Geoscience Reference
In-Depth Information
4.2 Uncertainty
Uncertainty
arises from the fact that tree-ring analysis is usually applied as a ret-
rospective study of radial tree growth formed in an unobserved, uncontrolled, and
steadily changing environment over some years in the past. Consequently, we rarely
know with much certainty the kinds and details of the environmental signals con-
tained in tree rings prior to analysis. And even after our best analyses, we can only
make our best 'guess' of what those signals are. This acknowledgment of uncer-
tainty ought not be viewed negatively, because this is the way science works. The
great Nobel Prize-winning physicist Richard Feynman viewed it as a necessary part
of science: 'Nothing is certain or proved beyond all doubt. And as you develop more
information in the sciences, it is not that you are finding out the truth, but that you
here we have a clear statement that scientific 'truth' is really probabilistic and, thus,
always associated with some degree of uncertainty, which in Feynman's view should
be embraced as central to scientific inquiry.
This fact does not mean that dendrochronologists cannot make useful a pri-
ori inferences about the likely environmental signals in tree-ring series. Thus, we
might be able to accurately infer the most likely dominant signal(s) expressed in
the growth variations of an annual ring width series based on the tree species being
studied (e.g., Douglas-fir [
Pseudotsuga menziesii
] or white spruce [
Picea glauca
]),
the growth metric being studied (e.g., ring width or maximum latewood density),
and the growth environment in which it is growing (e.g., lower or upper forest bor-
der limits). For example, ring widths of semiarid site conifers growing at a lower
forest border limit site are likely to reflect variations in available soil moisture sup-
of high-elevation conifers are likely to reflect variations in growing season temper-
inferences are nothing more than educated guesses that must be modeled and ver-
we could still be wrong. In addition, we should always be ready for surprises; i.e.,
unexpected discoveries that may point the way to rich new research opportunities.
Thus far, this discussion has dealt essentially with biological uncertainty because
we are dealing with the problem of identifying environmental signals in a biologi-
cal time series, again with an emphasis here on climatic influences on tree growth.
However, a completely different form of uncertainty also exists in tree-ring data
that is associated with the development of annual tree-ring chronologies most fre-
quently used for further study. This uncertainty is statistical rather than biological,
although some of the statistical uncertainty in a chronology may be generated by
the tree growth properties of a particular tree species (e.g., variable ring boundaries
and poor circuit uniformity). Typically, such chronologies are mean-value functions
of crossdated tree-ring series from many individual trees on a site. Given that the
tree-ring data used in the mean-value function all crossdate to an acceptable degree,
