Airborne and ground-based radar have been used extensively in the past to measure ice thickness and to investigate the internal structure of ice sheets in terms of layering. The main reflection mechanisms for internal reflections are changes in density, conductivity, and crystal orientation fabric, which alter thepermittivity of the ice. Linking the different mechanisms to the individual reflection horizons enables thededuction of glaciological parameters like accumulation rates or age-depth estimates. If no sample material from snow pits or ice-cores are available, multi-frequency and multi-polarization measurements must be applied to distinguish between the different reflection mechanisms. The backscattered power of horizons caused by changes in conductivity varies with the center frequency whereas in the case of horizons originating from changing crystal orientation the backscattered power is dependent on the polarization plane of the carrier signal.In this study we examine a sample data set near the German summer station Kohnen (drill site for theEPICA-EDML ice core) on the Antarctic plateau. The data were acquired with an airplane sliding on ground, producing varying incident polarization with a circular profile and several cross profiles with different headings. We find that the backscattered power changes with varying antenna orientation (i.e. polarization). In the upper third of the ice column the backscatter has two maxima with a 180° symmetry. The maxima align with the direction of minimal surface strain. At approximately 900 m depth the anisotropy is shifted by 90° in heading azimuth, with the maxima now being parallel to the maximum in surface strain. This dataset is unique, as airborne systems (primarily designed for the sounding of ice thickness) are usually not used for ground-based applications. The observed anisotropy appears clearly and is intriguing as the reason for it is entirely unknown. As primary suspects we consider the role of changing crystal orientation and ellipsoidal shaped air bubbles. The effect is visible from 200 1400 m. It appears distributed along the entire interval, and not restricted to individual layers. It seems that the polarization dependence becomes visible by a changing background level of the acquired signal, which is otherwise largely dominated by layer-like, polarization independent reflections. Hence we apply a (semi-analytical) volume scattering model in order to understand the different reflection mechanisms better. From ice-core measurements it is known that the crystals in the upper hundred meters are only weakly aligned (if at all), and it is unclear how the crystal orientation changes overshort depth intervals (~10 m). The rotation of the anisotropy coincides with the clathrate transition in the ice core and thus we first focus on the effect of anisotropic air bubbles. In an in-coherent approach we treat the ice matrix as a random medium and use the vector radiative transfer theory to incorporate boundary conditions. In a second step we model the effect of crystal orientation to estimate both, the degree of alignment and the statistical variance in the permittivity tensor needed to generate the observed pattern in backscatter. Doing so, we eventually aim at pinning down the mechanisms for the anisotropy in the upper interval, lower interval and the interrelation of the two by a shift of 90°.Anisotropic air bubbles as well as aligned crystal orientation allow to deduce stress and strain rates and a potential change thereof along depth. So far it is largely unclear, how surface strain rates relate with strain rates within the ice. If one of the two suspected mechanisms can be excluded or confirmed, this study may serve as a case study for future polarimetric surveys with low-frequency radars, in order to supply ice-sheet modelling with adequate boundary conditions - including changes in the internal structure of ice sheets along depth.