Virtual water and its effect in Huaihe River basin: an analysis using a dynamic multi-region computable general equilibrium model

A dynamic multi-region computable general equilibrium (CGE) model is used to analyze virtual water and its effect in the Huaihe River basin, which includes Jiangsu, Anhui, Shandong and Henan provinces. The baseline scenariowas first constructed to describe the trade situation and virtual water among different regions and sectors from 2009–2020. Then two hypothetical policy scenarios were set to analyze the economic, water consumption and virtual water trade impacts under different policy shocks. The results show that (1) the Huaihe River basin was a net virtual water outflow region, with 13.977 × 10 m virtual water transferred outside the river basin in 2009, and 26.653 × 10 m VW transferred to outside in 2020. Jiangsu is the largest virtual water outflow province (mainly flows to outside), followed by Anhui. Anhui suffers from severewater shortage, but provides a net virtual water outflow to Henan, Jiangsu and Shandong and outside; (2) reducing Anhui’s agricultural output (S1 scenario) and Anhui’s agricultural virtual water outflow (S2 scenario) can not only reduce virtual water trade but also savewater, but S1 scenario reduced net virtual water outflow by more, and savedmore water. S1 had a positive effect on gross domestic product (GDP), but reduced the employment rate.


INTRODUCTION
Many countries are suffering from water scarcity and it affects a third of the people in the world. Water scarcity is a serious problem in China and has been getting worse in recent years. China started construction of the South-to-North Water Transfer Project in 2002 to alleviate the uneven spatial and temporal distribution of water, and the 'sending grains from the north to the south' policy was formulated and has been carried out since 1990. So China's net virtual water trade (VWT) pattern is roughly the opposite of the distribution of its water resources.
VWT has often been recognized for its ability to improve physical and economic access to water in water-scarce regions, allowing regions to save domestic water through the import of water-intensive products (Kumar ). The term 'virtual water' (VT) was first proposed in 1994 by Allan (). Virtual water is defined as the water embodied in goods during the production process (Hoekstra ).
VWT has been much studied in recent years. Some studies have focused on the micro-level, such as the virtual water content of products or consumers; for example, Hoekstra & Chapagain () and Chapagain & Hoekstra () have researched the VW content of coffee, tea, rice and meat. Some studies have focused on the national and Recently, VW calculation methods have also attracted much attention. The methods mainly fall into two categories: one is the production tree method (Hoekstra et al. ), and the other is the input-output (IO) method. The IO model is a top-down method, which uses water use data and input-output tables to calculate the direct water coefficient and total water coefficient for each sector or region (Velázquez ). The IO model has been utilized as an effective method to assess resource flows and transfers. Some researchers, such as Wang et al. Other researchers such as Jiang et al. () and Deng et al. (), have used the multi-regional IO analysis method to analyze VWT among various regions. Their studies have verified that the IO model clearly quantifies inter-regional and inter-sectoral VW flows, and represents both direct and indirect water input during production processes. Other studies have used a computable general equilibrium (CGE) model to investigate VWT; for example, Roson & Sartori () applied a CGE to the Mediterranean region to assess to what extent VW is a viable adaptive option for water-scarce regions. Their results suggested that VW could help reduce the impacts of water scarcity, but only marginally. Berrittella et al.
() also used a CGE model to research the role of VW and water scarcity in the context of international trade. They found that VWT could help balance the supply and demand of water resources in regions suffering from water scarcity. The CGE model provides a systemic and disaggregated data representation of multi-regional and national economies. They fully account for circular income flows, inter-sectoral and market linkages. The CGE model has been used as a means to identify the hidden water consumers along the whole supply chain.
Despite the vast body of literature on VWT, however, (1) VWT research have focused more on agriculture, and the VW of industry and services has hardly been studied; (2) structural changes in the economy will lead VWT changes over time. Most work thus far has focused on assessing VWT over a specific year. The effects of longterm changes in VWT have barely been studied; (3) some studies have used the IO model to compute VWT, but the IO model is a static model that considers a specific year.
To the best of our knowledge, no one has studied VW flows using a dynamic (over time) multi-regional CGE model. This study endeavors to bridge these gaps by using a dynamic CGE model, the SinoTERM model (the enormous regional model), which includes 137 sectors in 31 regions on the Chinese mainland, to analyze VWT and its effect on the Huaihe River basin. The concept of VW was incorporated into the CGE model by adding the VW module to the CGE model. The CGE model helps realize a dynamic simulation through annual capital accumulation and regional labor market dynamic adjustment. The baseline scenario simulation is based on 12 years of data (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020). This paper discusses the trade situation and VW features of the Huaihe River basin from 2009-2020. It sets two policy scenarios (the policy simulation result analysis is from 2019 to 2020), to analyze the economic, water consumption and VWT impact under different policies.

