A sensor invariant atmospheric correction method for satellite images

Land surface reflectance is the fundamental variable for the most of earth observation (EO) missions, and corrections of the atmospheric disturbs from the cloud, gaseous, aerosol help to get accurate spectral description of earth surface. Unlike the previous empirical ways of atmospheric correction, we propose a data fusion method for atmospheric correction of satellite images, with an initial attempt to include the uncertainty information from different data source. It takes advantage of the high temporal resolution of MODIS observations to get BRDF description of the earth surface as the prior information of the earth surface property, uses the ECMWF CAMS Near-realtime as the prior information of the atmospheric sates, to get optimal estimations of the atmospheric parameters. It guarantees the correction is consistent cross different satellites image tiles and even cross different sensors. The validations against the AERONET sites are also show high correlation at around 0.9, with a RMSE of about 0.02.