Vegetation dynamics and dominant driving factors in a semi-arid coal mining area under a disturbance–restoration zoning framework

Understanding vegetation dynamics in arid mining ecosystems remains limited, despite the extensive disturbances imposed by mining activities. Integrating remote sensing and field observations, we applied the Mann–Kendall test, Theil–Sen slope, the BFAST01 algorithm, Boosted Regression Trees, and Partial Least Squares Structural Equation Modeling to quantify vegetation trajectories and their drivers across four disturbance–restoration zones in the Shendong mining area. Between 2000 and 2022, 88.4% of the region experienced significant vegetation improvement (p < 0.01), accompanied by pronounced spatial heterogeneity in transition patterns. Zone 1 exhibited the fewest interruption-related declines, zone 2 showed the highest frequency of transitions from increase to decrease, zone 3 displayed the lowest transition frequency, and zone 4 experienced more frequent interruptions. Non-climatic factors accounted for 86.6% of vegetation improvement, with dominant drivers varying among zones. These findings underpin a zoning–grading–customized restoration framework that provides targeted guidance for ecological rehabilitation in arid mining landscapes.