Soil erosion estimation has attracted considerable attention from the scientific community and governments because of its importance to sustainable regional development. In karst areas, the heterogeneous environment and rocky desertification create difficulties in determining the influencing factors and spatial patterns of soil erosion. A quantitative analysis of karst soil erosion distribution was conducted by applying the revised soil loss equation model and the geographical detector method of attribution identification, which was based on spatial variance analysis. The results show that soil erosion was most severe in areas with an elevation of 1200-1800 m and intense anthropogenic activity. When the vegetation coverage was below 0.5-0.6, soil erosion showed characteristics of a source-limited regime and increased with the increasing vegetation coverage. When the vegetation coverage was higher than 0.5-0.6, soil erosion followed a transport-limited regime and decreased with the increasing vegetation coverage. The factor detector showed land use to be the dominant factor, explaining 51% of soil erosion distribution. Among various land use types, dry land had the greatest vulnerability to soil erosion. Slope served as a controlling factor at large scales, especially when combined with annual precipitation exceeding 1500 mm, and in dry and grassland areas. From the attribution analysis of multiple factors, the combination of land use and slope was the controlling interaction factor explaining 68% of soil erosion distribution. The methods and results of this research could serve as scientific references for decision makers and researchers exploring the characteristics of soil erosion to develop effective measures for its control.
Keywords: Geographical detectors; RUSLE model; Soil erosion determinants; Spatial consistence; Spatial distribution.