关键词: 改进型遥感生态指数, 宁夏回族自治区, 土壤侵蚀模型, 生态环境质量, 驱动因子

Abstract: Objective　The ecological environment is the cornerstone of human survival and socioeconomic development, and its quality level directly reflects the coordination degree between human activities and natural ecosystems. With the rapid economic growth in China, the pressure on regional ecosystems is increasing, particularly in areas with fragile ecological environments. The Ningxia Hui Autonomous Region is characterized by arid climate, scarce precipitation, and high sediment content in surface soil, all of which make the region highly susceptible to soil erosion. As the only provincial-level administrative region entirely located within the Yellow River Basin, Ningxia is an important ecological zone in the upper and middle reaches of the Yellow River, making it strategically important for the ecological security and sustainable development of the region. Despite extensive ecological assessments have been conducted on Ningxia, they are mostly limited to local areas or rely on individual indicators, resulting in insufficiently comprehensive evaluations that account for the region's distinctive ecological characteristics. The traditional widely used Remote Sensing Ecological Index (RSEI), which integrates greenness (NDVI), wetness (WET), dryness (NDBSI), and heat (LST), provides an effective framework for assessing ecological quality but cannot directly account for soil erosion intensity, a major ecological constraint factor in Ningxia. Therefore, an improved RSEI (IRSEI) is developed in this study by incorporating the Revised Universal Soil Loss Equation (RUSLE) into the traditional RSEI framework to better capture the severe soil erosion conditions in Ningxia. The study aims to quantitatively assess the spatiotemporal variation in ecological environmental quality in Ningxia from 2001 to 2022 and identify the principal factors associated with these changes using the Geodetector model. The findings are expected to provide data support and a scientific basis for regional ecological restoration and sustainable environmental management in the region. Methods　Based on the Google Earth Engine (GEE) cloud-computing platform, the traditional RSEI was enhanced by incorporating RUSLE-derived soil erosion indicators, resulting in a five-component IRSEI framework comprising greenness, wetness, heat, dryness, and soil erosion intensity. Principal Component Analysis (PCA) was then applied to integrate these indicators and reduce dimensionality. Before using PCA for transformation, each indicator was normalized to a scale of 0–1 to eliminate differences in measurement units and ensure weight rationality. PCA effectively condensed key information from multiple variables, with the first principal component (PC1) representing the dominant ecological information. The resulting IRSEI values were classified into five grades (poor, relatively poor, moderate, good, and excellent) for spatial and temporal comparison. Four representative years (2001, 2008, 2015, and 2022) were selected to characterize long-term evolutionary trends in ecological environmental quality, and the differences between adjacent periods were analyzed to identify the areas of ecological improvement and degradation. The Geodetector method was applied to quantify the explanatory power (qstatistic) of driving factors, including precipitation, temperature, elevation, slope, population density, land use, and nighttime light. The factor detector and interaction detector modules were used to evaluate the individual and coupled effects of these factors, with significance test indicator p indicating whether the selected factor has a significant impact on the distribution of ecological environment quality. Results and discussion　PCA results showed that the contribution of PC1 in 2001, 2008, 2015, and 2022 reached 70.75%, 74.24%, 73.81%, and 75.70%, respectively, indicating that it can sufficiently represent the ecological environment quality reflected by various indicators. Among the five indicators, NDVI and WET contributed positively to PC1, indicating their positive impact on ecological environment quality, whereas LST, NDBSI, and RUSLE contributed negatively to PC1, reflecting their adverse effects. Spatially, high-quality ecological areas were concentrated in the urban areas of Yinchuan and Shizuishan, high-altitude forest areas in southern Ningxia, and areas along the Yellow River, while relatively poor ecological regions were mainly distributed in the mid-temperate semi-arid areas. Temporally, the overall ecological quality of Ningxia remained relatively poor but showed an improving trend. The mean IRSEI increased from 0.267 in 2001 to 0.336 in 2022, representing a 25.8% increase. The proportion of areas with poor or relatively poor ecological environment decreased from 80.65% in 2001 to 71.44% in 2008 and further declined in 2022, while the proportion of areas with moderate, good, and excellent ecological environment grades increased from 19% in 2001 to 31% in 2022. From 2001 to 2022, ecological improvement areas accounted for 45.86%, far exceeding degraded areas (18.78%). Significant improvement occurred in the southern mountainous areas and suburban zones, while slight degradation was observed in northern urban built-up areas and along the Yellow River economic belt, potentially associated with the extreme drought conditions in 2022. Geodetector results showed that all examined factors were statistically significant (p < 0.05). Land use was the dominant driving factor, with a mean q-value of 0.456, followed by elevation (0.340), whereas slope had the weakest explanatory power (0.026). The explanatory power of precipitation increased during the research period, while that of population density decreased. Interactive detection showed that two-factor combinations exhibited nonlinear or bivariate enhancement, especially the interaction between land use and precipitation, indicating that their synergistic effects provided stronger explanations of ecological differentiation. Among the examined factors, Land use, representing human activities, had the highest mean qvalue (0.456), indicating that it is the dominant factor associated with spatial variations in ecological quality. Conclusions　This study developed an improved RSEI by integrating the RSEI and RUSLE models, thereby improving the applicability of remote-sensing-based ecological assessment in regions affected by severe soil erosion. From 2001 to 2022, the ecological environmental quality of Ningxia remained generally low but showed an overall continuous and significant improving trend, with the ecological improvement area substantially exceeding the degradation areas, indicating an overall improvement in regional ecological conditions. Both natural and anthropogenic factors were associated with variations in ecological quality, and their interactions enhanced the explanatory power of individual factors. Among the driving factors studied, land use had the highest explanatory power, highlighting the importance of scientific land management and ecological restoration policies in improving regional ecological environment. These findings offer a scientific basis for ecological planning, soil erosion control, and sustainable development in Ningxia and other ecologically fragile regions in the Yellow River Basin.

Key words: improved remote sensing ecological index, Ningxia Hui Autonomous Region, soil erosion model, ecological and environmental quality, driving factors