Toward the detection of oil spills in sea ice-covered waters using C-band radar remote sensing

• Main Author: Asihene, Elvis
• Other Authors: Stern, Gary (Environment and Geography), Mojabi, Puyan (Electrical and Computer Engineering), Sanchez-Azofeifa, Arturo (University of Alberta), Isleifson, Dustin, Gilmore, Colin
• Language: English
• Published: 2024
• Subjects: Radar; Oil Spill; Sea Ice; Remote Sensing; Arctic; Polarimetric parameters; PSO; SPM; New Ice; Electromagnetic Modeling; Sea ice; ice covered waters
• Online Access: http://hdl.handle.net/1993/38082

The Arctic ice-covered waters are becoming more vulnerable to oil spills due to climate-driven sea ice loss, which has increased vessel traffic and natural resource extraction. In preparation for future incidents, this thesis presents the research undertaken in the area of Arctic crude oil spill response, with the ultimate goal of accurately detecting and characterizing such spills in sea ice-covered waters using C-band radar remote sensing. The research focuses on three interconnected objectives: observation, discrimination, and modeling of surface-based C-band scatterometer data collected during experiments involving spilled crude oil in newly formed sea ice (NI) at the University of Manitoba’s Sea-ice Environmental Research Facility. The observational study seeks to establish a definitive relationship between radar signatures and the physical properties of oil-contaminated NI. The results show that multipolarization radar signatures exhibit distinct responses when oil is encapsulated within the ice, up until the oil migrates onto the ice surface. For instance, when oil is encapsulated within the ice, a 13-dB local maximum in cross-polarization was observed with a coincidental 9-dB drop in co-polarization backscatters. The discrimination study uses radar polarimetric parameters (such as entropy, mean-alpha, copolarization correlation coefficient, conformity coefficient, and more) to accurately differentiate between uncontaminated and oil-contaminated NI. The findings reveal that a threshold classification plane of 0.3 entropy and 18° mean-alpha effectively distinguishes oil-contaminated ice. To minimize oil spill false alarms, the copolarization correlation coefficient and conformity coefficient emerge as the most reliable parameters for detecting spilled oil events in NI-covered waters. The modeling study applies the small perturbation method and particle swarm optimization in an electromagnetic inversion strategy to estimate the oil thickness on NI surface through radar simulations and observations. The ...