Typhoon waves play a major role in engineering and environmental problems. The long-term goal is to investigate variability and trends of typhoon waves over past 60years. However, the parameters in numerical wave models must to be calibrated before applying to any interesting region. In this paper, the most important parameters were selected from the source term of the SWAN wave model, and subjected them to sensitivity analysis to determine which parameter is the most sensitive. In order to find the optimal value of the parameter for improved modeling results, the ARS method was adopted to find the optimal value of the parameter to which the SWAN wave model is the most sensitive by comparing modeling results with observed data at three buoy stations in the northeastern Taiwan Sea near the coast. The procedure adopting the optimal parameter value of Cds in wave simulations did improve the accuracy of the SWAN wave model in comparison to default runs based on field observations at three buoys.