Bayesian Analysis for Inference of an Emerging Epidemic: Citrus Canker in Urban Landscapes

Neri, Franco M.; Cook, Alex R.; Gibson, Gavin J.; Gottwald, Tim R.; Gilligan, Christopher A.

Chief amongst the concerns of policy makers concerned with managing an emerging epidemic are: how far and how fast is the epidemic spreading? How reliable are future predictions of the epidemic severity? Does the epidemic merit the deployment of control, and how should this be optimised? Here we have focused on the first two questions about estimation and prediction, using a combination of Bayesian statistical inference and data for the spread of citrus canker in urban Miami. We assumed that little was known about the pathogen, using non-informative priors for the parameters and a selection of dispersal kernels. Our analyses have shown that the same spatio-temporal, stochastic model is able to capture the temporal trends and spatial statistics characterising the spread of infection in all four sites. Pathogen spread within sites is described by an exponential dispersal kernel with a time-varying transmission rate augmented by a small, time-varying rate of external infection. We show, therefore, that epidemics were not self-contained within sites but new foci of infection also arose from external inoculum, a phenomenon evident at all four sites.