Multi-temporal Analyses of Polarimetric Scattering from Snow, Firn and Ice.
Brown, Ian
Stockholm University, SWEDEN

Typical descriptions of synthetic aperture radar (SAR) scattering from the accumulation area of a glacier assume that snow grains and ice inclusions, such as lenses or pipes, are the principal scatterers (Fung, 1994; Forster and others, 1999; Brown and others, 2005). This approach correlates grain size and the number density of scatterers, usually a function of snow or firn density, with backscatter intensity. Grains are assumed to be spheres and scattering is modeled using a Rayleigh approach that sees the volume as a cloud of scatterers (Fung, 1994). The application of the Rayleigh model to the snow and firn of Greenland has been challenged (Hoen and Zebker, 2000). Hoen and Zebker (2000) argued that scattering is better modelled using a model comprised of surface scattering from multiple, closely spaced, layers within the snow and firn. These models were specifically developed to explain scattering from the Greenland ice sheet but should also be applicable to the other large ice sheets and icecaps a with dry snow zone.

Uncertainties therefore remain regarding scattering processes below the upper snow surface and the contribution of layers, inclusions and grains to backscatter amplitude. The depolarisation of waves in snow and firn is also poorly understood. Little has been published on the decomposition of backscatter from snow and firn. Floricioiu and Rott (2001) investigated the co- and cross polarised backscatter responses from SIR-C and -X, and AIRSAR images over the Austrian Alps. They applied backscatter modelling to separate volume and surface scattering components but did not derive decompositions. In one of the few comprehensive analyses of PolSAR data over glaciers, Sharma et a.( 2011) decomposed scattering in airborne SAR imagery into surface, volume and sastrugi components. They derived a specialised Freeman-Durden decomposition for glaciers. Stebler et al. (2005) derived Alpha, Entropy and Anisotropy decompositions from L- and P-band data over the Swiss Alps. They found strong relationships between polarimetric-interferometric (PolInSAR) measurements and firn temperature. That said, there are few analyses of scattering processes using polarimetric SAR data.

In this paper multi-temporal C-band Radarsat-2 quad-pol images are analysed to investigate scattering processes. Scatterer decomposition and particularly Eigenvector/Eigenvalue decompositions are used to explore the relative contributions of surface and volume scattering and to infer dominant scatterers in a complex facies zonation. Backscatter decomposition and analysis is made with reference to in situ observations of the surface facies from shallow coring and ground penetrating radar profiles.

Figure Caption: Comparison between a Pauli, H-Alpha-Anisotropy and H-Beta2-Anisotropy decomposition over Blåmannisen, Norway. The 2nd Beta angle (Beta2) strongly emphasises scattering from layered dense firn emerging at the firn line. This region appears as a bright zone on the eastern side of the glacier.