Earthquake surface deformation measurement with InSAR
Differential InSAR turns phase shifts between two SAR passes into centimetre-scale maps of coseismic ground displacement, revealing fault geometry and slip distribution within days of a major earthquake.
Sensors
- Sentinel-1 IW (ESA/Copernicus): C-band (5.6 cm wavelength), 250 km swath, 5 × 20 m ground range resolution in Interferometric Wide swath mode. Revisit 6 days at the equator with both satellites active, shorter at higher latitudes. Free and open archive from 2014. The workhorse for rapid coseismic interferograms worldwide.
- ALOS-2 PALSAR-2 (JAXA): L-band (23.6 cm wavelength), which maintains coherence in vegetated terrain where C-band decorrelates. ScanSAR mode covers 350 km at 100 m resolution; Stripmap reaches 3 m at 70 km swath. Revisit 14 days. JAXA activates emergency observation protocols for major earthquakes, typically releasing data within 24–48 hours of a request.
- RADARSAT Constellation Mission (CSA): C-band, three satellites, revisit as short as 4 days for a single look or daily for some latitudes in compact-polarimetry mode. Medium-resolution ScanSAR reaches 100 m; High-resolution Spotlight reaches 1 m. Useful for rapid repeat acquisitions in the first week after an event.
- NISAR (NASA/ISRO, expected ~2025): Dual-band L- and S-band system designed explicitly for deformation science. Planned 12-day global repeat at L-band. Once operational, NISAR will reduce the coherence-loss problem in tropical and forested earthquake zones that currently limits Sentinel-1 and even ALOS-2.
What a phase shift actually tells you
Synthetic aperture radar measures the two-way travel time of a microwave pulse to the ground and back. When the same patch of ground is imaged on two separate passes, any displacement along the satellite's line of sight changes that travel time by a fraction of the radar wavelength. Differential InSAR (DInSAR) isolates that change by subtracting the phase contribution of the pre-event topography (using a digital elevation model) from the interferometric phase. What remains encodes displacement. Each full colour cycle in the resulting fringe pattern represents half a wavelength of line-of-sight movement: roughly 2.8 cm for Sentinel-1's C-band, or 11.8 cm for ALOS-2's L-band.
A well-formed coseismic interferogram over a magnitude 6.5 or larger event typically shows concentric fringe lobes on each side of the fault, with fringe density highest in the near-field where displacement gradients are steepest. Counting fringes and inverting their geometry allows geodesists to estimate fault strike, dip, slip magnitude and depth, often to within a few kilometres, without a single field measurement. For emergency responders, the practical output is a displacement map available within 24–72 hours of a post-event SAR acquisition, showing which districts experienced the most ground movement.
Coherence: the thing that can make the measurement fail
DInSAR requires that the scattering characteristics of the ground surface remain stable between the two acquisitions. If they change, the phase difference becomes noise rather than signal. This loss of correlation is called temporal decorrelation, and it is the principal limit of the technique. Dense vegetation, freshly tilled agricultural fields, and loose sand all decorrelate rapidly, sometimes within a single satellite pass. A heavily forested epicentral region can produce an interferogram that is essentially uninterpretable in the near-field, precisely where displacement is largest.
L-band radar (ALOS-2) penetrates vegetation canopies more deeply and scatters from stable woody stems and soil, so it retains coherence in conditions that defeat C-band. That is why JAXA emergency acquisitions are often the product of choice in tropical or heavily vegetated earthquake zones. Even so, coherence is never guaranteed. Landslides, liquefaction, and surface rupture itself can destroy the scattering geometry entirely, creating holes in the deformation map that must be acknowledged rather than interpolated over.
Atmospheric artefacts are not a minor nuisance
The troposphere delays radar signals in proportion to water vapour content. Spatial gradients in humidity, particularly over mountainous terrain, can introduce apparent phase signals of several centimetres that are indistinguishable from tectonic deformation without correction. For a large earthquake this is manageable: fringes from a magnitude 7 event dwarf typical atmospheric noise. For smaller events, or for the subtle postseismic signals that follow in the weeks after the mainshock, atmospheric artefacts can be the dominant signal.
Standard correction approaches include using GNSS zenith total delay estimates from nearby continuously operating reference stations, applying ERA5 reanalysis weather model outputs, or using empirical phase-elevation correlations to separate topographically correlated atmospheric delay from genuine deformation. The COMET LiCSAR processing chain, developed at the University of Leeds and COMET (the UK Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics), automates Sentinel-1 interferogram production globally and applies GACOS (Generic Atmospheric Correction Online Service) tropospheric corrections as a standard step. LiCSAR has processed tens of thousands of interferograms and provides a publicly searchable archive. It is one of the most important open resources in operational earthquake geodesy.
Near-field rupture versus the broader deformation field
The near-field, typically within one to two fault lengths of the rupture, is where surface displacement is largest and where fringe rates can exceed the interferogram's ability to resolve them. If displacement between adjacent resolution cells exceeds half a wavelength (2.8 cm for Sentinel-1), the phase wraps ambiguously and standard unwrapping algorithms fail or introduce errors. This is called phase aliasing. Pixel offset tracking, which cross-correlates amplitude patches between pre- and post-event images rather than using phase, can measure displacements of metres in the near-field where DInSAR saturates. The two methods are complementary: DInSAR for the far-field centimetre-scale signal, pixel offset tracking for the near-field metre-scale rupture.
