Spatial-Temporal Variation and Driving Factors of Ecological Vulnerability in Nansi Lake Basin, China

Lake basins are one of the most significant areas of human–land interaction. It is essential for the region’s ecological protection and high-quality development to assess their ecological vulnerability (EV) and analyze the key driving factors of EV. Considering the characteristics of the lake basin, we chose 17 indicators to evaluate the EV of the Nansi Lake Basin based on the “sensitivity-resilience-pressure” (SRP) model. Then, spatial principal component analysis (SPCA) and a transfer matrix were used to analyze the spatial-temporal variation characteristics of the EV. Moreover, the optimal parameters-based geographical detector (OPGD) was applied to investigate the factors influencing the spatial heterogeneity of the EV. The results indicated that the EV of the Nansi Lake Basin was characterized by a circling spatial structure, with low values distributed in the Nansi Lake and its surrounding areas, as well as high values concentrated in the northwest. The EV of the Nansi Lake Basin decreased from 2010 to 2020, indicating that the overall ecological pressure in the Nansi Lake Basin decreased. Climatic factors, land use type, and habitat quality were the primary factors that influenced the spatial heterogeneity of the EV in the basin. Our findings can serve as policy guidelines for ecological management and the sustainable development of the Nansi Lake Basin and also contribute to the EV assessment of lake basins.


Introduction
With the increasing urbanization and deterioration of ecosystem functions, assessing ecological vulnerability (EV) has become one of the most crucial concerns for high economic and social growth as well as sustainable ecological development [1,2]. The connotation of EV, which evolved from the definition of vulnerability, was first introduced into ecology by Clements in the early 20th century [3]. It has since been adopted as a crucial research area by organizations and academic programs around the world, including the International Biological Program (IBP) and the Intergovernmental Panel on Climate Change (IPCC) [4,5]. According to the definition of the concept of EV based on the results of existing studies [6], it can be found that EV is an inherent property determined by its own structure. The EV assessment is often based on a specific spatial area and a specific time scale, and the dynamic changes in the degree of vulnerability of the research object were examined. Overall, the EV reflects the sensitivity of ecosystems to external disturbances and threats and their ability to self-regulate and recover [7,8].
The basins are complex ecosystems with strong interactions between natural processes and human activities [9]. Rapid urbanization and industrial land expansion have neglected the rational planning and protection of the basins' territorial space, reducing the carrying capacity of resources and the environment [10]. Consequently, an urgent issue of the EV appears in the basins, primarily manifesting in the encroachment and destruction of environmental spaces, severe water pollution, and the continued decline in ecosystem service functions [11,12]. The studies that have been conducted for the Nansi Lake Basin

Data Source and Processing
The study data primarily consists of topographic data, land use data, soil data, meteorological data, vegetation data, socio-economic data, and boundary data of the Nansi Lake Basin in 2010, 2015, and 2020 ( Table 1). The digital elevation model (DEM) data, land use data, GDP data, and population data are from the Resources and Environment Science and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn, accessed on

Data Source and Processing
The study data primarily consists of topographic data, land use data, soil data, meteorological data, vegetation data, socio-economic data, and boundary data of the Nansi Lake Basin in 2010, 2015, and 2020 (Table 1). The digital elevation model (DEM) data, land use data, GDP data, and population data are from the Resources and Environment Science and Data Center of the Chinese Academy of Sciences (https://www. resdc.cn, accessed on 6 July 2022), with GDP and population data for 2020 replaced by 2019 due to data availability. The soil data are from the National Cryosphere Desert Data Center (http://www.ncdc.ac.cn/portal/, accessed on 2 July 2022) with a spatial resolution of 1 km. The temperature and precipitation data for the basin and its surrounding meteorology were obtained from the China Meteorological Data Service Center (https://data.cma.cn/, accessed on 5 October 2021). The National Ecosystem Science Data Center (http://www.nesdc.org.cn/) provides the normalized difference vegetation index (NDVI) data with a spatial resolution of 30 m. The net primary productivity (NPP) data is from the MOD17A3 data product of the Nation Aeronautics and Space Administration (NASA) (https://ladsweb.modaps.eosdis.nasa.gov/, accessed on 8 July 2022). The Nansi Lake Basin boundary data was obtained from the Nanjing Institute of Geography & Limnology Chinese Academy of Sciences (http://www.niglas.ac.cn/, accessed on 10 May 2021). With the support of the ArcGIS (version 10.8) software platform, all spatial data were unified to a 1 km spatial resolution and the projected coordinate system of Krasovsky 1940 Albers.

