Gas flux across the air-sea interface is often determined by the product of gas transfer velocity k) and the difference of concentrations in water and air. k is primarily controlled by wind stress on the air-sea interface, thus all parameterizations ofk involve wind speed, a rough indicator of wind stress, as one of the independent variables. We attempted to explore the spatial and temporal variations of k in the East (Japan) Sea using a database from Naet al. (1992). Three different parameterizations were employed: those of Liss and Merlivat (1986), Wanninkhof(1992), and Wanninkhofand McGillis (1999). The strong non-linear dependence of k on wind speed in all parameterizations leads us to examine the effect of time resolution, in which the binned wind speeds are averaged, on the estimation ofk. Two time resolutions of 12 hours (short-term) and one month (long-term) were chosen. The mean wind speeds were fed into the given parameterizations, resulting in six different transfer velocities of $CO_2$ ranging from 12 to 32 cm/h. In addition to the threefold difference depending on the choice of parameterization, the long-term average of wind speed results in a value ofk up to 20% higher than the short-term (12 hours) average of wind speed due to the non-Rayleigh wind distribution in the East (Japan) Sea. While it is not known which parameterization is more reliable, this study proposes that the time-averaged wind speed should not be used in areas where non-Ralyleigh wind distribution prevails such as the East (Japan) Sea. The net annual $CO_2$ flux was estimated using the value of k described above and the monthly ${Delta}fCO_2$ of Oh et al. (1999); this ranges from 0.034 to 0.11 Gt-C/yr.