Interseismic strain accumulation on active faults
Between earthquakes, locked fault segments store elastic strain that will eventually release violently. InSAR time series and GNSS networks can measure this slow accumulation and constrain where slip deficits are growing.
Sensors
- Sentinel-1 (C-band SAR, ESA): 5.6 cm wavelength; 5 x 20 m ground range resolution in IW mode; 6-day repeat at mid-latitudes with both satellites active. Penetrates cloud and darkness. Archive from 2014 enables multi-year time series essential for separating interseismic signal from seasonal atmospheric noise.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength; 3 x 6 m resolution in spotlight mode, 10 x 10 m in stripmap. L-band coherence is better than C-band over vegetated terrain and across long time intervals, making it preferable for fault segments crossing forests or agricultural land. Repeat cycle 14 days.
- GNSS/GPS continuous networks: Sub-millimetre annual velocity precision at individual monuments when multi-year solutions are computed. Provides absolute reference frame and 3-D displacement vectors that InSAR alone cannot supply. Dense networks such as the Plate Boundary Observatory (now EarthScope) and the Turkish national CORS network are publicly archived.
- COSMO-SkyMed (X-band SAR, ASI/Italian MoD): Wavelength 3.1 cm; spotlight mode delivers 1 m resolution, useful for resolving near-fault deformation gradients in urban areas. Shorter wavelength increases atmospheric sensitivity, so atmospheric correction is more demanding than for C- or L-band. Commercial tasking allows flexible acquisition geometry.
What the ground is doing between earthquakes
A locked fault behaves like a compressed spring. The tectonic plates on either side keep moving; the locked interface does not. Elastic strain accumulates in the surrounding crust, expressed at the surface as a smooth velocity gradient that decays with distance from the fault trace. On a classic strike-slip fault such as the San Andreas, surface velocities transition from near-zero on the locked side to the full plate-motion rate kilometres away. The width of that gradient encodes the locking depth.
Measuring this gradient precisely enough to constrain hazard models is genuinely difficult. The signal is typically 2 to 20 mm per year across tens of kilometres. Atmospheric water vapour introduces path-delay noise in individual InSAR interferograms that can reach 10 to 20 mm, easily swamping the interseismic signal in a single scene. Only multi-year stacking, combined with independent atmospheric correction using ERA5 reanalysis or GNSS zenith delay estimates, brings the interseismic velocity field into view.
From interferograms to coupling coefficients
The standard processing chain starts with a time-series algorithm. SBAS (Small Baseline Subset) and PS-InSAR (Persistent Scatterer) are the two dominant published approaches. SBAS works well over distributed targets such as bare rock and agricultural fields; PS-InSAR suits urban areas where individual scatterers remain coherent across years. Both produce mean line-of-sight velocity maps, typically at 20 to 100 m posting, after stacking dozens to hundreds of interferograms.
Those velocity maps are then decomposed into fault-parallel and fault-perpendicular components using the satellite look geometry and, where available, ascending and descending pass data combined. The resulting surface velocity field is fed into an elastic half-space model, most commonly a variant of the Okada (1985) dislocation model, to invert for fault coupling coefficients and slip-deficit rates at depth. Published studies of the North Anatolian Fault using Sentinel-1 have resolved coupling variations along strike at roughly 10 to 20 km resolution. Studies of the Himalayan Main Frontal Thrust using ALOS-2 have estimated geodetic locking depths of 15 to 20 km and slip-deficit rates of 14 to 20 mm per year, broadly consistent with geological shortening estimates.
Coupling coefficients range from zero (fully creeping, no strain stored) to one (fully locked, all relative motion stored as elastic strain). Segments with coupling near one and long elapsed time since the last major rupture carry the highest unrelieved moment deficit.
Where the method struggles
Honest accounting matters here. The interseismic velocity signal on a moderately coupled fault can be as small as 2 to 3 mm per year in line-of-sight. Sentinel-1 time series over three to five years can approach 0.5 to 1 mm per year precision under favourable conditions, but that precision degrades sharply over vegetated terrain where coherence drops, in arid regions with strong tropospheric turbulence, and at high latitudes where orbital geometry reduces sensitivity to horizontal motion.
Atmospheric correction is the single largest source of systematic error. Even after ERA5-based correction, residual tropospheric noise at the 1 to 2 mm per year level is common. This means that on slow-slipping faults accumulating strain at rates below 5 mm per year, a multi-year GNSS solution is not optional; it is the only way to anchor the InSAR velocity field and separate tectonic signal from artefact.
There is also a fundamental model ambiguity. The same surface velocity pattern can be reproduced by different combinations of locking depth and coupling coefficient. Resolving this requires dense spatial sampling of the velocity gradient, ideally from both ascending and descending satellite passes, plus independent constraints on fault geometry from seismicity or geological mapping. A single-pass InSAR solution without GNSS integration should be treated with caution when the coupling inversion is the primary output.
