84
Chapter 5: Interpreting High-Resolution Proxy Climate Data
5.5 "Standard~zation" and Its Implications
for J udging Theoretical Signal
5.5.1 Theoretical Chronology Signal
In our discussion until now, we have recognised only a statistically-manifest
signal contained in each of a number of tree-ring series and expressed in a
mean chronology, but it is important also to appreeiate a different concept:
that of theoretical signal in the mean chronology. This is an expression of
some foreing which is of particular interest to the investigator and will differ according to the aims of the study. This theoretical signal may be very
different from the empirical signal represented in the data. Top of the dendroclimatologist's list of important theoretical tree-growth signals is obviously
that of (some form of) climate, but other possible signals include the effects of
growth fertilization (e.g. through regional nitrogen deposition or increasing
atmospheric CO2), pollution, insect defoliation, fire and management practices. Each of these may simultaneously affect the growth of all trees within a
given site or region. These processes operate on different timescales and one
may mask or amplify the effects of others. Careful consideration of the nature
of the signal under investigation, particularly its time-series characteristics,
can suggest ways of modifying original tree-growth measurement series so as
to enhance the signal by removing variance known not to be assoeiated with
it.
5.5.2 Standardization of "Raw" Data Measurements
The process of selective removal of unwanted variance in raw measurement
series, prior to their being averaged to form mean chronologies, is known
in dendroclimatology as "standardization". Expansive discussions of the rationale and techniques used to standardize tree-ring series may be found in
Fritts (1976); Fritts and Swetnam (1986); Briffa et a1. (1987); Cook et a1.
(1990a) and many references therein.
Standardization is most commonly employed to remove the effects of tree
ageing that are almost universally apparent in tree-ring (ring-width or ring
densitometric) measurement series. Simply stated, where they are not suppressed by competition, older trees generally put on thinner (and less dense)
rings than younger trees as net primary productivity is progressively spread
around an increasing ring eircumference. The pattern of decreasing ring
width often approximates to a negative exponential in ring-width series and
to a more linear trend in maximum latewood density series (see Bräker, 1981;
Briffa et a1., 1992a). Decadal and interannual variability is superimposed on
this clearly non-climatic trend.
Were it possible to average data from an unlimited (or at least very large)
number of tree-ring series, randomly distributed through time, this age bias
would cancel out. This is rarely, if ever, the case, however, and many "raw-
Chapter 5: Interpreting High-Resolution Proxy Climate Data
5.5 "Standard~zation" and Its Implications
for J udging Theoretical Signal
5.5.1 Theoretical Chronology Signal
In our discussion until now, we have recognised only a statistically-manifest
signal contained in each of a number of tree-ring series and expressed in a
mean chronology, but it is important also to appreeiate a different concept:
that of theoretical signal in the mean chronology. This is an expression of
some foreing which is of particular interest to the investigator and will differ according to the aims of the study. This theoretical signal may be very
different from the empirical signal represented in the data. Top of the dendroclimatologist's list of important theoretical tree-growth signals is obviously
that of (some form of) climate, but other possible signals include the effects of
growth fertilization (e.g. through regional nitrogen deposition or increasing
atmospheric CO2), pollution, insect defoliation, fire and management practices. Each of these may simultaneously affect the growth of all trees within a
given site or region. These processes operate on different timescales and one
may mask or amplify the effects of others. Careful consideration of the nature
of the signal under investigation, particularly its time-series characteristics,
can suggest ways of modifying original tree-growth measurement series so as
to enhance the signal by removing variance known not to be assoeiated with
it.
5.5.2 Standardization of "Raw" Data Measurements
The process of selective removal of unwanted variance in raw measurement
series, prior to their being averaged to form mean chronologies, is known
in dendroclimatology as "standardization". Expansive discussions of the rationale and techniques used to standardize tree-ring series may be found in
Fritts (1976); Fritts and Swetnam (1986); Briffa et a1. (1987); Cook et a1.
(1990a) and many references therein.
Standardization is most commonly employed to remove the effects of tree
ageing that are almost universally apparent in tree-ring (ring-width or ring
densitometric) measurement series. Simply stated, where they are not suppressed by competition, older trees generally put on thinner (and less dense)
rings than younger trees as net primary productivity is progressively spread
around an increasing ring eircumference. The pattern of decreasing ring
width often approximates to a negative exponential in ring-width series and
to a more linear trend in maximum latewood density series (see Bräker, 1981;
Briffa et a1., 1992a). Decadal and interannual variability is superimposed on
this clearly non-climatic trend.
Were it possible to average data from an unlimited (or at least very large)
number of tree-ring series, randomly distributed through time, this age bias
would cancel out. This is rarely, if ever, the case, however, and many "raw-