SAR interferometric surface deformation mapping
InSAR compares the phase of repeat SAR passes to map ground displacement at millimetre scale, revealing subsidence, uplift and fault creep invisible to optical sensors. Coherence, atmospheric delay and imaging geometry all shape what the technique can and cannot see.
Sensors
- Sentinel-1 (ESA): C-band (5.6 cm wavelength) SAR operating in Interferometric Wide Swath mode; 250 km swath, 5 × 20 m resolution in IW mode, 6-day repeat at the equator with both satellites active. Free and open archive from 2014. Workhorse of global InSAR monitoring.
- ALOS-2 PALSAR-2 (JAXA): L-band (23.6 cm wavelength) SAR; 14-day repeat. Longer wavelength penetrates vegetation canopy and maintains coherence over forested and agricultural terrain where C-band decorrelates badly. Stripmap mode reaches 3 m resolution.
- ICEYE SAR constellation: X-band (3.1 cm wavelength), sub-metre resolution in spotlight mode, revisit configurable to hours over a target through tasking. High coherence over urban hard targets. Shorter wavelength increases sensitivity to atmospheric water vapour noise.
- Capella Space SAR: X-band spotlight SAR with published 50 cm resolution. Rapid tasking enables short temporal baselines, which reduces decorrelation and atmospheric artefacts. Useful for time-critical infrastructure monitoring where archive depth is less important than immediacy.
What phase difference actually measures
Every SAR image records two numbers per pixel: amplitude (how strongly the surface scattered the radar pulse) and phase (where in the wave cycle that echo arrived). A single phase image is noise. Two images of the same patch of ground, acquired from nearly the same orbital position at different times, are not. Subtract one phase from the other and you get an interferogram: a map of how the path length between satellite and ground changed between passes.
One full colour cycle in that interferogram, called a fringe, corresponds to half the radar wavelength of range change. For Sentinel-1 at C-band, one fringe equals roughly 2.8 cm of displacement along the satellite's line of sight. Algorithms can resolve a fraction of a fringe, pushing detectable displacement down to a few millimetres over stable, coherent surfaces. That sensitivity is why InSAR can track slow aquifer depletion, permafrost thaw settlement and tectonic interseismic strain that no field survey would catch in time.
Coherence: the precondition nobody advertises enough
The interferogram only works if the ground scatterers in the two images are arranged the same way. Coherence is the statistical measure of that similarity, ranging from zero (pure noise) to one (perfect). Urban hard targets, bare rock and dry soil hold coherence well over weeks. Agricultural fields can decorrelate completely between a six-day Sentinel-1 pair as crops grow or are harvested. Dense tropical forest is almost always incoherent at C-band; the canopy moves with wind and the scattering geometry changes daily.
This is not a solvable problem with better processing. It is a physics constraint. L-band systems such as PALSAR-2 partially circumvent it because the longer wavelength penetrates the canopy and reflects from the more stable trunk and ground layer beneath. Even so, dense wet forest at L-band can decorrelate over 46-day baselines. Analysts working over vegetated terrain must account for this honestly rather than treating the resulting phase noise as signal.
The atmosphere is not transparent to radar
Tropospheric water vapour slows radar pulses and adds apparent path length. A column of humid air over one part of the scene and dry air over another creates a spurious phase ramp or blob that looks, in the interferogram, exactly like ground deformation. Over mountainous terrain, where water vapour correlates with elevation, this atmospheric signal can mimic or mask real displacement of several centimetres.
Correction methods exist. Generic Atmospheric Correction Online Service (GACOS) uses weather model outputs and GPS zenith total delay data to estimate and remove tropospheric delay. Ionospheric delay, significant at L-band at high latitudes, can be estimated from the range split-spectrum method. Neither correction is perfect; residual atmospheric noise typically limits single-interferogram accuracy to roughly 1 cm. Time-series methods such as Persistent Scatterer InSAR (PSI) and Small Baseline Subset (SBAS) average over many interferograms and push that floor down to 1 to 2 mm per year over stable point targets, but they require a long archive and assume the deformation is slow and temporally coherent.
Imaging geometry and the line-of-sight ambiguity
Sentinel-1 IW mode acquires in both ascending (roughly north-west looking) and descending (roughly north-east looking) passes. InSAR measures displacement only along the satellite's line of sight, not vertically or horizontally in isolation. A purely vertical subsidence of 10 mm and a purely horizontal east-west motion of similar magnitude can produce indistinguishable phase signals on a single pass geometry.
