Explorations of the Annual Mean Heat Budget of the Tropical Indian Ocean. Part I: Studies with an Idealized Model

Annual mean net heat fluxes from ocean general circulation models (OGCMs) are systematically too low in the tropical Indian Ocean, compared to observations. In the models, only some of the geostrophic inflow replacing southward Ekman outflow is colder than the minimum sea surface temperature (MINSST). Observed heat fluxes imply that much more inflow is colder than MINSST. Since inflow below MINSST can only join the surface Ekman transport after diathermal warming, the OGCMs must underestimate diathermal effects. A crude analog of the annual mean Indian Ocean heat budget was generated, using a rectangular box model with a deep “Indo–Pacific” gap at 7°–10°S in its eastern side. Wind stress was zonal and proportional to the Coriolis parameter, so Ekman transport was spatially constant and equaled Sverdrup transport. For three experiments, zonally integrated Ekman transport was steady and southward at 10 Sv (Sv ≡ 106 m3 s−1). In steady state, a 10 Sv “Indonesian Throughflow” fed a northward western boundary current of 10 Sv, which turned eastward along the northern boundary at 10°N to feed the southward Ekman transport. Most diathermal mixing occurred within an intense eddy in the northwest corner. Some of the geostrophic inflow was at temperatures colder than MINSST (found at the northeast corner of the eddy); it must warm to MINSST via diathermal mixing. Northern boundary upwelling exceeded the 10-Sv Ekman transport. The excess warms as it recirculates around the eddy, apparently supplying the heat to warm inflow below MINSST. In an experiment using the “flux-corrected transport” (FCT) scheme, diathermal mixing occurred in the strongly sheared currents around the eddy. However the Richardson number never became low enough to drive strong diathermal mixing, perhaps because (like that of other published models) the present model’s vertical resolution was too coarse. In three experiments, the dominant mixing was caused by horizontal diffusion, spurious convective overturn, and numerical mixing invoked by the FCT scheme, respectively. All three mixing mechanisms are physically suspect; such model problems (if widespread) must be resolved before the mismatch between observed and modeled heat fluxes can be addressed. However, the fact that the density profile at the western boundary must be hydrostatically stable places a lower limit on the area-integrated heat fluxes. Results from the three main experiments—and from many published OGCMs—are quite close to this lower limit.