Agricultural crop trade alleviates China’s water shortage but redistributes water value unevenly

Freshwater resources in China are acutely scarce relative to demand, posing a systemic challenge to sustainable development¹. China possesses about 21% of the world’s population but only 6% of global freshwater resources². Per capita renewable water availability (2100 m³) is approximately one-quarter of the global average³. Similarly, per capita arable land (0.08 ha) is less than 40% of the world average^4,5, necessitating that China support a large population with far less water and land than other countries⁶. Increasing food demand, driven by population growth, urbanization, and dietary change, further intensifies pressure on water resources^7,8,9,10. Consequently, water scarcity has become a critical constraint on China’s food security and economic development¹¹.

China also exhibits a profound spatial decoupling between water resources and cropland distribution. Although northern China comprises approximately 60% of the country’s arable land, it accounts for only 19% of national water resources^12,13. Following the onset of economic reforms, the shift of agricultural production toward the north has exacerbated regional imbalances: northern breadbasket provinces, despite their aridity, became grain-exporting centers, while water-abundant southern regions import grain^14,15,16. The government’s long-standing grain self-sufficiency policy compelled water-scarce provinces to sustain high production, often aggravating local water stress¹⁷. Consequently, China’s internal grain flows are frequently misaligned with water availability, and their net impacts on China’s water scarcity are complex¹⁸. Domestic trade patterns in China are predominantly dictated by economic drivers and policies rather than regional water resource endowments⁴. Historically, production decisions prioritized yield and economic growth, largely overlooking inherent water resource disparities¹⁹.

Virtual water trade, as the exchange of water‑intensive goods such as crops, offers a viable mechanism for reconciling spatial asymmetries between water endowment and food demand^{20,21,22,23,24}. When water‑scarce regions import food, they conserve the water that would have been required for irrigation, whereas water‑abundant regions exporting that food effectively use water on behalf of the importer^25,26,27. Prior studies demonstrate that interregional and international trade can extend resource management beyond local boundaries and help alleviate water stress, particularly by shifting production to rain‑fed regions^{28,29,30,31,32}.

However, trade flows are shaped by economic factors and policy rather than by water availability; consequently, virtual water does not always move from water‑rich to water‑poor regions³³. Indeed, water‑scarce northern provinces sometimes export large volumes of virtual water to humid southern regions, deepening their own stress³⁴. Moreover, trade not only reallocates water geographically but also redistributes the economic value derived from that water^35,36. Recent research has begun to explore how virtual water trade influences the equity of water consumption and the distribution of water benefits and burdens across regions and social groups^37,38,39.

Despite important progress, two key gaps remain. First, most assessments of water scarcity and water value are conducted at administrative scales and seldom resolve sub-provincial heterogeneity^40,41. High-resolution, grid-based analysis can reveal localized hotspots of water stress and value disparity that are invisible at provincial scales^42,43,44. Second, the spatiotemporal dynamics of water scarcity and water value under trade remain insufficiently understood, especially with respect to the interaction between blue and green water. In particular, it remains unclear how trade-driven water redistribution alters both physical scarcity and the distribution of water value over time.

To address these gaps, we leverage recent advances in hydrology and trade modeling to develop an integrated framework that couples a 1 km grid‑based crop‑water model with interprovincial and global trade matrices. This approach allows us to track changes in blue and green water consumption relative to availability and to quantify the economic value generated per unit water across China’s diverse landscapes. By comparing scenarios with and without trade for the years 2000, 2010, and 2014, we isolate the causal impacts of trade on regional water stress, map virtual water flows, and measure shifts in water value. We further analyze the relationship between these shifts and regional income levels to reveal emerging inequities—for example, low‑income, water‑scarce exporters earning little per unit water versus water‑secure regions capturing higher returns. This analysis provides a national diagnosis of the trade–water nexus and highlights policy levers to advance both water sustainability and equity.

