by Paulo C. Ventura, Allisandra G. Kummer, André B. B. Wilke, Jagadeesh Chitturi, Megan D. Hill, Chalmers Vasquez, Isik Unlu, John-Paul Mutebi, Susanne Kluh, Steve Vetrone, Dan Damian, John Townsend, Maria Litvinova, Marco Ajelli
Aedes-borne diseases represent a major public health threat and mosquito control operations represent a key line of defense. Improving the real-time awareness of mosquito control authorities by providing reliable forecasts of the relative abundance of mosquito vectors could greatly enhance control efforts. To this aim, we developed an analytical tool that forecasts Aedes aegypti relative abundance 1 to 4 weeks ahead. Forecasts were validated against mosquito surveillance data (2,760 data points) collected over multiple years in four jurisdictions in the US. The symmetric absolute percentage error was in the range 0.43–0.69, and the 90% interquantile range of the forecasts had a coverage of 83–92%. Our forecasts consistently outperformed a reference “naïve” model for all analyzed study sites, forecasting horizon, and for periods with medium/high Ae. aegypti activity. The developed tool can be instrumental to address the need for evidence-based decision making.