The third-generation wave prediction model, WAM Cycle 5, can simulate ocean waves with the assimilation of the model parameters. The background estimation errors may be ignored through the minimization of the evaluation function value.
　The model for assimilating wave spectra on the boundary of a computational domain was developed in this study with the adjoint method (4-dimensional assimilation method). For the assimilation of wave spectrum components at many computation grid points, the background errors should be counted in the evaluation function. It is preferable to take observed wave data at many ports at short interval time. The off-diagonal elements of the covariance matrix of the background errors should be taken into consideration for precise assimilation, however the elements may be ignored when the stable assimilation can be carried out with sufficient observed data.
　Ocean waves in four sea areas around Japan during five storms were simulated using wave observation data at NOWPHAS points. The effect of the observed data at each port on the final estimation values was evaluated.