Geoscience Reference
In-Depth Information
in the regions to which these studies apply. Even so, it is important to develop strate-
gies for detecting whether such divergences have occurred in the past. This could
best be done by examining whether completely independent types of natural archive
yield proxy records that continue to be mechanistically and climatically consistent
with one another for the periods of interest, as in the comparison of tree-ring-derived
as was mentioned above. Although the issue of unstable signal, and divergence, has
mainly been discussed in the context of tree rings as archives of past temperature,
this concern must apply equally to moisture-related records.
2.5 The Quest for Unbiased Chronologies
2.5.1 The Problem
Annual rings formed early in the life of a tree tend to be wider than those formed
later. The main cause of this is the tree's mode of growth. Each year's new annual
ring (which consists of xylem) is laid down outside the previous years'. Thus, if
approximately the same volume of wood is laid down each year, and if the pro-
portionality between radial and height growth does not change radically, the newer
rings must, in general, be thinner than the older rings. As a result, there is usually
an 'age-size-related trend' in ring width. In many cases, rings get larger in the first
few decades of the tree's life, presumably as the tree's canopy is built and as it
emerges from the shade of its neighbors. For other reasons that are likely related
more to the mechanical function of wood, there is usually also a downward trend in
maximum latewood density with increasing distance from the pith. These age-size-
related trends can contaminate any record of climate variability to be found in the
annual rings, and so must be discounted in some way.
There are several other non-climatic sources of variability in tree-ring width and
maximum latewood density. The overwhelming majority of dendroclimatic recon-
structions published to date have been based on measurements of TRW or MXD.
In the case of trees growing close enough to influence one another's growth, and in
dense enough stands for fire and other disturbances to affect growth, 'disturbance
pulses' leave strong imprints on the course of growth, modifying or even completely
obscuring the general age-size-related trend. In the early years of the development
of dendroclimatology, primarily in the semiarid American Southwest, the trees sam-
pled most commonly fit the description of 'open-grown'—that is, because trees were
far apart there was minimal interaction between them, and disturbance pulses played
only a small role. Thus, the preparation of tree-ring chronologies for use as climate
records consisted mainly of detrending each sample with a simple deterministic
model, such as a linear trend model or a modified negative exponential curve (Fritts
series per tree) to produce a mean chronology. The intention was that averaging
many replicate samples from each site should result in a relative strengthening of
