from: Keith Briffa <k.briffaatXYZxyz.ac.uk>
subject: Re: [Fwd: esper]
to: Tom Wigley <wigleyatXYZxyz.ucar.edu>
sorry , but Examiners meeting kept me out of office all day - no easy answer - but
basically smoothed data and simple rescaling (not regression) against smoothed Man curve
produces this amplitude - been working a while to explain the anomalous cold implied by
Esper curve in first half of first millennium (but turns out they did not do what the paper
says) - now 2 other series show large amplitude also to do with selection and method of
scaling - can I call to discuss on phone as have to go now and no time to type this ? When
good time to call you ? Malcolm Hughes here tomorrow and I have to leave about 3 pm.also.
Have to rush now to pick up Kirsten and rescue car from garage
At 21:29 20/06/2005, you wrote:
Sorry to bug you, but I am being pressured on this. Have you had time
to think it over??
Thanks for your help.
-------- Original Message --------
Date: Tue, 14 Jun 2005 10:18:57 -0600
From: Tom Wigley <email@example.com>
To: Keith Briffa <firstname.lastname@example.org>, Phil Jones <email@example.com>, Tim Osborn
<firstname.lastname@example.org>, Caspar Ammann <email@example.com>
CC: Peter Foukal <firstname.lastname@example.org>
No doubt you have thought through this, but what particular choice of
input proxies makes the Esper curve in 1600-50 different from others
What is interesting is that Keith's curve is the only other one to show this.
Briffa and Esper also are similar for dips around 1350, 1470, 1820 -- so
I presume they have data in common that is not expressed in the other
curves. I note, however, that Briffa and Esper are opposite in the second
half of the 17th century. Any idea why there is this contrast with the early
I realize that Esper is made up of different bits -- but it does have some
very odd behavior. For example, if I lowpass his annual data, then the
amplitude of the low-frequ fluctuations that I get is noticeably less that
what he has (i.e., as in the second attached plot). I guess there is some
scaling done somewhere -- which of course is statistically bogus.
Since you have compared all these things before, I'm sure you have some
answers. It seems to me that the radical differences between different data
sets (notwithstanding the multiple reasons for differences) do not engender
confidence in any of them. Comparisons with model results do not make
things much better. These points seem to be glossed over in the literature --
please tell me if this is a false impression on my part (since I would not
want to propogate bad press in our review paper).
Many thanx for your help and insights.
Professor Keith Briffa,
Climatic Research Unit
University of East Anglia
Norwich, NR4 7TJ, U.K.