Agricultural water use efficiency and spatial spillover effect considering undesired output in China

Water is the most critical input for agricultural production and also an important input for food security. Around the globe, almost 70% of the freshwater is used for agricultural production. Moreover, it will require a 50% increase in agricultural production to feed 9 billion people by 2025, and that 50% increase in agricultural production will require a 15% increase in water withdrawal (Khokhar, 2017). Agricultural production is composed of three elements: labor, land, and water resources. Sustainable agriculture is responsible for ecological changes and must be economically viable. Rapid population growth, as a major underlying force for environmental degradation, is a potential threat to the sustainable use of water resources and other natural resources (Maja & Ayano, 2021). At the same time, water scarcity has challenged the sustainability of agriculture, especially in arid and semi-arid regions, due to climatic effects. (Masoumeh & Ezatollah, 2011). Many scholars think about how to alleviate water shortage from the perspective of water evaporation (Valipour, 2017; El Shinawi et al., 2022) or water pollution management (Halsema & Linden, 2012; Abd-Elaty et al., 2022). So growing water scarcity and increasing demands for agricultural products have created a debate on improving agricultural water use efficiency and productivity.

Water resources are an important input for social and economic development. Being a largely agricultural country, China consumes a large amount of water resources in agriculture. In this modern era of science and technology, innovation in agricultural technology is continuously increasing and is being applied in the agricultural production process. Efficient agricultural machinery, pesticides, and fertilizers have improved the output of agricultural products and economic output as well. The utilization efficiency of new technology affects the improvement of the agricultural economy and social development. However, as one of the sources of nonpoint source pollution, agricultural activities have a negative effect on agricultural water quality (Badrzadeh et al., 2022), and due to the insufficient management efficiency of water resources and the increase of undesired output, the situation of agricultural water utilization is becoming increasingly severe. The overall plan for ecological civilization requires that regional development should not affect water resources. Therefore, it is imperative to ensure the sustainable and efficient utilization of agricultural water.

Aijun & Xianming (2010) and Guiliang et al. (2020) used the three-stage data envelope method (DEA) model and the Malmquist index method for the evaluation of China's water resources’ utilization efficiency. Shixiang et al., (2008) found the impact of the operation of the water rights trading market on the water resources’ utilization efficiency. Rilong & Mengyuan (2021) applied the meta frontier data envelopment analysis to measure the agricultural water efficiency based on the sample of 30 provinces in China from 2000 to 2017. Yaqing & Shipeng (2015) used the ratio analysis method to discuss the water resources’ use efficiency in China. Xiaobo & Boqiang (2017) revealed that improvement in technical efficiency in the agricultural sector has a significant impact on water demand and the spread of water-saving irrigation technologies (Caizhi et al., 2014). Based on the interprovincial water footprint and gray water footprint used for the interprovincial water resources’ utilization efficiency that is based on expected output and undesired output, Yu et al. (2007a, 2007b) and Xuhui et al. (2018) found through empirical research that expanding the proportion of food crops, increasing the rural labor force, and increasing the income of rural residents can effectively restrain the increase in agricultural water consumption. There is a positive correlation between the irrigated area of arable land and agricultural water consumption. Caizhi et al. (2017) investigated the influence of factors such as irrigation water-saving technology, engineering water-saving technology, water rights system, water price system, and water resources management system effect on the utilization efficiency of agricultural water resources. The results show that ecological benefits have become important factors affecting the level of regional water use efficiency. In terms of spatial research, Guoji et al. (2020) used DEA to measure China's water resources’ utilization efficiency by sector and stage, and also used Moran's index and impulse response function to analyze the spatial interaction. Feng et al. (2018) used the global Moran's I for spatial spillover effect of industrial water resources’ utilization efficiency tested with the Durbin model. Caizhi et al. (2014) and Gang et al. (2018) used exploratory spatial data analysis based on the perspective of water footprint to conduct water resource utilization efficiency analysis in the Yangtze River Economic Belt.

