An integrative analytical framework and evaluation system of water environment security in the context of agricultural non-point source perspective

Agricultural non-point source pollution (ANPSP) caused a contradiction between economic growth and water environmental security protection. In order to understand the trade-off between social-economic development and water environmental security in the context of agricultural non-point source pollution, a Driving force-Agricultural non-point source pollution-Pressure-State-Response (DAPSR) model framework was proposed, and 23 indicators were selected to construct the evaluation system of water environment security in this study. And we take Ya’an City, China as case study from 2017 to 2019, the characteristics of water pollution was analyzed, and the water environment security was evaluated by method of particle swarm projection pursuit. The results show that: (1) Agricultural non-point source pollutant discharge in Ya’an generally shows a decreasing trend. (2) The agricultural non-point source pollution subsystem and the response subsystem have a great impact on water environment security. (3) According to the values of water environment security, Yucheng, Hanyuan, Tianquan and Lushan are basically safe in level III, Mingshan is unsafe in level IV, Yingjing is safe in level II, Shimian and Baoxing are safe in level II. (4) The degree of agricultural non-point sources is highly correlated with the water environment security. This study shows that the DAPSR model is feasible and practical, and can provide a scientific basis for the decision-making of regional agricultural non-point source pollution prevention and water environmental security protection.


Introduction
Water security is one of the most important hot topics facing human beings in the 21st century, and it is closely related to people's life, property and ecology (Jensen and Wu 2018, Krueger et al 2019, Khan et al 2020. Global water use has increased nearly 6-fold over the past 100 years (Wada et al 2016), and about 60% of the world's people already live in a state of water scarcity (Mekonnen and Hoekstra 2016). Almost all rivers in Africa, Asia and Latin America have worsened since the 1990s (United Nations Environment Programme UNEP 2016). Soaring water consumption and water pollution threaten the water environment security and increase the uncertainty of future sustainable development.
Water environment security (WES) is a common index reflecting the coordination of the 'social economyenvironment' system that is used to evaluate and manage water environment (La Notte et al 2012). At present, there is no widely accepted definition of water environment security. Narain (Narain et al 2013) pointed out that water environment security is closely related to the process of urbanization and the health status of residents around the city. (Zeitoun 2011Bakker 2012Zhang et al 2019) define water environment security as the sustainable access of human beings to sufficient water resources and a good water environment for socioeconomic development, which is the outcome of various factors coming from nature, society and humanities. Water environment security evaluation is very significant to the local harmonious development among environment, society and economy (Green et al 2015), and it is the basis of water environment security protection decision-making. To study various aspects of water environment security, Giné Garriga (Giné Garriga and Pérez Foguet 2010) proposed the Water Poverty Index (WPI) for water security assessment, which consists of five components: resources, approaches, utilization, capacity and environment. Ahmadi (Ahmadi et al 2012) pointed out that water environment security mainly is decided on water quality safety by studying issues of social and water quality in the comprehensive planning of land use and water resources allocation. Varis (Varis Fraboulet-Jussila 2002) established a Bayesian network model to analyze the utilization of water resources in the Senegal River Basin in Africa, and concluded that the rapid and disorderly urbanization process will further aggravate the huge conflict with the water environment. Hamilton (Hamilton et al 2006) based on hazard analysis and critical control points, proposed the use of various detection methods to optimize WES risk management with a simple mechanism. Ip (Ip et al 2009) from the perspective of public health protection and uses the grey relational method to assess water environment security. The water environment security assessment objectively and comprehensively reflects the water environment security status, which has many design factors and is a complex and macro-environmental system. Therefore, in the water environment security evaluation, the interaction of many factors will also be involved.
Appropriate indicators and reasonable system are the premises of WES comprehensive evaluation (Jiang 2015, Jensen and Wu 2018, Krueger et al 2019. Due to the complexity of comprehensive evaluation, the selection of indicators cannot be uniformly recognized. the assessment models of water environment used in recent researches include: WR-WEC (  , DPSIRM (Driving force-Pressure-State-Impact-Response-Management) conceptual model (Chai et al 2020). Jia (Jia et al 2018) constructed a comprehensive index system from the aspects of water environment and water resources bearing status, water environment vulnerability and development and utilization potential. Ren (Ren et al 2017) established a cloud-like model to simulate the evolution mechanism of complex regional water security systems. Şener (Şener et al 2017) applied the water quality index (WQI) to select 21 water quality indicators in the Aksu River and 24 sampling sites to evaluate the water quality safety of the water environment and its suitability as a drinking water source.