Trade‑driven virtual water and water-value transfer

Building on the analysis of domestic trade, we further examine the impacts of international trade on China’s virtual water flows and associated economic values. As China has become a major agricultural trader, cross‑border transfers of virtual water and its economic value have grown rapidly. Net virtual water imports rose from 21.0 billion m³ (2000) to 244.1 billion m³ (2010), then grew 25.6% further by 2014 (Fig. 1). Net value imports climbed from USD 16.6 billion (2000) to USD 265.8 billion (2014). The divergence between physical volume and economic value reflects the combined effects of expanding trade scales, international commodity price volatility, shifts in crop composition, and gains in agricultural productivity. Our analysis focuses on the aggregate spatiotemporal evolution of these flows rather than decomposing the marginal contribution of each underlying driver. The net trade volume of virtual water and its value for major crops in China are detailed in Supplementary Table S1. Supplies are concentrated in the Americas—Brazil, the United States, and Argentina—whose share of China’s net virtual water imports surged from 76.9% (2000) to 95.9% (2014), and of net virtual water value from 69.2% to 88.4%. In 2014, Brazil became China’s largest supplier (134.9 billion m³, USD 70.9 billion), overtaking the United States. Shifts in trade partners—such as Japan reducing imports from China and South Korea increasing imports twenty‑fold—reflect evolving global demand and supply patterns. Japan’s net virtual water imports from China and their value contracted by factors of 4.82 and 7.25 during 2000–2014, while South Korea’s rose by 24 times volume and 16.25 times value, and Malaysia’s by 6.27 times and 3.36 times. India’s net exports to China declined by 4.28 times volume and 2.58 times value. Thailand, alongside Western European countries such as the Netherlands and Germany, shifted from net exporters to net importers. Ukraine and Kazakhstan steadily increased exports, reaching an aggregate 3.8 billion m³ and USD 2.23 billion in 2014. Australia’s net exports expanded by 13.5 times volume and 27 times value over 2000–2014.

Within China, the interprovincial pattern of virtual water trade is characterized by a pronounced coastal–inland gradient. The eastern coastal provinces are net importers, while western and northern provinces are net exporters. Between 2000 and 2014, both import volumes in net‑importing provinces and export volumes in net‑exporting provinces expanded substantially. Interprovincial virtual‑water trade and its value shifted markedly from 2000 to 2010 and again by 2014 (Fig. 2). For example, Guangdong’s net virtual‑water imports surged from 8.0 to 14.9 billion m³, and Beijing’s net virtual‑water value imports rose from USD 1.3 to 5.1 billion, signaling growing reliance on nonlocal water. Coastal hubs such as Guangdong, Jiangsu, Shandong, and Zhejiang led the rise in net imports, whereas Tibet, Xinjiang, and Qinghai remained net exporters with relatively stable volumes. The volume of interprovincial virtual water transfer nearly doubled from 2000 to 2010 and grew further by 2014, while the total value transferred increased more than tenfold. Trade redistributed water consumption toward more water-abundant and higher-productivity regions. While technical and economic parameters were treated as consistent constraints within our framework, this redistribution effectively optimized overall efficiency under the given technological baselines. Conversely, inland exporting provinces often remained or became more water-stressed as they transferred water-intensive crops to other regions. The blue and green water footprints of each province in China are presented in Supplementary Tables S2 and S3, while the blue and green water value footprints are provided in Supplementary Tables S4 and S5.

Effects on water scarcity

Trade substantially redistributes virtual water across regions, with mixed impacts on local water scarcity. Overall, WSI eased because production shifted to rain‑fed or water‑abundant regions and because China imported water‑intensive commodities (Fig. 3). These WSI changes should be interpreted as biophysical indicators of resource stress rather than as direct measures of regional economic outcomes or social welfare. In the arid regions of North and Northwest China, including Beijing, Tianjin, Inner Mongolia, Shanxi, Xinjiang, and Gansu, virtual water trade has effectively alleviated water scarcity, with significant improvements noted in 2010 and 2014. This may be attributed to the role of trade in reducing local precipitation dependency via virtual green water imports, which consequently enables the use of rainfall resources located elsewhere.

By contrast, in the southeastern coastal regions, including Guangdong, Guangxi, Fujian, and Zhejiang, both blue and green water have been virtually imported through domestic and international trade, resulting in a sustained reduction in water scarcity from 2000 to 2014. Several provinces—including Jilin, Sichuan, Hubei, Hunan, Anhui, and Jiangsu—experienced the opposite pattern: as net exporters of agricultural products, they faced worsening scarcity. At the national scale, expanding virtual-water trade—particularly green-water flows—significantly alleviated water stress, with green water contributing more than 80% of the mitigation in northern China.

Population exposure and trade-driven inequities in water value

Using a water scarcity index defined as the ratio between total water use (agricultural, industrial, and domestic) and renewable water supply minus environmental flows, we estimate how trade shifts population exposure to water stress. This assessment is based on provincial-average WSI and therefore assumes uniform exposure within each province; the results should be interpreted as a macro-scale assessment of regional physical stress rather than as a direct inference of social impacts at the individual level. From 2000 to 2014, trade systematically shifted people out of higher-scarcity classes (Fig. 4). The post-trade share of population living with WSI < 1 rose from 23.0% to 31.4%, while the share with WSI ≥ 2 (severe water stress) fell from 15.8% to 13.2%. The number of people in provinces experiencing a drop from WSI ≥ 1 to WSI < 1 increased by 68 million. Northern China saw especially large improvements. For instance, the population under severe stress in Hebei declined by 8.5 million from 2000 to 2014 as trade alleviated local water deficits. Yet trade benefits were unevenly distributed: some inland provinces (e.g., Gansu) experienced rising WSI and more people in stress categories post-trade, as their exports increased. Provinces like Heilongjiang and Inner Mongolia remained heavily water-stressed even while exporting water-intensive commodities that eased scarcity elsewhere.

Agricultural trade not only relieves irrigation pressure but also reshapes who benefits economically from water use. We assess equity using ΔCI, where positive values indicate benefits accruing relatively more to wealthier regions. The index is used here as a descriptive measure of the association between water-value flows and regional income levels rather than as a direct metric of fairness or welfare loss. Blue-water value became slightly more unequally distributed. ΔCI for the western economic zone rose from 0.15 (2000) to 0.21 (2014), indicating a stronger alignment between blue-water value gains and higher-income regions (Fig. 5). The central zone’s blue-water ΔCI also increased (from 0.09 to 0.14). By contrast, green-water value distribution grew more equitable: ΔCI for the eastern economic zone decreased from 0.22 to 0.08 (2000 to 2014). This trend suggests that the expansion of trade was associated with a narrowing gap in green-water value distribution across these regions after trade (Fig. 6). Thus, while aggregate water-efficiency gains occurred, the distribution of those gains has favored better-off provinces for blue water, even as green-water value disparities narrowed over time.

Our findings demonstrate that agricultural trade has a double-edged impact on China’s water sustainability. On the one hand, trade increases the overall efficiency of water use by allowing production to shift to regions with higher water productivity or more abundant rainfall. Virtual-water imports effectively relieve pressure on China’s scarce blue-water resources by substituting external water use for domestic irrigation demand. This has helped stabilize national water supply and support economic growth despite domestic limitations^45,46,47. However, these physically defined efficiency gains exhibit pronounced spatial asymmetry. From the perspective of trade theory, participation in trade is generally associated with economic gains for the involved parties; yet our results suggest that such gains may be accompanied by the spatial externalization of environmental pressures. On the other hand, the benefits of these efficiency gains are unevenly distributed and can externalize environmental costs. Water-rich importers concentrate high-value production locally while outsourcing water-intensive, low-value activities to poorer, drier regions⁴⁸. In effect, trade can displace water scarcity from wealthier coastal provinces to less-developed inland provinces. Changes in the spatial configuration of resource use do not necessarily imply a loss of aggregate economic welfare, as exporting provinces may receive economic compensation or development opportunities through agricultural trade. Overall, trade is associated with increased national water-use efficiency, while potentially concentrating environmental pressures in vulnerable exporting regions^49,50,51.