From the above review, the current academic research on agricultural water use efficiency and its spatial spillover effect mainly have the following shortcomings: the research on water use efficiency mainly focuses on the overall use of water resources and industrial water use. There are few studies related to agricultural water use. In previous studies, most scholars ignored the impact of undesired output on water use efficiency, although a few scholars selected undesired output indicators but could not reveal the impact of agricultural water resources on use efficiency. The actual loss is caused by the resources and environment, and the calculation results are not representative, which is not conducive to the accuracy of the evaluation of water resources’ utilization efficiency. The existing literature used the traditional DEA model to measure water resources efficiency without considering the ‘slack variable’ in it, and the output changes caused by the changes in technical efficiency have not been taken into account, making the calculation results of efficiency values unrepresentative.

The objective of this study based on the above-mentioned reality is to achieve the sustainable development goals in agriculture and improve agricultural water efficiency. In this paper, first, we will incorporate the undesired output into the efficiency evaluation index system and use the super efficiency of the SBM model to evaluate agricultural water efficiency from 1998 to 2018 in China; this paper also discusses the status quo of China's agricultural water use efficiency, by using Moran's I on the spatial correlation analysis, and illustrates the spatial agglomeration situation between the provinces. Then, the spatial Durbin model was used to analyze the spatial spillover effect of agricultural water use efficiency, and the factors affecting agricultural water use efficiency were discussed from the aspects of economy, society, and environment. Compared with previous studies, the possible marginal contributions of this paper are as follows:

(1) Taking agricultural water use as the research objective, measuring the water resources’ use efficiency has enriched the research into agricultural water. (2) Incorporating carbon emissions as an undesired output into the calculation system of agricultural water resources’ utilization efficiency, and to more truly and effectively understand the agricultural water resources’ utilization efficiency of various provinces. (3) In the measurement of agricultural water use efficiency, a Super-SBM model of undesired output is constructed. This model considers the non-radial and non-angle problems, and can effectively solve the slack variable caused by too many input variables. In addition, it can effectively solve the efficiency sorting problem when the decision unit efficiency value is greater than 1. (4) To seek the factors that affect China's agricultural water use efficiency and its interprovincial spatial spillover effects, this paper constructs a spatial econometric model and discusses agricultural fixed asset investment, labor transfer, fiscal expenditure on agriculture, forestry and water affairs, rural residents’ education and the effect of age, disposable income of rural residents, and crop sown area on agricultural water use efficiency.

DEA is an efficiency analysis method proposed by Charnes and Cooper in 1978 (Charnes et al., 1978). This method is different from the previous research methods. It can measure the relative efficiency of the research object through input-oriented and output-oriented research directions. It does not need to preset the specific function form between input and output, and also avoids parameters and weights. The subjective judgment of aspects is widely used in the evaluation of multiple input and output decision-making units. The traditional DEA model belongs to the radial and angular measurement methods, which can only be used for the research into expected output indicators in the social economy that cannot measure the data with the unexpected output indicators, and then the measurement efficiency of the evaluated object may be affected. Tone proposed the SBM model (Tone, 2001) based on traditional DEA. The measurement method of this model is non-radial and non-angular, which effectively solves the above problems. This paper focuses on the impact of carbon emissions, an undesired output, on agricultural water use, so an undesired output SBM model from the perspective of CRS (fixed expected output) has been adopted.

In the above formula, p is the value of China's agricultural water use efficiency, are the slack of input elements, expected output, and undesired output, respectively, are the vectors of input elements, expected output, and undesired output, respectively, and is the weight.

In the above formula, is the geographic distance between regions, represents the elements in the regional GDP matrix.

