Spatiotemporal evolution and driving mechanisms of ecological quality in shanxi province based on XGBoost-SHAP

Accurately assessing the spatiotemporal evolution of regional ecological quality is critical for advancing China's ecological civilization and achieving its carbon peaking and carbon neutrality targets. As a typical resource-based region, Shanxi Province faces significant ecological fragility and spatial heterogeneity resulting from the interplay of an arid/semi-arid climate and intensive coal mining activities. This study utilized the Remote Sensing Ecological Index (RSEI) from 2000 to 2025 as a core indicator. The Theil-Sen slope estimator and Mann-Kendall test were employed to identify trends, while the Hurst exponent was used to assess the persistence of changes. Furthermore, the XGBoost-SHAP framework was applied to disentangle the driving factors. The results indicate that: (1) From 2000 to 2025, the ecological quality of Shanxi exhibited a fluctuating upward trend, with the mean RSEI increasing by approximately 8.7%. (2) Trend analysis indicated that the proportions of areas exhibiting upward and downward trends in RSEI were 42.3% and 35.64%, respectively, while 22.05% of the region remained stable. Significant improvement areas were mainly concentrated in the Lüliang and Taihang Mountains. (3) Future trend predictions reveal that stable areas comprise the majority of the province (55.89%). However, the areas of persistent improvement and persistent degradation account for 9.54% and 10.88%, respectively, indicating a characteristic of "polarization" in ecological environmental quality. (4) Regarding driving mechanisms, anthropogenic factors exerted a stronger influence than natural factors. Precipitation was the primary positive natural driver, whereas coal mining intensity and population density were key negative stressors. Crucially, coal mining intensity exhibited a non-linear interaction with precipitation, significantly weakening the ecosystem's positive response to rainfall. This study provides a scientific basis for targeted ecological restoration in resource-based regions.