Pseudo-3D Back-Analysis and Predictive Modeling of Landslides Caused by the 2015 Lefkada Earthquake

The occurrence of landslides during major earthquake events poses a major threat to infrastructure, property and way of life in mountainous areas. Accurate prediction of landslide area, volume and location during strong ground motions is critical to assessing the risk posed by landslides and their consequences. Although reliable back-analysis of an individual landslide is feasible in geotechnical engineering practice using advanced numerical methods and reliable input parameters, such approaches do not easily scale to regional-level assessments due to the high computational costs and numerical complexity. In this study, we propose a novel pseudo-3D method for back-analysis and forward (i.e., predictive) modeling of landslides based on 1-D Newmark-type analysis for landslide triggering, followed by a geomorphic projection of the landslide on the landscape to derive its geometric characteristics (area and volume). An inversion scheme is employed to back-analyze large numbers of landslides considering their location, area and volume, where the results are then aggregated to derive regional Mohr-Coulomb parameters. The model is applied to several hundreds of landslides that occurred during the Mw 6.5 2015 Lefkada earthquake in Greece and regional estimates of strength are derived for the entire coastline. Subsequently, the geospatially derived strength parameters are used as input to compare the results of the analysis in a forward modeling scenario. Several factors assumed to affect model predictive capacity are investigated, such as input strength, DEM resolution and seismic loading. We find greater accuracy in the prediction of landslide location with the increase in the resolution of the digital elevation model (DEM), but also an overprediction of small landslides in areas that landslides did not occur. The results show the ability and limitations of the proposed pseudo-3D back-analysis and prediction model as a tool for regional landslide assessment following an earthquake.