Climatologically significant effects of space-time averaging in the North Atlantic sea - air heat flux fields

Differences between “classical” and “sampling” estimates of mean climatological heat fluxes and their seasonal and interannual variability are considered on the basis of individual marine observations from the Comprehensive Ocean–Atmosphere Data Set. Calculations of fluxes were done for intramonthly averaging and for 1°–5° spatial averaging. Sampling estimates give in general 10% to 60% higher values of fluxes than do classical estimates. Spatial averaging has a larger effect than temporal averaging in the Tropics and subtropics, and temporal averaging is more effective than spatial averaging in midlatitudes. The largest absolute differences between sampling and classical estimates of fluxes are observed in middle latitudes, where they are 15 to 20 W m−2 for sensible heat flux and 50 to 70 W m−2 for latent heat flux. Differences between sampling and classical estimates can change the annual cycle of sea–air fluxes. There is a secular tendency of increasing “sampling- to-classical” ratios of 1% to 5% decade−1 over the North Atlantic. Relationships between sampling-to-classical ratios and parameters of the sea–air interface, the number of observations, and the spatial arrangement of samples are considered. Climatologically significant differences between sampling and classical estimates are analyzed in terms of the contribution from different covariances between individual variables. The influence of different parameterizations of the transfer coefficients on sampling minus classical differences is considered. Parameterizations that indicate growing transfer coefficients with wind speed give the larger sampling minus classical differences in comparison with those based on either constant or decreasing with wind coefficients. Nevertheless, over the North Atlantic midlatitudes, all parameterizations indicate significant sampling minus classical differences of about several tens of watts per square meter. The importance of differences between sampling and classical estimates for the evaluation of meridional heat transport shows that differences between sampling and classical estimates can lead to 0.5–1-PW differences in meridional heat transport estimates