Landslide Susceptibility Mapping and Driving Mechanisms in a Vulnerable Region Based on Multiple Machine Learning Models

Landslides can cause severe damage to both the environment and society, many statistical, index-based, inventory-based methods have been developed assess landslide susceptibility; however, it is still challenging choose most effective method properly identify major driving factors for specific regions. Here, we applied four machine learning algorithms, adaptive boosting (AdaBoost), gradient-boosting decision tree (GBDT), multilayer perceptron (MLP), random forest (RF), predict susceptibility at 30 m spatial scale based on thirteen conditioning (LCFs) in a landslide-vulnerable region. Based inventory points, classification results were evaluated, indicated that performance of RF (F1-score: 0.85, AUC: 0.92), AdaBoost 0.83, 0.91), GBDT 0.88) significantly better than MLP 0.76, 0.79) method. The further areas with high very risk (susceptibility greater 0.5) accounted about 40% study All models matched well predicted similar distribution patterns susceptibility, mostly distributed western southeastern Daoshi, Qingliangfeng, Jinnan, Linglong towns highest risk, mean levels 0.5. leading contributing slightly different models; population density, distance road, relief amplitude generally among top towns. Our provided significant information highly landslide-prone decision-makers policy planners, suggested should take unique precautions mitigate or avoid from events.