Geoscience Reference
In-Depth Information
4.7 Additional Response Function Interpretations
Table 4.4 lists the actual eigenvectors used in each response function in order of
entry into the models. Given the disappointing verification results, these tree species
have still clearly 'voted' for the most important cross-taxa climate eigenvectors. For
six of the seven species, eigenvector #3 enters first; for the remaining species it
enters second (refer to Table 4.4 ) . The second most important is climate eigenvector
#6, which enters either first or second in five of seven species models, all deciduous
hardwoods (refer to Table 4.4 ) . The third most important is climate eigenvector #10,
which often enters third or fourth into the models (refer to Table 4.4 ) . After that, the
selected eigenvectors vary much more between species and models.
We stated earlier that the more important climate eigenvectors probably have
some physical meaning due to the intercorrelations between monthly temperature
and precipitation, but they are not constrained to have any biological meaning at
all. While this is true, the results of our response function analyses also indicate
that eigenvectors #3, #6, and #10, at least, are likely to have significant biological
meaning to the trees, given their associations with ring width. For this reason, we
will examine these climate eigenvectors for some biological meaning.
Climate eigenvector #3 (Fig. 4.6a ) reveals an oppositional pattern between
monthly temperature and precipitation that is strongest during the May-July current
growing season months. This pattern is probably physically based in the sense that
higher rainfall in the warm-season months should result in lower temperatures due to
increased cloudiness and vice versa. It also makes biological sense as a predictor of
tree growth, because our trees are growing on well-drained sites where they should
have a natural sensitivity to moisture supply and evapotranspiration demand during
the growing season. Thus, above-average precipitation and below-average tempera-
ture should jointly contribute to above-average radial growth, and this model applies
more or less equally well to all seven tree species. Interestingly, there is a rever-
sal in the precipitation/temperature association during the preceding March-April
months that presages the start of the radial growth season. This observation indi-
cates that warm/dry spring conditions tend to precede cool/wet late-spring/summer
conditions at Mohonk Lake and vice versa. The physical meaning of this spring-
summer pattern is unclear and its biological significance is even less certain. The
March-April pattern could be totally unrelated to actual tree growth and simply car-
ried into the response function because of its association with the more biologically
meaningful May-July pattern in eigenvector #3. Conversely, warm/dry March-April
conditions could also help in initiating early physiological activity in the trees prior
to the radial growth season; i.e., it may have some phenological significance. The
fact that the March-April pattern greatly diminishes in the final hemlock response
function (cf. Fig. 4.4a , d ) supports the former argument, but the highly significant
positive March temperature response is still probably real for this species (Cook and
Cole 1991 ) . In contrast, the May-July pattern remains largely intact through all four
steps of the response function calculation. Prior to March, the eigenvector loadings
are uniformly lower, which means that they are more likely to be there by chance
alone.
Search WWH ::




Custom Search