The Dependence of the PolInSAR Degree of Coherence on Forest Parameters
Marco Lavalle(1), Domenico Solimini(1), Eric Pottier(2) and Yves-Louis Desnos(3)
(1) University Tor Vergata, Via del Politecnico 1, 00133 Rome, Italy
(2) University Rennes 1, 263 Avenue General Leclerc, 35042 Rennes, France
(3) ESA-ESRIN, Via Galileo Galilei 1, 00144 Frascati, Italy
Modeling backscattering from natural targets is a challenging task in the simulation of Synthetic Aperture Radar (SAR) polarimetric and interferometric images. A promising application of PolInSAR is the estimate of tree height of forests. To this end, the availability of a reliable reference model for the PolInSAR observables is crucial: first, the model can be used to identify the biophysical parameters that have more impact on the observables; second, a parametric analysis helps to design an efficient inversion approach. The degree of interferometric coherence between a pair of polarimetric SAR images (PolInSAR) is an observable of major importance for forest parameters retrieval.
The prediction of the PolInSAR coherence of forest stands has been addressed by two approaches. The first is the Random Volume over Ground (RVoG) model  which combines incoherently the ground contribution with a random canopy contribution. The model is based on four input parameters: tree height, mean canopy extinction, ground-to-volume amplitude ratio and ground topography. The second approach simulates PolInSAR images by modeling the observed surface, including forest, bare soil and grass. The ESA toolbox PolSARPro  provides this simulator that coherently adds the direct contributions from vegetation and soil, the one from the ground-vegetation interaction and that from the ground-vegetation-ground interaction, bot for forest and grass .The input parameters are derived from the forest characteristics, the soil surface and the SAR parameters and acquisition geometry.
We have studied the dependence of the complex coherence versus these parameters, in particular trees height, terrain slope and forest density. Diffent incident angles and bands (L-band and P-band) have been also considered in the analysis. We first use the total SAR return and show the results for the main polarization states and in terms of coherence region in the complex plane. Secondly, we consider only the Direct Vegetation return and the Direct Ground return. This is particularly useful to compare the simulation output with the single contributions of the RVoG model. The importance of this analisys is twofold: from one side it offers guidelines to improve the RVoG model or to design new forward coherence models; on the other side, it serves as basis for a new inversion procedure that will be illustrated as further step of our work.
 K.P. Papathanassiou and S.R. Cloude, "Single-Baseline Polarimetric SAR Interferometry," Geoscience and Remote Sensing, IEEE Transactions on, vol. 39, no. 11, pp. 2352–2363, Nov. 2001.
 M. Williams and Al., "Forest Coherent SAR Simulator within PolSARPro: an Educational Toolbox for PolSAR and PolInSAR Data Processing," Asian Conference on Remote Sensing, Kuala Lumpur, Malaysia, 2007.
 M.L. Williams, "The Theory of a forward SAR Model: Implementations, Applications and Challenges," EUSAR 2006, Proceedings of, Dresden, Germany, 15-18 May 2006.