cc: dpierce@ucsd.edu, JKenyon , Myles Allen , Nathan , Phil Jones , David Karoly , Knutti Reto , Toru Nozawa , Tom Knutson , Doug Nychka , Claudia Tebaldi , Ben Santer , Richard Smith , Daithi Stone , "Stott, Peter" , Michael Wehner , Francis Zwiers , Hans von Storch date: Wed, 27 Aug 2008 09:33:33 +0100 from: Gabi Hegerl subject: Re: comments on AR5 experimental design - reply by Aug 28 to: tbarnett-ul@ucsd.edu Thanks Tim! We'll have another round later, confirmed by Tim, when we discuss storage and documentation - probably should try before WGCM meeting so that David can present results. the 'near term prediction' is a mip all by itself, so there will be some guidance coming up hopefully! In terms of ensemble size: for the stuff I was involved in, even one run from a model was good since it increased the overall ensemble size for multi model means and estimates of variance - did you analyze models individually? I would be keen to hear from the group: is say a single 20th c run, single natural only run, single ghg run a) useless b) much better than nothing? | vouch for b) for things I was involved in but it would be good to know for which applications its a! Gabi Tim Barnett wrote: > hi gabi..in real haste.....people will use the AR5 data set for impact > studies no doubt about it. so what will they find when they jump > in....same as we did trying to do the western D&A work with AR4....a very > disparate set of numbers. > 1.some models don't give the data one would like. > 2.some models have only 1 realization...which makes them useless. we > found that with multiple realizations one can do statistics with ensemble > techniques which give a lot more statistical power. suggesting 10 member > ensembles. with less the S/N can be small...e.g. we could not use the > GFDL runs very well as they were so noisey and had few (5) realizations) > 3. daily data is required. storage is cheap these days so at least daily > data for order 100 years is desired. otherwise it is finageled a la the > current downscaling methods (save one). > 4. the 20th century runs need to go to 2015 as suggested by IDAG. we had > to stop at 1999 and lost 8 years we would well like to have studies. > 5. some of the variables we needed to compare with satellite obs were > largely missing, e.g. clouds information. > 6. to Mike's point....just what data is going to be saved? > 7. i hope potential users of the data aside from the modeling groups get > a say in what is archived. we are to the point now where policy makers > want our best guesses as to what will happen in the next 20 years. the > people who will make those 'guesses' are most likely not in the major > model centers. > > I invite David Pierce to chip in here as he spend alot of time in the > details of the data sets and associated problems. > > sorry to be so hasty but such is life at the moment. best, tim > > > > >> Hi IDAG'ies, >> >> As you probably know, a proposal for the AR5 experiments is being >> circulated in the moment, with comments due by September 1. This will >> then be presented at the working group for coupled modelling (WGCM) >> meeting in Paris, which David Karoly will attend. >> Peter Stott and I discussed the draft when I visited last week, and we >> drafted a response and suggestions from IDAG (attached) Please let me >> know if you are ok with this (if I dont hear back I assume you are), >> if you suggest changes and if you want us to add another topic/concern. >> >> I would need this by next thursday to add it to a comment 'from IDAG' >> to be sent in time, and then hopefully David can present this also in >> Paris at the WGCM meeting. >> >> hope you all had a nice summer, and still remember our next meeting in >> planning, and your IDAG tasks :)) >> >> Gabi >> >> >> p.s. we were wondering also about forcing, and if the forcing issue >> (how stored, synchronized?) should be added. However, given even some >> 'rich' modelling groups worry about getting the mandatory experiments >> through we should however not hope that groups will run more than 1 >> single forcing set for the 20th century, and arguments against >> synchronizing are that its not feasible for many forcings (eg >> aerosols) and that we loose quite a bit of information if only a >> single, for example, set of solar forcings were used and with this >> open the AR5 up for criticism. Ideally, of course, one center would >> systematically explore all the forcings - but I am not sure somebody >> is planning to do this - in that case, a common set of 20th century >> forcings may be an advantage. Based on some EU project, forcings are >> synchronized for some European modeling centers - we could draw >> attention to that if you feel strongly about this...anyway, I hesitate >> to start a discussion about this... >> >> >> -- >> Gabriele Hegerl >> School of GeoSciences >> University of Edinburgh >> http://www.geos.ed.ac.uk/people/person.html?indv=1613 >> >> -- >> The University of Edinburgh is a charitable body, registered in >> Scotland, with registration number SC005336. >> >> >> > > > -- Dr Gabriele Hegerl School of GeoSciences The University of Edinburgh Grant Institute, The King's Buildings West Mains Road EDINBURGH EH9 3JW Phone: +44 (0) 131 6519092, FAX: +44 (0) 131 668 3184 Email: Gabi.Hegerl@ed.ac.uk The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.