58
Chapter 4: The Instrumental Data Record
(CRU /UKMO) (Jones, 1988; Jones et a1. 1991; Folland et a1. 1990, 1992),
the Goddard Institute for Space Studies (Hansen and Lebedeff, 1987, 1988)
and the State Hydrological Institute in St. Petersburg (Vinnikovet al. 1990).
The differences between the groups arise both from the data used and from
the method of interpolation. Intercomparison of the resuIts of the three
groups can be found in various publications (e.g. Elsaesser et a1. 1986; Folland et a1. 1990, 1992; Jones et al. 1991 and Elsner and Tsonis, 1991). Here
we consider only CRU /UKMO because they are the only group to incorporate
marine data.
Time series of the mean hemispheric annual and seasonal temperature
anomalies are shown in Figures 5.1 and 5.2. Both series have been widely
discussed (Wigley et al. 1985, 1986; Jones and Wigley, 1990; Folland et a1.
1990, 1992). The accuracy of the individual seasonal and annual estimates
undoubtedly varies with time. The most reliable period is from ab out 1950.
Various techniques have tried to assess error estimates, particularly for years
prior to the 1920s. Frozen grid analyses (Jones et a1. 1985, 1986a,b, 1991;
Bottomley et a1. 1990) indicate that hemispheric estimates can be weIl estimated from the sparse data available during the late nineteenth and early
twentieth centuries. Individual annual estimates appear to have about twice
the variability of the post-1950 data but the level of temperatures during the
late nineteenth century is weIl approximated by the sparse network. This
result has been confirmed by more sophisticated techniques which also use
GCM results to estimate the additional uncertainty associated with the lack
of data for the Southern Oceans (Trenberth et al. 1992; Madden et al. 1993).
Gunst et al.(1993) estimate uncertainty using optimal statistical sampling
techniques.
In Figures 4.1 and 4.2 warming begins in the hemispheric series around the
turn of the century in all seasons. Only in summer in the Northern Hemisphere are the 1980s not the warrnest decade. In this season the 1860s and
1870s were of a comparable magnitude to the most recent decade. In the
Northern Hemisphere, interannual variability is greater in winter compared
to summer. No such seasonal contrast is evident in the Southern Hemisphere.
The hemispheric annual time series and their average (the global series), Figure 4.3, represent the major observation al evidence in the "global warming"
debate (Jones and Wigley, 1990).
Though hemispheric and global series are heavily relied upon as important
indicators of past climatic change, they mask the spatial details of temperature variability across the globe. On apriori grounds there are only weak
arguments to suggest why any regional series should be more indicative of
global scale change than any other of comparable size. The issue has been
addressed by a number of workers showing spatial patterns of temperature
change (see e.g. Barnett, 1978; Jones and Kelly 1983; Folland et al. 1990,
1992; Jones and Briffa 1992 and Briffa and Jones, 1993).
Most of the focus of past and future temperature change has been con-
Chapter 4: The Instrumental Data Record
(CRU /UKMO) (Jones, 1988; Jones et a1. 1991; Folland et a1. 1990, 1992),
the Goddard Institute for Space Studies (Hansen and Lebedeff, 1987, 1988)
and the State Hydrological Institute in St. Petersburg (Vinnikovet al. 1990).
The differences between the groups arise both from the data used and from
the method of interpolation. Intercomparison of the resuIts of the three
groups can be found in various publications (e.g. Elsaesser et a1. 1986; Folland et a1. 1990, 1992; Jones et al. 1991 and Elsner and Tsonis, 1991). Here
we consider only CRU /UKMO because they are the only group to incorporate
marine data.
Time series of the mean hemispheric annual and seasonal temperature
anomalies are shown in Figures 5.1 and 5.2. Both series have been widely
discussed (Wigley et al. 1985, 1986; Jones and Wigley, 1990; Folland et a1.
1990, 1992). The accuracy of the individual seasonal and annual estimates
undoubtedly varies with time. The most reliable period is from ab out 1950.
Various techniques have tried to assess error estimates, particularly for years
prior to the 1920s. Frozen grid analyses (Jones et a1. 1985, 1986a,b, 1991;
Bottomley et a1. 1990) indicate that hemispheric estimates can be weIl estimated from the sparse data available during the late nineteenth and early
twentieth centuries. Individual annual estimates appear to have about twice
the variability of the post-1950 data but the level of temperatures during the
late nineteenth century is weIl approximated by the sparse network. This
result has been confirmed by more sophisticated techniques which also use
GCM results to estimate the additional uncertainty associated with the lack
of data for the Southern Oceans (Trenberth et al. 1992; Madden et al. 1993).
Gunst et al.(1993) estimate uncertainty using optimal statistical sampling
techniques.
In Figures 4.1 and 4.2 warming begins in the hemispheric series around the
turn of the century in all seasons. Only in summer in the Northern Hemisphere are the 1980s not the warrnest decade. In this season the 1860s and
1870s were of a comparable magnitude to the most recent decade. In the
Northern Hemisphere, interannual variability is greater in winter compared
to summer. No such seasonal contrast is evident in the Southern Hemisphere.
The hemispheric annual time series and their average (the global series), Figure 4.3, represent the major observation al evidence in the "global warming"
debate (Jones and Wigley, 1990).
Though hemispheric and global series are heavily relied upon as important
indicators of past climatic change, they mask the spatial details of temperature variability across the globe. On apriori grounds there are only weak
arguments to suggest why any regional series should be more indicative of
global scale change than any other of comparable size. The issue has been
addressed by a number of workers showing spatial patterns of temperature
change (see e.g. Barnett, 1978; Jones and Kelly 1983; Folland et al. 1990,
1992; Jones and Briffa 1992 and Briffa and Jones, 1993).
Most of the focus of past and future temperature change has been con-