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.

We consider emerging epidemics, arising, e.g., when a new pathogen is introduced in a host population. In face of the new threat, crucial control measures have to be implemented quickly, yet prior knowledge of the parameters underlying pathogen spread and transmission is often missing. Predictive modelling can greatly help in informing decision making by estimating those parameters from early observations of the outbreak. The important questions are then: can a modeller characterise the disease ''soon enough,'' i.e., within a useful time frame, in order to enact the proper control measures? At what stage of the outbreak can the future epidemic progress be reliably predicted? We analyse an outbreak of citrus canker, a wind-spread bacterial disease of citrus, in urban Miami. The model succeeds in capturing the main epidemiological features of the disease, but we find contrasting answers. The spatial scale of disease spread can be identified quickly and accurately from early observations. However, the rate of spread is rapidly changing in time, driven mainly by rare thunderstorms with very short-time predictability, which frustrates epidemic prediction.