However, these researches lack the consideration of the important influence of agricultural non-point source pollution on water environment security. Agricultural production is part of economic development and living food sources, but it also threatens water environment security. Agricultural non-point source pollution (ANPSP) has led to environmental problems such as eutrophication of water bodies, destruction of water ecosystems, and threats to the safety of drinking water for humans and animals (Bo et al 2012;Abler 2015). According to the national pollution census in China, the total nitrogen (TN) and total phosphorus (TP) loads from agricultural non-point sources accounted respectively for 46.5% and 67.2% of the total emissions in 2020. In the European Union, approximately 50% of surface water and 25% of groundwater bodies were in less than good condition in 2016 (EEA 2020), most of which were due to ANPSP (Plunge et al 2022). Water environment degradation caused by non-point source pollution is an urgent issue since it is greatly restricting both urban and rural development sustainability (Zhang et al 2016, exploring the coordinated development path under the agricultural non-point source pollution is the key to future socio-economic development and water safety protection. Water environment security in a certain area is always limited, and excessive agricultural non-point source pollution discharge is very likely to break through the regional water environment security, and even destroy the regional ecosystem. For example, Ya'an, China, there has an important position among the vital ecological protective screens in the Yangtze River with its superior water ecological environment. (Mei Bopeng 2020, Zhou et al 2021) However, till now, there are still many water monitoring sections in Ya'an with water quality of Class IV or V (according to China's environmental quality standards for surface water GB 3838-2002, the quality of surface water is divided into 5 grades from I to V from good to bad, 'Class IV or V' means that the water sources are not good enough to be used as drinking water sources). And coincidentally, these sections are mostly concentrated in districts and counties with a large scale of planting and livestock. Therefore, it is necessary to consider agricultural non-point source pollution into the water environment security evaluation process, which helps decisionmakers explore the regional socio-economic coordinated development path. In this study, based on the above models, fully considered the connotation of water environment security, and followed the principles of scientific, systematic, comprehensive, accessible, and representative (Yu and Li 2011), a Driving force-Agricultural non-point source pollution-Pressure-State-Response (DAPSR) model was proposed, and a water environment security evaluation index system covering 5 levels of driving force, agricultural non-point source pollution, pressure, state, response and 23 indicators was constructed.
In water environment security assessment, the evaluation method is also very important for obtaining objective and reasonable results. In study of environment assessment, the methods include principal component analysis (Demšar et al 2013), analytic hierarchy process (Shabbir and Ahmad 2016), entropy method (Sun Pingjun and Zhongzhi 2010) fuzzy comprehensive evaluation (Zheng et al 2019), Delphi method (Filyushkina et al 2018), etc. Projection pursuit (PP) is a suitable non-linear high-dimensional data processing method that effectively avoids the influence of subjectivity on assessment results, is applied to environment assessment (Han and Guo 2012). And for optimizing the evaluation results, intelligent optimization algorithm greatly enriches the optimization technology and provides a feasible solution to those combinatorial optimization problems that are difficult to deal with by traditional optimization techniques. The commonly used intelligent optimization algorithms include ant colony optimization (ACO) (Wan et al 2016), genetic algorithms (GAs) (Li et al 2020), differential evolution (DE) (Guan et al 2021) and particle swarm optimization (PSO) (Nguyen et al 2019, Zhang et al 2021. Specifically, the PSO algorithms has good search performance and fast convergence speed , it has been applied to field of sustainable development policy research (Zarghami andHajykazemian 2013, Wang et al 2022).
Based on above, in order to build a model which can objectively evaluate WES under the background of ANPSP, which could facilitate regional decision-making on ANPSP controlment and water environment security protection, this paper takes the districts and counties in Ya'an, China as research case: (1) the water pollution characteristics in Ya'an were calculated and analyzed.
(2) based on the investigation and analysis, the degree of ANPSP was considered in the construction of the WES evaluation index system. (3) A DAPSR model was proposed, and the WES of each district and county in Ya'an was calculated and discussed.
2. Evaluation indicator system based on DAPSR framework 2.1. DAPSR framework Constructing an evaluation system is the key point to objectively evaluating water environment security. Considered ANPSP into construction of evaluation system, a DAPSR model was proposed in this paper. The model framework is shown in figure 1.

