Soil Moisture Characterization using Multi-Angular Polarimetric Radarsat-2 Datasets
Wang, Hongquan1; Allain, Sophie2; Meric, Stephane1; Pottier, Eric2
1Institut National des Sciences Appliquees de Rennes, FRANCE; 2University of Rennes 1, FRANCE

The objective of this study is to examine the potentials of multi-angular polarimetric Radarsat-2 datasets for characterizing soil moisture over bare surface. Physical scattering, semi-empirical and empirical models are achieved in the past to interpret the depolarization of land scattering and to separate the soil moisture effect from surface roughness effect. Compared with conventional retrieval approaches that evaluate the total backscattering signature, the innovative soil parameter inversion algorithms focus more on scattering mechanism separation using target decomposition technique. It is demonstrated that at high frequency, the dominant polarimetric scattering angle alpha1 is only influenced by soil permittivity and incidence angle, after canceling out the influence of surface roughness [1]. Moreover, the correlation between HH and VV polarization is also reported to depend more on soil moisture than roughness. Thus, based on these kinds of polarimetric parameters, the direct relationships can be established for soil moisture characterization. Compared with single incidence angle image, multi-angular configuration takes the advantage to increase the information dimension, and to reduce the ambiguities of parameters estimation. Thus, it is anticipant that the exploration of datasets acquired from multi-angular condition may be achievable to characterize soil information with an improved robustness. In this study, we propose to evaluate the polarimetric parameter difference between two incidence angle observations, with the expectation to derive its sensitivity to dielectric constant and to restrain the influence of surface roughness.

In this research, Radarsat-2 images acquired in the end of March (October) / beginning of April (November) with four different ranges of incidence angle over test site Pleine-Fougères (in France) are explored using polarimetric decomposition methodology for polarimetric information extraction. The corresponding meteorological records are examined to guarantee the soil moisture is nominally invariant among multi-angular SAR observations. Moreover, in situ land cover investigation and soil moisture measurements are carried out for a further quantitative sensitivity analysis. Theoretical scattering model IEM is meanwhile compared with the real data in order to interpret the multi-angular behaviors of polarimetric parameters derived from Radarsat-2 data.

The results show that alpha1 derived from Radarsat-2 decreases with soil moisture increasing, and its dynamic range become wider when using a higher incidence angle configuration. Nevertheless, as the incidence angle increases, the entropy also increases that restrain the soil moisture discrimination abilities. Therefore, entropy and high incidence angle should make a compromise for the selection of incidence angle. The alpha1 difference between a small incidence angle and another big incidence angle is analyzed and found to decrease with soil moisture. It is noted that as the incidence angle difference becomes larger, the sensitivity of multi-angular parameters to soil moisture becomes higher. The IEM model simulation demonstrates that for smooth surface, alpha1 increases with soil moisture, indicating the influence of roughness on polarimetric parameter alpha1 can not be negligible for smooth field. In contrary, for rough surface, the behavior of alpha1 in IEM simulation is in accordance with the Radarsat-2 datasets except for different magnitude level. In addition, the correlation between HH and VV shows an increasing variation trend when the soil moisture increases using IEM model simulation, but when we examine this correlation coefficient in Radarsat-2 dataset, an inverse phenomenon occurs. It is also observed that as the incidence angle increases, the correlation between HH and VV decreases, suggesting a moderate incidence angle is suitable to characterize soil moisture under the condition that both alpha1 and correlation coefficient are used. It should be noted that the difference on HH and VV correlation between two incidence angle observations also increases with soil moisture, indicating that multi-angular polarimetric datasets have a potential to characterize soil moisture.

In conclusion, this paper demonstrates the potential of the combination of various incidence angles to help us get a better characterization of soil moisture.

[1]Sophie Allain. Caracterisation d'un sol nu a partire de donnes SAR polarimetriques etude multifrequency et multi-resolutions. PhD thesis, Universite de Rennes 1, 2003.