Inversion techniques in radar remote sensing of agricultural field : case studies on sugar beet and winter wheat

This thesis is an attempt to gain insight in the retrieval of the soil moisture content and the vegetation water content from the radar backscatter of agricultural fields. Two crops have been selected: sugar beet and winter wheat. For a retrieval of the two agricultural parameters two approaches have been adopted.The first approach is based on an inversion of existing models, which describe the electromagnetic interaction with a vegetated surface on basis of the radiative transfer theory. A solution to this inversion problem is suggested involving the inversion of the simple Cloud model, which is connected to the complex MIMICS model.In the second approach a semi-empirical algorithm is developed, which decomposes the covariance matrix of polarimetric radar data into a vegetation, soil, and vegetation-soil covariance matrix. It is shown how centrical symmetry and the Brewster angle effect can be used in this decomposition technique. The procedure for an assessment of the three covariance matrices has resulted in two solutions for the algorithm. A choice between these solutions can be made by means of three polarimetric tools. A sensitivity analysis reveals that decomposition results at C- and L-band are not always sensitive enough to changes in the agricultural parameters.The validity of the two radiative transfer models is examined using backscatter measurements and accompanying ground data. Because of the complex structure of winter wheat the Cloud model has been applied to sugar beet data only. The backscatter of sugar beet and winter wheat at C- and L-band is predicted with MIMICS.The performance of the Cloud model could not be tested properly because of two reasons. Firstly the temporal resolution in most of the data sets is too low and secondly a systematic decrease occurred in practically all backscatter data, which complicates the use of the Cloud model.The inversion scheme with MIMICS could not be applied to the examined data, because there are disagreements between model predictions and measurements. In particular the predicted extinctions with MIMICS are too low compared with the experimental observations. Polarimetric backscatter is investigated with the three tools. Sugar beet backscatter at C- and L-band exhibits centrical symmetry. A significant Polarimetric Phase Difference ( <em>PPD</em> ) is found in the backscatter of winter wheat at L-band. This <em>PPD</em> can be explained by means of simulations with a radiative transfer model.The semi-empirical algorithm has been applied to sugar beet and winter wheat polarimetric backscatter at C- and L-band. The decomposition results so obtained are then considered in relation to trends in the agricultural parameters. The analysis yields promising results since the predicted trends are in agreement with the few available ground data. It appears that the calibration accuracy of the measurements is a limiting factor for the sensitivity of the results for the two agricultural parameters, and for soil moisture content in particular. Of the two frequencies considered the L-band gives the best information.The available radar backscatter data could be applied in a vegetation study because of the synergy with optical data. In spite of the large amount of investigated data, there is still a need for detailed experiments, e.g. tower-based measurements. With such experiments the validity of backscatter models can be further evaluated and a better interpretation of phenomena occurring in the polarimetric backscatter becomes possible.