Wednesday morning was the main reason I was here - the symposium on "Agent-Based Modelling of Land-use Effects", organised by [http://www.macaulay.ac.uk/staff/staffdetails.php?garypolhill Gary Polhill].
It started with Gary giving us a quick run overview of ABM "so the rest of the presenters don't have to". A couple of references of interest were:
* [http://jasss.soc.surrey.ac.uk/12/2/2.html Nikolai and Madey 09] - a recent review of ABM toolkits
* [http://dx.doi.org/10.1016/j.geoforum.2007.05.005 Parker et al 08] - talking about dealing with complexity in human land-use models
He also pointed at some journals: [http://jasss.soc.surrey.ac.uk/ JASSS] being the main one, also [http://www.ecologyandsociety.org/ Ecology and Society] and [http://www.elsevier.com/wps/find/journaldescription.cws_home/422921/description#description Environmental Modelling & Software]
Next, Derek Robinson gave a talk about coupling ABM with the BIOME BGC biophysical model. It included some interesting results about which bits of forest to protect for the greatest conservation benefits. Pongchai Dumrongrojwatthana gave a truly inspiring talk about using [http://commod.org/ Companion Modelling] to work with herders and forresters in northern Thailand. Overall, the modelling process, which makes heavy use of roleplaying games with the stakeholders, provided mediation between the two groups, and inspired a lot of behavioural change in the stakeholders. (Anecdotally, other ComMod projects have found the need to go back for follow up work, because of the unexpected level of change which occurred after their visit!). Wenwu Tang talked about implementing huge (10^9) ABMs using supercomputers and GIS-aware middleware (GISSolve). David Bennet then talked about capturing the complexity in Land Use Systems: spatiotemporal data structures to represent the traces of many runs, and enable searching for bifurcations, convergences and regions of chaos within the simulation space. Also, the need to look at the provenance of the runs.
Tom Evans talked about balancing social and ecological complexity in ABM, looking at reforestation. James Millington talked about his model which looks for the possibility of human interference on cells, and runs a decision module if so, or a biophysical model otherwise, looking at the effects of wildfire. He'd tried to validate the results of the models by talking to the stakeholders and asking them to draw maps of their predictions, which was partially successful, except they were mostly interested in identifying their own farms, and seeing what would happen to them. Innocent Bakani's talk (given by Gary) looked at simulating cap and trade schemes for agriculture, and found that when there was a high degree of emission reduction, the price of permits was likely to fall, removing the incentive for compliance.
Alessandro Gimona talked about using ABM and Metacommunity models to look at the effects of Agri-envrionmental policy on species conservation, with some stylised land uses and species. It looked like "clustered rewards" were quite important, where landowners are given rewards if actions are carried out on blocks of squares. Randall boone talked about the process of coupling an ABM with a biophysical model, using a multi step process, where habitat suitability, spatial distributions, energetic/metabolic factors and mortality are passed back and forth between the ABM and BPM. Scott Heckbert presented a model of cap and trade for fertiliser application; it was highly participatory, and he showed rational action from his 10 year old daughter in the face of fertiliser being washed away by rain. He gave a whole bunch of great references, which I'll put up here when I have time.
In the synthesis at the end, there were a collection of themes and questions that came up, some of which were:
* How can we make our work most useful to others?
* How to deal with scale issues?
* Creating large scale simulations
* Edge effects
* What to do around the assumption of rationality?
* Approaches to take to uncertainty
* The need for descriptive frameworks (but see ODD!)
* how do we keep things grounded in social sciences? what can we take from that literature?
* How to use diverse knowledge sources
* What can we take from AI techniques, particularly in the realm of learning and autonomy
* What do we want from social scientists? Answers included: wider contexts, identifying missing processes, help communicating our results to stakeholders
* how to we couple with more traditional models? The value of traditional models which can break out into ABM at certain points.
* Easy ways to couple models (see the [http://www.openmi.org/reloaded/ OpenMI] project)
* How to find other people's models (see [http://ispoc.cscs.lsa.umich.edu/ SPOC] and [http://www.openabm.org/site/ OpenABM])
And that was it. We all had sushi, and sat in the hot-tub on the roof in the snow.