Towards near real-time forecasting of rainfall-induced landslides

A study focused on integrating trigger information with underlying static susceptibility for developing a statistical model capable of predicting rainfall-induced landslides. Northwestern Vietnam is used as a test-site for utilizing the built model for possible application in a Landslide Early Warning System (LEWS). A Generalized Additive Model (GAM) is built with rainfall as a dynamic trigger and the model is tested in external sites as well as a spatial, temporal and random spatio-temporal domain. Moreover, a suitable probabilistic cut-off is selected to visualize the possible LEWS on Google Earth Engine (GEE) Apps by allowing the user to select a date for extracting dynamic information as available on GEE Catalog. Conclusively, a framework is proposed that accounts for the orographic effect in the context of dynamic landslide susceptibility and allows for uncertainty estimation within a unified model.