A machine learning approach to retrieve aerosol single scattering albedo using joint satellite and surface visibility measurements

Abstract:

Aerosol single scattering albedo (SSA) measuring the ratio of scattering to extinction is a critical parameter in determining aerosol radiative effect. However, the retrieval of SSA is limited by comprehensive monitoring requirements of both direct solar radiance and scattered sky radiance. Most existing passive satellite sensors such as MODIS and VIIRS only provide the measurements of reflected solar radiation at the top of the atmosphere (TOA), which are only used to retrieve aerosol optical depth (AOD) on the basis of assumed SSA. On the other hand, we would be able to retrieve SSA using satellite measurements with known AOD. In this study, a machine learning based algorithm is developed for the retrieval of SSA using joint visibility and satellite measurements. With meteorological and ancillary information, surface visibility can be converted to column AOD. Then combining this converted AOD with MODIS measured TOA apparent reflectance, we retrieve SSA at over 1000 stations worldwide. Our results show a great consistency with AERONET retrieved SSA but have a higher spatial resolution than it. We also applied our method to surface PM2.5 measurements and obtained satisfactory results. Our work generates a global aerosol SSA dataset with extensive coverage over land, which can be used for the improvement of climate models and the estimation of aerosol radiative forcing.