A new simulation model of snow avalanche accelerated by GPGPU technique

  • TSAI, I-Chen (Tokyo Institute of Technology)
  • Nakamura, Takashi (Tokyo Institute of Technology)

Please login to view abstract download link

The objective of the study was to build General-purpose computing on graphics processing units (GPGPU)-based snow avalanche model in the global coordinate. The model successfully recreated Mt. Nasu snow avalanche event in Tochigi Prefecture, Japan. The simulation results of the flow regime and deposit showed a good agreement with the site investigation report. With the velocity and the snow depth data at the victim's position, the victim's moving path was simply evaluated by Runge-Kutta method. It helped to estimate the final position of the snow avalanche victim, which is sufficient information for mountain rescue. Besides, the calculation cost was compared in this study. When the mesh size is 1m uniform in both x- and y- directions, the elapsed time for calculation under a one-core CPU and GPU was 8hr and 3min, respectively. The GPGPU application speeded up at least 150 times faster. With the high-performance calculation of GPGPU, snow avalanches simulation in several initial conditions without massive calculation costs are expected. For 15 different snow avalanche cases simulation, it may only take 45min by GPU, much faster than 5-days by a one-core CPU. The GPU application will help to create a hazard map in a short time in the future.