Supervised by:

Saeid Homayouni (Associate Member)

Research project description

Mapping Sea Ice in the Resolute Bay: A RADARSAT-2 PolSAR Approach

Introduction

In the dynamic environment of the Resolute Passage, precise observation of sea ice is crucial, as its constantly changing state is an essential indicator of climate change. Moreover, to ensure safe maritime navigation, it is imperative to have a good understanding of this complex northern landscape through accurate monitoring of sea ice. Despite advances in remote sensing technologies, significant challenges remain in mapping sea ice, particularly regarding the automatic selection of optimal parameters for Polarimetric Synthetic Aperture Radar (PolSAR) for accurate classification. This difficulty arises from the variability and complexity of different types of sea ice and their interactions with radar signals, complicating the task of effective classification

Objectives

The primary objective of this study is to analyze the sea ice conditions in the Resolute Passage, located in the Canadian Archipelago. To enhance classification accuracy, this study aims to identify the optimal set of Polarimetric Synthetic Aperture Radar (PolSAR) parameters for effective differentiation among these sea ice types.

Study Sites

Resolute Passage, situated in the Canadian Archipelago, is a crucial marine corridor characterized by its dynamic and complex polar ecosystem. This region is predominantly comprised of drifting sea ice, with patches of fast ice forming along the shores, coves, and bays. The passage experiences significant seasonal variations, resulting in diverse ice types, including new, young, first-year, and multi-year ice2. These varying ice conditions are influenced by local climate patterns, ocean currents, and temperature fluctuations, making the area a vital indicator of climate change.

Material and methods

This study uses six RADARSAT-2 SLC images acquired on September 20th, September 30th, and October 14th, 2017, with a resolution of approximately 10 meters and incidence angles ranging from 26.1° to 33.7°. Pre-processing steps included calibration, speckle filtering, polarimetric processing, and terrain correction. Building on research by [1], which analyzed visual interpretation through entropy and volume-to-surface scattering, we employed various polarimetric parameters such as SPAN, Entropy-Isotropy, Alpha, and the Yamaguchi decomposition for enhanced sea ice analysis. A manually labeled dataset identified different sea ice types based on World Meteorological Organization (WMO) nomenclature, including open water, new ice, and grey/grey-white ice. An explainable AI approach with SHapley Additive exPlanations (SHAP) is used to optimize the PolSAR parameter selection, evaluating their combined effects and providing insights into their contributions to sea ice type classification.

References

1] Shokr, M., Dabboor, M., Lacelle, M., Zagon, T., Deschamps, B. (2022). Observations from C-Band SAR Fully Polarimetric Parameters of Mobile Sea Ice Based on Radar Scattering Mechanisms to Support Operational Sea Ice Monitoring, Canadian Journal of Remote Sensing, 48:2, 197-213,

Research Site Coordinates