Improved MODIS Aerosol Retrieval Using Modified VIS/SWIR Surface Albedo Ratio Over Urban Scenes

Abstract

In this paper, we demonstrate a regional approach to Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD) retrieval over urban scenes. By using simultaneous MODIS and AErosol RObotic NETwork (AERONET) sky radiometer data, we show that for urban scenes, the surface reflectance ratios between the visible (VIS) (0.47- and 0.66-¿m) and shortwave-infrared (2.12-¿m) channels used in the MODIS Collection 5 (C005) retrieval are significantly underestimated. This leads to an underestimate in the VIS ground albedos resulting in a significant overestimate of the AOD in the VIS channels. Since the dependence on scattering geometry is weak, revised surface reflectance ratio maps at spatial resolutions down to 1.5 km can be developed, permitting aerosol retrievals at these resolutions. Applying this procedure results in significant reductions in MODIS-retrieved AOD bias in comparison to AERONET. As an example, for New York City, at the operational 10-km resolution, the MODIS-AERONET regression using C005 is ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MODIS</sub> = 1.24 ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">AERONET</sub> + 0.11, <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.85 , while for the regional method, we apply results in ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MODIS</sub> = 1.02 ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">AERONET</sub> + 0.04, <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.92 . This improved performance is maintained even for high-resolution retrieval down to 1.5 km with a regression of ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MODIS</sub> = 0.96 ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">AERONET</sub> + 0.02, <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.88. In addition, we show that this method can be extended to other urban scenes such as Mexico City with similar results and discuss the implications of C005 AOD overestimates on PM2.5 monitoring. Furthermore, comparisons with distributed PM2.5 sensors in the region confirm that the high estimates of AOD from C005 cannot be explained by excess urban emissions.