Moreover, international trade allows China to tap into global water resources by importing water-intensive commodities such as soybeans. These imports have helped alleviate China’s domestic water scarcity—for example, soybean imports (primarily from the Americas) embody vast volumes of water that would otherwise strain domestic supplies. However, these imports mean that water demands and environmental impacts are exported abroad. They also create dependencies on foreign supply and expose China to external risks (such as trade disputes or climate shocks affecting exporters)^52,53,54. Thus, while trade improves immediate water security, it introduces new dimensions of risk management beyond China’s borders. In short, agricultural trade has enhanced China’s water-use efficiency and food security, but it has also spatially redistributed water stresses and created external vulnerabilities.

Another implication is the importance of domestic food trade policy in tandem with water policy. Historically, grain self-sufficiency mandates and regional development strategies in China have largely overlooked inherent water-resource disparities^33,55. Our results suggest that greater coordination is needed so that water-rich regions are incentivized to produce more of the water-intensive crops. Recalibrating subsidy schemes and procurement policies to factor in regional water endowments could encourage a shift in crop production toward more humid regions when other meteorological and soil conditions are suitable. For instance, expanding maize and wheat cultivation in the relatively water-abundant northeast, while reducing it in arid northwestern provinces, would align food production with hydrological supply, although such adjustments also depend on broader constraints, including agronomic and climatic suitability, as well as policy, infrastructural, and socio-economic drivers.

In the context of climate change, which is expected to alter water availability patterns, maintaining flexibility in both domestic and international sourcing will be crucial^56,57. Climate variability could exacerbate regional water scarcity in ways that static trade patterns cannot address. Adaptive management—encompassing strategic reserves, diversified import partners, and dynamic allocation of domestic production—can therefore buffer against climate-driven shocks. Overall, integrating water considerations into agricultural planning and trade decisions offers a pathway to a more water-secure and equitable future.

The distributional impacts of virtual‑water trade are deeply intertwined with China’s socio‑economic geography. Wealthier coastal provinces, which are typically more urbanized and have higher per capita incomes, import large amounts of virtual water and thus free up local water for high‑value industrial and service sectors⁵⁸. In contrast, poorer inland provinces specialize in water‑intensive, low‑value crops and export their limited water resources, potentially reinforcing regional disparities and posing challenges to long-term rural development^59,60. From an interpretative standpoint, these patterns may reflect issues of environmental equity, implying that regions with lower economic resources might disproportionately experience local environmental impacts⁶¹. Future policies must recognize that trade can entrench inequality if not accompanied by investment in economic diversification and social support⁶².

An integrated governance approach linking water management and trade policy could help balance efficiency with distributional concerns. We propose that compensation mechanisms or payments for ecosystem services be evaluated as potential policy options, whereby importing provinces or countries provide financial or technological support to water-scarce exporting regions⁶³. Such schemes would internalize hidden water costs in commodity prices and provide resources for investments in water‑saving technologies and ecosystem restoration⁶⁴. Targeted investments, such as modern irrigation infrastructure and drought‑resistant crop varieties, can reduce withdrawals and enable reallocation of saved water to higher‑value uses, though the irrigation paradox warrants careful management⁶⁵. Water pricing reforms—using higher charges or quotas in water‑scarce exporters while protecting smallholders—could discourage low‑value, water‑intensive exports and encourage a shift toward less water‑intensive or higher‑value crops.

Because China relies on foreign water embedded in key commodities, global water sustainability and China’s food security are intertwined⁶⁶. Engaging with major suppliers to promote sustainable farming practices, diversifying import sources, building strategic grain reserves and planning for climate extremes can enhance resilience⁶⁷. As climate change alters water availability, maintaining flexibility in domestic and international sourcing will be crucial.

