Satellite Earth Observations Identify Arbovirus Transmission Hotspots in an Urban Landscape

Mosquito-transmitted viruses (arboviruses) are an emerging human health threat of worldwide significance. Given rising temperatures associated with climate change and increasing mobility via global transportation networks, there is concern that arboviruses will emerge more regularly in temperate regions, including the United States. Understanding when and where outbreaks are likely to occur will be essential for anticipating and effectively responding to arbovirus emergence.

Mosquitoes and the viruses they transmit are highly sensitive to temperature and other meteorological factors. An emerging technique for risk assessment involves linking mechanistic epidemiological models based on temperature-trait relationships with meteorological and climatological datasets. However, a limitation of this approach is that the local microclimates experienced by mosquitoes can differ substantially from the meteorological conditions measured at nearby weather stations or represented by large cells in gridded data products. There is a need for new approaches to arbovirus monitoring and forecasting that match the scale of environmental measurement to the scale of the mosquito.

We integrated satellite remote sensing data and in-situ monitoring of microclimates with an epidemiological model of disease transmission to map patterns of potential dengue risk in Athens, GA. To generate a temperature microclimate map, we used temperature data collected from 54 temperature and relative humidity data loggers (RFID Track-it, Monarch Instruments) with radiation shields distributed across nine sites selected to encompass a gradient of impervious surfaces and tree cover in the surrounding landscape. We used a random forest machine learning algorithm with Landsat and Sentinel-2 data to generate maps of tree cover (12.9% MAE) and impervious surfaces (9.1% MAE). We used a linear model to predict the minimum and maximum daily temperatures measured at each microclimate logger as a function of minimum and maximum temperatures from GridMET combined with land cover variables from the impervious surface and tree cover maps. The resulting models predicted daily minimum and maximum temperatures with cross-validated MAE of 0.89 °C and 1.72 °C and observed-predicted correlation of 0.78 and 0.87, respectively.

We used these microclimate temperatures to predict the environmental suitability for dengue transmission with a model based on laboratory-derived temperature-trait relationships for mosquito biting rate, infection probability, transmission probability, mosquito lifespan, and extrinsic incubation period. Mosquito abundance was incorporated using an empirical model derived from field measurements of Aedes albopictus. The resulting maps highlighted the locations with the highest dengue transmission potential in Athens. Temperatures and transmission potential at these locations differ considerably from measurements based on data from coarse-resolution climate grids or meteorological stations. This proof-of-concept study has demonstrated that it is feasible to track variation in mosquito microenvironments within cities using satellite Earth observations. These microclimates reveal localized patterns of disease transmission risk that would not be detectable using coarser-grained environmental datasets.