date: Fri Jun 4 16:18:46 2004
from: Phil Jones
subject: Re: hi, a quick question
to: Gabi Hegerl
Gabi,
The numbers Myles has may come from the paper Folland et al (2001) in GRL. In
this paper OA was used to get error bars and these were smaller than the ones I
got, by a factor of 1.5 to 2. I still think my numbers are nearer my gut feeling. They
may be slightly smaller , but not half. This 2001 paper also incorporated error
estimates on the basic obs, but I still have difficulty with some of these and how
they relate to different timescales. Some errors are random, but some are biases
so more systematic.
Happy to read something when you have it in a state to send. I think the GKSS
runs have used a too large a value for the range of change in solar output. Also the
volcanic forcing seems to get too much cooling. I suspect this occurs in winter as
well as summer and it should only be in summer.
Cheers
Phil
At 14:24 03/06/2004 -0400, you wrote:
Thanks Phil!
I don't know where Myles got his numbers, but your 97 paper says for decadal error std
dev
annual 0.068 for nh mean, and 0.05 to 0.06 for spring, summer autumn averages (first
half
of century). This is still an ok ballpark number, right?
Does the 0.033 number ring a bell with you? I am going to go with your paper,
unless I hear differently!
(for what I am doing, small changes
around the error estimate are not too bad, since we are just checking consistency of the
residual
between paleo and instrumental with our error assumption, which is quite an uncertain
test)
Thanks so much for your help, see you in Trieste!
I think there are some cases now that paleo reconstruction variance cannot
be straightforwardly compared with instrumental since its something a bit different.
Hans has done some work on that with his runs, too.
I'll keep in touch - and am keen on your opinion if you disagree with some of the
things said!
Gabi
Phil Jones wrote:
Gabi,
I too am not sure what I've let myself in for ! I will find out soon in Trieste.
I think you should try and compute what you want from the 1998 paper with the
three of us (you, me and Tim Barnett). You can then use the options about
autocorrelation used there.
All the above is from the instrumental record. Paleo data will lose some variance,
but
this is difficult to estimate as it will depend on timescale. For decades I would
expect
it would need enhancing by a factor in the 1.2 to 1.5 range. So 1.3 is about the number
I'd expect.
My paper in J. Climate in 1997 may be useful to look at as well.
Hope this helps. May have misunderstood some things.
Cheers
Phil
At 11:30 02/06/2004 -0400, you wrote:
Dear Phil,
Congratulations (or commiserations) on the observations CLA!
(got me for detn, and francis...hope this doesn't mean I can kiss science
myself goodbye for a while...).
I am checking some amplitude problems between paleo data and instrumental
with Myles, and we are trying a tls approach since I need a good amplitude for
my sensitivity. Myles' program can choose the noise level, so what we'd need
is an estimate of the 1 std deviation uncertainty in
decadal mean instrumental data 1880-1980, 30-90N annual (if 20-90N land growing
is available, too, that would be great... and 5-yr averages), but any number
would help.
Or should I try compute it from the error estiamte in the 98 paper?
What we found so far suggests that we should enhance the variance in paleo
reconstructions a bit (not hugely, eg for Tom's reconstructions factor 1.3, will be
more for less correlated data like Esper)
So far, we assumed that the 1 std dev uncertainty is 0.033.... but that may be global?
Greetings
Gabi
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
----------------------------------------------------------------------------
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
----------------------------------------------------------------------------