampled to the required 3 km × 3 km resolution using the ArcGIS 10.8 software (https://desktop.arcgis.com/zh-cn/desktop/index.html). The 3 km × 3 km grids were generated using the Create Fishnet tool in ArcGIS. Furthermore, because some grids may be split among two or more cities, the grid cells must be allocated among cities using the Identity tool according to the percentage of each grid cell that falls within the boundaries of each city.

The principle of allocating a CWF based on crop acreage is shown in Eqs. (11) and Eqs. (12).

11

12

where , denote the GWF and BWF consumed by crop i in the production process of grid k of city j, respectively; denotes the planting area of crop i in grid k of city j; m denotes the number of grids contained in city j; and denote the total amount of green water and blue water consumed by crop i in the production process of city j, respectively.

Data sources

This study focuses on three staple crops in China: rice, wheat, and maize. Spatial distribution of crops with a resolution of 1 km × 1 km for three crops in China from 2000 to 2020 is provided by the Resource and Environmental Science Data Centre of the Chinese Academy of Sciences (RESDC) (https://www.resdc.cn). The CROPWAT software is employed to calculate the city-level CWF. To perform this calculation, it requires specific input data, including agrometeorological and crop parameters. Regarding the data sources, statistics on crop acreage (ha) and yield (t) are obtained from provincial and municipal statistical yearbooks. Meanwhile, agrometeorological data, such as average temperature, light hours, wind speed, relative humidity and rainfall, along with soil data, are obtained from the China Meteorological website (https://www.cma.gov.cn). Table 1 shows the main data information and sources.

Table 1.

Data types, spatiotemporal resolution, and sources.

Data type	Temporal resolution	Spatial resolution	Source
Crop acreage	Yearly	City-level	Provincial Statistical Yearbook and the Municipal Statistical Yearbook
Crop yield
Air temperature	Daily	0.0625°	China Meteorological Data Network (https://data.cma.cn)
Solar radiation
Wind speed
Relative humidity
Precipitation
Soil heat flux
Distribution of rice planting	Yearly	500 m	National Ecosystem Science Data Center (https://www.resdc.cn)
Distribution of maize planting		1 km
Distribution of wheat planting		1 km

Results and discussion

Interannual variation of CWF in China

Figure 3 shows the consumption structure and annual changes of GWF and BWF of China’s three staple crops from 2000 to 2020. In the terms of the annual average level, the GWF of rice production (GWF_r), BWF of rice production (BWF_r), GWF of wheat production (GWF_w), BWF of wheat production (BWF_w), GWF of maize production (GWF_m) and BWF of maize production (BWF_m) in China were 1.49 × 10¹² m³, 1.01 × 10¹² m³, 5.14 × 10¹¹ m³, 6.93 × 10¹¹ m³, 1.14 × 10¹² m³ and 4.04 × 10¹¹ m³, accounting for 28.41%, 19.22%, 9.79%, 13.18%, 21.72%, and 7.68% of the total CWF of the three crops respectively. Rice production water consumption accounted for 47.30% of the total production water consumption of the three crops, which is similar to the results of previous studies⁵¹. The study obtained that the proportion of GWF consumed by the three types of crop production in China was 59.9%, which is slightly lower than the estimate of Cao, et al.⁵² that the GWF of Chinese agricultural production accounted for more than 65% of the total CWF. Despite the differences in the study results, they reflect the pivotal role of green water resources in Chinese agricultural production.

Fig. 3.

The consumption structure and annual changes of GWF and BWF of China’s three staple crops (rice, wheat, and maize) from 2000 to 2020.

