Integrating Machine Learning and StoDyM for Check Dam Site Selection on the Loess Plateau, China

Severe soil erosion in the Loess Plateau has led to significant land degradation, threatening agricultural productivity and ecosystem stability. Check dams, as a critical soil–water conservation measure, have proven effective in mitigating sediment loss and reclaiming farmland. However, their long‐term sustainability depends on precise site selection. Traditional regional‐scale approaches, such as multicriteria decision analysis (MCDA), often rely on subjective weight allocation and static environmental assessments, limiting their ability to address dynamic hydrological response processes of check dam systems at the watershed scale. To overcome these limitations, this study developed a framework combining machine learning (ML) and the storage dynamics model (StoDyM) for check dam placement achieve dynamic optimization of site selection and sustainability assessment. Five ML algorithms—Support Vector Machine (SVM), Random Forest (RF), AdaBoost, XGBoost, and the ensemble algorithm Stacking to map suitability zones, validated by ROC–AUC metrics. After all suitable dam sites were identified, the StoDyM model was then used to screen and simulate all check dams with suitable lifespans. The results showed that: (1) the RF and XGBoost outperformed other models (ROC–AUC = 0.96). Revealing four dominant controlling factors: gully proximity, soil erosion intensity, slope gradient, and topographic wetness index. (2) The newly selected check dams demonstrated superior lifespan performance, with no instances of excessively short or long service lives in the simulations. Indicating five suitable dam sites in Fangta, 19 in Majiagou, and four in Zhifanggou. (3) The new check dam system (newly selected check dams + current dams) demonstrates a simulated 35% reduction in sediment yield in Fangta, 59% in Majiagou and 63% in Zhifanggou. This framework advances precision in check dam planning, offering a scalable, data‐driven strategy for sustainable watershed management in erosion‐prone regions globally.