West Nile Virus Forecasting

Predictive modeling of disease outbreaks

West Nile virus (WNV) is the most common mosquito-borne disease in the United States, with more than 53,000 reported cases and 2,500 deaths since its introduction in 1999. The northern Great Plains have consistently experienced some of the highest WNV incidence rates in the country, with cases concentrated in particular geographic areas and outbreak years (Wimberly et al. 2013). Because WNV transmission is highly seasonal, weather conditions during spring and early summer influence virus amplification in mosquitoes and birds and the subsequent risk of human infection (Wimberly et al. 2014). Building on these findings, we developed regional forecasting models that initially used satellite observations of vegetation greenup and temperature-based degree days to predict WNV incidence (Chuang and Wimberly, 2012). Subsequent research demonstrated that integrating environmental indicators with mosquito infection surveillance improves forecasting accuracy, providing a stronger foundation for anticipating WNV outbreaks (Davis et al. 2017, 2018).

WNV forecast time series
ArboMAP forecasts of WNV in South Dakota, including historical calibration as well as predictions outside the range of the training data (Wimberly et al. 2022).

We translated these advances into the Arbovirus Monitoring and Prediction (ArboMAP) system, a set of tools for implementing WNV forecasting in public health agencies. ArboMAP has supported WNV forecasting in South Dakota since 2016, and was extended to Louisiana, Michigan, and Oklahoma (Nekorchuk et al. 2023). Retrospective validation demonstrated that models driven by environmental observations and mosquito surveillance data enabled accurate forecasts early enough to inform disease prevention and mosquito control before the peak transmission season (Wimberly et al. 2022). This work demonstrates the potential of integrating Earth observations and public health surveillance to move from reactive disease monitoring toward proactive outbreak preparedness. We are now developing ArboMAP 2.0, a next-generation forecasting platform that incorporates additional environmental data streams, improved modeling capabilities, and enhanced visualization tools to support broader implementation across multiple states.

References