Aligning water conservation with food security and equity requires coordinated action at multiple scales. Domestic policies—such as water‑rights trading, ecological compensation, and differentiated water pricing—must be integrated with international trade agreements that promote sustainable production in exporting countries. The analytical framework developed here offers an empirical basis for aligning domestic instruments—such as water rights trading and ecological compensation schemes—with international trade arrangements. Global supply chains linking China to Brazil, the United States, and other partners offer opportunities to co‑develop water‑saving technologies and to address deforestation and biodiversity impacts of crop production. As climate change alters precipitation patterns and intensifies droughts, building resilient import portfolios and strategic grain reserves becomes increasingly important. These policy implications, grounded in biophysical water-stress analysis, are intended to inform more adaptive and balanced resource management strategies.

Several methodological limitations should be acknowledged. First, the analysis focuses on five major crops and three benchmark hydrological years, which capture staple food security and contrasting water conditions but do not represent continuous interannual variability or the dynamics of higher-value cash crops. Second, the framework assumes technological stationarity within each benchmark year and therefore does not capture rapid changes in irrigation efficiency, crop breeding, or management practices⁶⁸. Third, the residual value method is sensitive to input-output data and to volatility in non-water costs. Fourth, using province-level income rankings to calculate concentration index (CI) necessarily smooths intra-provincial differences⁶⁹. These constraints should be kept in mind when interpreting the robustness and generalizability of our results⁷⁰.

The apparent neutrality of technical and economic factors in some scenario comparisons reflects standardized assumptions used to isolate trade-driven effects, not the intrinsic unimportance of those factors. In practice, technical efficiency and economic valuation remain central to water governance. Future work should incorporate spatially differentiated technology pathways and explicit economic responses to better resolve their interaction with trade.

The autarky scenario is an intentionally simplified counterfactual designed to isolate the potential influence of trade. Production structures are held fixed: crop yields, technologies, and cropping patterns remain as observed. In reality, provinces would likely adjust crop portfolios, substitute toward less water-intensive products, or adopt alternative technologies under no-trade conditions. The autarky results should therefore be interpreted as an indicative upper bound on the influence of trade rather than as a realistic alternative system.

Observed trade flows are also treated as exogenous, even though trade patterns are themselves shaped by comparative advantage, prices, and water scarcity. This may overstate the extent to which trade reduces scarcity, because some observed flows are likely responses to hydrological constraints. Endogenizing trade decisions in integrated assessment or equilibrium models would allow future research to capture feedbacks between scarcity, markets, and trade.

The proportional allocation of post-trade virtual water and water value to grid cells assumes intraprovincial homogeneity. In provinces with strong internal gradients in irrigation infrastructure, climate, or scarcity, this rule may smooth or misplace local hotspots. Higher-resolution logistics information or sub-provincial input-output data would help refine these spatial attributions. Sensitivity analyses using alternative environmental-flow assumptions (60%, 70%, and 80% of runoff; Supplementary Table S6) indicate that the main distributional findings are robust, although absolute exposure levels vary. This provides additional confidence that our central conclusions do not depend on a single environmental-flow threshold.

Our high‑resolution analysis shows that trade alleviates national water scarcity by shifting production toward wetter regions and importing water‑intensive commodities, but it also widens disparities as arid, less‑developed provinces export virtual water and gain little economic return. Policymakers should integrate water, trade, and development strategies—through compensation schemes, targeted investments, pricing reforms, and international cooperation—to ensure that the benefits of trade do not come at the expense of already vulnerable regions. Our framework can be adapted to other countries and resources, offering a blueprint for managing the nexus of water, food, and trade under climate and socio‑economic change.

