results were obtained by Parker and Folland (1988) from historical sea level pressure analyses. The SST (Figure 7b) and surface air temperature (Figure 7c) anomaly patterns are grossly similar: Both exhibit maximum positive anomalies (about 1°) along the Gulf Stream east of Cape Hatteras, and positive anomalies of smaller amplitude in the eastern ocean. There is some indication of a poleward amplification of the air-temperature anomalies, but the data are sparse north of about 55°. (Stations on Greenland, Iceland, and Scandinavia experienced a warming of approximately 2° during the 1920s and 1930s (Jones and Kelly, 1984).) Although the SST (and to a lesser extent the air-temperature) records suffer from discontinuities around World War II (cf. Jones et al., 1986c), the similarity between the air and sea temperature changes lends credence to the patterns in Figure 7. The ''corrected" SST data set of Bottomley et al. (1990) yields a similar pattern, although the warming along the Gulf Stream does not extend as far eastward as in the COADS (not shown).5
The wind/temperature relationships are different from those associated with the dipole pattern discussed above. While there is some evidence of wind anomalies overlying SST anomalies in the eastern portion of the basin, the largest surface-temperature anomalies occur in the west, upstream of the atmospheric-circulation anomalies. Whether the temperature anomalies along the Gulf Stream are connected with changes in the strength or position of the current is unknown. However, it is worth noting that a northward shift of the Gulf Stream of only 50 km would be sufficient to produce the observed warming. An alternative explanation for the warming along the Gulf Stream is a reduction in frequency or intensity of cold air outbreaks. This would occur if the temperatures over eastern North America were elevated or the offshore winds were reduced.
To examine the robustness of the patterns and to test whether the patterns are artifacts of data discontinuities around World War II, we computed the linear least-squares trends during the period 1917 to 1939. The data were first smoothed in time with a 3-point binomial filter to enhance the low-frequency signals. The surface wind and sea level pressure fields (Figure 8a) exhibit a cyclonic circulation trend in the central North Atlantic, similar to the results from the epoch analysis. Over the western North Atlantic, the trend analysis shows easterly wind anomalies, which are consistent with the pressure field, whereas the epoch analysis shows weak westerly anomalies. The trends in SST (Figure 8b) and air temperature (Figure 8c) are similar to the results from epoch analysis; the largest warming occurs
along the Gulf Stream, over the eastern subtropical Atlantic, and near Iceland. The warming along the Gulf Stream is not as continuous, and the warming in the eastern Atlantic not as pronounced, in the trend analysis as in the epoch analysis. As in the epoch analysis, the largest SST trends along the Gulf Stream occur remotely from the wind trends.
The time series of air temperature along the Gulf Stream [44°-36°N, 76°-42°W] and sea level pressure in the region bounded by [46°-34°N, 48°-20°W] and at Ponta Delgada, Azores [40°N, 28°W] are shown in Figure 9. Surface warming along the Gulf Stream occurred during the period 1920 to 1950 (amplitude about 2.5°C). Note that the temperatures
Geographically and seasonally varying corrections are incorporated in the Bottomley et al. (1990) SST data set. The corrections are based on a simple model of air-sea heat transfer from an uninsulated bucket and are designed to minimize the difference in the annual cycle of SST before and after 1941.