MODELING FRAMEWORK
A dynamically multi-regional CGE model of the Chinese economy (also called SinoTERM) was developed by the The features of the SinoTERM model include full inputoutput databases, inter-regional trade matrices and behavioral equations for each province. The SinoTERM database contains input-output data for 137 sectors in 31 regions of China's mainland. It includes detailed interregional trade, consumption, import, and export data. It includes detailed information under 'margins', covering rail, road, water and air freight, plus pipelines, warehousing, trade (retail plus wholesale) and insurance. The database includes detailed tax matrices, allowing for the possibility (either in the original data or as part of a simulation) that tax rates on a particular good might vary both between regions and users. The equations in the model are linearized for simplicity, which, combined with accuracy via multi-step solution methods, ensures model efficiency through the use of GEMPACK software. This model provides a disaggregated representation of national, regional and multiregional economies, which fully accounts for circular income flows, inter-sectoral and market linkages. The corresponding resource and environmental impacts resulting from consumption activities in one region can be traced to specific production sectors in other regions through the inter-regional supply chain (Robson et al. ).

Dynamic mechanisms
Dynamic mechanisms have been added to SinoTERM (Wittwer & Horridge ). The main enhancement of dynamics was to link capital and investment over time.
(1) Physical capital accumulation Capital in industry j accumulates according to: K j (t) is the quantity of capital available for use in industry j during year t; I j (t) is the quantity of new capital created for industry j during year t; D j is the rate of depreciation, treated as a known parameter.
The mechanism for determining j's investment can be represented by: where E t denotes an expectation held in year t; ROR j (t) is the rate of return on industry j's investment undertake in year t; Q j (t þ 1) is the rental on j's capital in year t þ 1; r is the rate of interest; C j (t) is the cost of an extra unit of capital installed for industry j in year t; and f jt is a non-decreasing function.
Equation ( (2) Regional labor market dynamics are also introduced into the SinoTERM model. This allows partial adjustment in regional labor markets to occur over time in response to an economic shock. The regional labor market adjustment mechanism in SinoTERM, in level form, is given by: If the policy shock weakens the labor market in region r at period t relative to forecast, real wages W r t in policy scenario will fall relative to forecast Wf r t . In addition, there will be an initial enlarged gap between the labor market demand EMP r t and supply LS r t relative to forecast levels EMPf r t and LSf r t . In successive years, the gap between demand and supply will gradually return to forecast levels through a further decline in real wages. The speed of labor market adjustment is governed by a positive parameter α.
With the dynamic approach, the long-term effect of VWT is represented year-by-year in a simulation.

Virtual water module
The VW module assigned is the water used by final consumption rather than intermediate consumption, which allows the researcher to assess both direct and indirect water requirements of the final consumption pattern of a nation, region, lifestyle group, or household.
The SinoTERM model includes R provinces and N economic sectors, so the total output (X r j ) of sector j in province r can be expressed as intermediate input Equation (5) be rewritten as: where I is the identity matrix and has the same order as the matrix A r ij . The vector of direct water consumption is extended as: where W r j is the direct water consumption vector of region r.
The water consumption coefficients (water supplied by region r to regions s to generate one monetary unit of final demand) can thus be derived by: where D rs j denotes the total water supplied by region r to other regions to generate one monetary unit of final demand in sector j of region s.
The virtual water from region s to region r can be calculated as: where VW rs ij is the virtual water volume from i sector in region r to j sector in region s and Y rs j the output of j sector in region s.
Thus, the total virtual water inflow and outflow of region r can be calculated as: where VW r in is the total virtual water inflow of region r, and VW r out is the total virtual water outflow of region r. The SinoTERM model realizes dynamic simulation by capital dynamic accumulation and regional labor market

