Section 5.7: Discussion
93
only for regression estimates that fall within the range of the observational
(predictand) data. Estimates that fall outside of this range are extrapolations
and their accuracy is dependent on the assumption that such extrapolation
is valid. The more extreme the extrapolation the greater is the likelihood
that these estimates are in error (Graumlich and Brubaker, 1986).
Verification of the theoretical (or assumed) temporal signal is almost nonexistent outside of dendroclimatology (though it could be argued, with some
justification, that no other proxy science is as vulnerable to the dangers of
overcalibration of regression models). Nevertheless, other forms of proxy
data have been interpreted on the basis of calibration of multiple regression
statistics with little or no formal testing of the regression veracity.
5.7 Discussion
The intention in writing this chapter was not to catalogue al1 of the numerous mathematical techniques which are used to crossdate and standardize
tree-ring chronologies, or to calibrate and verify dendroclimatic reconstructions. To do so would have resulted in many pages of abstract equations
with little apparent relevance for students in other fields. Those who seek
the mathematical corpus may delve into the numerous references cited. Neither is this chapter an attempt to list even a cross section of the hundreds
of reconstructions of climate-related variables that form the product of the
science to date. Arecent selection is provided in Bradley and Jones (1992).
Instead, this chapter has discussed several concepts that underpin dendroclimatologyand which have important implications for viewing published work
in the context ofwider palaeoclimatic research. It is hoped that the relevance
of these concepts to other fields of palaeoclimatic research is also apparent.
Lack of space has prevented discussion of some questions listed at the start of
the chapter: specifically of seasonality and spatial representativeness of data.
For some discussion on these points see Briffa and Jones (1993) and Hughes
and Diaz (1994).
Before closing the discussion, however, it is necessary to highlight a fundamental tenet of dendroclimatic research which, while always assumed, is
frequently difficult, perhaps impossible to demonstrate in practice: that of
uniformitarianism. In a dendroclimatic context, this states that "the processes which control tree growth at present have operated unchanged in the
past".
Even when sampling living trees one can envisage the possibility that the
characteristics of the site (evolving soil fertility; changes in water holding
capacity of the soi! etc.) may have changed during the life of the trees and
modified their response to climate. The actual process of growth itself (root
and foliage development, changing competitional status etc.) can impart
age-related bias in a tree's climate response. Fortunately, this bias is small
and replicate sam pies of different age will overcome it. However, the age of
93
only for regression estimates that fall within the range of the observational
(predictand) data. Estimates that fall outside of this range are extrapolations
and their accuracy is dependent on the assumption that such extrapolation
is valid. The more extreme the extrapolation the greater is the likelihood
that these estimates are in error (Graumlich and Brubaker, 1986).
Verification of the theoretical (or assumed) temporal signal is almost nonexistent outside of dendroclimatology (though it could be argued, with some
justification, that no other proxy science is as vulnerable to the dangers of
overcalibration of regression models). Nevertheless, other forms of proxy
data have been interpreted on the basis of calibration of multiple regression
statistics with little or no formal testing of the regression veracity.
5.7 Discussion
The intention in writing this chapter was not to catalogue al1 of the numerous mathematical techniques which are used to crossdate and standardize
tree-ring chronologies, or to calibrate and verify dendroclimatic reconstructions. To do so would have resulted in many pages of abstract equations
with little apparent relevance for students in other fields. Those who seek
the mathematical corpus may delve into the numerous references cited. Neither is this chapter an attempt to list even a cross section of the hundreds
of reconstructions of climate-related variables that form the product of the
science to date. Arecent selection is provided in Bradley and Jones (1992).
Instead, this chapter has discussed several concepts that underpin dendroclimatologyand which have important implications for viewing published work
in the context ofwider palaeoclimatic research. It is hoped that the relevance
of these concepts to other fields of palaeoclimatic research is also apparent.
Lack of space has prevented discussion of some questions listed at the start of
the chapter: specifically of seasonality and spatial representativeness of data.
For some discussion on these points see Briffa and Jones (1993) and Hughes
and Diaz (1994).
Before closing the discussion, however, it is necessary to highlight a fundamental tenet of dendroclimatic research which, while always assumed, is
frequently difficult, perhaps impossible to demonstrate in practice: that of
uniformitarianism. In a dendroclimatic context, this states that "the processes which control tree growth at present have operated unchanged in the
past".
Even when sampling living trees one can envisage the possibility that the
characteristics of the site (evolving soil fertility; changes in water holding
capacity of the soi! etc.) may have changed during the life of the trees and
modified their response to climate. The actual process of growth itself (root
and foliage development, changing competitional status etc.) can impart
age-related bias in a tree's climate response. Fortunately, this bias is small
and replicate sam pies of different age will overcome it. However, the age of