Estimating open boundary conditions in a GCM

Open boundary conditions are implemented to the west of the Indonesian Archipelago and near 30°S (Fig. 1). The approach simultaneously optimizes the initial conditions of the hydrographic fields, surface fluxes, and the open boundary conditions (temperature, salinity, and horizontal velocities). The model configuration is similar to Lee and Marotzke (1997)(LM97).

Wind-driven surface currents

The estimated horizontal velocity is plotted in Fig. 1 for level 1 (12.5~m depth). The wind-driven surface currents are readily identified, like the westward South Equatorial Current between 10°S and 20°S, the eastward Equatorial Counter Current near the equator, the Somali Current and the anticyclonic Arabian Sea gyre. Compared to the closed-domain results of LM97, the estimated velocity field shows a marked improvement near the southern boundary, with a reasonably strong (for a model of this resolution) Agulhas current leaving the model domain. The Indonesian throughflow (ITF) is estimated as 2.7 Sv westward, which is on the low end of the range of previous estimates.

Figure 1
Figure 1


Meridional overturning

The zonally integrated meridional overturning stream function is plotted in
Fig. 2. A dominant counterclockwise cell is present in the upper 700 m. Its magnitude is 16 Sv and the maximum is located at 12°S and about 50 m depth. The northward flowing water upwells continuously north of 12°S, then returns to the wind-driven surface Ekman layer. We have confirmed that this overturning cell is indeed wind-driven, by separately turning off wind and surface buoyancy forcing of the optimized solution (following LM97). The shallow overturning carries equatorial surface water of high temperature southward and dominates the meridional heat transport; it is similar both in magnitude and depth with that estimated by LM97, who used a closed basin. Hence, we conclude that it was not the presence of closed boundaries that led to the differences between LM97 (weak deep inflow) and Toole and Warren (1993)(large deep inflow).

Figure 2
Figure 2


Open boundaries vs sponge layers

An optimization is carried out in which the southern and eastern boundaries are closed, and sponge layers are employed. The meridional overturning (
Fig. 3) shows a very strong (16 Sv) clockwise overturning cell in the sponge near the Indonesian Archipelago. A similar problem is seen in the southern sponge layer, and we attribute these overly large vertical velocities to the numerical formulation of the model, most likely the choice of a ``C'' grid. Our standard experiment (Fig. 2) shows that the solution with open boundaries behaves much better numerically, which is yet another argument for eliminating the closed-wall and sponge boundary conditions.

Figure 3
Figure 3


Freshwater transport

The divergent model freshwater transport (blue line) and the surface flux (green line) are reasonably well balanced between the northern sponge layer and the equator, as evidenced by the flat residual curve (red line) in
Fig. 4. Freshwater transport is weak in the northern Indian Ocean, reflecting the large differences in P-E between the evaporative Arabian Sea and the large precipitation over the Bay of Bengal. Between the equator and about $10^\circ$S, a southward freshwater transport reaches 0.29 Sv, which is reduced to about zero by the strong evaporation over the southern subtropics.

The cyan dotted line represents the freshening of the mean mass flux associated with the Indonesian throughflow (ITF), and the magenta dotted line is the ITF freshwater transport relative to the ITF section mean salinity. Both terms are smaller than the residual.
Hence, in our model, the ITF plays a negligible role in the freshwater budget, counter to widespread expectation but consistent with Macdonald and Wunsch (1996) who found their global linear inverse model to be relatively insensitive to widely varying assumptions about ITF strength.

Figure 4
Figure 4

Figure 4 Back to the index