Select Evaluation Indicators
Ecological vulnerability is affected by human activities and natural conditions. The SRP model is a comprehensive evaluation model that takes into account not only the structural characteristics and functions of ecosystems but also the external pressures to which they are subjected. This model covers the components of EV and has been widely used in the study area at different scales, which provides a more comprehensive description of the evolutionary pattern of the EV [20]. Considering the complex environmental characteristics in the Nansi Lake Basin, we selected 17 indicators from ecological sensitivity, resilience, and pressure (Table 2). Thus, the vulnerability evaluation system was developed based on the SRP model according to the results of previous research [1,47]. High-intensity human activities and dramatic climate change have affected lake basins' runoff and water security, leading to an increasing threat of ecological risk [10,20]. Therefore, climate and human disturbance should be considered essential aspects of the assessment index system. (1) Ecological sensitivity Ecological sensitivity reflects the sensitivity of the ecosystems in the basin to external disturbances. It includes terrain conditions, surface conditions, and climatic factors. Elevation, slope, and topographic relief can reflect the topographic conditions that positively impact the EV. Different land use types have varying influences on the EV [1], and they were graded based on previous research findings (Table 3). High soil erosion suggested that the surface was more prone to damage and that the EV was higher [48]. The degree of soil erosion was calculated using the revised universal soil loss equation (RUSLE) [49][50][51] (see Supplementary Materials S1.1). Rainfall erosivity, soil erodibility, slope length, slope gradient factor, vegetation cover factor, and erosion control practice factor should be calculated in the RUSLE. Among them, the erosion control practice factor of the cultivated land was assigned under different slope ranges (see Supplementary Materials, Table S1). Then, the results of the soil erosion were assigned a graded value under the National Soil Erosion Classification and Grading Standard (Table 3). In addition, climatic factors were negative indicators of EV. Areas with high temperatures, abundant precipitation, and high dryness typically had a wetter climate, more vigorous vegetation growth [52,53], richer biodiversity, and lower EV. The average annual temperature and precipitation were obtained by the ordinary kriging interpolation of meteorological station data using ArcGIS (version 10.8) software; based on this, dryness was calculated using the de Martonne method (see Supplementary Materials S1.2). (2) Ecological resilience The ability of the ecosystem to recover to its original state under disturbance is called ecological resilience. The NDVI, NPP, landscape diversity and habitat quality are all negative indicators of EV. Because densely forested areas are rich in biological resources, habitat quality is high, and ecosystems are relatively stable [22,53]. The SHDI (Shannon's diversity index) was calculated by Fragstats (version 4.2) software based on land use data from each period to respond to landscape diversity. The habitat quality index was calculated for the Nansi Lake Basin using the habitat quality module of InVEST (version 3.10.2) software (see Supplementary Materials S1.3), which required the input of threat factor attributes and the sensitivity of habitat types to each threat factor (see Supplementary Materials,  Tables S2 and S3). More importantly, the hydrological condition has an important impact on the ecological security of lake basins. We used water yield services and water purification services to characterize hydrological conditions. Among them, water yield services played a key role in improving the hydrological condition of the basin and regulating the regional water cycle [54]. Areas with a low water yield tend to have low vegetation cover, which in turn leads to poor biological survival conditions and lower habitat quality. Water purification services reduce nitrogen and phosphorus nutrients in surface runoff, regulate regional non-point source pollution, and are important indicators of the health of the water environment [9]. The overload of nitrogen and phosphorus output exacerbates the degradation of habitat quality and makes the ecological resilience capacity weak. We measured the water yield and nitrogen and phosphorus output of the Nansi Lake Basin using the InVEST (version 3.10.2) software (see Supplementary Materials S1.4 and S1.5), where the calculation of water purification services requires nitrogen and phosphorus output-related parameter values (see Supplementary Materials, Table S4).
(3) Ecological pressure Ecological pressure responds to the intensity of disturbance to the ecological environment caused by population and economic activities characterized by population and GDP density. Areas with a high population density and economic development have a higher demand for and exploitation of resources, exacerbating the deterioration of the ecological environment. Hence, population density and GDP density are positive indicators of EV.

