Parameterization of density-driven downsloping flow for a coarse-resolution ocean model in z-coordinate

In the World Ocean, densest waters found on the continental shelves induce density driven downsloping currents that can influence the deep ocean water masses properties. This process is poorly represented in z-coordinate ocean models, especially in Ocean General Circulation Model (OGCM) with coarse resolution in both horizontal and vertical directions. Consequently, continental shelves appear to be too isolated from the open ocean, whereas the density remains too low in the deep ocean. This study presents a simple parameterization of downsloping flow designed for z-coordinate, coarse resolution ocean model. At the shelf break, when the density on the shelf is higher than that in the neighbouring deep water column, a downsloping current is set up. This current is linearly related to the horizontal density gradient between the two adjacent boxes, using a prescribed coefficient. For simplicity, a uniform value of the coefficient is used here, although it should ideally vary in space. From the shelf, the downsloping flow is assumed to go downward along the slope until it reaches a level of equal density. An upward return flow of equal magnitude maintains the conservation of mass. This parameterization has been implemented in an OGCM and two experiments, with and without this scheme, have been integrated until equilibrium using restoring boundary conditions. The impact of the downsloping parameterization on the global ocean is dominated by the improvement of the Antarctic bottom water circulation and water mass properties. The parameterization increases the density of the deep ocean and tends to reduce the intensity and depth of the North Atlantic deep water circulation, which is in better agreement with observations. As a result of a higher exchange with the open ocean, the properties of continental shelf waters are also improved, with a marked reduction of the Antarctic shelves salinities. Therefore, this simple parameterization leads to a significant improvement of the model results, at little computational cost.