sm2: yamal region trw analysis - university of east anglia · figure yt07 – a) tree indices for...
TRANSCRIPT
08/04/2013 Page 1 of 12
SM2: Yamal Region TRW Analysis
YT1. Raw measurement data (Table YT1, YT2)
YT2. Inter-group common signal (Figures YT01 to YT04)
YT3. Khadytla site problem (Figures YT05-YT08)
YT4. Multiple RCS curves (Figures YT09-YT14)
YT1 Raw Measurement Data
Yamal Trees
The CE portion of the Yamal data set of 2002 (Hantemirov and Shiyatov) consisted of 265
sub-fossil samples and a selection of 17 living-tree samples from 5 sites across the region.
Further work by Russian scientists has resulted in living-tree samples from several more sites
and many more sub-fossil samples. Where multiple cores existed, these were averaged to
produce mean-tree series for all trees. Those modern samples which were less than 40 years
old were removed from the chronology because they were outside the age range of the sub-
fossil samples (all placed in file “young.rwl”). The file “yml-old.raw” is made up of a few
samples from the sites with insufficient samples to consider separately. Only a few trees from
the two earliest living-tree sites had been digitised and crossdated and these trees are in the
“yml-old.raw” file. The current data set (yml-all.raw) which we use here has samples from
473 sub-fossil trees and 142 living trees and includes sub-fossil samples with rings in the
period 500 B.C. to present though we only advise use of the resulting chronology from 1 CE.
All initial crossdating of samples was carried out by Rashit Hantemirov and Stepan Shiyatov
at the Institute of Plant and Animal Ecology, Ural Division of the Russian Academy of
Sciences, Ekaterinburg though this dating was subsequently re-examined as a prelude to the
work described here. Crossdating reports (...Corr.prn) and statistical reports of trees
(...Stats.prn) are available at http://www.cru.uea.ac.uk/cru/papers/briffa2013qsr/.
Khadytla Site
In 1991, Fritz Schweingruber (W.S.L.) and a Russian team flew by helicopter to the valley
Khadyta River (Southern Yamal) where samples were collected first from all located sub-
fossil wood. Subsequently, a search of the local forest area was made in an attempt to locate
suitable living larch trees from which to extract increment cores. As the river valley was
covered with dense forest and tall bushes, it proved difficult to find a suitable place to land
the helicopter. Finally a site was found, but the trees in this nearby forest were not considered
ideal for dendroclimatic analysis. This stand was located on the sandy shore of river, where
the depth of permafrost is more than 2 m. In such conditions the trees developed a deep root
system. Where the sand bank of the river had migrated away, a thick layer of moss developed,
and the permafrost was much closer to the surface (up to 20-30 cm) and the roots of trees
could be observed within the frozen ground. The trees at this location have reduced growth
and appear to be dying. Despite this, a set of modern samples from 18 trees was taken from
this site “Khad.raw” (Schweingruber and Briffa 1996). However, we consider that the
chronology produced from this site is not suitable for dendroclimatic analysis. These data
were included in the analysis described in Briffa and Melvin (2009). This analysis amounted
to a sensitivity study, testing the robustness of the Yamal chronology as previously published.
The work showed that the Khadytla data were anomalous in terms of 20th
century growth
trend. However, specifically to demonstrate the robustness of the Yamal chronology signal, a
version of the Yamal chronology was constructed that did contain the Khadytla data. This
was not substantively different from the previous RCS version. Nevertheless, the site report
(and statistical evidence) demonstrating the anomalous “signal” in the Khadytla data lead us
to omit them from the new Yamal chronology constructed here (see SM5 for details).
08/04/2013 Page 2 of 12
Table YT1. Some details of the sites/sub-groups of Yamal data explored in this analysis:
altitude (m.a.s.l.), coordinates (degrees and minutes), start and end year and species (LASI is
Larix sibirica Ledeb.).
File Site Name Alt North East Start End Species
Yml-sub.rwl Yamal 30 6720 7000 -764 1984 LASI
TNL.rwl Tanlovayakha 25 6723 6916 1686 1999 LASI
KHAD.raw1 Khadytayakha
1 15 6712 6950 1782 1990 LASI
HDT.rwl Khadytayakha 20 6710 6955 1626 2000 LASI
CCC.rwl Yadayakhodyyakha 30 6726 7048 1824 2005 LASI
PM0.rwl Yadayakhodyyakha 30 6726 7049 1823 2005 LASI
PVX.rwl Yadayakhodyyakha 30 6726 7049 1909 2005 LASI
YX0.rwl Yadayakhodyyakha 30 6726 7049 1886 2005 LASI
YAD.raw Yadayakhodyyakha 30 6726 7050 1803 1996 LASI
POR.raw Portsayakha 30 6725 7058 1580 1994 LASI
JAH.raw Yadayakhodyyakha 25 6723 7100 1573 1991 LASI
Notes:
1. The KHAD.raw data from Khadytayakha is called Khadytla here (following Fritz
Schweingruber and to distinguish it from HDT.raw), but it has also been called “Khadyta
River”.
