Estimation of the Vertical Structure of a Tropical Forest Using Basis Functions and Parametric SAR Tomography

SAR tomography represents a unique way to characterize forested areas from their 3-D density of reflectivity. As shown by numerous studies, coherent 2-D SAR images with intermediate horizontal resolution may be processed through spectral analysis techniques in order to focus 3-D cubes of reflectivity, and to provide an electromagnetic description of forets which is generally sufficient to estimates its main features, such as underlying ground topography, tree height and above-ground biomass. Nevertheless, a refined analysis of the vertical structure of a forest is usually highly limited by the vertical resolution and by the presence of side-lobes and focusing artifacts whose separation from the actual response may be problematic. As shown in this paper, direct deconvolution of the vertical impulse response using high-resolution techniques is an underdetermined inverse problem with an infinite number of plausible solutions. The use of basis functions, such as those proposed by Aguilera et al. [1], as well as a set of signal properties likely to be well adapted to the reflectivity of forested areas, allows to drastically reduce the size of the solution domain. This paper iuses an efficient iterative technique to estimate the intrinsic vertical reflectivity profile of forest measured at L and P bands