A global ocean circulation model is constructed in which the Pacific, Atlantic, Indian, and Southern oceans are separately resolved, each being represented by zonally averaged equations expressing conservation of momentum, mass, heat, and salt. Results are presented and compared with relevant zonally averaged observations. The sensitivity of the stream function and the meridional fluxes of heat and water are examined as functions of the horizontal and vertical diffusion coefficients, and as functions of a closure parameter introduced in averaging the equations. The results are sensitive to changes in the vertical diffusion coefficient and the closure parameter. The sensitivities to changes in the vertical diffusion coefficient are similar to those of a three-dimensional ocean general circulation model. Results are relatively insensitive to the value of the horizontal diffusion coefficient provided it is of the order of 103 m2 s-1 or smaller.

However, for larger values, a northward heat flux throughout the Atlantic basin, as observed, cannot be obtained. Wind stress significantly improve the comparison with observational estimates of the meridional heat and water fluxes, particularly near the equator, where there are large flux divergences associated with the Ekman transport. Results are further improved when horizontal diffusion is modified to include a contribution proportional to the local current speed. The effects of the Mediterranean and Red seas are examined and shown to be important in redistributing salt vertically. The model is then generalized to allow for depth-integrated flow through the Bering Strait and the Indonesian passages. Both effects improve the comparison with observational estimates, but neither effect appears to be crucial in determining the global circulation or water mass properties in this model. The inclusion of a brotropic flow introduces the necessity to specify a reference temperature in order to calculate the heat flux directly from velocity-temperature correlations. Simply neglecting the barotropic flow may contribute to large discrepancies between flux estimates based on integrated surface fluxes and ''direct'' correlation methods. ¿ American Geophysical Union 1992