Figure 3 also reveals the changing characteristics of the CWF. From 2000 to 2020, the total CWF associated with the three crops shows an increasing trend. The total CWF increased from 4.80 × 10¹² m³ in 2000 to 5.97 × 10¹² m³ in 2020. During the study period, the GWF of rice and wheat showed a relatively consistent trend, with respective increase of 15.21% and 17.17%. However, their respective shares of the total CWF of the three crops decreased from 29.00% and 10.09% in 2000 to 26.90% and 9.52% in 2020. The BWF of rice and wheat showed a decreasing trend during the study period, with respective decreases of 0.03% and 12.61%. Consequently, their respective shares of the total CWF decreased from 21.83% and 17.04% in 2000 to 17.57% and 12.00% in 2020. In contrast to these two crops, the total GWF and BWF values of maize in China increased significantly during the study period, with respective increases of 106.76% and 54.07%. Consequently, their shares of the total CWF of the three crops increased from 15.76% and 6.27% in 2000 to 26.23% and 7.78% in 2020, respectively. The primary reason for this trend is that the demand for maize in China is over 220 million metric tons and continues to grow due to the rigid feed demand from the development of livestock and the increasing demand for industrial processing. To address this issue, the central government issued the first Central Document No. 1 on Agriculture in the 21st Century in 2004, proposing measures such as the abolition of agricultural taxes, the introduction of policies to support and promote agriculture, and a temporary maize storage system. With the release of the annual Central Document No. 1 on Agriculture, the increase in maize sown area and production between 2000 and 2020 was as high as 78.97% and 145.91%, respectively^53,54.

Spatial and temporal pattern of CWF at the city scale

This study selected five specific years, 2000, 2005, 2010, 2015, and 2020, and used ArcGIS software to map the spatial distribution of the GWF of rice, wheat, and maize in 356 cities in China, based on the results of CWF calculations. From 2000 to 2020, the GWF_r showed a spatial distribution of higher values in the south and lower values in the north (Fig. 4a), which was consistent with the current situation that the amount of water resources is mainly concentrated in the south. During the study period, the most significant augmentation in the GWF_r was in the northeast region, where the national contribution increased from 3.98% to 13.8%. Cities such as Shuangyashan, Baicheng, Hegang, Jiamusi, Jixi and Qiqihar experienced increases of more than 500%, which is intimately correlated to the region’s industrial structural adjustment and the government’s adoption of various types of incentives to promote agricultural development. The center of gravity of the GWF_w resided primarily in east China and central China, in the south-eastern direction of China. These two regions accounted for 69.03% of the total national GWF_w (Fig. 4b). Henan Province is a large wheat-producing province, characterized by extensive wheat cultivation areas, high production volumes, and a substantial GWF. Among the cities, Zhumadian, Nanyang, and Zhoukou exhibited highest GWF_w during the study period, with annual average GWF_w of more than 1.40 × 10¹⁰ m³. Between 2000 and 2020, the center of gravity of China’s GWF_m underwent significant changes (Fig. 4c), with a distinct trend of northern expansion. This was mainly contributed to the continuous increase in maize area and production in the north region. During the study period, the proportion of GWF_m in the northeast region increased from 19.06% to 34.97% of the national total, with Heihe, Shuangyashan, Jixi, and Hegang all increasing by more than 1000%. In contrast, the GWF_m in the southwest region decreased from 23.46% to 2.65%, with the most pronounced declines observed in Diqing, Baoshan, and Ya’an. Overall, more than 1/2 of the provincial-level administrative regions showed an increase in their GWF. The three northern provinces of Heilongjiang, Jilin, and Inner Mongolia demonstrated the most significant growth in total GWF over the study period, with increases exceeding 150%. Conversely, Beijing, Tibet, and Zhejiang exhibited the most notable reductions, with decreases exceeding 48% (Fig. 4d).

Fig. 4.

Spatial and temporal distribution patterns of city-scale GWFs for three crops (rice, wheat, and maize) (a-c) and percentage changes in total GWFs of the three crops by province (d) in the selected years of China in 2000, 2005, 2010, 2015, and 2020. (Source: author, generated using ArcGIS 10.8; https://desktop.arcgis.com/zh-cn/desktop/index.html).