Standardization of Indicators
Indicators are distributed at various scales with different units, making them unable to be compared or integrated. Therefore, the range method was applied to standardize the quantitative indicators to a uniform scale, except for the land use type and degree of soil erosion, before calculating the results of the comprehensive EV. To standardize the positive and negative indicators, the following formula can be used: where Y i is the standardized result of the index factor i, ranging from 0 to 1, X i is the original data of the index factor i, and X min and X max represent the minimum and maximum values of the index i, respectively.

Spatial Principal Component Analysis
The spatial principal component analysis (SPCA) method is based on the support of the ArcGIS (version 10.8) software to evaluate EV. This method can reduce the dimensions of 17 evaluation indicators and recombine them into mutually unrelated comprehensive indicators. It can effectively avoid the influence of the correlation between the original indicators on the EV assessment. The top m principal components with a cumulative contribution rate greater than 85% are extracted to replace the original indicators for analysis. The results are calculated using the following formula: where SPCA j is the value of the principal component j, Y i is the standardized value of the original indicator i of each principal component, Z ij is the eigenvector of each original indicator i of the principal component j, EVI is the EV index of the study area, Q j is the contribution rate of the principal component j and m is the number of principal components with a cumulative contribution rate greater than 85%.

The Classification of EV
To make it easier to compare the EV across years and regions, we standardize the results of EVI using the range method [20,47], which is calculated as follows: where SEVI is the standardized value of EVI, ranging from 0 to 1, EVI is the EV index and EV I max and EV I min represent the maximum and minimum values of EVI, respectively. Using the equivalence method, the SEVI was classified into five levels based on the previous research findings [20,47]. Level I is slight vulnerability (0 ≤ SEV I < 0.2); Level II is mild vulnerability (0.2 ≤ SEV I < 0.4); Level III is moderate vulnerability (0.4 ≤ SEV I < 0.6); Level IV is severe vulnerability (0.6 ≤ SEV I < 0.8); and Level V is extreme vulnerability (0.8 ≤ SEV I ≤ 1).
In order to compare and analyze the changes in the overall EV of the basin, a multiplier model was introduced to measure the comprehensive EV of the whole basin by year [47], which is as follows: where CEVI is the comprehensive EV index of the Nansi Lake Basin, G i is the classification level value (I, II, III, IV, V) of SEVI, A i is the area of the EV level corresponding to the results of SEVI, and S is the total area of the Nansi Lake Basin.

Transfer Matrix
The transfer matrix can be used to analyze the state of the system and its transfer changes at the start and end of the study period. It was introduced to analyze the transfer changes in the various levels of EV in the Nansi Lake Basin through the following equation: where S stands for the study area, n denotes the number of SEVI classification categories (n = 1, 2, · · · , 5), and i and j represent the SEVI classification categories at the beginning and end of the study period, respectively.

