date: Thu, 15 May 2008 15:10:50 UT
subject: Review Received by Geophysical Research Letters
Content-Transfer-Encoding: binary Content-Type: multipart/alternative;
boundary="_----------=_12108642502325251" MIME-Version: 1.0 X-Mailer: MIME::Lite 3.01
(F2.74; B3.07; Q3.07) Date: Thu, 15 May 2008 15:10:50 UT Message-Id: <8121086425042@gems>
Dear Dr. Briffa:
Thank you for your review of "Annual temperatures during 2.5 millennia in the Eastern
Tibetan Plateau inferred from Tree rings" by LIU YU, Zhisheng An, Hans Linderholm, Deliang
Chen, Huiming Song, Qiufang Cai, Junyan Sun, and Hua Tian [Paper #2008GL034169], which we
have safely received. A copy of this review is attached below for your reference.
Thank you for your time and effort!
Geophysical Research Letters
Science Category: Science Category 2
Presentation Category: Presentation Category C
Annotated Manuscript: No
The material on which it is based - with potential to produce the first more than
millennium , continuous tree-ring-based climate reconstruction for this area. However, IN
ITS PRESENT FORM IT IS NOT SUITABLE FOR PUBLICATION -please see detailed review.
I wish to introduce my comments by saying that the subject of the analysis described, i.e.
that of the development and initial analysis of a long tree-ring chronology for the Eastern
Tibetan Plateau (ETP) is certainly an important one and one that is relevant for
publication in this journal.
However, there are a number of issues, associated with the methodology of producing the
chronology, and the consequent validity of its interpretation as a record of past
temperature variability, that must be recognised and addressed before I would sanction much
of the interpretation placed on these data by the authors. The potential importance of the
material they have assembled is great and for this reason it is incumbent on them to make
explicit reference to these issues and preferably even undertake additional/alternative
analyses to prevent (possible) misinterpretation of their results as currently presented.
The issues to which I refer include first, Chronology Production, specifically the issues
of standardisation and variance correction; second, the specific interpretation of the
produced long series in terms of climate; and third, the authors' comparison of their data
with other series of reconstructions and the discussion on prospective solar and carbon
dioxide causes of the observed multidecadal timescale variability.
I will deal with each of these issues in turn:
The 'take home message' that the authors stress and hence the essence of their results in
that the 20th Century was not the warmest period in the last 1-2 millennia. This conclusion
is unsound when their tree-ring data are standardised using the option they have chosen
i.e. Cook's ARSTND software with so-called 'conservative' standardisation curves (negative
exponentials or straight line - especially any slope straight line). This approach
explicitly restricts the retention of low-frequency variance in the chronology and can
cause serious end-effect distortion to chronology indices where strong recent warming
trends may be affecting recent increasing growth trends. In other words, despite their
desire to retain long-timescale information (vital if their conclusions are to be valid),
their choice of data processing likely negates the basis for the claims they make. At the
very least they should acknowledge the probability - but more correctly, they should
re-possess their data
using 'Regional Curve Standardisation' or RCS (see Briffa et al. 1992: Climate Dynamics
7,111-119 and Cook et al.1995: Holocene 5,229-237).
Even using RCS may be problematic with their data because of the clearly different source
material (modern versus tomb-derived wood samples), but provided they apply the technique
with care, they will have a much firmer basis for comparing 20th with 5th to 9th century
growth rates - something which is most definitely compromised by the processing technique
they have adopted.
Another issue, and one which they allude to, is the very variable replication rate of the
sample material through time. With such large differences (from virtually one to more than
150 samples) in sample replication between the roughly medieval time and earlier and recent
times it is important to correct the chronology variance for changing sample size through
time (see Osborn at al, 1997: Dendrochronologia15,89-99). I am aware that the authors make
passing reference to the possibility of a problem of varying (replication-related) variance
at specific times in their text. However, this is such a potentially significant influence
that it can not be left as it is. It is not clear to what extent variance correction will
affect the large variability apparent in the 7th to 9th centuries, but it is not sufficient
to leave this question unexplored - particularly when much of the high variability in the
chronology is apparent at just these times.
Interpretation in Terms of climate
The authors do not provide sufficient information regarding the justification and
statistical validity of their interpretation of the chronology in terms of annual mean
temperatures. It is usual to provide a correlation matrix where individual monthly climate
variables or various seasonal averages are compared against the tree-ring data to
demonstrate, at least empirically, the potential associations between growth and possible
seasonal temperature and other (e.g. precipitation) forcing. No such evidence is given
here. Instead we are informed that the ring width is driven by prior year mean
temperatures. We have no data to gauge the relative importance or justification of this
choice. Nor are we told precisely what this season is (i.e. calendar year or 12-month
average from some month other than January?). There is also clearly a problem with the
strong common trend in both the observational and tree-ring time series (as shown in Figure
1 of the supplementary material). This
appears to show that the association is dominated by trend and it is hard to conceive that
the assumed link with temperature , if real, would not show up as a high interannual match
and not only in the low-frequency . It is therefore necessary to calculate separate
correlations for the interannual data to quantify the extent to which this is so. At the
very least, the significance levels reported for the correlations between various
meteorological stations and the ring-width indices are clearly overstated because of this
clear co-incidental trend in both data sets the statistical significance values quoted need
to be adjusted for the presence of common significant autocorrelation in the tree-ring and
climate series. This is also true of all the correlations quoted in Tables 1and 3 of the
Supplementary material. I am also highly dubious about the validity of the claimed
verification statistics (Table 2) based on bootstrapping when the series are clearly both
affected by strong
positive trends. In short much more attention and much more detail is required to justify
the interpretation in terms of annual temperatures , and to demonstrate an absence of
association with other (e.g. summer season temperature) climate drivers.
Comparison with other Reconstructions and Solar and CO2 series
This is perhaps the weakest and clearly superficial part of the paper and requires serious
reconsideration and possibly reanalysis. Many of the series against which the ETP series is
compared have been specifically processed using RCS methods (see earlier comments). They
also represent calibrations directly against large-scale temperatures and this may render
their amplitudes of variability incompatible for direct comparison with that of the ETP
series which is calibrated against a single, local temperature record. The comparison of
these series is anyway rather subjective and vague - though I note the correlations in
Supplement Table 4 - though again the significance levels are not valid and require
adjustment to allow for common autocorrelation.
I also find the discussion on the association between solar variability and carbon dioxide
totally subjective, and unconvincing.
I have taken some time to detail problems with this manuscript, primarily because I
consider the underlying data and potential implications of rigorous evaluation to be of
importance. So much so that I urge the authors to consider my comments and revise their
analyses and description of results accordingly.
My comments are intended to be constructive. I would be happy to discuss them further and
to this end I am happy for my identity to be revealed to the authors. I sincerely hope that
they will interpret my remarks in the manner in which they were intended.
K. R. Briffa