Improving Persistent Scatterer Interferometry Results for Deformation Monitoring: Case study on the Gardanne mining site

The persistent scatterer interferometry (PSI) is a relatively new technique in radar interferometry which has matured, within a last decade, to a practical technique for measuring the deformation of the Earth’s surface. PSI overcomes the main limitations of conventional InSAR methods by identifying radar targets, called persistent scatterers (PS), which have stable backscattering characteristics in time. In 2003, ESA (European space agency) initiated the PSI Codes Cross-comparison and Certification (PSIC4) project to produce reliable information about accuracy and validity of PSI methodologies. The results of the experiment clearly indicated that different PSI approaches resulted in different number of detected PS (or PS density). Especially in deforming areas with strong deformations, the experience shows the low PS density. This causes that main areas of subsidence could not be assessed and identified. This lack of PS in deforming areas raises the hypothesis of type-I errors (falsely rejected PS) due to imperfections in the mathematical model of PSI processing. The background of this thesis is formed by the idea of using the results of the PSIC4 study to evaluate algorithm performance and assess which approaches are most suitable to retrieve reliable deformation parameters at a high spatial density. In this study, the contribution of four different sources of type-I errors is investigated, and some approaches to improve the PS density are proposed. The Gardanne mining area (the test site for PSIC4) was chosen as the case study for validation of the proposed approaches. The following optimizations are addressed in this study: 1) Initial network optimization 2) Optimized atmospheric phase screen (APS) estimation 3) Azimuthal subpixel position estimation 4) Non-linear deformation modeling In addition to these optimizations, the new approach for the final selection of PS is presented. Spatio-temporal consistency (STC) is introduced as a quality assessment for the final selection of PS. The advantage of using STC is that it can reduce the dependency of the PS quality assessment to the assumed deformation model and therefore it is better indicator for the observations precision. Integration of all proposed optimizations together results in 236% improvement in the number of detected PS in the Gardanne deforming area, leading to the improved identification of the Gardanne deformation field. Results of the study show that application of the proposed optimizations can improve the PS density. However, detected PS in areas with fast deformation mechanisms suffer from the unreliable estimation of ambiguities (unwrapping errors), resulting in unreliable estimates of deformation time series. The main reason for this localized effect (as it occurs mostly in deforming areas) is the fast evolution of deformation mechanisms. For example, in the case study of this thesis, the deformations range from few centimeters up to some decimeters occurred in few months. These deformations are strong ”fast deformation” for the view point of C-band radar (with wavelength of 5:6 cm) with at most monthly acquisition of ERS data. Different methods with restricted or relaxed conditions for removing unwrapping errors are presented. The decision for removing unwrapping errors from final selected PS or relaxing the acceptance condition of unwrapping errors is dependent to the application of PSI results. However, in order to deliver the PSI results including unwrapping errors to end users, a PS quality description which contains the quality of estimated ambiguities is required.

Subject

persistent scatterer
radar interferometry
remote sensing
deformation monitoring

To reference this document use:

http://resolver.tudelft.nl/uuid:d1eed85e-6625-434d-aacb-538464557f78

Part of collection

Student theses

Document type

master thesis

Rights

Files

PDF

SSEsfahany_Thesis-3.pdf

54.89 MB

Close viewer