Selected Environmental Assessment Model and Spatial Analysis Method to Explain Correlations in Environmental and Socio-Economic Data with Possible Application for Explaining the State of the Ecosystem

Abstract

Regional ecosystem health is the basis for regular regional exploration, ecological protection, and sustainable development. This study explored ecosystem health at the southern end of the Hu Line (Sichuan and Yunnan provinces) using the pressure–state–response model and examined the spatial evolution of ecosystem health. The proportion of unhealthy and morbid cities decreased from 45.9% in 2000 to 35.1% in 2016. The imbalance of ecosystem health among cities has gradually increased since 2006, but more high-quality cities have emerged (Z of Moran’s Index < 1.96, p > 0.05). Overall, the regional ecosystem on the southeast side of the Hu Line was healthier than that on the northwest side. Differences in ecosystem health on both sides of the Hu Line showed decreasing trends over time except for the pressure score. The spatial pattern of ecosystem health moved along the Hu Line because the pressure and state scores of ecosystems were mainly determined by the natural environmental conditions. Based on the county-level assessment, the grade of imbalance within cities was divided, and those that were lagging were identified. To correct regional imbalances, a comprehensive and proactive policy framework for a smart development model was put forward in Sichuan and Yunnan.

1. Introduction

The dramatic increase in the range and intensity of human activity has rapidly changed the global ecosystem, and poses a severe threat to the survival and development of human society [1,2]. The contemporary world is increasingly industrializing and urbanizing, especially in developing countries [3]. China, which is the largest developing country in the world, has shown remarkable progress in its social economy. However, ecosystems and social development have a limited capacity to manage environmental pressure, and when such pressure exceeds threshold levels there may be adverse effects on local ecosystems. In such cases, regional climate, hydrological conditions, and biological diversity will undergo appreciable changes, profoundly affecting regional ecological process [4,5,6]. The improvement and maintenance of healthy ecosystems is essential to achieving sustainable socioeconomic development, as natural ecosystems provide necessary materials and ecological services for survival and development of human beings [7]. Ecosystem health assessment is identified as the foundation of environmental management, and reasonable and effective evaluation of ecosystem health can identify environmental or economic crises to avoid serious deterioration of ecosystem health because of rapid socioeconomic development [7,8,9].

The concept of ecosystem health, which originated in the 1980s, has grown rapidly in recent years, and it is now a common perspective applied to ecosystem assessment [10,11]. Many assessment frameworks have been developed to examine ecosystem health on single-ecosystem scales, such as farmland [12], forests [13], grasslands [14], wetlands [8], marine systems [15], cities [16], and mines [17]. The exploration of better means for quantifying the interaction between human activities and ecosystems and recovery mechanisms such ecological services has been a long-term objective to enable the existence and sustainable development of human beings [18]. Comprehensive investigation of the combination of micro-level and macro-level is an ideal method for evaluating ecosystem health. Indicator taxa [19,20,21] and indices [8,22] are commonly used to conduct such investigations. Many ecosystem health assessment endeavors have employed the vigor-organization-resilience (VOR) model and its derivatives, which are often associated with benign operation of any life system at any scale, and can be used to construct systems for assessment of urban ecosystem health from different aspects [23,24]. The vigor is measured in terms of activity, metabolism or productivity; organization can be assessed as the diversity and amount of interactions between system components; resilience is measured in terms of a system’s capacity to maintain structural stability in the presence of stress [25]. Unlike the VOR model, which has the inherent confines due to lack of socio-economic and human health factors, the pressure–state–response (PSR) model emphasizes that human beings are a part of the ecosystem and play a pivotal role in determining the environmental state. Here, pressure indicators describe the pressures on ecosystem health exerted by human activities, including resource pressures and social pressures. State indicators reflect the status quo of ecosystem health, such as the vigor, organization, and resilience of an ecosystem. Response indicators show the response degree to the changes of ecosystem health conditions, including changes from humans and the ecosystem itself [26]. The comprehensive and dynamic features of this model makes it more informatory [27].

Western China is largely lagging behind in their own development, and most of the economic activity in this region is concentrated on the southeastern side of the Hu Line [28], which is a classic theories of Chinese geography in which there is a hypothetical line dividing eastern and western China into roughly two parts [29,30]. To the southeast side of the line, 36% of the territory is inhabited by 96% of China’s population [31]. Correcting this regional imbalance has become one of the primary tasks of the Chinese government, which launched the Development of the West Region program in 1999 and the National New Urbanization Plan in 2014 [28,32]. Yunnan and Sichuan, located at the southern end of the Hu Line, has many complex characteristics including their landforms, human activity, climate change, and ecosystems. Simultaneously, they are in an important ecologically functional zone in China, with species diversity that has been recognized by the World Wildlife Fund and the International Union for Conservation of Nature [33,34]. Ecosystem health is not only the basis for ecological protection and sustainable development, but also a key point that must be addressed when breaking through the Hu Line during regional economic development. Hence, it is necessary to systematically explore the ecosystem health and its spatial evolution and regional differences in the western region, especially in the transitional zone traversed by the Hu Line.

