In this paper, the calculation of eddy covariance flux is reevaluated with newly-emerged capability of the wavelet transform. It is demonstrated that the wavelet transform clearly revealed the cospectral gap that separates turbulent and mesoscale motions with the data collected over a rice and a larch forest. It is also shown that the computed eddy covariance flux is insensitive to the method of decomposition of the turbulent fluctuation from the mean when averaging time is set at around the cospectral gap location. Therefore, information on the gap location removes the ambiguity in the calculation of eddy covariance flux.