InSAR Technology and Methods

We measure ground deformation with the full range of satellite radar interferometry (InSAR) techniques. Each method suits a different pattern of movement, from slow millimetre-scale settlement to fast metre-scale displacement, and we select the approach, or combination of approaches, that fits the site and the question being asked.

The methods below span conventional differential interferometry, speckle (pixel) offset tracking for large displacements, deformation time series built with PSI, SBAS, and multidimensional MSBAS, and the satellite radar data that feeds them all, from the historical archive of the early 1990s to today's Sentinel-1 and high-resolution commercial sensors. These methods underpin our ground deformation monitoring services - from reservoir and injection deformation to slope and subsidence mapping - have been applied at national scale to build wide-area deformation mosaics from large Sentinel-1 archives, and report a standard error for every measured rate.

Differential InSAR (DInSAR) compares the phase of two radar images taken at different times to map the ground movement that occurred between them. A single interferogram can reveal a deformation pattern across an entire scene at centimetre to millimetre sensitivity. It is the foundation of every other method here and is well suited to mapping individual events such as an earthquake, a sudden subsidence, or volcanic uplift.

DInSAR measures movement along the satellite line of sight and relies on phase unwrapping, which is reliable only for small to moderate deformation. Where the surface stays coherent between passes, it delivers the highest precision of any of these techniques.

When movement is too large or too fast for the phase to be unwrapped reliably, we track the displacement of image features between acquisitions instead. Pixel offset tracking measures motion in both range and azimuth and can follow displacements of metres, for example fast-moving landslides, glaciers, or the ground above an active mine, where conventional interferometry loses coherence.

Because speckle offset tracking does not require phase unwrapping, it captures motion that DInSAR cannot, including movement along the satellite flight (azimuth) direction. Its precision is lower than DInSAR and depends on the spatial resolution of the radar data, so the two are often used together.

A deformation time series turns a stack of individual radar measurements into a continuous history of how each point on the ground has moved. Rather than a single before-and-after pair, dozens to hundreds of acquisitions are inverted together to estimate the deformation rate between successive dates, which is then accumulated into a cumulative displacement record for every pixel. The result is not just how much the ground has moved, but how the movement has evolved, where it is accelerating, slowing, or reversing.

Because real radar data are only partially coherent, the inversion is regularised and a per-pixel quality (rank) check removes points where too few reliable measurements exist, so the time series reports movement only where it can be trusted. The temporal sampling can be adjusted, for example to annual or seasonal rates, to widen spatial coverage, shorten processing, or make seasonal signals clearer. PSI, SBAS, and MSBAS are all time-series methods built on this principle.

PSI follows a network of stable, strongly reflecting points, such as buildings, rock outcrops, and infrastructure, through a long stack of radar images. By isolating these persistent scatterers it reaches very high precision on individual targets and is ideal for monitoring structures, urban subsidence, and infrastructure over many years.

SBAS builds a deformation time series from many interferograms formed between image pairs with small separation in time and position. This preserves coherence over natural, non-urban terrain and produces spatially continuous movement maps, making it well suited to wide-area subsidence, landslides, and slow regional deformation.

MSBAS combines deformation measurements from two or more satellite tracks, typically ascending and descending orbits, and can use both DInSAR interferograms and speckle offset (range and azimuth) data together. Because each track and each measurement direction sees the ground differently, inverting them jointly recovers the actual direction of motion rather than movement along one line of sight alone. Depending on how many viewing geometries and what kind of input data are available, MSBAS produces one-, two-, three-, or four-dimensional deformation time series, optionally constrained by surface topography.

Combines overlapping data sets that share a similar viewing geometry into one densely sampled, one-dimensional time series along the radar line of sight. This increases how often the ground is measured and extends the record further back in time.

Combines ascending and descending data to separate movement into vertical and east-west horizontal components. It assumes negligible north-south motion, a good approximation for many subsidence and uplift settings, and is the most widely used form for groundwater, mining, and volcanic deformation.

When azimuth-offset measurements are added to the ascending and descending data, MSBAS resolves the full three-dimensional displacement field, north, east, and vertical, without assuming any direction is zero. This gives a complete, unconstrained picture of how the surface is actually moving.

A topographically constrained three-dimensional solution that assumes the surface moves parallel to the terrain, using a digital elevation model to relate the components. This recovers full 3D motion from fewer viewing geometries and is well suited to slow-moving deep-seated landslides and glacier flow, where material travels along the ground surface.

A related constrained solution in which horizontal motion is assumed to follow the downslope aspect direction taken from the elevation model. Like the surface-parallel form, it resolves three-dimensional displacement for deep-seated landslides from limited data by constraining the flow direction with the local slope.

Resolves north, east, and two separate vertical subcomponents, separating two superimposed vertical processes, for example a long-term trend and a seasonal or transient signal, and extends the result through time. This four-dimensional record shows how each part of the full displacement field evolves over the monitoring period.

Every result depends on the radar data behind it. We work across the full archive of civilian synthetic aperture radar (SAR) satellites, combining a deep historical record with the latest high-frequency and high-resolution sensors.

Spaceborne SAR suitable for interferometry reaches back to the early 1990s, through ERS-1 and ERS-2, Envisat, JERS-1, and ALOS PALSAR. This archive lets us reconstruct ground movement that occurred years or decades before a project began and establish a historical baseline for sites that were never instrumented, which is often the only way to understand how a slope, aquifer, or mine has behaved over the long term.

Most current monitoring uses Sentinel-1, the European Copernicus C-band mission, which images wide areas every few days and is openly and freely available - the workhorse for wide-area monitoring. For fine detail we use X-band from Germany's TerraSAR-X and Italy's COSMO-SkyMed, which resolve individual structures down to about a metre, and we draw on L-band from Japan's ALOS-2, choosing C-band, X-band, or L-band and the ascending and descending geometries each site needs. Longer-wavelength L-band such as ALOS-2 holds coherence over vegetation where shorter wavelengths decorrelate, which matters over forested and northern terrain.

There is a trade-off between coverage, resolution, and cost. Sentinel-1 is free, wide-area, and medium-resolution; commercial high-resolution sensors such as TerraSAR-X resolve fine detail on individual structures but are expensive and cover smaller areas. Our standard strategy is to begin with free Sentinel-1 data to screen the whole site, build the historical baseline, and locate where the ground is moving, and then commission higher-resolution commercial data only where finer detail or precision is justified.

Processing these large radar stacks is computationally intensive: hundreds of images are inverted pixel by pixel, so our software is parallelised to run on multi-node compute clusters. This is what makes wide-area, multi-year monitoring of an entire site practical.

In practice we often combine several of these techniques: a PSI or SBAS time series for the long-term trend, offset tracking where movement is large, and MSBAS to resolve direction by merging multiple satellites. The right combination depends on the speed and scale of the movement, the land cover, and how the results will be used. Our processing methods are grounded in peer-reviewed research published in international remote-sensing journals.

What is InSAR?

InSAR (Interferometric Synthetic Aperture Radar) measures ground movement from space by comparing the phase of repeated satellite radar images, mapping deformation across wide areas at millimetre to centimetre precision without any instrument on the ground.

What is the difference between DInSAR and offset tracking?

DInSAR measures small to moderate movement very precisely by comparing radar phase, but it requires phase unwrapping. Pixel (speckle) offset tracking measures much larger, faster movement, metres of displacement, by following image features without unwrapping, at lower precision. MSBAS can combine both.

What is the difference between PSI and SBAS?

PSI tracks a network of stable, strongly reflecting points at very high precision and works best in built-up areas, while SBAS produces spatially continuous maps and preserves coherence over natural terrain. Many projects use both.

What does MSBAS add over standard InSAR?

A single radar image measures movement only along the satellite line of sight. MSBAS combines several viewing geometries to recover vertical, east-west, and north-south motion, and can extend this through time as a four-dimensional record.

Which satellites do you use?

We primarily use Sentinel-1, which is free and covers wide areas every few days, together with high-resolution commercial sensors such as TerraSAR-X when fine detail is needed. We also draw on the historical SAR archive going back to the 1990s (ERS, Envisat, JERS, ALOS) to reconstruct past movement.

Can you measure deformation that happened in the past?

Yes. The satellite radar archive extends to the early 1990s, so we can often reconstruct years or decades of past ground movement at a site, even where no instruments were ever installed.