cc: "Martin Juckes" , , "Keith Briffa" , date: Fri, 25 Feb 2005 14:53:45 -0000 from: "Myles Allen" subject: Millennial temperatures to: "Eduardo Zorita" , "Jan Esper" , "Anders Moberg" Dear Eduardo, Jan and Anders, I have been asked (on rather short notice) to put together a team of international experts to perform an inter-comparison and evaluation of reconstructions of temperatures over the past millennium on behalf of the Dutch government. Essentially, they have allocated 80keuro to this task and are soliciting bids -- my understanding, reading between the lines of the e-mails from the project manager, is that we may be the only bidders, but if there are rival bids and if you happen to be putting together another one, then please accept my apologies, and rest assured I won't take offence if you simply ask me to go away. The resources offered are not large, and they would like deliverables on a six-month timescale, so we clearly can't break a lot of new ground with this. What I thought would be useful would be to put together a common piece software in an accessible language like IDL or Matlab to take a common set of input proxy series and reproduce reconstructions using several different algorithms. I'm sure we won't be able to do this for all the algorithms out there, but if we can do this even for just a couple, then we can put down a framework for other authors to make their own input to. This way we can document exactly whether reconstructions differ because of inputs, processing philosophy, or simply tunable parameters. Gabriele Hegerl and I did something like this (with Nathan Gillett) pulling together different detection and attribution results in the build-up to the IPCC TAR, and although it sounds rather mundane, it was actually incredibly valuable for the TAR because it meant the chapter authors understood the origins of the differences between the studies' results in a nice, impersonal environment. We would aim to make this code publicly available to accompany a review paper at the conclusion of the project. A nice extension would be to apply the various algorithms to some pseudo-proxy data from models, but this would depend on us not tripping over other pieces of research currently under way. My colleague Martin Juckes (at the Rutherford Appleton Laboratory) is pulling together the current version of our outline bid, which currently involves myself, Gabriele Hegerl (with whom I have been working on another regression based reconstruction with Tom Crowley), Keith Briffa and Tim Osborn (who have also been looking into regression-based approaches). Would you be prepared to participate in this exercise, specifically with Jan and Anders contributing the details of their respective reconstruction algorithms and Eduardo advising on the details of the MBH algorithm? At present we were envisaging salary support for Martin and Gabi, and travel support possibly supplemented with a small consultancy fee if regulations allow for everyone else. At the very least, this should fund us all getting together for a practical workshop sorting out the code -- I'm sure we can choose somewhere nice for the purpose! Looking forward to hearing from you (please copy Martin in on any response) and congratulations to Anders on the impact of your paper: well deserved. Yours sincerely, Myles Allen Climate Dynamics Group Atmospheric, Oceanic and Planetary Physics Department of Physics, University of Oxford Tel: 44-1865-272085/925 Fax: 44-1865-272923 E-mail: myles.allen@physics.oxford.ac.uk