Remote-sensing based risk mapping for mosquito-borne diseases: a spatial modelling study

Abstract

BackgroundWithin the USA, South Dakota is the state with the highest incidence rates of human West Nile virus disease. For effective disease prevention and mosquito control, it is crucial to identify potential transmission hot spots. Because disease occurrence is driven by the presence of avian hosts and mosquito vectors, understanding their habitats is important to predict the risk of cases of human West Nile virus. Geospatial environmental information about vegetation cover and water availability can help us to understand what drives the spatial distribution of disease outbreaks and to map the spatial pattern of the risk of West Nile virus. MethodsIn our study, we use Landsat imagery from 2003 to 2012 to predict the risk of West Nile virus transmission to human beings in South Dakota. We computed environmental indices such as the normalised differenced vegetation index (NDVI) and the normalised differenced water index (NDWI) as proxies for vegetation cover and water availability. For predictive mapping, we used boosted regression tree models to fit a model of disease risk based on geocoded human case data. FindingsIndices that measured interannual variability in vegetation and surface water were important predictors of West Nile virus risk, and identified locations with temporary water bodies that provided habitat for vectors and hosts. The resulting risk map accurately highlighted the major areas of high West Nile virus risk across South Dakota. InterpretationWe suggest that use of the temporal depth of the Landsat archive will help to develop more effective environmental metrics for mapping vector-borne diseases. FundingNational Institute of Allergy and Infectious Diseases NIH/NIAID (R01AI079411) and NASA Applied Science Public Health and Air Quality Programme.