A STUDY OF APPLICATION OF NEURAL NETWORK TO REAL TIME STORM SURGE FORECASTS AT THE SANIN COAST

Abstract

 In the study, an application of a neural network is examined in order to forecast a real-time storm surge at the Sakai Minato on the Sanin Coast. In the forecast experiments, sea level pressures, depression rates of sea level pressures, wind speeds, wind directions, tidal levels and storm surge levels during only one typhoon (T0418) and two typhoons (T0418 and T0314) were trained. A storm surge during Typhoon 0415 was forecasted using the neural network. From the results of the experiment, it is found that the forecasted storm surge level trained with the tidal level, the storm surge level, the depression rate of sea level pressure are close to the observed storm surge level. In addtion, the forecasting results are significantly improved in maximum storm surge level with the increase in the number of the observation stations and with the inclusion of the wind speed in training the neural network. From the experiment of forecast times, it is seen that the predictions in the 1 and 2 hour forecasts are more accurate than those in the 3, 4 and 5 hour forecasts.