A sensor-agnostic albedo retrieval method for realistic sea ice surfaces: model and validation

Abstract. A framework was established for remote sensing of sea ice albedo that integrates physics with high computational efficiency and can be applied to optical sensors measure appropriate radiance data. scientific machine learning (SciML) approach developed trained on a large synthetic dataset (SD) constructed using coupled atmosphere–surface radiative transfer model (RTM). The resulting RTM–SciML combines the RTM multi-layer artificial neural network SciML model. In contrast Moderate Resolution Imaging Spectroradiometer (MODIS) MCD43 product, this does not depend observations from multiple days single angular obtained under clear-sky conditions. Compared existing melt pond detection (MPD)-based retrieval, has advantage being applicable wide variety cryosphere surfaces, both heterogeneous homogeneous. Excellent agreement found between retrieval results measurements collected airplane campaigns. Assessment against pyranometer data (N=4144) yields RMSE = 0.094 shortwave while evaluation albedometer (N=1225) 0.069, 0.143, 0.085 broadband in visible, near-infrared, spectral ranges, respectively.