Global water stress mitigation achieved through international crop trade

Summary

Global water savings can be achieved by trading crops from countries with higher to lower water productivity. However, strengthening such water-saving trade links could intensify global water stress if exports come from water-stressed countries to less stressed ones. Here, we explore whether international crop trade can alleviate global water stress using a virtual scarce water saving/loss indicator and refined trade matrices for 109 crops across 150 countries. We further assess how differences in water productivity and stress between trade partners mitigate global water stress by categorizing different types of crop trade relationships. Our results indicate that while international crop trade generally helps mitigate global water stress, over half of the trade links still contribute to increased water stress. Scenario analysis suggests that enhancing crop water productivity among exporters involved in virtual scarce water loss trade links could convert up to 53% of these loss links into saving links.

Graphical abstract

Highlights

• •Crop trade contributes to both global water savings and scarce water savings
• •Though 52% of trade links raise water stress, their overall global impact is moderate
• •Raising exporter productivity may turn 53% of scarce water loss links into savings

Earth sciences; Agricultural science; Agriculture in international trade; Sustainability aspects of food production

Introduction

Crop production is the largest contributor to global freshwater use, accounting for 72% of the world’s water withdrawals.¹ Due to the worldwide uneven distribution of water resources, many countries grapple with meeting crop demand while ensuring the sustainability of their water resources. International crop trade enables countries to import virtual water through importing water-intensive crops,² offering a way to alleviate local water scarcity. For example, Egypt, Iran, and South Africa have adopted virtual water trade strategies to replace domestic production, thereby reducing local water use for agriculture.^3,4,5 Indeed, about a quarter of the crop produced by humans is traded internationally,^6,7 in which 13% of the global water is virtually embodied in traded crops.² While importing countries benefit from virtual water trade by saving local water resources and alleviating regional water stress,^8,9,10 exporting countries, however, face intensified water scarcity due to substantial crop exports.^{11,12,13,14,15} This trade becomes particularly unsustainable when virtual water flows from water-scarce exporters to water-abundant importers.¹⁶ It is found that approximately 43% of global virtual water exports originate from water-scarce regions,¹⁷ which drives the depletion of surface water^14,18 and groundwater in these areas.^19,20,21,22

While studies have highlighted the contradiction between the regional water-saving potential of importing water-intensive products and the aggregated water stress in exporting regions,^10,11 they also indicate that international crop trade may contribute to global water savings. Such savings are achieved when crops flow from countries with higher water productivity to countries with lower water productivity.^{23,24,25,26,27} Otherwise, the importing countries might use more water if they choose to produce the imported crops.²⁷ Many studies have demonstrated that international crop trade results in global water savings.^{28,29,30,31,32} However, these assessments do not take into account the regional differences in water endowments. When a trade link is established between the exporter facing extreme water scarcity and the importer with abundant water resources, the volumetric water savings may occur, but at the expense of exacerbating water scarcity in the exporting regions.^26,27Hence, the global water saving assessment based on the volumetric flows of virtual water may be misleading through encouraging the water-saving trade links, which may in turn increase regional or even global water stress. To address this issue, Pfister et al.^33,34,35 developed a regional water stress index to weight water volumes by water scarcity. The metric that integrates virtual water flows and water stress index is referred to as “virtual scarce water”.³⁶ A higher virtual scarce water value represents a greater potential for water resource depletion.^34,37 Many studies have since assessed the impacts of crop trade on regional and global water stress through the lens of virtual scarce water.^{36,38,39,40,41,42,43} Building on this metric, Zhao et al.²⁶ further quantified virtual scarce water savings from interprovincial virtual water flows in China. Their findings revealed that while interprovincial trade resulted in 14.2 km³ of water loss without considering water stress, the scarce water loss was only 0.4 km³ when using the virtual scarce water concept. More recently, Wu et al.⁴⁴ examined the global trade of wheat, maize, and rice and their impacts on virtual scarce water. They discovered that trade in wheat and maize contributed to global scarce water savings, whereas rice trade exacerbated global water stress. These studies highlight the importance of robust measurements, regional considerations, and the specific commodities involved in determining the impact of virtual water trade on scarce water resources.⁴⁵ It is thus important to direct international crop trade toward reducing water stress at both regional and global scales through a refined framework considering regional water endowment differences.

To achieve the water stress mitigation through international trade, an important way is to use water resources in water rich region more productively.⁴⁶ In other words, the “ideal trade links” should originate from regions with higher crop water productivity and lower water stress to those with lower crop water productivity and higher water stress.²⁶ However, there is still lack of analysis identifying the hotspots in international crop trade links that align with or deviate from the “ideal trade links”. Previous studies have partly addressed this issue at sub-national level²⁶ or with limited crop categories.⁴⁴ Moreover, while it has been found that variations in water stress levels and water productivity between trade partners affect the role of crop trade in alleviating global water stress, the impact of these two factors on water stress reduction across different trade links remains unexplored to date. Answering these questions is crucial for guiding global crop trade toward mitigating water stress through the identification of key trade links that contribute to increased global water stress and explore potential to reduce these negative impacts.

Here, we assess if international crop trade may mitigate global water stress and investigate the role of both water productivity and water stress differences in mitigating the stress through classifying different types of crop trade links. We further investigate the potential of directing the trade links toward global water mitigation through water productivity improvement. Our study is distinguished from previous frameworks by (1) covering a more detailed list of countries and crops (109 crops and 150 countries), which can help countries to develop more targeted water saving policies for different crops; (2) introducing the origin-tracing algorithm developed by Kastner et al.,^47,48 the compiled bilateral trade matrices are exempted from re-exports and therefore prevent the problem of double counting trade links; (3) classifying the trade links into six types to determine how trade links in different regions are affected by water productivity and water stress; (4) setting scenarios to reveal the potential of water productivity improvement in reducing virtual scarce water loss through international crop trade.

Results

Virtual scarce water flows of international crop trade

Globally, the virtual scarce water flows derived from international crop trade was 152.74 km³ in 2018, accounting for 16.29% of the water scarcity footprint of global crop production. In comparison, the virtual water flows were 249.01 km³, accounting for 18.25% of the water footprint of global crop production. The top net exporters of virtual scarce water were found in India (26.48 km³), the USA (18.62 km³), and Pakistan (13.9 km³) (Figure 1). The majority of India’s net virtual scarce water exports were embodied in seed cotton (53.82%), rice (19.35%), and other oil crops (17.29%), mainly flowing to China, Bangladesh, and Viet Nam. The net virtual scarce water exports from the USA were mainly to its four main trade partners China, Mexico, Japan, and Viet Nam via seed cotton (49.51%), rice (39.63%), and soybeans (10.7%). Pakistan exported virtual scarce water through the trade of rice and wheat products (71.77% and 23.38%, respectively), mainly to Indonesia, Afghanistan, and Germany.

Figure 1.

Global net virtual scarce water imports and major net virtual scarce water flows through international crop trade in 2018

Countries with net virtual scarce water exports are red and countries with net virtual scarce water imports are blue. Only large net virtual scarce water flows ($\ge$1 km³) are shown.

The largest net virtual scarce water importers were China (9.17 km³), Germany (6.73 km³), and South Korea (5.26 km³). China’s largest net virtual scarce water imports (90.40%) were embodied in the trade of seed cotton mainly from India, the USA, and Australia. Germany imported virtual scarce water via rice, vegetables and fruits, and nuts (46.38%, 11.55%, and 10.78%, respectively) mainly from Pakistan, Spain, and the USA. For South Korea, 70% of its net virtual scarce water was imported through the trade of rice, other oil crops, and pulses mainly from China, India, and the USA.

Global virtual scarce water saving from international crop trade

Assuming the imported crops were all produced by the imported countries, the total amount of virtual scarce water required was 348.07 km³. These crop products were, however, being produced using only 152.74 km³ of virtual scarce water in the exporting countries. As a result, international crop trade generated 195.33 km³ of virtual scarce water saving, almost 1.28 times the volume of virtual scarce water flows. Such results suggested that international crop trade has helped mitigate water stress globally. Comparatively, the volumetric virtual water savings through international crop trade without considering national water stress differences was 292.94 km³, also exceeding the volumetrically virtual water flows (1.18 times).

Exports from the USA (35.08 km³), Brazil (34.05 km³), Argentina (32.46 km³), Ukraine (28.44 km³), and Russia (28.11 km³) contributed significantly to global virtual scarce water savings, mainly through trade in soybeans, maize, and wheat (Figure 2A). These countries were the major crop producers with abundant water resources and significant water productivity advantages. In contrast, exports from Pakistan (−11.61 km³), Kazakhstan (−6.35 km³), and India (−5.85 km³) led to the largest virtual scarce water losses mainly through the trade of rice, soybeans, and seed cotton. These three countries were water-stressed and didn’t have water productivity advantages for the crops they exported.

Figure 2.

National contribution to global virtual scarce water savings/losses

(A) Top 20 countries generating virtual scarce water savings/losses through exports.

(B) Top 20 countries generating virtual scarce water savings/losses through imports. The bubble colors represent the dominant crop contributing the most to virtual scarce water savings or losses, and the bubble sizes indicate the volume of virtual scarce water savings or losses generated by the dominant crops. For simplicity, abbreviations are used to represent different countries, with the corresponding names are shown in Table S1.

Looking at the importing countries (Figure 2B), crop imports from water-stressed countries, such as Egypt (32.02 km³), Saudi Arabia (24.16 km³), and Iran (20.22km³) contributed to 39.11% of global virtual scarce water savings through trade of wheat, barley, maize, and soybeans. While water-abundant importing countries like Germany (−5.35 km³), France (−3.72 km³), and Japan (−3.32 km³) resulted in virtual scarce water losses. It is worth noting that the volumes of virtual scarce water losses associated with these imports were relatively moderate in scale.

