The fluxes as well as the parameters that enter the bulk formulae are evaluated against buoy observations from the Prediction and Research Moored Array in the Atlantic (PIRATA; data as available at http://www.pmel.noaa.gov/pirata/); since WHOI uses PIRATA observations to adjust errors in its flux estimates, we also use independent data from three buoy experiments that are not used in assimilation by WHOI. These are:  FETCH (the "flux, etat de la mer, et teledetection en conditions de fetch variable"), the Baltic Sea Swell Experiment (BASE), and ROMEO, which was part of a field program off the North Carolina coast of the United States. It is shown that the differences in latent heat flux estimates are likely due to differences in specific humidity used in deriving these two products while those in sensible heat flux are likely due to differences in air temperature (Figure 1) (Santorelli et al., 2011).

Figure 1. Annual mean difference between the IFREMER and WHOI datasets during 1996-2005 for (a) sea surface temperature, (b) specific air humidity, (c) surface wind speed and (d) air temperature (units  are degrees Celsius for  sea surface and air temperature, gkg-1 for specific air humidity and ms-1 for surface wind speed).

Experiments were conducted with satellite based estimates of specific air humidity and air temperature (Jackson et al. (2006, 2009)) in the IFREMER approach based on the COARE3.0 algorithm. Resulting estimates of latent and sensible heat fluxes at two buoy locations from the PIRATA array (12°N, 38°W and 4°N, 38°W) are illustrated (Figures 2 and 3). This resulted in significant improvements in both sensible and latent heat estimates by the IFREMER method.

Figure 2  Comparison of the 12°N, 38°W PIRATA weekly-averaged (a) latent heat and (b) sensible heat flux to IFREMER_1, WHOI and IFREMER_2, IFREMER SST and 10m wind speed and Jackson et al. (2009). 10m specific air humidity and air temperature) estimates. The time span is 2/1/99-12/26/05; there were 107 values available or comparison. A table of statistics, including bias, RMSD and correlation coefficient is included (units are Wm-2)

Figure 3  Comparison of the 4°N, 38°W PIRATA weekly-averaged (a) latent heat and (b) sensible heat flux to IFREMER_1,WHOI and IFREMER_2 (IFREMER_1 SST and 10m wind speed and Jackson et al. 10m specific air humidity and air temperature) estimates. The time span is 2/15/99-12/26/05; there are 138 values available for comparison. A table of statistics, including bias, RMSD and correlation coefficient is included (units are Wm-2).

*http://cersat.ifremer.fr/layout/set/print/news/products_informations/new_release_of_satellite_turbulent_fluxes_1992_2007

 **http://oaflux.whoi.edu/heatflux.html

Note: IFREMER_1 denotes results from the original IFREMER approach; IFREMER_2 denotes results from IFREMER_1 when temperature and humidity are replaced with values from Jackson et al. (2009).

Back to Results