Storm Surge Forecast Using a Neural Network-Case Study of Sakai Minato and Hamada, Japan

摘要

The present study aims at the development of after-runner storm surge (ARS) forecast model using an artificial neural network in Sakai Minato and Hamada on the Tottori Coast, Japan, that is made in 30 hours in advance. To develop an artificial neural network-based ARS forecast model, local meteorological and hydrodynamic parameters (surge level, sea-level pressure, depression rate of sea-level pressure and typhoon location of longitude and latitude) on the Tottori Coast are collected. A series of experiments is carried out to improve the ARS forecast model as varying the unit number from 13 to 130. As a result, it was found that the accuracy of ARS forecast model is improved as increasing the unit number. Then, we can obtain the best performance of the ARS forecast model with the given lead time. In addition, the optimal unit number for the given lead time can be determined.