Study on Soil Erosion Driving Forces by Using (R)USLE Framework and Machine Learning: A Case Study in Southwest China

Soil erosion often leads to land degradation, agricultural production reduction, and environmental deterioration, which seriously restricts the sustainable development of regions. Clarifying driving factors soil is premise preventing erosion. Given lack current research on factors/force changes in different regions or under intensity grades, this paper pioneered use machine learning methods address problem. Firstly, widely used (Revised) Universal Loss Equation ((R)USLE) framework was applied simulate spatial distribution Then, K-fold algorithm evaluate accuracy stability five algorithms for fitting The random forest (RF) method performed best, with average reaching 86.35%. Permutation Importance (PI) Partial Dependence Plot (PDP) based RF were introduced quantitatively analyze main geological conditions force each factor respectively. Results showed that drivers Chongqing Guizhou cover management (PI: 0.4672, 0.4788), while Sichuan slope length 0.6165). Under shows nonlinear complex inhibitory promoting effects value changing. These findings can provide scientific guidance refined erosion, significant halting reversing degradation achieving resources.