Differences between two estimates of air-sea turbulent heat fluxes over the Atlantic Ocean

Uncertainties in turbulent ocean-atmosphere heat flux estimates, both among the estimates and between them and ground truth, suggest that further comparisons are needed. We analyze estimates from the French Research Institute for Exploitation of the Sea (IFREMER) and the Woods Hole Oceanographic Institution's Objectively Analyzed air-sea Fluxes (WHOI OAFlux). The IFREMER products are based on satellite observations and the WHOI OAFlux ones on data from satellites, buoys, and ships assimilated into numerical analyses. We focus on the Atlantic sector (70 degrees W-30 degrees E, 45 degrees S-45 degrees N) during 1996-2005, where the variables that enter the bulk formulae for computing fluxes (wind speed, sea surface and air temperature, and specific humidity) can be evaluated against buoys in the Prediction and Research Moored Array in the Atlantic (PIRATA). Since WHOI assimilates PIRATA observations, we have added two independent buoy data sets: FETCH and ROMEO. To examine how each variable contributes to the difference between estimated and buoy fluxes, the method of Bourras (2006) is applied. His so-called Q terms showed that specific air humidity and air temperature contributed the most to the biases of IFREMER latent and sensible heat fluxes, respectively, at both independent buoys. For WHOI OAFlux products, deviations from FETCH values were mainly due to wind speed and sea surface temperature differences, while in comparison with ROMEO fluxes, WHOI OAFlux biases were primarily due to specific humidity and sea surface temperature estimates. Modified estimates of turbulent fluxes with the IFREMER approach using the 10 m specific humidity and air temperature products of Jackson et al. (2009) show significant improvement in three test cases at PIRATA buoys.