Search for a command to run...
Landslides frequently result in human casualties and economic losses in mountainous regions, being particularly severely affected in the Himalayan areas. Mitigating hazards and risks serves as an effective means to the use of landslide susceptibility mapping (LSM). Previous studies have focused on the simple application of binary methods or the comparison of deep learning methods. This study is dedicated to exploring a more advanced model by integrating convolutional neural networks (CNN) with bivariate methods for LSM, aiming to combine the interpretability of bivariate statistics with the high predictive accuracy of deep learning. Four models were developed and compared: CNN, certain factors (CF) hybridized with CNN, frequency ratio (FR) hybridized with CNN and information value method (IV) hybridized with CNN. Initially, a total of 313 landslide were identified and systematically incorporated into a landslide inventory map, with 12 predisposing factors concurrently selected for subsequent analysis. Subsequently, the dataset was randomly partitioned into two subsets, wherein 75% was allocated for model training and 25% reserved for validation purposes. Finally, the performance of the models was validated and compared using area under the curve (AUC) and statistical metrics. The results showed that the IVCNN model demonstrated superior performance (AUC 0.974 and accuracy = 93.3%), compared with CNN model (AUC 0.91 and accuracy = 86.9%), CFCNN (accuracy = 88.6% and AUC = 0.959), FRCNN (accuracy = 88.6% and AUC = 0.959). In addition, the hybrid model retains the interpretability of the IV, clearly identifying maximum elevation difference exceeding 1,200 m as the top dominant factors controlling landslide occurrence in the study area. The integration of bivariate statistical methods and CNN can effectively enhance the accuracy and interpretability of LSM. The proposed IV-CNN framework provides a reliable technical tool for landslide risk assessment and management in high-altitude mountainous regions, offering valuable insights for decision-makers to formulate targeted disaster mitigation strategies.