The meridional structure of Fig. 3 is so similar to the combined structure of two leading EOFs. The amplitudes of regressed SST at maxima and minima exceed 0.6 ℃ in the tropics, 1.0 ℃ south of Greenland. At the same time, the amplitudes of correlations (not shown) exceed 0.6 at maxima and minima, which are above the 95% statistical significant level under the reliable degree of freedom.
We show simultaneous regression patterns of the sea level pressure (SLP) and surface wind fields onto the dipole index (Fig. 4a). Monthly 5 degree latitude-longitude resolution SLP, zonal and meridional wind speed fields for the Atlantic with 1950 through 1990 are computed from individual ship observations contained in the comprehensive ocean atmosphere datasets-compressed marine reports (Tanimoto et al 1997). In SLP regressions, a large positive and negative center of action appears in the southern and northern subtropics, which overlays the colder and warmer SST anomalies associated with the tropical dipole. In the North Atlantic, the positive phase of the NAO appears in the SLP field when the tropical dipole index is positive. In fact, the decadal component of the NAO index shows good agreement with that of the tropical dipole index in the present study (Fig. 4b). SST, SLP and wind patterns are all somewhat noisy, not satisfying the geostrophic relation. Further data collection is desired.
3. Dominant modes in the tropics
To estimate a percentage ratio of the tropical dipole mode to total variance of tropical Atlantic SST variability, we performed an EOF anaiysis for the domain within 20｡? and 20｡? (Fig. 5). The first EOF is characterized by an equatorial symmetric monopole structure and the second mode, by an equatorial anti-symmetric dipole structure. These two modes explain 40.2% and 32.7%, respectively. Such comparable variances indicate that monopole and dipole modes co-exist in the tropical Atlantic, which is consistent with the growth rate curve of dynamic coupled model (Xie et al, 1998, also in this volume). For the Atlantic of a zonal size, monopole and dipole modes have comparable growth rate, admitting them prosperity in the tropics.
The dispersion relation of this coupled model aiso shows that the frequencies of the dipole and monopole modes are well separated, which is consistent with observational results below. We made another scatter plots from bandpass (8-16 years) and highpass (＜ 5 years) filtered anomalies on each grid boxes with the last two digits of the calendar year superimposed on each data point (Fig. 6). On decadal variations, plots after late 1960s tend to have an out-of-phase relationship. In one case, the tilted track shows a clockwise rotation whose cycle is about 10 years. In the other case, it shows a counter-clockwise rotation, but a comparable cycle. Therefore, we cannot estimate a phase difference of SST variations between the north and the south of the equator. This phase modulation might cause decorrelation of the fourier components of SST anomalies between northern and southern tropics (see Mehta 1998).
In contrast, interannual SST variability is dominated by anomalies of like-sign across the entire basin. The longer first (shorter second) axis crosses from the first (second) quadrant toward the third (forth) one.
