Abstract
A stochastic typhoon model with three properties of central pressure, moving direction and moving velocity is constructed by using B-spline functions and multi-variate autoregressive models. This stochastic typhoon model generates hourly data of typhoon location, central pressure, moving direction and velocity. Predictors against the rainfall intensity and the storm surge at Osaka are selected by using a stepwise regression analysis. About 2820 typhoons for 600 years are generated, and rainfall intensities and storm surges associated with the simulated typhoons are generated at Osaka. The concurrent characteristics of rainfall and storm surge; such as the frequency distribution of time lags, the spatial distributions of typhoon locations, when the peak rainfall intensities and the maximum storm surges occurred. At the same time the joint return periods of heavy rainfall intensities and large storm surges, are evaluated by using observed and simulated data.
