An assessment of thin cloud detection by applying bidirectional reflectance distribution function model‐based background surface reflectance using Geostationary Ocean Color Imager (GOCI): A case study for South Korea

In this study, a new assessment of thin cloud detection with the application of bidirectional reflectance distribution function (BRDF) model‐based background surface reflectance was undertaken by interpreting surface spectra characterized using the Geostationary Ocean Color Imager (GOCI) over a land surface area. Unlike cloud detection over the ocean, the detection of cloud over land surfaces is difficult due to the complicated surface scattering characteristics, which vary among land surface types. Furthermore, in the case of thin clouds, in which the surface and cloud radiation are mixed, it is difficult to detect the clouds in both land and atmospheric fields. Therefore, to interpret background surface reflectance, especially underneath cloud, the semiempirical BRDF model was used to simulate surface reflectance by reflecting solar angle‐dependent geostationary sensor geometry. For quantitative validation, Cloud‐Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) data were used to make a comparison with the proposed cloud masking result. As a result, the new cloud masking scheme resulted in a high probability of detection (POD = 0.82) compared with the Moderate Resolution Imaging Spectroradiometer (MODIS) (POD = 0.808) for all cloud cases. In particular, the agreement between the CALIPSO cloud product and new GOCI cloud mask was over 94% when detecting thin cloud (e.g., altostratus and cirrus) from January 2014 to June 2015. This result is relatively high in comparison with the result from the MODIS Collection 6 cloud mask product (MYD35). Key Points Thin cloud detection algorithm over land area was suggested with geostationary ocean color satellite imagery Background reflectance of the surface underneath the cloud is simulated with bidirectional reflectance distribution function model Agreement between CALIPSO and new cloud detection algorithm was over 94% when detecting thin cloud (altostratus and cirrus) during study period