Radar backscattering over agricultural bare soils

Abstract

Agricultural surfaces are complex non-isotropic surfaces presenting a main direction due to the rows in combination with various smaller spatial scales (the clods). The applicability of two recent electromagnetic models, able to take account of these anisotropic surfaces, was tested under a variety of roughness encountered on an agricultural watershed (ploughing and sowing) by comparison with radar measurements made over fields with known surface parameters (soil moisture and soil roughness). The data set was collected during the "Orgeval 1989" campaign in France. Radar copolarized signatures were provided from the French helicopter-borne scatterometer ERASME in C and X bands. The Shin and Kong's model (1984) is in the frame of the Kirchoff approximation. The Integral Equation Model (Fung et al. 1992) overlaps the gap between Physical Optics and Small Perturbation models where are located in C band most of the Orgeval sowing fields, but it fails to be applicable in X band, except for ploughed field. For sowing fields the Integral Equation Model was found to yield good agreement with the backscattering measurements of this study. No discrimination of the various types of sowings was found by the copolarized radar signatures in C band due to an almost flat response of cross section with incidence angle (-5dB from 10 to 40 degrees). The direction of the soil practice is no more a key parameter. The high inner-field dispersion of the radar backscattering coefficients (3 to 5dB) may be predicted by this theoretical model by using the natural variabilities, especially at a small scale, of the two dimensional correlation function, rebuilt from the measured soil profiles parallel and perpendicular to the rows following the surface formulation of Shin and Kong. For the ploughed field, the great dispersion of radar measurements in both C and X bands (up to 10dB) shows the limitation of a comparison between measurements and simulations. The experimental inner-field dispersion could not be reproduced by IEM with the measured correlation functions (only 4dB). Further the Gaussian adjusted ones introduced in Shin and Kong's modelisation induce too extreme responses of cross sections with incidence angle which are not realistic for a surface scattering mechanism.