A 2 1/2-layer, thermodynamic numerical model is used to study the dynamics, thermodynamics and mixed-layer physics of Indian Ocean circulation. A surface mixed layer of temperature T(m) is imbedded in the upper layer of the model, and entrainment and detrainment in the mixed layer are determined by wind stirring and surface cooling. There is also detrainment w(d) through the base of the upper layer that models subduction. Monthly climatological data, including air temperature T(a) and specific humidity q(a), are used to force the model, and model sea surface temperature (SST), T(m), is used to determine the sensible and latent heat fluxes. With a few notable exceptions, our main-run solution compares well with observed current and SST data; this is particularly true for T(m), which typically differs from observed SST by less than 0.5-1.0-degrees-C. Our analyses focus on three topics: the relative importance of remote versus local forcing, the thermodynamic processes that determine the model SST field, and the development of meridional circulation cells. There are a number of examples of remotely forced circulations in our main run. During the spring a northeastward countercurrent flows against the prevailing winds along the Somali coast north of 4-degrees-N, and from October through February a southwestward Somali Undercurrent is present from the tip of Somalia to 3-degrees-N; both of these flows result in part from forcing during the previous Southwest Monsoon. From March through May there is another southwestward Somali Undercurrent south of 7-degrees-N, generated primarily by the propagation of a Rossby wave from the west coast of India. The currents along the west coast of India are either strongly influenced or dominated by remote forcing from the Bay of Bengal throughout the year. A northeastward flow is well established along the east coast of India in March, long before the onset of the Southwest Monsoon; it is remotely forced either by upwelling-favorable, alongshore winds elsewhere within the Bay of Bengal or by negative wind curl in the western Bay. Finally, the Agulhas Current is strengthened considerably in a solution that includes throughflow from the Pacific Ocean. To investigate the relative importance of thermodynamic processes, we carried out a series of test calculations with various terms dropped from the T(m)-equation. There is little effect on T(m) when the sensible heat flux is set to zero, or when the solar radiation field is replaced by a spatially smoothed version. When temperature advection is deleted, T(m) is most strongly affected near western boundaries since isotherms are no longer shifted there by the swift currents; the annual-mean, surface-heat-flux field QBAR is also changed, with QBAR becoming more positive (negative) to compensate for the absence of warm (cold) currents. Without entrainment cooling, T(m) never cools during the summer in the intense upwelling regions in the northern ocean, and the annual-mean heat gain through the ocean surface (the area integral of QBAR over the basin) reverses to become a net heat loss. In individual tests without entrainment cooling, with T(a)=T(m), and with q(a) set to 80% of its saturated value q(s), model SST warms near the northern and southern boundaries during their respective winters by about 1-degrees-C, indicating that several processes contribute to wintertime cooling. The T(m) field degrades considerably in a single test run with both T=T and q(a)=0.8q(s), so that one or the other of these external forcing fields is required to be able to simulate SST accurately. The annual-mean circulation has two meridional circulation cells. In the Tropical Cell, water subducts in the southern ocean, flows equatorward in the lower layer of the western-boundary current, and is entrained back into the upper layer in the open-ocean upwelling regions in the southern ocean. In the Cross-Equatorial Cell, the subducted water crosses the equator near the western boundary, where it is entrained in the regions of intense coastal upwelling in the northern ocean. The strength of the cells is directly related to the assumed magnitude of the subduction rate w(d), but their structure is not sensitive to the particular parameterization of w(d) used.
