COMSGOCI (Geostationary Ocean Color Imager) is the first geostationary ocean color satellite in the world launched by South Korea in June 2010, which includes eight bands from the visible to the infrared band. GOCI aerosol optical thickness (AOT) at 555nm was retrieved by atmospheric radiative transfer model based on two-stream approximation algorithm. Due to GOCI without near infrared band and has a high solar elevation angle, solar zenith angle must be recalibrated to solve the earth system albedo, and the surface reflectance solved by quack atmospheric correction and recalculated backward scatter coefficient. Evaluation of GOCIAOT with AERONET measurements showed that the average error becomes 0.107 from the original 0.393, that means GOCI aerosol optical thickness can be more accurately with the advanced two-stream approximation. Taking the eastern China in 3 and 4 December 2013 for example, comparing the GOCIAOT at 555nm, MODISAOT retrievals at 550nm, NPPAOT at 550nm and AERONET data products indicated that: take the AERONET data as reference, the error of three kinds of satellite data can be ordered as following: MODISAOT< GOCIAOT< NPPAOT and the GOCI-MODIS shows a bias of 0.02917 with the GOCI-NPP. GOCIAOT is 0.05714 generally bigger than that of MODISAOT. NPP-GOCI deviation is 0.10253. The deficiency of MODIS is its low spatial resolution and the high concentration of AOT will be mistaken for a cloud area. However, GOCI can well reflect the concentration and distribution of aerosols. Therefore, GOGI can provide real-time dynamic monitoring on China Eastern atmospheric environment and the accurate time event information of haze for each process can be obtained. Finally, applied GOCI to the "8.12 Tianjin bombings" and to monitor the migration and dispersion of pollutant.