Speaker
Description
SAR tomography is a useful technique for monitoring and estimating forest geophysical parameters, as demonstrated by airborne measurements [1]. The future ESA BIOMASS mission aims to map the above-ground biomass and tree height of tropical forest on a global scale and to estimate the underlying ground topography. To do so a tomographic imaging mode will be applied to multi-baseline PolInSAR data acquired by the mission. These geophysical descriptors are derived from an estimate of the reflectivity distribution in the vertical dimension, typically computed using non-parametric spectral estimation techniques such as beamforming or the Capon technique [1,2].
Parametric tomography:
However, these techniques have a limited resolution and can be affected by spurious peaks and sidelobes. To overcome these hindrances, adaptive estimation approaches have emerged that directly model the vertical reflectivity distribution using a combination of functional bases such as orthogonal wavelets [3]. As shown in [4], such a technique is very well adapted to the observation of tropical forests in P-band and models the vertical reflectivity of the forest with a very limited number of contributions. Nevertheless, this estimation approach is computationally intensive and requires hyperparameter tuning, which may be specific to a particular type of forest cover. In the same paper [4], the authors demonstrated that the vertical reflectivity profile can be effectively modelled with two contributions, confirming the arbitrary hypothesis derived in the framework of polarimetric SAR interferometry and polarimetric SAR tomography in [5,6]. In this paper, we propose to characterise the tropical forest using a single polarisation technique at P-band based on the use of two contributions: a very narrow reflectivity peak corresponding to the ground and a bell-shaped function representing the canopy. Therefore, several types of functions are investigated to model the volume response such as the exponential shape [5], the gaussian profile [7] with varying widths including an infinitively narrow width, or the rectangular profile.
Comparison of profile model performance:
This approach is applied to data sets acquired over two different continents. The first one was acquired by ONERA in 2009 in the frame of the TropiSAR campaign over French Guiana (South America) and the second one by the DLR F-SAR sensor in Gabon (Africa). In both cases, data sets are simulated with a degraded resolution equivalent to that of the future BIOMASS mission SAR images. In the high-resolution airborne cases, results show a good retrieval performance for ground topography and tree height estimation for all the functional profiles studied. A standard deviation of 1.5 m is obtained for the ground and 2.6 m for the tree height estimation. All investigated profiles are found to be suitable to adequately reconstruct the reflectivity distribution in height. The performances obtained are comparable to those obtained by PolTomoSAR at full resolution, and this approach is less complex. In the context of BIOMASS like resolution, the results are finer with this approach. This is because it doesn’t have to choose between the ground contribution and the volume contribution, which is modified by the BIOMASS resolution.
References:
[1] H. Aghababaei, G. Ferraioli, L. Ferro-Famil, Y. Huang, M. Mariotti D’Alessandro, V. Pascazio, G. Schirinzi, S. Tebaldini, “Forest SAR Tomography : Principles and Applications”, IEEE Geoscience and Remote Sensing Magazine, Volume 8, Issue 2, pp30 – 45, June 2020
[2] F. Gini, F. Lombardini, “Multibaseline cross-track SAR interferometry : a signal processing perspective”, IEEE Aerospace and Electronic Systems magazine, Volume 20, Issue 8, pp71 – 93, August 2005
[3] E.Aguilera, M. Nannini and A.Reigber, “Wavelet-based compressed sensing for SAR Tomography of forested areas”, IEEE Transactions on Geoscience and Remote Sensing, Volume 51, Issue 12, pp5283 – 5295, December 2013
[4] L. Ferro-Famil, Y. Huang and N. Ge, “Estimation of the vertical structure of a tropical forest using basis functions and parametric SAR Tomography”, in IGARSS 2022 – 2022 IEEE International Geoscience and Remote Sensing Symposium, July 2022, pp.599-602
[5] S.R. Cloude and K.P. Papathanassiou, “Three-stage inversion process for polarimetric SAR interferometry”, IEE Procedings – Radar, Sonar and Navigation, Volume 150, Issue 3, pp125 – 134, June 2003
[6] S. Tebaldini, “Single and multipolarimetric SAR tomography of Forested Areas : A parametric approach”, IEEE Transactions on Geoscience and Remote sensing, Volume 48, Issue 5, pp2375 – 2387, May 2010
[7] F. Garestier and T. Le Toan, “Estimation of the Backscatter Vertical profile of a Pine Forest Using Single Baseline P-band (Pol-)InSAR Data”, IEEE Transactions on Geoscience and Remote Sensing, Volume 48, Issue 9, pp3340-3348, September 2010
