FIGURE 2

(a) EOF 1 of North Atlantic surface air temperature anomalies based on un-normalized winter (November-March) means, 1900 to 1989 (bold contours). This mode accounts for 21 percent of the variance. Also shown is the climatological air temperature distribution (thin contours). (b) Time series of EOF 1 (dashed curve) and smoothed with a 5-point binomial filter (solid curve). (c) As in (a) but for EOF 2. This mode accounts for 17 percent of the variance. (d) As in (b) but for EOF 2.

accounts for 12 percent of the variance, exhibits a center of action east of Newfoundland at the boundary between the subtropical and subpolar ocean gyres. A weaker center of opposite polarity is located off the southeastern United States. The time series of E2 SST (Figure 1d) exhibits quasi-biennial and quasi-decadal fluctuations, as well as longer-term trends.

The first two EOFs of air temperature, shown in Figure 2, are similar to those of SST, but the order of the modes is reversed. E1 of air temperature (Figure 2a), which accounts for 21 percent of the variance, exhibits centers of action off Newfoundland and the southeastern United States; the correlation between the time series of E1 of air temperature and E2 of SST is 0.76.3 E2 of air temperature (Figure 2c), which accounts for 17 percent of the variance, exhibits maximum amplitude along the Gulf Stream and near western Europe. The time series of E2 (Figure 2d) is dominated by a long warming trend from the 1920s to the 1940s, followed by a cooling trend during the 1950s and 1960s. These trends follow those of global air temperature (Folland et al., 1984). Shorter-period fluctuations are also apparent.

The power spectra of E2 of SST and E1 of air temperature, based on the lag-correlation method, are shown in Figure 3. Both time series exhibit spectral peaks at approximately 10 to 15 years and approximately 2 to 2.5 years, consistent with our visual impression of the time series. E2 of SST also exhibits enhanced power at the lowest spectral estimate. The decadal and biennial peaks in El air temperature and the biennial peak in E2 SST are statistically significant at the 95 percent level (a priori) above the red-noise background spectra. The power spectra of E2 of air tempera-

3  

All of the correlation coefficients cited in this study are statistically significant at the 99 percent level or higher, based on a one-tailed Student's t-test and taking into account the autocorrelation in the time series according to Leith (1973).



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement