Evaluating recent advances in active microwave remote sensing for Arctic sea ice monitoring

  • Author / Creator
    Casey, John A
  • The objective of this research was to evaluate recent advances in active microwave remote sensing technologies in order to further refine the optimal radar parameters for sea ice monitoring in support of marine operations and climate research. To achieve this objective, data from recent active microwave remote sensing satellite missions, including RADARSAT-2, ALOS/PALSAR, TerraSAR-X and CryoSat-2, were analyzed to determine what new sea ice information could be derived from this current generation of satellite-borne active microwave remote sensing systems, relative to data from traditional single- and dual-polarized C-band synthetic aperture radar (SAR) systems. Specifically, this work focused on determining whether or not data from these sensors could be exploited to:
    i) differentiate between first-year and multi-year sea ice types during spring-summer melt, when wet snow and melt ponds on the ice surface can obscure the underlying ice;
    ii) differentiate between large-scale roughness features, which pose a substantial risk to marine operations, and small-scale roughness features; and
    iii) retrieve reliable estimates of sea ice thickness over both perennial (multi-year) and seasonal (first-year) ice regimes.
    Analyses of spatially coincident RADARSAT-2 and ALOS/PALSAR imagery acquired over landfast sea ice in the Canadian Arctic Archipelago during spring-summer 2009 indicate that C- and L-band SAR data provide complementary sea ice type information during the melt season, with each frequency providing improved separability between first-year ice (FYI) and multi-year ice (MYI) during different thermodynamic states. L-band data were found to provide enhanced ice type separability in the early stages of melt onset, where penetration through a slightly wet snow pack was achieved, and during advanced melt. The L-band data also provided improved capacity to delineate floe boundaries and to identify ridged ice throughout the melt season. It is therefore recommended that L-band data be acquired during the summer melt season to supplement C-band observations.
    Analysis of polarimetric RADARSAT-2 and TerraSAR-X data acquired over the Lincoln Sea, an area of predominantly thick deformed MYI, indicate that C- and X-band SAR data provide largely redundant sea ice information. Polarimetric SAR (PolSAR) parameters at both frequencies were found to be poorly suited for predicting the fraction of ridged ice, as the backscatter signatures of ridged ice (large-scale roughness features) and brash ice (small-scale roughness features) overlapped. However, numerous C-band PolSAR parameters were strongly correlated to the fraction of deformed ice, which includes both small- and large-scale roughness features. Multiple linear regression (MLR) models were developed to predict the fraction of deformed ice at the pixel scale from C-band data. Predicted deformed ice fraction values were used in a decision tree classification to separate areas of deformed ice from areas of level (undeformed) FYI and MYI. For pixels classified as level MYI, numerous C- and X-band backscatter intensity-based parameters were strongly correlated to MYI thickness. MLR models were therefore developed to predict MYI thickness from the PolSAR data. However, when assessed on independent test datasets (including in situ and airborne ice thickness measurements) the models were not found to be robust. Root-mean-square errors and mean biases were often > 2.0 m, indicating that the empirically derived MLR models are not suitable for predicting MYI thickness.
    Finally, analysis of CryoSat-2 data acquired over the Bay of Bothnia, a seasonal ice zone, indicated that the surface height measurements provided in CryoSat-2 SAR mode and SAR Interferometry (SARIn) mode level 2 data products are not suitable for use in deriving measurements of sea ice freeboard because the surface height profiles are strongly affected by snagging events. Analysis of CryoSat-2 level 1b (L1b) data indicate that reliable freeboard measurements can be retrieved from the L1b waveform data by implementing a threshold first maximum retracking algorithm. However, due to CryoSat-2’s coarse range resolution, minor snagging events may still occur. Ice thickness estimates were derived from the freeboard measurements according to the hydrostatic balance equation. For SAR mode, thickness estimates compared favourably to airborne ice thickness measurements and to thickness observations reported on ice charts. In contrast, SARIn mode data were not found to be suitable for the retrieval of ice thickness in this seasonal ice regime, due to the increased range uncertainty of the SARIn mode data.

  • Subjects / Keywords
  • Graduation date
    Fall 2018
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.