Y_it is the agricultural water use efficiency of the i region in year t; λ and ρ represent spatial regression coefficients; is the spatial lag coefficient of the independent variable; is the spatial weight matrix; is the parameter to be estimated; are items. When ⁠, then the model is the SLM, when then the model is the SEM, when ⁠, then the model is the spatial Doberman model (SDM); which model should be selected for research needs to be verified by testing. In this study, the LR test and the Wald test were selected to determine which model to choose, and the Hausman test results were used to determine whether to choose a random effect or a fixed effect.

The samples studied 30 provinces in Mainland China in this research. Hong Kong, Macao, Taiwan, and Tibet regions were not included in this empirical study due to the extremely special agricultural production conditions. The data collection period of time has been 1998–2018. Based on the principles of the feasibility of index quantification and the availability of data, all kinds of index data involving variables are from the ‘China Statistical Yearbook’, ‘China Rural Statistical Yearbook,’ and the provincial statistical yearbooks, ‘The First National Pollution Source Census’. Bulletin and other officially released statistical data, some missing data have been filled by linear interpolation, and 21 years of panel data of 30 provinces have been obtained after systematic sorting with the help of Excel tools. The data collected variables are shown in Table 1.

According to the KLEM model proposed by Jorgenson, the use of water resources in the process of economic growth can be divided into four types of inputs: energy, labor, capital, and intermediate inputs (Jun et al., 2004). The three basic indicators of labor input and capital investment have been used as input indicators. Agricultural water resources produce necessities for human survival and promote economic growth in the process of agricultural production, and at the same time, they also harm the ecological environment. Therefore, this paper involves output indicators. Considering two aspects of expected output and undesired output, we choose to take the agricultural economic output as the expected output indicator and agricultural carbon emission as the undesired output indicator.

Among these, the water resources input adopts the agricultural effective irrigation area as the secondary indicator, the labor input selects the primary industry employees as the secondary indicator, and the capital investment adopts the agricultural fixed investment as the secondary indicator. The carbon emission index refers to the ‘China Emissions Trading Network’. The carbon emission (E) of agriculture mainly includes the carbon emission of fertilizer production (E_f), the carbon emission of machinery use (E_m), and the carbon emission of irrigation (E_i), where E_t = E_f + E_m + E_i; E_f = G_f × A. Among them, G_f is the amount of fertilizer applied, A is the coefficient A = 857.54 kgC/t; E_m = (A_m × B) + (W_m × C), where A_m is the planting area of crops, W_m is the total power of agricultural machinery, and B and C are the conversion coefficients, B = 16.47 kgC/hm², C = 0.18 kgC/Kw; E_i = A_i × D, A_i is irrigation area, D is conversion coefficient, D = 266.48 kgC/hm². Index selection descriptive statistics are presented in Table 2.

The Super-SBM model of undesired output, based on the input-oriented perspective, Maxdea ultra 7.12, has been used to calculate the agricultural water resource utilization efficiency of 30 provinces in China from 1998 to 2018. To analyze its evolution process from a more comprehensive and detailed time perspective, the whole research period has been divided into two time periods: 1998–2008 and 2008–2018. Table 3 lists the efficiency values of 1998, 2008, and 2018, and measures the scale efficiency changes of 30 provinces in the two decades. It is divided into three different space types: efficiency increase area, efficiency decrease area, and efficiency stabilization area.

Table 3 shows that the utilization efficiency of agricultural water resources in China remains in a low state. In the three listed years, the average efficiency values remain at 0.496, 0.572, and 0.657, respectively. Although the values are low, they are in a state of slow rise. It can be seen that this is mainly caused by low technical efficiency. In Figure 1, we can see that interprovincial agricultural water resources’ utilization efficiency has many differences. Beijing, Hainan, Chongqing, Shandong, Zhejiang, Shanghai, and other provinces have relatively high utilization efficiency of agricultural water resources. It can be seen that these provinces belong to the forefront of China's economic development, have a good economic foundation and solid technical support in agricultural infrastructure, and some provinces are located in coastal areas, indicating the water resources endowment and economic development; all these situations have an important impact on the efficiency of agricultural water use. Some provinces with low efficiency, especially those in the lower ranks from 1998 to 2018, need more policy reforms and agricultural subsidies for resource configuration and facility construction.

