You are here:
Ecosystems and spatiotemporal mosquito-borne disease models across a gradient of urbanization
Citation:
Myer, M. AND JohnM Johnston. Ecosystems and spatiotemporal mosquito-borne disease models across a gradient of urbanization. 2018 ESA, ESC, and ESBC Joint Annual Meeting, Vancouver, CANADA, November 11 - 14, 2018.
Impact/Purpose:
Presented at the 2018 ESA, ESC, and ESBC Joint Annual Meeting
Description:
Mosquito-borne diseases are changing in geographic extent and increasing in incidence due to human-mediated ecological changes and a changing climate. In order to surveil, predict, and mitigate these diseases, it is important to study the ways in which ecosystems interact with the anthroposphere to determine the location, time, and severity of incidence. We used remote-sensed and local ecological, socioeconomic, and meteorological data to fit spatiotemporal models of West Nile virus incidence in Suffolk and Nassau Counties, New York, USA, based on Culex pipiens Linnaeus trapping data. In both cases, we identified a gradient of increasing spatial incidence probability that corresponded to increased urbanization, with our study area encompassing rural, exurban, suburban, and urban areas. Ecological findings included a positive West Nile incidence association in rural/exurban Suffolk County with on-site sewage disposal systems, and a negative association with open water and woody wetlands, indicating that healthy aquatic ecosystems may disrupt a step of the West Nile enzootic cycle. In suburban/urban Nassau County, older housing stock present in suburban areas was associated with increased West Nile incidence, while both dense urbanization and high vegetative index was associated with lowered incidence. This indicated that, in contrast to less urbanized areas, direct human-associated changes to mosquito habitat are a primary driver of West Nile incidence in suburban and urban areas, with a particular risk in ‘sweet-spot’ locations with intermediate urbanization and vegetative cover. Both models performed well in predicting disease incidence, with classifier performance between 75-80% on held-out data.
