An artificial intelligence-based assessment of soil erosion probability indices and contributing factors in the Abha-Khamis watershed, Saudi Arabia

Soil erosion is a major problem in arid regions, including the Abha-Khamis watershed Saudi Arabia. This research aimed to identify soil erosional probability using various erodibility indices, clay ratio (CR), modified (MCR), Critical Level of Organic Matter (CLOM), and principle component analysis based index (SEI). To achieve these objectives, study used t -tests an artificial neural network (ANN) model best SEI for management. The performance models were then evaluated R 2 , Root Mean Squared Error (RMSE), (MSE), Absolute (MAE), with CLOM identified as predicting erodibility. Additionally, Shapley additive explanations (SHAP) values influential parameters erosion, sand, clay, silt, organic carbon (SOC), moisture, void ratio. information can help develop management strategies oriented parameters, which will prevent erosion. showed notable distinctions between CR CLOM, where 25–27% contribution explained over 89% overall diversity. MCR indicated that 70% area had low erodibility, while 20% moderate 10% high range from 40% showing moderate, high. Based on T -test results, significantly different MCR, principal (PCA), PCA, PCA. ANN implementation demonstrated highest accuracy ( 0.95 training 0.92 testing) SOC, being most important variables. SHAP confirmed importance variables each four models. provides valuable regions. identification effective promote agricultural production. be by policymakers stakeholders make informed decisions manage