Interferometric synthetic aperture radar, abbreviated InSAR (or deprecated IfSAR), is a radar technique used in geodesy and remote sensing. This geodetic method uses two or more synthetic aperture radar (SAR) images to generate maps of surface deformation or digital elevation, using differences in the phase of the waves returning to the satellite or aircraft. The technique can potentially measure millimetre-scale changes in deformation over spans of days to years. It has applications for geophysical monitoring of natural hazards, for example earthquakes, volcanoes and landslides, and in structural engineering, in particular monitoring of subsidence and structural stability.
Technique
Synthetic aperture radar
Synthetic aperture radar (SAR) is a form of radar in which sophisticated processing of radar data is used to produce a very narrow effective beam. It can be used to form images of relatively immobile targets; moving targets can be blurred or displaced in the formed images. SAR is a form of active remote sensing – the antenna transmits radiation that is reflected from the image area, as opposed to passive sensing, where the reflection is detected from ambient illumination. SAR image acquisition is therefore independent of natural illumination and images can be taken at night. Radar uses electromagnetic radiation at microwave frequencies; the atmospheric absorption at typical radar wavelengths is very low, meaning observations are not prevented by cloud cover.
Phase
SAR makes use of the amplitude and the absolute phase of the return signal data. In contrast, interferometry uses differential phase of the reflected radiation, either from multiple passes along the same trajectory and/or from multiple displaced phase centers (antennas) on a single pass. Since the outgoing wave is produced by the satellite, the phase is known, and can be compared to the phase of the return signal. The phase of the return wave depends on the distance to the ground, since the path length to the ground and back will consist of a number of whole wavelengths plus some fraction of a wavelength. This is observable as a phase difference or phase shift in the returning wave. The total distance to the satellite (i.e., the number of whole wavelengths) is known based on the time that it takes for the energy to make the round trip back to the satellite—but it is the extra fraction of a wavelength that is of particular interest and is measured to great accuracy.
In practice, the phase of the return signal is affected by several factors, which together can make the absolute phase return in any SAR data collection essentially arbitrary, with no correlation from pixel to pixel. To get any useful information from the phase, some of these effects must be isolated and removed. Interferometry uses two images of the same area taken from the same position (or, for topographic applications, slightly different positions) and finds the difference in phase between them, producing an image known as an interferogram. This is measured in radians of phase difference and, because of the cyclic nature of phase, is recorded as repeating fringes that each represent a full cycle.
Factors affecting phase
The most important factor affecting the phase is the interaction with the ground surface. The phase of the wave may change on reflection, depending on the properties of the material. The reflected signal back from any one pixel is the summed contribution to the phase from many smaller 'targets' in that ground area, each with different dielectric properties and distances from the satellite, meaning the returned signal is arbitrary and completely uncorrelated with that from adjacent pixels. Importantly though, it is consistent – provided nothing on the ground changes the contributions from each target should sum identically each time, and hence be removed from the interferogram.
Once the ground effects have been removed, the major signal present in the interferogram is a contribution from orbital effects. For interferometry to work, the satellites must be as close as possible to the same spatial position when the images are acquired. This means that images from two satellite platforms with different orbits cannot be compared, and for a given satellite data from the same orbital track must be used. In practice the perpendicular distance between them, known as the baseline, is often known to within a few centimetres but can only be controlled on a scale of tens to hundreds of metres. This slight difference causes a regular difference in phase that changes smoothly across the interferogram and can be modelled and removed.
The slight difference in satellite position also alters the distortion caused by topography, meaning an extra phase difference is introduced by a stereoscopic effect. The longer the baseline, the smaller the topographic height needed to produce a fringe of phase change – known as the altitude of ambiguity. This effect can be exploited to calculate the topographic height, and used to produce a digital elevation model (DEM).
If the height of the topography is already known, the topographic phase contribution can be calculated and removed. This has traditionally been done in two ways. In the two-pass method, elevation data from an externally derived DEM is used in conjunction with the orbital information to calculate the phase contribution. In the three-pass method two images acquired a short time apart are used to create an interferogram, which is assumed to have no deformation signal and therefore represent the topographic contribution. This interferogram is then subtracted from a third image with a longer time separation to give the residual phase due to deformation.
