Determinants of Spatial and Temporal Variation of West Nile Virus Transmission in Texas

West Nile virus (WNV) is a zoonotic vector-borne virus that infects avian and mammal hosts. In Texas, WNV was first reported in 2002 in Harris County and has since been reported annually throughout the state. With variable funding available for mosquito surveillance in Texas, predictive modeling is an economical method for mosquito control, but has not been parameterized for major metropolitan areas of central and southeast Texas. Thus, this dissertation uses historical databases to create predictive models that are specifically tailored for major cities in Texas. To investigate the 2012 WNV epidemic in Dallas County, TX, logistic regression models identified an index of urbanization (composed of greater population density, lower normalized difference vegetation index, higher coverage of urban land types, and more impervious surfaces), lower elevation, and older populations as key factors in predicting the risk of WNV in Culex quinquefasciatus. Our model was then extrapolated as a risk map, which highlighted north and central Dallas County as areas of high risk for WNV-positive mosquitoes. A similar study for Harris County was conducted, where the best-fit model found that areas with higher elevation, more impervious surfaces, greater median income, and predominantly Hispanic populations will have higher vector indexes, which measure the average number of WNV-infected female Culex mosquitoes collected per trap night. The predictive map based on this model emphasized high-risk areas in central and north Harris County. Harris County’s long-term database was also used to investigate temporal patterns between vector abundance, WNV infection in Cx. quinquefasciatus, and weather patterns. A time-series analysis revealed correlations between abundance and environmental variability measurements, following our hypothesis of Schmalhausen’s law that states organisms are susceptible to mean (average) temperature and precipitation measurements as well as extreme or variability in weather. The infection rate model identified temperature with an 8-month lag as a significant covariate for WNV infection rates, highlighting the importance of overwintering temperatures preceding the WNV season. These models (landscape, demographic, and meteorological conditions) can be used by local mosquito control agencies to predict WNV infection in Cx. quinquefasciatus for proactive and effective control efforts