West Nile Virus Mosquito Abundance Modeling Using Nonstationary Spatiotemporal Geostatistics

The lack of spatial coverage and missing observations of adult mosquito surveillance data challenge the quantitative assessment of human exposure to West Nile virus (WNV). We developed a geostatistical spatiotemporal prediction model for missing WNV mosquito data. In the proposed Poisson generalized linear-mixed model, the effects of meteorological and physiographic conditions on mosquito abundance are modeled as a drift, and the spatiotemporal variations around the drift, possibly correlated, are captured by a spatiotemporal residual random field. The proposed model accounts for discrete counts of the mosquito surveillance data within a generalized linear-mixed model, and tackles the nonstationarity in WNV mosquito abundance data by restricting the decision of stationarity to a local neighborhood surrounding the target prediction point.