Neighborhood-scale modeling of Culex tarsalis abundance and West Nile virus infection prevalence

Abstract Prevention of West Nile virus (WNV) transmission relies heavily on vector surveillance and control. Current methods of population reduction are often reactive, guided by field surveillance. Vector control measures would benefit from neighborhood-scale, spatiotemporal predictions of WNV risk that could facilitate proactive, targeted surveillance to locations and times where vector and virus activity is expected to be highest. To determine the potential of predictions at this spatial scale, we approached model development in two phases: (1) inferential generalized additive models were developed using high-density mosquito data collected across Davis, CA that identified causal associations between neighborhood-scale spatial features and relative abundance or WNV infection prevalence in Culex tarsalis; (2) Bayesian predictive models for Cx. tarsalis abundance and WNV prevalence were constructed by incorporating neighborhood-scale spatial variables with temporal and climatic predictors, using historical vector surveillance data from Davis between 2006 - 2024. Inferential analyses revealed significant, time-dependent effects of nearby rice fields and distance from urban edge on Cx. tarsalis abundance and WNV prevalence. Final predictive models included terms for spatial, meteorologic, and temporal variables, and performed well for Cx. tarsalis abundance and WNV prevalence when challenged with novel data from Davis or the neighboring town of Woodland, CA. Results support the potential of neighborhood-scale models to predict WNV vector activity and inform vector control efforts, including routine operational control and placement of field trials.