A hybrid data-driven approach for rainfall-induced landslide susceptibility mapping: Physically-based probabilistic model with convolutional neural network
Citation
Cui, H. Z., Tong, B., Wang, T., Dou, J., Ji, J. (2025). A hybrid data-driven approach for rainfall-induced landslide susceptibility mapping: Physically-based probabilistic model with convolutional neural network. Journal of Rock Mechanics and Geotechnical Engineering. Link to paper
Abstract
Rainfall-induced landslides are among the most frequent and destructive natural hazards, causing significant economic losses and casualties worldwide. Accurate susceptibility mapping is crucial for effective landslide risk management and disaster prevention. This study proposes a hybrid model that combines the physically-based probabilistic model (PPM) with convolutional neural network (CNN), where the PPM captures the spatial distribution of landslides by incorporating the probability of failure considering the slope stability mechanism under rainfall conditions. CNN was used as a binary classifier to capture the spatial and channel correlation between landslide conditioning factors and the probability of landslide occurrence. The proposed PPM-CNN hybrid model presents a higher prediction accuracy, with an area under the curve (AUC) value of 0.85 based on the landslide case of the Niangniangba area of Gansu Province, China. The results demonstrate that the hybrid approach effectively integrates the strengths of both physically-based and data-driven methods, providing more reliable landslide susceptibility assessments for rainfall-induced landslides.