In terms of the crop category (Table 1), trade of maize (49.68 km³), soybeans (48.28 km³), and wheat (38.43km³) generated the largest virtual scarce water savings. While trade of seed cotton (−2.76 km³) and sugar crops (−1.55 km³) generated the largest virtual scarce water losses.

Table 1.

Crop-specific contribution to global virtual scarce water saving or loss (the detailed contribution of 109 crops can be seen in Table S2)

	Crops	Virtual water flows(km3)	Virtual scarce water flows (km3)	Virtual water saving(km3)	Virtual scarce water saving (km3)	Trade (Mt)
	Wheat	12.34	8.07	79.74	38.43	208.66
	Rice	59.40	35.14	12.30	4.05	113.57
	Barley	2.85	1.64	12.54	11.60	53.66
	Maize	12.21	4.53	53.35	49.68	183.24
	Other cereals	2.28	1.27	3.56	2.50	22.59
	Roots and Tubers	4.78	2.65	5.56	2.87	210.87
	Sugar crops	5.96	4.11	−0.79	−1.55	148.65
	Pulses	12.51	7.22	1.28	2.48	43.41
	Nuts	10.32	6.26	3.24	1.69	9.60
	Soybeans	16.69	8.70	73.98	48.28	208.22
	Seed cotton	61.78	41.13	−0.12	−2.76	52.39
	Other oil Crops	16.87	11.15	30.55	25.10	105.31
	Vegetables and fruits	23.93	17.29	10.98	6.65	185.21
	Stimulates	3.84	1.40	5.68	5.40	8.06
	Spices	1.77	1.58	1.18	0.76	2.35
	Other crops	1.47	0.60	−0.07	0.15	4.29
	Total	249.01	152.74	292.94	195.33	1560.1

Contribution of trade links to global virtual scarce water savings and losses

We investigated the crop trade links that contributed to global virtual scarce water savings or losses. Considering the net trade flows of each crop category between any two countries as a link, there were a total of 415,652 crop trade links globally. About 47% of the total trade links contributed to global virtual scarce water savings (286.77 km³). While more than half of the total trade links (52%) contributed to global virtual scarce water losses (−91.44 km³). The remaining 2,384 links neither contributed virtual scarce water savings nor losses.

When the water productivity of two trading partners is equal, trade flowing from regions with lower water scarcity to those with higher water scarcity results in scarce water savings, while the reverse leads to a loss. Similarly, when water scarcity levels are equal, trade from regions with higher to lower water productivity yields scarce water savings, whereas the opposite direction results in a loss. By jointly considering differences in both water productivity and water scarcity between trading partners, six distinct outcomes can be identified. Accordingly, we further classified trade links into six types based on the relative differences in water productivity and water stress, as illustrated in Figure 3.

Figure 3.

Tree map of different types of trade links

Links: the number of trade links. We exclude trade links that produce neither virtual scarce water savings nor virtual scarce water losses; trade: the total trade volume of trade links (Million ton); saving/loss: virtual scarce water savings or losses (km³). The sizes of different types are allocated by the number of trade links.

Among the six types, the ideal type for virtual scarce water saving through trade is the crops flows from regions with higher water productivity and lower water scarcity to regions with lower water productivity and higher water scarcity (type 4).²⁶ This type of links may effectively utilize water resources and alleviate regional water stress. In contrast, the least favorable trade type flows from regions with lower water productivity and higher water scarcity to regions with higher water productivity and lower water scarcity (type 1). Some links originate from regions with lower water productivity and lower water scarcity that flowing toward regions with higher levels of both. These trade links result in volumetric water loss due to the lower water productivity of the exporting regions. However, the outcomes in terms of scarce water savings vary. Type 6 scarce water saving occurs when the saving effect of exporting from regions with lower water scarcity outweighs the losses caused by their lower water productivity. Conversely, when this effect does not outweigh the losses caused by water productivity differences, a scarce water loss occurs (Type 3). Comparatively, trade links originating from regions with higher water productivity and higher water stress toward those with lower levels generate volumetric water savings. However, in terms of scarce water savings, these trade links may lead to two different outcomes. Type 5 scarce water saving occurs when the saving effect of exporting from regions with higher water productivity outweighs the losses caused by their higher water stress. Otherwise, when this effect does not outweigh the losses caused by water stress differences, a scarce water loss occurs, categorized as type 2. Examples of such shifts for types 2 and 6 can be found in Figure S1.

