Research >> Multiscale Comparison of Regional Rainfall-Induced Landslide Predictive Models
Research >> Multiscale Comparison of Regional Rainfall-Induced Landslide Predictive Models
Predicting rainfall-induced landslides across multiple spatial scales requires balancing broad regional coverage with the physical detail needed to characterize slope failure. While global empirical-probabilistic models can rapidly identify areas of elevated landslide hazard, high-resolution mechanistic models provide more detailed predictions of landslide occurrence and behavior within specific watersheds.
We compared two complementary approaches for Hurricane Maria (2017) in Puerto Rico:
Regional-scale prediction: The global LHASA model identifies landslide hazard hotspots across the island at a 1-km resolution.
Watershed-scale simulation: The Transient-Flow Pseudo-3D mechanistic model simulates rainfall-induced landslides at a 1-m resolution, predicting their locations, areas, volumes, and timing within a watershed in Utuado.
This multiscale framework connects regional hazard screening with detailed, physics-based analysis of critical watersheds, demonstrating how predictions at broader spatial scales can guide more targeted, high-resolution modeling.
Regional and high-resolution landslide models provide complementary information across spatial scales. Regional predictions help identify areas for further investigation, while fine-scale mechanistic simulations provide detailed insight into landslide locations, sizes, and timing. Linking these approaches offers a framework for moving from island-wide hazard assessment to detailed watershed-scale prediction.
Probability of at least one landslide per 1 km² cell predicted by the LHASA model in Puerto Rico for the day ending at the specified time (AST), with a zoomed-in view of the 2.5 km² area in Utuado, central Puerto Rico.
Modeled landslides at the end of Hurricane Maria predicted by the mechanistic model within the 2.5 km² study area, compared to the mapped landslides. Landslide outlines are exaggerated for visualization purposes.
Schematic workflow illustrating the multiscale landslide modeling framework, progressing from regional- to global-scale susceptibility analysis using LHASA, to basin-scale mechanistic modeling applied to the most critical watersheds identified by LHASA, and finally to advanced numerical analysis of individual critical hillslopes identified by the mechanistic model.