SENSITIVITY OF A WORLD OCEAN GCM TO CHANGES IN SUBSURFACE MIXING PARAMETERIZATION

The sensitivity of a coarse-resolution model of the World Ocean to par ameterization of subgrid-scale mixing is examined. The model is based on the GFDL code. Results are presented from a series of model runs wh ere the subsurface mixing parameterization is sequentially upgraded to ward a more physical representation. The surface forcing is the same f or all principal model runs and features a strong relaxation of surfac e temperature and salinity toward perpetual wintertime observed values . One model version is rerun with a full annual cycle of surface forci ng and verifies that use of the perpetual winter surface relaxation in troduces only minor biases in the essential characteristics of the sol ution. Runs 1 and 2 feature the diffusivity tensor in the traditional horizontal/vertical orientation, and examines the effect of different vertical diffusivity profiles on the solution. Results are compared wi th those of previous studies. In both cases, the water mass properties (especially the salinity fields) are rather poor. In runs 3-5, a stan dard parameterization is introduced that allows for enhanced diffusion along the isopycnal surfaces. Each of these runs feature a different prescribed profile of isopycnal diffusivity, though with the same prof ile of vertical diffusivity as for run 2. Introduction of isopycnal mi xing considerably improves the water mass structure, in particular by freshening and cooling water at intermediate depths toward realistic l evels. However, the vertical stratification and density fields are lit tle changed from mn 2. Likewise, the current structure and meridional overturning are little changed. Thus isopycnal mixing has a major effe ct upon the temperature and salinity fields, but very minor effect on the ocean dynamics. Isopycnal mixing is found to modestly increase pol eward oceanic heat transport in the midlatitudes via enhanced quasi-ho rizontal mixing of warm salty subtropical and cold fresh subpolar wate rs. In run 6, the isopycnal diffusivity of run 4 is retained, but the vertical diffusivity is instead allowed to vary as the inverse of the local Brunt-Vaisala frequency. However, the resulting solution is litt le changed from that of run 4. Reasons for this small change are discu ssed. Also discussed are the impact of numerical problems associated w ith the use of realistically small vertical diffusivity, and problems inherent in deep water formation in coarse-resolution models.