38
trated in Bradley et al. (1987). Whether data for individual stations are
capable of revealing interdecadal fluctuations in hemispheric mean temperature is more problematical. On the basis of an examination of twelve
seasonal and annual mean surface air temperature records 150 years or
more in length, together with results of a more comprehensive study by
Jones and Kelly (1983), Jones and Bradley (1992) concluded: "Although
there appears to be some agreement between warm and cold decades over
the major northern continents, ... no one region can be said to be representative of hemispheric-wide conditions. . .. the only way of producing a truly
representative time series for the Northern Hemisphere is to include station data from as many regions as possible. There is no short-cut method
using only a few stations ... " However, they went on to acknowledge that all
the sites that they examined showed warming over the period 1851-1980
and that the la-year lowpass filtered temperature records for 8 of the 12
stations that they examined were quite well correlated with fluctuations in
hemispheric mean temperature, with correlation coefficients ranging from
0.64 to 0.76.
The difficulties inherent in relating local temperature fluctuations to
hemispheric or global fluctuations are illustrated by Fig. 2. The top panel
shows monthly mean surface air temperature anomalies for a localized region of the midwestern United States centered in Iowa. The large monthto-month variability, with a standard deviation of several degrees, is an
order of magnitude larger than the observed fluctuations in hemispheric
mean surface air temperature on the interdecadal time scale. Even if these
month-to-month variations exhibited no persistence, the sampling variability of la-year running means would be as large as the observed decade-todecade variability of hemispheric mean temperature. It is only the stronger
trends that are sustained over several decades that might, at least in principle, be detectable in individual station records. Monthly mean surface
air temperature averaged over the entire United States (exclusive of Alaska
and Hawaii), shown in the middle panel of Fig. 2, exhibits consider~bly
less scatter because of the compensation between areas of positive and negative anomalies within this region. Upon close inspection, the distinctive
interdecadal signature of hemispheric mean temperature is evident in these
data, even without the temporal smoothing. When the spatial averaging
is extended to include the entire Northern Hemisphere continents (bottom
panel) the scatter is further reduced and the hemispheric interdecadal signal is much more clearly apparent. The similarity between regional and
trated in Bradley et al. (1987). Whether data for individual stations are
capable of revealing interdecadal fluctuations in hemispheric mean temperature is more problematical. On the basis of an examination of twelve
seasonal and annual mean surface air temperature records 150 years or
more in length, together with results of a more comprehensive study by
Jones and Kelly (1983), Jones and Bradley (1992) concluded: "Although
there appears to be some agreement between warm and cold decades over
the major northern continents, ... no one region can be said to be representative of hemispheric-wide conditions. . .. the only way of producing a truly
representative time series for the Northern Hemisphere is to include station data from as many regions as possible. There is no short-cut method
using only a few stations ... " However, they went on to acknowledge that all
the sites that they examined showed warming over the period 1851-1980
and that the la-year lowpass filtered temperature records for 8 of the 12
stations that they examined were quite well correlated with fluctuations in
hemispheric mean temperature, with correlation coefficients ranging from
0.64 to 0.76.
The difficulties inherent in relating local temperature fluctuations to
hemispheric or global fluctuations are illustrated by Fig. 2. The top panel
shows monthly mean surface air temperature anomalies for a localized region of the midwestern United States centered in Iowa. The large monthto-month variability, with a standard deviation of several degrees, is an
order of magnitude larger than the observed fluctuations in hemispheric
mean surface air temperature on the interdecadal time scale. Even if these
month-to-month variations exhibited no persistence, the sampling variability of la-year running means would be as large as the observed decade-todecade variability of hemispheric mean temperature. It is only the stronger
trends that are sustained over several decades that might, at least in principle, be detectable in individual station records. Monthly mean surface
air temperature averaged over the entire United States (exclusive of Alaska
and Hawaii), shown in the middle panel of Fig. 2, exhibits consider~bly
less scatter because of the compensation between areas of positive and negative anomalies within this region. Upon close inspection, the distinctive
interdecadal signature of hemispheric mean temperature is evident in these
data, even without the temporal smoothing. When the spatial averaging
is extended to include the entire Northern Hemisphere continents (bottom
panel) the scatter is further reduced and the hemispheric interdecadal signal is much more clearly apparent. The similarity between regional and