Learning to predict rainfall-triggered landslides
Deutsche Forschungsgemeinschaft (DFG)
Research Training Group NatRiskChange - Project I8
10/2018 - 09/2021
Rainstorms trigger a large number of landslides worldwide. But how does the amount and location of rainfall affect the size and spatial pattern of landsliding? Can we learn to better predict landslides based on forecasted rainfall, moving from static assessments of landslide susceptibility to dynamic assessments of landslide hazard? How might landslide hazard change in a changing climate?
In this project, we use two complementary approaches to address these questions: 1. statistical analysis of rainfall-triggered landslide inventories and 2. numerical slope stability modeling. In the first approach, we use machine learning techniques to analyze inventories of rainfall-triggered landslides. We combine landslide inventory data with observed rainfall measurements to learn if and how rainfall controls the spatial pattern of landsliding. In the second, we use numerical models of slope stability to understand how changes in rainfall affect the probability of landsliding.
In the context of a changing climate, this project aims to improve our ability to predict rainfall-triggered landslides and to advance practical landslide forecasting.
Prof. Dr. Jürgen Kurths , Potsdam Institute for Climate Impact Research (PIK)
Dr. Norbert Marwan , Potsdam Institute for Climate Impact Research (PIK)
Research training school NatRiskChange: www.natriskchange.de
PhD project I8: Learning landslide triggers from inventory data