We found that 30.46% of the crop trade links (125,895) were ideal virtual scarce water saving links (type 4), generating 270.8 km³ of global virtual scarce water saving. The top 20 trade links contributed 26% of the total virtual scarce water savings (Table S3), exhibiting clear geographic concentration with directional dominance: mainly exporting from major agricultural producers like Brazil, the USA, Argentina, and Russia, to water-stressed destinations including Yemen, Spain, Saudi Arabia, and Egypt. Soybeans and maize were the most traded crops within these trade links. The combined numbers of types 5 and 6 links accounted for 35% of the total virtual scarce water savings links, but they only contributed to 5% of the total scarce water savings. Such disproportion between the quantity of trade links and their scarce water saving contributions is predominantly constrained by exporters’ water productivity levels or water stress limitations. For example, China had the highest number of net export links (1,469) in type 5, primarily exporting crops to Eastern and Northern European. Despite demonstrating higher water productivity in exported crops such as fruits and vegetables, and cereals, China’s higher water stress levels relative to importers constrained its virtual scarce water savings to 0.06 km³. Similarly, in type 6, Tanzania, Kenya, Rwanda, and Guatemala collectively formed the majority of trade links (2,659), yet contributed only 0.07 km³ to virtual scarce water savings. This limited contribution stemmed from these countries’ low water productivity in exported crops (particularly fruits and vegetables), despite their abundant water resources (WSI between 0.01 and 0.02).

The number of virtual scarce water loss links (type 1, 2, and 3) were larger than that of virtual scarce water saving links, but their trade volumes and virtual scarce water losses were relatively small compared to the virtual scarce water saving links. In type 1, the top virtual scarce water loss links originated from India, Kazakhstan, the USA, and were directed to China, Vietnam, and Turkey, through the trade of seed cotton, soybeans (Table S3). Virtual scarce water losses in type 2 were also significantly contributed by the USA and India, and mainly through exports of rice and seed cotton. The exporting countries involved in these links not only faced severe or extreme water stress but also fell below the global average in water productivity. Despite challenges such as diverse climatic conditions, land management practices, and lack of irrigation equipment,^49,50 countries like India, Kazakhstan, and Pakistan continued to export substantial quantities of crops, driven by extensive farmland and strong economic incentives.^51,52 For type 3, the largest virtual scarce water losses occurred in trade links among Kazakhstan, Turkey, Australia, China, and the USA. Although these links satisfy the “low-to-high water stress” transfer condition, they predominantly involve either intra-trade between moderate water stress nations or flows to severely water-stressed countries, where exporters’ water stress advantages remain insignificant.

Potential for improving water productivity to reduce global virtual scarce water loss

We found a large number of virtual scarce water loss links had relatively small volumes of losses compared to the savings, and 92.63% of these losses were generated without satisfying the ideal water productivity differences (i.e., 84.7 km³, type 1 and type 3). Hence, we explored the feasibility of reducing virtual scarce water losses by improving crop water productivity of exporters in three scenarios. We did not consider changes in water stress situation due to methodological constraints and data availability, which was further specified in limitations. We adopted the highest global water productivity level for each crop to characterize the upper limit of potential improvements in crop water productivity, following the approached proposed by Mekonnen and Hoekstra.⁵³

First, we introduced scenario 1 to investigate the extent to which water scarcity can be alleviated through improving crop water productivity of exporters in the virtual scarce water loss links to the highest global level (links in type 1,2, and 3). A total of 200,842 trade links of virtual scarce water losses has the potential of increasing their crop water productivity for exporters. The improvement in crop water productivity would enable around 53% of the virtual scarce water loss trade links transform from losses to savings, resulting in an increase of 69.78 km³ in virtual scarce water savings. While the remaining 47% of virtual scarce water loss links cannot be converted into savings, improving the water productivity of exporters could still reduce virtual scarce water losses by 29.68 km³. Hence, the total global virtual scarce water saving under scenario 1 would thus amount to 291.79 km³ (Figure 4, Table S4). Improving crop water productivity in India, the USA, Kazakhstan, Pakistan, and Egypt could yield the most significant scarce water saving benefits. Specifically, the number of trade links originating from these five countries account for only 1% (9,551) of all convertible links that transition from virtual scarce water losses to savings; however, they contribute 62% (42.98 km³) of the total virtual scarce water savings generated by such links (Figure S2A). This is because these links are associated with large crop export volumes, where enhanced water productivity translates to greater water-saving potential. In terms of crops, only 2% of trade links capable of transitioning from virtual scarce water losses to savings derived from seed cotton (Figure S2B). However, once enhanced water productivity of exporters (particularly India), these seed cotton trade links could achieve remarkable virtual scarce water savings (29.76 km³). For virtual scarce water loss links related to soybean, barley, and wheat, water productivity improvements demonstrate greater effectiveness, with 94%, 86%, and 84% of improvable loss links in these categories successfully transitioning to savings, respectively.

Figure 4.

Virtual scarce water saving/loss of different scenarios

Scenario 1: enhance the crop water productivity of exporters in Type 1,2,3 trade links to the highest global level; scenario 2: enhance the crop water productivity of exporters in Type 1,2,3,6 trade links to the highest global level; scenario3: improve the crop water productivity of exporters in type 1,2,3 trade links until zero virtual scarce water loss achieved. VSW: virtual scarce water.