Table 4 lists three different spatial evolution trends. Among them, Beijing, Tianjin, Shanghai, and Hainan have become areas for increasing agricultural water use efficiency due to their sound agricultural development foundation and abundant resource stock. Qinghai and Ningxia also have good policy protection and are in a state of efficiency improvement. However, Guangxi is in the efficiency reduction zone due to its fragile ecological environment and backward technological and economic levels. Although Guizhou, Yunnan, Gansu, Xinjiang, and other provinces are in the efficiency stability zone, it can be seen that the efficiency values in the two time periods have been extremely low, which is also due to their relatively backward economic level. At the same time, it is worth noting the agricultural industries such as Hebei, Jiangsu, Shandong, Anhui, Hubei, Hunan, Liaoning, and Heilongjiang. The efficiency of the main production areas is also in the efficiency reduction area.

The level of agricultural water resources’ utilization efficiency not only depends on the province's measured own water resources’ utilization and agricultural economic development, but is also affected by the other regions and provinces, especially neighboring regions and provinces with convenient communication. The previous article has studied the changing trend of China's agricultural water use efficiency in the time dimension. On this basis, this paper will use the spatial Durbin model to continue to analyze its spillover effect in the spatial perspective.

Before carrying out the regression analysis of the SDM, the spatial autocorrelation test has been carried out on the utilization efficiency of agricultural water resources in China through the global Moran's I and the local Moran scatter plot. The results are shown in Table 5. It is shown in Table 5 that the estimation results under the three matrices are the same. Except for 2016, 2017, and 2018, the Moran values are positive numbers, and most of the Moran's P values are less than 0.1, indicating that China's Agricultural water use efficiency showed a significant correlation at the 10% level, and some of them have less than 0.05 and passed the significance test at the 5% level, indicating that the spatial correlation of agricultural water use efficiency from a spatial perspective is considered very necessary.

To observe the changing trend of the spatial correlation of China's agricultural water use efficiency in the time dimension more comprehensively and intuitively, the spatial local Moran test was carried out on the agricultural water use efficiency in China in 1998 and 2018. The economic geographic distance matrix is more significant; therefore, the economic geographic distance matrix is used for testing.

Table 6 shows the specific distribution of Moran's I in each province at two time points. In the spatial local autocorrelation test, the first quadrant is ‘High-high agglomeration type’ (HH), which means that its agricultural water resources’ use efficiency value is high, and the adjacent agricultural water resources’ use efficiency value is also high; the second quadrant is ‘low-high agglomeration type’ (LH), indicating that its agricultural water use efficiency value is low but surrounded by provinces with high agricultural water resources’ use efficiency value; the third quadrant is ‘low-low agglomeration type’ (LL), indicating that its agricultural water resources’ use efficiency value is low and surrounded by the surrounding areas with the low value of agricultural water use efficiency; the fourth quadrant is ‘high-low agglomeration type’ (HL), which means that the value of its agricultural water use efficiency is high but the surrounding agricultural water use efficiency value is lower. Surrounded by regions, among the provinces in the four quadrants, the first and third quadrants indicate that there is a positive correlation between the observation object and its related regions, whereas the regions located in the second and fourth quadrants have a negative correlation with their related provinces. Table 6 indicates the local spatial autocorrelation test of China's agricultural water resources’ use efficiency; it can be found that during the period time from 1998 to 2018, it remained in the first place. Provinces in the first and third quadrants decreased, and moved to the second and fourth quadrants, indicating that the positive correlations of adjacent provinces tended to decrease while the negative correlations tended to increase.

