I first encountered IBMs at workshops in 2003 and 2004 where Chris Topping showed their value compared with classical population approaches. Classic approaches do not take account of spatial heterogeneity, but real landscapes are fragmented and consist of contiguous patches of different types, and this can be taken into account using an IBM approach.Chris and I and others subsequently collaborated on a number of papers, and then the EU CREAM project provided an opportunity to develop the approach with scientists across Europe. This led me to develop a novel way of representing the energy budgets of individuals which we are now implementing across a range of species, from earthworms to elephants. Several projects aimed at improved environmental risk assessment have been collaborations with Dr Pernille Thorbek of Syngenta.
Development of IBMs has to take account of all available data, and I have become interested in how this should be done. In classical statistics models are routinely fitted to data, and it turns out the same can be done with IBMs using Approximate Bayesian Computation, ABC. Many of the issues are the same as those encountered in classical statistics – can discrepancies between model outputs and data be assumed to be normally distributed; can models with different structures be compared statistically. The answers to my mind are very encouraging.
My University web page is here.