Furthermore, we noticed that though type 6 achieved virtual scarce water savings, the water productivity of exporting countries in this type remained lower than that of importing countries. Therefore, we proposed scenario 2 to further enhance the crop water productivity of exporters in type 6 trade links to the highest global level, building on the foundation of scenario 1. Compared to scenario 1, scenario 2 increases virtual water scarce savings by an additional 2.3 km³, resulting in a total global virtual scarce water savings of 294.09 km³. Notably, just two trade links—maize exports from the USA to Mexico and seed cotton flows from Turkey to Syria—account for 47% of the total incremental virtual scarce water savings achievable in type 6 through water productivity improvements.

Recognizing that maximizing water productivity for exporters is challenging to achieve across all countries. Alternatively, we set the goal to achieving zero losses, rather than focusing on reversing losses into savings. As a result, exporters who originally have the potential to shift from losses to savings will see a smaller increase in water productivity (scenario 3). Compared to scenarios 1 and scenario 2, the virtual scarce water savings generated by scenario 3 are smaller (250.57 km³). However, compared to scenario 1, 62% of the virtual scarce water loss links could experience up to 20% of smaller increase in required water productivity. This efficiency gain was particularly pronounced for exports from India, the USA, Kazakhstan, Pakistan, and Egypt, where necessary water productivity enhancements decreased by 16.8–20.2% on average—achieving substantial water savings while lowering financial and technological barriers faced by nations in improving water productivity.

Discussion

Identifying key trade links contributing to global water stress mitigation

International crop trade plays a pivotal role in alleviating hunger,^54,55 enhancing food diversity,^6,56 and closing the nutrient gap.^57,58 Our study demonstrates that international crop trade has alleviated global water stress, offering additional support for the benefits of trade. However, global challenges and geopolitical tensions, such as the COVID-19 pandemic,⁵⁹ the Russia-Ukraine conflict,⁶⁰ and US-China trade disputes,⁶¹ has increased in recent years, contributing to a rise in trade protectionism. In response to these challenges, advancing global crop trade governance is essential to ensure equitable access to international crop markets. Key strategies include: (1) establishing platforms for diplomatic resolution of trade disputes⁶²; (2) strengthening regional trade frameworks such as the African Continental Free Trade Area and the Regional Comprehensive Economic Partnership⁶³; (3) fostering resilient supply chains through improvements in infrastructure, digital connectivity, and storage capacities^64,65; and (4) promoting the trade of a diverse range of crops to enhance resilience against targeted protectionism.⁶⁶

We found that international crop trade alleviates global water scarcity, primarily due to two key factors: (a) exporting countries exhibit higher crop water productivity than importing countries; and (b) exporting countries experience lower water stress compared to importing countries. The scenario analysis further indicated that focusing solely on improving crop water productivity in exporting countries involved in virtual scarce water loss links could significantly mitigate the effect of international crop trade on water stress aggravation. Given that there are still many exporting countries exhibiting relatively lower water productivity in agricultural production,^67,68 it is important to enhance crop water productivity of these countries through advancements in agricultural technology and management, such as crop selection, soil fertility enhancement, weed and pest control, and unnecessary transpiration reduction (e.g., improving irrigation techniques, reducing tillage, and soil mulching).^69,70,71 In addition, extensive crop exports can expose exporting countries to greater risks of exacerbated water stress in the long term. To address these challenges these countries need to extend their water supply from multiple sources, such as desalination, inter-basin water transfer, rainwater harvesting, and water reuse.^72,73

Although enhancing crop water productivity and ensuring a reliable water supply for exporting countries present promising strategies for mitigating water stress through international crop trade, achieving these goals remains challenging. First, increasing crop water productivity and ensuring water supply require substantial investments in the extensive construction of water infrastructure and the development of water saving technologies, which are difficult to achieve for low- and middle-income countries.⁷⁴ Second, climate change compromises the effectiveness of improvements in crop water productivity. For example, rising temperatures may increase water intensity for certain crops.⁷⁵ Third, measures aimed at improving crop water productivity often entail unintended consequences. For example, efforts to enhance crop water productivity can paradoxically result in increased water consumption.⁷⁶

For crop importing countries, importing crops would alleviate their water stress no matter such importation may result in global water stress mitigation or not. Hence, for countries adopting a virtual water strategy—where water-scarce nations address their water shortages by importing water-intensive commodities—such imports may unintentionally exacerbate global water stress and the water stress in exporting countries if the importing nations are part of the virtual scarce water loss links. In addition, virtual water strategies can have other negative impacts on exporting countries, including biodiversity loss and reduced carbon sequestration caused by agricultural expansion in these regions.^77,78,79 Hence, it is crucial to identify imports that exacerbate water stress globally or/and in exporting countries and to develop policy instruments that hold importers accountable for the increased water stress resulting from their imports. Our study thus provides a framework for identifying specific trade links that contribute to exacerbating water stress globally or/and in exporting countries.

