Been quiet on the blog, partly because I've been doing so much real talking that I've not had much energy (or need...) for this outlet, but also because for the last week or so we have been on the road. Participants at the Isaac Newton Institute are encouraged to give seminars around the UK and we have taken advantage of this to visit a few different places, remind people that we still exist, and present our recent work on the multimodel ensemble.
First we went to Lancaster, which by complete coincidence is close to where jules' Dad lives, and much more importantly is where a former colleague who was in Japan now works, so we got him to arrange an invitation. The growing
LEC seemed to be on the up, and the audience were very perky - as well as interesting comments from environmental scientists, there were some attendees from the nearby
forecasting centre (part of the management school, so that's forecasting in the J. Scott Armstrong mold) and we had some interesting conversations. They seemed to appreciate that people were working meaningfully on climate model validation and testing, and we appreciated their interest and comments.
Then we went down to Bristol where we know a few people and have an ongoing collaboration with former-prostetnic reader Andy (he now reads for real). They have a big and very well regarded paleoclimate group so we focussed a bit more on jules' use of paleoclimate data for validation.
Then on Monday we were in Exeter at the Hadley Centre/UK Met Office. This was the best attended seminar of the three, with over 60 people present. I enjoyed the interjections from the NWP people (or perhaps I should say Persson) saying that (a) the truth-centred paradigm was obviously rubbish and (b) they had known that the ensemble mean was better than the ensemble members (and why) since the birth of ensemble methods, if not sooner. I certainly don't think the Hadley Centre people are particularly culpable in this (in fact AIUI they haven't really written on either subject) but given the common basis of climate and weather prediction, the cultural/personal barriers that have enabled these questions to remain unanswered all around the world for so long must be regarded as rather worrying. I recently came across the following in
a 2009 PNAS paper: "Although
MM's [multi-model mean] superiority has been found in previous studies focusing on the mean climate, the reasons for this have not generally been elucidated." Well, they have now!
Since then we have been staying at Exeter University for a workshop on probabilistic methods in climate prediction. These sort of things always involve a certain amount of talking in circles (or at cross purposes) but also some interesting interactions which have proved very useful, at least to me. Whether something coherent and concrete comes out at the end of it is another matter, but we've only had one day so far and there's another day and a half to go before we head back to Cambridge.