cc: Edward Cook <drdendroatXYZxyzo.columbia.edu>, David Frank <david.frankatXYZxyz.ch>, Jan Esper <esperatXYZxyz.ch>, "Rosanne D'Arrgio" <druidrdatXYZxyzo.columbia.edu>
date: Wed, 21 May 2008 07:40:20 -0400
from: Edward Cook <drdendroatXYZxyzo.columbia.edu>
to: Keith Briffa <k.briffaatXYZxyz.ac.uk>
Some brief, initial thoughts and points:
I am working on implementing a full bootstrap procedure for my principal components
regression program PCREG. Hopefully, it will be sufficiently completed for some meaningful
results to be shown at the workshop. Certain aspects of it are a bit complicated to
implement, so I am hedging my bets a bit here. What I am working towards are much more
realistic expressions of tree-ring reconstruction uncertainty than that normally presented
based only on calibration period RSQ and MSE and on the usual suite of verification period
statistics. Neither honestly tells us ANYTHING about the true uncertainty in the
reconstructions prior to the calibration/verification periods. So the goal is to have
uncertainties on the reconstructions back in time based on bootstrapped pseudo-samples
taken from the actual data being used for reconstruction and weighted by pseudo-sample
estimates of the betas over the calibration period. While this would only express the
internal uncertainty of the estimates conditioned on the tree-ring series being used
(NOTHING about the accuracy of the climate estimates back in time), it would still add to
our evaluation of reconstruction uncertainty I believe. This procedure would also ignore
the uncertainties in the tree-ring chronologies themselves, which would be extremely
difficult, if not impossible, to fully account for in bootstrap model I am working on now.
Some stats people I have talked to suggest that all (i.e. tree-ring chronology development
AND climate reconstruction!) could be incorporated together in a hierarchical Bayes
framework, but I think they are smoking the wrong stuff and talking largely out of
ignorance of what we actually do in dendroclimatology and tree-ring chronology development.
How does this sound for a brief, initial contribution?
Dr. Edward R. Cook
Doherty Senior Scholar and
Director, Tree-Ring Laboratory
Lamont-Doherty Earth Observatory
Palisades, New York 10964 USA
On May 15, 2008, at 9:47 AM, Keith Briffa wrote:
A General Call for Input to a Meeting on Palaeoclimate Uncertainties
PLEASE NOTE - this message has been sent to a representative selection of those working
in different tree-ring laboratories - please forward to those of your colleagues who
would be interested - THANK YOU
I have been tasked with drafting the White paper in the general topic of Reducing
Uncertainties, in my case with a focus on tree-ring data. This is meant as the basis for
discussion at a wider meeting dealing with various high-resolution proxy data, being
held in Trieste funded by PAGES/CLIVAR.
Hence I am asking for specific input from any of those among you who wish to contribute
specific points or stress, even briefly or as concepts, areas of concern regarding
present work or future requirements.
The context is general dendroclimatology and the use of tree-ring-derived climate
reconstructions specifically for establishing the precedence of instrumental
observations in a recent multi-millennial context.
The specific issues I have been asked to address include:
1) sources of climate interpretational uncertainty how can this be quantified and
2) strategies for reducing these uncertainties?
3) database / data archiving needs and ideas?
The white paper is only intended to be several pages long so specific ideas, concerns
etc. along the lines indicated, would be very welcome. I would then try to condense them
and draft the text.
I must complete this task in the next 2 weeks so brief, initial thoughts and points that
you consider must be included would be most welcome.
At present Ed Cook ,Rosanne D'Arrigo and Dave Frank are included among the participants
( Congratulations to Jan Esper on the recent arrival of a brace of beautiful girls -
provided they take after their mother that is) and I would particularly hope for input
from them but I know it is vital to get wider input from others working in this area of
dendroclimatology or who have real concerns with the issue of climate change detection
and attribution and the use of tree-ring data for model validation or work aimed at
quantifying transient climate sensitivity in the real world.
Any thoughts, specific text or important PowerPoint slides would be most welcome.
With very best wishes and thanks
15^th May 2008
Professor Keith Briffa,
Climatic Research Unit
University of East Anglia
Norwich, NR4 7TJ, U.K.