MSc Dissertation: Gavin Doyle


Adobe-PDF-downloadDoyle, Gavin. Three Applications of Satellite Borne Repeat Pass SAR Interferometry in Southern Africa. MSc Dissertation. Department of Electrical Engineering, University of Cape Town, 2000.



Synthetic Aperture Radar Interferometry (InSAR) is a relatively new remote sensing technique, which can be used to derive precise topographic height and height change informaton over large areas. It is a technique which has been used in a great variety of situations, from the topographic mapping of Venus to the detection of subtle ground deformations due to earthquakes and mining subsidence.

The InSAR technique involves using the phase information inherent in radar images to extract elevation ad elevation change information. The process requires very careful co-registration of a pair of complex images of the same scene, followed by the multiplication of the one image by the complex conjugate of the other. In this manner, a phase difference image, or interferogram is created, in which subtle differences in the ditances from the two imaging sensors to the ground are mapped, thereby making it possible to generate a height model of the ground surface.

An extension to the InSAR technique is that of differential InSAR (dInSAR), whereby it is possible to detect changes in elevation over time in the order of a fraction of a wavelength of the radar signal. The signal wavelength of the European ERS sensors, data from which was used extensively in this project, is 56 mm. Any change in the distance from the satellite to the ground of half this distance will cause a phase change of one whole cycle (because of the two-way path of the signal). Hence, each fringe cycle in a differential ERS interferogram represents 28 mm of range change.

The Radar Remote Sensing Group at the University of Cape Town has developed the first viable interferometric processing capability in Southern Africa. An important part of developing this capacity has been the testing of the InSAR technique in real situations. This thesis presents a look at some applications of InSAR and dInSAR to real geophysical and geodetic issues in Southern Africa. It is neither a pure geological nor a pure electrical engineering approach to the subject, but it does endeavour to describe the application of this technology to geophysical and geodetic issues. Three distinct applications are presented, in which the procedures, strengths and limitations of the technique are discussed. The three applications are:

1.   The development of a digital elevation model (DEM) of the Western Cape region, using a pair of ERS SAR images.

This project involved applying the InSAR technique to a pair of ERS images to derive a height map of the South Western Cape.

The two images, which were selected from a possible set of five, were selected on the basis of their acquisition times being close together, and their viewing geometry being appropriate for the task. The data selection process for interferometry is critical, as the technique is susceptible to phase decorrelation, which results from a combination of statistically significant changes in the Earth’s surface characteristics between image acquisition, and differences in the sensor flight paths.

The images were processed from raw data, and an interferogram was generated. The interferogram, which is a representation of the range difference between the ground surface and the sensors, but with modulo 2ð ambiguities, was filtered for noise, and then corrected for the side looking geometry of the SAR sensors. This side looking geometry causes the topographic differences to be superimposed on the differences in range from the near side of the image to the far side. The average component due to the side looking geometry is removed, leaving only that due to topography (and atmospheric effects). The resulting “flattened” interferogram was phase unwrapped where possible, to eliminate the modulo 2ð ambiguities. The unwrapped phase image was then used to calculate actual topographic heights, and the resulting height map was then re-sampled and georeferenced so that points on the map could be identified with their corresponding points on the ground.

The phase unwrapping process was incomplete in certain areas of the scene. This was because of information loss due to phase decorrelation in areas subject to radar shadow, layover, and the presence of water bodies. This resulted in there being numerous data gaps in the generated height map. Some of the specific regions affected in this way include all of the dams and reservoirs in the image, most particularly the Theewaterskloof and Steenbras dams. Regions affected by radar shadow tended also to be affected by layover, sometimes eliminating peaks in their entirety from the DEM. The Sentinal, Chapmans Peak, and Hangklip are clear cases of where this happened. Decorrelation due to the effects of wind on vegetation is evident on the slopes of most of the mountains, particularly Devil’s Peak, Signal Hill, and the forested areas around Jonkershoek and the Helderberg.

