Contrasting vegetation–topography controls on soil erosion in karst and non-karst areas: insights from the modified RUSLE model

Soil erosion is a major driver of ecological degradation in karst landscapes, yet effective management remains hindered by limited model applicability and unclear driving mechanisms. This study applied a modified RUSLE model to assess soil erosion dynamics from 1990 to 2020 in a typical peak-cluster depression basin, with model performance evaluated at basin, sub-regional, and plot scales. The model was modified by calibrating the rainfall erosivity (R) factor using a 16.5 mm erosive daily rainfall threshold and incorporating surface karstification rates into the slope length (L) factor as a power function coefficient. Results showed a significant decline in soil erosion after 1998, which was attributed to ecological restoration associated with the Natural Forest Protection Program (initiated in 1998) and the Grain for Green Program (fully implemented around 2000). The mean annual erosion rate decreased by 64%, from 4.95 t ha -1 yr -1 in 1990–1997 to 1.77 t ha -1 yr -1 in 1998–2020. Erosion rates remained much higher in non-karst than in karst areas in both periods, decreasing from 8.45 to 3.06 t ha -1 yr -1 and from 1.90 to 0.64 t ha -1 yr -1 , respectively. Factor analyses showed that vegetation coverage was the dominant driver in karst areas ( q = 0.28), whereas slope gradient dominated in non-karst areas ( q = 0.23). Notably, karst erosion was particularly sensitive to vegetation change, decreasing sharply once vegetation coverage exceeded 10%. These findings support spatially targeted soil conservation in karst basins.