Combining ascending and descending interferograms allows decomposition into vertical and east-west horizontal components. North-south displacement remains largely invisible because the satellite flies nearly north-south and has almost no sensitivity in that direction. This is a fundamental geometric limit, not an instrument deficiency. For applications such as fault slip monitoring where north-south motion dominates, analysts must supplement InSAR with GPS or offset-tracking methods applied to the SAR amplitude images.
From interferogram to operational product
Raw interferograms are not deliverables. Phase must be unwrapped (converting cyclic 0-to-2π values into continuous displacement), geocoded to a geographic coordinate system, and corrected for orbital errors, topographic phase (using a digital elevation model such as SRTM or Copernicus DEM) and atmospheric delay. Time-series stacking then separates linear deformation rates from seasonal signals and residual noise.
The outputs that matter operationally are displacement velocity maps (millimetres per year over a monitoring period), time-series plots for individual persistent scatterers anchored to stable reference points, and threshold alerts when a structure or slope exceeds a defined rate. For urban infrastructure monitoring, PSI can resolve individual buildings; published studies on cities including Shanghai, Mexico City and London have reported subsidence rates from a few millimetres to several centimetres per year using this approach. Satellize runs InSAR time-series processing on Sentinel-1 open-archive data and can add commercial X-band tasking where higher spatial resolution or shorter revisit is required for a client's specific infrastructure.
Where the technique reaches its limits
Rapid or large displacements can exceed one fringe per pixel, causing phase aliasing that unwrapping algorithms cannot resolve without additional constraints. Co-seismic displacement from a large earthquake can reach metres; standard InSAR fails and pixel-offset tracking on SAR amplitude becomes the fallback. Similarly, very steep slopes facing away from the radar are in shadow or layover and produce no usable signal at all.
Temporal resolution is the other honest constraint. Sentinel-1's six-day revisit means fast-moving landslides or sudden sinkhole events may not be captured mid-motion. Commercial constellations shorten this, but at cost. And the millimetre-scale precision headline applies only to PSI results over long time series on stable point targets. A single interferogram pair over a non-urban area is better described as centimetre-scale. Buyers who conflate the two figures will be disappointed.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 5 × 20 m (range × azimuth); multi-looked to ~14 m for interferometry |
| Spatial resolution (commercial X-band spotlight) | 0.5 to 1 m (ICEYE, Capella); enables building-level persistent scatterer density |
| Revisit period | 6 days (Sentinel-1 two-satellite); 14 days (ALOS-2); hours to days (commercial tasking) |
| Radar frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1); L-band 1.27 GHz / 23.6 cm (PALSAR-2); X-band ~9.6 GHz / 3.1 cm (ICEYE, Capella) |
| Minimum detectable displacement (single interferogram) | ~5 mm line-of-sight over coherent surfaces; practical accuracy ~1 cm with atmospheric correction |
| Minimum detectable displacement (PSI time-series) | 1 to 2 mm per year velocity over stable point targets with sufficient archive |
| Swath width | 250 km (Sentinel-1 IW); 70 km (ALOS-2 stripmap); 5 to 15 km (commercial spotlight) |
| Archive depth | Sentinel-1 from 2014; ALOS-2 from 2014; ICEYE/Capella from 2019 to 2020 |
| Delivery formats | GeoTIFF displacement maps, shapefiles of PS/DS point velocities, CSV time-series, NetCDF stacks |
| Cloud penetration | Full: SAR is unaffected by cloud cover or darkness |
Analytics Satellize can run
| Subsidence velocity map | SBAS or PSI time-series InSAR over Sentinel-1 archive | GeoTIFF raster of mm/year displacement rates, updated quarterly or on new acquisition |
| Persistent scatterer point network | PSI on urban hard targets; phase stable pixels selected by amplitude dispersion index | Shapefile or GeoPackage of PS points with velocity, time-series and uncertainty per point |
| Slope instability alert | SBAS displacement time-series with threshold trigger on acceleration or cumulative displacement | Automated alert (email or API push) when defined threshold is crossed at a monitored slope or structure |
| Co-seismic or volcanic deformation map | Single-pair interferogram with Goldstein filter, phase unwrapping and DEM-phase removal | Wrapped and unwrapped interferogram GeoTIFFs with line-of-sight displacement in metres, delivered within 24 hours of post-event SAR acquisition |
| Vertical / horizontal displacement decomposition | Ascending and descending geometry combination following published vector decomposition method | Two-band GeoTIFF: vertical and east-west horizontal velocity components with propagated uncertainty |
| Infrastructure settlement monitoring report | High-resolution X-band PSI over dam, bridge or building footprint using commercial tasking | Periodic PDF report with PS time-series plots per structure, anomaly flagging and comparison to engineering tolerance thresholds |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.