Synoptic and climatological dynamic studies generally rely on bulk aerodynamic flux formulae to describe air-sea heat and momentum exchange on synoptic and climatological scales. Barometric pressure maps (which involve an intrinsic spatial and temporal averaging of the wind) and wind roses provide two sources of spatial and temporal wind information for flux calculations. Several investigators have shown that time-averaged estimates of the fluxes based on vector-averaged winds systematically underestimate the actual time-averaged fluxes.Using surface meteorological observations from 9 weatherstations in the North Atlantic Ocean and 2 weatherstations in the North Pacific Ocean, the three hourly stresses, latent heat fluxes and sensible heat fluxes were calculated. The sampled data and the calculated fluxes were then averaged over periods up to 28 days. The estimates of the averaged fluxes based on the vector-averaged winds were then compared to the directly averaged values. An upper bound for the difference in the two stresses calculations was directly proportional to the sum of the x and y component wind variances lost through the averaging process (in agreement with Fofonoff, 1960) and inversely proportional to the square of the vector-averaged wind speed. The wind-averaged and directly averaged flux estimates were grouped according to the Beaufort wind speed category and the averaging period. A multivariate regression was then performed to optimize a transformation from the wind-averaged to the directly averaged case. For all fluxes, the transformation dramatically improved the wind-averaged estimates of the climatological means and variances of the directly averaged fluxes. The residual error between the two estimates was decreased up to a factor of 5 over the uncorrected case and the correlation coefficients showed a moderate increase. The regression cofficients showed similar values for all temperate latitude stations.An empirical formula was found which interpolated the wind speed and averaging period dependences and duplicated the multivariate regression results. The data from the ten temperate latitude stations were grouped and a single formula was found which only moderately increased the errors between the wind-averaged and directly averaged estimates. The geographically averaged formula was not applicable at Station N, located at the northern extremity of the North Pacific Trade Wind region.Analysis of the 28 day wind-averaged flux spectral estimates showed that they underestimated the 28 day directly averaged flux spectral estimates. Application of the specific ship empirical formula greatly improved agreement between the two spectral densities and reduced the residual series power density at all frequencies. High latent heat flux errors at Station N were reduced by application of a seasonal correction.