Predicting Dengue Incidence in Brazil Using Broad-Scale Spectral Remote Sensing Imagery

Infectious disease burden is continuing to increase around the globe. These diseases have increased, in part, due to globalization, human behavior, and environmental changes. There is an urgent need for improved prediction of their spread so that mitigation techniques and treatments can be administered proactively rather than just reactively. One of the challenges is that many regions of interest are in hard-to-reach locations, and as such, clinical surveillance data (reliant upon self-reporting) can be both sparse and lagging. Remote sensing imagery is an attractive data source to exploit for this application as it provides real-time information without having to physically be on the ground. Here, we derive standard indices from multispectral imagery, and explore how predictive they are for forecasting dengue incidence in Brazil. This is done on broad spatial and temporal scales, covering all of Brazil for multiple years. Results will show that the normalized difference vegetation index is a leading predictor for dengue incidence.