Postseismic viscoelastic relaxation and afterslip mapping
After large earthquakes, the crust deforms for months to years through afterslip and viscoelastic flow. InSAR time-series from Sentinel-1 and ALOS-2 separate these mechanisms and constrain the rheological models that underpin long-term seismic hazard assessment.
Sensors
- Sentinel-1 A/B (C-band SAR, ESA): 5.6 cm wavelength; Interferometric Wide Swath mode gives 250 km swath at 5 x 20 m resolution. Repeat pass every 6 days (12 days per satellite after Sentinel-1B failure in 2021). Sensitive to centimetre-scale line-of-sight displacement; phase coherence maintained over months in arid and semi-arid terrain.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength; penetrates vegetation canopy and maintains coherence in forested terrain where C-band loses it. ScanSAR mode provides 350 km swath at 100 m; Stripmap at 3 m. 14-day repeat. Essential for postseismic studies in vegetated settings such as the Kaikōura region.
- ARIA GUNW (processed Sentinel-1 unwrapped interferograms, NASA/Caltech JPL): Geocoded Unwrapped interferograms derived from Sentinel-1, provided as standardised NetCDF products. Removes the preprocessing burden; enables rapid time-series construction. Spatial posting 90 m. Archive extends to 2014.
- GNSS campaign and continuous networks: Three-dimensional surface velocity vectors at benchmark precision (sub-millimetre per year with long time-series). Used to resolve the vertical ambiguity inherent in single-geometry InSAR and to tie InSAR time-series to an absolute reference frame. Essential for separating afterslip from viscoelastic relaxation at depth.
- Sentinel-1 burst overlap intensity tracking: Along-track displacement component derived from azimuth pixel offsets at burst boundaries. Adds a second look geometry to constrain the full 3D displacement field without requiring a second satellite geometry, at roughly 40 m azimuth resolution.
Why the earthquake is not the end of the story
A magnitude 7.8 or larger rupture releases elastic strain accumulated over centuries, but it also loads the surrounding crust in ways that take years to equilibrate. Two distinct physical processes drive postseismic deformation. Afterslip is continued aseismic sliding on the fault plane, concentrated in the velocity-strengthening regions above and below the seismogenic zone. Viscoelastic relaxation is the slow, distributed flow of the lower crust and upper mantle as they adjust to the sudden stress change. Both produce surface displacement, but at different spatial wavelengths and decay timescales.
Distinguishing the two matters for hazard. Afterslip redistributes stress onto adjacent locked patches; viscoelastic relaxation can promote seismicity on nearby faults for decades. The 2015 Gorkha earthquake (Mw 7.8) produced measurable postseismic signal for at least three years, documented in multiple published InSAR studies using Sentinel-1. The 2016 Kaikōura earthquake (Mw 7.8) involved at least 21 fault segments and generated a complex postseismic field that required both Sentinel-1 and ALOS-2 to characterise, because C-band coherence failed in the forested Kaikōura ranges where L-band succeeded.
What the interferogram time-series actually measures
InSAR measures the change in range between the satellite and the ground along the radar line of sight. A single interferogram captures cumulative displacement between two acquisition dates. A time-series, built by stacking and inverting dozens of interferograms using methods such as Small Baseline Subset (SBAS) or the MintPy open-source package, resolves the temporal evolution of that displacement at each pixel. Typical postseismic signals range from a few centimetres in the first weeks after a large rupture to millimetres per year at the tail of the relaxation curve.
The fundamental ambiguity is geometric. A single ascending or descending pass constrains only the line-of-sight component. Combining ascending and descending geometries allows decomposition into approximate vertical and east-west horizontal motion, but north-south displacement remains poorly constrained by C-band because the satellite orbit is near-polar. ALOS-2's longer wavelength and wider azimuth bandwidth improve the along-track sensitivity. GNSS vectors, where available, resolve what InSAR cannot.
Atmospheric delay is the dominant noise source. Tropospheric water vapour introduces path-delay artefacts that can mimic or mask deformation signals at the centimetre level. Correction using ERA5 reanalysis fields or GACOS (Generic Atmospheric Correction Online Service) reduces this noise but does not eliminate it. In high-relief terrain, such as the Nepal Himalaya, residual topographic-correlated atmosphere remains a real limit on the minimum detectable signal.
Separating shallow afterslip from deep viscoelastic flow
The spatial pattern of surface deformation carries the rheological fingerprint. Afterslip on a shallow fault patch produces a displacement lobe that decays rapidly away from the fault trace, roughly inversely with distance. Viscoelastic relaxation in the lower crust or upper mantle produces a broader, longer-wavelength signal that can extend hundreds of kilometres from the rupture. In practice the two overlap, and separating them requires forward modelling against the time-series.
The standard workflow fits a kinematic or mechanical model to the observed displacement field. For afterslip, the fault geometry from the coseismic rupture model is held fixed and slip distribution is solved on the extended fault plane using constrained least-squares or Bayesian inversion. For viscoelastic relaxation, a layered Maxwell or Burgers rheology is assumed for the lower crust and mantle, and the viscosity is varied until the modelled surface velocity matches the observed time-series. Published viscosity estimates from the Gorkha postseismic studies range from roughly 10^18 to 10^19 Pa·s for the lower crust, depending on the assumed rheological model and the time window used.
The honest limit here is model non-uniqueness. Different combinations of afterslip distribution and viscosity structure can produce similar surface signals. GNSS data at multiple azimuths, and time-series extending several years, are the primary means of reducing that ambiguity. Short observation windows, or sparse GNSS networks, leave the solution underdetermined.
Coherence, coverage and the cases where InSAR fails
Coherence is the correlation of radar phase between two acquisitions. It degrades with vegetation growth, snow cover, heavy rainfall and long time intervals. C-band (Sentinel-1) loses coherence rapidly in tropical forests and in mountain snowfields; L-band (ALOS-2) is substantially more resilient. The Kaikōura case made this concrete: published studies found that Sentinel-1 coherence was insufficient over the forested ranges north of Kaikōura, and ALOS-2 data were required to map the full postseismic field.
Ionospheric delay is a second L-band-specific problem. At high latitudes or during geomagnetically active periods, dispersive ionospheric path delay introduces phase ramps that can be confused with tectonic signal. Split-spectrum correction methods exist but add processing complexity. For equatorial targets, the ionosphere is more variable and L-band data require careful correction.
Revisit rate sets the temporal resolution of the time-series. Sentinel-1's 6-day repeat (12-day since the 2021 Sentinel-1B failure) is adequate for capturing the rapid early postseismic phase. ALOS-2's 14-day repeat can miss the fastest transients. For the first days after a major rupture, the InSAR archive may have a gap simply because no acquisition was scheduled over the epicentral area.
From deformation map to hazard-relevant output
A postseismic InSAR time-series becomes operationally useful when it is translated into products that planners and engineers can act on. The primary analytic outputs are: a time-evolving displacement map at each pixel, a best-fit afterslip distribution on the fault plane, a constrained estimate of lower-crustal viscosity, and a Coulomb stress change map that shows how postseismic slip and relaxation have loaded adjacent fault segments.
Coulomb stress transfer is particularly relevant for aftershock hazard assessment. Published work after Gorkha showed that afterslip in the months following the mainshock loaded the updip portion of the Main Himalayan Thrust, a region of concern for future rupture. This is the kind of output that national geological surveys and civil protection agencies need, and it requires months of postseismic InSAR data, not just the coseismic interferogram.
Satellize can run SBAS time-series processing on ARIA GUNW products and Sentinel-1 SLC archives, integrating GNSS campaign data where clients supply it, and deliver fault-plane slip models and Coulomb stress layers as GIS-ready outputs. The workflow is the same open-science stack used in the published Gorkha and Kaikōura studies, applied operationally rather than academically.
What a programme actually needs to commission this work
The minimum viable input is a defined postseismic time window, a coseismic fault model (published or client-supplied), and a geographic bounding box for the area of interest. Sentinel-1 data back to 2014 are freely available through the Copernicus Data Space; ALOS-2 data require a JAXA research agreement or commercial licence. GNSS data from national networks, where available, dramatically improve the result and should be included in the scoping conversation.
Delivery timescales depend on the archive depth required. A 12-month postseismic time-series over a 200 x 200 km area typically involves 40 to 80 interferograms per geometry, processing time of days on cloud infrastructure, and QA review before delivery. Longer windows or larger areas scale accordingly. The output is not a single map but a stack of dated displacement rasters plus model parameters, which means the client needs GIS or Python capacity to use the full product. A summary report with interpreted Coulomb stress maps is the alternative for non-specialist audiences.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | 5 x 20 m (single look); typically multilooked to 40-80 m for time-series |
| Spatial resolution (ALOS-2 Stripmap) | 3-10 m; ScanSAR 100 m |
| Revisit period | 6 days (Sentinel-1 A+B when both operational); 12 days (Sentinel-1A only, post-2021); 14 days (ALOS-2) |
| Minimum detectable line-of-sight displacement | ~5 mm per interferogram in low-noise conditions; ~1-2 mm/yr in time-series after atmospheric correction |
| Radar frequency | C-band 5.405 GHz (Sentinel-1); L-band 1.2565 GHz (ALOS-2 PALSAR-2) |
| Swath width | 250 km (Sentinel-1 IW); 350 km (ALOS-2 ScanSAR) |
| Archive depth | Sentinel-1 from April 2014; ALOS-2 from 2014; ALOS-1 PALSAR legacy archive to 2011 |
| Atmospheric correction | ERA5 reanalysis or GACOS tropospheric delay models; residual error 1-3 cm in high-relief terrain |
| Delivery formats | GeoTIFF displacement rasters, NetCDF time-series stacks, Shapefile or GeoPackage fault-plane slip models, PDF interpreted report |
| GNSS integration | Client-supplied campaign or continuous GNSS vectors ingested for 3D decomposition and absolute reference frame tie |
Analytics Satellize can run
| Postseismic displacement time-series | SBAS or MintPy time-series inversion of Sentinel-1 or ALOS-2 interferogram stacks | Dated GeoTIFF rasters of cumulative line-of-sight displacement at each epoch, delivered as a NetCDF stack with metadata |
| Afterslip distribution on fault plane | Constrained least-squares or Bayesian inversion of surface displacement against fault geometry from coseismic model | GeoPackage of fault patches with slip magnitude and uncertainty; PDF figure for report |
| Lower-crustal viscosity estimate | Forward modelling of Maxwell or Burgers viscoelastic rheology fitted to time-series velocity decay | Rheological parameter table with uncertainty ranges; model-vs-observation comparison figure |
| Coulomb stress change map | Elastic dislocation modelling of afterslip and viscoelastic contribution projected onto receiver fault geometries | GeoTIFF Coulomb stress change layer; interpreted PDF showing stress loading on adjacent fault segments |
| Coherence and coverage quality report | Per-interferogram coherence statistics and temporal baseline analysis across the study window | PDF QA report flagging data gaps, low-coherence epochs and recommended supplementary sensors |
| 3D surface velocity decomposition | Ascending/descending geometry combination with optional GNSS constraint to resolve vertical and horizontal components | Separate GeoTIFF layers for vertical and east-west velocity; north-south component flagged as poorly constrained where GNSS absent |
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.