Model Drift

Because the model is run directly from the Visbeck climatology and zero velocity without any spinup time, a certain amount of model drift is expected. Interannual variability such as the NAO means that some drift does occur in the Labrador Sea from year to year and must be separated from model error. The model's uppermost 50 meters actually warm up 1 degree Celsius in one year (see Fig.3). Conversely, there is about 1 degree Celsius of cooling directly below that. This dipole structure extends to a depth of almost 600 meters. The drift in salinity has roughly the same vertical structure as temperature. Increases in the spatial mean temperature of 1 degree Celsius per year are much too large at the surface. The surface temperature becomes too large and the depth of the mixed layer is too small. One possible explanation for this model drift is the advection of abnormally cool and fresh Irminger Sea Water through the eastern boundary at depths of 200-700 meters. However, it first should be noted that there is no mixed layer physics scheme in this model run. The missing physics can explain the sign reversal in drift temperature in the upper ocean. Heat does not diffuse downward into the ocean deeply enough. Therefore, a lack of mixed layer physics probably contributes most to model drift. I believe that the dipole temperature structure reaches all the way to 600 meters since the average mixed layer depth at initialization is almost 500 meters. The depth averaged temperature drift is actually a cooling of 0.3 degrees C/yr. The net cooling of the entire Labrador Sea seems to be reasonable since the second half of the winter of 1996-97 had large heat losses to the atmosphere and spawned deep convection. There is almost no model drift in the deep levels of the model. The addition of the KPP (Large et al) mixed layer physics model is recommended to improve this GCM run, which will hopefully remove spurious model drift in the upper levels.