Coseismic surface displacement mapping after earthquakes

SAR interferometry and pixel-offset tracking can map coseismic surface rupture, fault geometry, and slip distribution within hours to days of a major earthquake, feeding both source-model inversions and emergency damage assessment.

Sensors

Sentinel-1 A/B (C-band SAR, ESA): 5.6 cm wavelength; Interferometric Wide Swath mode at 5 × 20 m resolution (range × azimuth) with 250 km swath. Repeat pass every 6–12 days depending on latitude; ESA activates rapid-revisit scheduling after major events. Free and open archive since 2014.
ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength penetrates vegetation and maintains coherence in near-field zones where C-band decorrelates. ScanSAR mode reaches 490 km swath at 60 m; Stripmap at 3–10 m. JAXA routinely tasks ALOS-2 for disaster response within 24–48 hours of a major earthquake.
COSMO-SkyMed (X-band SAR, ASI/Italian MoD): 9.6 GHz X-band; Stripmap at 3 m, Spotlight at 1 m. Constellation of four satellites enables revisit of under 12 hours. High resolution makes it well suited to near-field pixel-offset tracking and damage-proxy mapping in urban areas.
Capella Space (X-band SAR, commercial): Spotlight resolution down to 0.35 m. On-demand tasking with typical revisit of hours to a day. Useful for resolving fine-scale surface rupture geometry and co-registering with optical imagery for fault-scarp characterisation.

What the interferogram actually measures, and where it breaks

A coseismic interferogram compares the phase of two SAR acquisitions, one before the earthquake and one after, to measure the change in slant-range distance between the satellite and the ground. Each full colour cycle, or fringe, represents half the radar wavelength of displacement along the line of sight: roughly 2.8 cm per fringe for Sentinel-1 C-band, about 11.8 cm for ALOS-2 L-band. A magnitude 7 or larger event can produce hundreds of fringes in the near-field, mapping the displacement field across thousands of square kilometres with millimetre-class precision far from the rupture.

The method has a hard limit. Where displacement gradients are steepest, typically within a few kilometres of the surface rupture, fringe rates exceed one cycle per pixel and the phase becomes impossible to unwrap reliably. Coherence collapses. In those near-field zones, pixel-offset tracking takes over: sub-pixel cross-correlation of amplitude patches in the pre- and post-event images can resolve displacements of roughly one-twentieth to one-thirtieth of the pixel spacing, which at Sentinel-1 Interferometric Wide Swath resolution means detectable offsets of around 0.3–0.5 m. ALOS-2 L-band retains coherence closer to the rupture than C-band because longer wavelengths are less sensitive to surface change, but even L-band loses phase coherence at the fault trace itself.

The 2023 Kahramanmaraş sequence as a working example

The 6 February 2023 Kahramanmaraş earthquakes (Mw 7.8 and Mw 7.7) ruptured over 300 km of the East Anatolian Fault system. ESA activated emergency Sentinel-1 acquisitions within hours. Coseismic interferograms and pixel-offset maps produced by multiple groups within two to three days showed left-lateral surface displacements exceeding 5–6 m in the near field, consistent with field observations. ALOS-2 descending passes acquired by JAXA within 48 hours provided complementary line-of-sight geometry that, combined with Sentinel-1 ascending data, allowed three-dimensional decomposition of the displacement field.

The near-field pixel-offset maps resolved the along-strike slip distribution in segments where the interferogram was fully decorrelated, and those slip profiles fed elastic dislocation inversions that constrained fault geometry to within a few kilometres depth. The same displacement grids, differenced against a pre-event digital elevation model, generated damage-proxy maps that correlated with collapsed building clusters identified in subsequent optical imagery. The sequence is now a standard benchmark dataset for rapid coseismic mapping workflows.

From raw phase to source model: the processing chain

The standard workflow begins with co-registration of the pre- and post-event SLC (Single Look Complex) images to sub-pixel accuracy, followed by interferogram formation, flat-earth and topographic phase removal using a reference DEM such as SRTM or Copernicus DEM at 30 m, and adaptive filtering to improve phase signal-to-noise. Phase unwrapping, typically by minimum cost flow algorithms, converts the wrapped phase to continuous displacement. In parallel, amplitude cross-correlation on the same image pair produces range and azimuth offset maps covering the near-field zone where unwrapping fails.

Combining ascending and descending passes from the same or different sensors allows decomposition into east-west and vertical displacement components; north-south sensitivity is poor for polar-orbiting SAR because the satellite flies nearly north-south and the along-track displacement is largely invisible to range measurements. Pixel-offset tracking in azimuth partially recovers north-south motion, though with lower precision than range offsets. The resulting displacement grids are then ingested by source-model inversion codes, such as the widely used elastic half-space formulations, to estimate fault geometry, dip, and slip distribution at depth.

Processing latency depends on acquisition scheduling and archive access. With pre-positioned processing infrastructure and a pre-event baseline image already in hand, an initial interferogram can be published within six to twelve hours of a post-event pass becoming available. That latency is the realistic floor, not a guarantee: orbit repeat cycles, cloud cover (irrelevant for SAR but relevant for optical damage mapping run in parallel), and data downlink windows all introduce delays.

