In this study, the high-resolution EEQ datasets of China in 2000, 2010, and 2020 were obtained from the National Earth System Science Data Center. National Science and Technology Infrastructure of China (http://www.geodata.cn). The spatial resolution is 1 km × 1 km. Due to the unavailability of data for 2000, data from 2001 was used as a substitute. The original sources of this data were MOD09A1, MOD11A2, and MCD12Q1, which were calculated in the Google Earth Engine platform (Xu et al., 2021; Li et al., 2023). The verification results showed that EEQ dataset was largely consistent with the environmental index provided by China’s Ministry of Ecology and Environment, with an RMSE of 0.11 and an R² of 0.71 (Xu et al., 2021). EEQ dataset has filled the research gap in the field of EEQ monitoring in China. The population data for 2000, 2010, and 2020 were derived from the fifth, sixth, and seventh censuses of the county units, and the PD is the ratio of the number of people living in the county to the area of that county (unit: person/km²).

EEQ dataset was constructed using four indicators: humidity, greenness, heat, and dryness, each of which directly reflects the ecosystem status. Specifically, humidity, greenness, heat, and dryness were respectively characterized by humidity index, vegetation index, surface temperature, and bare soil index. After the standardization of different ecological indices, the principal component analysis was employed to reduce the dimensionality of the standardized indicators, and the contribution rate of each indicator to the first principal component was determined (Li et al., 2023). The specific calculation equation can be referred to Xu et al. (2021).

In this study, three sets of regression models were conducted to investigate the impact of population on EEQ change (Chi, 2010; Chi and Ho, 2018), including the ordinary least squares (OLS) model, three spatial regression models [spatial lag model (SLM), spatial error model (SEM), and spatial error model with lagged-dependence (SEMLD)], and geographically weighted regression (GWR) model. In the OLS model, population pressure was regarded as an independent variable, and its direct effect on EEQ change was reflected by the regression coefficient, without considering the spatial impact of neighboring units. In the SLM model, the influence of population pressure on EEQ change was not only reflected in the direct effect but also reflected in the indirect effect of population pressure from adjacent units on the current EEQ change by the spatial lag term. As for the SEM model, the spatial autocorrelation of the error term was incorporated into the model, thus improving the accuracy of the estimation. This model can reveal the potential spatial effects of population pressure on EEQ change. The SEMLD model included the spatial lag and spatial error models and spatial autoregressive models enhanced by adding spatial lag explanatory variables (Chi and Zhu, 2008). The GWR model was used to detect whether there was spatial heterogeneity in the dependent and independent variables in the study, and was mainly used as a tool for explanation (Videras, 2014). The influence of population pressure on EEQ change may be different in different spatial locations. By calculating the regression coefficient of each spatial unit, the degree of local influence of population pressure on EEQ change can be obtained. The GWR model can more accurately reflect the spatial heterogeneity of population pressure on EEQ change. The OLS and spatial regression models were implemented in Geoda095i and the GWR model was implemented in the ArcTool function of ArcGIS 10.3. The model calculation equations were shown in the corresponding references (Chi and Zhu, 2008; Videras, 2014).

Local Moran’s I was often used to estimate the spatial autocorrelation of adjacent features between geographical factors (Anselin, 1995). To identify the stress areas of the population on EEQ, bivariate local autocorrelation was introduced in this study to provide a feasible method for identifying population pressure areas (Chi and Ho, 2018). Through the analysis of local Moran’s I, the relationship between the change of population size and the change of EEQ was identified. The “high-low” type was defined as a population-EEQ improvement zone, indicating that a 1% decrease in population size corresponded to an increase in EEQ by more than 1%. The “low-high” type was defined as a population-EEQ pressure zone, showing that an increase in the number of population size in 1% of the area corresponded to a decrease in EEQ of more than 1%. The “high-high” type indicated that the increase in the number of regional populations corresponded to the increase in EEQ, and was defined as the improvement of the coordinated development zone. The “low-low” type indicated that the decrease of the regional population corresponded to the decrease of the regional EEQ, and was defined as an unsustainable zone. The calculation Equations 1–3 was as follows: I k l i = z k i ∑ j = 1 n   w i j z l j (1) z k i = X k i − X k ¯ σ k (2) z l j = X l j − X l ¯ σ l (3) where, w i j is the spatial matrix between units i and j , X k i is the value of property k of spatial unit i , X l j is the value of the property l of the spatial unit j , X k ¯ and X l ¯ are the average value of elements k and ⋅ l , respectively, σ k and σ l are the variance of attributes k and ⋅ l , respectively.

