2:15 PM - 2:30 PM
[HDS10-03] PCOMCOT-Landslide: A New GPU-parallelized Two-layer Model for Landslides and Tsunami Generation over Complex Topographies

Keywords:landslides, tsunamis, 3D topographies, numerical simulation, Parallel computing
Numerical modeling of tsunami generation by landslides remains a challenge, especially over complex 3D topographies. Landslides typically occur on steep slopes or curved surfaces, where vertical acceleration and centrifugal forces significantly influence landslide motion. Existing landslide models generally employ curvilinear coordinate systems or Boussinesq-type equations to capture these effects. However, curvilinear systems rely on a well-defined downslope direction, which makes them unsuitable for irregular topographies. Meanwhile, the Boussinesq-type equations for landslides are more complicated than those for tsunamis, involving nonlinear higher-order derivatives that are prone to numerical instabilities. In this study, we develop a new model for landslide tsunami generation over complex topographies. First, we adopt a novel method to simulate landslides, which accounts for vertical acceleration and centrifugal forces by modifying the source terms in the shallow water equations. Next, we couple the landslide model with the dispersive tsunami model PCOMCOT within a two-layer framework. The numerical schemes are parallelized on a GPU using the CUDA Fortran programming language. As a result, the developed model, PCOMCOT-Landslide, is capable of accurately simulating landslides and the subsequent tsunamis with high efficiency. We validate this model against a series of laboratory experiments and the 2018 Anak Krakatau volcanic event. Good agreement between simulations and observations is obtained for both the landslides and tsunamis.