On the time mean and seasonal cycles, we find that most simulations overestimate the time mean wind stress outside of the tropics. Along the equator the seasonal cycle in surface wind stress is somewhat exaggerated. We find weak wind convergence and underestimation of precipitation between the equator and 10ĒŠS in the western and central basin as well as the northern tropics.
At longer interannual and decadal periods the tropical Atlantic appears to support a variety of modes including an Atlantic Nino and an interhemispheric mode. Here we focus on the connection between the atmospheric response and local pattern of SST using a rotated principal component analysis. Examination of the atmospheric patterns associated with the Atlantic Nino reveals that the response to the Atlantic Nino is similar among simulations and COADS. Most show a strengthening of the southeast trade winds in the southeast and the northeast trade winds in the northeast, as well as an equatorward shift and strengthening of precipitation.
Our examination of the interhemispheric pattern shows greater diversity. In this pattern a northward shift of the trade wind systems and the ITCZ develops in response to a positive anomalous interhemispheric SST gradient, and results in drying of the southwest tropics and enhanced precipitation in the northwest. Only two simulations have precipitation and cross-equatorial wind anomalies as large as observed.
The atmospheric pattern includes changes in latent heat flux, which are believed to be critical in controlling the development of anomalous SST through a positive feedback on boundary layer winds. We find substantial differences in the sign and distribution of the feedback in the simulations. In an earlier study Chang et al. (2000) pointed out the presence of a zone of positive feedback in the western tropics and negative feedback in the east. We find similar behavior in four simulations. In the other two simulations, a broad region of negative feedback is evident, with no corresponding region of positive feedback, suggesting in particular that coupled models using these AGCMs will be unable to reproduce the dynamics of the interhemispheric mode. The cause of these differences can be traced to the reduced size of wind-driven latent heat flux anomalies and, in several simulations, the greater importance of humidity-driven latent heat flux anomalies.