Therefore, this study was conducted to (1) evaluate the ecosystem health from 2000 to 2016 at the prefecture level using the PSR model; (2) measure the spatial temporal variation of ecosystem health using primacy ratio, coefficients of variation, spatial autocorrelation analysis, spatial gravity center model and standard deviational ellipse; (3) diagnose the ecosystem health at county dimension based on the statistical downscaling model; (4) conduct a difference analysis of regional ecosystem health to determine the strengths or weaknesses of typical counties in social development. The findings from this study can provide references for policy-making regarding the regional economy and environmental protection.

2. Materials and Methods

2.1. Study Area

The study area is located at the southwestern end of Hu Line, which was further narrowed to Yunnan and Sichuan province (21°08’–34°19´ N and 97°21´–108°33´ E), which are the first and second terrain transition zones of China (Figure 1). Our study area included 37 prefecture-level cities and 312 counties (Appendix A and Appendix B). The southern end of the Hu Line is an ecologically fragile area that is also a hot spot for economic development and population growth. Located in Sichuan and Yunnan provinces, the southeastern part of the Tibetan Plateau is fragile and vulnerable to climate change [35]. Sichuan Basin is in the northeastern part of the study area, where the terrain is relatively flat and the soil is fertile. The Yunnan–Guizhou Plateau is located in the southern part of the study area, which belongs to a subtropical humid zone subjected to significant climate differences [36]. Sichuan and Yunnan Provinces have the most abundant biodiversity in China, and Hengduan Mountain in the west is even a global biodiversity hotspot [32]. The study area, in which the highest altitude is over 7000 m, has environmental and socioeconomic differences among the plain, mountain, and plateau areas. When compared with eastern China, Sichuan and Yunnan have not undergone the rapid urbanization, and its land use change is slow but marked.

2.2. Ecosystem Health Assessment Framework

A framework for information collection that embodies socio-economic, ecological and resource aspects and considers the interactions among these components is needed [37]. This framework must ascertain both human activities that exert pressure on the environment and the impacts of environmental change on human well-being [8,38]. Moreover, health is not the opposite of disability, while ecosystem health is an embodiment of ecological carrying capacity [39]. The deficiency of ecosystem services and management would lead to the decline of ecological carrying capacity, thus reducing the level of ecosystem health [1]. In this study, we set a desired indicator system for regional ecosystem health based on the PSR framework [8,27,40,41]. We then introduced indicators of human attributes into the assessment framework together with indicators of natural qualities. The developed indicator system consists of 17 indicators that reflect the pressure, state, and response in the assessment region to produce the ecosystem health scores. More details are shown in

Table 1. Indicator system used to evaluate ecosystem health at different scales.

1st-Level Indicator	2nd-Level Indicator	3rd-Level Indicator	Weight (Prefecture Level)	Weight (County Level)
Pressure	Agriculture	Planting area of crops (I1)	0.316	0.382
		Fertilizer application amount (I2)	0.126	0.170
	Population	Population density (I3)	0.205	0.071
		Natural population growth rate (I4)	0.181	0.062
	Nature	Percentage of temperature anomaly (I5)	0.055	0.064
		Percentage of precipitation anomaly (I6)	0.116	0.251
State	Agriculture	Total output of agriculture, forestry, animal husbandry and fishery (I7)	0.196	0.207
		Grain yield per unit of cultivated land (I8)	0.226	0.289
	Economy	Per capita GDP (I9)	0.198	0.353
		Rural per capita net income (I10)	0.228	-
	Nature	Normalized difference vegetation index (NDVI) (I11)	0.152	0.151
Response	Agriculture	Irrigation area (I12)	0.107	0.178
		Total power of agricultural machinery (I13)	0.245	0.220
	Society	Per capita investment in fixed assets of the whole society (I14)	0.193	0.144
		Per capita local government budget expenditures (I15)	0.174	0.168
		Total mileage of highway (I16)	0.191	0.154
		Tertiary industry proportion (I17)	0.089	0.135

.

Human activities, unsustainable resource consumption, or unreasonable economic structures exert stress on the natural environment, changing the quality and quantity of natural resources. But these changes would lead to responses in human organized behavior that adopt rational economic policies to restore or ameliorate the health of the environment and to mitigate or prevent ecosystem degradation.

