APPLICATION OF ONE WEEK AHEAD WAVE PREDICTION FOR USING GROUP METHOD OF DATA HANDLING

Abstract

 The present study aims to predict nearshore wave heights and periods for one week in advance on the Japanese coast using Group Method of Data Handling with actually distributed three global wave forecast data. The results indicate that the GMDH-based wave height prediction model can improve the prediction up to 60 % in mean square error, while the GMDH-based wave period prediction model can do it up to 70 %. As a result, it is found that the best performing combination of three global data for training the GMDH-based model depends on location.