SCENARIO DEVELOPMENT
A baseline was developed first. It is a business-as-usual scenario without any policies. Then a policy simulation was conducted, and a shock with the VWT policy was added.
The effects of VWT are measured by deviations of variables in the policy simulation from their baseline levels.

Research area
The Huaihe River basin is located in the eastern part of section.

Regions and sectors
The database of the SinoTERM model consists of value matrices of commodities, sectors, sources and regions. In order to focus on the VW issue and speed up solution times, we aggregated the sectors and regions to manageable dimensions while retaining detail in the sectors. We also aggregated the SinoTERM database of 137 sectors to 18 sectors. The 18 sectors are listed in Table S1 (Supplementary Information). The Huaihe River basin was divided into four regions: Henan, Anhui, Jiangsu, Shandong.

Macro economy
For the baseline scenario, macroeconomic data was divided into two periods, the first ( The predictions for economic growth are displayed in Table 1, and the agricultural added value and percentages are displayed in Table 2.

Inter-provincial trade flows
Inter-provincial trade data are obtained from the Inter-    Shandong has an obvious advantage in the petroleum and chemistry sectors, but for agriculture and some service sectors it relies on other regions. Table 3 shows the interprovince trade value flow. The total trade value of the Huaihe River basin was 8,886 × 10 9 yuan.

Water data
Water data are collected from various sources. The water use by sector is extracted from the water resource bulletin of Huaihe. Only blue water (i.e., fresh surface and groundwater withdrawals) is included in these statistics. Green water (i.e., precipitation stored in the soil) is not included in this study. indirectly. The VW content of agriculture is the highest, followed by food and electricity. The total water use coefficient of the same sector between provinces is different.
The flow of VW in 2020 is shown in Table 4. The Huaihe River basin will transfer 26,653 × 10 6 m 3 VW to other regions outside of the basin.

Other parameters
The labor demand elasticity was 0.243, according to the Chi-

Policy scenario
The net outflow VW region is mainly concentrated in Anhui, which is a water shortage region. In order to alleviate the water stress in the upper reaches of the Huaihe River basin, the outflow of water-intensive commodities should be reduced. Two policy scenarios were set: in S1 Anhui's agricultural output is reduced by 5%; in S2 Anhui's agricultural outflow demand is reduced by 5%.

Macroeconomic impact
The macroeconomic impact under the two policy scenarios is shown in Figure 3. In scenario S1, Anhui's agricultural output is reduced by 5%, so its GDP will decrease by 0.218%. GDP is defined as a function of underlying capital, labor and technology (generally unchanged), so this decrease reflects the combined effects of capital use will be 0.037% higher than the baseline scenario in the Huaihe River basin, which reflects the combined effects of capital use (0.403%) and labor adjustments (À0.396%).
Agriculture is a labor-intensive sector, employment will be negatively affected in and the whole Huaihe River basin. A decrease in the amount of agricultural commodities will cause an increase in price. The consumer price index (CPI) will be 0.481% higher in 2020 than the baseline. As a result, household real consumption will be 0.174% lower and investment will be 0.296% higher than the baseline in 2020. The price of Anhui agricultural commodities rises relative to outside; the outflow of Anhui decreases by 0.41%, and the inflows increase. Changes in the inflow and outflow of the Huaihe River basin are respectively 0.023% and À0.119%.  In scenario S2, Anhui's agricultural outflow demand reduces by 5%, Anhui's GDP is negatively affected, falling by À0.248 in 2020. The decrease of outflow mainly comes from Anhui's agriculture, which is a labor-intensive sector, and employment will be negatively affected (À0.342%) in Anhui.
In S2, more agricultural production is available in Anhui markets, the price of agricultural produce decreases, and CPI shows a downward trend. This effect can affect other regions through inter-regional trade links, Jiangsu and Shandong need to inflow agricultural commodities from outside.
As labor and capital move between regions, employment and capital will increase in Henan, Jiangsu and Shandong, leading to an increase in GDP in these three provinces.
Employment and capital in Henan will increase 0.094% and 0.013%, lead to a 0.018% increase in GDP. For the Huaihe River basin, the positive influence of the Henan, Jiangsu and Shandong cannot offset the negative influence of Anhui, and the GDP of the Huaihe River basin shows a downward trend. But the impact range of GDP is smaller than that of Anhui.

