A pseudo-3D methodology for regional-scale back-analysis of earthquake-induced landslides

Abstract

Major earthquakes can cause hundreds to thousands of landslides in mountainous areas, but the methods for regional earthquake-induced landslide analyses are currently not well developed. This paper proposes an algorithm for back-analyzing earthquake-induced landslide inventories. The result of the back-analysis (or “inversion”) is a modeled three-dimensional (3D) “best-matching” landslide for each mapped landslide in an inventory as well as an estimate of average shear strength along the landslide failure plane. The algorithm leverages seismic displacement models to detect the triggering location (i.e., triggering grid cell) of modeled landslides. The triggering grid cell is then projected to form a 3D cone-shape failure surface for each modeled landslide using geomorphic characteristics, i.e., flow direction and flow accumulation, and a geometric rule, a procedure termed here as pseudo-three-dimensional (pseudo-3D). The inversion algorithm uses the landslide location, area, and volume (if available) of each mapped landslide to find a modeled landslide that minimizes the “mismatch” between the two, expressed as total error, and outputs an associated triggering depth, landslide geometry, and strength. The estimate of strength for each individual landslide can then, if needed, be aggregated using a geospatial methodology (e.g., geologic unit information or a K-means clustering method) to derive regional-scale Mohr-Coulomb shear strength parameters. The potential bias of the proposed pseudo-3D methodology on the strength estimate is assessed against synthetic landslide inventories generated using a 3D limit equilibrium method. The back-calculated friction angles agree well with 3D results, while the cohesions are lower than the estimated cohesions from 3D slope stability analysis. In addition, the effect of different seismic displacement models on the results of the back-calculated strengths is investigated. The proposed methodology is implemented to the 2015 Lefkada earthquake event in Greece, which triggered more than 700 landslides, with the goal to demonstrate the efficiency in the inversion of many landslides and the generation of regional estimates of strength parameters. For this study region, the spatial distribution of the shear strength results indicates that areas with fewer landslides have higher shear strength while areas with more landslides have lower shear strength.