Résumé
Coseismic landsliding is an important contributor to the long-term erosion of mountain belts. The scaling between earthquake magnitude and the volume of eroded material has been known for decades. However, the resulting geomorphic consequences, such as divide migration or valley infilling, are still poorly understood. Indeed, the prediction of the location of co-seismic landslides sources and deposits is a challenging issue and algorithms accounting for triggering of landslides by ground shaking are needed. Peak Ground Acceleration (PGA) has been shown to control at first order the spatial density of landslides. PGA can trigger landslides by two mechanisms: the direct effect of seismic acceleration on forces balance, and a transient decrease in hillslope strength parameters. To test the accuracy of modelling seismic waves effect by a cohesion drop, we use SLIPOS, an algorithm of bedrock landsliding based on a simple stability analysis (Culmann criteria) applied at local scale. The model is able to reproduce the landslide area-volume scaling relationship and the area-frequency distribution of natural landslides. Co-seismic landslide triggering is accounted for in SLIPOS by imposing a cohesion decrease. We calibrate the relationship between PGA and model cohesion via the landslides density. The spatial distribution of PGA is modeled using empirical relationships between earthquake source and PGA. We run the model on earthquake scenarios (Mw 6.5 to 8) applied to the Alpine fault of New-Zealand to predict the volume of landslides associated with large magnitude earthquakes. We demonstrate that simulating the effect of ground acceleration on landsliding using a cohesion drop leads to a realistic scaling between the volume of sediments and the earthquake magnitude.