Background Soil erosion is one of the environmental problems worldwide. Chongqing is located in the southwest purple soil area, with severe erosion. Exploring the possibility of using GeoDetector to determine the leading factors of soil erosion spatial differentiation in county-level administrative regions, and mastering the leading factors affecting the spatial heterogeneity of soil erosion in each district and county is the basis for effective soil loss control.

Methods In this study, China Soil Loss Equation (CSLE) model was used to calculate soil erosion intensity in Hechuan district and Yunyang county of Chongqing, and GeoDetector was used to explore the dominant factors and interaction factors affecting the spatial heterogeneity of soil erosion. The CSLE model is based on the USLE and RUSLE. It is widely used because of its simple structure and easy access to parameters. GeoDetector is a statistical model based on spatial autocorrelation, which detects and analyzes spatial heterogeneity. It has four functions of factor detection, interaction detection, risk detection and ecological detection. This article used factor detection and interaction detection for analysis. Factor detection was used to reveal the influence of each factor on dependent variables. Interaction detection was used to judge the relationship between two different factors acting together on the dependent variable.

Results 1) The slope and biological measure factor were the leading ones for the spatial variability of soil erosion in Hechuan district and Yunyang county, respectively. The affecting factors of the spatial differentiation of soil erosion in Hechuan district were ranked in descending order of explanatory power, namely, slope factor S, slope length factor L, tillage measure factor T, biological measure factor B, engineering measure factor E, soil erodibility factor K, and rainfall erosivity R. The affecting factors of Yunyang county were ranked in descending order of explanatory power, namely, the factor B, L, T, S, E, K, and R factor. 2) The interactive detection results showed that the effect of any two factors on the spatial difference of soil erosion was enhanced. The largest interaction q values were terrain factors interacting with other factors in the two counties. The top three interaction types in Hechuan district were L∩B (60.1%), S∩B (56.5%), and L∩S (54.1%). The top three interaction types in Yunyang county were L∩B (78.5%), L∩E (59.0%), and S∩B (56.9%).

Conclusions These results suggest that GeoDetector can be used to identify key factors in the spatial differentiation of soil erosion in county-level administrative regions. Due to the different geographical environments in different regions, the dominant factors affecting the spatial differentiation of soil erosion were regionally different. The determination of the dominant factors can provide targeted opinions for different districts and counties to control soil erosion.