Water use data and modeling

We focused on five staple crops—wheat, maize, rice, soybeans, and potatoes—which collectively underpin China’s food security strategy and account for the majority of agricultural water use and virtual-water trade. To capture longitudinal shifts under contrasting hydrological conditions, we selected 2000 (relatively dry), 2010 (relatively wet), and 2014 (representing average conditions) as study years. This selection also aligns high-resolution gridded data with provincial statistics and multi-regional trade information, allowing a consistent assessment of water–trade linkages across spatial scales. Agricultural water consumption for these crops in China was quantified using a grid-based crop water model^71,72,73. The virtual water content of the five crops in major trading countries was obtained from Mialyk et al.⁷⁴. For China’s five crops, blue and green water values were determined using the residual value method, a widely used approach in water value assessments^{36,75,76,77,78}. The water value of crops in trading countries followed the approach proposed by D’Odorico et al and Zhuo et al.^36,77. The virtual water content and water value of different crops in domestic and international contexts are presented in Supplementary Figs. S1-S12. The spatial distribution of harvested areas for irrigated and rainfed crops was derived from the Global Agro-Ecological Zones (GAEZ) database and the GAEZ + 2015 dataset⁷⁹. The long-term average monthly reference evapotranspiration (ETo; mm/month; 10 arcmin; 1961–1990) was obtained from FAO and resampled to a 1–km resolution. It is divided by the number of days in the month to convert monthly values into daily values⁷¹. Daily precipitation data were sourced from the CN05 grid dataset by the Climate Change Research Center, Chinese Academy of Sciences⁸⁰.

Operating at a high spatial resolution (1 km), the model accounts for local climate, soil moisture, and irrigation conditions. Crop-specific blue water use (irrigation) and green water use (effective rainfall utilized by crops) were estimated based on crop water requirements and soil water balance simulations.

For each day, crop, and grid cell, green water consumption for rainfed crops (ET_a,i,t,c,g) and potential evapotranspiration for irrigated crops under no water stress (ET_a,i,t,c,ir) are calculated. For irrigated crops, blue water consumption (ET_a,i,t,c,b) is determined by subtracting the evapotranspiration of rainfed crops (ET_a,i,t,c,g) from that of irrigated crops (ET_a,i,t,c,ir)^71,81. The total blue and green water volumes (CWC_i,c,b and CWC_i,c,g) for crop i during the entire growing season in grid cell c are calculated as follows:

Crop evapotranspiration refers to the total water released into the atmosphere by crops through transpiration and evaporation. The daily (t) actual evapotranspiration (ET_a,i,t,c, mm/day) for crop (i) in grid cell (c) at time t was calculated⁸²:

where k_c,i,t is the crop coefficient of crop i on day t (Davis et al.⁷¹), ET_i,t,c represents potential evapotranspiration on day t in grid cell c, indicating the crop water requirement without water stress. ET_o,t,c is the reference evapotranspiration on day t in grid cell c. For irrigated crops, no water stress is assumed, and k_s,i,t is set to 1⁷².

For rainfed crops, the daily water stress coefficient k_s,i,t is calculated using the following formula⁸³:

where S_i,t represents the depth-average soil moisture (mm), calculated with a vertical soil water balance model. S_{max, i} is the maximum effective soil moisture in the root zone for crop i, and p_i is the fraction of S_max,i that crop i can extract from the root zone^73,82.

Soil moisture (S_i,t, mm) for each crop i and grid cell c is calculated using a daily soil water balance equation:

where S_{i, t-1} is the soil moisture from the previous time step, Δt is one day, P_{eff, t} is the effective precipitation (mm/day), I_{i, t} is additional irrigation water for irrigated crops (mm/day), and D_{i, t} is deep percolation below the root zone when soil moisture exceeds field capacity.

To evaluate the relationship between crop yield and evapotranspiration, the Doorenbos and Kassam equation is used⁸³, illustrating the linear relationship between crop yield and water consumption:

where K_y is the yield response factor (water stress coefficient), Y_{a, i, c} is the actual harvested yield of crop i in grid cell c during the growing season (ton ha⁻¹), and Y_{m, i} is the maximum yield (ton ha⁻¹). The maximum yield for each crop is determined by multiplying the national average yield by a factor of 1.2^84,85.

