A global dataset and model of earthquake-induced landslide fatalities

Abstract

In this study, we present a newly developed, comprehensive dataset of 196 historical earthquakes, including 127 events with known landslide fatality counts, which we use to examine the impact of earthquake-induced landslides. Using this dataset, we assess the potential impact of seismically induced landslides on affected populations, based on the exposure of population to expected landslide occurrence. We calculate predicted landslide probabilities out of a global statistical landslide model in the area surrounding each earthquake using USGS’s ShakeMap model of strong ground motion. We identify estimated population exposure by comparing the predicted probability grid with a time-corrected global population dataset to determine a predicted landslide exposure index. We compare these values to the number of actual fatalities for 91 training events and use these comparisons, together with the United Nations Human Development Index as a vulnerability proxy, to develop a multiple regression model that can potentially be used to provide order-of-magnitude estimates of human impact of earthquake-induced landslides. We observe a significant positive correlation between predicted and observed fatalities, but with very high variability in fatality rates for similar exposure levels, suggesting that other factors (e.g., landslide type and density, the effect of urbanization on population exposure) should be used to improve this estimate. We also demonstrate that these estimates of landslide impacts can be used in a deterministic mode, in collaboration with use of scenario earthquakes, to assess the potential impacts of anticipated future events. With further development, these landslide scenarios can be used by vulnerable communities to improve land use planning, structural design, and emergency response in landslide-prone areas.