Background to Satellite Radar Interferometry (InSAR)
InSAR is a technique in which the phase component of the returning radar signals of two or more synthetic aperture radar (SAR) scenes of the same location are processed to allow the detection of ground movements. The relative low cost and non-invasive application has led to them being used in numerous scientific disciplines. One such discipline - natural hazards, and in particular ground motion hazards, fall under two categories - natural and man-made, with a degree of crossover between the two. Since 1992 conventional 2-pass Differential InSAR (DifSAR) has been used by the remote sensing community to further the study and understanding of specific ground deformation hazards. These hazards can result from natural phenomena such as seismic events, volcanic deformation, landslides and ground collapse; and from more subtle processes such as subsidence, all of which have direct economic, environmental and human safety impacts. The need for more accurate and temporally dependent results has led to the inevitable technical advances seen over the last few years, including PSI.
Persistent Scatterer interferometry (PSI)
PSI is a non-invasive surveying technique used to calculate fine motions of individual ground and structure points over wide-areas covering urban and semi-urban environments. The technique uses an extensive archive of satellite radar data (dating back to 1992) to identify networks of persistently scattering (i.e. radar reflecting) features such as buildings and bridges, or natural features such as rocky outcrops, against which precise millimetric motion measurements are calculated retrospectively over the time spanned by the data archive. The unique benefit of PSI is its ability to provide both annual motion rates and multi-year motion histories for individual scatterer points.
The PSI technique take conventional InSAR a step further by correcting for atmospheric, orbital and DEM errors to derive very precise displacement and velocity measurements at specific points on the ground.
PSI uses a series of co-registered satellite synthetic aperture radar (SAR) images to identify time persistent radar scatterer points and to derive an atmospheric phase screen for each scene. These persistent scatterers are naturally occurring or pre-built features on the ground that reflect back to the satellite reliably and persistently over many years. They correspond to parts of buildings, lampposts, antennas, exposed rocks or other features that are persistent reliable radar reflectors. The exact location of the stable points cannot be predicted in advance of processing but over urban areas their densities exceed 200/km2 and may be as high as 600/km2.
PSI software analyse the phase responses of the point targets and are able to separate the ground displacement, height and atmospheric contributions of the phase for each point. The phase contribution due to atmosphere is estimated for each scene in the dataset and these contributions are removed the results. Derivation and accounting for the atmospheric effects in the processing produces much finer measurements than the conventional DifSAR technique. In addition, the precise height of the points can be determined thus removing the phase contribution from the measurements. As a result of these corrections, the levels of measurement precision achieved on the annual displacement rates are of the order +/- 0.1mm/year.
PSI maps wide-area relative ground movements with sub-millimetric accuracy along the satellite’s line-of-sight (LOS) and its precision in the vertical domain is beyond that achievable with GPS. The absolute spatial accuracy is about 15m, while the relative spatial accuracy is about +/- 5m in East-West and +/-2m in North-South direction. PSI therefore represents a cost and time effective measure of ground motion over large areas.
PSI analyses use linear displacement models (i.e. to detect ground motion that has occurred at a broadly constant rate over time) to extract points on which measurements can be made. This approach has the advantage of being able to economically measure displacements over medium sized regions (up to 2000km2) at a time. Some degree of non-linear displacement events can be detected since many of these can be reasonably modelled using a linear displacement rate. For each persistent scatterer, a motion history is available for the time span of the available data (from 1992 onwards).
- Data quantities
PSI requires a large number of SAR scenes and for some locations outside of Europe, not enough data may be available to apply this technique.
- Persistent Scatterer density and locations
A feature of the technique is that the number and location of persistent scatterers cannot be predicted before processing, and measurement success can only be guaranteed over built-up urban areas or over dry and rocky regions; and assuming that the data are suitable for interferometric processing. Furthermore, it should be noted that it is the movement of the persistent scatterer that is measured and not that of the ground.