Evaluation indicator system
The main elements of the model framework include population and economic driving forces, water environment and water pressure, water quality and quantity state, and response measures such as sewage treatment and green construction. Based on this, and followed the basic principles of scientificity,  representativeness, and operability, 23 indicators were selected to construct the indicator system, which is shown in table 1.
(1) The Driving force(D) subsystem represents the impact of population and economic development on environmental protection. Therefore, population density and per capita GDP were selected to reflect the driving force of population and economy.
(2) According to ANPSP mainly comes from planting industry and livestock ( (3) The Pressure(P) subsystem represents the pressure level of water resource consumption and water pollution controlling. Therefore, the impact factor of domestic sewage discharge, chemical oxygen demand (COD) and ammonia nitrogen emissions of domestic sewage, industrial wastewater discharge, COD and ammonia nitrogen emissions of industry were selected to reflect the impact of wastewater discharge on water safety. Meanwhile, agricultural and industrial water consumption reflect the economic benefits of water resources, and daily domestic water consumption per capita to reflect the water consumption of residents.
(4) The State(S) subsystem represents the state of water resources and water quality condition. The indicators proportion of water quality of grade III and above and water quality compliance rate of water function zone reflects the quality of the water environment, annual total water supply reflects the capacity of regional water supply.
(5) The Response(R) subsystem represents the response of the government, enterprises or social organizations to water environmental protection, such as water environment security protection policy formulation and greening construction. Thus, the indicators per capita garden green area and greenery coverage of urban area reflect the regional ecological environment status, industrial wastewater treatment rate, industrial COD removal rate and industrial ammonia nitrogen removal rate reflect the water pollution prevention and control capacity.
The standard of each evaluation grading was determined by: (a) national standards, local standards or policy documents of relevant indicators, (b) references which have listed the relevant evaluation grading standards, (c) ranked the index data of each province in the country from 2010 to 2019, the 10th percentile and the 90th percentile has chosen as the optimal value and the worst value for classification.

Particle swarm projection pursuit model
As data of indicators have high-dimensional, nonlinear and complex characteristics, the WES comprehensive evaluation model can be constructed by projection pursuit method which can efficiently process nonlinear, non-normal and high-dimensional data (Friedman and Tukey 1974;Holland 1992). The basic idea of the projection pursuit method is to project the high-dimensional data onto the low-dimensional subspace, and reflect the comprehensive features of the original data through the scattered structure of the projected data on the low-dimensional subspace (Peng and Deng 2020). The basic idea of applying it to the modeling of comprehensive evaluation of water environment security showing as follows , Ouyang et al 2021: Set the data y ij as n samples with m indicators (i=1, 2K, n; j = 1, 2, K, m):

Normalization of indicators
Eliminate the influence of the various physical dimensions of indicators, the indicator data of needs to be dimensionless, so that the target value is between 0 and 1. Let the jth indicator of the ith sample in the sample indicator be x ij (i = 1, 2K, n; j = 1, 2, K, m). The indicator attributes are divided into positive and negative. Positive means bigger is better, while negative means smaller is better. Calculate according to the following formulas: positive indicators: where x jmax is the maximum value of the ith index, and x jmin is the minimum value of the ith index.