2.3. Data Acquisition and Processing

To quantify ecosystem health changes, we collected many relevant data at the city and county level from 2000 to 2016 from Sichuan and Yunnan statistical yearbooks and regional economy statistical yearbooks. These societal, economic, agricultural, and ecological data were used to derive or calculate ecosystem health indexes for ecosystem health assessment. Chinese statistical yearbooks at the county-level were used to supplement some missed data not available in the former two types of yearbooks, while the rest of the anomalies or missed data (0.93%) could be revised and calculated by regression equations. Temperature and precipitation data were provided by the China Meteorological Administration [35]. The meteorological data at a 5 km × 5 km resolution were obtained by using the kriging method to interpolate data from 109 meteorological stations around and within the Sichuan and Yunnan province. The moderate resolution imaging spectroradiometer (MODIS) data was downloaded from the NASA website [32]. The monthly NDVI dataset was compiled by the maximum value composite method to minimize the impacts of atmosphere scan angle, solar zenith angle, and cloud contamination.

2.3.1. Climate Change Index

Climate change is associated with species abundances and distributions, as well as one species-level extinction [42]. Natural disasters such as floods and droughts pose serious threats to the environment, social development, and human life [43,44]. Sichuan and Yunnan are often the sites of natural disasters and have been specified as a global biodiversity hotspot [32]. The percentage of precipitation anomalies can intuitively reflect anomalies of precipitation and are commonly used to evaluate drought events [45]. Therefore, using the percentage of temperature anomaly and precipitation anomaly, we evaluated the natural pressure for the study area caused by climate change as follows:

$$P_a=\frac{P-\overline{P}}{\overline{P}}\times 100\%$$

(1)

$$T_a=\frac{T-\overline{T}}{\overline{T}}\times 100\%$$

(2)

where P_a and T_a are the percentage of temperature anomaly and precipitation anomaly, respectively; P and T are the total precipitation and mean temperature of a year, respectively; and P and T are the long-term mean annual precipitation and temperature, respectively. The term was from 2000 to 2016.

2.3.2. Data Acquisition at the County Level

Ecosystem assessments require some socio-economic data of spatial and temporal resolutions, which are currently not available from remote sensing images. Because yearbook data are missing or given in low resolution, it is important to downscale them to the required spatial resolution in practical case studies. The indicators used in the assessment process at the county level were consistent with those used at the city level. The data acquisition methods and approaches mentioned above were used to collect most of the data, but some indicators lack data of some or a large number of counties (

Table 2. Ratio of missing indicator data.

Indicator	Missing Ratio
I2	58.7%
I7	41.3%
I12	41.3%
I13	16.7%

).

There is usually a demand to translate information from a large spatial scale to finer geographic scales while keeping consistency with the raw dataset [46]. This process is called spatial downscaling, and it is a common method to use existing data with correlation to estimate unknown data [47,48]. Based on the correlation analysis of the indicator data with land cover data and socio-economic data, a multiple linear regression equation was established to calculate the indicator data that were missing. Moreover, the indicator of planting area of crops was replaced by the indicator of sown area of grain crops. Rural per capita net income data were not available directly or indirectly; therefore, this indicator was excluded. The missing data at the county level were calculated by the formula:

$$Y=a_1{A}_1+a_2{A}_2+\dots +a_m{A}_m$$

(3)

$$Y_{county}=Y\cdot \frac{Y_{city}}{{\displaystyle \sum _{i=1}^{n}Y}}$$

(4)

where Y is the coefficient of a county-level indicator, A is the socio-economic indicator or area of land cover type of the county, and a is the coefficient of the corresponding indicator. Additionally, Y_county is the indicator data needed to acquire the county; Y_city is the total value of the city in which the county is located, and n is the number of counties in the city. Regression analysis was calculated using SPSS 16.0 and the results are shown in

Table 3. Relationships between county-level indicators and other data.

Indicator	Regression Model	R2 (Regression Equation)	R2 (Fitted Equation)
I2	Y = 49010.2A1 − 37950.7A2 + 19857.5 A3 − 292.3	0.72	0.87
I7	Y = 795200A4 + 0.067	0.97	0.98
I12	Y = 17288.7A3 + 39586.6A4 + 9099.1A5 + 1255.7	0.86	0.92
I13	Y = 33.538A3 + 23.88A4 + 25.428A5 − 17.429A6 + 7.149	0.60	0.82

. A₁ is the population of the county; A₂ is the GDP of the county; A₃ is crop area of the county; A₄ is the gross value of primary industry of the county; A₅ is the impervious area of the county; and A₆ is the area of Bareland. To verify the accuracy of the data, scatter plot analysis was performed with the calculated data and the actual data. Although the correlation coefficient of the total power of agricultural machinery was lowest, its missing ratio was also the lowest (Table 2 and Table 3).

