Prediction of road blockages caused by rainfall-induced landslides

Abstract

This paper develops and evaluates an approach to predict road blockages caused by rainfall-induced landslides, using an indicator for landslide triggering (landslide probability, rainfall) and modelling landslide runout via the viewshed (area visible from the road). Based on the landslide inventory of 2023 ex-tropical Cyclone Gabrielle, the study investigates the prediction potential of this approach, focussing on the influence of different digital elevation model (DEM) resolutions. Findings suggest that coarser DEMs (20 or 25 m) slightly outperform finer resolutions (5 or 10 m), likely due to the other input variables presenting similar resolutions. While the viewshed approach effectively identifies larger road blockages, it fails to predict smaller blockages. Results also indicate a tendency to overestimate the extent of blockages, which can be addressed by evaluating landslide hazards at a road link scale (e.g. between intersections) rather than individual 100 m-segments. Uncertainties arise from using precipitation as a triggering variable as the precise time of landslide initiation is unknown, preventing the accurate calculation of the accumulated rainfall. Despite limitations and the need for further research, the viewshed approach presents a valuable tool for the prediction of road blockages during future rainfall events, providing critical information for emergency planning and mitigation efforts.