On 27 October Dr Veronica Bowman, Principal Statistician at Dstl, spoke about Hazard Assessment Simulation and Prediction (HASP). The seminar focused on Bayesian inference and uncertainty modelling, with specific reference to methods for predicting the effects of a chemical or biological release (CBR) using computational modelling. Predictions are made and assessed through a chain of models, many of which will take as input parameters the outputs of previous models. Modelling of CBR must consider factors such as the weather, dispersion properties, detection of the release and responses to different dosages of the released material. Each of these models contains uncertainty, so to understand the reliability of the final prediction we need to understand how these uncertainties will combine with one another.
The seminar introduced several of the types of models associated with CBR modelling, including virtual battlespace modelling, dispersion modelling and models for sensor placement and source term estimation. Topics such as physical protection techniques and the wider area of knowledge management were also covered. The talk ended by introducing current research into statistical emulation for dispersion modelling, comparing three possible approaches to this challenge.
Diagram of the effects of a CBR.