2.3.3. Comprehensive Assessment

Positive indicators denote that the ecosystem health score is declining when indicator values decrease; negative ones indicate that the ecosystem health score is improving when indicator values decrease. In this study, planting areas of crops and all state and response indicators were positive. The extremum difference method was used to normalize each index [49]. Entropy is an objective method of measuring the information uncertainty by probability theory that implements quantitative analysis for indicators [50]. In this study, we used the entropy weight method to express the decision information of each indicator and assign the weight for indicators. The pressure, state, and response scores were found by weighted overlaying corresponding index layers, and the composite ecosystem health assessment result was obtained by overlaying these three criterion. The formula is as follows:

$$S_i=\left|\sqrt{{\displaystyle \sum _{j=1}^{k_i}{\left(W_{ij}{R}_{ij}\right)}^2}}\right|$$

(5)

$$H=\left|\sqrt{{\displaystyle \sum _i^3{S}_i{}^2}}\right|$$

(6)

where S_i is the assessment score of the i items (pressure, state, and response); k_i is the number of assessment indicators in the items i; W_ij is weight of indicator j in i items; R_ij is the normalized value of each index; and factor H is the ecosystem health scores.

2.4. Analysis of Overall Evolution Characteristics

Based on assessment of the ecosystem health in the study area at the city level from 2000 to 2016, its overall evolution characteristics were analyzed. The primacy ratio, coefficients of variation, spatial autocorrelation analysis, and Herfindahl coefficient methods [51,52,53] were introduced to explore the distribution, equilibrium degree, spatial autocorrelation trends, and agglomeration degree of ecosystem health in the entire region over the study period. Because there are many references [51,52,53] describing how to use these methods, only the spatial autocorrelation analysis is covered here. Global Moran’s Index, which is a commonly used method of spatial autocorrelation analysis, was used in this study to examine the spatial relationship with ecosystem health in each region. Global Moran’s Index was greater than 0, indicating that the ecosystem health of each region had a positive spatial autocorrelation. A smaller index indicated a stronger spatial dispersion of the assessment results. The formula was as follows:

$$I=\frac{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{W}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}}}{S^2{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{W}_{ij}}}}$$

(7)

$$S=\frac{1}{n}{\displaystyle \sum _{i=1}^n{\left(x_i-\overline{x}\right)}^2}$$

(8)

where n is the number of spatial units (prefecture-level cities); x_i and x_j are the health scores for units i and j, respectively; and W is the spatial weight matrix. If units i and j are adjacent, then W = 1, otherwise, W = 0. The global Moran’s Index was tested for significance using the following formula:

$$Z\left(I\right)=\frac{I-E\left(I\right)}{\sqrt{Var\left(I\right)}}$$

(9)

2.5. Analysis of Local Evolution Characteristics

We obtained the annual ecosystem health distributions over the last 20 years, after which the spatial change rate of ecosystem health was obtained by using the least squares method:

$$slope=\frac{n{\displaystyle \sum xy-{\displaystyle \sum x{\displaystyle \sum y}}}}{n{\displaystyle \sum x^2}-{(x)}^2}$$

(10)

where n is the number of years, and x and y are the year and ecosystem health scores (or pressure, state, and response scores), respectively. The concept of a gravity center is derived from physics, and the gravity model has been widely utilized in the fields of economic geography, land use science, urban planning, and ecosystem services value, etc. [54]. The variation track of the gravity center of assessment result value can well reflect the regional difference of the results in changes [55]. The migration process of ecosystem health status in space was revealed using the gravity model and the gravity center coordinate of ecosystem health is given as follows:

$$\overline{X}=\frac{{\displaystyle \sum _{i=1}^n{H}_i\times X_i}}{{\displaystyle \sum _{i=1}^n{H}_i}},\overline{Y}=\frac{{\displaystyle \sum _{i=1}^n{H}_i\times Y_i}}{{\displaystyle \sum _{i=1}^n{H}_i}}$$

(11)

where n is the number of small areas (spatial units); X_i and Y_i are the center coordinate of area i; and H_i is an attribute value (health, pressure, states or response scores) of area i. A standard deviation ellipse was used to reflect the spatial distribution of ecosystem health and identify changes in average position and the moving direction [36]. The standard deviation ellipse consists of three elements: average location, orientation, and dispersion (or concentration). The long axis of the ellipse delineates the direction and trend of spatial distribution of ecosystem health. In ArcGIS10.4, three standard deviations can be used to describe the standard deviation ellipse and contain about 68%, 95% or 99% centroids of all input features, respectively. We chose the first one to explore the evolution characteristics of ecosystem health at the prefecture level.

3. Results

3.1. Global Features of Ecosystem Health Condition at the City Level in the Study Area

The overall characteristics of ecosystem health at the prefecture level of the study area from 2000 to 2016 are shown in Figure 2 and Figure 3. In 2000, the average ecosystem health score was 0.60, which increased by 15% to 0.69 by 2016 (Y = 0.0058X–11.01, R² = 0.95, p < 0.0001) (Figure 2). Specifically, the average pressure scores of the ecosystem showed a fluctuating trend, and they generally slowly increased (Y= 0.0008X–0.92, R² = 0.70, p < 0.0001). The highest pressure period in Sichuan and Yunnan was in 2000, when the score was 0.720, while the lowest was in 2014, when there was a score of 0.735. Both the state score (Y = 0.0089X–0.17.34, R² = 0.92, p < 0.0001) and response score (Y = 0.0089X–17.34, R² = 0.96, p < 0.0001) of the study area increased rapidly with a slope of 0.0089 per year. These findings indicate that three aspects of the ecosystem were improving, as was the overall ecosystem health score.

