For the characterization of Arctic scenario, synthetic aperture radar (SAR) sensing plays a key-role. SAR satellites provide measurements regardless of weather conditions and natural illumination with high spatial resolution. In particular, the Sentinel-1 mission is considered an important operational data source for sea ice classification. Sea ice classification from SAR relies on the fact that different ice types have different signatures in the image. These are generally related to the physical nature of the ice surface, such as roughness or dielectric properties, or system and geometric properties, such as frequency, polarization, noise level, and incidence angle (IA). Unfortunately, signal ambiguities and additive system noise might vary across each single scene, making the automatic interpretation of the scenes very cumbersome, which is why ice charting is still a partly manual process.
It is then paramount to address the operations of denoising in SAR data analysis, so to enable an accurate understanding of the Arctic scenario. This would then allow to provide a precise characterization of the physical phenomena occurring over the regions of interest. To achieve this goal, it is also important to obtain a robust understanding of the physical models causing the signals to be analyzed, so to improve the effectiveness of the aforesaid denoising phase. In this respect, considering multiple data sources (e.g., optical remote sensing, in situ measurements) could help in designing the automatic data analysis schemes. Finally, it is important to address the uncertainty of the outputs of the data analysis scheme, i.e., to provide a degree of confidence on the analysis outcomes. COSMIC will address these points, by taking advantage of the expertise gained by the group at Telecom Paris in efficient and effective SAR-based data analysis, and of the expertise of the group at UiT in characterization of Arctic scenarios by multimodal data analysis.