A comparison made between the DEM and independently (optically) derived height information, showed a number of low spatial frequency anomalies in the InSAR derived DEM. The presence of these anomalies has been attributed to refractive variations within the atmosphere at the time of image acquisition. These sorts of aberrations are only prevalent in repeat pass interferometry, where the two images are acquired at different times. It is possible with a dual antenna system, where the antennas are well spaced, to acquire the two images simultaneously, thus eliminating the effects of atmospheric perturbation.

This project was initially intended as a test of the Gamma software suite, but it has become an important demonstration of the effects of atmospheric refraction on the interferometric technique. Work is currently under way to develop techniques for combining multiple data sets, so as to minimise such errors, and to fill the gaps resulting from incomplete phase unwrapping.

2.   Katse basin deformation mapping by satellite radar interferometry.

This project involved the use of differential InSAR to map the deformation field associated with the 2 billion tonne load imposed on the Earth’s crust by the Katse reservoir in Lesotho. The results were to be used to constrain geophysical models of the Earth’s crust in the region, and in so doing, assist reservoir designers in assessing the seismic risk associated with the Lesotho Highlands Water Project.

Predictions, based on an elastic half-space model, indicated that the added load of the water in the Katse reservoir would cause the Earth’s crust to be depressed by more than 10 cm. This amount of ground movement is well within the limits of detection by differential InSAR, and the technique had the potential to provide a high resolution map of the deformation and any structural inhomogeneities associated with the loading.

ERS SAR image data of the area was obtained from before, during and after the filling of the Katse reservoir. Where geometric constraints allowed, interferograms were generated, and an attempt was made to produce an InSAR DEM for the purposes of removing the topographic effect from interferograms that should have included the effects of deformation. The alpine type terrain in Lesotho causes significant loss of information in ERS SAR images, mainly through layover and shadow effects. It was therefore not possible to generate such a DEM, and use had to be made of a DEM which was produced by combining an existing, low resolution DEM with one derived by radargrammetrically processing a pair of overlapping images. The viewing geometry used was however inappropriate, and the stereo reduction of a pair of ERS images was only partly successful.

A single interferogram, which included the effects of reservoir loading, was produced from a pair of images acquired with a three year time separation. The fact that it was possible to produce this interferogram represented an indication of the viability of SAR interferometry in areas where the ground surface does not change significantly even over long time periods. The spatial separation of the sensors for this interferogram was small (20 m), and any deformation effects should have been immediately visible, but no such differential effects were apparent. Subsequent to the processing of the interferograms, information obtained from conventional geodetic monitoring of the dam site indicated that the deformation had in fact been almost an order of magnitude less than that anticipated through the elastic half-space modeling. This put the deformation outside the limit of detectability by differential InSAR.

Although no ground deflection was detectable by interferometry, the technique worked far better than expected, given the terrain and time separation between images. This bodes well for the use of this technique in similarly non-ideal conditions.

3.   The mapping of ground surface deformations associated with a severe rock-burst in the Welkom gold fields in April 1999.

This project made use of the same techniques as those used in the Katse project, but in much more favourable conditions. The project was concerned with the mapping, by differential InSAR, of the surface deformations caused by an earthquake in the Welkom gold fields in April 1999.

The earthquake caused two fatalities and considerable damage to the Eland shaft of the Matjhabeng mine. The magnitude of the event was measured by seismographs in the area as being M4.5 (local), and was determined to have been induced by the mining activities themselves. Seismicity induced through rockburst is a problem which is becoming more prevalent as mining takes place at deeper and deeper levels. Consequently, it is currently the subject of active research by geophysicists and mining companies. The differential InSAR technique had been used with spectacular success in other parts of the world in mapping deformations due to, among other things, earthquakes and mining induced subsidence. Through this project, it has also been demonstrated to offer valuable input for research into rock failure dynamics in the deep gold mines.

Using combinations of ERS radar images acquired before and after the April 1999 event, a differential interferogram was produced, in which the ground deformation that resulted from the earthquake is very clearly quantified. The result coincides very closely with the location of the fault plane where the movement took place, and also with the mine shaft where the worst of the damage was reported.

Leave a Reply