Sea-Ice Surface Types Characterization and SAR Volume Backscattering Properties in Response to Sea States Interactions Using Structural Feature Fusion

Abstract

Despite extensive research in polar remote sensing, there is still no consensus on how sea-ice evolves in response to interactions with ocean waves, nor is there agreement on the most effective way to characterize it based on fluctuations in height profile, both of which compromise interpretations of its texture properties when observed by synthetic aperture radar (SAR). This study aims to address these knowledge gaps by investigating the sea-ice modeling, marginal ice zone (MIZ) categorization and identifying an optimal approach to extract its freeboard profile and relevant backscattering properties under varying sea state conditions. To accomplish this, a numerical approach for modeling random polar media (RPM) is proposed, benefiting from transform-domain fusion method decompositions in conjunction with a convolutional neural network to jointly reconstruct a reference MIZ model in response to wind waves. To ensure consistency in surface structure, the fusion scheme incorporates a point-wise processing principle technique. The reconstructed reference sea-ice model is then subjected to an inverse problem solution analysis to generate the SAR volume backscattering dataset. The simulation results show that the entire routine, encompassing RPM modeling, MIZ characterization, structural fusion, and SAR raw data generation, yields promising results in the realm of sea-ice remote sensing. It reduces dependence on empirical methods, which typically necessitate polar field experiments that are limited by access constraints, or incorporate nonuniform and time-varying media characteristics into large-scale theoretical observation.