The distribution map of China’s PD shows that, from 2000 to 2020, the spatial pattern of “more in the southeast and less in the northwest” remained largely unchanged (Figure 1). Densely, populated areas were mainly located in the eastern coastal regions, the Huang-Huai-Hai Plain (HHHP), the Yangtze River Delta (YRD), the Sichuan Basin (SCB), the middle reaches of the Yangtze River (MRYR), and the Pearl River Delta (PRD). On the contrary, in the northwest region, except for a few provincial capitals such as Urumqi and Xining with high PD, most areas had sparse populations.

The percentage of county units with population decreased between 2000–2010 and 2010–2020 were 35.8% and 55.5%, respectively. From 2000 to 2010, 6.8% of the counties had a population decrease with a magnitude greater than 50 person/km², 22.1% of the counties had a population increase greater than 50 person/km², and the proportions were 13.3% and 21.9%, respectively, from 2010 to 2020. Between 2000 and 2010, the regions with significant population reduction were mainly distributed in central Sichuan, southern Hubei, southeastern Henan, Anhui, and northern Jiangsu, while the regions with significant population growth were mainly distributed in the central areas of the three major urban agglomerations of Beijing-Tianjin-Hebei, YRD, and PRD.

During 2010–2020, the number of counties with population decline had increased evidently, and the main distribution area had gradually spread to the middle and lower reaches of the Yellow River and south of the Yangtze River (Figure 2). Moreover, the areas of PD decline in some parts of Xinjiang and Northeast China had further expanded, and the magnitude of the decline had further increased. In addition, the areas with a slight decrease in PD had the largest area and the widest distribution. However, the main distribution areas of counties with increasing populations also expanded, and the PD in Hainan Province shifted from decreasing to rapidly increasing. Other regions with sporadic distribution of counties exhibited increased PD.

In 2000, 2010, and 2020, China’s average EEQ scores were 0.393, 0.391, and 0.391, respectively, indicating a trend of decreasing and then slightly stabilizing in EEQ (Figure 3). Overall, China’s EEQ showed a pattern of low in northwestern China and high in southeastern and northeast China. Moreover, EEQ in China demonstrated a clear trend of spatial clustering. Due to unfavorable natural conditions, the northwest region exhibited a low level of EEQ clustering. Due to high forest coverage, some sub-regions in the south and northeast exhibited high-level clustering of EEQ. Due to the intensive agricultural activities in the Northeast Plain, North China Plain, and SCB, EEQ showed moderate-level clustering of EEQ.

During 2000–2010, EEQ of 57.0% of counties in China decreased, especially in Tibet, southern Qinghai, Hunan-Guangxi-Guangdong-Jiangxi-Fujian border area, YRD, and HHHP. In 23.6% of the counties, EEQ decreased by more than 0.028. However, there are also some areas where EEQ has shown an improvement trend, such as the Northeast region, northern Hebei, Shanxi, Shaanxi, southern Henan, and some areas in Xinjiang. From 2010 to 2020, the proportion of units with reduced EEQ was 43.4%, and more than 13.9% of the county EEQ decreased by more than 0.028. The area with improved EEQ increased, mainly in eastern Inner Mongolia, the southern Qinghai, the southern part of Gansu, and Yunnan, Guizhou, Guangxi, Guangzhou, and Fujian in southern China. There was a decrease in all other regions (Figure 4). Overall, there were certain differences in the trend of EEQ changes between the two decades of 2000–2010 and 2010–2020. The EEQ scores of some counties in the northwest border, northeast and central China had changed from increase to decrease. However, the EEQ scores of some counties in southern coastal provinces, Xizang and Qinghai had changed from decrease to increase.

