How RADARSAT-2 ScanSAR Cross-Polarization Imagery can Enhance Sea Ice Motion Detection.
Komarov, Alexander; Barber, David
University of Manitoba, CANADA

Over the past three decades Arctic sea ice extent and thickness have drastically decreased. These changes in sea ice volume lead to increasing ice mobility. High resolution sequential synthetic aperture radar (SAR) images can be effectively used to accurately (within 1 pixel) detect sea ice motion in a large geographical area. Conventionally, ice motion has been derived from sequential co-polarization images (HH or VV). With the launch of Canadian RADARSAT-2 in 2007, a cross-polarization ScanSAR channel (HV) is available equally with the co-polarization (HH) one making the dual-polarization mode (HH-HV). The HH signal is sensitive to small scale roughness of sea ice compared to the HV signal which is primarily generated on large scale changes in ice topography. In terms of automated ice motion tracking, the cross-polarization channel may provide stable features undetectable by the HH image. Following this rationale our objective was to investigate how the HV imagery may enhance the ice motion tracking derived from HH images only.
First, we present main details of our sea ice motion tracking algorithm operating with sequential SAR images. The system was operationally implemented at the Canadian Ice Service. Then, we analyzed ice motion tracking results derived from ten RADARSAT-2 dual-polarization image pairs acquired over the Beaufort Sea area. Our analysis revealed that if the difference between the HV signal and the noise equivalent sigma zero (NESZ) in linear units exceeds some threshold, then the HV imagery is able to significantly enhance ice motion tracking results derived from the HH images. Otherwise, the HH channel is preferable to the HV channel for ice motion tracking. Furthermore, the additional ice motion vectors derived from the HV images have primarily medium or high levels of confidence. We also found, that our ice motion tracking system is not sensitive to the noise floor stripes parallel to the satellite track. The results suggest that further possible improvements of noise floor characteristics of future SAR platforms (e.g. Sentinel and RADARSAT Constellation) would result in even more reliable ice motion detection using cross-polarization imagery.