A Study on Applicability of Ensemble Kalman Filter for Wave Prediction Model

The Routes of ships are mainly determined based on their speed and roll angles. The routes are also determined based on the estimated weather and wave situation predicted using numerical simulation, economic and political situations, and the captain’s experience. Many researchers have used data assimilation techniques for improving the accuracy of wave prediction. However, there are few studies that focus on data assimilation using ensemble Kalman filter (hereinafter, EnKF) for wave prediction. In this study, we focused on one factor, namely wave prediction for improving the accuracy of the wave prediction. We constructed a data assimilation system that applies EnKF to a wave prediction model SWAN, in order to make wave predictions using real-time observational data obtained from ships. We discuss its applicability by targeting typhoon No. 23 which occurred on October 13, 2004. We also performed sensitivity analysis for examining the accuracy of estimation by changing the number of ensemble members and the parameters of random number generation. In the cases using EnKF, the errors in the significant wave height and peak wave periods between the observational data and simulation results were smaller than those in the original case. It was shown that the accuracy of wave prediction both offshore and near the coast can be improved by using NOWPHAS wave data. Thus, the proposed wave prediction model has enough accuracy to predict the wave situation. Moreover, in cases where a small observation error was assumed, the error in the estimated significant wave height could be reduced by about 93.8 % (at the maximum) compared to the case without using data assimilation.