Optimal Parameters-Based Geographical Detector
The geographical detector model is a technique that is extensively used for spatial stratified heterogeneity analysis [42]. In the traditional GD model, factor detection measures the influence of independent variable X on the spatial heterogeneity of the dependent variable Y in terms of the q value; interaction detection measures influence the two-two interactions of different independent variables on the spatial heterogeneity of the dependent variable Y in terms of the q value, using the following equation: where q denotes the explanatory power of a single or interacting independent variable, with larger values of q indicating more substantial explanatory power, and h denotes the stratification of a single or interacting independent variable or dependent variable. N h and N are the numbers of cells in stratum h and the whole region, σ 2 h and σ 2 are the variances of the Y values for stratum h and the whole region, respectively. SSW and SST are the sum of the variances within the stratum and the total variance of the whole region, respectively.
However, spatial data discretization has lacked accurate quantitative assessment in previous studies. An optimal parameters-based geographical detector (OPGD) model was developed to address this issue for a more accurate spatial driving analysis of the driving factors [44]. The OPGD model chooses the set of parameters (discrete method and the number of intervals) with the highest single factor q value to determine the best spatial discretization technique for continuous variables. Compared with the GD model, the OPGD model can determine the best spatial data discretization method through quantitative evaluation, which effectively improves the accuracy of spatial data analysis [44]. Meanwhile, the identification of drivers can differ at different spatial scales [55]. Therefore, based on the research results of existing scholars [20,56], we chose the 1 km grid scale to explore the driving mechanisms of EV in the Nansi Lake Basin at different times. With the support of the ArcGIS (version 10.8) software platform, 1 km × 1 km fishing nets were created to generate the fishing net points. Raster data from the SEVI and 17 evaluation indicators of the Nansi Lake Basin were extracted to the fishing net points for each period, yielding a total of 27,947 grids that served as the base data for the OPGD model. The parameters at the highest q value of the continuous variable were calculated using a combination of discrete methods (quantile method, geometric method, standard deviation method, natural breakpoint method, and equal method) and break numbers (3-6 categories) with the aid of the GD package in the R (version 4.2.1) software. The main influencing factors of EV changes in the Nansi Lake Basin were identified using factor detection and the interaction detection of the OPGD model.

The Spatial Distribution of EV
The SPCA was used in ArcGIS (version 10.8) software to calculate EVI for the years 2010, 2015, and 2020. Specifically, we extracted the top six principal components with a cumulative contribution rate greater than 85% instead of the original indicators to calculate EVI. The eigenvalues and contribution rates of principal components are shown in Table 4. After standardizing the data of EVI, the SEVI was obtained and divided in accordance with the classification criteria in order to better compare and analyze the spatial distribution of EV. The spatial distribution of SEVI in the Nansi Lake Basin in 2010, 2015, and 2020 is shown in Figure 2. The overall EV showed significant regional differences that increased from the southeast to the northwest and exhibited a circling spatial structure. The ecological conditions in the basin were not good in 2005, mainly in the moderate vulnerability category accounting for approximately 34.93% of the whole basin (Table 5). Furthermore, only 10% or less of the area was slightly and mildly vulnerable, concentrated in Nansi Lake and its surroundings. In 2015, moderate vulnerability decreased by about 6.30% compared to 2010, but the area of severe vulnerability expanded significantly, accounting for 43.69%. Combined with the land use status, we also found that the extreme vulnerability was more scattered in the construction land. In 2020, the percentage of severe vulnerability decreased significantly, only accounting for 2.33%, with severe vulnerability down roughly 22.22% from 2015. In general, the CEVI has decreased by 20.48% from 2010 to 2020, indicating that there was a greater improvement in the basin's ecosystem. It can be seen that the unique ecological restoration and treatment (with Nansi Lake as the core) carried out in recent years has been quite effective and contributed to the improvement of the ecosystem function of Nansi Lake and its surrounding areas. However, the ecological problems in the northwest part of the Nansi Lake Basin cannot be ignored. vulnerability was more scattered in the construction land. In 2020, the percentage of severe vulnerability decreased significantly, only accounting for 2.33%, with severe vulnerability down roughly 22.22% from 2015. In general, the CEVI has decreased by 20.48% from 2010 to 2020, indicating that there was a greater improvement in the basin's ecosystem. It can be seen that the unique ecological restoration and treatment (with Nansi Lake as the core) carried out in recent years has been quite effective and contributed to the improvement of the ecosystem function of Nansi Lake and its surrounding areas. However, the ecological problems in the northwest part of the Nansi Lake Basin cannot be ignored.

Dynamic Changes of EV
In order to better comprehend EV change from 2010 to 2020, we used the ArcGIS

Dynamic Changes of EV
In order to better comprehend EV change from 2010 to 2020, we used the ArcGIS (version 10.8) software to calculate the transfer matrix of SEVI during the periods 2010-2015 and 2010-2020 (see Supplementary Materials, Tables S5 and S6). The results were visualized using chord diagrams in Origin (version 2022b) software ( Figure 3). During 2010-2015, moderate and severe vulnerability dominated the basin. Except for the mutual transfer of the same category, the conversion from moderate vulnerability to severe vulnerability was the largest, covering about 3441 km 2 , implying that the ecological environment of some moderate vulnerability continued to deteriorate. Additionally, the slight and mild vulnerability remained stable. Compared with 2010-2015, the vulnerability levels of 2010-2020 changed significantly. There were 4903 km 2 regions experiencing the transition from moderate vulnerability to mild vulnerability, followed by a shift from severe vulnerability to moderate vulnerability, indicating that the overall ecological condition of the Nansi Lake Basin improved.

