Revisiting the problem of the Gulf Stream separation: on the representation of topography in ocean models with different types of vertical grids

The difficulty of simulating a realistic Gulf Stream (GS) that separates from the coast at Cape Hatteras has troubled numerical ocean modelers for a long time, and the problem is evident in different models, from the early models of the 1980s to the modern models of today. The source of the problem is not completely understood yet, since GS simulations are sensitive to many different factors, such as numerical parameterization, model grid, treatment of topography and forcing fields. A curious result of early models is that models with terrain-following vertical grids (e.g., sigma or s coordinates) seem to achieve a better GS separation than z-level models of similar resolution, so the impact of the vertical grid type on GS simulations is revisited here. An idealized generalized coordinate numerical model is used to compare between a sigma-coordinate grid and a z-level grid while maintaining the same numerical code and model parameters. Short-term diagnostic-prognostic calculations focus on the initial dynamic adjustment of the GS from a given initial condition and imposed boundary conditions. In diagnostic calculations, wherein the three-dimensional flow field is adjusted to time-invariant temperature and salinity data, the GS is quite realistic independent of the grid type. However, when switching to prognostic calculations, the GS in the z-level model tends to immediately develop an unrealistic GS branch that continues along the continental slope instead of separating from the coast at Cape Hatteras. The GS is more realistic in either a sigma-coordinate model or in a z-level model with a vertical wall replacing the continental slope. Increasing the vertical resolution in the z-level model reduces numerical noise, but it does not solve the GS separation problem. Vorticity balance analysis shows that the Joint Effect of Baroclinicity and bottom Relief (JEBAR) and its associated bottom pressure torque are very sensitive to the choice of vertical grid. A stepped topography grid may disrupt the local vorticity balance near steep slopes; this vorticity balance may be important to develop a counterclockwise circulation north of the GS that pushes the GS offshore. Therefore, the study suggests that a smooth representation of bottom topography in ocean models by using either a terrain-following coordinates or a z-level grid with partial cells may allow a more realistic treatment of flow-topography interactions and potentially a better simulation of the GS. (C) 2016 Elsevier Ltd. All rights reserved.