Once the ground, orbital and topographic contributions have been removed the interferogram contains the deformation signal, along with any remaining noise (see Difficulties below). The signal measured in the interferogram represents the change in phase caused by an increase or decrease in distance from the ground pixel to the satellite, therefore only the component of the ground motion parallel to the satellite line of sight vector will cause a phase difference to be observed. For sensors like ERS with a small incidence angle this measures vertical motion well, but is insensitive to horizontal motion perpendicular to the line of sight (approximately north–south). It also means that vertical motion and components of horizontal motion parallel to the plane of the line of sight (approximately east–west) cannot be separately resolved.
One fringe of phase difference is generated by a ground motion of half the radar wavelength, since this corresponds to a whole wavelength increase in the two-way travel distance. Phase shifts are only resolvable relative to other points in the interferogram. Absolute deformation can be inferred by assuming one area in the interferogram (for example a point away from expected deformation sources) experienced no deformation, or by using a ground control (GPS or similar) to establish the absolute movement of a point.
Difficulties A variety of factors govern the choice of images which can be used for interferometry. The simplest is data availability – radar instruments used for interferometry commonly don't operate continuously, acquiring data only when programmed to do so. For future requirements it may be possible to request acquisition of data, but for many areas of the world archived data may be sparse. Data availability is further constrained by baseline criteria. Availability of a suitable DEM may also be a factor for two-pass InSAR; commonly 90 m SRTM data may be available for many areas, but at high latitudes or in areas of poor coverage alternative datasets must be found.
A fundamental requirement of the removal of the ground signal is that the sum of phase contributions from the individual targets within the pixel remains constant between the two images and is completely removed. However, there are several factors that can cause this criterion to fail. Firstly the two images must be accurately co-registered to a sub-pixel level to ensure that the same ground targets are contributing to that pixel. There is also a geometric constraint on the maximum length of the baseline – the difference in viewing angles must not cause phase to change over the width of one pixel by more than a wavelength. The effects of topography also influence the condition, and baselines need to be shorter if terrain gradients are high. Where co-registration is poor or the maximum baseline is exceeded the pixel phase will become incoherent – the phase becomes essentially random from pixel to pixel rather than varying smoothly, and the area appears noisy. This is also true for anything else that changes the contributions to the phase within each pixel, for example changes to the ground targets in each pixel caused by vegetation growth, landslides, agriculture or snow cover.
Another source of error present in most interferograms is caused by the propagation of the waves through the atmosphere. If the wave travelled through vacuum, it should theoretically be possible (subject to sufficient accuracy of timing) to use the two-way travel-time of the wave in combination with the phase to calculate the exact distance to the ground. However, the velocity of the wave through the atmosphere is lower than the speed of light in vacuum, and depends on air temperature, pressure and the partial pressure of water vapour. It is this unknown phase delay that prevents the integer number of wavelengths being calculated. If the atmosphere was horizontally homogeneous over the length scale of an interferogram and vertically over that of the topography then the effect would simply be a constant phase difference between the two images which, since phase difference is measured relative to other points in the interferogram, would not contribute to the signal. However, the atmosphere is laterally heterogeneous on length scales both larger and smaller than typical deformation signals. This spurious signal can appear completely unrelated to the surface features of the image, however, in other cases the atmospheric phase delay is caused by vertical inhomogeneity at low altitudes and this may result in fringes appearing to correspond with the topography.
Persistent scatterer InSAR
Persistent or permanent scatterer techniques are a relatively recent development from conventional InSAR, and rely on studying pixels which remain coherent over a sequence of interferograms. In 1999, researchers at Politecnico di Milano, Italy, developed a new multi-image approach in which one searches the stack of images for objects on the ground providing consistent and stable radar reflections back to the satellite. These objects could be the size of a pixel or, more commonly, sub-pixel sized, and are present in every image in the stack. That specific implementation is patented.
Some research centres and companies, were inspired to develop variations of their own algorithms which would also overcome InSAR's limitations. In scientific literature, these techniques are collectively referred to as persistent scatterer interferometry or PSI techniques. The term persistent scatterer interferometry (PSI) was proposed by European Space Agency (ESA) to define the second generation of radar interferometry techniques. This term is nowadays commonly accepted by scientific and the end user community.