Uncertainty, Emergence, and Statistics in Dendrochronology - Dendroclimatology - page 81

for tree-ring chronologies. This is easily done by using parametric methods or the
data-adaptive bootstrap (Cook 1990 ) .
Quantifying generic statistical uncertainty in annual tree-ring chronologies is
standard practice now in dendrochronology, principally through the use of the
RBAR, EPS, and SSS statistics. The methods for doing so are theoretically sound
and well tested, but unfortunately they still tell us exactly nothing about the true
strength of the environmental signal(s) in any given tree-ring chronology. It is up to
the dendrochronologist to determine what those signals are, but considerable bio-
logical uncertainty still exists in knowing what they truly are and how they are
expressed in the ring widths. Process-based forward models of cambial growth
based on first principles of tree physiology have successfully modeled the effects
of climate on radial growth of certain tree species (e.g., Fritts et al. 1991 ; Fritts and
Shashkin 1995 ; Fritts et al. 1999 ; Shashkin and Vaganov 1993 ; Anchukaitis et al.
2006 ; Evans et al. 2006 ; Vaganov et al. 2006 ) . The results to date are very promis-
ing, but the challenge remains to make these models more adaptable to the likely
presence of biological uncertainty and emergence in many tree-ring studies.
4.3 Emergence
Emergence is the greatest source of biological uncertainty in dendrochronology
because it represents a property or signal in a tree-ring series that cannot be
predicted, even in principle, from our best understanding of the fundamental physi-
ological and environmental processes that control radial growth in trees. Emergent
properties arise in tree rings as a function of the inherent complexity of trees as
living things and the ways in which they interact with and are constrained by their
operational environment (Fritts 1976 , pp. 48-50).
Examples of emergence in the fields of dendroclimatology and dendroecology
are the almost universal positive correlation of March temperatures with subsequent
radial growth of eastern hemlock ( Tsuga canadensis ), which appears to be indepen-
dent of site characteristics and location within that species' range (Cook and Cole
1991 ) , the almost universal positive correlation between December-January tem-
peratures and subsequent radial growth of high-elevation red spruce ( Picea rubens )
in the northern Appalachian Mountains (Cook and Johnson 1989 ) , the importance
of winter temperatures on annual radial growth of six northern range margin tree
species (Pederson et al. 2004 ) , and the phylogenetic separation of Quercus species
by subgenus (Erthrobalanus vs. Leucobalanus) in the West Gulf Coast forests of
Texas and Louisiana, even when these oak subgenera grow together on the same site
(Cook et al. 2001 ) . None of these well-replicated statistical properties would have
been predicted based on previous studies, and it is unlikely that they would ever be
deduced from our best understanding of the underlying physiological processes that
directly result in annual ring formation. The interactions between tree genetics and
the tree's operational environment render such emergent properties inherently diffi-
cult, if not impossible, to predict. It also does not matter that some of these emergent
climate signals, like the March temperature response in eastern hemlock, are of no