to force the temperature (Gill and Niiler, 1973; DC3): SST is coolest in late winter following the highest ocean-to-atmosphere fluxes. For the flux-forcing-SST case to be valid, the SST anomaly must play a negligible role in determining the flux, so that the flux is driven by atmospheric conditions—i.e., w, Ta, and q variations dominate the variations of SST. Indirectly, this scenario has been investigated by relating SST anomalies to the large-scale atmospheric circulation (Namias, 1972; Davis, 1976; Wallace et al., 1990). While all of these studies established a definite connection between the monthly or seasonal mean circulation and the SST anomalies, they could only infer the role of anomalous heat flux in forcing SST anomalies. When atmospheric fields are related to contemporaneous monthly SST anomalies, it is difficult to determine whether SST is forcing the atmosphere, both fields are being forced by some other agent, or the atmosphere is forcing the SST.
In testing the phase relationship between anomalies of the estimated flux and SST, the dominant linkage exhibited in the extratropics is that the latent and sensible flux anomalies force the SST anomalies, rather than SST anomalies' forcing the flux anomalies (DC3). Historical time series of flux data provide the basis for directly testing the forcing. Correlations between bulk formulae fluxes and anomalous temperature variations have been documented over the North Pacific by Clark et al. (1974) and Frankignoul and Reynolds (1983), and in the western tropical Pacific by Meyers et al. (1986). As was reported in DC3, the latent and sensible heat-flux anomalies proved to strongly affect monthly changes in SST anomalies over a large portion of the world oceans, in particular the North Atlantic and the North Pacific. A simple thermodynamic model was adopted to relate the flux anomalies to the tendency (time rate of change) of the SST anomalies. More discussion of the full mixed-layer thermodynamic equation is given below, and a comprehensive treatment is provided by Frankignoul (1985). Note that the temperature equation predicts the tendency of the SST anomaly , not the anomaly itself; the tendency is represented here by its finite-difference forms, . The flux parameterizations do not contain knowledge of the SST tendency, so the relationship of fluxes to SST-tendency anomalies is an independent test of the influence of the fluxes.
Having a higher-frequency character than the anomaly itself, has nearly as many independent samples of as there are months; four decades of records contain approximately 120 independent December-January-February samples.
Spatial Distribution of Flux versus SST Tendency Anomalies
The geographical distribution of the correlations between Q'l+s and observations for winter months is examined in Figure 2. The 0.3 level is used as a threshold of statistical significance. Meaningful correlations on this map are almost everywhere positive, as they were for zonal averages of the two fields (DC3). Strongest correlations (= 0.5) are found mostly between about 25°N and 40°N within the anticyclonic subtropical gyres of the two oceans. For the North Pacific, strongest correlations are east of 180° and extend to the California Current. Though significant in many locations, correlations are weaker in the western North Pacific and from the tropics to 30°S. In the central North Atlantic, the flux and anomalies have strong correlations in the central subtropical gyre as well as in the high-variance western North Atlantic region. Seasonally, most of the regions have strongest correlations in winter.
Near the equator between 5°N and 5°S in all three basins, Q'l+s and are not well correlated. Correlations between zonal average flux and SST anomalies (DC3; not shown) indicate that in the tropics, the flux and SST anomalies (not the tendencies) tend to be negatively correlated. This suggests that the flux is driven by SST, presumably because equatorial SST anomalies are governed more by internal ocean processes than by the air-sea heat exchange.
versus Fluxes during Strong Atmospheric-Circulation Modes
The organization of the latent-plus-sensible flux anomalies (Q'l+s) by the anomalous atmospheric circulation can be used to test the consistency of large-scale links between the flux and SST tendency anomalies . If the SST tendency anomalies are caused by the fluxes, they should have corresponding patterns. In DC3, the organization by the circulation was exploited by using the dominant SLP EOF modes as an index to compare flux-anomaly and corresponding SST-anomaly tendency patterns.
To identify strong winter-circulation months, extreme positive and negative EOF amplitudes were chosen for each SLP EOF (DC2 and DC3). Using this criteria, several (10-30) months of each of the extreme EOF modes (positive and negative amplitudes) were selected. Composites of Q'l+s and were formed by averaging the fields during the respective extreme months. For brevity, the composites were expressed as the difference between averages of positive (strong) and negative (weak) phase months of the SLP EOFs.
In both northern oceans, the patterns of are well aligned with the flux anomaly patterns. In the North Pacific, composite differences and Q'l+s differences corresponding to positive-minus-negative extremes of the PNA pattern are shown in Figure 3. Remarkably, the flux and signatures of the PNA are marked out more than halfway around the hemisphere. Major centers of are closely matched to those of Q'l+s, which confirms that