Damage-proxy maps: a related product with different physics

Damage-proxy maps use the same pre- and post-event SAR pair but exploit coherence loss rather than phase. Where buildings collapse, the scattering geometry changes dramatically between passes, and coherence drops sharply. Pixels with low coherence in built-up areas are a statistical proxy for structural damage. NASA JPL's ARIA team operationalised this approach after the 2016 Kumamoto and 2018 Lombok earthquakes, and it has since been applied routinely by multiple agencies.

The method is fast but imprecise. It flags areas of probable damage rather than individual buildings, and it conflates building collapse with any other surface change, including post-event vehicle movement or emergency works. Spatial resolution is the binding constraint: at Sentinel-1's 20 m azimuth resolution, a single pixel covers an area larger than most individual structures. COSMO-SkyMed or Capella Spotlight imagery, at 1–3 m, resolves individual buildings and produces far more actionable damage maps, at the cost of narrower swath and commercial tasking fees.

Honest limits and what they mean for emergency response

Three constraints matter most for a buyer planning an emergency response workflow. First, SAR is weather-independent but not acquisition-independent: if no satellite passes over the affected area within the repeat cycle after the event, there is no data. Sentinel-1 repeat is 6–12 days at mid-latitudes; commercial constellations reduce this to hours but require pre-arranged tasking agreements. Second, the near-field displacement field, the zone most critical for understanding fault rupture, is precisely where InSAR is least reliable. Pixel-offset tracking fills part of this gap but at coarser displacement precision. Third, three-dimensional displacement recovery requires at least two independent look geometries, which means waiting for both ascending and descending passes, potentially from different sensors with different wavelengths, introducing systematic biases that must be corrected.

Source-model inversions derived from these data carry their own uncertainties. Elastic half-space models assume a homogeneous crust, which is never true. Fault geometry ambiguities, particularly dip direction for oblique-slip events, can produce multiple models that fit the surface displacement equally well. Publishing uncertainty bounds alongside the best-fit model is standard practice in the scientific literature and should be expected in any operational product.

Satellize processes Sentinel-1 and ALOS-2 open-archive data for coseismic displacement and can arrange commercial tasking on COSMO-SkyMed or Capella through client licence agreements, integrating both into a single displacement product and source-model summary.

Typical figures

Spatial resolution (interferogram)	5–20 m (Sentinel-1 IW); 3–10 m (ALOS-2 Stripmap); 1–3 m (COSMO-SkyMed / Capella Spotlight)
Line-of-sight displacement precision (far field)	~1–5 mm per interferogram under good coherence conditions
Pixel-offset tracking detection threshold (near field)	~0.3–0.5 m at Sentinel-1 IW; ~0.1–0.2 m at COSMO-SkyMed Stripmap
Repeat pass / revisit	6–12 days (Sentinel-1, latitude-dependent); 14 days (ALOS-2 standard); <12 hours (COSMO-SkyMed constellation); hours to 1 day (Capella, on-demand)
Processing latency after post-event pass	6–12 hours (realistic floor for initial interferogram with pre-event baseline in hand)
Radar frequency / wavelength	C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band ~3 cm (COSMO-SkyMed, Capella)
Swath width	250 km (Sentinel-1 IW); 50–490 km (ALOS-2 mode-dependent); 40 km Stripmap / 10 km Spotlight (COSMO-SkyMed)
North-south displacement sensitivity	Poor from range measurements alone; partially recovered via azimuth pixel-offset tracking at reduced precision
Archive depth	Sentinel-1: from 2014; ALOS-2: from 2014; ALOS-1 (predecessor): 2006–2011
Delivery formats	GeoTIFF displacement grids, wrapped/unwrapped phase rasters, shapefile fault-trace picks, PDF source-model summary

Analytics Satellize can run

Coseismic interferogram and unwrapped displacement map	Two-pass DInSAR with Copernicus DEM topographic correction; minimum cost flow phase unwrapping	GeoTIFF grids of line-of-sight displacement (ascending and descending geometries) with coherence mask
Near-field pixel-offset map	Sub-pixel amplitude cross-correlation (normalised cross-power spectrum) on pre- and post-event SLC pairs	Range and azimuth offset GeoTIFFs covering the decorrelated near-field zone, with uncertainty estimates
3-D surface displacement decomposition	Combination of ascending and descending line-of-sight and azimuth offsets to resolve east-west, north-south, and vertical components	Three-component displacement GeoTIFF stack with propagated uncertainty layer
Fault trace and surface rupture mapping	Gradient analysis of displacement field and visual interpretation of offset fringes and pixel-offset discontinuities	Shapefile of mapped surface rupture with displacement magnitude attributed per segment
Elastic dislocation source model	Okada-type elastic half-space inversion of displacement grids; Monte Carlo uncertainty sampling	PDF report with best-fit fault geometry, slip distribution map, and uncertainty bounds; GeoTIFF slip model
Damage-proxy map	Coherence change detection between pre- and post-event SAR pairs in built-up areas	GeoTIFF damage-probability layer classified into low / moderate / high change, with caveat annotation
Rapid displacement bulletin	Automated interferogram generation triggered on new Sentinel-1 acquisition; quality-checked by analyst before release	Structured alert (JSON + PDF) with peak displacement, affected area extent, and data-quality flags, issued within 12 hours of pass

Sources

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.