According to the theoretical basis of the previous model selection, the LR test and Wald test have been carried out on the utilization efficiency of agricultural water resources in China. Since the test results rejected both Hypothesis 1 and Hypothesis 2, the SDM has been selected for spatial effect analysis. On this basis, according to the results of the Hausman test, the χ² value is 24.41, and the Prob > χ² value is 0.0010. Therefore, the fixed effect is selected to estimate the SDM. In order to ensure the accuracy of the estimation results, this paper reports the estimation results under the geographic distance matrix and the economic geographic matrix at the same time, as shown in Table 7.

Table 7 shows the estimation results of the geographic distance matrix and the economic geographic distance matrix under the fixed effect of the SDM of agricultural water resource utilization efficiency in China. The spatial ρ values were −0.6294 and −0.0819 under the two-way fixed effects, respectively. Except for the spatial fixed effect under the geographic distance matrix, the spatial ρ value is significant at the 5% test level and is negative, indicating that China's agricultural water resources’ utilization efficiency on the interprovincial level shows a negative spillover effect. The improvement of agricultural water use efficiency in neighboring provinces has not been conducive to the improvement of agricultural water use efficiency in this province.

In addition, it can be seen from Table 7 that in the fixed-effect regression of the Durbin model under the two distance matrices, the LogLik values of the time fixed effects are 34.073 and 21.101, respectively, and the LogLik values of the spatial fixed effect are 340.601 and 343.247, respectively, and the LogLik values of the two-way fixed effect are, respectively, 34.073 and 21.101, and 363.023 and 364.302, and the two-way fixed effect LogLik value is the largest; on this basis, in the geographic distance matrix, the AIC and BIC values under the time fixed effect are −40.145 and 22.095, respectively, and the AIC and BIC values under the space fixed effect are −653.203 and −590.962, the AIC and BIC values under the two-way fixed effect are −698.046 and −635.806, respectively; in the economic geographic distance matrix, the AIC and BIC values under the time fixed effect are −14.201 and 48.039, and −658.494 and −596.254, respectively, and the AIC and BIC values under the two-way fixed effects are −700.603 and −638.363, respectively; whether it is a geographic distance matrix or an economic geographic distance matrix, the two information criteria of the two-way fixed effect, AIC and BIC, have the smallest. Therefore, the two-way fixed effect in the Durbin model is the best choice. To further analyze the spatial spillover effect of China's agricultural water use efficiency, the direct, indirect, and total effects of the explanatory variables on the explained variables are estimated. The direct effect represents the spatial spillover effect of the explanatory variable in the province on the explained variable in the province, the indirect effect represents the spatial spillover effect of the explanatory variable in the neighboring province on the explained variable in the province, and the total effect is the sum of the two. The specific results under the geographic distance matrix and the economic geographic distance matrix are shown in Table 8.

Table 8 shows the direct effect, indirect effect, and total effect results under the two distance matrices. The estimated results under the two matrices are the same. Among the direct effects, fixed investment in agriculture, fiscal expenditure in agriculture, forestry and water affairs, years of education of rural residents, and crop planting area are positively correlated with agricultural water use efficiency. Among them, the coefficient estimation of agricultural fixed investment is 0.0577, and the coefficient estimation of rural residents’ years of education is 0.4629 (0.4778), both of which are significant at the 1% test level. When agricultural fixed investment and rural residents’ years of education increase, agricultural water use efficiency will increase accordingly, within the range of cognitive knowledge. On the contrary, the estimated value of the labor transfer coefficient is −0.0925 (−0.0891), and the estimated value of the disposable income coefficient of rural residents is −0.0739 (−0.0745), which is negatively correlated with agricultural water use efficiency in theory. Among them, the labor transfer index is significant at the 1% test level.