2. CCC, PM0, PVX, YX0 and YAD are plots located close to each other. POR, HDT and
TNL are larger areas and the coordinates given are approximately the site centre.
Table YT2 – Basic statistics for Yamal sub-groups:
Indices created by standardising with a 30-year high-pass spline.
Corr - the mean correlation of each index series with the chronology (excluding that series).
Rbar - the mean inter index-series correlation.
MnRaw - the mean value of TRW measurements for the site.
Name File Trees Start End Years Rings Corr RBar MnRaw
CCC ccc.rwl 12 1824 2005 182 1328 0.81 0.71 1.016
HDT hdt.rwl 17 1626 2000 375 3929 0.81 0.70 0.601
JAH jahm.raw 23 1573 1991 419 4243 0.79 0.66 0.634
KHA khad.raw 18 1782 1990 209 2368 0.84 0.72 0.683
PM0 pm0.rwl 12 1823 2005 183 943 0.82 0.71 0.849
POR por.raw 12 1580 1994 415 3503 0.80 0.69 0.466
PVX pvx.rwl 12 1909 2005 97 902 0.81 0.69 1.127
TNL tnl.rwl 21 1686 1999 314 3087 0.80 0.68 0.879
YAD yad.raw 10 1803 1996 194 1338 0.76 0.67 0.880
YX0 yx0.rwl 9 1886 2005 120 794 0.80 0.71 0.796
Old yml-old.raw 14 1628 2003 376 2929 0.75 0.60 0.500
SUB yml-sub.rwl 473 -764 1984 2749 71235 0.78 0.63 0.636
ALL yml-all.raw 633 -764 2005 2770 96599 0.78 0.62 0.649
young.rwl 15 1968 2005 38 412 0.61 0.44 1.475
08/04/2013 Page 3 of 12
YT2 Inter-site common signal
Figure YT01 – To examine the consistency of the medium to high frequency signal in the
sub-sets of data the raw data were standardised using a 100-year high-pass spline, signal-free
standardisation. High-pass filtered chronologies for each site are shown (a). The same data
after filtering with a low-pass 10-year spline are shown in b). Thick lines show sample
counts >3. A high degree of medium to high frequency common variability is apparent
between all sub-groups.
Figure YT02 – As for YT01 above but using signal-free RCS, each site processed
independently. The low-frequency signal is noisy and less consistent, partly because sample
counts at each site are low and secondly, because the overall slope of each of these
chronologies is poorly defined (see Briffa and Melvin 2011 for detailed discussion). The
means of each chronology will be 1.0, independently of the common forcing over the life of
each chronology. Chronologies are shown smoothed with a 20-year spline.
08/04/2013 Page 4 of 12
Figure YT03 - The various site data sets were combined into one set and standardised using
one-curve signal-free RCS. Sub-chronologies are shown here for each site (a), and also
shown filtered with a low-pass 20-year spline for display (b). Thick lines show sample
counts >3. There is a much wider spread of values when the low-frequency is retained
(compare with YT01). For each chronology the slopes are better constrained (than was the
case in YT02b) because the inclusion of sub-fossil data and because the means are set relative
to the average of all trees. In YT02b POR looks anomalous in the 20th
century while KHA
appears to be consistent with the other series. In YT03b POR fits reasonably well while KHA
exhibits an anomalous decrease (relative to the other sites) in the late 20th
century (see later
Figures).
Figure YT04 –Sample counts by ring age (a) and by calendar year (b) for each site/group.
The sub-fossil numbers are frequently outside the plotted area because of large sample counts.
Thick lines show sample counts >3.
08/04/2013 Page 5 of 12
YT3 Khadytla site problem and justification for removal of these data
Figure YT05 - The RCS curves for each site (mean of signal-free measurements by ring age
smoothed using an age related spline) are plotted for trees standardised using signal-free RCS
using separate sites (a), one-curve RCS for all trees (b), and two-curve RCS for all trees (c).