The spatial distribution of GWF for the three crops in 2000–2020 was consistent with the trend of GWF (Fig. 5). However, significant variations were observed in both the GWF and BWF among the various crops. Generally, the GWFs of rice and maize exceeded their corresponding BWFs, with ratios of approximately 1.5:1 and 2.8:1, respectively. Conversely, the BWF of wheat was about 1.4 times higher than its GWF. Between 2000 and 2020, the proportion of BWF_r in the northeast region increased from 10.9% to 22.2% (Fig. 5a). Notably, Jiamusi experienced the largest increase in BWF_r during the study period, with a rise of 518.8%. Henan, Shandong, and Hebei emerged as the leading contributors to the BWF_w (Fig. 5b), accounting for over half of the national aggregate, and all ten cities that exhibited the highest BWF_w during the study period were located within these three provinces. The BWF_m displayed a distinct spatial pattern characterized by “high in the north and low in the south” during 2000–2020 (Fig. 5c). Inner Mongolia, Heilongjiang, and Xinjiang were the three provinces with the highest BWF_w. Specifically, Chifeng in Inner Mongolia had the highest BWF_w throughout the study period, about 1.25 × 10¹⁰ m³ a⁻¹. The total BWF of the three crops across all provinces during the study period was comparable to that of GWF, but the increase was slightly smaller compared to GWF. Additionally, 1/3 of the provincial-scale administrative regions exhibited an increase in their BWF. Heilongjiang, Xinjiang, and Jilin emerged as the three fastest-growing provinces, experiencing increases exceeding 20%. Conversely, Beijing, Zhejiang, and Shanghai, three economically developed provincial-level administrative regions, witnessed the most significant decrease, by more than 49%.

Fig. 5.

Spatial and temporal distribution patterns of city-scale BWFs for three crops (rice, wheat, and maize) (a-c) and percentage changes in total BWFs of the three crops by province (d) in the selected years of China in 2000, 2005, 2010, 2015, and 2020. (Source: author, generated using ArcGIS 10.8; https://desktop.arcgis.com/zh-cn/desktop/index.html).

The results of this study demonstrate that the GWF of rice and maize exceeded that the BWF, whereas the GWF of wheat was less than the BWF. Maize exhibited the highest ratio of green water consumption among the three crops under study. Maize cultivated in the provinces of Sichuan, Guizhou, Yunnan, Chongqing, Guangxi, Jiangsu, Hunan, Hubei, and Anhui was entirely rain-fed, with the proportion of maize’s green water requirement approaching 100%. This is due to effective precipitation fulfilling the water requirement for the growth and development period of the maize. In the northern region of China, maize cultivation predominantly occurs during the rainy season, with precipitation meeting 70–80% of the crop’s water requirements⁵⁵. The double-season rice cropping areas in South China (Guangxi, Guangdong, Fujian, etc.) and the single-season and double-season rice cropping areas in Central China (Jiangsu, Hunan, Zhejiang, etc.) represent the primary rice-producing regions in China, collectively accounting for 86.4% of the country. Annual precipitation in the majority of these regions falls between 1,300 and 1,600 mm, sufficient to satisfy over 60% of the water demand for rice cultivation^56,57. Wheat cultivation in China is primarily concentrated in the North China Plain, particularly in western Shandong, southern Hebei, central-eastern Henan, and northern Anhui. Winter wheat in the North China Plain is predominantly cultivated during the dry season and the main growing period, which overlaps with the rainfall period, despite a relatively low water demand during the crucial March–May period. Since most of the precipitation in the North China Plain during the period is below 80 mm, the effective precipitation is much lower than the wheat water demand, so the region’s winter wheat is mainly dependent on groundwater irrigation^58,59. It is noteworthy that recent studies have indicated an increase in the demand for irrigation water for crop production in recent years. This trend is attributed to intensified crop evapotranspiration, which is a consequence of warmer temperatures and reduced precipitation. Moreover, this trend will intensify in the foreseeable future^60,61.