Concerning the indirect effect, it can be seen that the estimated values of agricultural fixed investment, labor transfer, and fiscal expenditure on agriculture, forestry, and water affairs are −0.0198 (−0.0508), −0.0570 (−0.0663) and −0.0495 (−0.1119), respectively, and the coefficient sign is negative. It means that the surrounding provinces’ agricultural fixed investment, labor transfer, and financial expenditure in agriculture, forestry, and water affairs increase, and reduce the efficiency of agricultural water use in this province. In Table 8, the estimated coefficient of rural residents’ years of education is 0.5440 (0.8851), and the estimated coefficient of rural residents’ disposable income is 0.1827 (0.1969). The estimated coefficient of crop sown area was 0.6637 (0.6053). With the increase of years of education, the increase of disposable income of rural residents, and the increase of crop sown area in neighboring provinces, the utilization efficiency of agricultural water resources in this province was improved. The radar map of agriculture water use efficiency can be seen in Figure 2.

The study measures the utilization efficiency of agricultural water resources in China based on the perspective of unexpected output. The results show that the utilization efficiency of agricultural water resources in China always remained at a low level but in a state of slow rise. As is indicated from measured data results, it is mainly caused by the low technical efficiency which has been mainly affected by the input factor and management system, and the low technical efficiency of agricultural water resources’ utilization in China is mainly caused by the loss of water input factors. The comparative benefits have been relatively low, especially in the planting industry. Therefore, farmers, especially those in water-scarce areas, have invested water resources in aquaculture and other industries with higher returns. On the other hand, the construction of irrigation and water conservancy facilities is the key to the effective use of agricultural water resources. China's water conservancy infrastructure construction is weak, especially for irrigated agriculture in the central and western regions. It is affected by the climate, terrain, and water scarcity; irrigation technology has not been mature; and agricultural water use efficiency has a low trend. In addition, Shicheng (2011) believes that there are still some problems in China's water conservancy facilities, such as inadequate market-oriented reform, unreasonable proportion in the structure of the water conservancy facilities system, and lagging reform of management system and backward irrigation technology.

In terms of method selection, the traditional DEA model does not consider the problem of slack variables and cannot measure the existence of undesired output, and the measurement of efficiency will be biased. Therefore, this study selects the SBM model that can measure the undesired output. Just using the ordinary SBM model, there will be cases where the efficiency is all 1 and there is no obvious fluctuation. In this study, the Super-SBM model was finally selected to measure the agricultural water use efficiency, but this is not perfect. This study assumes that all inputs have equal weights. Subsequent research needs to weigh the subjective and objective weights scientifically and seek greater breakthroughs in the model.

From the perspective of the measurement index system, in the past, in the measurement of agricultural water use efficiency, the focus was on the expected agricultural output, and the undesired output was ignored. For example, Guoji et al. (2020) did not consider undesired output to measure agricultural water use efficiency and concluded that the efficiency value was high, the efficiency value was higher than 1 in most provinces, and the conclusions were significantly different. In consideration of ecology and low carbon, the results obtained by incorporating carbon emissions into the output perspective to measure agricultural water use efficiency are more accurate. The differences in research results are particularly prominent in Beijing, Shanghai, and other provinces. These provinces have more developed agricultural technologies and more significant carbon emissions. If the undesired output of carbon emissions is not considered, the efficiency obtained is higher than the actual level. Therefore, this paper incorporates carbon emissions as an expected output in the analysis, which will make the results more reasonable.

In terms of spatial correlation, the results show that the positive correlation among the Chinese provinces shows a weaker trend, while the negative correlation shows a strengthening trend. This shows that there are few exchanges between provinces in China in terms of agricultural water use efficiency. The reason may be that the unique characteristics of water resources are easy to flow and difficult to transport which makes cross-regional exchanges more complicated. For example, the South-to-North Water Diversion Project is to improve economic development and the optimal allocation of water resources in China. The improvement of water use efficiency has made outstanding contributions. But its implementation, from approval to trial operation, took more than 10 years, and the time cost is quite large. In addition, in the South-to-North Water Diversion Project, the implementation of water resources has a great impact on the ecological environment of the Yangtze River Basin. This also reflects to a certain extent that China's agricultural water resources’ utilization efficiency has difficulties with spatial communication.

