Difference in thickness of the 1000-500 hPa layer, 1981-1990 minus the period 1951-1980. (a) Wintertime means, (b) annual means. Contour interval 10 m (0.5K). Negative values are dashed. (After Wallace et al., 1993; reprinted with permission of the American Meteorological Society.)
pattern, local wintertime temperatures generally tend to be less coherent with the mean temperature of the Northern Hemisphere than those observed during other seasons. For example, in the EOF analysis of Barnett (1978) cited in the previous section, the leading mode of wintertime surface air temperature is dominated by regional-scale features reminiscent of the PNA pattern, in sharp contrast to the monopolar structure of the leading mode of annual mean temperature. And when the annual mean time series shown in Figure 1 are broken down by season (Table 1), strong positive correlations are observed between the China series and the hemispheric mean series for every season except winter.
Since the low-frequency variability during the warm season is less strongly influenced by circulation anomalies, hemispheric mean temperatures observed during the warm season should be of particular interest from the standpoint of detecting thermodynamically induced hemispheric or global trends. Seasonal mean time series of hemispheric mean surface air temperature presented by Jones and Briffa (1992) are suggestive of a more favorable "signal-to-noise ratio" for the detection of interdecadal variability in the plots for the warm season. This distinction is brought out clearly in Figures 3 and 4, which show time series of monthly mean surface air temperature anomalies averaged over the extratropical Northern and Southern Hemispheres,3 respectively, with the months May to October and November to April plotted separately. These figures, and several of the subsequent figures in this paper, are based on the data set compiled by Jones et al. (1985, 1986a) under the sponsorship of the U.S. Department of Energy. The "DOE" identifier on individual curves also refers to these data. Analogous plots (not shown) based on 3-month means (May-through-July and August-through-October versus November-through-January and February-through-April) also exhibit substantially more scatter during the cold seasons.
Hence, it can be concluded that the recent upward trend in surface air temperature averaged over the Northern Hemisphere is a reflection of a genuine, year-round hemispheric warming. The more dramatic rises in wintertime temperatures over western Canada and Siberia are of considerable interest in their own right, but it seems unlikely that they are the fundamental cause of the hemispheric warming, which shows up just as clearly in the warm-season time series as in the cold-season time series.
The abrupt increase in variance of the Southern Hemisphere time series in 1958 coincides with the establishment of stations in Antarctica.