A Modelling Tool for Rainfall-triggered Landslide Susceptibility Mapping and Hazard Warning based on GIS and Machine Learning

Abstract

Abstract The landslide susceptibility mapping and hazard warning are widely adopted tools by the government, stakeholders and the public for landslide disaster preparedness and emergency planning. This study presented a modelling tool based on geographic information system (GIS) and machine learning to aid the two-step modelling procedure. The machine learning methods including artificial neural networks, support vector machines, and logistic regression were integrated into the GIS environment for modelling landslide susceptibility to simplify and automate the routines of model training, verification and prediction. Then, the meta-element model was employed to take the landslide susceptibility, antecedent effective rainfall and 24-hour forecasted rainfall as inputs to determine the landslide hazard level. The architecture to deploy the established meta-element model for real-time landslide hazard warning was also proposed. A study case in Chunan, China was selected to demonstrate the applicability of the modelling tool to aid landslide susceptibility mapping and real-time hazard warning in response to a typhoon event. The developed modelling tool was desired to evolve into cloud computing architecture to facilitate easy-to-reuse and uplift its scalability.