West Nile Virus (WNV) is the leading cause of mosquito-borne disease in the United States, yet transmission activity remains difficult to predict. The present study used 20 years of digitized WNV seroconversion data from 526 sentinel chicken coops across Florida to develop spatiotemporal models with landscape and climate variables to predict WNV seroconversion at monthly and seasonal timescales. We found several environmental predictors hypothesized to impact WNV transmission were important at both timescales. Seroconversion of WNV was predicted to decrease with higher maximum temperatures during the sampling month and increase with higher precipitation and minimum temperature two months prior to sampling. In the seasonal model, intermediate values of cumulative precipitation in the previous season predicted higher WNV seroconversion. High accuracy in out-of-sample predictions at both timescales demonstrated the utility of our models toward ecological forecasting of virus activity. Monthly models had higher precision than the seasonal model, but both timescales have potential to inform management decisions. Monthly predictions could guide targeted control efforts during active transmission seasons, while seasonal predictions provide a lead-time to improve preparedness and inform resource allocation. Retrospective statewide predictions across the 20-year time period provided qualitative correlations between areas of high predicted WNV transmission hazard among humans and equines, while also providing insights into WNV transmission ecology following its introduction in 2001. Overall, our framework provides a step forward in the use of spatiotemporal ecological modeling for public health and vector-borne disease ecology and management.
Baecher et al. (Tue,) studied this question.