Sensitivity Analysis of Model-based Pol-TomoSAR Decomposition for Ground-Volume Separation and Soil Moisture Retrieval over Agricultural Areas

Abstract

One method for estimating soil moisture is through the use of radar systems, such as Synthetic Aperture Radar (SAR), due to their high spatial resolution. However, this presents a significant challenge when the soil is situated beneath a vegetative canopy, as the vegetation exerts a considerable influence on the backscattered signal. In order to distinguish between vegetation and the ground, this thesis concentrates on the utilisation of interferometric multi-baseline airborne F-SAR data for the acquisition of vertical information and polarimetric airborne F-SAR data for the estimation of soil moisture. To this end, the problem is initially divided into two sub-problems. The initial issue concerns the separation of the ground and vegetation layers. This entails defining the vertical reflectivity profile of the ground layer and the polarimetric volume signature of the vegetation, as well as calculating the corresponding polarimetry matrix for the ground. This is achieved through optimisation, which enables the imposition of a high degree of variability in additional constraints. The second sub-problem concerns the estimation of soil moisture, employing physical models. This study demonstrates that even when ground and volume separation is acceptable, it is not possible to uniquely calculate soil moisture using physical models. Consequently, we demonstrate that by integrating the physical models directly into the layer separation, the solution aligns with the models but is not unique. By constraining the powers of the components, we are able to estimate soil moisture with sufficient precision. The findings indicate that it is feasible to estimate soil moisture with the imposed constraints, while also highlighting the importance of accurately defining the corresponding power constraints.