volcanic signature shows up much more clearly in the globally averaged time series than in the time series for the tropical belt or for individual hemispheres. Over a wide range of frequencies, much of the large, dynamically induced temperature variability involves compensating warming and cooling in different latitude belts. For example, the correlation between month-to-month temperature changes in the tropics (20°N to 20°S) and extratropics is -0.79 (-0.86 when the 3 months subsequent to the Pinatubo eruption are excluded). The dynamically induced temperature perturbations are almost entirely filtered out by the global averaging, leaving only the smaller radiatively induced perturbations (Yulaeva et al., 1994).
During the period of record 1979 through 1992, the globally averaged MSU-2 and MSU-4 time series are virtually uncorrelated. In the presence of the large ENSO signal in tropospheric temperatures, the cooling signatures associated with the El Chichon and Pinatubo eruptions do not stand out clearly in Figures 5 and 7. As pointed out by Bradley (1988) and Mass and Portman (1989), the Katmai eruption in the Aleutians in June 1912 was followed by a pronounced dip in Northern Hemisphere monthly mean temperatures during the warm season (Figure 3, upper plot). There was an analogous, although somewhat weaker, dip following the eruption of Ksudach (also in the Aleutians) three years earlier. There is no clear signature of the ENSO cycle in the global-mean stratospheric time series, although Randel and Cobb (1994) and Yulaeva and Wallace (1994) have shown evidence of a geographically localized signature.
The ENSO and volcanic signatures are reflections of natural variability in the earth system: The former is generated by processes within the more limited atmosphere-ocean climate system, and the latter is externally forced. The remaining variability in the climate record is predominantly interdecadal in time scale. It is not clear how much of it is due to natural causes and how much is a consequence of anthropogenic influences. Depending on whether aerosols or greenhouse gases predominate, and the strength and polarity of various feedbacks, anthropogenic influences might serve either to warm or to cool the troposphere. Hence, the absence of a pronounced upward trend in tropospheric temperatures as sensed by the MSU-2 instrument during the past decade does not necessarily preclude the possibility that a detectable anthropogenic signal exists. A gradual increase in the optical thickness of the atmosphere is, in fact, suggested by the secular decrease in static stability observed over the course of the past 40 years in the Northern Hemisphere extratropics and over at least the past decade in the Southern Hemisphere extratropics, coupled with the marked increase in nighttime temperatures relative to daytime temperatures at many land stations. Increases in greenhouse gases and aerosols could both be contributing in the same sense to these trends. Such a signature would be consistent with the results of many general-circulation models' simulations of the response to increasing CO2, which show the largest temperature rises at the earth's surface at higher latitudes, in regions subject to frequent low-level temperature inversions.
The sharp rise in Northern Hemisphere surface air temperatures from 1920 to 1940 (Figures 3 and 4) coincided with a northward shift of the Gulfstream (Bjerknes, 1959; Deser and Blackmon, 1995). The subsequent dip during the 1960s and 1970s may be related to a reduction in the salinity of the surface waters of the subpolar North Atlantic that was observed during this period (Dickson et al., 1988; Delworth et al., 1995; Dickson, 1995; Mysak, 1995), which might have tended to suppress the thermohaline circulation,