A Study on the background surface reflectance retrieval of near-UV wavelength using GK-2B/GEMS data

 Surface reflectance is the product of removing atmospheric scattering and absorption effects from the Top-Of-Atmosphere (TOA) radiation using the Radiative Transfer Model (RTM), and it refers to the reflectance according to the solar and satellite zenith angles at the time of observation. Surface reflectance is an essential input data for other Level-2 calculation algorithms such as aerosol, cloud, ozone, gas tracers, etc. Therefore, if the surface reflectance data has missing value, it will lead to missing other products that use it. However, when there are clouds in the satellite image, there is a problem with that blank pixels are generated because the surface reflectance cannot be calculated. Therefore, in this study, we conducted an algorithm to calculate background surface reflectance (BSR) without missing values with high accuracy using GK-2B/Geostationary Environment Monitoring Spectrometer (GEMS) data. The BSR is an estimate of the surface reflectance under specific observation conditions (solar and satellite zenith angles) and is a product that avoids the calculation precedence dilemma between AOD and surface reflectance. In many studies, the BSR is mainly calculated using the minimum reflectance method, but it has limitations in not considering the angular conditions at the time of observation and the reflectance characteristics of the ground surface. To overcome these limitations, a realistic BSR calculation was performed considering the anisotropic reflectance characteristics of the surface according to the observation conditions through bi-directional reflectance distribution function (BRDF) modeling.

 Surface reflectance, which is an input variable for BRDF modeling, was calculated based on the Look-Up Table (LUT) generated using the Second Simulation of Satellite Signal in the Solar Spectrum (6SV) RTM. At this time, LUT interpolation was additionally performed through the 6d-interploation technique to resolve discontinuities that may occur in LUT-based atmospheric correction. For BRDF modeling, the kernel-based Roujean model was used, and the optimal synthesis period for BRDF modeling considering the characteristics of the GEMS satellite was selected. To evaluate the accuracy of BSR, the simulated BSR through the BRDF model and the observed surface reflectance were compared, and it was confirmed that the BSR showed higher accuracy than the minimum reflectance method. In the future, the BSR produced through this study is expected to have a great impact on improving the calculation accuracy of aerosol and atmospheric products of GEMS satellites.