Hybrid method for rainfall-induced regional landslide susceptibility mapping

Landslide susceptibility maps can provide important information for managing regional landslide risks. Traditionally, data-driven and physically-based models are widely used for rainfall-induced landslide susceptibility mapping, but each method has limitations. In this study, a hybrid method that integrates a data-driven model and a physically-based model is proposed for rainfall-induced landslide susceptibility mapping, where the uncertainty in the soil properties can be explicitly considered. The proposed method is illustrated with landslide susceptibility mapping in Shengzhou County, Zhejiang Province, China. Logistic regression is used as the data-driven model, and the regional assessment of rainfall-induced landslides model (RARIL) is used as the physically-based model. Three hybrid models are developed. Hybrid model I, which considers soil parameters uncertainty, is compared with hybrid models II and III, which do not consider it. Results indicate that all the three hybrid models outperform the conventional logistic regression and RARIL models. Notably, hybrid model I, which considers the soil parameters uncertainty, outperforms hybrid models II and III, which do not consider it.