Different Administrative Regions of EV
The mean SEVI of counties in the Nansi Lake Basin was extracted from vector d of county-level administrative units using the zoning statistics function of ArcGIS (vers 10.8) software (Table 6). Combined with Figure 2, it was clear that counties located on northwestern were under consistently high vulnerabilities in the ecosystem, while southeastern cities were greatly reduced. As shown in Table 6, the mean SEVI of Weishan County has always been in the mil vulnerable zone. For a long time, the county has been committed to ecological restorati actively carrying out projects such as converting farmland to wetlands and greening b

Different Administrative Regions of EV
The mean SEVI of counties in the Nansi Lake Basin was extracted from vector data of county-level administrative units using the zoning statistics function of ArcGIS (version 10.8) software (Table 6). Combined with Figure 2, it was clear that counties located on the northwestern were under consistently high vulnerabilities in the ecosystem, while the southeastern cities were greatly reduced. As shown in Table 6, the mean SEVI of Weishan County has always been in the mildly vulnerable zone. For a long time, the county has been committed to ecological restoration, actively carrying out projects such as converting farmland to wetlands and greening barren hills. It also conducted large-scale ecological restoration work for the Nansi Lake, effectively improving its ecological carrying capacity. The EV of the counties surrounding the Nansi Lake, such as Peixian, Fengxian, and Yutai, had been reduced to a mildly vulnerable level under the strengthening of ecological protection and restoration with the Nansi Lake as the core. In contrast, Wenshang, Liangshan, and Yuncheng in the northwest were almost always in the severely vulnerable zone, which means that they had lower ecological resilience to disturbance. Among them, Liangshan had a particularly severe EV problem, with a 0.8543 mean SEVI in 2015, placing it in the extremely vulnerable zone. This county was rich in mineral resources, but the over-exploitation of non-coal mineral resources, including single-point limestone mines, seriously damaged the ecological landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods.
"△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○" represents the natural breakpoint method. Table 7 shows that the q value of X12 (habitat quality) in 2010 was 0.748, which is more significant than the other variables, indicating that it was the dominant factor affecting the EV in the Nansi Lake Basin that year. The variables X4 (land use type) and X13

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○" represents the natural breakpoint method. Table 7 shows that the q value of X12 (habitat quality) in 2010 was 0.748, which is more significant than the other variables, indicating that it was the dominant factor affecting the EV in the Nansi Lake Basin that year. The variables X4 (land use type) and X13 The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each ind (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosion ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth △" is by classification criteria, "◆" is by professional experience, method, "□" represents the geometric method, "•" represents the stan represents the natural breakpoint method. Table 7 shows that the q value of X12 (habitat quality) in more significant than the other variables, indicating that it was t ing the EV in the Nansi Lake Basin that year. The variables X4

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○" represents the natural breakpoint method. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○" represents the natural breakpoint method. Table 7 shows that the q value of X12 (habitat quality) in 2010 was 0.748, which is more significant than the other variables, indicating that it was the dominant factor affecting the EV in the Nansi Lake Basin that year. The variables X4 (land use type) and X13 (water yield) were followed, with q values of 0.746 and 0.537, respectively. In 2015, the top

Factor Detection Results
The OPGD model was used to investigate the impact of 17 e X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each ind (Table 7). Among them, X4 (land use type) was spatially discreti use/cover change (LUCC) classification standard, and X5 (soil er tized by the professional experience of the National Soil Erosion ing Standard. Then, the rest indicators were discretized by the parameters (discrete method and the number of intervals) when Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth "△" is by classification criteria, "◆" is by professional experience, method, "□" represents the geometric method, "•" represents the stan represents the natural breakpoint method. Table 7 shows that the q value of X12 (habitat quality) in more significant than the other variables, indicating that it was t ing the EV in the Nansi Lake Basin that year. The variables X4 (water yield) were followed, with q values of 0.746 and 0.537, res