Structure of comprehensive eigenvalue Z i
Project the m-dimensional WES evaluation data onto one-dimensional projection value with the projection direction a j ( j=1, 2, K, m), the formula as follows: where y ij is the sample index value, a j is the projection direction parameter, a j ä [-1, 1], following the constraint condition:   a =1.

Construction of projection index function Q (a)
Optimize Z i , its scattered structure is required to have the characteristics that the local projection points are as dense as possible, and the overall projection point clusters are as dispersed as possible. Therefore, the inner-class density d(a) and the between-class distance s(a) are introduced to construct the projection objective function Q(a).
The d(a) is larger, the classification of samples is more significant; the s(a) is larger, the samples are more dispersed; the Q(a) is larger, the Z i is better represents y ij , and when the Q(a) reaches its maximum value, means that the optimal projection direction.
Inner-class density: Between-class distance: where r ik is the distance between Z i and Z k ; R is the window radius of the local density, r max + m/2 R 2m, r max is max r ik ; f (R-r ik ) is 1 when R > r ik , otherwise it is 0;Z is average value of Z i .

Optimization of projection direction a
Set the objective function as max Q(a). Constraint condition:   a =1. When the constraints are satisfied, the maximum value of Q(a) will be solved, and the optimal projection direction a will be found at the same time. There are many methods to optimize the projection direction. In this study, particle swarm optimization algorithm (PSO) was used for optimization.

Particle swarm optimization algorithm
The basic idea of particle swarm optimization algorithm is: firstly, initialize a group of random particles (random solution), and then find the optimal solution through iteration. In each iteration, the particle updates itself by tracking two extreme values (p best , g best ), where p best is the optimal solution (individual extreme value) found by the particle itself, and g best is the current optimal solution found by the entire population (global extrema) (Kennedy and Eberhart 1995).
Search the target in a D-dimensional space, which includes N particles. The ith (i=1, 2, K, N) particle is a D-dimensional vector, and the velocity and location of the ith particle are: The optimal location is: p p p , , ..., 11 best g g gD 1 2 = After p best and g best were found, the velocity and position of the ith particle were updated according to the following formulas: where w is inertia weight, usually be 0.9, 1.2, 1.5, 1.8; c 1 and c 2 are learning factors, usually be c 1 = c 2 =2 or c 1 =1.6, c 2 =1.8 or c 1 =1.6, c 2 =2; rand() ä (0,1), which is a random number. Furtherly, we perform functional tests on the PSO algorithm. PSO algorithm was used to independently optimize the test function for 100 times, the test functions are shown in formulas (14)~(18). The analysis results of the extreme values of the test functions and the optimization results listed in table 2.
Test 1: Find the maximum value of a multimodal positive function. The theoretical maximum value is: x= 0.0000, y max = 1.0000. The function curve is shown in figure 2(a), and the function expression is as follows: The function curve is shown in figure 2(b), and the function expression is as follows:

Calculation process
The values of all indicators were calculated, especially ANPSP and domestic and industrial source pollution in Ya'an was analysed first. According to formula (1)~(13) and (20), the particle swarm projection pursuit programming calculation was realized by MATLAB 7.0 software, to find the optimal projection direction, to calculate the comprehensive evaluation value of WES, and to draw the grade distribution map with ArcGIS 10.6. PSO programming iterative operation was performed by MATLAB 7.0, and the parameter settings of PSO are shown in table 3. Note: Inertia weight * , which decreases linearly with the number of iterations.

