done in 1987 by Gordon Fleming at the Naval Postgraduate School. He was supervised by John Walsh, who is here. One of the interesting calculations in the thesis is the cross-correlation analysis of the COADS SST in the Labrador-Greenland-Barents Sea region and the position of the ice edge. Fleming found the two to be inversely correlated, not unlike Dr. Deser's findings for the Davis Strait. However, I believe Fleming was able to show that in some regions the SST actually leads the ice edge and therefore could be used as a predictor.
A recent paper by me and Manak (Atmosphere-Ocean, 1989) studying 32 years of sea-ice concentration data clearly showed the existence of a decadal oscillation of the sea-ice cover in the Labrador Sea and in the Barents Sea. The signal in the former region is the same as that shown in Dr. Deser's Figure 10a. In the Greenland Sea, variations in the sea-ice extent seemed to have a much longer time scale, because there the sea ice was dominated by a big ocean event during the 1960s and early 1970s that we have now associated with the Great Salinity Anomaly.
In that paper Manak and I suggested something that might set up an oscillation of 10 to 12 years, namely, the circuit time of the subpolar gyre in the northern North Atlantic. While 12 to 14 years was the circuit time for the GSA, there was nothing to reinforce it within the subpolar gyre region. Thus we abandoned the idea that there might be self-sustained oscillations of this period in this region.
Other papers by Mysak et al. (Climate Dynamics, 1990) and Marsden et al. (JGR, 1991) showed that salinity and sea-ice extent are highly correlated in the Greenland and Labrador seas, and that there is advection of both ice and salinity anomalies from the Greenland Sea into the Labrador Sea, with an advection time of a few years.
In terms of modeling studies, I first want to mention M. Ikeda (Atmosphere-Ocean, 1990), who developed an ocean model for the decadal oscillations in the Barents-Greenland seas. He emphasized that there were feedbacks between the atmosphere and the ocean in the Barents Sea. In 1991, Weaver and Sarachik published in Atmosphere-Ocean the first model evidence of a decadal oscillation in the thermohaline circulation in the North Atlantic. The physical location of the oscillation is somewhat further south than the Greenland-Labrador Sea, because it is centered between the subtropical and subpolar gyres. In more recent modeling studies with an improved geometry, Weaver and his coworkers have found that the decadal oscillation now extends to 20 years.
There has also been the more recent modeling work by Delworth and his colleagues at GFDL, who find a somewhat longer-time-scale fluctuation in the thermohaline circulation in a coupled atmosphere-ocean model. The period is 30 to 50 years, and the oscillation is found to have a dipole-like structure in the subsurface waters in the central North Atlantic. As for recent data studies looking at decade-to-century time scales, a paper in Climatic Change (Stocker and Mysak, 1992) presents a spectrum of the Koch ice index. (The Koch ice index is the number of weeks per year that ice affects the coast of Iceland.) A spectrum of 360 years of this index shows peaks at 90, 27, and 14 years.
Finally, a well-known time series of SST anomalies in the North Atlantic (45°N to 55ºN) is presented by Bryan and Stouffer in J. Marine Systems. This time series also shows interdecadal variability in the northwest Atlantic. In addition, in the Centre for Climate and Global Change Research's Report No. 91-1 (1991), I suggested that heavy ice conditions in the Greenland Sea lead cold SST anomalies in the northwest Atlantic (i.e., east of Newfoundland) by 4 or 5 years. Similarly, warm SST anomalies there were preceded by generally light ice conditions in the Greenland Sea during the previous 5 years.
Let me now refer you to two figures, Figures 10 and 12a in Stocker and Mysak. The first is a spectrum of the Koch ice index as defined above. During heavy ice conditions in the Greenland Sea, you have a high index; during mild conditions, a low index. This spectrum clearly shows a significant peak at 27 years, which indicates an interdecadal climate cycle in the northern North Atlantic. The second, a spectrum of the famous Central England temperature, shows peaks at around 15 and 7 years and an interdecadal peak at 24 years.