POLARIMETRIC ANALYSIS FROM COMPACT-POL MEASUREMENTS: POTENTIAL AND LIMITATION M-L. Truong-Loï 123 , P. Dubois-Fernandez 1 , E. Pottier 2 , S. Angelliaume 1 , and J-C. Souyris 3 1. ONERA, Centre de Salon de Provence, France 2. IETR CNRS 6164, Université Rennes1, France 3. CNES, Toulouse, France ABSTRACT Global warning is now known to be the major environmental issue mankind will have to face in the next decade. Monitoring of vegetation and biomass is clearly an essential piece of information required at all levels ranging from the scientific studies to understand and forecast, to the political actors and government leaders responsible for drafting remediation policies and evaluating their impact. Microwave remote sensing with the low-frequency SAR technique can provide a useful characterization of forest (spatial coverage, species, density, height…) at a global scale, relying on the all-weather imaging capabilities of SAR linked with the significant penetration of the low-frequency EM wave in the canopy. The published techniques for forest characterization from low frequency SAR data include radiometry inversion, polarimetric inversion based on the anisotropy parameters and PolInSAR Random Volume Over Ground inversion. In this paper, we will more specifically concentrate on the PolSAR technique and the impact of ionospheric effect. Keywords - SAR, biomass, compact polarimetry mode, ionosphere, calibration, bare surfaces, soil mositure. 1. INTRODUCTION Compact polarimetry has been shown to be an interesting alternative mode to full polarimetry when global coverage and revisit time are key issues [1-2-3]. It consists on transmitting a single polarization, while receiving on two. Several critical points have been identified, one being the Faraday rotation correction and the other the calibration. When a low frequency electromagnetic wave travels through the ionosphere, it undergoes a rotation of the polarization plane about the radar line of sight for a linearly polarized wave, and a simple phase shift for a circularly polarized wave. In a low frequency radar, the only possible choice of the transmit polarization is the circular one, in order to guaranty that the scattering element on the ground is illuminated with a constant polarization independently of the ionosphere state. This will allow meaningful time series analysis, interferometry as long as the Faraday rotation effect is corrected for the return path. In full-polarimetric (FP) mode, two techniques allow to estimate the FR: Freeman method using linearly polarized data [4], and Bickel and Bates theory based on the transformation of the measured scattering matrix to a circular basis [5]. In CP mode, an alternate procedure is presented which relies on the bare surface scattering properties. The bare surfaces selection is therefore essential for Faraday rotation estimate and will be done with the conformity coefficient. Using CP data this coefficient is FR invariant. This coefficient is compared to two published FP decompositions, Cloude-Pottier and Freeman-Durden [6-7], to assess its potential in distinguishing three different scattering types: surface, double-bounce and volume. The performances of the bare surfaces selection and FR estimation are evaluated on PALSAR and RAMSES data. Once the bare surfaces are selected and Faraday angle estimated over them, the correction can be applied over the whole scene. The algorithm is benchmarked against both FP techniques. Then, we show that an application is possible directly with CP data without resorting to FP data reconstruction. In the last part of the paper, we study how to calibrate the CP system and correct for both cross-talk and Faraday effects. 2. SELECTING BARE SURFACES The selection of bare surfaces from CP data can be achieved based on the conformity coefficient which has been shown to be invariant with Faraday rotation. This coefficient, used in CP mode (ȝ CP ) as well as FP mode (ȝ FP ), allows discriminating the three main scattering types (surface, double-bounce and volume scattering) and is expressed as: ( ) CP VV HV HH HV VV HH FP RV RV RH RH RV RH RV RV RH RH RV RH CP S S S S S S S S S S S S M M M M M M μ μ μ ≅ + + − = + = + = 2 2 2 2 * * * * * * * 2 ) Re( 2 Im 2 Im 2 (1) where M RH and M RV are the measured scattering elements, considering a right circular transmission and two linear V - 1 978-1-4244-3395-7/09/$25.00 ©2009 IEEE IGARSS 2009