Limitations of the study

Our study is subject to several limitations and uncertainties. Firstly, we utilized the global crop virtual water contents (VWCs) from 1996 to 2005, as provided by Mekonnen and Hoekstra.⁸⁰ This dataset contained numerous missing values and did not accurately represent updated crop VWCs in 2018. Secondly, we assessed the potential to reduce virtual scarce water losses through improvements in crop water productivity among exporters at the national level. However, crop water productivity within a single country may exhibit substantial spatial heterogeneity, influenced by geographic factors such as soil properties,⁸¹ climatic conditions,⁸² and topographic features.⁸³ Incorporating such spatial variability into trade links would require high-resolution, crop-specific datasets and would increase the complexity of our trade model,⁸⁴ which is beyond the scope of the current study. However, we acknowledge the importance of accounting for these geographic dimensions to translate theoretical water-saving potentials into nationally actionable policies—a key focus of our future research. Thirdly, we assumed the regional water stress remain unchanged in scenario settings, which may underestimate the mitigation of water scarcity through crop trade. Water stress is defined by the ratio of total annual freshwater withdrawals to water availability. However, the water stress index (WSI) provided by Pfister et al.³³ is based on average multi-year precipitation conditions and is not sensitive to changes in water availability. When agricultural water consumption is reduced by decreasing crop VWCs, the available water in the region increases, resulting in an overall decrease in regional water stress. The WSI we used did not reflect this dynamic, leading to an underestimation of the reduction in virtual scarce water loss. Finally, our study does not account for feed crops and livestock. Trade in livestock products contributes 17% of international virtual water flows and have be found to contribute to virtual water saving.^25,85 A critical aspect of evaluating water resources involved in livestock trade is converting livestock products into equivalent feed crops. However, previous studies often assumed that livestock are fed using domestic resources,^86,87 whereas in reality, some countries may rely on imported crops for feeding. Consequently, identifying the origin of livestock producers and their environmental impacts proves complex.

Resource availability

Lead contact

Further information and requests for resources and data should be directed to and will be fulfilled by the lead contact, Xu Zhao (xuzhao@sdu.edu.cn).

Materials availability

This study did not generate new unique materials.

Data and code availability

• •This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the National Key R&D Program of China (2023YFE0113000); the National Natural Science Foundation of China (no. 72074136, 72104129, and 72033005); Major grant in National Social Sciences of China (23VRC037 and 24VHQ018); the Taishan Scholar Youth Expert Program of Shandong Province (no. tsqn202103020, tsqnz20221106); Shandong Provincial Natural Science Foundation (ZR2024MG008).

Author contributions

X.Zhao designed the study. Y.S., X.Zhao, and R.Z. performed the experiments and computational analyses. Y.S., X.Zhao, X.Zhang drafted the manuscript. X.Zhao, M.R.T., and H.Z. supervised this study. All authors discussed the results and revised the manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Crop production data	Food and Agriculture Organization Statistical Database (FAOSTAT)88	http://www.fao.org/faostat/en/#data/QC
Bilateral trade data	Food and Agriculture Organization Statistical Database (FAOSTAT)88	https://www.fao.org/faostat/en/#data/TM
Technical conversion factors	Technical Conversion Factors for Agricultural Commodities89	https://www.fao.org/fileadmin/templates/ess/documents/methodology/tcf.pdf
Virtual water content	Mekonnen and Hoekstra80	https://www.waterfootprint.org/resources/Report47-WaterFootprintCrops-Vol1.pdf
Water stress index	Pfister et al.33	https://pubs.acsS.org/doi/full/10.1021/es802423e
Software and algorithms
Python software	Version 3.9.12	https://www.python.org/downloads/release/python-3912/

Experimental model and study participant details

We focused on the global virtual scarce water flows in 2018 in order to avoid potential interference from the COVID-19 pandemic. We analyzed international crop trade for 461 crops, including primary crops and processed crops, across 150 countries. Trade data reported by the FAO can appear twice—once by exporters and once by importers. Discrepancies between import and export volumes may occur due to time lags or variations in the quality of national reporting. In this study, we considered import data to be more reliable.⁹⁰ The import and production data for the crops were obtained from FAOSTAT.⁸⁸

We aggregated all processed crops into 109 equivalent primary crops using extraction rate downloaded from the FAO. To simplify the analysis, we classified these primary crops into 10 categories based on the classification by Mekonnen and Hokestra⁸⁰: cereals, roots and tubers, sugar crops, pulses, nuts, other oil crops, vegetables and fruits, stimulants, spices, and other crops. Six crops with substantial trade volumes are examined in detail: wheat, rice, barley, maize, soybeans, and seed cotton (Table S2).

Virtual water content (VWC) measures the volume of water used to produce per ton of crop product.⁹¹ VWC can be divided into blue VWC, green VWC, and grey VWC. We only considered blue VWC for each crop in this study. Data on crop VWCs was obtained from Mekonnen and Hokestra,⁸⁰ who provided the blue VWCs for 145 crops from 1996-2005. This data is widely cited in crop-related water resources studies such as Kummu et al.,⁹² Tamea et al.,⁹³ and Halpern et al.⁹⁴ For missing VWC values for certain countries and crops, we used regional average VWCs to fill the gap.

