To estimate long period ship motions in harbors, specific wave information like long wave spectrum is more useful than the mean value like information such as the significant wave height and the significant wave period. As for wave field calculations in harbors, so-called Boussinesq's model seems to be widely used because of its versatile features. However, it hasn't been clarified yet about how large the sponge layer should be taken to simulate open-sea condition; how long the computation should be continued to obtain statistically stable spectrum; what size of the computational area should be considered, etc. In this study these basic conditions were investigated by conducting wave field simulations by Boussinesq's model. Useful information was obtained such as: the data length, from which the spectrum is calculated, need to be longer than at least 5 to 6 hours to get a stable spectrum; the length of the sponge layer could be much less than the length expected from one-dimensional calculations; the size of computational area doesn't affect the results so much if it is as large as the one empirically taken in computations.
