Search

With the large amount of free Synthetic Aperture Radar (SAR) datasets that have been made available in recent years, wide-scale mapping of land surface displacements with differential SAR interferometry (DInSAR) has become increasingly popular. It can provide a more comprehensive picture of the spatial distribution of vertical surface motion of an entire country and help identify areas of deformation on local and larger spatial scales. The key aim of this thesis is to assess the application of DInSAR for detection and wide-scale monitoring of vertical land motion in Scotland. This is the first attempt to produce a spatially continuous deformation map for the country, providing a nationwide means of investigating land motion due to different causes. The different vertical surface deformation phenomena that can be expected in Scotland are all rather subtle, with slow and low-magnitude motion characteristics. On the one hand, this includes signals from ‘soft’ land surface covers, such as the extensive peatlands that comprise more than 20% of the Scottish landmass. On the other hand, ‘hard’ surfaces, such as urban or exposed rock areas can show motions due to, for example, mining subsidence, neo-tectonic fault displacements or large-scale crustal motion associated with glacio-isostatic uplift. A limited precision of the explored SAR datasets and the presence of residual orbital and atmospheric artefacts, make it difficult to discern actual ground deformation signals from noise. Therefore, this thesis is of exploratory character to assess what can and cannot be detected. This study uses archived L-band ALOS PALSAR and C-band Envisat ASAR data to assess the possibilities of wide-scale mapping of vertical land motion in Scotland. Seven ALOS satellite tracks with 8-12 SAR epochs approximately between 11/12/2006 and 07/03/2011, as well as two Envisat tracks with 15 dates each between 05/11/2002 and 03/04/2007, spanning the country from north to south, were processed to form multi-temporal interferogram stacks. Since most of Scotland is of a rural nature, the Small Baseline InSAR technique (SBAS) is chosen to invert the interferograms into displacement time series and mean velocity maps. Before time series analysis is performed, a general assessment of the applicability of DInSAR in Scotland is made. Geometrical limitations as well as the quality of the SAR data in terms of interferometric coherence in the two frequency bands are evaluated. Foreshortening, layover and shadow do not pose any significant restrictions to large-scale DInSAR mapping across Scotland, especially in the Highlands. A detailed coherence study for L-band and C-band emphasizes the differences in mapping capabilities of both wavelengths. For ALOS, a Scotland-wide good coherence well above 0.2 can be achieved, which delivers a high density of ground measurement points. Even the wide-spread grassy surface covers in Scotland, such as dwarf shrub heath and acid grassland, lead to relatively high interferometric coherence over longer temporal baselines of several months. With the applied Envisat data, C-band is only coherent on surfaces with stable properties, such as urban areas in the Midland Valley and rock platforms in the north of Scotland. Major limiting effects for SAR interferometry are image artefacts that are essentially a by-product of the SAR image acquisition process and subsequent interferometric procedure. Orbital inaccuracies in the repeat-pass geometry, as well as temporally and spatially changing atmospheric conditions, are both well-known and frequently-researched DInSAR problems. Different processing strategies are explored in this study in an attempt to reduce these artefacts and extract the ground deformation signals. For the correction of residual orbital effects, interferogram-based network techniques are tested and an extension proposed that enhances the quality of the corrected interferograms. This extended version introduces additional constraints to the estimation of the orbital planes by employing phase loop triplets. Atmospheric issues are addressed in a twofold manner. Tropospheric artefacts are corrected with High Resolution ECMWF (European Centre for Medium-Range Weather Forecasts) Zenith Total Delay maps, derived from Newcastle University’s GACOS (Generic Atmospheric Correction Online Service for InSAR) system. The ALOS data over Scotland, but also to a smaller extent the Envisat data, are significantly affected by ionospheric distortions. They are addressed with a statistical correction technique that was devised for this study and applies Principal Component Analysis to subsets of interferograms that share a mutual acquisition date. Five different processing strategies (A-E) are tested, each consisting of a different combination of these various image artefact correction methods. The optimal one is chosen for processing the entire SAR datasets over Scotland. The extensions to the different orbital and atmospheric correction techniques, that were devised here, are implemented into the SBAS time series processing chain within the in-house ‘InSAR TS+AEM’ software package, allowing them to be potentially used by other users in the InSAR community. In the final deformation results, the most obvious surface displacement signal that can be discerned with the evaluated ALOS and Envisat datasets is subsidence (in radar line-of-sight, LOS). This signal is consistent across extensive peatland areas. The detection of surface height changes in Scotland’s widespread peatlands has implications for a more comprehensive assessment of possible peat shrinkage and/or erosion and carbon losses. Within the Midland Valley, peatlands showed average subsidence trends of ca. 2.2 and 3.7 mm/yr for two ALOS scenes. Some time series extracted from peatland sites in that area indicated peat subsidence of 8 to 9 mm/yr. While in the Central Belt of Scotland these peat surfaces are spatially confined, they can make up much larger and continuous areas elsewhere, as in Sutherland and Caithness. Here again, the ALOS DInSAR measurements mostly indicate subsidence in LOS in the low mm-range. With the present datasets, it is not possible at this stage to achieve a precision that is good enough to detect underlying long-wavelength crustal deformation signals on a nationwide scale across the range of Scotland. On a more local scale, however, uplift signals in LOS are evident in the Envisat and ALOS datasets, though without more detailed (in situ) investigations, it is not possible to attribute this uplift definitively to a geophysical process. If it is associated with hard surface covers, such as exposed rock in the Northwest Highlands or man-made structures in the vicinity of some GPS stations, this upwards vertical motion might indicate a glacial rebound signal, if other motion factors can be neglected. On these surfaces, Envisat and ALOS data showed uplift rates in LOS of ca. 1-3.5 mm/yr. The precision of the derived deformation products is limited, i.e. areas that are seemingly reliable are intermixed with areas that contain too abundant remaining noise and artefacts. However, an uncertainty analysis and validation of the DInSAR data with independent adjacent and overlapping satellite tracks, as well as external continuous GPS data across Scotland is promising. Deformation trends between sensors and between independent tracks of the same sensor are mostly consistent and relatively low root mean squared errors between DInSAR deformation measurements and the assumed linear deformation model could be achieved.