Classification of Earthquake-triggered Landslide Events - Review of Classical and Particular Cases

Seismically induced landslides often contribute to a significant degree to the losses related to earthquakes. The identification of possible extends of landslide affected areas can help to target emergency measures when an earthquake occurs or improve the resilience of inhabited areas and critical infrastructure in zones of high seismic hazard. Moreover, landslide event sizes are an important proxy for the estimation of the intensity and magnitude of past earthquakes in paleoseismic studies, allowing us to improve seismic hazard assessment over longer terms. Not only earthquake intensity, but also factors such as the fault characteristics, topography, climatic conditions and the geological environment have a major impact on the intensity and spatial distribution of earthquake induced landslides. Inspired by classical reviews of earthquake induced landslides, e.g. by Keefer or Jibson, we present here a review of factors contributing to earthquake triggered slope failures based on an `event-by-event' classification approach. The objective of this analysis is to enable the short-term prediction of earthquake triggered landslide event sizes in terms of numbers and size of the affected area right after an earthquake event occurred. Five main factors, `Intensity', `Fault', `Topographic energy', `Climatic conditions' and `Surface geology' were used to establish a relationship to the number and spatial extend of landslides triggered by an earthquake. Based on well-documented recent earthquakes (e.g. Haiti 2010, Wenchuan 2008) and on older events for which reliable extensive information was available (e.g. Northridge 1994, Loma Prieta 1989, Guatemala 1976, Peru 1970) the combination and relative weight of the factors was calibrated. The calibrated factor combination was then applied to more than 20 earthquake events for which landslide distribution characteristics could be crosschecked. We present cases where our prediction model performs well and discuss particular cases where it does not. These are e.g. cases of far distant, delayed or ancient earthquake induced landslides.