| Summary: | Over recent decades, there have been significant changes in Arctic sea ice, marked by a shift from thicker to thinner ice, and reductions in extent and volume. These changes, along with extended melt seasons, have accelerated sea ice melt and the ice-albedo feedback. This dissertation leverages Synthetic Aperture Radar (SAR) to analyze sea ice dynamics and thermodynamics at various scales. Focusing on the advanced melt season, this dissertation examines the relationships between sea ice geophysical properties and SAR backscatter signatures at C- and L-bands. Our analysis uses RADARSAT-2 (C-band) and ALOS-2/PALSAR-2 (L-band) imagery in different polarimetric modes to assess sea ice type separability and the effects of incidence angle and melt ponds on backscatter. We discover that C-band SAR is more effective early in the melt season, while L-band SAR provides clearer delineation of ice features later in the season. Additionally, the co-pol ratio (VV/HH) was found to be a consistent indicator of the melt ponds for FYI, despite environmental changes, such as wind-roughened melt ponds, for both frequencies and incidence angles in the near and far range. Utilizing our findings, we developed a dual frequency, dual-incidence angle sea ice classification approach using a random forest classifier. We achieved over 70% accuracy in sea ice type classification, validated by airborne measurements. This research underscores the value of a dual-frequency approach in improving sea ice classification, particularly for first-year and multi-year ice during advanced melt stages, while considering the evolution of deformed ice types in the advanced melt season. The findings contribute significantly to climate studies and operational services and are pertinent to future dual-frequency SAR missions. Graduate
|
|---|