Dendroclimatology in High-Resolution Paleoclimatology - Dendroclimatology - page 26

alpine and subarctic locations, they concluded that 'during the second half of the
twentieth century, the decadal-scale trends in wood density and summer tempera-
tures have increasingly diverged as wood density has progressively fallen . ' These
results apply to their network of sites chosen to be optimal for using wood density
as a proxy of summer temperature. Further, the 'divergence' they report is in the
decadal, not the interannual, component of the response, and this divergence has the
opposite sign in parts of the middle latitudes and in the western regions of North
America and of Europe (their Fig. 2.1). Bearing this in mind, Briffa and his col-
leagues, working with this dataset, have chosen to end their calibration period in
1960, so that the divergence could not affect their reconstructions.
Subsequently, several authors have sought to identify the cause or causes of this
divergence (for example, D'Arrigo et al. 2008 b ; Vaganov et al. 1999 ) . Proposed
causes fall into three general classes—the emergence of new combinations of
climate conditions in the late twentieth century, the intervention of non-climatic
factors, such as changing atmospheric or soil composition or changed UV radia-
tion, and previously undetected artifacts in the tree-ring chronologies introduced by
the methods used to remove non-climatic variations ('standardization'), along with
associated biases related to the calibration and verification periods used. This pos-
sibility of artifacts introduced by standardization methods is raised by Melvin and
Briffa ( 2008 ) and is discussed by Briffa and Melvin ( Chapter 5 , this volume). Grudd
( 2008 ) found that increasing the number of relatively young modern trees in a mul-
timillennial density chronology from northern Sweden caused a previously present
late-twentieth-century divergence to disappear, lending support to the suggestion
that at least some of the instances of divergence are artifacts of some combination
of sample structure and standardization treatment.
Setting this last possibility aside for the moment, we see that the question arises,
might such a divergence have occurred in earlier times, and, if so, how would we
know? The Hughes et al. ( 1984 ) and Conkey ( 1986 ) studies cast light on the last
200-300 years, and in neither case is there any evidence of such instability over
that time before about 1970. Büntgen et al. ( 2008 ) examined a large network of
chronologies from across the European Alps, all carefully prepared to preserve low-
frequency variability, in the context of climate data covering the late nineteenth
and twentieth centuries. They found no evidence for any unusual late-twentieth-
century divergence problem, and indeed concluded that the warmth of the late
twentieth/early twenty-first century was unique in the last millennium. Esper et al.
( 2009 ) reported similar findings for a network of ring-width and density chronolo-
gies across northern Siberia, concluding that there is no divergence there between
tree-ring data and temperature when appropriate methods of standardization and
calibration are used.
In fact, tree-ring-network-based reconstructions of temperature over the past
several centuries are remarkably consistent with those derived from independent
records such as historical sources (Luterbacher et al. 2004 ) and the gross dynamics
of large numbers of mountain glaciers (Oerlemans 2005 ) . These results suggest that,
in the context of the last several centuries, apparent divergences that survive analyses
such as that reported by Esper et al. ( 2009 ) are unique to the late twentieth century