Uncertainty, Emergence, and Statistics in Dendrochronology - Dendroclimatology - page 103

3-5 years in oaks (Trumbore et al. 2002 ) . The relative importance of this climate
eigenvector in the response functions may also reflect some phylogenetic differ-
ences in the climate response of hardwoods and conifers. For example, the implied
greater dependence on above-average precipitation during the March-April months
may be related to critical early springwood vessel enlargement, particularly in the
Quercus species. The other interesting feature of eigenvector #6 is the prominent
positive loading in January. This phenomenon may be associated with the winter
temperature sensitivity of certain tree species described by Pederson et al. ( 2004 ) .
Climate eigenvector #10 (Fig. 4.6c ) is more complicated (perhaps due to orthogo-
nality constraints) and, therefore, more difficult to interpret from a tree physiological
perspective. However, its presence in most of the response functions indicates that
it does have some true biological meaning. It mainly emphasizes monthly climate
variability during the late prior growing season, mainly during the prior August-
October months. As a predictor of tree growth, it also appears to be more important
for the hardwoods.
This evaluation of the three most important climate eigenvectors has
revealed a well-understood dependence between tree growth and moisture sup-
ply/evapotranspiration demand during the current growing season of trees growing
on well-drained sites, especially eigenvector #3. This common signal among all
seven tree species is why they all crossdate significantly; from COFECHA (Holmes
1983 ) , the mean correlation between all series is 0.55 (range: 0.30-0.69). Therefore,
a forward model of cambial growth parameterized to model this basic water rela-
tions' effect on tree growth should produce useful first-order estimates of these
tree-ring series (see Vaganov et al. Chapter 3 , this volume). Details at the genus
and species level might be missed by the forward model, however, unless additional
climate response information suggested by eigenvector #6 (and perhaps eigenvector
#10) is part of the model. Interestingly, the most poorly correlated series among the
seven is pitch pine (PIRI), whose response function does not include eigenvector
#6. The lack of this signal, common to all other species, may help explain why this
pitch pine chronology crossdates relatively weakly with the others ( r = 0.30).

4.8 Some Implications for Climate Reconstruction

The response function model results indicate that the seven tree species tested have
statistically significant climate information in them, but with varying degrees of
strength and fidelity. Based on the examinations of the most important climate
eigenvectors entered into the response function models, it appears that the domi-
nant common signal among all species is a May-July growing season response to
above-average precipitation and below-average temperature. Climate eigenvectors
#3 and #6 also indicate that this response occurs in both the current and prior grow-
ing seasons. Consequently, it should be possible to exploit both years of climate
information in the tree rings for reconstructing past growing season climate. We
have investigated this potential by reconstructing May-July total precipitation at
Mohonk Lake using the mutual climate information in all seven tree-ring series. We