Analysis of the primacy ratio, variation coefficient, Herfindahl coefficient, and Moran‘s Index indicated that there was a tipping point in the evolution of ecosystem health at the year 2005 or 2007, since which all values except Moran’s Index have shown increasing trends (Figure 3). A Moran’s Index value greater than 0.15 appeared before 2005, reflecting a clustering distribution trend in ecosystem health, and the p value and Z value of Moran’s index were less than 0.05 and larger than 1.96, respectively (Figure 3d). Moran’s Index showed a reverse pattern (Z > 1.96, p < 0.05) after 2005, with a decrease turning to a random distribution.

3.2. Spatio-Temporal Pattern Evolution of Ecosystem Health Condition at the City Level

3.2.1. Spatio-Temporal Pattern Evolution

The pressure, state, and response scores as well as the spatial change rate (slope) of each region were calculated. The results were divided into four categories using the natural breaks method in ArcGIS 10.4. The natural breaks method is designed to determine the natural clustering of attribute values through seeking to minimize average deviation within the class while maximizing average deviation among the classes, and the method has good adaptability and high precision in dividing geographical environment units [56,57]. This division of relative results is still valid for long time series, because all units would not have essential changes or leaps at the same time during the study period. The spatial distribution of pressure and state scores across the study area showed lower scores on the northwest side of the Hu Line than on the southeast side (Figure 4Aa,Ab). The highest pressure scores (to withstand minimum pressure) were distributed in Nanchong, Mianyang, Dazhou, and Liangshan in Sichuan, and Qujing and Wenshan in Yunnan. The spatial distribution of the response scores was not obvious, and scores of the provincial capital (Chengdu and Kunming) were slightly higher than those of other areas (Figure 4Ac). Analysis of ecosystem health changes over the past 20 years revealed that the negative slope of pressure scores was spread over the Sichuan Basin, while the positive slope was spread over Yunnan, with values ranging from −0.0036 yr⁻¹ to 0.0062 yr⁻¹. (Figure 4Ba). Higher slopes of state scores were found in the provincial capital and surrounding areas, all on the southeastern side of the Hu Line (Figure 4Bb). The lower slope of response scores mimicked the distribution of lower response scores and the slope of response scores in Sichuan was superior to that in Yunnan in general (Figure 4Bc).

In 2000, nearly 50% of the prefecture-level cities in the study area were below the unhealthy level (Figure 5a). The ecosystem health of four cities were at a morbid level, almost all of which were located on the northwest side of the Hu Line. The areas of health level were Chengdu, Mianyang, Dazhou, and Nanchong in Sichuan Basin. In 2016, the number of areas below the unhealthy level decreased by four and the number of morbid cities decreased by 50% (Figure 5b). Although the health of cities on the northwest side of the Hu Line have improved, the region is still unhealthy, especially Lijiang and Nujiang.

There were differences in ecosystem health on both sides of the Hu Line, and the ecosystem health on the southeast side of the Hu Line was better than that on the northwest side (Figure 6). However, the scores for different 1st-level indicators showed different trends over time. From 2000 to 2016, the state scores gap (Y = 0.001X–1.99, R² = 0.53) among regions on the two sides increased continuously and the difference in the pressure scores (Y = −0.0008X + 1.67, R² = 0.44) and health scores (Y = −0.00038X + 0.63, R² = 0.38) among the two sides declined. There was a tipping point in the year 2007, with the differences in response scores increasing before then, and rapidly decreasing after. The support of the area on the northwest side of the Hu Line was apparently stepped up.

3.2.2. Results of the Spatial Gravity Center Model and Standard Deviation Ellipse

The gravity center of ecosystem health for the entire study area was located to the east of Liangshan, and moved 6.64 km to the southwest from 2000 to 2016 (Figure 7). The gravity center of the ecosystem pressure (pressure scores) moved 11.14 km to the northeast (southwest) along the Hu Line. From 2005 to 2007 it moved 5.72 km, which was much farther than in other years. Similarly, the gravity center of the ecosystem state scores moved 11.43 km along the direction of Hu Line. The gravity center of the response scores moved north 5.21 km (3.77 km along the Hu Line and 3.65 km perpendicular to Hu Line). In general, the gravity center of the ecosystem health, including pressure scores and state scores, moved almost exclusively along the Hu Line.

The regions with healthier ecosystems were mainly distributed in the Sichuan Basin and eastern Yunnan. Furthermore, the direction and trend of standard deviation ellipses were dominated by these regions. The center of standard deviation ellipses was at a central location between the two provincial capitals (Chengdu and Kunming).

