ENSO effect on the northern tropical Atlantic

Observational studies (e.g., Enfield and Mayer, 1997) and model simulations (e.g., Huang et. al., 2002) showed that tropical Atlantic temperature variability is correlated with Pacific El Niño-Southern Oscillation (ENSO) variability in several regions. The major region affected is the North Atlantic area of NE trades west of 40W along 10N-20N and extending into the Caribbean. There, about 50-80% of the anomalous temperature variability is associated with the Pacific ENSO, with Atlantic warming occurring 4-5 months after the mature phases of Pacific warm events, i.e., in spring and even early summer (Enfield and Mayer, 1997).

To investigate temporal changes in this lagged ENSO effect on the tropical Atlantic, STARS was applied to two time series: 1) December sea-surface temperature in the Niño 3.4 region (N3.4) for the period 1948-2013, and 2) June surface air temperature in the eastern part (15-20N, 60-65W) of the Caribbean basin (EC), 1949-2014. Both series are anomalies from the 1981-2000 base period normalized by their standard deviations. The first step was to check for change points in the mean level of fluctuations of these time series and remove stepwise trends if necessary. As shown in Fig. 1, two shifts were detected in N3.4, one in 1977 (up), and the other in 1999 (down), both statistically significant at p = 0.1. The 1977 shift is part of the Great Pacific Climate Shift, associated with a warm phase of the Pacific Decadal Oscillation (PDO). The changes in the late 1970s, however, were not limited to the Pacific Basin; they had a truly global scope. For example, as shown in Fig. 2, the 1977 shift is a dominant feature of the time series of 700-hPa height normalized anomalies (z700) for the entire tropical belt between 25N and 25S, with the p-value reaching 5*10^-12. Note the lack of auto-correlation in z700 (AR1 = 0.07), just a powerful step in the mean level.

The 1999 shift in N3.4, which marks the beginning of a relatively cool background temperature in the equatorial Pacific, is also associated with the cool phase in the PDO. It does not reveal itself, however, in the z700 series for the Tropics (Fig. 2). When the stepwise trends are removed from both N3.4 and z700, the correlation coefficient between their residuals is 0.8.

Two change points were also detected in the eastern Caribbean (Fig. 3), one in 1979 (up), and the other in 2005 (up again). The timing, magnitude and direction of the first one are about the same as for its counterpart in N3.4. The second shift, however, reflecting a positive phase in the Atlantic Multidecadal Oscillation (AMO), is in the direction, opposite to that in N3.4. Therefore, the temperature difference between N3.4 and EC has substantially increased in recent years.

After the stepwise trends were removed from both N3.4 and EC, the further analysis of the residuals reveals that there is more than just the temperature difference that has changed in recent years. Figure 4 shows a sharp decline in the correlation coefficient between the two series from 0.62 during 1949-1998 (years of EC) to 0.19 during 2000-2014. This shift is statistically significant at p = 0.09.

This decoupling between the Pacific and Atlantic may have had significant implications for the global climate system, although more research is needed. One thing is certain: it has substantially diminished the boreal spring temperature predictability for the tropical Atlantic.