Multi-temporal interferometry (MTI) techniques for ground deformation mapping using Sentinel-1 SAR data
MetadataShow full item record
The multi-temporal interferometry (MTI) techniques, such as persistent scatterer interferometric synthetic aperture radar (PS-InSAR) and small baseline (SBAS) have proved to be powerful geodetic tools to measure deformations of the Earth’s surface in space and time using a stack of SAR images. The techniques have been widely used by the scientific community to measure the displacement velocity and time-series related to subsidence/uplift, landslide, tectonic, and volcanoes. In this dissertation, one successful example of applying the MTI technique using multi-mission, multi-frequency, and multi-track SAR datasets for detecting surface subsidence due to groundwater overexploitation is presented. The InSAR data then is used for hydrogeological interpretations to analyze the reason of subsidence. Although the techniques can potentially provide the information at high level of accuracy (sub-cm), their performance is largely affected by the density of the measurement points, i.e, persistent scatterers (PS) and distributed scatterer (DS) pixels. Traditionally, the PS-InSAR method has been formulated and applied to single-polarimetric SAR data probably because of the limitations of available multi-polarimetric SAR images. The recent launch of the Sentinel-1 mission with capability to obtain acquisitions on dual-polarized (HH+HV, VV+VH) channels helps to increase the spatial density of PS points through the polarimetric optimization method. The method performs a search over the available polarimetric space in order to find a linear combination of polarization states, which yields the optimum PS selection criteria. This thesis evaluates the performance of the method using dual-polarized (VV+VH) Sentinel-1 images. To increase the density of the DS pixels, SqueeSAR applies a spatially adaptive filter on images and optimizes the DS pixels by the maximum-likelihood technique. Although it improves the final results, it is a time consuming process for large stacks of SAR data. Moreover, the whole procedure of DS selection should be repeated as soon as a new SAR acquisition is made, which is challenging considering the short repeat-observation of missions such as Sentinel-1. This thesis follows SqueeSAR and instead of using Kolmogorov-Smirnov (KS) test for DS analysis, implements a new approach using two-sample t-test to more efficiently identify neighboring pixels with similar behaviour. Another advantage of Sentinel-1 mission is providing wide spatial coverage (250 km), which opened new perspectives for large-scale InSAR analysis. However, the spatiotemporal changes in troposphere limits the accuracy of these measurements for operational monitoring at a wide scale. Due to the coarse node spacing of the tropospheric models, like ERA-Interim and other external data like the Global Navigation Satellite System (GNSS), the interpolation techniques are not able to well replicate the localized and turbulent tropospheric effects. This thesis, proposes a new technique based on machine learning (ML) regression approach using the combination of small-baseline interferograms and GNSS derived zenith total delay (ZTD) values to mitigate phase delay caused by troposphere.