Item

Abstract

Using a suite of simulations with the Max Planck Institute Ocean Model (MPIOM) at resolutions of about 0.1°, 0.4° and 1.5°, we study the impact of resolved and parameterized vertical eddy fluxes on the long-standing biases obtained when running MPIOM at low resolutions. In the 0.1° simulation, the eddy heat and salt fluxes have three features in common. First, their horizontal area averages are both upward, counteracting the downward fluxes due to time-mean circulations. Second, their divergences at intermediate depths are both negative, acting to cool and to freshen water masses, thereby reducing the major long-standing warm and saline biases of the low-resolution MPIOM at these depths. Third, both the heat and salt budgets are dominated by a balance between the divergence of eddy flux and that of mean flux. The vertical profiles of the tendency forcing due to parameterized eddies resemble those due to resolved eddies. This resemblance does not guarantee a bias reduction, as the tendency forcing terms are much less well compensated in the 0.4°- and 1.5°-simulation than in the 0.1°-simulation. When concentrating on the eddy-induced transports, we identify two situations in which the eddy effect is not appropriately represented by the GM-parameterization. One emphasizes the importance of the mean tracer distribution and the other the importance of the simulated isoneutral slope in determining the eddy-induced transports. Given the mean salinity distribution in the Southern ocean, characterized by a tongue of fresh Antarctic Intermediate Water, the salinity advection via eddy-induced transport tends to strengthen, rather than to weaken, the saline biases. Due to the density biases in a widened region of the Agulhas current in the low-resolution runs, the isoneutral slope vectors are erroneous and the large parameterized eddy-induced transports do not occur where they should. © 2016 Elsevier Ltd