A MACHINE LEARNING-BASED SYSTEM AND METHOD FOR REAL-TIME LANDSLIDE PREDICTION AND EARLY WARNING

Abstract

An integrated landslide early warning system combines a machine learning model with a robust monitoring network and a centralised data analytics platform. The platform integrates real-time data from a network of automatic rain gauge field stations using redundant transmission and deduplication. An XGBoost model utilises eight key rainfall and slope features to generate slope-specific landslide probabilities. The system processes the rainfall measurements and forecast data through dedicated analytical and visualisation modules to compute dynamic territory-wide numbers of reported landslides and produce a dynamic, territory-wide landslide risk map. This enables the evaluation of landslide risk against a threshold, supporting the timely issuance of public warnings and activation of emergency response protocol.

Inventors

Kei Hong CHU
Dong Sheng CHANG
Chak Lam CHAN
Kin Ho CHAN
Ling Ho KIU

Assignees

The Government of the Hong Kong Special Administrative Region (Civil Engineering And Development Department)

Dates

Publication Date: 20260508
Application Date: 20260114