A near real-time forecasting system for shallow landslides at regional scale based on distributed modelling and weather predictions

Abstract

This work presents the development of a forecasting system for shallow rainfall-induced landslides, designed to provide daily near–real-time outputs at a regional scale. The system integrates hydrological and slope stability distributed modelling with daily rainfall forecasts from a weather prediction model to generate daily shallow landslide initiation forecasts for an area of hundreds of square kilometers. Research activities focused on: i) the optimization of the HIRESSS model to enhance computational efficiency and physical consistency for real-time forecasting; (ii) the calibration of the system based on the hydrological, geotechnical, morphological and climatic characteristics of the study area and (iii) developing algorithms enabling the continuous use of meteorological forecasts and the production of both distributed (grid-based) and aggregated (sub-basin) outputs. The operational system based on the HIRESSS model has been active since June 2024 for the Alert Zone B of the Aosta Valley Region. It uses precipitation forecasts from the ICON-CH1 weather prediction model as dynamic input. The system generates landslide initiation susceptibility maps in terms of failure probability, with a 10 m spatial and 3-hour temporal resolution, producing forecastsfor the current and following day. Additionally, it provides aggregated probabilities at sub-basin level through a calibrated threshold system, supporting early warning activities. Recent developments have focused on optimizing the real-time dissemination of model outputs through an open-data platform ensuring transparency, accessibility, and operational usability for regional authorities and decision-makers. The developed system represents a significant step toward a fully operational landslide early warning system at the regional scale based on physically based distributed modelling and meteorological forecasting.