How should we model complexities in human-environment systems

Up until now, we have recognized that human-environment interactions are complex through those diagrams where every component is connected to everything else. The intuitive next step is how to model complex systems in the real world. Among tools and approaches, I found agent-based modeling great potential for such tasks. Agent-based modeling (ABM) is a bottom-up method that simulates actions of individual “agents” (e.g., persons or households) and their interactions with the environment to produce the aggregated macro-level patterns and processes. When dealing with complexities in human-environment interactions, previous studies often pointed to ABM as a powerful tool for exploring such complexities. It seems to me, however, that there are not many working examples that found non-trivial differences between findings from the use of ABM and those from traditional approaches (e.g., regression-based analysis). Could anyone provide working examples of using ABM with non-trivial findings of explaining human-environment interactions?

This is a great question, to which I unfortunately have little in the way of answers.  The topic is close to one of 10 priority research needs for sustainability science identified at a recent workshop that Prof. Simon Levin of Princeton and I ran at the request of NSF, and that was attended by several participants in this seminar.  The full workshop report is listed as one of the readings for this seminar's last session on "Grand challenges....".  Here is the relevant text from the Executive Summary:

"How can theory and models be formulated that better account for the variation in types or trends of human-environment interactions? Many properties of human-environmental systems can be adequately captured with conventional statistical or system-dynamic models. But the complex dynamics, inter-sectoral and multi-scale interactions, emergent properties and uncertainty that characterize many of the human-environment systems most relevant to sustainability concerns have proven very difficult to deal with using such approaches. Advances in agent-based and network approaches to the modeling of complex adaptive systems offer promise of doing better, as do several approaches to the qualitative analysis of non-linear systems and the development of interdisciplinary, multi-scale scenarios. But that promise has not yet been fulfilled in more than a handful of cases. Part of the problem is that most empirical scientists who understand the causal structure of human-environment systems are not expert in the new modeling approaches, while modeling experts seldom have access to more than “toy” systems and simple data sets. The workshop concluded that much could be gained from a concerted effort to compare the ability of a suite of promising modeling approaches to shed light on a few well-understood human-environment systems. Reciprocally, sustainability science would certainly benefit greatly from developing its own suite of “model systems” to play the roles that stalwarts such as Drosophila, E. coli, and lynx-hare interactions have played for other sciences. Such model systems – including long term, spatially explicit data sets of key variables, a summary of key causal relationships, and a catalogue of models already developed for them – would attract the attention of new families of complex system theorists and modelers to the field. A good start might be made with fish-stock/fishery fleet systems of the sort recently reviewed by C.W. Clark (2006) and the lake/agricultural pollution systems developed by Carpenter (Brock and Carpenter 2007)."

I would be especially interested in others' ideas on this, either here or in the discussion of the "Core questions" session on Dec 6.

Also take a look also at the 'Events' posting on this site, where a workshop on  related topic is listed.

Previously Xiaodong Chen wrote:

Up until now, we have recognized that human-environment interactions are complex through those diagrams where every component is connected to everything else. The intuitive next step is how to model complex systems in the real world. Among tools and approaches, I found agent-based modeling great potential for such tasks. Agent-based modeling (ABM) is a bottom-up method that simulates actions of individual “agents” (e.g., persons or households) and their interactions with the environment to produce the aggregated macro-level patterns and processes. When dealing with complexities in human-environment interactions, previous studies often pointed to ABM as a powerful tool for exploring such complexities. It seems to me, however, that there are not many working examples that found non-trivial differences between findings from the use of ABM and those from traditional approaches (e.g., regression-based analysis). Could anyone provide working examples of using ABM with non-trivial findings of explaining human-environment interactions?