We used the water stress index (WSI) data for 150 countries, calculated by Pfister et al.³³ The WSI was adjusted to a logistic function with continuous values between 0.01 and 1. According to Pfister et al.,³³ WSI can be classified into four levels: Minor (0.01-0.09), Moderate (0.09-0.5), Severe (0.5-0.91), and Extreme (0.91-1).

Method details

Adjusting international crop trade data

First, we convert the processed products into their equivalent primary crops based on the extraction rate.²¹

$$\mathrm{R}=\left(1/\mathrm{R}\left(\mathrm{c}\right)\right)\times \left(\mathrm{R}\left(\mathrm{c}\right)/\sum _{{\mathrm{c}}^{\prime }\in \text{branch}}\mathrm{R}\left({\mathrm{c}}^{\prime }\right)\right)$$ (Equation 1)

where, $\mathrm{R}$ is the conversion factor from processed products to primary products, $\mathrm{R}\left(\mathrm{c}\right)$ is the extraction rate of commodity $\mathrm{c}$, which represents the amount of the processed product obtained from processing the primary product. The second term on the right-hand side is designed to prevent double-counting of processed products from the same category. For example, wheat can be processed simultaneously into wheat flour, wheat bran, and wheat germ, with extraction rates of 0.79, 0.18, and 0.02, respectively. Then the conversion factor of these three products into wheat is 1.01. The extraction rates for processed products can be downloaded from the FAO⁸⁹ (see key resources table), and the corresponding conversion factors can refer to Table S5.

Second, we compiled the bilateral trade matrices of crops and applied the origin-tracing algorithm to adjust trade data. The raw trade data from the FAO faces the issue of distinguishing between exports and re-exports, which can lead to double counting of re-exports. The algorithm used in this study addresses this issue by assuming that imports and domestic production contribute proportionally to exports and domestic consumption. The specific applying procedure is described in Kastner et al.^47,48

Quantifying global virtual scarce water saving/loss

This study applied the approach proposed by Zhao et al.²⁶ to quantify the global virtual scarce water saving/loss through international crop trade. It can be shown as follows:

$$\text{SW}{\mathrm{S}}^{\mathrm{g}}=\sum _{\mathrm{r}\ne \mathrm{s}}\sum _{\mathrm{i}}[\left(\text{HVS}{\mathrm{W}}_{\mathrm{i}}^{\mathrm{s}}-\text{VS}{\mathrm{W}}_{\mathrm{i}}^{\text{rs}}\right)+\left(\text{HVS}{\mathrm{W}}_{\mathrm{i}}^{\mathrm{r}}-\text{VS}{\mathrm{W}}_{\mathrm{i}}^{\text{sr}}\right)]$$ (Equation 2)

Where $\text{SW}{\mathrm{S}}^{\mathrm{g}}$ means global virtual scarce water savings or losses through international crop trade. The positive value of $\text{SW}{\mathrm{S}}^{\mathrm{g}}$ means global virtual scarce water savings, whereas the negative value represents additional virtual scarce water losses. $\text{HVS}{\mathrm{W}}_{\mathrm{i}}^{\mathrm{s}}$ is hypothetical virtual scarce water use in country $\mathrm{s}$ if country $\mathrm{s}$ chooses to produce the traded product $\mathrm{i}$ domestically, and it can be calculated as $\text{HVS}{\mathrm{W}}_{\mathrm{i}}^{\mathrm{s}}=\text{WS}{\mathrm{I}}^{\mathrm{s}}\times \text{VW}{\mathrm{C}}_{\mathrm{i}}^{\mathrm{s}}\times {\mathrm{T}}_{\mathrm{i}}^{\text{rs}}$. $\text{WS}{\mathrm{I}}^{\mathrm{s}}$ is the water stress index of country $\mathrm{s}$. $\text{VW}{\mathrm{C}}_{\mathrm{i}}^{\mathrm{s}}$ is the virtual water content of product $\mathrm{i}$ in country $\mathrm{s}$. ${\mathrm{T}}_{\mathrm{i}}^{\text{rs}}$ is the trade of product $\mathrm{i}$ from country $\mathrm{r}$ to country $\mathrm{s}$. $\text{VS}{\mathrm{W}}_{\mathrm{i}}^{\text{rs}}$ refers to real virtual scarce water flows from country $\mathrm{r}$ to country $\mathrm{s}$, and $\text{VS}{\mathrm{W}}_{\mathrm{i}}^{\text{rs}}=\text{WS}{\mathrm{I}}^{\mathrm{r}}\times \text{VW}{\mathrm{C}}_{\mathrm{i}}^{\mathrm{r}}\times {\mathrm{T}}_{\mathrm{i}}^{\text{rs}}$.

Setting the highest crop water productivity level

To determine the highest global crop water productivity level, we adapted the water footprint benchmark developed by Mekonnen and Hoekstra.⁵³ This indicator provides information for formulating water use reduction targets in crop production, and has been widely used in cases on China,⁹⁵ Iran,⁹⁶ and the USA.⁹⁷

For each crop, we ranked the VWC values for all relevant countries from smallest to largest and plotted these values against the cumulative percentage of the corresponding trade volumes. A case illustrating the highest water productivity level for rice is shown in Figure S3. From the graph, we could read the VWC values at different trade volume percentiles. We considered VWC at the 10% trade volume percentiles as representing the highest global crop water productivity. The selection of the 10% threshold is cautious. Studies have shown that there is no clear correlation between crop VWC and a country’s income or aridity.⁵³ While high-income countries may have greater capacity to implement best available water saving technology and practices, less developed and tropical countries may inherently exhibit lower VWC levels in specific crops. Therefore, achieving VWC within the top 10% of current global trade volume is technically feasible everywhere.

Scenario setting

Three scenarios are introduced to explore the feasibility of reducing virtual scarce water losses by improving crop water productivity of exporters. The specific scenarios are set up as follows:

Scenario1: to investigate the extent to which water scarcity can be alleviated through enhancing the crop water productivity of exporters in the virtual scarce water loss links to the highest global level (links in Type 1,2, and 3);

Scenario2: further enhance the crop water productivity of exporters in Type 6 trade links to the highest global level in the basis of Scenario 1 (links in Type 1,2, 3, and 6);

Scenario3: set the goal to achieving zero losses instead of pursuing the reverse from losses to savings (links in Type 1,2, and 3).

Quantification and statistical analysis

Effects of reexport adjustments

To assess the effects of reexport adjustments on global trade analyses, we compared the global export volumes acquired from directly reported trade flows and reexport-adjusted trade flows. With FAO’s report data, the global total export volume is 1,594 Mt, However, after adjusting for re-exports, this volume decreases to 1,560 Mt. Large export attribution difference can be found in the Austria, Belgium, Malaysia, Netherlands, and the UK (Figure S4.). These countries usually use import crops for re-export, but are counted by FAO as the country of origin of the crop.

It is important to notice that there is also a change in the virtual scarce water saving linked to exporting. Typically, the original planting countries possess advantages in crop water productivity and water endowments compared to the re-exporting countries. This further increases the global water savings with contributions from exporting countries. For example, compared to the reexport-adjusted virtual scarce water savings, the savings calculated using FAO’s report data for exporting countries like Argentina, Brazil, China, Ukraine, and the USA have been underestimated. Overall, the misallocation of re-exports affects the results of water savings for individual countries, which in turn affects the water policy guidance of the countries concerned. The algorithm we used, to some extent, compensates for this shortcoming and provides refined data support for crop management.

Brief overview of data comparisons with the literature

Differences in hydrological models and the handling of trade data account for the various results in water-saving accounting. To facilitate comparisons with previous water savings results, we provide volumetric water flow and volumetric water savings data for selected focus crops without considering water stress, as shown in Table S6.

We estimated that in 2018, 249 km³ of virtual water flows were embodied in the international crop trade and generated 293 km³ of virtual water savings. For individual crops, Mekonnen and Hoekstra,⁸⁰ Liu et al.,³² Konar et al.,⁹⁸ and Fader et al.²⁹ reported that the trade of wheat, maize, barley, and soybeans led to virtual water saving, which is consistent with our estimate. However, there is debate regarding the impact of the rice trade on virtual water savings. Mekonnen and Hoekstra⁸⁰ and Konar et al.⁹⁸ reported that the rice trade resulted in virtual water losses. Our study, along with Liu et al.,³² found that the rice trade may help in virtual water savings. Previous studies rarely included seed cotton. Mekonnen and Hoekstra⁸⁰ estimated it in their comprehensive study and found that the trade of seed cotton saved volumetric water, but our study found that it actually resulted in virtual water loss. As a water-intensive crop mainly export from water-scarce regions like India and Pakistan, the trade of seed cotton has the possibility to cause virtual water losses.

Cases of trade links in type 2/6

We identified 22,058 trade links that generated 4.12 km³ of virtual water losses but resulted in 1.81 km³ of virtual scarce water savings (Type 6). For example, maize trade between the USA and Mexico (Figure S1A) initially showed a virtual water loss of 0.03 km³ without accounting for water stress. However, considering water stress, it generated a virtual scarce water saving of 0.26 km³. In this case, the USA and Mexico had similar water productivity, but the USA experienced moderate water stress while Mexico faced severe water stress. Consequently, maize exports from the USA to Mexico contributed to global virtual scarce water savings from a scarce water perspective.

In contrast, 53,880 trade links generated 6.74km³ of virtual scarce water losses but resulted in 15.53km³ of virtual water savings without considering water stress (Type 2). Take the India’s net exports to Bangladesh through the trade of seed cotton for an example (Figure S1B). This link generated 1.03 km³ of virtual water saving but 0.29 km³ of virtual scarce water loss. As one of the major seed cotton exporters, India had a higher water productivity than Bangladesh, which led to virtual water saving. However, India was under extreme water stress whilst Bangladesh had moderate water stress, more virtual scarce water embodied in seed cotton was exported from India to Bangladesh than under hypothetical condition, resulting in virtual scarce water loss.

Published: June 13, 2025