A Physics-Based TOA Forward-Simulation Framework for Radiometric Stability Monitoring of Optical Satellites Over Forest Environments

Abstract

Reliable radiometric monitoring of Earth observation satellites is often constrained by reliance on bright desert reference sites, which do not represent dark and heterogeneous vegetation targets. This study presents a physics-based top-of-atmosphere forward-simulation framework that integrates an automated forest ground radiometry network (KoFReN), directional reflectance homogenization using the RossThick–LiSparse–Reciprocal (RTLSR) model, and Py6S radiative transfer simulation constrained by collocated AERONET observations. Rather than providing a direct on-orbit validation of CAS500-4, this study evaluates the applicability of the framework through a Sentinel-2–based proxy assessment prior to CAS500-4 launch. The framework links hemispherical ground measurements to satellite-view geometry for radiometric evaluation over complex forest environments. Validation over five RadCalNet sites yielded gain ratios close to unity under homogeneous benchmark conditions, supporting the internal consistency of the forward-modeling chain. Application across eight Korean forest sites showed clear scene dependence, with terrain-driven representativeness and viewing geometry emerging as major limitations in heterogeneous landscapes, while geometry-aware screening improved NIR agreement under favorable conditions. A site-level representativeness analysis further showed that within-ROI reflectance heterogeneity, terrain variability, and illumination geometry help explain the magnitude of residual errors and their reduction after site-specific QC filtering. A data-driven uncertainty analysis indicated that internal processing-chain uncertainty is dominated by cosine-response correction and RTLSR-based BHR-to-BRF directional conversion. These results suggest the potential of distributed forest radiometry networks to support proxy-based prelaunch and early operational radiometric stability assessment over vegetation-dominated landscapes.