Assessing framework of rainfall-induced landslide hazard considering spatiotemporal asymmetry in extreme precipitation indices under climate change

Landslides triggered by extreme rainfall events often cause losses of life, property damage, and environmental alterations. While past studies have assessed landslide hazards using various indices, how to select rainfall indices in rainfall-induced landslide hazard assessment is still a challenge due to the spatiotemporal asymmetry of rainfall indices. In this study, we employed three machine-learning models, namely the Random forest (RF), Support vector machine (SVM), and logistics regression models, and developed an extreme rainfall index-based model to evaluate rainfall-induced landslide hazards. To eliminate the effect of spatiotemporal asymmetry in indices, we selected six extreme rainfall indices that are highly correlated with rainfall-induced landslides and tested 63 combinations. Over the past four decades, extreme rainfall events have become more frequent and intense. Both the number and type of rainfall indices affected the assessment results of landslides in the study area. The RF model showed a better accuracy in landslide hazard assessments than did the other two models. To better predict rainfall-induced landslide hazards, an optimal model based on three extreme rainfall indices, i.e., PSSPTOT, R25mm, and Rx5day, was proposed for the study area. With climate change, the study area may encounter more intense rainfall events and experience high levels of rainfall-induced landslide hazards. Compared to the baseline, landslide hazards in the study area are projected to increase by 9.9% and 11.9% in the 2030s (2021–2050). Areas with high- and very high- levels of landslide hazards will account for more than 50% of the study area and will be mainly distributed in the central and eastern parts of the study area. This study suggested an optimal combination of extreme precipitation indicies and provided scientific information about rainfall-induced landslide hazard management under climate change.