The regression analysis for the periods of 2000–2010 and 2010–2020 revealed that the Breusch-Pagan and Koenker-Bassett test statistics were significant at the 0.001 level, suggesting that the data for the independent variables did not exhibit heteroscedasticity. When the spatial effects of adjacent counties were not considered, the OLS regression results exhibited a negative relationship between population change and EEQ change in EEQ in China from 2000 to 2020 (p < 0.001, Tables 1, 2). This suggested that population growth was strongly linked to the decline in EEQ. Comparing the results of the four models, it was found that the SEMLD model had the highest Log likelihood value, the lowest AIC and SC values, and the best fitting effect, making it the most suitable model for this analysis. According to the SEMLD model results, there was a negative association between population change and EEQ change from 2000 to 2010. Specifically, a 1% increase in population was associated with a 0.029% decrease in EEQ. From 2010 to 2020, a 1% increase in population was associated with a 0.069% decline in EEQ (Tables 1, 2).

The regression coefficient obtained by OLS and spatial regression models was the effect of population pressure on EEQ change in the whole study area, but the impact degree may be different from county to county. It is more realistic to assume that the population change and EEQ change between regions had spatial heterogeneity. Therefore, the GWR model was used to generate local variation coefficients. According to the GWR regression results, only 0.8% of counties were severely underestimated during 2000–2010 with the standard deviation of the residual less than −2.5, and only 2.0% of counties were severely overestimated with the standard deviation of the residual greater than 2.5. In contrast, 43.1% of the counties’ standard deviations of the residuals was between −0.5 and 0.5, suggesting that GWR results can more accurately reflect the relationship between EEQ change and population change. During 2010–2020, only 0.8% of counties were extremely undervalued with a standard deviation of the residual below −2.5, and only 1.1% were extremely overvalued with a standard deviation of the residual above 2.5. In contrast, the standard deviations of 42.0% of the county residuals were between −0.5 and 0.5, further indicating that GWR results can well reflect the relationship between EEQ change and population change (Figure 5). GWR results showed that EEQ decreased with population increase during 2000–2010, especially in western Inner Mongolia, southern Gansu, Ningxia, Shanxi, Shaanxi, Henan, Hubei, Anhui, and other regions. The increase in population in the west of Xinjiang and Tibet and the Heilongjiang region of northeast China during 2010–2020 led to an EEQ increase, and a population increase in other regions led to a decrease in EEQ.

Bivariate local Moran’s I was used to identify population pressure zones, and based on the Z-test (p = 0.01), bivariate local spatial autocorrelation Local Indicators of Spatial Association (LISA) aggregation plots were drawn for changes in EEQ and population size in China during 2000–2010 and 2020–2020 (Figure 6). The “low-low” zoning, which refers to areas with decreasing EEQ and continuously decreasing population, is considered an unsustainable development area. During 2000–2010, it was mainly distributed in most areas of Xinjiang, Xizang and Yunnan, as well as the border areas of Sichuan, Qinghai and Gansu. However, during 2010–2020, the region mainly shifted to Xinjiang and Northeast China. Therefore, the region needs to strengthen the dual management of the economy and environment. The “low high” zone refers to the area where EEQ decreases while PD increases, and the increase in PD was the main factor causing the decrease in EEQ. This area belongs to a typical population pressure zone. During 2000–2010, it was mainly distributed in the East China region, including Liaoning, Shandong, Jiangxi, Jiangsu, Anhui, Shanghai, Zhejiang, Guangdong, Fujian, and other areas. Distributed in Hubei, Zhejiang, Hunan, Jiangxi, Shanghai, Anhui, southern Jiangsu, parts of Shandong, Henan and other areas during 2010–2020. Therefore, the region should strengthen the management of human activities and balance the relationship between population and environment. Then there were the “high-high” type zones, which were dispersed in the eastern part of China from 2000 to 2010, including northern Hebei, parts of Liaoning, northern Anhui, and parts of Jiangxi and Zhejiang. From 2010 to 2020, it was dispersed in parts of north China, southern Jiangxi, Fujian, Guangdong, Guangxi, and Hainan. The population and EEQ in these areas increased at the same time, which was defined as the improvement of coordinated development areas. The second was the “high-low” type area. During 2000–2010, the “high-low” type areas were widely distributed, mainly in the western region, including the western part of Xinjiang, Qinghai, Gansu, Shaanxi, Sichuan, Chongqing, Guizhou, and Yunnan, and the western part of Hubei. From 2010 to 2020, there were few high-low type areas, and the discrete distribution was in the western region. These areas were mainly concentrated in mountainous areas, where the population was mainly reduced and EEQ was mainly improved. A decrease of 1% of the population corresponded to an increase of >1% of EEQ, which was a population-EEQ improvement zone.

