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.

Strategic decisions about how to react to emerging epidemics are inevitably made early on, when few data are available. However, it is strongly suspected [28] that the main drivers of the epidemic (responsible for the fluctuations and the final slowing down of transmission rates found in our post hoc analyses, cf. Figures 3B-I and related discussion) were major weather events that could not be known at the beginning of the outbreak. Such lack of knowledge affects epidemic forecasts made from the early stages of the outbreak. In the following, we investigate three different hypothetical scenarios for early prediction: when no prior information is given about the future conditions of the epidemic (scenario A), and when some prior knowledge is assumed (scenarios B and C). For each scenario, the parameters were estimated using observation windows of increasing size, all starting at t = 0, and then used to predict future trajectories of the epidemics up to 18 months (i.e. for the pre-drought period; see above). The results are shown in Figure 6 for one of the Miami Dade sites (D1), with observation windows of 3, 6, and 9 months.