Spatiotemporal Heterogeneity and Driving Mechanisms of Ecological Quality Based on Modified Remote Sensing Ecological Index and XGBoost–SHAP Analysis

Abstract

Ecological environment change is a critical issue in global environmental protection research. Understanding the spatiotemporal dynamics and drivers of regional ecological environment quality (EEQ) is essential to support sustainable ecosystem management. To evaluate the spatiotemporal dynamics of EEQ in the Qilian Mountain National Nature Reserve (QMNNR) from 1986 to 2023, this study constructed a modified remote sensing ecological index (MRSEI) using Google Earth Engine (GEE) and incorporated the patch‐generating land use simulation (PLUS) model. A coupled explainable machine learning model (XGBoost–SHAP), along with a multivariate regression residual approach, was used to quantify the contributions of climate variability and anthropogenic activities to EEQ dynamics. This study presents the following key findings: (1) the MRSEI effectively integrates information from multiple variables, enhancing model robustness for long‐term ecological monitoring; (2) from 1986 to 2023, EEQ in the reserve underwent overall improvement. A Moran's I index of 0.83 indicated significant spatial clustering along both latitudinal and longitudinal gradients; (3) under the natural development scenario, the PLUS model predicts that by 2035, the proportion of EEQ area classified as improved areas (20.46%) will be lower than that of degraded areas (21.60%); (4) climate change contributes only slightly more to EEQ variations in the reserve (50.11%) compared to anthropogenic activity (49.89%). The primary factors influencing EEQ are land use, followed by precipitation, temperature, population density, night lights, and geographic coordinates (longitude and latitude). This study provides novel insights into regional EEQ monitoring, driving factor analysis, and ecological environment protection strategies.