Key points are not available for this paper at this time.
Abstract Landslides pose a severe threat to the safety of mountainous regions, and existing landslide susceptibility assessment methods often suffer from limitations in data quality and methodology. This study focused on Wushan County, China, combining machine learning algorithms with InSAR data to improve the accuracy of landslide susceptibility mapping. Employing seven machine learning models, the investigation identified CNN, LR, and RF as the most effective, with AUC values of 0.82, demonstrating their ability to predict landslide-prone areas. Key influencing factors for landslides included digital elevation model (DEM), rainfall, lithology, normalized difference vegetation index (NDVI), terrain curvature, roughness, and distances to roads and rivers. Integrating InSAR data significantly enhanced the accuracy of landslide susceptibility mapping, particularly in areas with high deformation, refining assessments and reducing misclassifications. Slope analysis and InSAR monitoring provided insights into instability mechanisms, highlighting InSAR's potential for early warning systems. The study concludes that combining InSAR with machine learning holds promise for improving landslide susceptibility mapping. Future research should explore advanced machine learning techniques and other remote sensing data to address the impacts of climate change and seasonal variations on slope stability, ultimately supporting disaster risk management and sustainable land-use planning.
Cui et al. (Sun,) studied this question.