- High displacement rates
InSAR phase measurements are recorded as wrapped phase in the range -? to +?. Phase unwrapping solves this ambiguity by calculating the correct number of phase cycles that need to be added to each wrapped phase measurement so that the correct slant range distance can be computed. However in areas where more than one phase cycle of movement has occurred between sampling points (PS points) within the span of a single interferogram the phase unwrapping may fail. Such areas of very high displacement rate can appear as regions devoid of PS points.
- Significantly non-linear motion
Similar gaps in coverage can occur if the motion that has occurred is significantly non-linear (e.g. a dramatic land movement after a period of stability). TRE have an Advanced PSInSAR Analysis (APSA) algorithm that utilises non-linear displacement models and can overcome this limitation. The analysis is however much more time consuming and is not offered as a standard Terrafirma product. A user may request this product if they believe that the ground motion that has occurred is significantly non-linear.
- Point target movement vs ground movement
PS points or point targets are features that provide a strong and consistent radar reflectance. They can be located on the ground on the side or on the top of buildings or structures. They may not therefore provide data on ground movement. Therefore in a given area some points may be showing the movement of a building and others of the ground while a third group may be showing movement of a building relative to the surrounding ground movement. For this reason it is important to consider the displacement data provided by the PS points in context. The results in some areas may show two sets of motions overlain – the gross movements of the ground superimposed by the movements of buildings.
- Reference Persistent Scatterer
The reference Persistent Scatterer is the point within the study area against which all ground velocity measurements are relative to. As such, it is crucial that this persistent scatterer is itself not moving over time.
Moreover, since the accuracy of scatterers’ velocity measurements are proportional to their distance from the reference scatterer it is best to locate it as near as possible to the centre of the study area being processed.
- PSI Resolution and Accuracy
- Displacement rate resolution: 0.1mm/year.
- Displacement rate precision: The precision of the relative average annual displacement rate between two neighbouring points is of the order of +/- 0.1mm/year.
- Displacement accuracy:
Absolute displacement accuracy is dependent on the specific dataset parameters: number of images, time interval covered, baselines of data used, DEM accuracy and the nature of the site. The error due to the uncompensated atmospheric component increases with distance from the reference point. The standard deviation of the average annual displacement rate for each PS point is provided within the PS data.
- Absolute X, Y accuracy: 15m - dependent on reference layer (Landsat ETM+ as standard)
- Relative X, Y accuracy: +/-5m East-West, +/- 2m North/South.
The resolution cell of an ERS / Envisat SAR image is about 20 meters in the East-West direction, and 4 meters in the North-South direction. Depending on the reference layer (Landsat ETM+ as standard), the absolute spatial accuracy is about 15m, while the relative accuracy is about +/-5m in the East-West and +/- 2m in the North-South direction. Enhanced spatial accuracy could be met if the reference layer for geo-correction is highly accurate, e.g. LIDAR or aerial ortho-photo.
- PSI Capabilities
- High precision differential measurements: relative average annual displacement rate between two neighbouring points is in the order of +/- 0.1mm/year; single event relative PS displacements typically as little as 1 mm along the Line Of Sight (LOS) and 1 cm in the East-West direction
- The maximum rate of motion that can be detectable is determined by the wavelength of the SAR - for ERS-1 / ERS-2, deformations of up to 14 mm in 35 days.
- Higher precision with respect to GPS technique for vertical measurements
- Cost and time effective measure of ground motion over medium sized regions (up to 2000km2) at a time
- Lower costs for monitoring as demanding less in-situ surveys
- Richness of data available in the ESA-ERS archive (from 1992), meaning it is possible to reconstruct the past history of the area of interest.
- High PS density in urban areas (up to 400 PS/Km2)
- Deliverables easily importable into GIS software
- To be used complementary to existing methods of monitoring (GPS, optical levelling, etc).
This text was provided by Terrafirma, a project initiated in 2003 by ESA under the GMES Service Element programme with the aim to support geohazard risk assessment using EO based measurements.