Development of Calibration Methods For A Ku-Band Polarimetric Terrestrial Radar Interferometer

Imaging radar systems are a fundamental tool for remote sensing: by using the sensitivity of electro- magnetic waves to the geometrical and dielectric properties of objects they interact with, these de- vices permit to observe many natural phenomena. These devices have been exploited for a range of applications in the geosciences; for example to monitor glacier flows, the inflation of volcanoes, to de- tect land subsidence or to estimate environmental parameters. Spaceborne imaging radars are used in the reconstruction of post-seismic displacements, for the observation of land subsidence, in mar- itime surveillance, for the estimation of vegetation height and growth stage, the production of digital elevation models and many other applications. In most of the above cases, spaceborne —or to a lesser extent airborne— synthetic aperture radar (SAR) sensors are employed to acquire the data. These systems acquire images with excellent spatial resolution and coverage; their main drawback —particularly in the case of spaceborne SAR— is the inflexible acquisition schedule, which does not allow to image the same region more often than every few days at best. However, the study of many natural processes requires a denser temporal sampling to capture their dynamics. Terrestrial radars —also known as ground based radars or ground based SAR— are complementary to space- and airborne SAR systems for the observation of fast changes of limited spatial extent since they permit greater flexibility in the acquisition mode and scheduling. This dissertation concentrates on KAPRI a Ku-Band polarimetric terrestrial radar interferometer based on the Gamma Portable Radar In- terferometer II (GPRI-II), which was modified by the manufacturer to add polarimetric imaging capabilities. The thesis focuses on two aspects: on the one hand the data processing and calibration techniques needed to obtain radar data free from distortion caused either by the device’s design and environmental effects; on the other hand the applications and the suitability of such data for the observation of changes in the natural environment. These two aspects are elaborated and assessed with three publications that are the main contributions of this research work; these papers focus respectively on data preprocess- ing and polarimetric calibration, phase calibration for differential interferometry and the analysis of polarimetric signatures of natural terrain. The first investigation deals with the data processing and calibration needed to convert the raw data acquired by the radar into properly focused polarimetric radar images. Preprocessing methods to mitigate the geometric distortion caused by the antenna frequency response inducing beam squint and cross-range phase ramp induced by the polar acquisition geometry were developed, followed by a polarimetric calibration model adapted to the system’s antenna design. These methods were validated by studying the response of corner reflectors, whose radar reflectivity and polarimetric signature is well- known. The results show that applying the proposed methods permits to obtain radar images with an imaging resolution close to the nominal resolution, free of cross-range phase ramps and with minimal residual polarimetric distortion. The second investigation is dedicated to the statistical modeling and correction of the atmospheric phase screen (APS), a significant obstacle to precise deformation monitoring using radar interferometry: by exploiting the sensitivity of the radar signal’s phase to variations in the distance between sensor and targets, this technique allows to measure displacements with a precision potentially limited only by the wavelength and by various sources of phase noise. However, other effects also modify the radar phase; among them variations in the distribution of atmospheric water vapor in the scene, which can be of a magnitude comparable to the displacement signal and may severely reduce the precision of the displacement estimates. The spatio-temporal statistical behavior of the APS and its mitigation has been extensively studied: the main assumption being that the APS is correlated in space and uncorrelated in time for the typical acquisition scenarios. However, there is lack of research on whether these results can translate to terrestrial radar interferometry (TRI), where the spatial extension of the scene is typically smaller and the acquisition repeat time can be as short as a minute. In this study, these questions are assessed with the help of an interferometric stack acquired using KAPRI near the Bisgletscher glacier, in the Swiss Alps. The analysis shows that the APS is correlated both in space and time but that its spatiotemporal statistics are approximately separable, so that given a stack of interferograms a fraction of the phase variation can be corrected on individual interferograms by spatial interpolation from a set of locations known not to be affected by deformation, while the remaining residual phase variation is only correlated in time and can be corrected by a temporal filter operating on individual pixels along the stack reducing the computational complexity of the correction. The third and last investigation returns to the polarimetric aspects of KAPRI and aims to assess the suitability of Ku-Band polarimetric data for the observation of natural terrain: while the polarimetric response of natural surfaces at L- and X- band are widely investigated thanks to the availability of space- and airborne sensors operating at these wavelengths, much less information is available at Ku- Band. Two datasets covering alpine terrain and an agricultural areas are used to this end by studying the polarimetric signatures of different terrain types, showing a consistently high cross-polarized power and Cloude-Pottier entropy for all land cover types but for individual highly reflective scatterers such as radar reflectors or single buildings. These observations suggest a high level of scattering randomness, likely due to the presence of depolarizing scattering from random media and to the combination of a short wavelength and a relatively large resolution cell size