A fully automated method of locating building shadows for aerosol optical depth calculations in high-resolution satellite imagery

摘要

Vincent (2006) developed a technique for remotely measuring Aerosol Optical Depth (AOD) using commercial high-resolution satellite imagery. This technique measured the radiance difference between a building shadow and an adjacent sunlit region with the same surface reflectance to calculate Total Optical Depth (TOD). AOD is then determined by subtracting Rayleigh scattering from TOD. The procedure for making this calculation was time consuming, particularly locating suitable shadows within the region of interest. This paper outlines a fully automated method of performing the AOD calculation and examining shadow properties. The automated method relies on high-resolution Digital Surface Model (DSM) data collected using a Light Detection and Ranging (LIDAR) sensor coupled with sun and satellite geometry to identify shadow regions. Configuration settings allowed for specific regions in the shadow and sunlit area to be selected before determining their respective radiances. Finally, a technique for aligning the satellite and DSM pixels was developed to correct for small differences between the datasets. Results from the automated method were compared with AERONET data for validation. The automated method using WorldView-1 and QuickBird imagery worked best at Solar Village, Saudi Arabia and an area northeast of Washington, D.C., which included the Goddard Space Flight Center. Testing of IKONOS multispectral imagery suggested the resolution is inadequate in urban settings. Testing in areas that included downtown regions in Houston, TX and Baltimore, MD identified weaknesses in the alignment algorithm.