Abstract: Taking Shanxi Province as the research area, based on multi-source remote sensing and climate data from 2001 to 2023, the RUSLE model was constructed, and the main controlling factors of soil conservation changes from both spatial and temporal dimensions were systematically identified by integrating Theil-Sen Median trend analysis method, random forest model, and pixel-by-pixel regression method. The results showed that the total soil conservation in Shanxi Province showed a significant upward trend from 2001 to 2023, with an average annual increase of 0.149 t/hm². The high-value areas of changes in potential soil erosion and conservation were concentrated in regions with large topographic relief and significant ecological engineering intervention, showing an obvious geomorphology-climate coupling effect. The results of driving factor identification indicated that soil moisture, wind speed, and potential evapotranspiration were the key control variables in the spatial dimension, and the maximum contribution to the random forest model was 20.0%, 16.3% and 15.5%, respectively. In the temporal dimension, the drought index was the dominant factor, reflecting the long-term regulatory effect of water deficit on the interannual changes of soil conservation.

Key words: soil conservation, climate change, RUSLE model, spatio-temporal response mechanism, geomorphic sensitive areas, Shanxi Province