Main content area

Coherence-based SAR tomography for spaceborne applications

Nannini, Matteo, Martone, Michele, Rizzoli, Paola, Prats-Iraola, Pau, Rodriguez-Cassola, Marc, Reigber, Andreas, Moreira, Alberto
Remote sensing of environment 2019 v.225 pp. 107-114
climate change, data collection, ecosystems, forests, ice, image analysis, remote sensing, tomography, variance covariance matrix
Future SAR missions will provide three-dimensional images of semi-transparent media, such as vegetation and ice, through SAR tomography. Access to information on the internal structure of these volume scatterers is a key factor for a better understanding of ecosystem dynamics and climate change. Because of this, several concepts are nowadays examined to implement SAR tomography in a spaceborne framework.In order to do that, it is necessary to gather different observations of the area of interest. Unfortunately, a consequence of the time that elapses between acquisitions is that the electromagnetic properties of the medium may vary. This implies that, there may be inconsistencies in the acquired data, leading to errors in the final inversion. A solution to partially cope with this temporal decorrelation, is to acquire data employing two or more sensors operating with a reduced (or even absent) temporal gap and then to collect several acquisitions at different time instants. By means of this imaging concept, the required line-of-sight diversity is granted and the desired resolution in the height direction ensured. In this way, sets of temporal decorrelation-free interferometric coherences can be built and the vertical scattering profile can be retrieved via coherence-based tomography.This contribution analyzes a two-sensor system like TanDEM-X (Krieger et al., 2007), Tandem-L (Moreira et al., 2015), or SAOCOM-CS (Davidson et al., 2014). In particular, the potential of coherence-based tomography are shown with data acquired with the TanDEM-X sensors for boreal and Amazon forest. In addition, a technique to partially cope with temporal decorrelation through covariance matrix filtering is also presented in the paper.