Multispectral aerosol optical depth retrievals from high-resolution satellite imagery
October 15, 2009
Phil Durkee (Naval Postgraduate School)
Hosted by Thomas Vonder Haar
Abstract
(with contributions from Kurt Nielsen, Nick Vincent, Jack Evans, Ryan Dombrock, Brian Rivenbark, and Brian Belson)The advancement and proliferation of commercial high-resolution imaging satellites presents a new opportunity for overland aerosol characterization. Current aerosol optical depth retrieval methods typically fail over areas with high surface reflectance, such as urban areas and deserts, since the upwelling radiance due to scattering by aerosols is small compared to the radiance resulting from surface reflection. An exception is the NASA Deep Blue product that relies on absorption properties of dust at blue wavelengths. The method demonstrated here uses shadows cast on the surface to exploit the differences between radiance from the adjacent shaded and unshaded areas of the scene. Shaded areas of the scene are primarily illuminated by diffuse irradiance that is scattered downward from the atmosphere, while unshaded areas are illuminated by both diffuse and direct solar irradiance. Given uniform surface reflectance for the shaded and unshaded areas, the difference in reflected radiance measured by a satellite sensor is related to the direct transmission of solar radiation. Aerosol optical depth can then be estimated from its contribution to the total atmospheric optical depth following a correction for molecular Rayleigh scattering. Results based on QuickBird imagery of several sites of varying surface and aerosol type are presented and compared with ground truth from nearby AERONET measurements. Since the technique applies to all solar wavelengths, analysis of particle size and gaseous absorption are also discussed.