3 km × 3 km gridded inventory of CWF

Based on the methodology introduced in section"Methodology", the WF of each crop production at the city scale was allocated to the 3 km × 3 km grids by high-resolution rice, wheat, and maize cultivation distribution maps, thereby enhancing comparability among different regions⁶². Figure 6 and Fig. 7 show the GWFs and BWFs of rice, wheat, and maize on a 3 km × 3 km grid for the selected years. As can be seen in Fig. 6a-c, the vast majority of the crop grid-scale GWF for the selected years were less than 1.00 × 10⁷ m³ and the overall GWF of each crop unit grid showed GWF_r (1.36 × 10⁷ m³·a⁻¹) > GWF_m (1.00 × 10⁷ m³·a⁻¹) > GWF_w (7.39 × 10⁶ m³·a⁻¹). The GWF per unit grid of rice was higher than that of the other two crops. This result is related to the cropping pattern of the crops themselves, in addition to the climatic conditions in the main growing areas of the three crops described above and the degree of coupling of the crop growth cycle with the rainfall period. Rice is typically cultivated using a flooded method, which requires the field to maintain a certain water depth throughout its growth cycle. Conversely, wheat and maize are grown in drylands, where wheat has a longer growth cycle and maize is grown on farmland with higher drainage requirements.

Fig. 6.

Spatial and temporal patterns of GWF for (a) rice, (b) wheat, and (c) maize at 3 km × 3 km resolution in the selected years of 2000, 2005, 2010, 2015, and 2020. (Source: author, generated using ArcGIS 10.8; https://desktop.arcgis.com/zh-cn/desktop/index.html).

Fig. 7.

Spatial and temporal patterns of BWF for (a) rice, (b) wheat, and (c) maize at 3 km × 3 km resolution in the selected years of 2000, 2005, 2010, 2015, and 2020. (Source: author, generated using ArcGIS 10.8; https://desktop.arcgis.com/zh-cn/desktop/index.html).

Throughout the study period, the primary regions exhibited larger GWF_r encompassed Lishui and Wenzhou in Zhejiang Province, along with Suizhou in Hubei province. Similarly, significant GWF_w was concentrated in Zhejiang, Anhui, Hubei, and Inner Mongolia provinces. For maize production, the grids with notable GWFs spanned Shandong, Gansu, Hebei, and Shaanxi provinces. Notably, the per-unit grid GWFs demonstrated substantial growth trends for all three crops, with rice, wheat, and maize experiencing respective increases of 21.77%, 59.26%, and 33.74%. The grid-scale GWFs of various crops exhibited considerable variability, which can be attributed to the differing characteristics of each crop. Generally, the distribution of grid-scale GWFs among the three crops showed similarity, all adhering to lognormal distributions but with varying shapes and maximum widths at around the lower quartile. Nonetheless, the box plot and median values of grid-scale rice GWFs were higher compared to the other two crops, with a correspondingly wider range (Fig. 8a).

Fig. 8.

Temporal variability of (a) GWF and (b) BWF for rice, wheat, and maize at 3 km × 3 km resolution in the selected years of 2000, 2005, 2010, 2015, and 2020.

Figure 7 shows that the vast majority of the crop grid-scale BWF for the selected years were less than 5.00 × 10⁶ m³ and the total BWF of each crop unit grid exhibited a hierarchy, with BWF_w (9.59 × 10⁶ m³·a⁻¹) > BWF_r (8.98 × 10⁶ m³·a⁻¹) > BWF_m (3.21 × 10⁶ m³·a⁻¹). The spatial distribution of grids with higher BWF_r was concentrated in Yunnan, Inner Mongolia, and Guangxi provinces, whereas those for wheat production were distributed in Inner Mongolia, Ningxia, and Shandong. Similarly, grids with notable BWF_m were distributed in Ningxia, Xinjiang, and Gansu provinces. Throughout the study period, the BWF_r and BWF_w per unit of grid experienced varying rates of increase, but the increase was much lower compared to the GWF, which was 6.14% and 14.29%, respectively. The BWF_m per unit grid decreased slightly, by 0.94%. In terms of the overall distribution of the data, the BWFs at the crop grid scale exhibited slightly differences compared to the GWFs (Fig. 8b). Furthermore, the BWFs of rice and maize at the grid scale were narrower in scope, yet they demonstrated great consistency with GWFs. The distribution of grid-scale BWFs for wheat was larger and more dispersed than that of GWFs. The BWF of crop production shown by the results does not match the yield space, with northern China producing about 60% of the crop but only 17.72% of the water resources. In contrast, South China, which holds 82.82% of China’s water reserves, produced only 40% of cereals in 2017. Therefore, seeking harmonization between water resources and crop production is crucial for sustainable crop production in China.

