Dendroclimatology in High-Resolution Paleoclimatology - Dendroclimatology - page 25

2.4 Stability of the Climate Signal
2.4.1 Temporal Stability
The existing climate reconstructions made using tree rings are based on statisti-
cally derived transfer functions using (largely) twentieth-century instrumental data
for calibration and validation. Two questions arise, although the difference between
them is not always recognized. First, it is necessary to test whether the transfer
function is statistically stable. Second, there is a need to know if the fundamen-
tal relationship between climate and tree-ring formation is likely to have varied
over the period to be reconstructed. A particularly advantageous situation arises
when local climate data cover a sufficiently long period to allow for the develop-
ment of statistically stable transfer functions that cover non-overlapping periods.
Hughes et al. ( 1984 ) were able to perform a rigorous test of the stability of the
tree-ring/temperature relationship back to the early nineteenth century, and less
rigorously into part of the eighteenth century, by using two of the longest contin-
uous instrumental records in the world—the Edinburgh, Scotland, meteorological
record, and Manley's central England monthly temperature series. These were com-
pared with tree-ring data from the Scottish Highlands. The tree-ring/temperature
relationship was stable over at least 160 years, and probably over 250 years, as com-
pared with the central England temperature record. Not only are the reconstructions
derived from the separate 80-year calibrations almost identical, but the structures of
the statistical models for each are similar (Hughes et al. 1984 , p 341). It would, of
course, be of great interest to see if this relationship stayed stable since 1970.
Conkey ( 1986 ) similarly checked her ring-width and density-based reconstruc-
tions of New England summer temperatures, calibrated and validated against
twentieth-century temperature records by comparison with late eighteenth- through
early nineteenth-century temperature data derived from early measurements and
diaries, and found them to have been stable over the two-century period. Whereas
Hughes et al. ( 1984 ) examined their seasonal transfer functions for stability, Carrer
and Urbinati ( 2006 ) tested the stability of monthly correlations and response func-
tions in larch in northern Italy against long climate data extending back to the
beginning of the nineteenth century. They found tree-ring growth to be consistently
most sensitive to June temperatures, but the strength of this relationship varied rather
strongly. From the point of view of reconstructing past temperatures by using their
tree-ring data, it would be interesting to know if this variation in sensitivity would be
enhanced, unaffected, or damped when it is examined in a seasonal transfer function
mode, rather than in the monthly response function mode they investigated.
2.4.2 Recent Reports of Divergence Between Temperature
and Tree-Ring Density and Width
It is against this background that the results of Briffa et al. ( 1998 ) emerged. Working
with hundreds of ring-width and maximum latewood density chronologies from