Monday, March 5, 2012


date: Thu Mar 18 14:12:06 2004
from: Phil Jones <>
subject: Re: minor alteration
to: "Tas van Ommen" <>

Will keep in touch. Will send paper when we get a pdf/reprint.
I'll look at the annual data when I have some time. Meetings all next week need preparing
for and then the April weeks after Easter as well. Not sure when I'll get to it, but I
At 22:07 18/03/2004 +1100, you wrote:

G'day Phil,
Sounds like you are on firm ground if you used the series I sent last year - 4 year
resolution will not harm the decadal scale smoothing even if it makes annual correlation
somewhat questionable. You need not be worried if you only used my 2003 data - it is
based on the best data we have, which are indeed a multi-core stack over much of the
calibration period (the header notes from the file give details).
This still leaves a point of question over your relatively high r value, and this may be
because of the underlying 4 year smooth in your "annual" data. I would recommend that
you look at the annual data I sent earlier today and confirm what is going on.
As a point of interest, my calibration exercise is revealing that the Law Dome d18O is
capturing comparable variance in the mid-latitude SLP field in the eastern Indian Ocean
to the local temperature signal near Law Dome. Not surprising when one appreciates the
role of cyclonic transport.
Do keep in touch - I am happy to help however I can to ensure the best fidelity in our
contribution to the paleo-reconstruction.
-----Original Message-----
From: "Phil Jones"<>
Sent: 18/03/04 20:13:27
To: "Tas van Ommen"<>
Subject: Re: minor alteration

> Tas,
Attached is the series I've used. It does look quite smooth, so it
seems to be
the 4-yr smoothed one you mention. So it seems the annual correlation I
talk about
for 1957-80 is based on this against the HadCRUT2v series. Both then get
with the 10-yr Gaussian filter I talked about yesterday. The fact that
the Law Dome series
I have is already smoothed won't make too much difference to this.
I have the old series also, but never compared the two. I say I have
it, but it is in an
old directory somewhere. Just found it and it has years 1304-1987. Didn't
use it though
as it was elsewhere and I gave it a different name then.

So, I think we're OK. If I understand correctly the del18O data for
1957-99 is based on
3 series and maybe more for this time. What I've been using is just one
of these, which goes
back 2K years - ends in 1995, and is smoothed.

The reason HadCRUT2v isn't exactly like Jo Jacka's series is the
variance correction
that gets done. This uses the residuals from a 30-year filter for each
month separately, reducing
the anomaly (from the filter) from a single obs in the box to the
infinitely sampled one. Values
further from the filter get reduced more than one closer. Upshot will be
that the annual series
will differ a little, but as the 30-year filtered series get added back,
the smoothed series will
get closer and closer to the original as you smooth more. All in a paper
in J. Climate in 1997.

Still doesn't explain why my r-values are as high as they are.
Smoothing must be the
answer and the period 1957-80.

I wasn't worried as I knew I'd used the latest one from last May.
Should I be worried?


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


[Message truncated. Tap Edit->Mark for Download to get remaining portion.]

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

No comments:

Post a Comment