Analysis on water pollution characteristics in Ya'an 4.4.1. Agricultural non-point source pollution
Ya'an is located in the southwest of Sichuan Province, and the land used for grain crops and cash crops meets the standard farmland standards. According to the relevant calculation parameters of The National Environmental Protection Plan for Drinking Water Sources and the source intensity coefficient of farmland runoff pollutants published by the China Ecological Environment Administration, the chemical oxygen demand emission coefficient of standard farmland is 1.5 t/km 2 /year, and the ammonia nitrogen is 0.3 t/km 2 /year, total nitrogen is about 0.345 t/km 2 /year, and total phosphorus is about 0.018 t/km 2 /year. According to the National Pollutant Discharge Standard for Livestock Breeding Industry, the livestock and poultry breeding volume including chickens, cattle, sheep, rabbits and other living poultry is converted into pigs according to the conversion coefficient. And then, according to the accounting method updated by Dr Genyi Wu (Wang et al 2013) based on The Pollution Production Coefficient and Pollution Discharge Coefficient of Livestock and Poultry Breeding Industry, the pollution production coefficient is determined as follows: COD is 200 kg/piece/year, total nitrogen is 18 kg/piece/year, total phosphorus is 2.4 kg/piece/year, and ammonia nitrogen is 6.4 kg/piece/year. According to the above pollution production and discharge coefficient, the discharge of non-point source pollutants from planting industry and from livestock and poultry breeding in each district and county of Ya'an was calculated. The results are shown in table 2. In order to reflect the pressures of preventing and controlling non-point source pollution per unit of arable land and per unit of arable land area, the calculation results were converted into the non-point source pollutant discharge of planting per hectare of arable land area (unit: kg/hm 2 ) and the non-point source pollutant discharge of livestock and poultry breeding per square kilometer land area (unit: t/km 2 ). The result is shown in figure 4.    Therefore, the Yucheng government should pay more attention to the prevention and control of non-point source pollution in the planting industry, while the Baoxing government should pay more attention to the prevention and control of non-point source pollution in livestock and poultry breeding. Hanyuan and Mingshan should consider formulating policies to prevent and control planting and livestock and poultry breeding at the same time. While Hanyuan and Mingshan implementing the reduction of chemical fertilizers and pesticides, they should also implement policies for the collection, treatment and harmless utilization of livestock and poultry manure. And in Yingjing, Shimian, Tianquan and Lushan, based on effectively controlling agricultural non-point source pollution, they should start to think about how to further accelerate the construction of ecological agriculture, for establishing a zero-emission mode of agriculture non-point source pollution as soon as possible.

Domestic and industrial source pollution
In the survey statistics, the data of domestic sewage discharge and industrial wastewater discharge and their pollutant emissions came from 2018-2020 Ya'an City Statistical Yearbook, 2018-2020 Ya'an City Water Resources Bulletin, 2017-2019 Ya'an City Water Environment Quality Bulletin and the environmental statistics data from Ya'an City Statistics Bureau. In order to reflect the economic benefits behind wastewater discharge and pollutant emissions, the data results were converted into discharge per and pollutant emissions per 100,000,000 CNY. The results are shown in figure 5.
From figure 5, it can be seen that, among all administrative districts, during study period, on the whole: (1) Hanyuan has the largest domestic sewage discharge per 10,000 CNY, followed by Yucheng and Mingshan, while Shimian has the lowest.  (5) Shimian faces smaller pressure on the treatment of domestic wastewater discharge, but bigger pressure on the treatment of industrial source wastewater. (6) In Tianquan, the overall change trend of sewage discharge and pollutant emissions from domestic source are relatively stable, but that from industrial source are increasing. (7) In Lushan, the overall changing trend of sewage discharge and pollutant emissions from domestic source are downward, but that from the industrial source are increasing. (8) In Baoxing, the overall changing trend of sewage discharge and pollutant emissions from domestic source have a small fluctuation, while that from industrial source are obviously rising but overall in a relatively lower level in Ya'an City.
Therefore, when regional decisionmakers formulate water environment security protection policies: (1) Mingshan should appropriately expand and retrofit the sewage disposal plant to cope with the dual pressure of high discharge of domestic sewage and industrial wastewater, and makes efforts on wastewater discharge monitoring and management in industrial areas, encourage enterprises to adopt measures such as cleaner production and wastewater reuse for reducing the pressure of industrial pollution at the source through.
(2) In the case of industrial source pollution being effectively reduced, Yingjing should pay attention to monitoring the signs of possible rebound of pollution discharge. (3) Hanyuan should appropriately expand and retrofit the domestic sewage treatment plant to ensure that the treatment capacity can cope with a large amount of domestic sewage discharge. (4) On the basis of stable discharge of domestic and industrial pollution sources, Shimian should consider how to further reduce pollution from the aspects of sewage treatment technology upgrade. (5) Yucheng, Tianquan, Lushan and Baoxing should realize that industrial source pollution is gradually increasing, and pay attention to strengthen the investigation of industrial pollution sources then take corresponding measures to effectively control pollution.
From figure 6, it can be seen that the top five indicators have the greatest impact on the WES of Ya'an are in order: a 3 , a 8 , a 20 , a 9 , a 21 , where a 3 is pollutant emissions from planting per hectare of arable land, a 8 is COD emissions of domestic sewage per 100,000,000 CNY, a 9 is ammonia nitrogen emissions of domestic sewage per 100,000,000 CNY, a 20 is greenery coverage of urban area, a 21 is industrial wastewater treatment rate. It shows that the water environment security in Ya'an can be significantly improved by controlling the discharge of nonpoint source pollutants from the planting industry, enhancing the treatment capacity of domestic sewage pollutants, strengthening the greening construction, and improving the treatment rate of industrial wastewater.

