Machine Learning-Powered Landslide Forecasting: From Initiation to Mobility


Te Xiao
The Hong Kong University of Science and Technology, Hong Kong
Li-Min Zhang
The Hong Kong University of Science and Technology, Hong Kong


Prompt prediction of landslide occurrence and movement in a future rainstorm is one of the most effective manners to cope with the increasing landslide risk in a changing climate. Despite the rapid development of many machine learning algorithms, most studies stay on landslide susceptibility mapping because of the challenging time-unknown and terrain-unmatched issues in landslide forecasting. This study proposes two novel machine learning strategies to predict the spatio-temporal distribution of landslides considering both initiation and mobility. Hong Kong is taken as an example to demonstrate the capacity of city-scale landslide forecasting using machine learning. The spatio-temporal evolution of both man-made slope failures and natural terrain landslides in a rainstorm can be well predicted using machine learning models, which can provide a powerful real-time decision-making tool for landslide early warning and risk management.

December 30, 2023
Online ISSN