temperatures at the earth's surface (Wallace et al., 1993). The wintertime variability of the 1,000 to 500 hPa thickness field, a measure of the mean temperature of the lower troposphere, is dominated by a PNA-like pattern (Figure 9, upper panel). The summertime variability is dominated by a relatively featureless ''background field" (Figure 9, lower panel) whose expansion coefficient is highly correlated with the time series of hemispheric mean thickness. Even though the wintertime mode exhibits higher amplitude, the leading EOF of annual mean thickness closely resembles the leading summertime pattern, which explains a higher percentage of the hemispherically integrated variance than its wintertime counterpart, and is evidently present year round. For fluctuations with periods of 5 years or longer, variations in hemispheric mean thickness account for half the hemispherically integrated variance of the local thickness time series.
There is some evidence of systematic differences in the long-term trends in surface air temperature versus temperatures aloft. Figure 10 compares hemispherically averaged temperature time series based on (a) surface air temperature over land, (b) mean tropospheric temperature as sensed by Channel 2 of the MSU, and (c) mean virtual temperature of the 1,000 to 500 hPa layer based on the operational analyses of the National Meteorological Center (NMC), all for the area poleward of 20°N. All three time series have been smoothed by taking 3-month running means, and the DOE climatology has been adjusted to eliminate the annual march in the 1979 to 1990 period of record. Neither of the upper-air time series exhibits as pronounced a warming trend as that evident in the time series of surface air temperature, nor do the hemispheric mean time series of Angell (1988) or Oort and Liu (1993), which are based on radiosonde data. The corresponding MSU-2 time series of