What a well-constrained strain model actually tells a client
A slip-deficit map is a probabilistic input, not a deterministic prediction. It tells a national geological survey or a catastrophe risk modeller where moment is accumulating and at what rate, which feeds directly into probabilistic seismic hazard analysis (PSHA). The Himalayan frontal thrust example is instructive: geodetic estimates of slip deficit, combined with historical rupture chronologies from palaeoseismic trenching, suggest that several segments have not ruptured in 500 to 700 years and have accumulated sufficient deficit for magnitude 8 or greater events.
For infrastructure owners, the practical output is a ranked list of fault segments by unrelieved moment deficit per unit length, updated as new InSAR epochs are processed. That ranking informs where to invest in strong-motion network densification, where to apply stricter setback rules in land-use planning, and where to commission targeted palaeoseismic investigations to constrain recurrence intervals.
Archive depth and operational cadence
Sentinel-1's archive from April 2014 (Sentinel-1A) and April 2016 (Sentinel-1B, now decommissioned, with Sentinel-1C launched in late 2023) provides a decade-scale baseline over most seismically active regions. A five-year stack of 6-day repeat acquisitions over a well-covered fault can yield 300 or more interferogram pairs for SBAS processing. ALOS-2 archives from 2014 complement this, particularly over vegetated fault zones in Southeast Asia and the Himalayan arc.
Satellize processes these open constellations and can add commercial COSMO-SkyMed tasking where near-fault urban areas require higher spatial resolution. The analytic pipeline follows the published SBAS and Okada-inversion methods described above.
Typical figures
| Spatial resolution (velocity map) | 20 to 100 m posting typical for Sentinel-1 SBAS; 10 to 30 m for ALOS-2 stripmap; 5 to 10 m for COSMO-SkyMed spotlight |
| Velocity precision (line-of-sight) | 0.5 to 2 mm/yr over 3 to 5 year stacks under good coherence; degrades to 3 to 5 mm/yr over vegetated or high-turbulence terrain |
| Minimum detectable interseismic rate | Approximately 2 to 3 mm/yr in LOS with multi-year Sentinel-1 stack and atmospheric correction; GNSS anchor required below this threshold |
| Revisit (Sentinel-1) | 6 days at mid-latitudes with two active satellites; 12 days with one satellite |
| Revisit (ALOS-2) | 14 days standard; emergency observation mode can reduce this |
| Radar frequency / wavelength | C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band 3.1 cm (COSMO-SkyMed) |
| Archive depth | Sentinel-1 from 2014; ALOS-2 from 2014; ERS/Envisat back to 1992 for historical context (reduced precision) |
| Coverage | Sentinel-1 systematic global coverage of seismic zones; ALOS-2 and COSMO-SkyMed by tasking |
| Coupling inversion resolution along strike | 10 to 20 km patch size typical, constrained by data density and model regularisation |
| Deliverable formats | GeoTIFF velocity grids, GeoJSON fault-segment coupling tables, NetCDF time-series cubes, PDF technical report |
Analytics Satellize can run
| Mean LOS velocity map | SBAS or PS-InSAR time-series analysis on multi-year Sentinel-1 or ALOS-2 stack with ERA5 atmospheric correction | GeoTIFF raster, millimetres per year in satellite line-of-sight, with uncertainty layer |
| Fault-parallel horizontal velocity profile | Decomposition of ascending and descending LOS velocities into horizontal and vertical components using orbital geometry | CSV velocity profiles perpendicular to fault strike, suitable for direct input to elastic dislocation models |
| Coupling coefficient map | Okada elastic half-space inversion of surface velocity field, regularised by fault geometry from published catalogues and seismicity depths | GeoJSON or shapefile of fault patches with coupling coefficient, slip-deficit rate (mm/yr), and 1-sigma uncertainty |
| Moment deficit accumulation rate | Integration of slip-deficit rate over locked fault area, using published shear modulus for crustal rock (typically 30 to 33 GPa) | Table of fault segments ranked by moment deficit rate (N·m/yr), formatted for input to PSHA workflows |
| GNSS-InSAR integrated velocity solution | Weighted combination of continuous GNSS station velocities (from public networks) with InSAR LOS velocities to constrain absolute reference frame and 3-D motion | Merged velocity grid in ITRF reference frame, GeoTIFF and NetCDF |
| Atmospheric noise assessment | Variogram analysis of residual interferogram phase after ERA5 correction; identification of epochs with anomalous tropospheric contamination | QA report flagging unreliable epochs; corrected vs uncorrected velocity comparison figure |
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.