Abstract
Potential density referenced to sea level pressure $(\sigma\sb0$) has shown its usefulness as vertical coordinate in ocean models in many ways, but there are problems with $\sigma\sb0$ (potential density referenced to sea level) coordinates in the deep ocean: $\sigma\sb0(z)$ can be multivalued, leading to coordinate folding, and $\sigma\sb0$ surfaces can deviate from the so-called neutral surfaces, which are the surfaces along which turbulent lateral mixing takes place in a stratified medium. The reason for both of these problems is that most isopycnal models regard seawater as uniformly compressible. However, the effect of water temperature on compressibility cannot be ignored.In this study a two-pronged approach is taken to improve the model accuracy. First, since the effects of compressibility variation are proportional to the difference between the local and the reference pressure, we replace the model's traditional $\sigma\sb0$ coordinate by $\sigma\sb2$ (potential density referenced to 2000 dbar). This step eliminates many of the coordinate folding problems associated with $\sigma\sb0$ and generally reduces the difference between coordinate and neutrally buoyant surfaces. Second, we split the compressibility coefficient into a pressure- and a temperature-dependent part and, recognizing that the former is dynamically passive, retain only the effect of the latter in the governing equations. This is accomplished by introducing a new variable called "active density"--the density with the pressure-related compressibility removed. Therefore, $\sigma\sb2$ is adopted as vertical coordinate, but active density is used to express the seawater density within the layers.The above changes are applied in a near-global, 16-layer, 2$\sp\circ$ x 2$\sp\circ$cos (lat.) Miami Isopycnic Coordinate Ocean Model (MICOM). The model is driven by observed atmospheric conditions. MICOM modified in this fashion produces realistic meridional mass and associated heat fluxes in the three major ocean basins. A realistic formation rate of a few major water masses, including the North Atlantic Deep Water and the Antarctic Bottom Water, is also obtained. To summarize the various circulation features obtained by the model, a three-dimensional mass transport diagram spanning four density classes is constructed. Many features in it compare well with those revealed by observed hydrographic data.