Coastal wave forecasting by dynamical model coupled with a statistical method

The coupled coastal wave prediction scheme ,which is a combination of a multi-scale numerical model and a statistical method,is proposed in order to avoid the limitations of one single scheme. By ocean wave model,the wave energy density spectrum of the computational grid in the coastal model is forecasted. We have defined a transfer coefficient matrix for thewave energy density spectrum between the computational grid and the coastal forecasting point. A statistical model for the prediction of this transfer coefficient is established using empirical orthogonal function (EOF) and Kalman filtering method. This statistical model is then coupled with the numerical model. The wave energy density spectrum of computational grid is optimized using the observed coastal buoy data. The coastal wave forecasting are validated by the observations of NAHA station for one year,indicating that this coupled method significantly improved the prediction power compared with the numerical model on its own. The rootmeansquare error of the significant wave height reduces about 0.16mand the average relative error is reduced by about 9%. It is also found that the forecasting accuracy of this method is limited within 24 hours; the principal components decomposed from the wave energy density spectrum reflect the main characteristics of local wave climate; and the change transfer coefficient of the spectrum reflects the seasonal variation of the wave climate.