Factor Detection Resul
The OPGD model was u X17) on the spatial heteroge determine the optimal spatia (Table 7). Among them, X4 ( use/cover change (LUCC) cl tized by the professional exp ing Standard. Then, the res parameters (discrete method Taking 2015 as an example, vation), when divided into s 0.066 □ "△" is by classification criteri method, "□" represents the geo represents the natural breakpoi Table 7 shows that the more significant than the oth ing the EV in the Nansi Lak (water yield) were followed,

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○" represents the natural breakpoint method.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○"

Factor Detection Results
The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each in (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosio ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth ■ "△" is by classification criteria, "◆" is by professional experience, method, "□" represents the geometric method, "•" represents the stan

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. "△" is by classification criteria, "◆" is by professional experience, "■" represents the quantile method, "□" represents the geometric method, "•" represents the standard deviation method, "○"

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each in (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosio ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth

Factor Detection Results
The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each ind (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosion ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each ind (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosion ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each ind (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosion ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each in (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosio ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each in (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosio ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. landscape and functions. Meanwhile, the ecological restoration and treatment of the geological environment of mines was expensive and took a long time.

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. The OPGD model was used to investigate the impact of 17 X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2 determine the optimal spatial discretization method for each in (Table 7). Among them, X4 (land use type) was spatially discret use/cover change (LUCC) classification standard, and X5 (soil e tized by the professional experience of the National Soil Erosio ing Standard. Then, the rest indicators were discretized by th parameters (discrete method and the number of intervals) whe Taking 2015 as an example, using the standard deviation metho vation), when divided into six categories, were greater than oth The OPGD model was used to investigate the impact o X17) on the spatial heterogeneity of SEVI in 2010, 2015, an determine the optimal spatial discretization method for eac (Table 7). Among them, X4 (land use type) was spatially dis use/cover change (LUCC) classification standard, and X5 (so tized by the professional experience of the National Soil Ero ing Standard. Then, the rest indicators were discretized b parameters (discrete method and the number of intervals) w Taking 2015 as an example, using the standard deviation m vation), when divided into six categories, were greater than Table 7. Spatial discretization methods and factor detection result

Factor Detection Results
The OPGD model was used to investigate the impact of 17 evaluation indicators (X1-X17) on the spatial heterogeneity of SEVI in 2010, 2015, and 2020, respectively. We can determine the optimal spatial discretization method for each indicator through the model (Table 7). Among them, X4 (land use type) was spatially discretized according to the land use/cover change (LUCC) classification standard, and X5 (soil erosion degree) was discretized by the professional experience of the National Soil Erosion Classification and Grading Standard. Then, the rest indicators were discretized by the optimal combination of parameters (discrete method and the number of intervals) when maximizing the q value. Taking 2015 as an example, using the standard deviation method, the q values of X1 (Elevation), when divided into six categories, were greater than other discrete methods. The OPGD model was used to investigate the imp X17) on the spatial heterogeneity of SEVI in 2010, 20 determine the optimal spatial discretization method fo (Table 7). Among them, X4 (land use type) was spatia use/cover change (LUCC) classification standard, and tized by the professional experience of the National S ing Standard. Then, the rest indicators were discreti parameters (discrete method and the number of inter Taking 2015 as an example, using the standard deviat vation), when divided into six categories, were greate  Table 7 shows that the q value of X12 (habitat quality) in 2010 was 0.748, which is more significant than the other variables, indicating that it was the dominant factor affecting the EV in the Nansi Lake Basin that year. The variables X4 (land use type) and X13 (water yield) were followed, with q values of 0.746 and 0.537, respectively. In 2015, the top three driving variables (X6-X8) were all climatic factors, which showed a strong impact on the spatial pattern of EV. In 2020, the most influential factors in EV were variables X7 (average annual precipitation), X8 (dryness), and X13 (Water yield). In short, it can be seen that over time, habitat quality, average annual temperature, average annual precipitation, and water yield have become the main drivers of EV spatial heterogeneity. Noteworthily, the influence of climatic factors such as average annual precipitation and dryness has gradually increased.