Water consumption
In S1, the water consumption of Anhui decreases by 2.32% in 2020 (see Table 5), driving agricultural water consumption in the Huaihe River basin to drop by 0.65%. The agricultural output increases in Henan, Jiangsu and Shandong have a large demand for water, so water consumption is 0.08%, 0.10% and 0.69%, respectively, in 2020. Water consumption in most other industries shows a downward trend. Food processing and manufacturing is a downstream industry of agriculture, which results in a decrease in production due to the lack of agricultural products in local markets in Anhui.
In S2, agricultural water consumption reduces by 1.56%, resulting in agricultural water consumption reducing by 0.3% in the Huaihe River basin. Agriculture being the main water consumer, reductions in agricultural water consumption will relieve the water stress of other sectors, and the water consumption of other sectors will increase.
Water consumption in food processing and manufacturing will increase by 1.96% due to its upstream industry output increasing. The agricultural output of Jiangsu and Shan will increase, so the water consumption of agriculture increase.

Virtual water flow
In S1, the VW of Anhui decreases by 322 × 10 6 m 3 in 2020.
The VW outflow to outside the basin decreases by 178 × 10 6 m 3 (see Table 6). Because of Anhui decreasing outflows of agricultural commodities to Henan, Jiangsu and Shandong provinces, these three provinces need to inflow agricultural commodities from elsewhere. The VWT from Henan to Jiangsu and Shandong decreases by 25 × 10 6 m 3 and 10 × 10 6 m 3 , respectively, and the net VW outflow of Henan decreases by 23 × 10 6 m 3 . The net VW outflow of Jiangsu and Shandong increases by 78 × 10 6 m 3 and 89 × 10 6 m 3 , respectively, in the Huaihe River basin. But VW In S2, net VW outflows decrease by 89 × 10 6 m 3 in the basin, and the VW outflows to other regions outside of the basin decrease by 116 × 10 6 m 3 (see Table 7). Jiangsu and Shandong increase the inflows of agricultural commodities from Henan, and Anhui decreases agricultural commodities inflows from Henan. The VW transfer from Henan to Jiangsu and Shandong increases by 24 × 10 6 m 3 and 15 × 10 6 m 3 , respectively. The VW transfer from Henan to Anhui decreases by 7 × 10 6 m 3 . The net VW outflow of Henan decreases by 80 × 10 6 m 3 . The net  VW outflow of Jiangsu and Shandong decreases by 43 × 10 6 m 3 and 2 × 10 6 m 3 . The total VW outflow of the Huaihe River basin decreases by 241 × 10 6 m 3 .

DISCUSSION Water management and policy considerations
Unlike previous studies, this study uses a dynamic multiregion CGE model to provide an explanation of the macroeconomic, water consumption and VWT impacts on different regions and sectors. Two policy scenarios were set up, and S1 (Anhui's agricultural output reduced by 5%) can effectively reduce VW outflow and save more water resources. S1 causes an increase of 0.112% in GDP in the Huaihe River basin in 2020, but GDP in S2 decreases by 0.013%. However, employment declines in S1 and remains unchanged in S2.
This finding calls for attention to the comprehensive impact (such as GDP, employment, consumption, food security) of inter-regional VW policies.
There is an interesting phenomenon: commodity trade and VWT flow from underdeveloped areas (Henan and Anhui) to relatively developed areas (Jiangsu and Shandong) and flow from areas with severe water scarcity to relatively water-rich areas. This phenomenon is a deviation from the concept of water-related sustainable development. The main causes are that many factors (such as coal and land) determine the distribution of economic production and productivity. The upstream areas of the Huaihe River basin have a large coal base and a major arable area, and these resources provide comparative advantages. In addition, the upstream areas occupy an absolute advantage in other mining, food, papermaking, building materials and metallurgy industry (these are water-intensive and pollution-intensive industries). Water trade is not only used as a water management strategy, but also uses a land or resource management strategy. VWT improves global landuse efficiency (Seekell ). This finding should call attention to the impact of inter-regional VW transfer on local water stress and ecological environment.

