<p>An early identification of catchments with potential for debris flow (debris flow catchments) is a fundamental step in the investigation and early warning of debris flow disasters. However, when identifying potential debris flow catchments at regional scales, certain methodological challenges exist, such as the subjective determination of drainage area thresholds during catchment delineation, the lack of systematic criteria for selecting negative samples (non-debris-flow catchments), and low model accuracy. To address these issues, we employ the river network density method to achieve a more reasonable delineation of catchment units compared to traditional methods. Additionally, a Self-Organizing Map (SOM) is employed to improve the identification of non-debris-flow catchments. Factors closely associated with debris flow formation related to topography, material sources, and climate are selected as input features. By leveraging the stacking ensemble method, which includes the Random Forest (RF), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) algorithms, a model for identifying potential debris flow catchments is produced. To evaluate the performance of the model, the Wenchuan earthquake area in the northwestern Sichuan Basin was used, where potential debris flow catchments were extracted from the ensemble and compared with the results of individual logistic regression (logit), RF, KNN, SVM models. The comparison of Receiver Operating Characteristic (ROC) curves, Area Under the Curve (AUC), accuracy, precision, recall, and F1 score reveals that the stacking ensemble model achieves values of 0.967, 0.918, 0.918, 0.895, and 0.906, respectively, all of which are higher than those for the individual models. The identification of potential debris flow catchments using this ensemble model can help to screen areas for the deployment of debris flow forecasting and early warning, thereby supporting disaster prevention and mitigation efforts.</p>