Toggweiler, J. R., and B. Samuels, 1995: Effect of sea ice on the salinity of Antarctic bottom waters. Journal of Physical Oceanography, 25(9), 1980-1997.

Abstract: Brine rejection during the formation of Antarctic sea ice is known to enhance the salinity of dense shelf waters in the Weddell and Ross Seas. As these shelf waters flow off the shelves and descend to the bottom, they entrain ambient deep water to create new bottom water. It is not uncommon for ocean modelers to modify salinity boundary conditions around Antarctica in an attempt to include a "sea ice effect" in their models. However, the degree to which Antarctic salinities are enhanced is usually not quantified or defended.

In this paper, studies of shelf hydrography and delta¹⁸O are reviewed to assess the level of salinity enhancement appropriate for ocean general circulation models. The relevant quantities are 1) the salinity difference between the water masses modified on the shelves and the final offshelf flow and 2) the flux of salt (or freshwater) that gives rise to this salinity difference. Onshelf/offshelf salinity changes in the Weddell and Ross Seas appear to be fairly small, 0.15-0.20 salinity units. The quantity of brine needed to produce this salinification is equivalent to the salt drained from <0.50 m of new sea ice every year.

Salt fluxes and salinity distributions from three GCM simulations are then compared. The first model has its surface salinities simply restored to the Levitus observations. Levitus restoring produces a slight freshening in the area of the Weddell and Ross Sea shelves. The global-mean bottom-water salinity in this model is 34.57 psu, which 0.16 units less than observed. The second model includes a very modest salinity enhancement in the area of the Weddell and Ross Sea shelves. This produces a salt flux equivalent to the formation of ~ 0.50 m yr^-1 of new sea ice. Even though this amount of salt input is close to the amount observed, global-average deep salinities in the second model are only 0.02 units greater than the deep salinities in the first model. The third model includes a large salinity enrichment, which is applied throughout the Weddell and Ross embayments without regard to water depth. Its deep salinities are 0.18 units higher than the deep salinities in the first model, but the amount of salt pumped into the model greatly exceeds the salt flux in nature.

The authors conclude that salt from sea ice is probably not a major influence on the salinity of Antarctic bottom waters. Predicted salinities in ocean GCMs are too fresh because of circulation deficiencies, not because of inadequate boundary conditions. Models that employ large salinity modifications near Antarctica run the risk of grossly distorting the processes of deep-water formation.