Towards Forest Vertical Structure Monitoring From Space: First Experiments With Multi-Baseline TandDEM-X Data

Abstract

The estimation of the vertical structure of forests from multibaseline SAR data becomes particularly challenging when addressed in terms of a space borne mission implementation. In particular, the lack of multiple space borne SAR configurations able to perform simultaneous acquisitions, combined with (temporal) scene decorrelation, reduce the number of suitable acquisitions in a realistic space borne scenario drastically. The acquisition and analysis of Pol-InSAR data without the disturbing effect of temporal decorrelation on a global scale is now possible by means of the TanDEM-X mission, which provides for the first time single pass (single- and dual-) polarimetric interferometric data from space. After two years of mission, first time series with variable baseline over the same forest site are available, allowing to (1) explore their information content, (2) assess penetration capabilities, (3) asses scattering model assumptions, and (4) estimate vertical structure and monitor its dynamics. Particularly crucial is the penetration of the X-band electromagnetic waves into the forest layers. This aspect was evaluated in a previous paper by analysing dual-pol data and by calculating the phase difference between the polarization with lowest and highest ground contributions. Beyond heterogeneity of forest stands and seasonal effects (e.g. leaf fall, freezing conditions …), results demonstrated enough penetration capabilities at X-band to allow Pol-InSAR inversion. Given these penetration potentials at X-band, time has come to evaluate in more detail the tomographic capabilities offered by a set of tandem acquisitions with variable baseline. As a set of temporal decorrelation-free coherences corresponding to different vertical wavenumbers are available for processing, the estimation of the vertical profile can be recast in the framework of coherence tomography. The simplest way to obtain the vertical profiles consists in calculating the Fourier transform of the available coherence set. However, the Fourier imaging suffers from inflated (positive and negative) sidelobes due to the non-uniformity of the coherence samples in the wavenumber domain. As a consequence, different algorithms should be set up in order to obtain an imaging with higher constrast and height resolution. For instance, the vertical structure could be either extracted by model based inversion from the interferometric coherences or by approximating the structure function through a weighted sum of a series of (orthogonal) basis functions, as in the Polarization Coherence Tomography (PCT). Alternatively, one could also apply deconvolution-based algorithms, which do not need to assume particular scattering models. This work is focused on the derivation of vertical vegetation profiles from TanDEM-X data. First of all, a performance comparison of different coherence tomography algorithms (model-based and not) will be carried out, and pros and cons evaluated. Furthermore, the obtained tomograms will be analysed in order to derive first insights about further potentials of the TanDEM-X mission for the characterization of the 3-D structure of forests.