HYBRID REAL TIME PREDICTION OF STORM SURGES CAUSED BY TYPHOON WHICH USES NEURAL NETWORK METHOD AND NUMERICAL MODEL

Abstract

The purpose of this research is a real-time forecasting of a storm surge caused by typhoon. The lead time of forecasting was set six hours in consideration of time that it was possible to take shelter. Various parameters of the typhoon model used in the storm surge analysis were presumed by the neural network technique (BP method). The typhoon data which had passed the surrounding of Shikoku between 1991 and 2004 was used for the study data in the BP method. The numerical analysis of the storm surge for the T0514 typhoon was executed based on the predicted parameters by the BP method and the practicality of this forecasting system was examined by the comparison between observed storm tide and calculated one.