Decadal Variability of the Tropical Atlantic Ocean Surface Temperature in Shipboard Measurements and in a Global Ocean-Atmosphere Model

Abstract

Numerous analyses of relatively short (25?30 years in length) time series of the observed surface temperature of the tropical Atlantic Ocean have indicated the possible existence of decadal timescale variability. It was decided to search for such variability in 100-yr time series of sea surface temperature (SST) measured aboard ships and available in the recently published Global Ocean Surface Temperature Atlas (GOSTA). Fourier and singular spectrum analyses of the GOSTA SST time series averaged over 11 subregions, each approximately 1?106 km2 in area, show that pronounced quasi-oscillatory decadal (?8?20 yr) and multidecadal (?30?40 yr) timescale variability exists in the GOSTA dataset over the tropical Atlantic. Motivated by the above results, SST variability was investigated in a 200-yr integration of a global model of the coupled oceanic and atmospheric general circulations developed at the Geophysical Fluid Dynamics Laboratory (GFDL). The second 100 yr of SST in the coupled model's tropical Atlantic region were analyzed with a variety of techniques. Analyses of SST time series, averaged over approximately the same subregions as the GOSTA time series, showed that the GFDL SST anomalies also undergo pronounced quasi-oscillatory decadal and multidecadal variability but at somewhat shorter timescales than the GOSTA SST anomalies. Further analyses of the horizontal structures of the decadal timescale variability in the GFDL coupled model showed the existence of two types of variability in general agreement with results of the GOSTA SST time series analyses. One type, characterized by timescales between 8 and 11 yr, has high spatial coherence within each hemisphere but not between the two hemispheres of the tropical Atlantic. A second type, characterized by timescales between 12 and 20 yr, has high spatial coherence between the two hemispheres. The second type of variability is considerably weaker than the first. As in the GOSTA time series, the multidecadal variability in the GFDL SST time series has approximately opposite phases between the tropical North and South Atlantic Oceans. Empirical orthogonal function analyses of the tropical Atlantic SST anomalies revealed a north-south bipolar pattern as the dominant pattern of decadal variability. It is suggested that the bipolar pattern can he interpreted as decadal variability of the interhemispheric gradient of SST anomalies. The decadal and multidecadal timescale variability of the tropical Atlantic SST, both in the actual and in the GFDL model, stands out significantly above the background ?red noise? and is coherent within each of the time series, suggesting that specific sets of processes may be responsible for the choice of the decadal and multidecadal timescales. Finally, it must be emphasized that the GFDL coupled ocean-atmosphere model generates the decadal and multidecadal timescale variability without any externally applied force, solar or lunar, at those timescales.