Implementation of Non-Spherical Dust Model Improves Dark Target Ocean Dust Retrieval

Over the ocean, the Dark Target (DT) aerosol retrieval algorithm for MODIS has long-standing bias in its retrievals of Aerosol Optical Depth (AOD), Angstrom Exponent (AE) and Fine Mode Fraction (FMF) for dust aerosols because non-sphericity of dust particles is not properly treated. These biases are scattering-angle-dependent. A two-step strategy is designed to improve the DT retrieval of dust over ocean. First, a dust detection algorithm is constructed following a comprehensive survey of existing dust detection algorithms and a combination of spectral tests from near-UV (deep blue), visible, and thermal infrared (TIR) wavelength spectral tests are selected to detect the dusty pixels. Then a new dust model is built assuming a mixture of spheroid particles to represent global dust optical properties. Comparison with AERONET identified dust measurements indicates significant reduction of AOD bias (from 0.06 to 0.02) while improving the fraction of retrievals within expected error (%EE) from 64% to 82%. At the same time, the overall bias in AE is reduced from 0.13 to 0.06, and the scattering-angle-dependent AE bias is largely eliminated. A thorough evaluation of the new retrieval implemented within the operational Dark Target algorithm shows a general improvement of dust retrieval quality, with an overall reduction of AOD, FMF and AE in the major dusty regions in the world.