The broader deformation field, extending tens to hundreds of kilometres from the fault, captures elastic strain release in the surrounding crust. This is directly useful for estimating the earthquake's source parameters and for identifying areas of increased Coulomb stress that may be primed for aftershocks. Emergency managers increasingly use these maps not just to assess immediate damage but to anticipate where secondary hazards, including aftershock-triggered landslides, are most likely.
Processing timelines and honest operational limits
Sentinel-1's systematic acquisition strategy means that for most seismically active regions a pre-event image already exists in the archive. The critical variable is when the next post-event pass occurs. In the best case, a Sentinel-1 pass crosses the epicentre within 12 hours of the earthquake and a processed interferogram is available within another 12 hours. In the worst case, the next pass is six days away, the orbit geometry is unfavourable, or cloud cover has already triggered a competing demand for optical imagery (which InSAR does not need, since radar penetrates cloud). Latency for ALOS-2 emergency products depends on JAXA's observation scheduling and data distribution protocols, typically 24–72 hours from acquisition to product delivery.
The minimum detectable displacement is theoretically a few millimetres under ideal coherence and atmospheric conditions, but in practice the noise floor for a single interferogram over rough terrain is closer to one to two centimetres. Stacking multiple interferograms (time-series InSAR, using methods such as SBAS or PS-InSAR) reduces noise substantially, but that approach requires weeks to months of post-event acquisitions and is more relevant to postseismic monitoring than to immediate response. Satellize integrates Sentinel-1 and ALOS-2 processing pipelines into sovereign disaster-response workflows, applying GACOS atmospheric correction and delivering georeferenced displacement GeoTIFFs and fault-trace shapefiles within agreed service-level windows. The Tonga crop-estimation programme is a separate analytics engagement, but the underlying infrastructure for systematic open-constellation processing is shared.
What the product cannot do
InSAR measures line-of-sight displacement, not three-dimensional movement. Recovering the full three-dimensional displacement vector requires combining ascending and descending orbit interferograms, and ideally adding pixel offset tracking in the along-track direction. Even then, the north-south component is poorly constrained by polar-orbiting SAR geometry and must be supplemented by GNSS.
The technique gives no information about building damage directly. A district showing two centimetres of coherent subsidence may have experienced severe structural damage or none at all, depending on soil type, construction quality and whether the displacement was uniform or differential. Building-damage assessment from InSAR coherence loss is a related but distinct method covered in the sibling page on SAR coherence change detection. Finally, InSAR cannot see through dense forest to a buried fault with no surface expression. Where the rupture is entirely blind, the deformation field may be detectable but the fault trace itself must be inferred from the displacement pattern rather than observed directly.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 5 × 20 m (ground range × azimuth); interferometric products typically multilooked to 40–80 m for phase quality |
| Spatial resolution (ALOS-2 Stripmap) | 3–10 m depending on mode; ScanSAR 100 m |
| Revisit (Sentinel-1, both satellites) | 6 days at equator; 1–3 days at mid-to-high latitudes due to overlapping tracks |
| Revisit (ALOS-2) | 14 days standard; emergency tasking can deviate from standard cycle |
| Radar frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1, RCM); L-band 1.236 GHz / 23.6 cm (ALOS-2, NISAR) |
| Minimum detectable LOS displacement (single interferogram) | ~1–2 cm under good coherence and after atmospheric correction; theoretically a few mm with time-series stacking |
| Phase aliasing limit (Sentinel-1) | Fringes unresolvable where LOS displacement gradient exceeds ~2.8 cm per resolution cell; pixel offset tracking used beyond this |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); global systematic acquisition from 2016 |
| Latency (rapid response product) | 12–72 hours from post-event SAR acquisition depending on sensor and processing pipeline |
| Delivery formats | Georeferenced GeoTIFF (wrapped and unwrapped phase, displacement in metres LOS), coherence raster, fault-trace shapefile, PDF report |
Analytics Satellize can run
| Coseismic displacement map | Two-pass DInSAR with SRTM or Copernicus DEM topographic removal; GACOS tropospheric correction | Unwrapped LOS displacement GeoTIFF with colour-scale legend, delivered as GIS layer and PDF summary within agreed SLA |
| Fault rupture trace delineation | Phase gradient analysis and pixel offset tracking on amplitude imagery to locate near-field discontinuities | Shapefile of inferred surface rupture or maximum displacement corridor, with confidence classification |
| Earthquake source parameter estimate | Elastic dislocation modelling (Okada formulation) fitted to unwrapped displacement field | Tabular fault model (strike, dip, rake, slip, depth, length) with uncertainty ranges, delivered as structured report |
| Atmospheric-corrected interferogram stack | SBAS or PS-InSAR time-series using post-event Sentinel-1 acquisitions; ERA5 or GACOS delay correction per epoch | Time-series displacement GeoTIFFs at 12-day intervals showing postseismic relaxation, as GIS-ready archive |
| Priority zone overlay for field teams | Displacement magnitude and gradient thresholding combined with population grid (WorldPop or GPW) | Ranked district list and map showing population exposed to displacement above defined thresholds, as PDF and GeoJSON |
| Coherence change detection (complementary) | Pre- to post-event interferometric coherence differencing to flag areas of surface disruption | Coherence-loss raster flagging likely landslide, liquefaction or building-collapse zones, as GIS layer and alert |
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.