Agent-based models of coupled social and natural systems

Jiaqi Ge and Gary Polhill

fig4Agent Based Models provide the ability to simulate the overall behaviour of a complex system from simple components which interact (“agents”).

They are useful for exploring the logical consequences of policy assumptions with dynamic implications that are beyond the capacity for human reasoning or tractable mathematical analysis. For example the inter-connectedness of ecological and social systems means that policy makers should not only consider the impact of a policy on the system directly affected by the policy. They should also consider the indirect effect of the policy on other related systems. A locally implemented policy can trigger strategic responses from neighbouring areas and led to unintended consequences in the neighbouring areas.

Applications reviewed include how :

  • Individual farmer behaviour and interactions can be modelled explicitly.
  • Agricultural landscape can be modelled in a spatially explicit way.
  • Landscape planning can be simulated and tested spatially.
  • Interactions between neighbouring land patches are modelled explicitly. Spatial correlation and network effects can be studied.
  • Transport can be modelled in a spatially explicit way. Planned construction in the future can be simulated and tested explicitly.
  • Individual route choosing behaviour can be modelled explicitly.
  • Impact of culture or policy on behaviour can be included.
  • Systemic effect such as how the level of congestion and pollution emerge from individual route-choosing behaviour.
  • Human decision making can be explicitly modelled.
  • The whole environmental system can be included to study the systemic effect of management practice.
  • Any side effects and unintended consequences may be identified.
  • Hypothetical scenario analysis can be conducted for different management systems and policies.

Recognised difficulties include :

  • Detailed Expert knowledge regarding the system required
  • Could be computationally expensive
  • Might encounter problems with integrated modelling approach
  • Detailed information may be required
  • Spatial information required could be computationally expensive
  • Microsimulation of transport has been shown to be very computationally cxpensive.
  • Information on individual behaviour required

Back to Summary / Visit