Speaker
Description
Tomographic Synthetic Aperture Radar (SAR), TomoSAR, is an electromagnetic imaging technique able to map the 3D reflectivity of complex environments. It is generally implemented using radar signals acquired over a 2D aperture, i.e. using a set of coherent 2D SAR images measured from slightly shifted trajectories. Forested environments can be efficiently characterized using TomoSAR acquisitions operated at lower frequency bands, such as P or L bands, as larger wavelength values ensure both a deep penetration of dense forest covers, as well as a better preservation of signal coherence over time. Polarimetric TomoSAR (PolTomoSAR) uses polarization diversity in order to further discriminate signals originating from the canopy or from the underlying ground of a forest. Several applications of PolTomoSAR advanced signal processing techniques on data acquired during airborne campaigns, have demonstrated that this type of configuration could be exploited in order to estimate some key parameters of forest covers, such as tree height, ground topography, biomass, structure . . .
The imminent launch of the European Space Agency (ESA) BIOMASS spaceborne mission, that includes a polarimetric SAR device operating at P band, will provide a unique opportunity to accurately study densely forested regions, and in particular tropical regions. The estimation of the performance bounds, i.e. the best achievable precision for a given observed medium and given acquisition conditions, reveals a fundamental importance for both the evaluation of the different existing estimation approaches, but also for the validation of the mission products.
This paper proposes to first set up a methodological background for the computation of PolTomoSAR performance bounds that remains valid for single-polarization observations, multi-polarized ones, in PolinSAR configuration, or using more than two baselines. The performance of different modes is evaluated using Bayesian inference [1], and statistical processing classically used for the resolution of inverse problems [2]. Among all the descriptors of a forest cover, three main parameters are considered: the Digital Terrain Model (DTM), i.e. the topography of the underlying ground, the Canopy Height Model (CHM) or tree height and the Above Ground Biomass (AGB). As AGB is not directly estimated by PolTomoSAR imaging, various proxies are investigated [3].
The second part of this work concerns the influence of spatial resolutions, i.e. in range and azimuth directions, on the performance of PolTomoSAR. The particular case of BIOMASS with resolution of 12.5 m in azimuth and 25 m in slant range is investigated and compared with airborne measurements with a resolution of approximately 1m. Deteriorated resolution properties are accounted for using two approaches. The first one is based on an low-resolution adaptation of forest structure description with and unchanged theoretical derivation of the performance limits. The second technique extend the modeling of the forest reflectivity to include an additional axis in the inference process.
The validity and accuracy of the proposed methods are illustrated using the PolTomoSAR data set acquired at P band over the tropical forest test site of Paracou, in French Guiana, during the TropiSAR campaign in the summer 2009, using the ONERA’s SETHI system. It consists of six polarimetric and interferometric images having a spatial resolution of 1.5m in azimuth and 1.2m in range [4]. From this airborne data set, images having the spatial resolution of spaceborne BIOMASS acquisitions, 12.5m in azimuth and 25m in range [5, 6], are generated.
[1] George Casella and Roger Berger, Statistical Inference, Duxbury Resource Center, June 2001.[2] Albert Tarantola, Inverse Problem Theory and Methods for Model Parameter Estimation, Society for Industrial and Applied Mathematics, USA, 2004.
[3] D. Ho Tong Minh, T. L. Toan, F. Rocca, S. Tebaldini, M. M. d’Alessandro, and L. Villard, “Relating p-band synthetic aperture radar tomography to tropical forest biomass,” IEEE Transactions on Geoscience and Remote Sensing, vol. 52, no. 2, pp. 967–979, 2014.
[4] Pascale Dubois-Fernandez, Hélène Oriot, Colette Coulombeix, Hubert Cantalloube, Olivier Plessis, Thuy Le Toan, Sandrine Daniel, Jerome Chave, Lilian Blanc, and Malcolm Davidson, “TropiSAR, a SAR data acquisition campaign in French Guiana,” 07 2010, pp. 1 – 4.
[5] Maciej J. Soja, Shaun Quegan, Mauro M. d’Alessandro, Francesco Banda, Klaus Scipal, Stefano Tebaldini, and Lars M.H. Ulander, “Mapping above-ground biomass in tropical forests with ground-cancelled p-band sar and limited reference data,” Remote Sensing of Environment, vol. 253, pp. 112153, 2021.
[6] L. Ferro-Famil, Y. Huang, L. Villard, T. Le Toan, and T. Koleck, “Comparison of biomass acquisition modes for the characterization of forests,” in 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, 2021, pp. 1545–1548.
