Several artificial constraints on the model tend to minimize differences among the different resolution cases. Nevertheless, for quantities of interest to global climate studies the simulations show significant changes as resolution increases. These changes generally but not always bring the model into better agreement with observations. Differences are typically more noticeable when comparing the 4ƒ° and 2ƒ° runs than when comparing the 2°ƒ and 1°ƒ runs or the 1°ƒ and 1/2ƒ° runs. A reasonable conclusion to draw for current studies with coupled ocean-atmosphere GCMs is that the ocean grid spacing could be set to about 1°ƒ to accrue the benefits of enhanced resolution without paying an excessively steep price in computer-time cost.
The model's poleward heat transport at 1/2°ƒ grid spacing peaks at about 1 x 10^15 W in the Northern Hemisphere and 0.5 x 10^15 W in the Southern Hemisphere. These values are significantly below observations, a problem typical of ocean GCMs even when they are less constrained than in the present study (e.g., when they are coupled to an interactive atmosphere). This problem is alleviated somewhat in the 1/4° run. In this case, however, the eddies resolved by the model generally act to counter rather than to reinforce the heat transport of the mean flow. Improved heat transport may result less from enhanced resolution than from other changes made in this version of the model, such as more accurate wind forcing. (pdf file)