Monday, April 9, 2012

3212.txt

cc: Thorsten Kiefer <thorsten.kieferatXYZxyzes.unibe.ch>, Christoph Kull <christoph.kullatXYZxyzes.unibe.ch>, Heinz Wanner <wanneratXYZxyzb.unibe.ch>, Nick Graham <ngraham@hrc-lab.org>, Tim Osborn <t.osbornatXYZxyz.ac.uk>, Keith Briffa <k.briffaatXYZxyz.ac.uk>, philip.brohan@metoffice.gov.uk, Francis Zwiers <francis.zwiersatXYZxyzgc.ca>
date: Wed, 09 Aug 2006 16:16:48 +0100
from: philip.brohanatXYZxyzoffice.gov.uk
subject: Re: Challenge
to: Caspar Ammann <ammannatXYZxyzr.edu>

Caspar et al.

A few comments on the proposal.

1) Reconstructions of the real climate differ widely in their results -
they can't all be correct, and it's likely that all existing
reconstructions are in error in some important aspect. I think the
challenge needs to propose a reconstruction task that is sufficiently
difficult that we can reproduce this state of affairs in the challenge
world. In other words, the purpose is not to find which reconstruction
method is best; the aim is to allow each team doing reconstructions to
find out what the limitations of their method are.

2) Making realistic proxies will be difficult and time consuming, also
it will be difficult to make many different sets of proxies where the
errors are realistic but independent of other sets, so we should limit
the number of 'full blown' challenge experiments done. (Perhaps to 2: 2A
and 2C?). Also, it needs to be easy for people to participate in the
challenge, which means the number of reconstructions should be limited.

3) It's vital to separate the reconstruction group from those making the
pseudoproxy data, but it's less important to separate the model group
from the pseudo-proxy group. combining these two groups would make the
project easier to run and reduce the chances of a mistake due to a
misunderstanding between modellers and proxy people.

4) I like the proposal as it stands, and if we can find the resources
we should do it as described. But I'm also tempted by a simpler
approach:
Make just one blind model from set 2A, and a set of full-complexity
proxies for it. Invite reconstructions based on these proxies. With a
bit of luck the reconstructions will differ wildly from one another and
significantly from the true result. Then distribute the true model data
and information on the errors introduced into the proxies - each
reconstructor can then work out where their method went wrong and try to
improve it. When enough people have done this, make another blind model
from set 2A (or 2C) and repeat the whole process.

Regards,

Philip

On Wed, 2006-08-02 at 17:28, Caspar Ammann wrote:
> Hi everybody,
>
> here the further updated draft of the Challenge description so nicely
> named by Mike "PR Challenge" (Paleoclimate Reconstruction Challenge)
>
> Again, this is a draft and I welcome any comments and suggestions. I
> added a time table and questions for discussion at the end. Please let
> me know what you think about this setup. Particularly the funding we
> would have to discuss rather quickly and see what we need and how to
> approach the agencies. Any input on that side would be highly welcome.
>
> Caspar
--
Philip Brohan, Climate Scientist
Met Office Hadley Centre for Climate Prediction and Research
Tel: +44 (0)1392 884574 Fax: +44 (0)1392 885681
Global climate data sets are available from http://www.hadobs.org

No comments:

Post a Comment