The difference in meridional scale between the Tpac and Tatl regressions is better illustrated in Fig. 1c that contrasts their zonal averages. The Pacific sees a distinct equatorial peak while the Atlantic SST anomalies have a flat structure within 20｡?atitudes.
In addition to the above latitudinally symmetric monopole mode, the Atlantrc displays an antisymmetric mode of SST variability, a unique feature that further distinguishes itself from the Pacific. This is so-called dipole mode characterized by out-of-phase relation between the northern and southern tropics. In a companion paper in this volume (see also Xie et al. 1998), Tanimoto presents the strongest ever evidence in support of the existence of such a dipole mode in the tropical Atlantic. Performing separate EOF analysis in the North, and South Atlantic Oceans, he shows that their respective leading EOFs constitute a dipolar configuration in the Tropics and their time series are highly correlated, despite the statistical independence of the datasets for the two oceans.
The purpose of this paper is to provide a physical basis for the empirical monopole-dipole decomposition. We will show they are indeed two independent physical modes of the tropical coupled ocean-atmosphere system. Our solutions to the coupled model display differences in SST variability as outlined above, in their meridional scales and dipole's preference of the Atlantic.
2. Theoretical backgrounds
The tropical ocean-atmosphere system contains two basic modes of interaction: the Bjerknes (1969) and wind-evaporation-SST (WES; Xie and Philander 1994) feedbacks. The Bjerknes feedback involves ocean-atmosphere interaction in the zonal direction within the equatorial upwelling zone (Fig. 2a). A SST increase in the east induces anomalous westerly winds, leading to a thermocline depression and weakened upwelling. Their combined effects cause further warming in the east. The WES feedback, on the other hand, is an interaction in the meridional direction (Fig. 2b). With a positive SST anomaly to the north of the equator and a negative one to the south, we have southerly winds. The Coliolis force will turn the southerly winds to the east to the north of the equator. In the Southern Hemisphere where the Coriolis paramelter is negative, the Coriolis force produces easterly wind anomalies Because in the basic state we have easterly trade winds, so we got weaker trade winds to the north and stronger winds to the south of the equator. This wind speed difference will further enhance the initial SST difference between the Northern and Southern Hemispheres This is the same feedback mechanism that operates to keep the ITCZ to the Northern Hemisphere. The Bjerknes feedback dominates the equator, while the WES feedback should become important off the equator where the general Ekman downwelling prevents thermocline variability from affecting SST.
Previous theoretical studies have been carried out along two separate lines: each focusing on either the Bjerknes (see McCreary and Anderson 1991 and Neelin et al. 1998 for reviews) or WES (Xie 1996) feedback. Here we construct a unified model that includes the both. This is achieved by adding a wind speed-dependent latent heat flux term to the linearized Zebiak-Cane ocean model. The atmosphere is a dynamic Matsuno-Gill model forced by SST anomalies. In this coupled model, both the temporal and spatial characters of a mode is determined by internal dynamics. So there is no a priori restriction on how a coupled mode should organize itself.
