Retrieval of Aerosol Optical Depth (AOD) from the Landsat8 Oli Observations Over Beijing

The deep blue (DB) algorithm is usually used for retrieval of AOD on high reflectivity surfaces, which cover complex surface types. However, due to low spatial resolutions, it shows limited information details in urban areas and cannot fully display the aerosol variation characteristics. Now, a new aerosol retrieval algorithm with a priori Land Surface Reflectance (LSR) database support (HARLS) is proposed for Landsat-8 (OLI) images at 1km spatial resolution to use the look-up table (LUT) method to retrieve AOD. The continental aerosol type defined in the 6S model is modified by AERONET single scattering albedo (SSA) parameter to correct the LUT for different seasons, and deep learning neural network is used to train the samples based on the corrected LUT and retrieve AOD. The retrieved AOD are then validated against the aerosol robotic network (AERONET) AOD ground measurements from five stations in Beijing and compared with MOD04 C6.1 10km AOD product. The overall results demonstrate that the AOD retrieved is highly consistent with the AOD derived from ground measurements.