A comparison of L- and P-band PolSAR observations of a sub-polar ice-cap

Abstract

Polarimetric SAR observations at long-wavelengths (e.g. L- and P- band) provide the possibility to characterize the near-surface structure of glaciers and ice sheets. The knowledge of such information is crucial for understanding and monitoring the dynamic of ice masses [1]. With dry snow conditions, SAR signal penetrates several meters through snow and ice, interacting with buried features, like ice lenses, air bubbles, etc. The structure of subsurface layers determines the penetration depth of the SAR waves. In a polarimetric SAR (PolSAR) sensor, the penetration capability of microwaves is combined to the additional information brought by the polarization diversity, which provides high sensitivity to scatterers properties [2]. This paper addresses the interpretation of L- and P- band PolSAR measurements over the Austfonna ice-cap, Svalbard, by using the polarimetric scattering model presented in [3]. In detail, two test sites are investigated, which are located in different glacier zones [4] of the ice-cap. Experimental SAR data were collected by the airborne E-SAR sensor of the German Aerospace Center at L- (1.3 GHz) and P-band (0.35 GHz) over the percolation zone and the superimposed ice region of Austfonna, in spring 2007 during the ICESAR campaign. First, a qualitative polarimetric data analysis is performed based on a set of descriptors which includes backscattering coefficients, co-polarization ratio, co-polarization phase difference, entropy and mean alpha angle. Differences appearing between the two sites at both frequencies are discussed and related to the differences in subsurface structure with the support of ground measurements. A polarimetric decomposition, based on the scattering models proposed in [3], is also applied to the SAR data to provide a quantitative estimation of physical parameters at both locations. A final discussion addresses the incongruences observed by comparing the results obtained at L- and P-band over each site. The difference in wavelength turns out to generate significant changes in the observed scattering properties, suggesting that the adopted scattering model fits better the L-band case. In fact, one of the assumptions made in [3] is that only single scattering occurs in order to consider the different scattering components uncorrelated. At P-band, this hypothesis might not hold as the radar wavelength is about 86cm and the scatterers might be not sparse enough to avoid multiple interactions, even in the case of the percolation zone where very few ice lenses and pipes per square meter are expected [4]. Furthermore, additional scattering mechanisms generated in deeper layers, which are not sensed at L-band, might significantly contribute to the total P-band backscattering. REFERENCES [1] W.S.B. Paterson, The Physics of Glaciers, 3rd ed. Oxford, U.K.: Pergamon Press, 1994. [2] S. Cloude, Polarisation: Application in Remote Sensing. London, U.K.: Oxford Univ. Press, 2010. [3] G. Parrella, I. Hajnsek and K. Papathanassiou, “On the interpretation of L- and P- band PolSAR signatures of polythermal glaciers”, in Proc. Of 6th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry (PolinSAR), Frascati, Italy, 2013. [4] C.S. Benson, “Stratigraphic studies in the snow and firn of the Greenland ice sheet”, Ph.D. dissertation, California Institute of Technology, Pasadena, California, 1960.