Interaction Detection Results
In our study, the OPGD model was also used to detect the effects of bivariate interactions on the EV spatial heterogeneity in the Nansi Lake Basin. The interaction results were visualized with the help of heat maps in Origin (version 2022b) software (Figure 4).

Interaction Detection Results
In our study, the OPGD model was also used to detect the effects of bivariate interactions on the EV spatial heterogeneity in the Nansi Lake Basin. The interaction results were visualized with the help of heat maps in Origin (version 2022b) software ( Figure 4). From 2010 to 2020, the results of the bivariate interactions were either non-linearly enhanced or bi-factorially enhanced. This means that the interaction of any two variables was more significant than the effect of a single variable on the EV. In 2010, the interaction between variables X12 (habitat quality), X4 (land use type), and other variables significantly influenced the EV spatial heterogeneity. The largest q value was the interaction between variable X12 (habitat quality) and X6 (average annual temperature) at 0.914. In 2015, the interaction between variable X12 (habitat quality), X4 (land use type), and climate factors (X6-X8) was greater than 0.85, most likely relating to single-factor detection results. In 2020, the interaction between the variables X6 (average annual temperature) and X7 (average annual precipitation) was the strongest, with a q value of 0.955. Moreover, the q values of the interactions between X7 (average annual precipitation), X8 (dryness), and other variables were higher than 0.85, indicating that these two variables were significant determinants of EV in that particular year.
In general, the interaction between X12 (habitat quality), X4 (land use type), and climatic factors (X6-X8) better explained the EV spatial heterogeneity. It also suggested that climatic factors such as average annual temperature, average annual precipitation, and dryness were critical drivers of EV in the Nansi Lake Basin.

Discussion
Green basin governance is a fundamental strategic requirement for promoting highquality regional development. The Nansi Lake Basin is an essential agricultural production base and green ecological security barrier in Shandong Province. The EV assessment of the basin can provide a scientific reference for ecological governance decisions.
Our study evaluated the EV in the Nansi Lake Basin using the SRP model. The results showed that EV exhibited a circling structure that is consistent with the spatial characteristics of lake basins [24]. The areas with lower EV were mainly concentrated in the Nansi Lake and the low hilly areas to the east, which were usually rich in biodiversity and had high habitat quality [32]. In recent years, governments have continued to strengthen ecological co-protection and management in the Nansi Lake Basin. The provincial government uses Nansi Lake as the core area for ecosystem protection and carries out ecological From 2010 to 2020, the results of the bivariate interactions were either non-linearly enhanced or bi-factorially enhanced. This means that the interaction of any two variables was more significant than the effect of a single variable on the EV. In 2010, the interaction between variables X12 (habitat quality), X4 (land use type), and other variables significantly influenced the EV spatial heterogeneity. The largest q value was the interaction between variable X12 (habitat quality) and X6 (average annual temperature) at 0.914. In 2015, the interaction between variable X12 (habitat quality), X4 (land use type), and climate factors (X6-X8) was greater than 0.85, most likely relating to single-factor detection results. In 2020, the interaction between the variables X6 (average annual temperature) and X7 (average annual precipitation) was the strongest, with a q value of 0.955. Moreover, the q values of the interactions between X7 (average annual precipitation), X8 (dryness), and other variables were higher than 0.85, indicating that these two variables were significant determinants of EV in that particular year.
In general, the interaction between X12 (habitat quality), X4 (land use type), and climatic factors (X6-X8) better explained the EV spatial heterogeneity. It also suggested that climatic factors such as average annual temperature, average annual precipitation, and dryness were critical drivers of EV in the Nansi Lake Basin.

