Integrating Environmental Monitoring with Public Health Surveillance for Arbovirus Forecasting

West Nile virus (WNV) remains a persistent public health hazard in many parts of the U.S., and there is a need for better information about when and where WNV outbreaks will occur to target public health responses. To address this need, we developed and implemented the Arbovirus Monitoring and Prediction (ArboMAP) system in South Dakota, the U.S. state with the highest incidence of WNV. ArboMAP produces weekly, county-level forecasts of WNV risk using environmental data combined with entomological and human case surveillance. We made prospective WNV forecasts in South Dakota during 2016, 2017, and 2018.

ArboMAP uses a data-driven modeling approach in which human WNV cases were modeled as a function of meteorological conditions and mosquito infection status. Gridded air temperature, vapor pressure deficit, and precipitation data were derived from NASA's North American Land Data Assimilation System. Mosquito data were obtained from trapping programs throughout the state, and were pooled and tested for WNV on a weekly basis. We used distributed lags to model delayed effects of environmental fluctuations and a log-linear function to model growth of the mosquito infection rate in the early transmission season.

Weekly forecasts of WNV transmission risk were made based on observed meteorological conditions and mosquito infection status during the current transmission season. The model was able to predict interannual variation in the incidence of WNV cases, as well as shifts in the start of the WNV season. The model predictions also highlighted a change in the usual geographic pattern of WNV cases that occurred in 2017. Integration of environmental data with mosquito infection data was necessary to achieve the highest prediction accuracy. The forecasts filled an important information gap in the early WNV season, when vector abundance is not an effective predictor of risk and most human cases are still unreported. ArboMAP has been implemented using the R software environment for data processing, modeling, and reporting combined with a Google Earth Engine application for environmental data access. Current research and development is focused on generalizing the underlying models and evaluating forecasts in new locations with different transmission cycles, including Oklahoma and Louisiana.