Oh yes, I forgot to mention that one reason for our slumming it with the miserable slice of pizza on Tuesday night was that we went to Eric Wolff's excellent medal lecture. Anyway, Wednesday kicked off with paleoclimate modelling, which included Pascale Braconnot's medal lecture about PMIP/PMIP2, and also a lot of stuff well outside the scope of those projects. The paleo bunch seem very friendly and cooperative and I enjoy their meetings. I think they have (at least historically) felt a bit overlooked by the 100y prediction IPCC juggernaut (and maybe it has something to do with the unusually large proportion of women involved?). However there are now plenty of people explicitly pointing out the value of paleo simulations in helping toward future predictions, something that we also try to contribute to (even if not everyone is fascinated by it). We managed to scrounge another free lunch by being in the right place at the right time.
After lunch I caught a useful talk on limitations of particle filters, which I'm currently playing with. Then I enjoyed an amusing take-down of the Roe and Baker paper, "Why is Climate sensitivity not doomed to be unpredictable?" This very much had the flavour of a rejected comment (of course I'm well acquainted with those). It pointed out in excruciating detail that R&B's main result that the "uncertainty" in climate sensitivity would not decrease much even if the uncertainty in 1/sensitivity is reduced, was based on a very unorthodox and specific definition of "uncertainty" and was not actually true for any reasonably standard usage of the term. In detail, they were defining "uncertainty" as the probability of sensitivity lying in the interval 4.5-8, (apparently under the belief that values above 8 were nonsense anyway, even though their inverse gaussian assigned substantial probability to that region). Of course, the blogosphere got there first, although that linked post doesn't explicitly talk about the 4.5-8 weirdness but it does point out the odd truncation of the distribution (which contradicts the premise of the research, and is maybe a bigger problem than the odd analysis). Of course R&B is published in a peer reviewed journal, and the various critical analyses are not, which makes the former correct, I suppose...
Talking of Stoat, at some point I noticed he had a poster up. OK, it wasn't really his poster, but he was listed as a co-author. I wonder if his current employers know he is still moonlighting as a climate scientist? Obviously he's finding it harder to give up than he thought...
After lunch I caught a useful talk on limitations of particle filters, which I'm currently playing with. Then I enjoyed an amusing take-down of the Roe and Baker paper, "Why is Climate sensitivity not doomed to be unpredictable?" This very much had the flavour of a rejected comment (of course I'm well acquainted with those). It pointed out in excruciating detail that R&B's main result that the "uncertainty" in climate sensitivity would not decrease much even if the uncertainty in 1/sensitivity is reduced, was based on a very unorthodox and specific definition of "uncertainty" and was not actually true for any reasonably standard usage of the term. In detail, they were defining "uncertainty" as the probability of sensitivity lying in the interval 4.5-8, (apparently under the belief that values above 8 were nonsense anyway, even though their inverse gaussian assigned substantial probability to that region). Of course, the blogosphere got there first, although that linked post doesn't explicitly talk about the 4.5-8 weirdness but it does point out the odd truncation of the distribution (which contradicts the premise of the research, and is maybe a bigger problem than the odd analysis). Of course R&B is published in a peer reviewed journal, and the various critical analyses are not, which makes the former correct, I suppose...
Talking of Stoat, at some point I noticed he had a poster up. OK, it wasn't really his poster, but he was listed as a co-author. I wonder if his current employers know he is still moonlighting as a climate scientist? Obviously he's finding it harder to give up than he thought...
1 comment:
> moonlighting as a climate scientist
I think they'll recognize that's product development; as those little transmitters get small enough and cheap enough, they'll be very useful.
Just picked up Sterling's _Caryatids_ at the library, and it starts along those lines. He sez elsewhere:
"... I’m trying to write some science fiction that was literally inconceivable in the 20th century. .... a lot of internet-of-things stuff ... used to be pretty inconceivable. It’s also got a lot of the down-and-dirty stuff that we’ve learned lately about climate collapse.
What was the hardest part of writing it?
That was the part when I discovered that I knew so much about ubiquitous computing that people wanted me to teach design school.... Nowadays I teach lots of things in lots of different design schools.....
The thing that’s cool about design students is, when you tell them about something goofy and speculative, they immediately want to build one...."
http://futurismic.com/2009/02/11/interview-bruce-sterling-on-caryatids-viridian-and-the-death-of-print/
________
Small radios for climatology, coming soon.
Post a Comment