Thursday, April 11, 2013

EGU review part 1

As jules has already mentioned, it is EGU time again. It's been a bit hectic for the first three days but should calm down a bit now.

We didn't go last year so were anticipating it more than usual. Somehow despite being about half the size of the AGU meeting, the program manages to be packed with interesting things to see. The EGU app is also far better than the AGU equivalent, good enough for me to plan my program without reference to the web site.

As usual, we turned up a day early to get over jet-lag. I think this is still officially allowed by JAMSTEC rules, but anyway we can claim the meeting starts on Sunday with the registration and reception. We got to our favourite hotel (which has the best breakfast buffet imaginable) in good time and thanks to tripadvisor's recommendation, had possibly the best beer and ribs in Vienna. To be honest it was probably not quite as good as the Yokohama equivalent, but about half the price so we're not complaining. On Sunday, I'd planned to go on a guided tour of the Opera House, being one of the few attractions we've not yet visited, but the tours were off due to a ballet. So we spent the morning in the Albertina and then went to the ballet, as they had some v cheap tickets. Not really my cup of tea but it made for a relaxing day off. Then it was time for registration and a square foot of schnitzel (well it was round) to build up our strength for the week ahead.

Monday morning started with nonlinear time series analysis and probabilistic forecasting, which introduced some new (to me at least) methods which may be useful. I was talking after lunch on climate sensitivity, which seemed quite well attended. It seemed to go ok, there was a certain amount of scepticism but that was only to be expected. I find it amusing that on the one hand, people will argue that based on observational analyses, the GCMs have a worrying probability of excluding the truth (not just for sensitivity, although that's the poster child) but then when confronted with more observational data at the fringe of the ensemble, they insist that this doesn't really mean anything and the models aren't that bad after all! Anyway, it made for a lively afternoon. I'll probably post more about my talk later. My poster in this session was not until Wednesday. The Climate division organises things much better with posters on the same day as talks, which helps in facilitating discussion.

Tuesday morning was jules' newly invented "Past2Future" session, not to be confused with the EGU project Past4Future. Unfortunately due to a failure of scheduling it suffered from a clash with another paleo session, which must have hurt the attendance. But there were still quite a lot of people and there now seems to be much more interest in really using paleo modelling and data to improve and constrain future predictions - rather than just vaguely hinting at it, which has sometimes been the case. Then there was a big session on the response to orbital forcing, including a good medal lecture by Paillard. Jules' poster session (including our LGM reconstruction and related sensitivity estimate) was that evening. The poster sessions have been improved since we last attended, with much better lighting and the addition of moderate quantities of beer to add to the nasty wine, but we had to rush off a bit early to a concert I'd booked earlier.

Wed morning jules was speaking in a paleo modelling session, I was mostly sitting in the decadal prediction session. It still seems like there is very little skill in these methods, and in fact someone has just presented a simple statistical method that seems to generally outperform the initialised GCM systems. There is a "new" DePreSys system from the UKMO based on the newish model HadGEM2, but the presenter didn't mention that, let alone explain why, the forecast from the last one had fared so badly. Ed Hawkins showed that recent data lie outside the Smith et al forecast from 2007, which is a picture I was about to plot myself. He said he thought it might just be luck, but I need some convincing of that.

Later on Wednesday we had the data assimilation session. It's usually rather technical and I enjoy hearing about all the newest methods being devloped. Someone was talking about parameter estimation in a GCM, and came out with "and of course Annan did this ten years ago" citing a 2004 paper. Yes, that really is about as good as it gets for a minor scientist like me.

In the evening poster session we had two posters in different rooms, so jules and I took care of one each. Actually the sensitivity-related posters were only thinly attended (probably would have been better on Monday), so I also spent a fair bit of time upstairs in the paleo session. For some reason I'd not actually submitted my poster on the LGM temperature to any session. but we made the poster last week before realising this omission! However there were as always spaces due to withdrawal so it ended up being displayed on both Tuesday and Wednesday and attracted a surprising amount of attention. Then out the back of the conference centre and across the park for a huge Chinese banquet to celebrate the end of our official duties.


EliRabett said...

The lesson from air quality models is that simple statistical methods will always outperform ab initio ones except when something changes, so over a decade pick the statistical models absent a huge volcano or the sun winking out.

Anonymous said...

The best decadal forecasters were Tsonis and Swanson.