3.3. Ecosystem Health at the County Level in Sichuan and Yunnan

Because of the large number of counties in the study area (number = 312), the ecosystem health evaluation results of each county were divided into six categories using the natural breaks method in ArcGIS 10.4 (Figure 8). The distribution of ecosystem health of the study area as of 2016 is shown in Figure 8. There was a significant difference between counties on the southeast and northwest sides of Hu Line with regards to the intensity of ecosystem pressure, ecosystem state, and ecosystem health. On the southeastern side of the Hu Line, the pressure, state, and health performance was significantly higher than that on the northwestern side, while areas with high response scores were scattered throughout the study area. The overall health based on the county-level evaluation is consistent with that of the city-level evaluation. In 2016, the county with the highest ecosystem health index was Xuanwei (in the southeastern Hu Line area) of Qujing, while Dege of Ganzi (in the northwestern Hu Line area) had the lowest score. Moreover, 16.2% of the prefecture-level cities were unbalanced in their development, and the ecosystem health condition levels of each county in these cities differed greatly (

Table 4. Equilibrium degree of internal development within prefecture-level cities. Typical prefecture-level cities have counties leading in ecosystem health or lagging in ecosystem health.

Balance Degree	Proportion	Overall Health	Typical Cities	Lagging Counties	Leading Counties
Balanced (<3) *	13.50%	Healthy	Ziyang
		Relative healthy	Bazhong, Suining
		Unhealthy	Nujiang, Ganzi
Relatively balanced (>2, <5)	70.2%	Healthy	Meishan	Danling
			Chengdu	Pidu
		Relative healthy	Mianyang		Santai
			Chuxiong		Chuxiong
		Unhealthy	Diqing	Weixi	Shangri-La
			Aba		Lixian
Unbalanced (>4)	16.2%	Healthy	Qujing	Malong
		Relative healthy	Liangshan		Xichang
			Dali	Jianchuan

* Numbers indicate how many types of ecosystem health levels there are in the region. There were no extremes in our study area. The healthiest ecosystem and the unhealthiest ecosystem will not be located in a region at the same time if the region contains only two levels of ecosystem health. Therefore, we considered regions that contained only one or two ecosystem health levels to be balanced.

). Overall, 13.5% of the prefecture-level cities had a balanced development, and the ecosystem health levels of the counties were basically the same. The ecosystems of some cities were healthy as a whole, but with that of one county lagging behind. For example, the ecosystem of Pidu District was the unhealthiest county in Chengdu, indicating that greater attention needs to be given to this county.

4. Discussion

4.1. Assessment Methodology

Human society is bound to ecosystems, and the health of ecosystems underpins human sustainability [58]. When human activities exert pressure on ecosystems, the state of the ecosystem does not change immediately. Their interactions, which combines the complex cumulative effects of temporal and spatial patterns, are highly dynamic [37]. Owing to the complexity of ecosystems, it is not easy to develop an integrated framework and model that equally blends both natural and human factors to assess ecosystem health. Accordingly, a common and unifying approach to the process is critical [59,60]. The PSR model emphasizes that human beings are a part of the ecosystem and that human activities play a pivotal role in determining the environmental state. These comprehensive and dynamic features makes it a more informatory model [27], as indicated by the results of studies in China and other countries [61,62].

