A Reduced-Gravity, Primitive Equation, Isopycnal Ocean GCM: Formulation and Simulations

A reduced gravity, primitive equation, ocean GCM with an isopycnal vertical coordinate is developed. A “buffer” layer is introduced to allow the mixed layer to detrain mass at arbitrary densities without the coordinate drift or the heat loss suffered by other isopycnal models. The diapycnal velocity is derived from the thermodynamic equation. Negative layers are removed by a heat- and mass-conserving convective adjustment scheme. The model formulation on a β plane employs an A grid and allows irregular coastlines and local grid stretching. Simulations with climatological winds and surface heat fluxes based on observed sea surface temperatures (SSTs) are presented for the Atlantic, the Pacific, and the Indian Oceans. The surface mixed layer is modeled as a constant depth layer, and salinity effects are neglected in this version. The surface heat flux parameterization used here leads to errors in model SSTs, which are reasonable in the Tropics but are higher in the western boundary current regions. The seasonal dependence of the currents compare reasonably well with the available observations and other model results, though there are differences in the amplitudes of the currents. The model thermocline reproduces the observed slopes, troughs, and ridges in the Tropics. Neglecting salinity effects and lack of a variable depth mixed layer affect the model simulation of the thermocline at higher latitudes. The cold tongue in the eastern Pacific is also affected by the assumption of a constant-depth mixed layer, but the warm pool in the west and the zonal slope of the thermocline correspond well with the observations. The Gulf Stream and the Kuroshio have reasonable current speeds but separate slightly earlier than observed, in contrast to most models that separate late. Seasonal reversal of the Somali Current in the Indian Ocean and the South Equatorial Current in the Pacific Ocean are reproduced, and the Equatorial Undercurrent is stronger than in most models with comparable grid resolution. Effects underway to improve model performance are listed along the way.