Climate variability on decadal and longer time scales is a subject of increasing interest and relevance. Concern over anthropogenic effects on global climate provides a strong impetus to describe and understand the natural modes of variability of the climate system. In this study, we focus on the low-frequency climate fluctuations over the North Atlantic Ocean since the turn of the century.

The North Atlantic is a region of particular importance to the global climate system. The formation of bottom water at high latitudes of the North Atlantic drives a trans-equatorial thermohaline circulation. Changes in the rate of deep-water production south of Iceland (and hence in the strength of the thermohaline circulation) can have a profound effect on global climate, as can be inferred from paleoclimate data (Broecker et al., 1986; Lehman and Keigwin, 1992) and as is shown by the modeling studies of Rind et al. (1986) and Manabe and Stouffer (1988). Recent ocean modeling experiments indicate that self-sustained oscillations of the thermohaline circulation can occur on decadal and longer time scales (Weaver and Sarachik, 1991).

Low-frequency climate fluctuations in the North Atlantic since World War II have been investigated in several studies. Levitus (1989) used subsurface measurements of temperature and salinity to document a major shift in the ocean circulation in the North Atlantic between the late 1950s and the early 1970s. Knox et al. (1988) showed that the Northern Hemisphere atmospheric circulation experienced an abrupt transition during the early 1960s (see also Flohn, 1986; Shabbar et al., 1988). One of the best-documented examples of low-frequency variability at high latitudes of the North Atlantic is the Great Salinity Anomaly, a freshwater mass that was observed to travel around the subpolar gyre from 1968 to 1982 (Dickson et al., 1988). The freshening of the surface waters was sufficient to halt temporarily deep-water formation in the Labrador Sea (Lazier, 1980). Mysak et al. (1990) discuss the relationship between the Great Salinity Anomaly and interdecadal oscillations in the Arctic climate system.

Studies dealing with the longer (100-year) marine records have tended to emphasize globally averaged surface-temperature variations (cf. Paltridge and Woodruff, 1981; Jones and Kelly, 1983; Barnett, 1984; Folland et al., 1984; Jones et al., 1986c). These studies find that the dominant signal in worldwide temperatures is a warming trend from about 1920 to 1940, with the largest amplitudes at high latitudes.

Bjerknes (1959, 1961, 1962, 1964) investigated air-sea interaction in the North Atlantic on time scales ranging from interannual to interdecadal. Using data from 1890 to 1940, Bjerknes provided compelling evidence that interannual fluctuations in SST are largely governed by wind-induced changes in latent and sensible heat fluxes at the sea surface. However, the long warming trend during the first quarter of this century appears to be linked to a change in the ocean circulation rather than to air-sea energy fluxes (Bjerknes, 1959). Similar conclusions regarding interannual and interdecadal SST variations are reached by Kushnir (1994).

The purpose of this study is to describe the variability of the winter climate over the North Atlantic Ocean, using 90 years of surface marine data. The North Atlantic is the only ocean basin with data coverage that is sufficiently dense for a regional study of climate change since 1900. We focus on the winter season because (1) the atmosphere-ocean coupling is most vigorous and (2) the SST variations reflect significant heat-content anomalies in the upper ocean. We are particularly interested in whether the climate system exhibits preferred time scales of variability, and whether the atmosphere-ocean relationships change with spectral frequency. Our results are based on an objective (empirical orthogonal function) analysis of four components of the climate system: SST, air temperature, wind, and sea level pressure. We will show that the surface climate over the North Atlantic exhibits coherent decadal fluctuations that resemble the variations on interannual time scales. The decadal fluctuations appear to be closely related to sea-ice extent in the Labrador Sea. We will also show that the Gulf Stream was apparently involved in the long-term warming trend during the 1920s to 1930s.

The data and methods are described; the two leading modes of surface-temperature variability during the period 1900 to 1989 and their relation to the atmospheric circulation are documented; and the results are discussed.


The surface wind, sea level pressure, SST, and surface air temperature data used in this study are from the Comprehensive Ocean-Atmosphere Data Set (COADS), an extensive compilation of weather observations from merchant ships (Woodruff et al., 1987). We have used data gridded by 4° latitude/longitude squares for the period 1900 to 1989. The COADS data are uncorrected for changes in instrumentation, observing practice, ship type, etc. Spurious trends resulting from these changes have been reported in surface wind speed and SST over the tropical oceans (Ramage, 1987; Wright, 1988; Cardone et al., 1990). Our approach to the issue of whether the routine ship-based measurements are sufficiently accurate and homogeneous to allow the detection of real climate signals is to demonstrate physical consistency among independent variables: wind patterns are compared to pressure distributions, and SST fields to surface air temperatures.


The monthly mean data were converted to anomalies by subtracting long-term monthly mean values for the period 1900 to 1989. These monthly anomalies were then averaged

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement