Monday, 30 June 2008


Another AMEMR conference ticked off (23rd to 26th June). As conferences go, I've never really been to anything quite as singularly focused on modelling as the AMEMR series. Most other conferences slip modelling in alongside more observational work, or at most have a specifically modelling session, but AMEMR is wall-to-wall modelling. Admittedly with only a single session (~18 talks per day for 4 days).

So it should be something like modeller's heaven, right? Well, at best a mixed bag. The first AMEMR conference in 2005 has left little impression on me beyond a loathing for super-narrow regional modelling (e.g. a dreary discussion of a bog-standard model in a boring provincial region). If a theme arose out of it, then it was a questioning of the justifiability of the complex end of marine ecosystem models. Up until that point, the continual increase of model complexity was seen as natural and right, with nary a second thought given to whether the resulting models could be robustly validated. Furthermore, when it came to validation, simply showing a side-by-side comparison between a highly-abstracted observational field and a carefully (= dubiously) colour-intervalled model field was a fairly acceptable strategy. Some speakers didn't even do that, and were content to simply show model output as if it were gospel truth.

While AMEMR 2008 also suffered somewhat from regional talks, and from some weak model-data comparison (of which, my poster was something of an example!), things were generally better. There were echoes of the last meeting's anti-complex model sentiments, but the absence from the conference of one of the major offenders on this count kept this particular "discussion" more low-key. An interesting development this time was the inclusion of more higher trophic level modelling. In aquatic circles, this specifically means fish models. Which, given the socio-economic importance of fisheries made for more anthropocentric sessions towards the end of the conference.

Although only a few fisheries modellers made an appearance at AMEMR, they had a couple of good talks and two excellent keynote speakers. The best research speaker summarised what must have been a massive piece of work with an emsemble size of thousands of individual simulations (for reference, Sarah Gaichas). Applying a generic fisheries model of the North Pacific to three specific locations, she was able to show the consequences of perturbations of different fish species on the whole food web, and to demonstrate that the three locations behaved differently to management despite the imposition of the same strategy everywhere. A real tour-de-force. The two keynote speakers both brought home the necessity of modellers to take in the bigger picture, and not to lose themselves in the detail. While they recognised the importance of advances in model realism, the socio-economic problems that they find themselves dealing with simply cannot wait until all of the modelling details can be resolved. They advised that modellers need to bear the wider context in mind, and to explore model behaviour (or "have fun", as one of the speakers put it; Beth Fulton) while recognising the limits of their models. Stirring stuff.

In summary, and to paraphrase terribly, there were a couple of take-home messages from this year's AMEMR:
  • From the biogeochemical modelling end: Be aware of model limitations and perverse/pathological details
  • From the fisheries modelling end: Be aware that the world cannot wait for perfection, and that second best models may have to suffice

On a more social note, I was able to catch up with both MEB and SME on several evenings. Mark, as ever, was doing sterling work with his size-based model. Meanwhile, Steve retains his animosity to models because of their inadequacy as a means of representing reality. We had some philosophical fun thrashing that one about.

No comments: