Sea ice type and open water discrimination using dual co-polarized C-band SAR

Abstract

We investigate the utility of dual co-polarized C-band synthetic aperture radar (SAR) imagery for discriminating sea ice types and open water during winter. We base our analysis on ENVISAT ASAR alternating vertical and horizontal polarization (VV, HH) medium-resolution imagery of thin sea ice, first-year sea ice, multiyear sea ice, and open water in the Canadian Arctic. We introduce a methodology for generating colour composite imagery (based on VV, HH, and co-polarization ratio channels) that substantially enhances visual discrimination. We statistically compare sampled vertical and horizontal backscatter from sea ice types and open water to assess backscatter magnitudes and polarization differences. The latter are presented as co-polarized ratio (VV/HH) values in decibels (dB). We conclude by presenting a decision-tree classifier using estimated statistical thresholds. Open water is unambiguously discriminated (>99% accuracy) from all sea ice types, except thin sea ice, using a co-polarized ratio threshold of 2 dB. Thin sea ice exhibits high variability in backscatter magnitude and polarization difference, making its discrimination from multiyear sea ice, rough first-year sea ice, and open water ambiguous. Thin sea ice is effectively discriminated (93% accuracy) from smooth first-year sea ice using a co-polarized ratio threshold of 1.3 dB. Smooth snow-covered first-year sea ice exhibits polarization differences that are attributed to volume scattering mechanisms within the brine-wetted snow cover.