Thursday, May 10, 2012


date: Wed, 03 Aug 2005 14:57:45 +0200
from: Stefan Rahmstorf <>
subject: [Fwd: Last Millennium section 6.5 - comments by SR]
to:, Tim Osborn <>

Hi Keith (and Tim),
Eystein told me you're back in the office tomorrow, and, to simplify things for you, he
asked me to resend my comments on your section so you will find them easily.
One point we'll perhaps need to cover better in our chapter is the model tests of the proxy
methods. David Ritson and Caspar Amman both independently seem to have concluded that
Storch et al. did not implement the MBH method correctly in their model, and that is why
the errors are so large (see the very different trends in the calibration period, where the
proxy method should be scaled to match the climate trend).
I also attach again my own little note that is currently under consideration at Science,
concerning another aspect of the VS04 publication.
If the ECHO-G result falls apart, the conclusion would be that all the other model tests of
the proxy method confirm that it works within its stated error limits.
Cheers, Stefan
-------- Original Message --------

Subject: Last Millennium section 6.5 - comments by SR
Date: Tue, 26 Jul 2005 12:09:33 +0200
From: Stefan Rahmstorf [1]<>
To: Keith Briffa [2]<>
CC: [3], Jonathan Overpeck [4]<>, Eystein Jansen
References: [6]<>

Hi Keith and all, (please everyone have a look at point (4))
I think section 6.5 is in remarkably good shape (certainly compared to my own..).
There are some comments from me:
(1) About the new proxy reconstructions, the section says: "Most of these are shown..." in
the Figure. This immediately raises the question: why not all? Which one is not shown? This
section will be scrutinised with great suspicion by some people, so we need to be careful.
Can you clarify which one you left out, and why? Or can we just write: "These are shown..."
That would be much nicer.
(2) Several times you say "simply scaled" - would "scaled" do as well? The "simply" in this
context sounds a bit like we criticise that.
(3) Is "predictand" a word that everybody knows? I'd never seen it before.
(4) Now here is my biggest question, that I think we need to discuss in the whole group.
Figure 6.5.2-1 shows simulations of the past millennium, relative to 1500-1899 means. Is
this really the best reference period?
Contra: it differs from how we show the data reconstructions, i.e., relative to 1961-1990.
Everyone knows what that climate actually was, since there are good instrumental data for
1961-1990, so that it makes sense to look at changes relative to that period. Nobody knows
what the real 1500-1899 mean was, so this is a fictitious baseline.
Pro: it gets rid of "end effects", i.e. model initialisation problems at the beginning (as
in Von Storch 04), and different anthropogenic forcings used at the end (e.g. some ignore
aerosols); the simulations look closer together in this way (right?)
I have not formed a clear opinion on what is best.
(5) Also on the figures: I like the grey bands, but here's a suggestion for improvement:
instead of leaving the core region between those two bands white, I think they should also
be shaded - either the same grey, or a darker shade of grey. This makes it more clear that
we are talking about one, wide uncertainty band here, not about two seperate things. It had
me confused at first when I saw it, even though I was there when we discussed this in
Final point: we need to keep an eye on developments concerning the model tests of the proxy
method, there seem to be several important things in the pipeline there.
Cheers, Stefan
To reach me directly please use: [7]
(My former addresses are read by my assistant Brigitta.)

Stefan Rahmstorf

To reach me directly please use: [10]
(My former addresses are read by my assistant Brigitta.)

Stefan Rahmstorf

Attachment Converted: "c:\documents and settings\tim osborn\my

No comments:

Post a Comment