Polarimetric indices for crop monitoring based on model simulations and satellite observations
Croix du Sud 2/16,
A polarimetric radiative transfer model has been developed at UCL for the simulation of polarimetric radar observation of rough soils and vegetated areas. The so called Polscat model is a fully polarimetric model and allows to compute the complete Mueller matrix from which the backscattering coefficients in various polarizationís can be obtained. For the soil, the Integral Equation Method (IEM) has been implemented. This surface scattering model requires as input the dielectric properties of the soil and the geometric properties of the surface (i.e. rms height and slopes). For the vegetation, the volume electromagnetic scattering model is based on the radiative transfer theory. It includes size distributions, it accounts for the reciprocity effect, it allows vertical distributions within the canopy and provides a precise description of the vegetation component. The soil parameters measurements are widely discussed in the literature. The rms height and the correlation length are commonly measured using linear profilometer. In our study, the 3-D geometry of the soil surface is determined using stereoscopic pairs of photographs acquired at 4- meter high over 2 crop types (i.e. sugar beet and spring barley) along the first stage of the growing period (when the soil is sparsely covered by the vegetation). From the modeled soil surfaces, 2-D profiles are generated in various directions (relative to the crop row orientation), positions and using different sampling distances. The roughness parameters are deduced from this different configurations and the corresponding co-polarized (HH and VV) signals are modeled. The effects of the sampling distance, the profile length and the roughness anisotropy on the output simulated signal are systematically investigated. In a second part, the partial validation of the surface scattering model is presented. This was made using simultaneous SAR images acquisitions (ERS, RADARSAT and JERS) and fields measurements from the 1995 to the 2000 growing seasons over sugar beet and cereal fields. Simulations were made for each configuration of mentioned sensors and compared to the recorded co-polarized signal. Based on the simulations in C-band, polarimetric indices are proposed to separate soil and vegetation effects from the backscattered signal. Temporal signal variation and the anisotripic effect of row orientation thanks to the satellite look direction are also presented.