On Sea Ice Characterization by Multi-Frequency SAR Systems
Grahn, Jakob1; Eltoft, Torbjorn1; Brekke, Camilla1; Holt, Benjamin2
1UiT, NORWAY; 2JPL, UNITED STATES

Full-polarimetric synthetic aperture radar (SAR) systems allow us to link scattering processes to the physics of the medium imaged. We aim at identifying suitable polarimetric analysis strategies that let us compare the scattering mechanisms in sea ice at different radar frequencies.

This study will be based on two SAR data sets, which are both multi-frequency and multi-polarization. In 1987, the NASA AIRSAR system became operational. The system is now retired. The polarimetric mode became available in 1988. The AIRSAR system operated in fully polarimetric mode at three frequencies (wavelengths) P- (0.68 m), L- (0.25 m) and C-band (0.057 m) simultaneously. Results from a current investigation of a series of quad-polarimetric multi-frequency AIRSAR scenes, recorded from the Beaufort Sea on 2 December 2004, will be presented at the conference. The scenes contain multiple recently frozen sea ice leads [1, 2]. Age and thickness information about the sea ice types is derived from a coincident RADARSAT-1 time series and metrological information. We also have available multi-frequency measurements from the ESA ICESAR campaign in 2007 with the German E-SAR system. This dataset contains several sets of simultaneous measurements on C- and L-band. The L-band is quad-polarimetric, whereas the C-band is dual-polarization only. From this campaign we have limited amount of ground truth data.

The noise equivalent sigma nought (NESZ) is a measure of the radar cross-section equivalent to the instrument noise floor, and is a function of incidence angle. For AIRSAR, the NESZ lies in the range of -32 to -30 dB for C-band, -47 to -43 dB for L-band and -50 to -49 dB for P-band [4]. The signal level, as compared to the NESZ, is an important parameter to take into account when analyzing newly frozen sea ice types causing low backscatter signatures in SAR images. This will be carefully considered in our analysis.

First, a feature-based segmentation method, developed for multi-channel SAR data within our research group at University of Tromso, will be applied to define regions with distinct polarimetric properties in the AIRSAR data. Then the frequency dependencies of the different polarimetric signatures will be investigated.

A preliminary study of the AIRSAR co-polarization ratio indicates a benefit of combining multi-polarization and multi-frequency measurements for discrimination among the various new and young lead ice types [3]. Other features, such as the co-polarized phase difference, are also found useful [2]. More advanced target decomposition techniques aim at providing an interpretation of distributed targets and natural media in terms of dominant scattering mechanisms. Finally, we will look at the applicability of both model-based approaches, such as Freeman-Durden [5], and eigenvector-based decompositions, such as van Zyl [5] for the purpose of characterizing Arctic sea ice.

References:

[1] Bäck, D., "Analysis of Polarimetric Signatures of Arctic Lead Ice Using Data from AIRSAR and RADARSAT", MSc Thesis, Dept. of Radio and Space Science, Chalmers Univ. of Tech., Goeteborg, Sweden, 2008.
[2] Bäck, D. Holt, B., Kwok, R., "Analysis of C-band polarimetric signatures of arctic Lead Ice Using Data from AIRSAR and RADARSAT-2", in proc. of SEASAR, Frascati, Rome, Italy, 21-25 January 2008.
[3] Brekke, C., Holt, B., Jones, C., and Skrunes, S., "Towards Oil Slick Monitroing in the Arctic Environment", in proc. of PolInSAR, Frascati, Italy, 28 January - 1 February 2013.
[4] Lou, Y., "How Accurate Are My AIRSAR Data?", in Proc. of AIRSAR Earth Sci. and Appl. Workshop 2002.
[5] Lee, J.-S. and Pottier, E., "Polarimetric Radar Imaging. From Basics to Applications", CRC Press, Taylor & Francis Group, 2009.