Before I retired (last Oct) I had been involved in two large scale complex adaptive models of human/ecosystem interaction, both of which included primary productivity. Our models were agent-based models comprised of agents which represented entities ranging from "individuals" through super-individuals (schools/shoals) through to regional populations of marine organisms and "field" representations, as well as the "human" side of the simulation (fisheries management, fishers, industry, tourism, etc). Many of the agents were designed to react to the state of other agents and the state of system variables, as well as to interact in exchanges (booking tourist activities, for example).
The link below is a sort of table of contents for the reports associated with the more complex model. Also google for "NWS-JEMS" and pick the ones with authors like Randall Gray (me), Elizabeth Fulton, or Rich Little.
Hope this helps,
Randall
Oops, doesn't seem to want to let me add another link, so here is a Tech Report on the older study:
That's great, I need to look into in detail. Shame I missed your reply. Had one of our young minds with Beth Fulton all of Feb 2014 (Artur Palacz). Hope you had a chance to meet.
I was publishing the question above to see if we can find colleagues out there that might be having parallel headaches with ANN validation through observations, so that these guys learn from each others headaches.