Abstract
The accuracy of surface heat flux estimates from the NCEP–NCAR and ECMWF atmospheric model reanalyses is assessed by comparison with Woods Hole Oceanographic Institute research buoy measurements made during the Subduction Experiment in the Northeast Atlantic. Each of the reanalyses persistently underestimates the ocean heat gain in this region, the array-averaged net heat gain being less than the corresponding buoy value by 32 ± 9 W m−2 for ECMWF and 35 ± 12 W m−2 for NCEP–NCAR. The model biases are primarily due to a combination of underestimated shortwave gain and overestimated latent heat loss. They are similar in sign and magnitude but show a greater spread between the various buoys than was found in an analysis of operational model output by Moyer and Weller. The tendency for the reanalyses to overestimate the latent heat loss in this region is consistent with the results of other studies that show that a bias of this sort is to be expected given the choice of bulk flux algorithm in the models. The poor performance of the reanalyses contrasts with estimates based on ship meteorological reports in the Southampton Oceanography Centre (SOC) flux dataset. The array-averaged net heat flux from the SOC dataset agrees with the buoy value to within 10 W m−2. Similar results are obtained when the comparison is restricted to winter, which is the period most relevant to studies of subduction. The December–February array-averaged net heat flux is −52 W m−2 from the buoys, −57 W m−2 for SOC, −78 W m−2 for NCEP–NCAR, and −93 W m−2 for ECMWF. The results from the buoy comparisons reinforce the need for basin-scale evaluations of surface fluxes to be supplemented by local comparisons against high quality flux measurements.
Corresponding author address: Dr. Simon A. Josey, Southampton Oceanography Centre, James Rennell Division for Circulation, Room 254/31, Empress Dock, Southampton SO14 3ZH, United Kingdom.