The overall pattern of population in China has remained relatively stable throughout the study period (2000–2020), characterized by high population density in the southeast and low density in the northwest of China. A combination of factors is responsible for this pattern, including social, economic, natural, and historical influences (Xu et al., 2021). Due to its geographical advantages and favorable national policies, the eastern region of China was the first to undergo reform and opening up, which has significantly accelerated urbanization and industrialization. Consequently, China’s regional economic development has displayed a distinct spatial pattern of “high in the east and low in the west.” Various economic production factors have continuously shifted from the central and western regions to the east, from rural to urban areas, and from less developed to more developed regions. This concentration of resources and factors has driven the spatial shifts in population distribution (Liu et al., 2022; Qi et al., 2022). During the study period, China’s EEQ has increased on the whole, and the proportion of county units with increased EEQ has increased evidently from 2000 to 2010 to 2010–2020. This indicated that EEQ in China has improved significantly in the past two decades. This was consistent with the results of a number of studies (Chen et al., 2024).

The regression results showed that population change exerted great pressure on EEQ. The relationship between population changes and EEQ change in China has a strong spatial spillover effect, which also suggested that future regional EEQ governance and protection and population adjustment should be carried out in the whole region, rather than in a single administrative region. Establishing a cross-administrative management mechanism can improve the efficiency of resource use and management, as well as coordinate the sharing of responsibilities and benefits across the region (Chen W. et al., 2020).

The results of pressure zone identification showed that the “low-low” zone was the most widely distributed, and the number of populations in these zones was decreasing, but EEQ was constantly deteriorating, which was an unsustainable development zone, and the management of EEQ needed to be strengthened. This is followed by the “low-high” type, which was a population-EEQ pressure area. The “high-high” and “high-low” types represented the improved coordinated development area and the population-EEQ improvement area, respectively, but the number of these type areas was small.

The study uncovered that the continuous increase in population and the rising frequency of human activities have placed unprecedented pressure on EEQ. Based on assessing the pressure of population growth on EEQ, the study identified the areas in China experiencing significant ecological stress. These findings lay a scientific foundation for optimizing land use in China and help guide regional population redistribution. Based on these insights, the study offered the following countermeasures and recommendations.

This study used counties as the research units to examine the relationship between changes in EEQ and population in China from 2000 to 2020, employing an EEQ index derived from multi-source remote sensing data to represent EEQ in China. Note that different ecological indicators may affect research results in different ways. Future research could explore the relationship between changes in other ecological indicators and population size. The indicators used in this study’s model were somewhat simplified, and future studies could incorporate factors such as transportation infrastructure, socioeconomic conditions, and natural elements. While such additions might increase the model’s complexity, they could offer clearer insights into the interplay between population and EEQ changes.

This study primarily focused on the effects of population activities on EEQ within China. Although it provides a certain level of representativeness, it may not fully capture the relationship between population dynamics and EEQ on a global scale. Furthermore, insufficient consideration of the unique characteristics and differences among countries and regions could limit the generalizability of the findings. Future research should emphasize the global relationship between population and EEQ, uncovering both universal patterns and distinct differences through comparative studies across various countries and regions. Additionally, in-depth studies of specific areas and particular population groups should be strengthened to enhance the relevance and practical application of the research.