easyWave-GPU computes tsunami wave propagations in a fraction of minute on GPU-powered machines. Thus , a single simulation, or several simulations with varying granularity, covering several hours of tsunami wave propagation, and tailored to the actual situation can be computed within seconds.
easyWave-GPU computes tsunami wave propagations in a fraction of minute on GPU-powered machines. Thus , a single simulation, or several simulations with varying granularity, covering several hours of tsunami wave propagation, and tailored to the actual situation can be computed within seconds. Since main uncertainty in the early warning tsunami forecasting is attributed to the uncertainty of source parameters, fast on-demand computations allows fast forecast re-assessment in case of their update.
The initial OpenMP version of easyWave was re-engineered to parallel multi-GPU processing. Finally, various optimization techniques were applied for further code acceleration.
The GPU-accelerated easyWave is relevant for applications dealing with fast tsunami forecasting like early warning and probabilistic hazard analysis. In this context the easyWave-GPU version is, for example, integrated as a service in the TRIDEC Cloud.
easyWave algorithm solves shallow water equations in long-wave approximation in spherical coordinates using leap-frog explicit time-stepping on a staggered finite difference grid. Numerical implementation follows the well-known TUNAMI-F1 algorithm. Boundary conditions presume full reflection at the shoreline. Typical grid resolution for trans-oceanic applications is 30 arc seconds to 2 arc minutes. Since coarse grid resolution may not resolve the wave shoaling process, Green's law can be optionally applied to estimate peak coastal tsunami amplitudes from the wave heights at offshore positions.
easyWave produces both 1D- and 2D-outputs. Wave time series could be recorded at any given location or point of interest (POI) in a simple text-table format. Hazard relevant data such as estimated time of arrival (ETA) and estimated wave height (EWH) are provided as well.
Additionally, 2D outputs of wave propagation snapshots are stored at regular time interval as binary GRD-files (Golden Software grid file format).