Spatiotemporal Variations and Influencing Factors of Soil Erosion in the Qingyi River Basin (Southwest China) Based on the CSLE Model: Implications for Sustainable Watershed Management

Soil erosion is a major environmental issue that threatens watershed ecological security and the sustainable use of land resources. Quantifying its spatiotemporal variability and associated environmental controls is important for sustainable land use planning and watershed management. To characterize the spatiotemporal variation in CSLE-simulated soil erosion and the relative explanatory contributions of environmental variables in the Qingyi River Basin, this study integrated rainfall, soil type, digital elevation model, land use, and vegetation coverage data for six observation years from 2000 to 2025 with GIS spatial analysis and the Chinese Soil Loss Equation (CSLE). Geodetector and CatBoost–SHAP were further applied to evaluate the explanatory contributions and interaction patterns of the selected environmental variables on the simulated erosion results. The results showed the following: (1) Woodland and cropland dominated the land use structure of the basin, while construction land increased from 90.95 km2 to 164.21 km2. Land use patterns differed markedly between the upstream and downstream areas, with woodland and grassland dominating the upstream area and cropland and construction land accounting for higher proportions in the downstream area. (2) Across the six observation years, the mean soil erosion modulus ranged from 135.74 to 378.39 t·km−2·a−1, indicating generally low erosion levels, with the highest value occurring in 2015 and the lowest in 2025. (3) Soil erosion intensity was mainly characterized by slight and mild erosion, which together accounted for more than 95% of the basin area, whereas areas of moderate erosion and above were mainly concentrated in downstream mountainous areas and along both sides of river valleys. (4) The explanatory analysis showed that elevation, land use, and vegetation coverage made relatively high contributions to the spatial variability in the CSLE-simulated erosion results. Topographic and vegetation-related variables showed higher explanatory contributions in the upstream area, whereas land use showed a higher contribution in the downstream area. These findings provide a quantitative basis for soil erosion monitoring, the identification of priority areas for soil and water conservation, sustainable land use optimization, and region-specific watershed management in the Qingyi River Basin.