Thick lines show sample counts >3. The CCC (green) chronology, consisting of cohorts of
young (about 75 years) and older (about 150 years) trees, appears anomalous when processed
separately (YT05a) but when processed with all the other trees is not especially unusual. The
Khadytla trees (red) show up as having an anomalous slope in Figures (b) and (c) relative to
all the other sites. In Figure (a) the slope problem will have been mitigated by the removal of
the slope of the chronology (built from living trees) from each individual tree (see Briffa and
Melvin 2011, Figure 5.2 and associated explanation).
08/04/2013 Page 6 of 12
Figure YT06 – Chronologies created from sub-groups of Yamal data (see Table YT1 for site
details) using one-curve RCS separately on each site (a). These site data were also pooled
into one data set and standardised using either one-curve signal-free RCS (b) or two-curve
signal-free RCS (c). Thick lines show sample counts >3. The tree indices for each sub-group
were then averaged together to produce separate sub-chronologies. All chronologies were
smoothed with a 20-year spline for display. The recent spread of the site chronologies is
smaller for two-curve RCS than for one-curve RCS.
08/04/2013 Page 7 of 12
Figure YT07 – a) Tree indices for the Yamal chronology (black) were created using two-
curve signal-free RCS using all the Yamal raw measurement data, including those from the
Khadytla site. A separate Khadytla sub chronology (red) was created by averaging the indices
from the Khadytla trees. The common high-frequency signal can be seen in both sets of trees.
However, a large reduction in growth rate of the Khadytla trees after 1950 relative to the
overall Yamal chronology (which includes the Khadytla trees) is clearly shown. Individual
series of tree indices for the Khadytla trees, after 30-year low-pass smoothing, are plotted in
(b). These show a decline in many trees after 1950. The Khadytla site data appear anomalous
with respect to all the other Yamal tree data used here, possibly because it is not located next
to the river (see YT08 and http://www.cru.uea.ac.uk/cru/people/briffa/yamal2009/). The
Khadytla data were removed from consideration in subsequent analysis.
08/04/2013 Page 8 of 12
Figure YT08 – Photograph of the trees at the Khadytla site ( sampled by Fritz
Schweingruber) in 1991. These trees look unhealthy. Some of the trees on this site were
dying and some were depressed likely because of changes to the permafrost layer.
08/04/2013 Page 9 of 12
YT4 Multiple RCS curves
Figure YT09 - The site data (henceforth omitting the Khadytla site measurements) were
combined into one data set and alternatively standardised using one-curve signal-free RCS (a)
and two-curve signal-free RCS (b). The raw data for each site were averaged together to
produce mean-ring width by age curves and these were smoothed with a 20-year spline for
display. Thick lines show sample counts >3. Distribution of site RCS curves shows little
difference between one and two RCS curve chronologies.
Figure YT10 - The site data were combined into one data set (yml-all.raw) and standardised
using one-curve signal-free RCS (a) and two-curve signal-free RCS (b). The tree indices for
each site were averaged together to produce site sub-chronologies and these were smoothed
with a 20-year spline for display. Thick lines show sample counts >3. The spread of the
recent site chronologies is smaller using two-curve RCS.
08/04/2013 Page 10 of 12
Figure YT11 - The site data were combined into one data set (yml-all.raw) and standardised
using one-curve signal-free RCS. The tree indices for each site were averaged together to
produce site chronologies (a) and these chronologies were rescaled to have the same mean
over their full length as the count-weighted mean of the chronology over their common
period (b). Chronologies were smoothed with a 20-year spline for display and thick lines
show sample counts >3.
08/04/2013 Page 11 of 12
Figure YT12 – as for Figure YT11 but created using two-curve signal-free RCS. Rescaling
makes little difference when using two RCS curves.
Figure YT13– as for Figure YT10 but created using three-curve signal-free RCS. Again
rescaling makes little difference when using three RCS curves.
08/04/2013 Page 12 of 12
Figure YT14 - The Yamal data (yml-all.raw) were standardised using three-curve signal-free
RCS (compare these results with those shown for two-curve RCS in Figure 4 of the main
text). Mean ring-width by age (a) is plotted for the three separate RCS curves and the average
of all values. The three separate chronologies by growth rate (b) each have a mean of
approximately 1.0 and show that these independent chronologies generally produce the same
common signal over time, with the exception that chronology sections with lower sample
counts (below 8 shown by thin lines) are less reliable. Chronologies were also created (c)
after rescaling each series of tree indices to have the overall mean value that it would have
had if a single RCS curve had been used. The three chronologies reflect the relative overall
growth rates of the trees. Chronologies were smoothed with a 50-year spline for display
purposes.