Assessment of forest biomass retrieval from Compact-pol SAR data

International audienceIn this paper, the potential of compact polarimetric data at low frequency for aboveground biomass retrieval is explored on an extensively studied maritime pine plantation forest situated on flat topography. Aboveground biomass was calculated at 55 stands using site-specific allometric equations. The site biomass level is up to 120 t/ha with a mean observed biomass of 66.7 t/ha. The P-band synthetic aperture radar (SAR) dataset acquired by the Office National d'Etudes et de Recherches Aerospatiales' airborne SAR system, known as RAMSES, is fully polarimetric and can be used to compute the compact polarimetric data associated with a circularly polarized transmit polarization and two mutually coherent independent polarizations. Information acquired with a compact polarimetric (CP) system is reduced compared with full polarimetry (FP) (i.e., the information in CP is represented by a two-dimensional complex vector resulting in four parameters; whereas, in FP, considering a monostatic case, the representation requires a three-dimensional complex vector with six independent parameters). The linear cross-polarized channel horizontal-vertical (HV) has been shown to be of significant relevance for forestry studies but is not directly measured in the standard CP modes. The challenge of this CP study was then to find alternate indicators specific to CP allowing accurate biomass retrieval. Such indicators are proposed for the compact mode for which the transmit polarization is circularly polarized and their sensitivity to biomass are quantified through a correlation analysis. The corresponding coefficients of determination are observed to be of the order of 0.9 comparable with the number obtained for FP data. Various estimation algorithms based on the regression curves were proposed and evaluated on the 55 test stands. The results obtained with CP data agree quite closely with those obtained with FP data with a measured accuracy in root mean square error within 6 t/ha. This shows that for biomass estimation, in the case of a flat topography, CP can be used instead of HV making it an interesting mode of operation for future spaceborne SAR such as the European BIOMASS mission