Three-stage inversion process for polarimetric SAR interferometry

The authors provide a new geometrical approach for the inversion of a two-layer coherent scattering model, widely used for the interpretation of polarimetric interferometric SAR data. It has been shown in several recent publications that, by using interferograms in multiple polarisation channels, estimation of vegetation height, underlying ground topography and mean extinction is possible. Furthermore, this can be achieved with a single frequency sensor without the need for a separate reference DEM, other a priori information or the use of data-specific regression formulas. The authors first review the details of this approach and then develop a three-stage inversion procedure to illustrate the steps involved in parameter estimation. They then consider several possible sources of error in the inversion. In particular, they concentrate on the effects of vertical tree structure and on the effects of temporal decorrelation on inversion accuracy. It is shown that the former leads to errors, mainly in the extinction estimation, while the latter does not change the model structure but reduces the available parameter set and increases the variance of the parameter estimates. Finally, the new algorithm is applied to simulated vector coherent SAR data for a random canopy.