Inferring the effect of plant and soil variables on C- and L-band SAR backscatter over agricultural fields, based on model analysis

Abstract

The goal of this study was to extract from dual-frequency satellite SAR signatures consistent information about moisture in soils and about various features of plants for analyzing crop growth conditions in any agricultural region. The study was carried out on Polish agricultural regions but it is hoped that it will be applicable anywhere on the planet. During a satellite overpass on a particular date, the ground-based measurements required such as soil moisture (SM), Leaf Area Index (LAI), and biomass were collected from 10 to14 May 1998. The backscattering coefficients at various frequencies were collected from ERS-2.SAR (C-VV) on May 10, 1998 and from JERS-SAR (L-HH) on May 14, 1998. The applicability of three different vegetation descriptors to the semi-empirical water-cloud model was investigated. The contribution to the backscatter values of vegetation features such as leaf area expressed in the Leaf Area Index and the dielectric properties of leaf surface expressed in the Leaf Water Area Index (LWAI) and the Vegetation Water Mass (VWM) was examined in order to reveal the best fit of the model. It was found that in C-band, which had an incidence angle of 23°, the soil moisture contribution to the sigma value was predominant over the vegetation contribution. When the canopy cover increases, the sensitivity of a radar signal to dry soil conditions ( SM < 0.1) decreased. The sigma value was the most sensitive to vegetation descriptor VWM which described the amount of water in vegetation. Attenuation of soil signal by the canopy was found in all three vegetation descriptors types; the strongest attenuation effect was observed in the case of VWM. In L-band (where the incidence angle was 35°), the dominant signal to total σ o value comes from volume scattering of vegetation for LAI > 3. When LAI < 3 the vegetation contribution to total σ o value appeared in two-way attenuation. The results gave us the possibility of comparing the modeled with the measured soil and vegetation parameters.