Hierarchical Bayesian models: Epidemiology and data for defining pest extent

22 - 24 September 2009

Area freedom cannot be proven; only its probability can be estimated. Current informal approaches to delimiting pest extent rely upon expert ecological interpretation of presence / absence data over space and time. Hierarchical Bayesian models provide a cohesive statistical framework that can formally integrate all of the available information from both epidemiology and data. They allow inference to be made on the probable extent of pests by breaking down surveillance systems into a series of simpler component models.

A model is presented to demonstrate the hierarchical Bayesian approach to estimating area freedom. The method involves constructing an observation model for the surveillance data, conditional on the hidden extent of the pest and uncertain detection sensitivity. The extent of the pest is then defined by a dynamic invasion process model that includes uncertainty in epidemiological parameters. Markov chain Monte Carlo techniques allow the probable extent of the pest to be estimated from the combined observation and process model, given the surveillance data and stated uncertainty.

The methodology can assist decision-making across a range of plant biosecurity surveillance activities including early detection, market access and incursion response. Outputs include not only the estimate of area freedom, but also the likely location of pests and epidemiological characteristics of the invaders. Risk maps derived from these models can be incorporated into GIS systems and updated as new data arrive. The tools developed by this project can be used by biosecurity regulators to direct surveillance towards current risks and to assess the performance of surveillance programs.