Accelerating generation of stochastic cyclone routes with GPU programming

Typhoon/Cyclone is an important part of the risk assessment of natural catastrophes. To estimate probabilistic cyclone damage for a very long time (e.g. 1,000 years, 10,000 years), the first step of catastrophe risk assessment model is to generate a stochastic set of routes derived by historical cyclone routes, which takes a long time. In order to accelerate the generation, we try to use GPU programming, and gain about 2X speedups.