The improved thin cloud detection using BRDF model based background reflectance from GOCI geostationary satellite imagery

In this study, the improved algorithm of thin cloud detection for geostationary ocean color imager (GOCI) satellite was developed to classify the thin cloud area over land area. The new cloud mask approach of GOCI satellite is required to expand its ocean dedicated application to other applications such for vegetation in land or aerosol optical properties (AOPs) in atmosphere due to its attractive shortwave wavelength bands of ocean color sensors. However, when trying to apply the advantages of the ocean color bands to the land area, only visible spectral bands of GOCI make it difficult to expand the land application the other way due to its limitation of cloud detection for relatively bright land surface. Furthermore, the geostationary of GOCI satellite has highly sensitive to geometry location of sun, meaning that high effective (Bidirectional Reflectance Distribution Function) BRDF effects make it also difficult to detect cloud mask in land surface due to its anisotropically scattered surface reflectance. In this paper, cloud mask algorithm of GOCI is proposed to consider those limitations by mainly using background surface reflectance from BRDF model. Therefore, minimum difference in reflectance between TOA and land as baseline of clear atmosphere and background surface reflectance underneath cloud were estimated from BRDF model. In conclusion, our new thin cloud detection is effectively detect the thin cloud over land surface area under limited ocean color bands of GOCI. The improved thin cloud detection algorithm of GOCI will be not only useful for following on instruments such as GOCI-II of Geo-KOMPSAT-2B and Sentinel 3 Ocean and Land Color Instrument (OLCL), but also applicable for existing geostationary satellites such as Geo-KOMPSAT-2A AMI, Himawari, and GOES-R as alternative cloud masking approach.