An alternative surface air-sea buoyancy flux estimate for the Southern Ocean (SO, south of 20S) has been obtained using an oceanic state estimation: Southern Ocean State Estimate (SOSE, Mazloff, 2008). Heat, freshwater flux and wind forcing were adjusted to make results of a numerical simulation of the SO circulation consistent (within the prescribed error limits) with all available oceanic observations that were assimilated in the model (Mazloff, 2008).The initial forcing fields were prescribed by NCEP-NCAR Reanalysis 1 (NCEP1) data. The calculation was performed at eddy-permitting horizontal resolution of 1/6 degree; the results for years 2005 and 2006 have been analyzed. The resulting heat and freshwater flux estimates are compared with the numerical weather prediction model flux estimates (NCEP1, ECMWF operational model), air-sea heat and freshwater flux estimates by Large and Yeager (2008; LY hereafter) and the final report of Working Group on Air-Sea Fluxes (WGASF, Taylor et al., 2000). The quality of this new air-sea buoyancy flux estimate was assessed by:

(1) evaluating the extent to which SOSE adjustments of NCEP 1 forcing fields correct the corresponding biases estimated by WGASF, Taylor et al. (2000),

(2) comparing SOSE adjustments of atmospheric forcing fields to corresponding LY adjustments of atmospheric forcing fields (note that LY heat flux estimates are indendent of NCEP1 heat flux estimates because, although both SOSE and LY adjust the meteorological variables given by NCEP1, but whereas SOSE uses NCEP1 heat and freshwater fluxes as the initial forcing, LY flux estimates do not use NCEP1 flux data in any way).

(3) using all the above noted air-sea buoyancy flux estimates, including SOSE, as input to a Walin (1982) analysis to estimate Subantarctic Mode Water (SAMW) formation. Accurate buoyancy flux estimates are of crucial importance for SAMW formation estimates since in the formation region (on the equatorward side of the Subantarctic Front) the net buoyancy flux is determined by the difference between the buoyancy loss due to cooling and the buoyancy gain due to the freshwater flux.

The results of the comparison of SOSE fluxes with other buoyancy flux products are not easily summarized, but some main features emerge.

(1) Where the SOSE adjustments to NCEP1 are of greatest magnitude, they almost always act to reduce NCEP1 biases as reported by Taylor et al. (2000).

(2) SOSE adjustments tend to agree qualitatively well with LY adjustments in the large scale pattern although quantitative differences remain.

(3) The density ranges within which Southern Ocean mode waters are formed are more faithfully estimated using SOSE buoyancy fluxes than either NCEP1, ECMWF operational model or LY fluxes.

By geographical region the most significant large scale adjustments in net heat flux are: (i) in the polar region (south of 40 S) where the net ocean heat loss was increased; (ii) in the subtropical gyre of the South Indian ocean (west of Australia) where SOSE adjustments decreased net ocean heat loss; (iii) south of Cape Aghulas where SOSE adjustments reduced the net annual ocean heat loss; (iv) over the Brazil/Malvinas Falkland Current where SOSE decreased the net ocean heat loss.

The largest SOSE adjustments in precipitation were: SOSE decreased evaporation in the subtropical gyre and decreased the precipitation in the polar region. The SOSE adjustment of precipitation was very similar to LY adjustment of precipitation both in pattern and the amplitude.

The estimated SAMW formation rates were 4.6 Sv of 26.8 sigma-theta Southwest Indian SAMW and 5.3 Sv of 27.0 sigma-theta Pacific SAMW for 2005; 3.5 Sv of 26.6-26.8 sigma-theta Southwest Indian SAMW and 6.0 Sv of 26.9 sigma-theta Pacific SAMW for 2006.