Relationship between water use pattern and production performance of crops

In addition, this study has carried out in-depth analyses of the GWD, unit area yield (UAY) and WUE for rice, wheat, and maize, averaged annually from 2000 to 2020 (Fig. 9). Figure 9a demonstrates that in southern regions (e.g., Jiangxi, Guangxi, Hainan), rice exhibited high GWD (> 60%), with UAY generally below 6 t/ha. In northern irrigated regions (e.g., Xinjiang and Ningxia), irrigation typically satisfied more than 90% of water demand during the growing period (GWD < 10%), leading to higher UAY (> 8 t/ha). However, high evapotranspiration rates reduced the irrigation water-use efficiency, resulting in higher WUE (> 10,000 m³/t). Heilongjiang was an exception, where a highly efficient irrigation system enabled simultaneous achievement of low WUE (8,263.66 m³/t) and high UAY (10.51 t/ha). Figure 9b shows that the UAY of wheat was significantly lower (< 2.5 t/ha) in regions with high GWD (> 95%; e.g., Hunan, Jiangxi), whereas in highly efficient irrigated regions (e.g., Henan, Shandong, Tibet), an optimal balance between UAY (> 5.5 t/ha) and WUE (< 10,000 m³/t) was achieved under a moderate GWD (35%−40%). These findings emphasize the high reliance of wheat on irrigation. Figure 9c reveals that maize exhibited the most advantageous water-use pattern, characterized by high GWD (> 80%), high UAY (> 5 t/ha), and low WUE (< 9,000 m³/t) in most provinces. The synergistic improvement of these three indicators was particularly evident in Sichuan, Shandong, and Henan through an optimized cropping system.

Fig. 9.

Three-dimensional coupling analysis of GWD, UAY, and WUE for (a) rice, (b) wheat, and (c) maize based on annual averages from 2000 to 2020.

Significant differences were found among crops. Rice was the most irrigation-dependent, with high yields but high blue water use due to high-intensity irrigation in northwestern production areas (e.g., Xinjiang). Wheat was generally irrigation-dependent (with an average green water use of only 42.1%). Maize was stable in most environments, particularly in southwestern areas (e.g., Sichuan), where more than 94% of the green water was used. These results confirm the dual advantage of maize as both a water-saving and high-yielding crop.

Uncertainties and limitations

Several factors may contribute to uncertainties in high resolution CWF inventory. Various factors, including activity data, allocation factors, completeness, missing data, and measurement mistakes, can provide varying degrees of uncertainty. Nonetheless, certain uncertainties remain unquantifiable owing to technical limitations. The primary source of uncertainty in this study stems from the distribution coefficients, which will be analyzed and discussed subsequently.

Thirty groups of tests were conducted to assess the uncertainty associated with the allocation factor. In each group, three provinces (approximately 30–40 cities) were randomly selected to establish a corresponding between crop acreage and the accompanying city-scale CWF. The results are shown in Fig. 10, which demonstrates that the coefficients of determination for each group of data range between 0.9119 and 0.9956, with an average of 0.9721. This indicates a strong correlation between the crop acreage and CWF, affirming the scientific validity of the downscaling method in this study. Notably, the coefficient of determination obtained from this test fitting is lower than what would be observed in the real-world scenarios. This discrepancy arises from the larger variations in climatic conditions and crop cultivation habits among randomly selected cities are larger than the differences between grids within the same city.

Fig. 10.

Correlation between crop acreage and CWF in 2000, 2005, 2010, 2015 and 2020: (a) GWF and (b) BWF.

Furthermore, we acknowledge the primary limitations of our study. Firstly, the analysis of the CWF only considered the consumption of water resources, neglecting the impact of crop production on water quality. This omission limits the holistic understanding of water-related environmental pressures. Secondly, due to the absence of high-resolution ET₀ data, our study was unable to directly perform grid-scale accounting. Instead, it chosed to use allocation factors to distribute city-scale WFs to the grid scale, ignoring variations in conditions such as climate, soil types, and elevations among grids within a city. Additionally, the study did not fully incorporate yield data across varying regions, which could result in an overestimation of water pressure in high-yield areas or an underestimation of the efficiency limitations in low-yield areas.

