An Operational System for Surveillance and Ecological Forecasting of West Nile Virus Outbreaks

Abstract

Mosquito-borne disease surveillance has traditionally focused on tracking human cases along with the abundance and infection status of mosquito vectors. Vector and host population dynamics are also sensitive to climatic factors, including temperature fluctuations and the availability of surface water for mosquito breeding. There is a potential to strengthen surveillance and predict future outbreaks by monitoring environmental risk factors using broad-scale sensor networks that include earth-observing satellites. The South Dakota Mosquito Information System (SDMIS) project combines entomological surveillance with gridded meteorological data from NASA’s North American Land Data Assimilation System (NLDAS) to generate weekly risk maps for West Nile virus (WNV) in the north-central United States. Critical components include a mosquito infection model that smooths the noisy infection rate and compensates for unbalanced sampling, and a human infection model that combines entomological risk estimates with lagged effects of meteorological variables from the North American Land Data Assimilation System (NLDAS). Two types of forecasts are generated: long-term forecasts of statewide risk extending through the entire WNV season, and short-term forecasts of the geographic pattern of WNV risk in the upcoming week. Model forecasts are connected to public health actions through decision support matrices that link predicted risk levels to a set of phased responses. SDMIS accurately predicted higher-than-normal WNV transmission in 2016 and lower-than-normal transmission in 2017. Our experiences highlight several important lessons that can inform future efforts at disease early warning. These include the value of integrating climatic models with recent observations of infection, the importance of automated workflows to facilitate timely integration of multiple data streams, and the need to link forecasts with specific public health responses.