Detection of surface surface deformation in cities using Persistent Scatterer Interferometry (2012)
The mesurement of surface deformation as a consequence of natural or man-made changes of the environment is a typical task in surveying. Traditional techniques used for this purpose are leveling, tachymetric surveys and since the last two decades GPS-measurements. All these methods provide high accuracy measurements. However they are limited in a way, that their measurements are confined to a relatively small number of monitoring points. If the areal characteristic of a deformation process has to be determined, a huge amount of monitoring points is needed. As a consequence the costs increase and the whole project may get uneconomical.
One alternative for measuring the areal characteristics of a deformation phenomenon is the differential radar interferometry. This method utilizes high resolution radar images obtained with the Synthetic Aperture Radar technique (SAR). A differential interferogram is then computed within a specific processing using these two complex valued SAR images. The phase of the obtained interferogram contains the surface deformation in the viewing direction of the radar sensor. Theoretically the deformation can be inferred from these phase with millimeter accuracy. However, in practice the deformation is masked by two disturbances rendering the estimation of the deformation with the aforementioned accuracy difficult:
Changes of the reflectivity within the scene between the two acquisition dates lead to a decorrelated signal. This means, that it is impossible to extract a deformation signal from the measured phase. This problem is inherent to vegetated terrain. Generally speaking the evaluable area of the interferogram decreases with increasing difference between the two acquisition dates. This effect is illustrated by an example in Fig. 1. The upper part shows the amplitudes of two complex SAR images. In the lower right the phase of an interferogram (color coded) calculated from these SAR images is displayed. It is apparent, that both amplitude images look very similar. The interferogram shows however no information, which is due to the decorrelation.
The SAR principle is based on the time of flight principle of emitted microwave pulses. The speed of this wave depends mainly on the refraction index of the medium of propagation. Since the water vapor content is the main factor influencing the refraction index for radiowaves in the troposphere, different weather conditions during the acquisition dates lead to significant change in the signals time of flight. This change causes also changes in the interferometric phase and can completely mask the deformation signal.
Fig. 1: Effect of temporal decorrelation using the example of the Staßfurt (south of Magdeburg ) scene. In the lower right part a differential interferogram is displayed. The temporal baseline of the SAR images used amounts to 630 days. The amplitudes of the complex SAR images used for interferogram generation are displayed in the upper part. In the lower left a Google Earth view of the scene is illustrated.
One approach for mitigating the aforementioned limitations of differential interferometry is the Persistent Scatterer Interferometry (PSI). PSI is based on two ideas. The first idea is to use a stack of differential interferograms. This enables the separation of atmosphere and surface movement exploiting their different spatio-temporal behavior. Typical stacks consisting of 20 up to 100 interferograms. The second idea is the restriction of the analysis to points showing a constant backscattering behavior over time. Such points are called Persistent Scatterer (PS). Under favorable conditions the estimation of the deformation velocity with an accuracy of one millimeter per year is possible. A PSI result for the Staßfurt scene is shown in Fig. 2. This result is in good agreement with the available reference data.
Fig. 2: PSI result for the Staßfurt scene. A stack of 43 ERS1/2 images has been used. The deformation in the viewing direction of the sensor is shown. The deformation evolving in the center of the city is clearly visible. The results are in good agreement with reference data; i.e. a mined subsurface space is located below the zone of deformation.
So far the data used for PSI have been acquired by the European radar satellites ERS1, ERS2 and ENVISAT. These satellites provide a ground resolution of approximately 20 meters. Recently data of new generation radar satellites has become available, which provides a much better ground resolution. One example is the German TerraSAR-X, which provides a resolution of one meter in the mode of highest resolution. This high resolution data opens new possibilities to the PSI technique, that are to be examined within this project. One effect is for instance the increase of the PS density leading to a improved accuracy of the deformation estimates. Furthermore a matching of objects, leading to PS, to building structures (balconies, windows, chimneys) is conceivable. This would enable a detailed description and modeling of the deformation of a specific building or parts of a building respectively.
Even, M.; Schunert, A.; Schulz, K.; Soergel, U. (2009): PSInSAR: detection of localized surface deformations with a modified StaMPS-algorithm: Proc. SPIE, Vol. 7477 (2009) more