Conclusions and policy implications

This study quantified the WFs of the three major crops (rice, wheat, and maize) in China at the city and 3 km × 3 km grid scale in 2000, 2005, 2010, 2015, and 2020. The primary results show that:

Regarding the total national CWF of each crop over the study period, rice, maize, and wheat ranked in terms of water consumption, accounting for 47.63%, 29.20%, and 22.96% of the total CWF of the three crops, respectively. Notably, maize exhibited the most rapid increase in CWF during the study period, with an increase of 91.77%. In terms of crop water use composition, many provinces had more than 80% GWF, while rice and wheat showed a stronger trade-off between yield and irrigation needs. Overall, China’s major crop production had increasingly relied on green water during the study period, with the proportion of GWF consumption rising from 54.85% to 62.66%. Under this trend, the promotion of rain-fed synchronized cropping systems and the construction of rainwater harvesting systems can further optimize the use of green water. In particular, maize, with its dual advantages of water conservation and high yield, is a priority crop for agricultural restructuring.

From a regional perspective, the GWF and BWF of the three crops in China showed significant temporal and spatial disparities. The GWF of crop production was low in the west and high in the east, with the focus had gradually migrated from the southwest to the northeast throughout the study period. The areas with the highest BWF for crop production were predominantly located in eastern, central, and northeast China. To address these spatial patterns, it is recommended that locally adapted water conservation strategies be adopted to promote decoupling of agricultural production from blue water dependence: 1) Implementing precision irrigation technologies in Yangtze River mid-lower reaches (e.g., Suizhou, Lishui)³⁷; 2) Cultivating drought-resistant cultivars in southern double-cropping rice systems (e.g., Nanning, Zhanjiang); 3) Adopting moisture-retention mulching practices in North China’s groundwater depletion areas (e.g., Hengshui, Zhangjiakou); and 4) Establishing maize-soybean rotation systems in northeastern black soil regions (e.g., Harbin, Jiamusi)⁶³.

At the 3 km × 3 km grid scale, the ratios of GWFs among rice, wheat, and maize were 1.8:1.4:1, while the ratios of BWFs were 2.8:3:1 for rice, wheat, and maize, respectively. The WFs of rice, wheat, and maize per unit area increased to varying degrees between 2000 and 2020, with the largest increase in the WF being for wheat, at 31.07%. Over the study period, maize demonstrated the highest increase in GWF, reaching 59.26%, while it was the only crop among the three to exhibit a decrease in BWF of 0.94%. Based on the CWF results at the grid scale, precise irrigation and data-driven governance are key measures to achieve refined agricultural water management. These include: 1) Deploying IoT-enabled soil moisture sensors coupled with AI-optimized irrigation systems in high-BWF grids (e.g., Weihai and Taizhou with BWF_w > 1 × 10⁸ m³/grid); 2) Developing a national 3 km WF monitoring platform enabling real-time visualization of WF hotspots (e.g., Longnan and Tangshan exhibiting GWF_m > 5 × 10⁸ m³/grid) for dynamic resource allocation.

The results of this study provide valuable insights into the spatial and temporal dynamics of water resource utilization in China’s agricultural production, facilitating the adoption of precision agricultural management practices, the formulation of locally tailored agro-ecological management, and providing a scientific rationale for advancing sustainable agricultural development. Additionally, the study highlights potential novel research avenues and innovative management strategies for future endeavors in agricultural water resource management.

Future research should incorporate GrWF metrics to quantify water quality degradation while improving spatial precision through high-resolution ET₀ datasets. Concurrently, policymakers must support open-access hydrological data platforms and region-specific incentives linking subsidies to WF reductions, enabling sustainable intensification across diverse agroecological contexts.

Author contributions

J.Z.: Writing – original draft, Methodology, Investigation. L.W.: Data curation, Writing – review & editing. X.W.: Writing – review & editing. Y.Y.: Writing – review & editing. K.H.: Writing – review & editing, Supervision, Investigation.

Funding

This research was supported by National Key R&D Program of China (2023YFC3205600).

Data availability

The data used or analyzed during this study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.