Analysis on each subsystem
According to the optimal projection direction of each index, the projection eigenvalues of each subsystem in each district and county in Ya'an from 2017 to 2019 were calculated. Furtherly the contribution rate of each subsystem to the evaluation level of water environment security was obtained, as shown in figure 7. And the variation trends of each subsystem in different districts and counties of Ya'an are analyzed, as shown in table 5.
From figure 7 and table 5, it can be seen that, from 2017 to 2019: (1) The contribution rate of the Response subsystems in Yucheng, Yingjing, Shimian and Tianquan shows an obvious growth trend, it indicates that the governments and enterprises in these regions were strengthening the response to water environmental protection from 2017 to 2019; (2) The contribution rate of the Driving force subsystems in Mingshan and Hanyuan shows an obvious growth trend, it indicates that the water environment security of these two districts were gradually adapting to the pressure on social and economic development from 2017 to 2019; (3) The contribution rate of the ANPSP subsystem in Baoxing shows an overall obvious growth trend, it indicates that Baoxing's attention to in water environmental protection was gradually increasing from 2017 to 2019, especially in improving agricultural non-point source pollution.
On the whole, from 2017 to 2019, the important factors affecting the comprehensive evaluation results of the water environment security in Ya'an were the response subsystem and the agricultural non-point source pollution subsystem, the contribution rate of the Driving force subsystems on the water environment security shows a gradually rising trend, and that of the Pressure system shows a gradually weakening trend, it indicates that: (1) Ya'an municipal government, enterprises and social groups were increasing their investment in water pollution prevention and water environment improvement from 2017 to 2019; (2) In the context of population density reduction and GDP per capita continuous increase in each district and county of Ya'an from 2017 to 2019, the pressure caused by the population on the water environment security has decreased, and the increase in per capita GDP also contributes to the water safety, the coordination between economic gain and water environment protection was increasing from 2017 to 2019.
Through the analysis of the subsystems above, the relationship between the subsystems was furtherly explained, which helps to understand the DAPSR model framework, as shown in figure 8. From table 4 and figure 9, it can be seen that during the study period:

Analysis of the index values of the WES
(1) The overall ranking of water environment security of the districts and counties in Ya'an from high to low is: Baoxing, Shimian, Yingjing, Tianquan, Lushan, Yucheng, Hanyuan, Mingshan.
(2) The index values of water environment security of Yucheng, Hanyuan, Tianquan and Lushan are stable on the whole, they are basically safe in level III.
(3) The index values of water environment security of Mingshan is unsafe in level IV, it is because the water quality in Mingshan is class V and inferior class V from 2017 to 2019, and the level of agricultural non-point source pollution is high. However, Mingshan's WES values have a tendency to increase slowly, they benefit from not only the decrease in population density, which has eased the pressure of water environment safety on the population scale, and also the rise of the economic development level that can contribute to environmental governance and protection.
(4) The index values of water environment security of Yingjing is fluctuates greatly. It is safe in level II in 2017 and 2019, while it is basically safe in level III in 2018. The main reason is that the response(R) subsystem contributes less to WES in 2018, indicating that the effectiveness of water environmental protection measures in 2018 is relatively weaker than in 2017 and 2019.
(5) The index values of water environment security of Shimian and Baoxing are safe in level II, because the nonpoint source pollution in these two places is low, and the response(R) subsystem has a high contribution to the improvement of WES. Meanwhile, Shimian's per capita GDP ranks in the forefront of Ya'an city, Baoxing has the lowest population density in Ya'an, high economic level and low population pressure are also conducive to maintaining good water environment security. However, the WES of Baoxing in 2019 decreased significantly, which is due to under the trend that the pressure(P) subsystem contribution rate in Baoxing reduced year by year from 2017 to 2019, and the pressure(P) subsystem index in 2019 decreased significantly, social and economic development has increased the pressure on water resources consumption and wastewater pollutant discharge, and has shown the result of weakening the water environment security.

Analysis of the correlation between ANPSP and WES
Spearman correlation analysis of the ANPSP subsystem with its indexes and the comprehensive evaluation index of WES was carried out. The results are shown in figure 11, it can be seen that the correlation between each agricultural non-point source index and the ANPSP subsystem is from 0.86 to 0.96, and the correlation between the ANPSP subsystem and the WES is 0.9, indicating that the correlation between ANPSP and the WES is very high. The aforementioned analysis 4.3.1 also shows that the degree of ANPSP has a high contribution to the comprehensive evaluation results of water environment security. These conclusions show that it is particularly important to take into account the background of ANPSP in the comprehensive evaluation of WES. Therefore, in order to improve the security of water environment, all districts and counties in Ya'an City should strengthen the implementation of pesticide and fertilizer reduction actions and the special action plan for livestock and  Pressure of water resources situation and consumption subsystem (P) State of water resources and environmental situation subsystem (S) Note: Z, overall increasing trend; [, overall decreasing trend; -, overall stable trend; * , represents that the subsystem has the largest contribution rate to the comprehensive evaluation of the water environment security of the corresponding area.
poultry breeding pollution controlling, continuously improve the ANPSP control system, and create ecological agricultural products and green agriculture, promote the coordinated development of high-quality agricultural development and high-level environmental protection. Note: x 3 , pollutant emissions from planting per hectare of arable land; x 4 , pollutant emissions from livestock and poultry breeding per square kilometer of land area; x 5 , fertilizer consumption per hectare of arable land; x 6 , pesticide consumption per hectare of arable land.

Conclusion
(1) From 2017 to 2019, the discharge of agricultural non-point source pollutants in Ya'an City generally show a downward trend year by year. Hanyuan, Mingshan, Yucheng, and Baoxing are the key areas for the prevention and control of ANPSP in Ya'an City.  (2) From 2017 to 2019, Ya'an City's domestic sewage and industrial wastewater discharge per 100,000,000 CNY varies greatly from region to region. Hanyuan and Yucheng should attach more importance to the treatment of domestic sewage, Yingjing, Shimian and Lushan should strengthen the treatment of industrial wastewater, while Mingshan needs to handle the dual pressure of domestic sewage and industrial wastewater treatment.
(3) The important factors affecting the comprehensive evaluation results of the WES values of Ya'an City are Response(R) subsystem and Agricultural non-point source pollution(A) subsystem. Therefore, in order to improve the level of WES, Ya'an government has to notice the response of environment protection policies and control of ANPSP.
(4) From 2017 to 2019, the WES values of all districts and counties in Ya'an City are: Yucheng, Hanyuan, Tianquan and Lushan are basically safe in level III, Mingshan is unsafe in level IV, Yingjing is usually safe in level II, Shimian and Baoxing are always safe in level II.
(5) ANPSP is a key factor affecting the comprehensive evaluation results of WES value, and it is highly correlated with WES. The DAPSR model is feasible and practical, it can be used for providing a scientific basis for regional management of the water environment and control of ANPSP.