The residual value method distinguishes water’s total contribution from other agricultural production inputs⁸⁶. During production, if all other inputs have known prices, the remaining value of the product is attributed to water⁷⁵. The production value function for agricultural products is expressed as follows:

where Y_i is the yield of crop i (t/ha), P_i is the price of crop i (USD/t), CWC_i,w is the water consumption of crop i (m³/ha), P_i,w is the price of water for crop i (USD/m³), and P_i,v is the price of all other inputs except water (USD/t), including seeds, fertilizers, pesticides, machinery, technical services, labor, maintenance, and taxes.

Since water price is the only unknown variable, the price of water can be isolated from the product’s total value. According to the residual value theory, the price of water (its residual value) represents water’s contribution to the crop’s total value, denoted as V_i,w:

This method calculates the blue and green water values for various crops at each grid cell. For crop i in grid cell c, the blue and green water values are determined as follows:

where V_i,c,b and V_i,c,g represent the blue water (irrigation water) and green water (effective rainfall) values for crop i in grid cell c (USD/m³).

To assess the impact of domestic and international trade on water scarcity for five key crops in China, we quantified the Water Scarcity Index (WSI) for each grid cell based on regional water supply and consumption¹⁵:

where W_Ir is the agricultural water consumption, W_In is the industrial water consumption, W_Li is domestic water consumption; R is the total runoff, and EF is the environmental flow, set at 80% of the total runoff¹³.

The WSI was calculated before and after trade, with water scarcity classified into three levels: WSI < 1 indicates no or low water stress; 1 ≤ WSI < 2 indicates moderate water stress; and WSI ≥ 2 indicates severe water stress.

To evaluate the inequity of virtual water value due to crop trade in China, we constructed concentration curves and calculated the Concentration Index (CI) by plotting the cumulative proportion of virtual water value (blue and green water) across provinces against the cumulative population proportion ranked by per capita GDP, both before and after trade. A CI in the range of −1 ≤ CI < 0 indicates that agricultural trade alleviates the virtual water value burden on poorer populations, while a CI in the range of 0 < CI ≤ 1 suggests that trade benefits are skewed toward wealthier populations. The CI is calculated:

where L_w(p) denotes the concentration curve, n denotes the population size, i denotes the order of GDP per capita rank, W_i denotes the cumulated share of water use of the top i people, and P_i denotes the cumulated share of population of the top i people.

Trade data and scenarios

We integrated the water use model with crop trade data to assess how domestic and international trade reallocates water use across regions. For international trade matrices, we collected import data from the Food and Agriculture Organization (FAO), as customs authorities are motivated to record imports for taxation purposes. We compiled international trade matrices for five major crops—encompassing 29 agricultural commodities—to evaluate crop trade flows. Since many of these commodities are derived from the processing of primary products, we aggregated them into five staple crops to construct international trade matrices accordingly^66,87.

The international trade matrix for agricultural crops between China and other countries encompasses both primary agricultural products and their processed derivatives. The trade matrix formula is defined as:

where x represents the exporting country, y represents the importing country, j represents a processed agricultural commodity (derived from the primary crop), i represents the primary crop, t represents the trade year, T_x,y,j,t is the annual trade volume of processed commodity j from country x to country y in year t (tons), T_x,y,i,t is the total trade volume of primary crop i from country x to country y in year t (tons), and E_j is the proportion of processed product j derived from the primary crop. The extraction rates of primary products from processed goods are consistent across countries. i is derived from the set of agricultural commodities j. The study uses FAO import data for the trade matrix because customs authorities are financially incentivized to collect it for tax purposes.

