73
The linear regression regression calculation that yielded the temperature pattern Fig. 15 and the fitted expansion coefficient time series in Fig.
16 was repeated, using the time series of global-mean SST in place of TL
and the global SST field in place of the extratropical Northern Hemisphere
surface air temperature field. The resulting global SST pattern whose amplitude time series is most highly correlated with the time series of globalmean SST (Fig. 30) exhibits an ENSO-like tropical signature reminiscent
of the bottom panel of Fig. 20. Its expansion coefficient time series (Fig.
31) traces out all the major warm and cold episodes of the ENSO cycle,
as represented in Fig. 22, as well as the interdecadal-scale features noted
above. It also exhibits a shift in the opposite direction (i.e., toward colder
equatorial Pacific SST) in 1942. Apart from a few brief positive excursions
during the major EI Nino events, the time series exhibits negative values
throughout the epoch 1942-76.
Also shown in Fig. 31 is the time series of the expansion coefficient of
the leading EOF of the global SST anomaly field (hereafter referred to as
G; comparable to time series presented by Hsiung and Newell 1983, Nitta
and Yamada 1989, and Parker and Folland 1991), defined on the basis of
the anomaly deviation field; i.e., the difference between the SST at each
gridpoint and the global-mean SST for the same month. G is highly correlated with global-mean SST, with indices ofthe Southern Oscillation, and
with interdecadal climate variability over the extratropical North Pacific
(Zhang et al. 1996).
9 Interpretation of the variability of global-mean temperature
From the foregoing, it is evident that regardless of how hemispheric-mean
surface air temperature may be responding to increasing concentrations of
greenhouse gases and aerosols, it is also varying on the time scale of years
to decades in response to the sampling fluctuations associated with the
month-to-month variability of the atmospheric circulation patterns that
determine the value of the "COWL index" defined in section 5. Superimposed upon this random variability are interdecadal-scale ENSO-like SST
variations in the Pacific and Indian Oceans and fluctuations in the amplitude and polarity of North Atlantic Oscillation, which mayor may not be
deterministic.
The linear regression regression calculation that yielded the temperature pattern Fig. 15 and the fitted expansion coefficient time series in Fig.
16 was repeated, using the time series of global-mean SST in place of TL
and the global SST field in place of the extratropical Northern Hemisphere
surface air temperature field. The resulting global SST pattern whose amplitude time series is most highly correlated with the time series of globalmean SST (Fig. 30) exhibits an ENSO-like tropical signature reminiscent
of the bottom panel of Fig. 20. Its expansion coefficient time series (Fig.
31) traces out all the major warm and cold episodes of the ENSO cycle,
as represented in Fig. 22, as well as the interdecadal-scale features noted
above. It also exhibits a shift in the opposite direction (i.e., toward colder
equatorial Pacific SST) in 1942. Apart from a few brief positive excursions
during the major EI Nino events, the time series exhibits negative values
throughout the epoch 1942-76.
Also shown in Fig. 31 is the time series of the expansion coefficient of
the leading EOF of the global SST anomaly field (hereafter referred to as
G; comparable to time series presented by Hsiung and Newell 1983, Nitta
and Yamada 1989, and Parker and Folland 1991), defined on the basis of
the anomaly deviation field; i.e., the difference between the SST at each
gridpoint and the global-mean SST for the same month. G is highly correlated with global-mean SST, with indices ofthe Southern Oscillation, and
with interdecadal climate variability over the extratropical North Pacific
(Zhang et al. 1996).
9 Interpretation of the variability of global-mean temperature
From the foregoing, it is evident that regardless of how hemispheric-mean
surface air temperature may be responding to increasing concentrations of
greenhouse gases and aerosols, it is also varying on the time scale of years
to decades in response to the sampling fluctuations associated with the
month-to-month variability of the atmospheric circulation patterns that
determine the value of the "COWL index" defined in section 5. Superimposed upon this random variability are interdecadal-scale ENSO-like SST
variations in the Pacific and Indian Oceans and fluctuations in the amplitude and polarity of North Atlantic Oscillation, which mayor may not be
deterministic.