Impact of the South-North Water Transfer Project on VWT
The South-North Water Transfer Project (SNWTP) represents the largest-scale attempt to solve the uneven spatial-temporal distribution of water resources and relieve the North China Plain's water problems (Barnett et al. ). By March 20, 2019, the middle line of SNWTP had transferred a total of 20.5 × 10 9 m 3 water. Henan province received the largest amount of water: 6.942 × 10 9 m 3 . Annual water transfer is increasing year by year. SNWTP alleviated the water shortage in Henan, but total VW outflows will reach 4.3 × 10 9 m 3 in 2020, which will further exacerbate water stress in Henan.
The east line of SNWTP had continuously transferred a total of 3 × 10 9 m 3 water to Shandong, but total virtual water outflows will reach 1.8 × 10 9 m 3 in 2020, which will further exacerbate water stress in Shandong. SNWDP has a limited impact on alleviating water shortage, but it may exacerbate the VW outflow in water-scarce regions.

Limitations
There are several limitations that require future research.
Firstly, only blue water was taken into consideration, and Huaihe River basin À46 89 À32 À21 0 À241 À241 Note: 10 6 m 3 cumulative deviations from the baseline scenario. Source: simulation results.
green and grey water (or water pollution) has been ignored.
Secondly, 18 sectors were considered, but agriculture was not further split into crops, livestock and other agricultural products, forestry, and fishing, which could be corrected in the future by analyzing different agricultural sub-sectors.
Thirdly, surface water and groundwater were not distinguished. It is important that the sectoral dependence on surface water and groundwater are distinguished in the future.

CONCLUSIONS
This study has progressed the following aspects. Firstly, the study is a meaningful attempt to analyze the role of VWT by using a dynamic multi-regional CGE. 18 sectors and four regions (Henan, Anhui, Jiangsu and Shandong) in the Huaihe River basin are included in the model. This allows the reader an in-depth appreciation of the impact of different VWT on the sectors and regions. The CGE is based on 12 years of data (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020), which means long-term changes in spatial patterns of production, consumption, trade and VW flows can be studied. Secondly, the VW study is not only focused on agriculture, but also includes industry and service sectors. Thirdly, two hypothetical scenarios were set to analyze the economic, water consumption and VWT impact.
Our results show that the Huaihe River basin is a net VW outflow region, with a transfer of 26.653 × 10 9 m 3 VW in 2020; agriculture VWT accounts for 63% of the total VWT and therefore often decides the direction of VWT among regions. The industry sector is an important part of VWT, and affects the scale of VWT among the regions.
The service sector's water use coefficient is small and does not have a great impact on VWT patterns. Jiangsu is the largest VW outflow province (main outflow is to outside), followed by Anhui. The VW outflows of these two provinces respectively are 12.072 × 10 9 m 3 and 8.372 × 10 9 m 3 in 2020.
Anhui is a region with severe water stress, but is a net VW outflow province to Henan, Jiangsu and Shandong and outside of the Huaihe River basin. Henan and Anhui are net virtual water outflow areas, while Jiangsu and Shandong are net virtual water inflow areas in the Huaihe River basin. This reality contradicts the concept of sustainable water development. S1 (Anhui's agricultural output reduced by 5%) can effectively reduce VW outflow and save more water resources. S1 also causes an increase of 0.112% in the GDP of the Huaihe River basin in 2020, but S2 causes GDP to decrease by 0.013%. Employment declines in S1 but is unchanged in S2.