Determining the indicator and its weight are two important steps in the evaluation process [63]. In actual application, it is impossible to incorporate all factors affecting regional ecosystem health in an assessment model. In this study, the indicator system for regional ecosystem health was built based on the PSR framework. Human impacts and natural factors were introduced as indicators into the assessment framework. Overall, this framework consisted of 17 indicators reflecting the pressure, state, and response of each region to produce ecosystem health scores: (1) Our study area has limited land resources with over 90% mountain cover [32]. Farmland has been shrinking for various reasons since 1957 [64]. In 2013, the per capita cultivated land area of 12 cities, such as Chengdu, Panzhihua, and Luzhou, was lower than the critical level issued by the United Nations (per capita cultivated land should be no less than 0.80 mu). Food supplies require continuous soil fertilization; however, the application of chemical fertilizers can cause environmental pollution and soil nutrient imbalance, such as an increase in heavy metals and toxic elements, decrease in soil microbial activity, and soil acidification. The northwest part of the study area is part of the Tibetan Plateau, which has a fragile environment that is sensitive to climate change [35]. Continuous drought also occurs in many places in Yunnan Province. Precipitation anomalies commonly cause drought events [45]. After 1949, the population of the study area increased substantially, with the permanent population exceeding 130 million in 2016. These factors of the study area are represented by the pressure indicator of the planting area of crops, fertilizer application amount, percentage of temperature anomaly, percentage of precipitation anomaly, population density, and natural population growth rate. (2) Agriculture production, economic situation, and ecological condition were selected as components of the state indicator to reflect ecosystem function and environmental status. Agriculture is a mainstay for human survival and development and, globally, the advances in agriculture is seen as an important means of economic prosperity and human well-being. Land resource vitality serve as a valuable indicator for measuring primary productivity and ecosystem activity [65]. Therefore, agricultural state, such as total output values for agriculture, forestry, animal husbandry, and fisheries and grain yield per unit of cultivated land, were selected as components of the state assessment indicators. Economic state was reflected in the use of rural per capita net income and per capita GDP as state indicators. To a certain extent, social and economic activities can improve people’s quality of life, enhance environmental protection awareness, and promote the coordinated development of humans and land. Healthy ecosystems are more resilient to adverse effects, such as disturbance from excessive human activities or natural disasters. For instance, a good forest ecosystem can regulate climate and preserve water and soil, and NDVI has been successfully used to reflect ecosystem vitality and monitor habitat degradation [66]. The NDVI is also commonly used as an assessment indicator of ecosystem state [8,67,68,69]. Generally, the higher the vegetation coverage (NDVI), the better the quality of the regional environment and the ability of the ecosystem to self-regulate. (3) Ecosystem responses to pressures can adjust a system’s state and agricultural and social countermeasures were used to indicate this response that was transformed into specific outcomes acting on the social and natural environment. For example, local government expenditure is an important measure for governments to coordinate regional economic development. In this study, total power of agricultural machinery, irrigation area, per capita local government budget expenditures, per capita investment in fixed assets of the whole society, tertiary industry proportion, and total mileage of highway were selected from social and agricultural responses to assess the responsiveness of the ecosystem.

By combining a range of driving forces of ecosystem changes with the health status and responsiveness of the ecosystem, the PSR model was able to reflect the nature of ecosystem evolution more comprehensively. In addition to the traditional evaluation framework reflecting ecosystem quality, such as pressure, state, and response, health assessment is also related to biophysical processes and human ecological services. The responsibility for each indicator of ecosystem pressure, state, and response should be highlighted. The ownership of each index is non-definitiveness and subjective decision, whereas related areas are prone to confusion, especially ecosystem state and ecosystem response. As such, it is essential to monitor the state and sources of risk as well as to enforce timely regulation of probable sources of stress.

4.2. Dynamics of Ecosystem Health in Sichuan and Yunnan

Evidence shows that many human-dominated ecosystems have become highly intense [70]. Regional ecosystems have been under such tremendous stress that it was difficult to obtain higher pressure scores. The average pressure scores of ecosystems in Sichuan and Yunnan have slowly increased at a decreasing rate since 2000 (Figure 2). Development of the west region promoted the economic development of the western provinces through a large amount of state capital investment, which was a great strategic idea that began in 2000. Because macro-development was adjusted and controlled by the public policies [71], both the average state and response scores of the ecosystem have improved significantly, which were sensitive to administrative polices. As a result, even under low pressure scores, the average health scores of the ecosystem in the study area increased gradually. Increases in the primacy ratio and coefficient of variation indicated that the regional development imbalance has been further aggravated and the areas with higher health scores showed an expansion of health advantages after 2006 (Figure 3a,b). As shown in Figure 3d, there has been a more random spatial distribution of healthy ecosystems across the study area since then. The Chengdu Economic Circle, with its prosperous regional economy, took the lead in realizing rapid economic development by taking advantage of its innate strengths [72]. Traditional industries spread all over mountainous areas, and their initial economic growth had a lot to do with the exploitation and utilization of resources [73]. This has led to tremendous pressures on the ecosystems of regions undergoing rapid economic development. Therefore, the regional health balance in the study area increased before 2006 (before implementation of the resource protection policies). The Moran Index shows that more regions in the study area were well developed, but they were far less developed than the core regions, failing to correct regional imbalances.

There were large differences in landforms, climate, ecology, and populations among the two sides of the Hu Line [30,74]. Overall, the pressure, state, response, and comprehensive health scores were all poorer on the northwest side of the Hu Line than the southeast (Figure 4 and Figure 5). The distinct topography and climate on the southeast side creates a harsh and fragile geographical environment [35], which makes regional resources scarce and slows social and economic development. Although the superior natural conditions of the Sichuan basin brought about rapid social and economic development, the area has faced severe challenges such as intensive farmland reclamation, overexploitation of forest and mineral resources, river pollution, and habitat destruction. These regions had lower and lower pressure scores over time (Figure 4a). Anthropogenic determinants involving social responses and concerns about environmental change have a strong impact on response scores [27]. Government decision-making focuses on supporting some core or hotspot regions, thereby improving their response scores. The ecosystem health has improved in most areas, but the situation in Lijiang appears to be very serious (Figure 5). Other studies have also shown that the ecosystem health in much of the region is deteriorating [1]. Jianchuan County, Yunlong County, Yongping County, and Yangbi County were found to be the four worst counties in the Dali ecosystem (Figure 8), which is in accordance with the results of previous study [40]. The vegetable industry in Pidu District is well developed, and it has suffered from agricultural non-point source pollution for a long time due to the excessive application of fertilizer, and the ecosystem of Pidu District was the unhealthiest county in Chengdu (Table 4). With the passage of time, the difference in ecosystem response scores between the southeast and northwest sides of the Hu Line have narrowed (Figure 6), indicating that more attention has been paid to ecologically fragile areas [75]. The gap in the pressure and health scores has been narrowing as the economy of the region on the southeast side of the Hu Line has grown. The long-term existence of the Hu Line is conditional on the comprehensive natural geographical conditions, which will not change in the near future [76]. This means that the center of gravity and spatial pattern of ecosystem health can only move along the Hu Line (Figure 7).

