Multi-Frequency Polarimetric SAR Tomography for the 3-D Characterization and Monitoring of Agricultural Crops

Abstract

Monitoring the temporal variation of soil and plant parameters of agricultural crops is of high interest. Since Synthetic Aperture Radar (SAR) measurements are sensitive to dielectric and geometric properties of the observed scattering scenario, they provide key observables for monitoring the temporal variation of biophysical parameters. However, the scattering mechanisms occurring in agricultural vegetation in dependency of biophysical parameters are highly complex and simultaneous dynamics of the soil and the vegetation are difficult to differentiate. By utilizing several horizontally separated SAR acquisitions, SAR Tomography, as demonstrated for forest volumes, is a powerful tool able to estimate vertical profiles of the backscattered power and to resolve and interpret scattering mechanisms along height. The fact that tomographic SAR techniques are, in principle, independent of scattering models makes their application very promising towards a better understanding of the highly dimensional scattering scenarios of agricultural vegetation. Challenges are however the high vertical resolution required in order to be sensitive to the low plant heights, and the possibly present anisotropic propagation effects of the vegetation volume limiting the application of state-of-the art tomographic ground and volume separation algorithms. The Crop Experiment (CROPEX) campaign in 2014, which plays a key role in this PhD thesis, fills the gap in the availability of fully polarimetric multi-frequency and tomographic SAR data over agricultural crops providing high vertical resolution capability and covering different dates of the phenological cycle. The main objective of the campaign is to foster the physical understanding of the influence that changes in soil and plant parameters have on the ground and volume scattering component as a function of crop type, polarization and frequency. The interpretation of changes on the ground and in the volume from vertical backscatter profiles is limited and can only be quantified by separating the ground and volume scattering components. Without posing model assumptions to the vegetation volume, this separation becomes ill-posed. In the thesis, this is addressed by applying a separation algorithm which overcomes this ambiguity by integrating the a priori knowledge of the ground height as a given parameter and is able to provide robust estimates of the multi-baseline volume coherences and the ground and volume powers. The biggest novelty of this work is the quantitative analysis of the distinct ground and volume powers for agricultural scattering scenarios, particularly taking into account their temporal variation as a function of varying soil and plant parameters. The temporal variation of the estimated powers provides an unambiguous quantification of scattering changes on the ground and in the vegetation. Since the center of mass of the vertical backscatter profiles correspond - at least at the first order - to the interferometric phase center that can be estimated by means of only two acquired tracks, its analysis gives an understanding of the potential of a reduced observation scenario. In conjunction with in-situ measured soil and plant parameters, the sensitivity of the tomographic parameters across different frequencies, X-, C- and L-band, to dielectric (e.g. dynamics of the water content) and geometric (e.g. alignments of plant components) changes is demonstrated. The experimental results in this thesis underline the importance of the three-dimensional information. With the separation in height, the conclusions drawn on the changes of the incoherent scattering signature provide a basis for future research including the coherent ground and volume signatures. Scattering models based on the coherent polarimetric signature might provide promising opportunities towards the inversion of biophysical parameters.