Abstract: The surface ecology of coal mining areas is influenced by the spatiotemporal interplay of various factors such as topography, land type, vegetation, meteorology, and mining disturbances. Addressing the issue that existing research on vegetation Net Primary Productivity (NPP) often focuses on macro scales while lacking detailed analysis, this study takes the Shendong Mining Area and Shangwan Mine as research subjects. It establishes a multi-level evaluation framework of "mining area-coal mine-working face", incorporating ecological background characteristics such as high-density mining in arid and semi-arid regions and aeolian landforms. By employing an optimized CASA model and the probability integral method, and integrating 30-meter resolution remote sensing data with surface deformation parameters, the study systematically reveals the spatiotemporal evolution of vegetation NPP from 2000 to 2023 and its multi-scale driving mechanisms. The research findings indicate: NPP shows an increasing trend across multiple scales (2.8 % in the mining area, 2.9 % in the coal mine, and 3.2 % at the working face), though the working face scale exhibits periodic declines due to localized mining disturbances. At the mining area and coal mine scales, NPP is primarily driven by rainfall and NDVI (with r values up to 0.98), while at the working face scale, NPP shows a significant negative correlation with horizontal surface deformation (r=-0.66), elucidating the micro-mechanisms of direct physical damage from mining. A "zonal management" governance strategy is proposed: at the mining area scale, climate-adaptive vegetation restoration should be strengthened, while at the working face scale, priority should be given to controlling horizontal deformation, with the establishment of a three-year post-mining ecological restoration window.