date: Mon Feb 23 15:06:46 2009
from: Phil Jones
subject: Re: Fwd: pre-review Arctic 2k reconstruction
to: Ray Bradley
Ray,
Had a quick look and weak on statistics isn't the half of it!
The tree-ring stuff needs to mention RCS - otherwise you might as well drop the trees!
There needs to be a column on how they were standardized.
The Yamal chronology in this paper has no low frequency - the way it is
standardized loses it. The paper says so - and it is obvious if you look.
What you should be referencing is a paper of Keith's from Phil Trans in 2008. He
standardized the long Russian series with a consistent RCS approach.
Nauzbaev and Vaganov know how to do this as well, but their 2000 paper is an odd
one to use. Use their one from the Holocene in 2002.
Keith told me that Grudd 2008 got it wrong as well.
Also whilst we're on trees, if I was a skeptic I'd get fed up with people referencing
Mann et al (2008) for some North American tree sites. This is a red rag to a bull.
Again they should be to original sources and should say how they are standardized.
Finally, I've calculated numerous DW statistics in my time. I've never got one approaching
3.9. According to wikipedia it is possible (range is 0 to 4). However, what you're
supposed to
do with the values above 2 is to subtract them from 4. So this one should be 0.1.
What it then means is that there is very, very, very, very high negative autocorrelation
(because of the subtraction). It doesn't look as though there is, so I think
the calculation is screwed!
I think it does stink, but they just need to write a paper with more details!
The ice core interpretation depends on how the diffusion is allowed for - may
not make much difference on this timescale.
Also - they don't need all the ADs. They need to go before dates, but they aren't
needed!
Cheers
Phil
At 20:11 21/02/2009, Ray Bradley wrote:
Hi Phil,
If you have a moment to scan through this & give us some feedback...it would be
appreciated. As you will see there are some statistical weaknesses....! But I think
you will find it of interest.
Darrell Kaufman just wrote :
" Ray says that you'd tell us whether it stinks.
To lay it bare, we're focusing on 20-year means, and our calibration
period is only 100 years, so n = 5. The standard error of the regression is ± 0.04°C,
and the Durbin- Watson statistic, d = 3.9, indicating that autocorrelation is not
significant.
On the other hand, we can calculate annual proxy values for subset of 11 records that
are
resolved annually to show that the relation between proxy and
instrumental temperature is strong (Fig 2 in the attached pdf). ".
Thank you in advance for your suggestions.
Darrell
Darrell S. Kaufman
Professor of Geology and Environmental Sciences
Northern Arizona University
928-523-7192
[1]http://jan.ucc.nau.edu/~dsk5/
Raymond S. Bradley
Distinguished Professor
Director, Climate System Research Center*
Department of Geosciences, University of Massachusetts
Morrill Science Center
611 North Pleasant Street
AMHERST, MA 01003-9297
Tel: 413-545-2120
Fax: 413-545-1200
*Climate System Research Center: 413-545-0659
< [2]http://www.paleoclimate.org>
Paleoclimatology Book Web Site: [3]http://www.geo.umass.edu/climate/paleo/html
Publications (download .pdf files):
[4]http://www.geo.umass.edu/faculty/bradley/bradleypub.html
Please don't print this e-mail unless you really need to! *
Prof. Phil Jones
Climatic Research Unit Telephone +44 (0) 1603 592090
School of Environmental Sciences Fax +44 (0) 1603 507784
University of East Anglia
Norwich Email p.jones@uea.ac.uk
NR4 7TJ
UK
----------------------------------------------------------------------------