Assessing Buried Landslide Rupture Surfaces Using Genetic Algorithms and Dynamic Flow Modeling

Abstract. Estimating the landslide volume and rupture geometry remains a critical challenge, particularly for landslides whose toe of the rupture surface is buried by displaced materials. This geometrical ambiguity leads to significant uncertainties in hazard assessment. To address this issue, this study proposes an integrated framework that couples a geometric search method with a physics-based dynamic model. This study employ the Genetic-Algorithm Ellipse-Referenced Idealized Curved Surface (GA-ER-ICS) to generate candidate rupture surfaces. Unlike traditional geometric fitting, the optimal rupture surface is constrained not only by topographic fit but also by the dynamic behavior of the post-failure motion. The validity of the approxi- mated geometry is verified by simulating the subsequent flow paths and deposition patterns using a GPU-accelerated two-phase grain-fluid model (MoSES_2PDF). The proposed method is validated against the 2009 Hsiaolin landslide and applied to the 2022 Provincial Highway No. 7 landslide event in Taiwan. Results demonstrate that the integrated approach successfully approximates the buried rupture surface, achieving a deposition coverage accuracy of over 75 % and reducing the uncertainty in volume estimation compared with the estimates derived from the difference between pre- and post-event Digital Elevation Models (DEMs). This study highlights the potential of using dynamic flow calibration to resolve static geometric indeterminacy in back-calculation of the landslide failure surface.