When exploring the factors affecting agricultural water use efficiency, the Tobit model was used for correlation analysis in the past, but this method has a strong dependence on the distribution of data, and the results are not robust enough; it can only analyze the influencing relationship of the province and cannot explore the relationship between provinces. The spatial correlation of the obtained results cannot be compared horizontally. For example, in the analysis of factors affecting China's agricultural total factor water use efficiency, Tong Jinping et al. used the Tobit model to analyze the influencing factors of 30 provinces, which lacked spatial connectivity. This study adopts the spatial spillover effect analysis method. On the one hand, it can explore how the economic, social, and environmental perspectives have an impact on agricultural water use efficiency and then propose ways to improve efficiency; on the other hand, it can discover the mutual influence mechanism of neighboring provinces through indirect effects, strengthen cooperation in surrounding areas, and ensure the comprehensiveness of research. Of course, this method still has some limitations. When selecting explanatory variables, this study comprehensively considered and selected impact indicators from multiple perspectives. However, due to the complexity of agriculture, there are many factors affecting agricultural water use efficiency, practice-based maneuverability cannot incorporate all indicators into the study, and follow-up studies can be more detailed.

The study concluded that the comprehensive technical efficiency of China's agricultural water resources’ utilization efficiency has not been high in the past 21 years, but it is in a state of slow rise, and the low-efficiency state is mainly caused by the low pure technical efficiency. China's agricultural water use efficiency is in a state of spatial agglomeration as a whole, but there is a trend of positive correlation to the negative correlation between neighboring provinces, and China's agricultural water use efficiency shows a negative spatial spillover effect in terms of spatial spillover effect. From the perspective of the direct effect of agricultural water use efficiency, agricultural fixed asset investment, agricultural, forestry, and water affairs fiscal expenditure, rural residents, the increase in the number of years of education, and the sown area of crops have a positive impact on the efficiency of agricultural water use in the province. Meanwhile, the transfer of labor and the increase in the disposable income of rural residents has reduced the efficiency of agricultural water use in the province. From the perspective of indirect effects, the years of education, disposable income of rural residents, and crop sown area of rural residents in neighboring provinces have a positive correlation with the efficiency of agricultural water use in the province, which is manifested in the spatial spillover effect. Transfer and fiscal expenditure on agriculture, forestry, and water affairs will have a negative correlation, which is manifested as a negative spatial spillover effect. Although the overall water resources’ utilization efficiency has a negative spillover effect from a spatial perspective, it can be seen that the reason for the low water resources’ utilization efficiency in China is purely technical efficiency, and the construction of agricultural infrastructure, especially water conservancy facilities, is particularly important. On the other hand, the low-income problem caused by the natural characteristics of agriculture is also the reason for the low water efficiency. This problem is still the top priority of researchers, and it is also the fundamental reason to improve the utilization efficiency of agricultural water resources in China.

According to the above conclusions, the following inspirations are obtained: (1) The improvement of agricultural water resources’ utilization efficiency mainly depends on water resources endowment and infrastructure construction. (2) Increase investment in education, strengthen the supervision system for the implementation of 9-year compulsory education, comprehensively improve the quality of agricultural producers, and on this basis, focus on the promotion and application of agricultural technology, especially irrigation technology. (3) The reform of agricultural supply side can turn agricultural development into green and healthy development, and accelerate the pace of ecological civilization construction. (4) Starting from the top-level design, interprovincial exchanges and cooperation in agricultural water resources’ utilization technologies, and rationally optimize water resources allocation. (5) Reasonably set water prices to reduce waste caused by large-scale flood irrigation and the loss of agricultural water resources.

This research work was supported by the National Social Science Foundation Project ‘Study on the Transformation of Ecological Modernization in Rocky Desert areas of Western China’ (Grant No.: 18CSH036).

We declared that we have no conflicts of interest in this work. We declare that we do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.