Impact of urban heat islands on dengue thermal transmission suitability across contrasting U.S. cities

Ectotherms such as mosquitoes are exquisitely sensitive to environmental conditions, including temperature. Dengue fever, caused by dengue virus and transmitted by Aedes mosquitos, is commonly thought of as a tropical disease but has recently emerged in small local outbreaks in the USA. Urban heat island effects are a well-known phenomenon, but measuring their impact on air temperature at the scale of urban environments is complicated by the limitations of meteorological observations. These observations are collected only at discrete coordinates in which weather stations are located, making it difficult to interpret spatial variation across a landscape. Although global gridded reanalysis data such as ERA5 are available at hourly frequency, the finest spatial resolution is 0.25 ∘ , and thus the details in variation are lost. In this project, we developed a framework for localized machine learning models to predict the air temperature at a fine-scale 30 meter resolution for a given study area. The models we present are trained on location-specific satellite imagery and local weather station data, but we provide automated code for use with freely available data, allowing them to be retrained for nearly any location. In this study, we used these predictions to assess the seasonality of dengue thermal transmission suitability for Aedes aegypti and Ae. albopictus in two US cities. Urbanized areas were associated with extended projected dengue seasons for both vector species. This highlights the need for attention to urban planning and vector control in order to reduce the burden of arboviruses in dense urban areas.