Monday, April 18, 2011

The EGU review

Well, jules has already covered the important bits. The hotel was great, and having an extra day at the start was a nice bonus, especially as the first conference day was a particularly busy one for us with my talk and a poster each to defend. It sometimes seems a bit boring visiting the same city over and over again - this is probably my 4th visit and jules' 5th - but on the plus side it makes the practicalities very easy to arrange, and the conference venue is hard to fault.

I started off with some decadal prediction, which still doesn't really work as far as I can tell. But people are having fun trying. Then there was stuff about volcanoes and geoengineering, which was mostly a bit too focussed on technical details to be really fascinating to me - plus I may have been distracted by thoughts of my (co-convened) session which was next. Two snippets of information I did get out of it was that the relationship between volcanic eruption strength and climate response is a rather complex one where seasonality plays a role, and that TiO2 may be a more efficient alternative to sulphate aerosols for artificial injection (though more research is needed....). I enjoyed my session, I thought there were a lot of interesting presentations which covered a range of approaches, both Bayesian and non-Bayesian, with recent obs or paleodata, all for the purposes of probabilistic climate prediction. People were mostly talking about recent work (that I was basically aware of) rather than future stuff though. Then straight into the poster sessions, which as usual were all run in the evening (at least for the climate division) along with the free paint-stripper.

Tuesday started with stochastic and statistical physics, which ranged from the purely theoretical to entirely ad-hoc. I'm never quite sure how important this is for climate modelling and prediction. Then Andrey Ganopolski gave a really good medal lecture. Sometimes these are rather routine surveys of past glories, but he included a lot of new stuff too relating to glacial-interglacial cycles and the carbon cycle. After lunch I spent the afternoon sessions running between the nonlinear geosciences session (mostly) with a bit of paloeclimate modelling. The former had lots of interesting time series analysis ideas including another medal lecture and a young scientist award lecture, both of which were good and gave me food for thought. The Young Scientist, Reik Donner, introduced some interesting ideas relating to complex networks, that I mean to look up some time. In the Lewis Fry medal lecture, Catherine Nicolis claimed to have invalidated the Maximum Entropy Principle, which I was very happy to hear, having decided some time ago to not invest any time in trying to understand it :-) I also enjoyed a talk about coastal erosion and the inevitability (or otherwise) of fractal coastlines. Jules had a talk in the paleo modelling session which was not news to me and I didn't see much of the other paleo talks but they all seemed to be having fun. By now we had finished all the real work so could enjoy the "10 years of open access publishing" party which somehow morphed into the President's Reception up on the secret 4th floor.

On Wednesday morning there was lots of last millennium stuff, mostly analysis of proxies with a little bit of modelling. Several people warning about the difficulties of inferring much about past climates based on limited data, and finally Martin Juckes gave a rather hammed-up deconstruction of the McShane and Wyner paper. I'm surprised he or anyone else thought it was a useful thing to do, really, and it was a bit toe-curling in places. At lunchtime the Climate Division had its business meeting at which everything passed off smoothly. No Exxon controversy to debate this time! I wonder what happened about that? The afternoon was a bit blank so I went to see part of Peter Challoner and Dan Cornford talking about emulators. These things are a great tool but sometimes somewhat oversold, since (as I have pointed out multiple times) if you only want to characterise the uncertainty of the output(s) of a model with multiple uncertain inputs, you can do a pretty good job with a modest O(100) simulations irrespective of the number of inputs.

I had another poster in the evening, relating to our own Last Millennium work. That's very much work in progress and I got a lot of ideas and motivation from the other talks and posters so hope to do some more on this soon. In fact I'd been meaning to do stuff in March but the earthquake and power cuts put paid to that - I ended up just taking the same poster I'd presented in Kyoto last December. Lucky I had it really. Work was still not over, we met Dan and Emma for some discussion of secret collaboration before adjourning to the local Chinese for a decent dinner.

Thursday had the highlight of the Inverse Problems and Data Assimilation session. Particle filters are all the rage these days. There is a basic and well-known problem of dimensionality (or equivalently ensemble size) but a number of tricks are being developed to make them work in practical applications. It is funny to see how the field has come full circle back to nudging, which was the first attempt at data assimilation over 50 years ago before there was any real theory to underpin it! Last session of the day was about feedbacks, Ray Bates gave a talk which initially seemed a little provocative and indeed did provoke some audience reaction. He argued, using a simple model, that a net positive forcing could generate a net negative temperature response even in a stable system. This initially sounds implausible, until you realise that this situation only requires a mild negative forcing over a region of high sensitivity, combined with a slightly stronger positive forcing over a region of rather lower sensitivity (having a relatively low coupling between the regions also helps). And indeed on checking his paper, this is really all he's done, though he's also undertaken a very detailed analysis. I suppose this could have some relevance to energy-balance analysis of the response to forcing with strong spatial patterns, such as negative aerosols over land and positive GHGs everywhere, but I can't help thinking that the presentation made it seem more complex and counterintuitive than it really is. Another prize lecture, this time by Andreas Oschlies, finished off the day. He was arguing strongly for more quantitative assessment of marine biogeochemical models, which seems like a good idea to us.

By Friday we were pretty worn out, so it was no bad thing that there were not many sessions of direct importance to our work. However, a special emergency session about the Tohoku earthquake and tsunami had been arranged at short notice, so we spent the day learning about what had happened. Apparently it has refuted most of the major theories about earthquake size, being far bigger than expected for its situation. There were even preliminary comments about the large aftershock that had happened the previous day, the location of which was also not in accord with expectation. The tsunami was about 15m most of the way along the coast, and had been measured to run up as high as 38m at one point! The following panel discussion also covered the nuclear problem, and I asked if they had any comments on the hysterical over-reaction to this. They seemed to accept the premise (at least as far as Europe and the USA was concerned) and thought that some clearer and more authoritative Europe-wide scientific assessment of the (low) risks could have helped. We also submitted a poster to this session, about which more later...

So, that was the end of it. We struggled along to the Convenors' party for a good feed and collapsed into bed. Unusually, this year we didn't even skive off any sessions to do some touristing, and only turned up a little late one morning when the 8:30 talk didn't seem that interesting. Overall, there was plenty to do and the whole conference re-confirmed our view that it's generally rather better than the equivalent AGU event. I haven't even talked about the people we saw doing all sorts of interesting things relating to carbon capture and storage and renewable energy. In contrast, the AGU was full of people hand-wringing ineffectually about "communication". On the other hand, San Francisco food is certainly better overall (Hotel Stefanie breakfast buffet notwithstanding). So it's a tough call :-)

The abstracts and program are on-line here, by the way.


Steve Bloom said...

Thanks for the thorough post!

Toe-curling in what sense?

James Annan said...

Oh it was just a bit too full of sarcastic commentary and nit-picking...based on the abstract I was expecting something of wider interest. I'm sure you know the paper has already been widely criticised (see here for coverage) and although I haven't kept abreast of the details I don't think there was anything much new in his talk.

The rest of the session was full of people actually getting to grips with the difficulties of reconstructing past climates, and was rather interesting.

Steve Bloom said...

Maybe he thought it was important to rub things in. I have the impression that more people are feeling that added derision and mockery may help as contrasted to the common past approach of being polite and even encouraging.

James Annan said...

But they (almost certainly) weren't there!

Steve Bloom said...

Oh, I'm sure you're right, their being North American and not earth scientists. But we can hope that getting material like this on-line will help discourage future such efforts.