4.3. Suggestions and Implications

China, the world’s second-largest economy, has achieved amazing economic progress, with an average annual growth of 9.8% over the past 30 years. However, there is a huge disparity in social development and ecosystem health between the southeast and northwest sides of the Hu Line in China. In 2013, Premier Li Keqiang of the State Council questioned whether the pattern of the Hu Line could be broken. In the same year, the National New-Style Urbanization Plan (2014–2020) was promulgated and implemented. Accordingly, it is necessary to determine a method through which eco-sustainable development can be applied to the areas on both sides of the Hu Line in the Sichuan and Yunnan regions. To develop a smart model for breakthrough of the Hu Line suitable for Sichuan and Yunnan, it is essential to seek a relationship between geographical conditions and economic development, and to adjust social development to local conditions. At the national level, it is necessary to make overall arrangements for the ecological sustainable development of the entire country. The state should make construction of ecological civilization the goal, and resolve various practical conflicts, especially those associated with multisectoral management. Provinces or regions should guide and coordinate the ecological construction of various cities and counties. For example, areas with better institutional capacity can provide aid to resource-poor areas, and the ecologically sound places should undertake more population residence and economic construction through centralized resettlement. Prefecture-level cities should make detailed implementation and arrangements for the requirements of their superiors, but they must be tailored to local conditions. The northwest side of the Hu Line makes full use of local advantages, further developing ecotourism, promoting a distinctive culture, and improving access to domestic and foreign international markets. Furthermore, full use should be made of the economic resources on the southeast side of Hu Line, while developing technical innovations to foster green energy use and avoid excessive resource consumption and environmental pollution. The key management counties should be divided to identify those with weak and superior ecosystem health in each city (Table 4). Counties with poor ecosystem health need to be emphatically managed, while those leading in ecosystem health should play a leading role. In practical terms, strategies should include promoting regional ecosystem health, extending green industries in towns and townships in the western region, and elevating the self-supporting capability and ecological protection consciousness of local residents through basic education and professional training. Finally, urban and rural integration strategies should be implemented to achieve rural revitalization. The mind map shown in Figure 9 provides insightful suggestions to enable decision makers to achieve a smart breakthrough in the Hu Line that leads to long-term stability and prosperity.

5. Conclusions

This study systematically revealed the spatial evolution of regional ecosystem health at the south end of the Hu Line (Sichuan and Yunnan) over the past 20 years. The ecosystem health of most prefecture-level cities has improved from 2000 to 2016, and the number of unhealthy cities has fallen by 25%, but the situation in Lijiang appeared to be unsatisfactory. Overall, the gap between cities was widening, but the Moran Index results indicated that more high-quality cities have emerged. Because the pressure on and state scores of the regional ecosystems was highly dependent on natural geographical conditions, the comprehensive health conditions of the ecosystem southeast of the Hu Line were slightly better than those in the northwest. The spatial pattern and change direction of ecosystem health were consistent with the direction of the Hu Line. At the county level of the ecosystem health evaluation, the imbalance in the development of the cities was divided, and counties within cities that were lagging were identified, such as Pidu District in Chengdu and Danling in Meishan. To correct regional imbalances, a comprehensive and proactive policy framework for development of a smart breakthrough model of the Hu Line in Sichuan and Yunnan was put forward. Our study focused on the spatial evolution of ecosystem health and contributed to the reasonable planning of ecological and environmental protection with the goal of identifying a path toward breakthrough in the Hu Line.

However, this paper has no substantial contribution to the innovation of methods and the improvement of models. Future regional ecosystem health assessment requires a comprehensive system based on different ecological and biological information scales that also takes into account human health and cultural factors. Although the NDVI has been selected in many studies, there may be other reasonable results without using NDVI. Many areas, such as plateau areas, have low scores of ecosystem. Perhaps the ecological environment is not unhealthy, it may be fragile or have low carrying capacity. It may be unreasonable or inappropriate to apply a gravity model to ecosystem health research in this paper, although it is really used in economic geography and has sense. Other suitable indicators need to be explored or the assessment model and analytical method need to be improved.