For domestic trade within China, we applied a linear programming approach to simulate interprovincial trade volumes of the five major crops⁸⁸. Crop demand and self-sufficiency rates for each province were estimated based on total grain production and per capita consumption. The model aims to minimize transportation costs while satisfying crop supply and demand constraints across provinces. Based on the calculated supply-demand balance, crop surpluses from the n self-sufficient regions were allocated to the deficit provinces⁸⁹. The optimal trade paths were then analyzed using a minimum-cost origin-destination (OD) matrix generated in ArcGIS. The optimal trade routes between regions were determined using an OD cost matrix⁹⁰.

The study uses linear programming to simulate the trade volumes of five key crops among China’s provinces. The objective function is defined as follows:

where x represents crop deficit regions, y represents crop surplus regions; 31-n is the total number of regions with crop demand; n is the total number of regions with crop surplus; C_xy represents the OD cost matrix; T_xy represents the trade volume between regions (t/y); E_xy represents the crop demand of region x; D_xy represents the crop surplus of region y.

Using this information, we constructed two primary scenarios

1. Autarky (No-Trade) Scenario (Pre-Trade): Each province (and China as a whole) consumes only what it produces, with no food crop trade between provinces or internationally. In this scenario, we reallocate final crop demand back to local production. Production structures are held fixed, meaning that crop yields, technologies, and cropping patterns remain the same as in the observed baseline. This represents a counterfactual baseline in which there is no exchange of virtual water and its value – all water use and the resulting economic value occur locally at the point of consumption. It is an idealized extreme scenario designed to isolate the maximum potential influence of trade rather than a realistic alternative in practice.

2. Actual Trade Scenario (Post-Trade): This scenario uses the observed domestic and international trade patterns. Provinces and countries can import or export crops, thus shifting water and its value spatially. The net virtual water and water value trade for each province—after trade—were allocated to grid cells based on the proportion of crop-specific water use and water value in each province relative to the national total. This disaggregation enabled the estimation of virtual water and water value trade volumes at the grid level.

Water scarcity and inequity metrics

We evaluated changes in water scarcity using several metrics²³. The primary measure is the WSI for each province (or grid cell) as defined above. The impact of agricultural trade on local water scarcity—both blue and green water—was assessed by subtracting the post-trade WSI from the pre-trade WSI for each grid cell. From this, we calculated the population exposed to water stress in each scenario by overlaying population distribution on the water scarcity map (assuming population within a region experiences that region’s average scarcity). We then identified how many people are in regions that see an increase or decrease in their water scarcity level due to trade. This approach captures the social impact of water scarcity changes.

The impact of blue and green water on grid-scale water scarcity due to agricultural trade was assessed by calculating the difference between pre-trade WSI and post-trade WSI:

where WSI_pre is the water scarcity coefficient before trade. WSI_post is the water scarcity coefficient after trade.

Post-trade agricultural water use at the grid level is calculated by distributing the net import volume of blue or green water per province based on each crop’s water use share in national production:

where ({NT}{B}_{{VWC},j}) represents the net import volume of agricultural water in a province, ({T}_{{Imp},i}) and ({T}_{{Exp},i}) represent the crop import and export volumes of the trade partner and the province, respectively; (overline{{VW}{C}_{{Imp},i,j}}) and (overline{{VW}{C}_{{Exp},i,j}}) represent the average virtual water content of the crops for the trade partner and the province.

To examine inequity in water value, we calculated the economic value generated per unit of water in each region and analyzed how trade alters the distribution of these values. We quantified changes by comparing the post-trade and pre-trade concentration indices, thereby capturing how trade shifts the relationship between water value and regional income rank.

The impact of trade on water value inequity is evaluated by subtracting the post-trade concentration index from the pre-trade index:

where CI_pre is the concentration index before trade, CI_post is the concentration index after trade.

The net value of blue or green water imports for each province is allocated to grid cells according to each crop’s production water value proportion at the national level:

where ({NT}{B}_{{VVW},j}) represents the net import volume of virtual water value in a province, (overline{{VV}{W}_{{Imp},i,j}}) and (overline{{VV}{W}_{{Exp},i,j}}) represent the average agriculture water value of crops for the trade partner and the province, respectively.