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.

Here we focus on the early stage of the epidemic in urban Miami and, in particular, how to estimate the inherent spatial and temporal scales of the epidemic in order to predict the future course of an epidemic in a spatially heterogeneous urban setting. Infection on these trees constituted a potent source of inoculum that must be controlled were the disease threat to plantations to be economically managed. Accordingly, the USDA Agricultural Research Service initiated a detailed census of susceptible trees and the occurrence of ACC in five sites in Broward and Dade counties in the Miami region. The sites ranged from 2.6 km 2 to 15.5 km 2 [28] . The data provide a full census of susceptible trees, with 24 successive monthly snapshots for the occurrence of new infections. Retrospectively, the outcome of the epidemic at each site is known. Here we use subsets of the data at different stages of the epidemic to recreate different levels of ignorance about the future course of the epidemic. Then using Bayesian statistical inference and a stochastic model we compare model predictions with the known course of the epidemic. Specifically we ask: By using the citrus canker outbreak to address these broad questions, we introduce and test methodologies that are applicable to a much broader class of spatially-and temporally-complex epidemics.