3.2.2. Effect of the GWR Model

In order to illustrate the applicability of GWR, this study first used the OLS model to simulate. As shown in (Table 6), the R<sup>2</sup> and Adjusted R<sup>2</sup> of the GWR model in the three years are larger than those of the OLS model, indicating that the simulation effect is more accurate and representative. Meanwhile, the AICc of the GWR model is smaller than that of the OLS model, and the difference is greater than 3.0, which demonstrated that the GWR model is more applicable. Therefore, the GWR model can accurately explain the relationship between WESV and each explanatory variable and the spatial heterogeneity.

**Table 6.** Statistical test of OLS and GWR in 2010, 2015, and 2020.


### 3.2.3. Per Capita GDP

Figure 4 displayed the spatiotemporal distribution of the per capita GDP impact on WESV. The darker the color, the greater the positive effect. In 2010 (Figure 4a) and 2020 (Figure 4c), WESV and Per Capita GDP were positively correlated. The magnitude of the coefficient indicated that, in general, the 2020 Per Capita GDP had a stronger impact on WESV. In 2010, the correlation relationship gradually increased from west to east, with Huanglong, Huangling and Yanchang in the southeast, Wubao in the east and Fugu in the northeast being the most influential. In 2020, the impact of Per Capita GDP on WESV gradually increased from southwest to northeast, and WESV in Fugu, Shenmu and Jia County was more sensitive to Per Capita GDP. Because Yulin is the base of energy and chemical industry in China, the economy had developed rapidly after 2008, and the demand for water had increased. The combined effect of the two led to changes in the spatial distribution from 2010 to 2020. Nevertheless, in 2015 (Figure 4b), the WESV of 11 districts and counties in the northern and central parts of the study area showed a negative correlation with Per Capita GDP, and the negative correlation effect was strongest in the northeast. This illustrated that during the period from 2010 to 2015, the protection of water resources and water environment in the northern region was neglected due to the great economic development, which was consistent with the reality of the reduction of the water area in this region. It also showed that from 2015 to 2020, the water resources condition in the northern part of the study area and the ecological services provided had been greatly restored.

**Figure 4.** Temporal and spatial distribution of the impact of Per Capita GDP on WESV in different years: (**a**) 2010; (**b**) 2015; and (**c**) 2020.

#### 3.2.4. Population Density

In a certain area, population density determines the demand for WES. Figure 5 displayed the spatiotemporal distribution of WESV response to Population Density. Throughout the study period, WESV exhibited a negative correlation with population density, and the negative correlation gradually weakened with time. From 2010 to 2015, the more northerly the geographical location is, the more sensitive WESV is to changes in population density. Compared with Figure 5a,b, although the overall negative correlation was slightly enhanced, the area of region with the lowest level of negative correlation was increased. As shown in Figure 5c, the spatial distribution pattern of the impact of population density on WESV had fundamentally changed in 2020, and the coefficient increased from the southeast to the outside. The larger the coefficient, the weaker the negative correlation.

**Figure 5.** Temporal and spatial distribution of the impact of Population Density on WESV in different years: (**a**) 2010; (**b**) 2015 and (**c**) 2020.
