Soil moisture retrieval over periodic surfaces using PolSAR data
Sandrine Daniel(1), Sophie Allain(1), Laurent Ferro-Famil(1) and Eric Pottier(1)
(1) IETR Laboratory, UMR CNRS, Campus de Beaulieu - 263 Avenue General Leclerc, 35042 Rennes Cedex, France
Polarimetric SAR data at L-band are known to be particularly well
adapted to the study of natural surface physical properties, like
moisture content and roughness. The analysis of polarimetric SAR
data acquired by the DLR's E-SAR sensor during two different campaigns,
over two agricultural sites (Alling in 2000 and Görmin in 2006,
both located in Germany), revealed that a non negligible proportion
of agricultural fields are affected by a so-called Bragg resonance
This effect, which influences strongly the polarimetric
response of crops, is due to the coherent summation of simultaneously
constructive or destructive contributions from a set of scatterers.
This Bragg phenomenon occurs over agricultural fields showing periodic or
quasi-periodic patterns, like plowing or irrigation rows and may strongly
influence the estimation of soil moisture from SAR data.
In this paper, agricultural fields whose response is corrupted by the
Bragg resonance phenomenon are first detected and an adapted method is proposed
to estimate soil moisture from resonant scattering patterns.
The Bragg resonance detection scheme is based on the fact that periodic
structures may show a resonant behavior for a very narrow domain in illuminating
frequency and azimuth look-angle. Indeed a field may resonate for only one part
of the frequency domain covered by the emitted chirp and for a given azimuth position with respect to the sensor.
The SAR images of the test areas are processed through a two-dimensional
continuous Time-Frequency (T-F) decomposition in order to estimate the spectral
polarimetric information over each pixel, i.e. the response of each pixel in the
frequency-azimuth look angle domain. The T-F stationary behavior is then tested
using polarimetric detectors, based on maximum likelihood ratios.
It is then shown that soil moisture estimation results using classical
methods are heavily affected by Bragg resonances. In order to understand the
polarimetric behavior of quasi-periodic soils, a randomly perturbed periodic
surface model has been developed, based on Kirchhoff's method with a
scalar approximation, that polarimetric scattering coefficients for various
dielectric and geometrical properties.
The joint us of the scattering model and the T-F approach mentioned above
highlight the relevance of the polarimetric parameter alpha1 angle for
estimating soil moisture even in case of Bragg resonance. Values of this
parameter are simulated and measured for various locations within the T-F plane
and demonstrate the robustness of this indicator to the presence of periodic