cc: email@example.com, firstname.lastname@example.org, email@example.com, Scott Rutherford <srutherfordatXYZxyz.edu>
date: Mon, 03 Nov 2003 12:16:00 -0500
from: "Michael E. Mann" <mannatXYZxyzginia.edu>
subject: Re: follow up as per Mike's comments earlier?
to: Keith Briffa <k.briffaatXYZxyz.ac.uk>, firstname.lastname@example.org, mhughesatXYZxyzr.arizona.edu
Thanks for your message.
I this as two related issues.
1) Building a bit on the initial response we've already submitted, I believe we can show
beyond a reasonable doubt that the reconstruction MM produced natural results from the
elimination of key predictors from our network--and that the resulting reconstructions,
unlike those we published, do not pass cross-validation.
We're talking to Nature right now about allowing us a formal response, and I believe this
may go forward...This alone will demonstrate the invalidity of the MM03 paper, and I'd like
to think that it will put the matter behind us as far as legitimate scientists are
concerned--the feedback I'm already getting indicates that our colleagues believe that we
have shown already strongly put in doubt the M&M result just with the limited analysis
2) What you suggest, however, would be immensely useful. It is a natural followup to the
paper that we all currently have in review in Journal of Climate w/ Scott as first author,
and I see that as the first step, which is a first stab at the intercomparison issue,
though it doesn't deconstruct the MBH98 predictor network. So I'd like to suggest we
proceed, as you suggest, in that spirit. We can begin to coordinate plans, and it would
seem natural to include Scott in this as he has really been carrying the actual work
forward for us...
3) In parallel, we have been working on a considerable expansion of the original MBH98
network for further reconstructions, and perhaps we should discuss how this might best mesh
w/ your efforts.
At 04:51 PM 11/3/2003 +0000, Keith Briffa wrote:
Mike , Ray, Malcolm
We three have been discussing the weeks shenanigans and thought we should start the
wider discussion on the concept and practical aspects of someone (perhaps us - perhaps
not us) doing the "independent" audit of your 98 and/or 99 work. It is clear that the
debate as regards the M and M results will now likely stall , until one or more people
undertake this - but it is unlikely to go away until such time as something is done. The
problem , however, is what this audit would set out to do.
If it constitutes taking the exact same data and exact implementation of your method -
there seems little value in doing it - the result will be identical to your result(s).
The question then , revolves around a fuller experiment in the use of various selection
criteria for retaining subsets of the predictors , and presumably working towards
gaining some better impression of the stability of the results , and the sensitivity to
the inclusion of particular predictors. We could perhaps also compare results with those
achieved using other methods (such as our Orthogonal Spatial Regression)? We would not
be attempting to do any new reconstructions.
The question then , stimulated again by Mike's message , and the message from the Editor
of E and E, is whether we should consider going ahead here, with some such work?
We are already motivated to look at the role of the tree-ring data (in collaboration
with Malcolm and Ed hopefully, looking at stuff like standardisation issues, the western
U.S. recent trend correction etc.) , but we feel some discussion among all of us would
help to clarify opinions and prospects for a wider look at the robustness of the
We have no particular axe to grind , but it is almost certain that there will be some
pressure for some such work, and we suspect that DEFRA here will be quizzed by various
bodies for their opinion on this. If so, why not us rather than others ?
It may be that anything we do here would not be seen as "independent" by the skeptics
anyway ( and we would not consider doing it without some appropriate level of
interaction with you lot) - but in the end , what counts, is what is published in the
peer-review literature.It was important to get your statement out , but it needs to be
followed up now by one or more studies by other groups. We could go with the candidate
predictor set you used and do a Monte Carlo approach to selection over different periods
, or add in other predictors or ....what? What do you think? We might need to go for a
very small amount of money from DEFRA ( to pay Harry or someone just to manipulate
palaeo data , and then after implementing the method(s) and deciding on the scheme, run
the numerous experiments and synthesize results), or we may be able to do it by
diverting his time from some other stuff anyway.
To get the discussion going , we wish to ask your opinion(s) on the concept, level of
interaction between us and you guys ( in planning , or also implementation , and
synthesis, writing up?). What about this issue of our perceived independence - do we
give a damn?
Keith, Tim, Phil
Professor Keith Briffa,
Climatic Research Unit
University of East Anglia
Norwich, NR4 7TJ, U.K.
Professor Michael E. Mann
Department of Environmental Sciences, Clark Hall
University of Virginia
Charlottesville, VA 22903
e-mail: mannatXYZxyzginia.edu Phone: (434) 924-7770 FAX: (434) 982-2137