Discussion
Green basin governance is a fundamental strategic requirement for promoting highquality regional development. The Nansi Lake Basin is an essential agricultural production base and green ecological security barrier in Shandong Province. The EV assessment of the basin can provide a scientific reference for ecological governance decisions.
Our study evaluated the EV in the Nansi Lake Basin using the SRP model. The results showed that EV exhibited a circling structure that is consistent with the spatial characteristics of lake basins [24]. The areas with lower EV were mainly concentrated in the Nansi Lake and the low hilly areas to the east, which were usually rich in biodiversity and had high habitat quality [32]. In recent years, governments have continued to strengthen ecological co-protection and management in the Nansi Lake Basin. The provincial government uses Nansi Lake as the core area for ecosystem protection and carries out ecological afforestation and greening along the lake's shoreline to build ecological corridors. The local governments have implemented a series of water environmental protection projects to improve the ecosystem functions of Nansi Lake. In consequence, ecosystem services such as water storage, water purification, soil conservation, and biodiversity maintenance have significantly enhanced and promoted the ecological security of the Nansi Lake Basin.
Our results showed that Weishan county had the lowest EV, meaning that it is ecologically safer with a higher capacity for ecological recovery and resistance. Similarly, the EV of counties around Nansi Lake has been reduced due to a series of ecological restoration policies for the Nansi Lake Nature Reserve [45]. Conversely, the northwest parts of the basin have higher EV values, indicating lower ecological security. The study of Lv et al. (2012) can indirectly verify this conclusion [57]. In order to improve the whole ecological safety of the basin, ecological protection and restoration policies should be considered and integrated, focusing on systemic integrity and strengthening common protection and joint management. It is necessary to promote the construction of ecological safety corridors in the whole basin and improve the ecosystem functions of the regions, such as water storage and biodiversity maintenance. On the one hand, governments should continue to enhance ecosystem protection with Nansi Lake as the core and carry out ecological afforestation and greening along the lake's shoreline. On the other hand, more efforts are needed to prevent further ecological deterioration in the northwest regions and increase the effectiveness of land use to lessen the ecological strain of human activities. In addition, it is also necessary to restore the forest landscape of the low hills and green barren slopes in the eastern parts for soil and water conservation.
Climate factors were the dominant factor and became increasingly important over time based on the results of the driving mechanism. Moreover, the Nansi Lake Basin suffered a drought in 2014, which threatened water quality safety, severely damaged the ecological structure and function of the lake, and affected the fisheries industry. Therefore, climate change is a challenging task that must be addressed concurrently with the ecological management of the lake's basin. The governments should strengthen the monitoring of meteorological indicators as well as ecological changes in the basin. Simultaneously, the cascade effects of climate change in ecological analysis, prediction, and risk warning should be of concern in order to improve the ecological security of the whole basin.

Conclusions
Based on the SRP model framework, the ecological characteristics of the lake basins were fully considered. There are 17 indicators selected to construct the EV assessment system for the Nansi Lake Basin, which mainly includes the terrain conditions, surface conditions, climatic factors, vegetation conditions, hydrological conditions, and human disturbance. Our study evaluated the spatial-temporal evolution characteristics of EV in the Nansi Lake Basin from 2010 to 2020. Then, we used the OPGD model to detect the drivers of EV spatial heterogeneity. The results showed that the spatial distribution of SEVI in the basin was higher in the northwest than in the southeast from 2010 to 2020 and exhibited a circling spatial structure. According to the results of spatial and temporal changes in ecological vulnerability for different categories, we can see that the ecosystem of the basin improved obviously. The EV changes of different county-level administrative units indicated that the mean SEVI of southeastern cities generally decreased from 2010 to 2020, primarily distributed around Nansi Lake. In contrast, most cities in the northwest always maintained high values, particularly in the county of Liangshan. Habitat quality, land use type, average annual temperature, average annual precipitation, and water yield were the main drivers of EV spatial heterogeneity. The key roles of climate factors demonstrated a clear growth in strength. Moreover, the EV was jointly influenced by many factors. From the results of interaction factor detection, the strongest degree of interaction was found for climatic factors, land use type, and habitat quality. Therefore, the impacts of land use and climate change on ecological security patterns need to be fully considered in the future environmental management of lake basins. In summary, the findings confirmed the applicability of the SRP model in the EV assessment of the lake basin and also provided support for ecological protection and restoration in the Nansi Lake Basin along with decision-making in the EV assessment of similar lake basins.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/ijerph20032653/s1. Table S1: P-value of cultivated land under different slope ranges; Table S2: Maximum distance and weights of the threats affecting habitat quality; Table  S3: The sensitivity of habitat types for each threat factor; Table S4: Parameter values related to nitrogen and phosphorus output;

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request and the approval of the data owner.

Conflicts of Interest:
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.