Three-Dimensional Tsunami Modeling Using GPU-SPHysics

Abstract

With the devastating effects of the 2004 Sumatra Tsunami, tsunami research is at an all time high. Tsunami forecasting and modeling has become exceedingly important in the anticipation of major disasters. Tsunami inundation modeling, how a tsunami invades a coastal area, is an extremely useful tool for the prevention of major disaster in tsunami laden zones. Using a new free-surface hydrodynamic modeling code called GPUSPHysics, accurate inundation and propagation models of tsunamis can be modeled at very high resolutions. GPU-SPHysics takes advantage of the extremely powerful computational power of a GPU (Graphics Processing Unit) and calculates the dynamics of fluids based on SPH (Smoothed Particle Hydrodynamics). The implementation of SPH on the GPU not only creates accurate, three-dimensional models but stunning visualizations of a tsunami wave breaking on beaches or other structures. Using the data from these models, coastal communities will be well prepared for any magnitude of tsunami that they may encounter by adjusting their infrastructure and disaster preparation to accommodate for this common disaster and potentially save many lives. To utilize GPU-SPHysics? models accurately, they must first be verified. NOAA (The National Oceanic and Atmospheric Administration) has provided benchmarks for tsunami inundation and propagation models. These benchmarks consist of analytic tests, laboratory tests and field tests. A key benchmark for GPU-SPHysics to be verified against is the solitary wave inundation on a sloping beach experiment. The solitary wave best represents the leading wave of a tsunami; hence it is vital to test other inundations that involve more complex structures than sloping beaches. Through visual analysis, the GPU-SPHysics solitary wave model, accurate to a small deviation, has been verified using the analytic calculation for maximum runup as provided by Synolakis. To verify other benchmarks provided by NOAA, GPU-SPHysics must be tested against multiple experiments. Once GPU-SPHysics has been verified for multiple data sets, it can be considered an accurate tool for hazard analysis.