Modelling ecological systems using Simile

This possible project involves working with the Simile simulation software, which has been developed by an Edinburgh University spinout company Simulistics.

Simile, in common with similar software such as POWERSIM, iThink and Stella, provides a visual interface to continuous-systems modelling using the System Dynamics (stock-and-flow) notation, but (unlike the other software) provides sophisticated features for handling disaggregation (e.g. spatial grids, populations of individuals, etc). The graphical system dynamics notation allows the underlying differential equations defining the simulation to be defined in an intuitive graphical way and then solved using inbuilt differential equation solvers. Simile generates C++ code for simulating model behaviour (hence is computationally efficient), and can be driven by a scripting language (TCL).

Simile's spatial modelling facilites make it well suited to modelling ecological systems, but it is also capable of doing the things other system dynamics packages can do.

Several projects are possible using this software.

• A. The Scottish Crop Research Institute in Dundee is developing a model in Simile of a crop ecosystem. The model they are building is very detailed with each plant, weed and invertibrate being represented individually. For different random seeds for the simulation the model will behave differently (probably slightly). The Crop Institute is interested in optimizing some aspects of the problem e.g. finding the optimum partitioning of energy between the crop and the non-crop ecosystem.
  • A.1 One version of this project would use the exisiting stochastic model as a black box and find a way of optimizing the average performance.
  • A.2 A second version would involve building a much smaller deterministic aggregate model and calibrating this to get as close as possible to the average performance of the stochastic model using a parameter estimation tool built incorporated in Simile and then optimize using the smaller model. Discovering how close a deterministic model could come to the stochasctic model is interesting in its own right.
• B. A set of models has already been build for small African farms, modelling crops and nutrient flows. There is work to be done developing these further.

The project would be based in Edinburgh with visits to Dundee for the A projects.

Note: It is possible to suggest some other ecological/environmental area - it doesn't have to be limited to the projects A and B listed above.