AI–DRIVEN Soil Moisture Extended Cohesive Damage Element Method for Forecasting Landslide Initiation and Propagation under Heavy Rainfall

Landslides cause significant human and financial losses worldwide. They are often triggered by heavy rainfall, which reduces soil cohesion along weak interfaces in sloped terrain. Accurate forecasting remains a global challenge, and the frequency of landslide events is rising. There is an urgent need to establish global digital networks capable of forecasting rainfall-induced landslides and identifying areas at risk. This PhD project aims to develop an AI–Soil Moisture Extended Cohesive Damage Element (AI-SMECDE) method to improve the understanding and forecasting of landslide initiation and propagation under heavy rainfall. The SMECDE initially developed by the primary supervisor will be further enhanced through the integration of an artificial intelligence (AI) algorithm, forming an AI-determined geotechnical slope model. This method will capture both lateral and vertical variations, as well as the duration effects of intense rainfall or elevated soil moisture, within a predictive framework. A correlation between soil moisture and forecasted rainfall intensity will be established to support accurate predictions. The SMECDE will estimate rainfall thresholds that trigger failures, based on soil moisture content or rainfall intensity–duration effects, and will simulate shear crack initiation and propagation in sloped terrain. An AI algorithm will be developed and trained using historical landslide data to validate and refine SMECDE predictions. This algorithm will optimise input parameters, including meteorological, geotechnical, geological, material, and fracture property data, within the SMECDE framework. Ultimately, the AI-SMECDE method will enable early warnings of landslide risks and the digital mapping of areas vulnerable to rainfall-induced damage.