Airborne imaging spectroscopy to monitor urban mosquito microhabitats David R. Thompson a , Manuel de la Torre Juárez a, ⁎, Christopher M. Barker b , Jodi Holeman c , Sarah Lundeen a , Steve Mulligan c , Thomas H. Painter a , Erika Podest a , Felix C. Seidel a , Eugene Ustinov a a Jet Propulsion Laboratory, California Institute of Technology, Pasadena CA, United States b Center for Vectorborne Diseases, University of California, Davis, CA, United States c Consolidated Mosquito Abatement District, Selma, CA, United States abstract article info Article history: Received 5 December 2012 Received in revised form 20 June 2013 Accepted 22 June 2013 Available online 28 July 2013 Keywords: Imaging spectroscopy Disease vector control West Nile virus Green swimming pools Matched filter detection Urban environments West Nile (WNV) is now established in the continental United States with new human cases occurring annu- ally in most states. Mosquitoes in the genus Culex are the primary vectors and exploit urban stagnant water and swimming pools as larval habitats. Public health surveys to monitor unmaintained pools typically rely on visual inspections of aerial imagery. This work demonstrates automated analysis of airborne imaging spec- troscopy to assist Culex monitoring campaigns. We analyze an overflight of Fresno County, CA by the Airborne Visible Infrared Imaging Spectrometer instrument (AVIRIS), and compare the spectral information with a concurrent ground survey of swimming pools. Matched filter detection strategies reliably detect pools against a cluttered urban background. We also evaluate remotely sensed spectral markers of ecosystem characteristics related to larval colonization. We find that commonly used chlorophyll signatures accurately predict the probability of pool colonization by Culex larvae. These results suggest that AVIRIS spectral data provide sufficient information to remotely identify pools at risk for Culex colonization. Published by Elsevier Inc. 1. Introduction Remote sensing has long contributed to infectious disease predic- tion and warning systems (Linthicum, Bailey, Davies, & Tucker, 1987; Washino & Wood, 1994). Spatial epidemiological studies have used satellite data to map environmental conditions associated with dis- ease vector habitats. They typically correlate environmental variables such as land use, vegetation indices, temperature, and elevation with relative vector abundance or pathogen transmission (Beck, Lobitz, & Wood, 2000; Kalluri, Gilruth, Rogers, & Szczur, 2007). Instruments used for this purpose include the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and Moderate Resolu- tion Imaging Spectroradiometer (MODIS). Urban areas pose a special challenge: they are heterogeneous collections of residential and com- mercial areas, parks, and other land use types with potential vector hab- itats on much smaller spatial scales (Reisen, 2010). Characterizing these sparse microhabitats requires different remote sensing techniques. A particular concern is West Nile Virus (WNV), which spreads in urban areas by transmission between birds and mosquitoes in the genus Culex. These mosquitoes often colonize stagnant water during their aquatic immature stages, and features such as open containers or unmaintained swimming pools provide key habitats (Caillouët, Carlson, Wesson, & Jordan, 2008; Reisen, Takashi, Carroll, & Quiring, 2008). Unmaintained residential swimming pools, or green pools, are especially problematic. These neglected pools become stagnant with accumulated organic matter, and often harbor Culex pipiens mos- quito larvae (Fig. 1). Current WNV mitigation efforts generally rely on street-level monitoring and treatment campaigns. Previous work has used remote sensing products such as airborne images to monitor individual pools, enabling more effective Culex treatment programs by identifying risk areas and flagging specific households for direct intervention. For in- stance, Reisen et al. (2008) demonstrate an airborne survey of Bakers- field, CA with high resolution color imagery. These images clearly reveal neglected green pools. However, manual inspection is necessary to catalog these habitats, so the approach is better suited for a single snapshot in time than sustained monitoring campaigns that track the evolution of the vector habitats. Since then, local providers have contin- ued to refine the manual image inspection using GIS pool catalogs and higher spatial resolution (Franklin, 2013). There have also been efforts to automate the image analysis. Kim et al. (2011) propose a fully auto- mated method to locate pools in GeoEye satellite images, detecting pools from a Normalized Difference Water Index (NDWI) score followed by a morphological classification. This is effective at finding pools, but assessing their condition still relies on manual inspection. To our knowledge no previous study has quantified a link between remotely sensed pool color and Culex colonization that would permit automated assessment of pool condition from airborne instruments. Remote Sensing of Environment 137 (2013) 226–233 ⁎ Corresponding author. Tel.: +1 8183544548. E-mail addresses: david.r.thompson@jpl.nasa.govemail (D.R. Thompson), mtj@jpl.nasa.govemail (M. de la Torre Juárez), cmbarker@ucdavis.edu (C.M. Barker). 0034-4257/$ – see front matter. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.rse.2013.06.015 Contents lists available at SciVerse ScienceDirect Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse