The automation of a physically-based slope stability model for real-time landslide forecasting

Rainfall-induced landslides are becoming a growing concern for disaster management due to the increasing frequency of high-intensity rainfall events. Identifying the space and temporal occurrence of such phenomena is paramount to ensure the development of reliable early warning systems and to effectively reduce the element exposed at risk. Conducting this analysis at the regional scale is a significant challenge due to the spatial variability of hydrological, geomorphological and geotechnical properties. Physically-based landslide models aim to identify potentially unstable areas during heavy rainfall by calculating the factor of safety (FS) across a spatial grid, integrating hydrological and geotechnical models.

Fully automated integration of such models into a Landslide Early Warning System (LEWS) is, however, still challenging due to complexities in real-time data acquisition, variability in model parameters, computational demands, and the need for accurate real-time forecasting. The proposed methodology uses meteorological forecasts, provided through meteorological Application Programming Interfaces (APIs), in addition to topographic and soil data to predict FS with an hourly resolution. These are visualized dynamically in real time on the ‘NGI Live’ data platform developed by the Norwegian Geotechnical Institute (NGI). Values of FS for each grid are uploaded to a cloud database as geotiff files and can be visualized in the form of maps in NGI Live. These prediction models, which are running at regular intervals to pull updated weather data from forecast APIs, are the model runners. Static input data for the models are kept in cloud storage, while API keys and other sensitive information are kept secure in a cloud secret store. The NGI Live dashboard offers a gateway to on-demand access to state-of-the-art predictions and historical data, and provide support for physics-informed decision-making relevant to disaster risk reduction and asset management.

This work is the result of collaboration between NGI and Universidade Federal do Rio de Janeiro, Brazil, through the project NATRISK (337241), ”Enhancing risk management & resilience to natural hazards in India, Brazil, & Norway through collaborative education, research, & innovation”, supported by the Research Council of Norway.