A Data-Driven Approach to Post-Seismic Landslide Hazard Assessment

Earthquake-induced landslides cause a significant threat to communities living in earthquake-prone areas, as they potentially worsen the destructive impact of an earthquake event physically and socio-economically. This hazard emerges as an aftermath of strong ground motion in mountainous areas, which disturbs the stability of the hillslope material and reduces its shear strength, leading to failure. This earthquake legacy effect is often called shear-strength reduction (RSS). An understanding regarding this matter is important, as it can be used for an immediate post-seismic response and long-term mitigation strategies. However, incorporating RSS for post-seismic landslide predictions remains challenging due to the complex interactions between the hillslope and the ground shaking, making it hard to quantify the RSS degree. Applying the same RSS estimation method used for the 2008 Wenchuan earthquake to the 2023 Turkey earthquake, this study aims to estimate the RSS caused by the earthquake and incorporate it into the post-seismic landslide prediction model.

The study uses a data-driven approach to develop the co-seismic landslides prediction model, utilizing the co-seismic landslide inventories and various predictor variables to see which variable most strongly contributes to the failure. The model was evaluated with random (RCV) and spatial cross-validation (SCV). Simulations will be conducted using a seismic hazard map as a ground-shaking predictor variable to estimate the spatial distribution of earthquake-induced landslides for future events.

Preliminary results of the developed co-seismic landslide model showed that most of the morphometric variables significantly contributed to the failure, as well as the seismic factor, where only the sediment and metamorphic lithology gave a positive contribution to the failure. The Area Under the Curve (AUC) value from the RCV and SCV showed a strong correlation between observed and predicted landslide areas. The RSS will be integrated into the simulation output to evaluate its impact on the post-seismic landslide estimation, which is expected to provide valuable insight into the earthquake-induced landslide predictions.