Increasing model efficiency by dynamically changing model representations

There are a number of strategies to deal with modelling large complex systems such as large marine ecosystems. These systems are often comprised of many submodels, each contributing to the overall trajectory of the system. The balance between the acceptable modelling error and the run-time often dictates the form of these submodels. There may be scope to improve the position of this balance point in both regards by structuring models so that submodels may change their algorithmic representation and state space in response to their local state and the state of the model as a whole.This paper uses an example system consisting of a single population of animals which periodically encounters a diffuse contaminant in a localised region as an example of such a system, and discusses the key issues that arise from the approach.

► Complex models, like ecosystem models, are often comprised of many smaller submodels.
► The form of these composite models is often dictated by run-time and acceptable error.
► Dynamically changing models (IBM↔EBM) gave a 3.6 times increase in test model speed.
► The example shows a successful way to improve the balance between speed and accuracy.