Slow slip event detection on subduction zones

Slow slip events release tectonic stress aseismically over days to months, producing surface displacements too small for seismometers but resolvable by InSAR time-series and GNSS. This page covers the physics, the processing chains, and the public datasets that have documented SSEs on Cascadia, Hikurangi and Guerrero.

Sensors

Sentinel-1 A/B (ESA): C-band SAR at 5.6 cm wavelength; Interferometric Wide Swath mode gives 250 km swath at 5 × 20 m resolution (range × azimuth). Repeat cycle 6–12 days depending on orbital geometry and hemisphere. Phase sensitivity to line-of-sight displacement is roughly 2.8 mm per half-wavelength fringe, making it the primary open-access source for SSE surface signals.
ALOS-2 PALSAR-2 (JAXA): L-band SAR at 23.6 cm wavelength. Longer wavelength reduces atmospheric phase delay sensitivity relative to C-band and penetrates vegetation canopy, improving coherence in forested subduction zones such as Cascadia and Hikurangi. ScanSAR mode covers 350 km swath; stripmap mode achieves 3 m resolution. Repeat cycle 14 days, though tasking is not open-access.
COMET LiCSAR (NERC/COMET): Automated Sentinel-1 interferogram processing chain covering all tectonic plate boundaries globally. Produces geocoded, multi-looked interferograms and coherence maps archived back to 2014. Removes the need for raw SAR processing in most SSE studies; outputs are publicly browsable and downloadable.
GNSS networks (GEONET, PBO/NOTA): GEONET (GNS Science, New Zealand) and PBO/NOTA (UNAVCO, western North America) provide continuous 3-D displacement time-series at individual stations with sub-millimetre vertical precision after post-processing. Used to validate InSAR line-of-sight projections and to constrain SSE depth and slip magnitude through joint inversion. Station spacing of 20–80 km limits spatial resolution in sparsely instrumented regions.
Sentinel-1 GACOS atmospheric correction (generic): Generic Atmospheric Correction Online Service for InSAR (GACOS) uses ECMWF ERA5 reanalysis and GNSS zenith total delay estimates to model tropospheric delay. Reduces atmospheric artefacts from roughly 5–10 cm RMS to 1–2 cm RMS in favourable conditions, critical for separating SSE signals of comparable magnitude.

What a slow slip event actually does to the ground

Subduction zones accumulate elastic strain as the overriding plate is dragged down with the descending slab. Most of that strain releases catastrophically in megathrust earthquakes. Some of it does not. Slow slip events occur when a patch of the fault interface slips aseismically, typically at depths of 25–45 km on the down-dip edge of the locked zone, over periods of days to months. The overriding plate surface moves seaward and subsides by millimetres to low centimetres, the opposite direction to interseismic loading.

The physical setting matters for remote sensing. Cascadia SSEs recur roughly every 12–15 months and produce surface displacements of 2–6 mm over 10–20 days. Hikurangi (North Island, New Zealand) hosts both deep SSEs with similar amplitudes and shallow SSEs offshore that can reach 15–20 mm of surface uplift. The Guerrero seismic gap in Mexico produces some of the largest documented SSEs, with displacements exceeding 40 mm over several months. These differences in amplitude and duration determine which sensor combination is appropriate.

Why InSAR struggles here, and what makes it work anyway

The fundamental problem is scale. An SSE signal of 5 mm distributed across a 200 km aperture sits comfortably within the tropospheric noise budget of a single interferogram, which can reach 5–10 cm in mountainous coastal terrain. A single interferogram cannot distinguish the two. Time-series methods change the arithmetic.

Persistent Scatterer InSAR (PS-InSAR) and Small Baseline Subset (SBAS) analysis stack tens to hundreds of interferograms and separate temporally coherent deformation from spatially correlated atmospheric noise. The SBAS approach, applied to Sentinel-1 data over Hikurangi by multiple published groups, has resolved SSE surface displacements of 10–15 mm with spatial detail that GNSS networks at 30–50 km station spacing cannot provide. GACOS atmospheric correction applied before stacking reduces residual tropospheric noise further. Even so, shallow SSEs in wet, forested terrain remain genuinely difficult: coherence drops in L-band over short baselines, and C-band decorrelates faster in dense canopy. Honest assessment: InSAR alone cannot reliably detect SSEs below roughly 5 mm surface displacement in high-noise environments. Joint inversion with GNSS is not optional in those cases.

The processing chain for a usable SSE product typically runs: coregistered SLC pairs, multi-look interferogram formation, phase unwrapping (SNAPHU or similar), atmospheric correction, time-series inversion (π-RATE, MintPy or equivalent), and finally geodetic modelling to estimate fault slip distribution. Each step introduces assumptions. Phase unwrapping errors in low-coherence zones can mimic or mask SSE signals, and users should inspect coherence masks before drawing conclusions.

The public record: what Cascadia, Hikurangi and Guerrero have taught us

Cascadia was the proving ground. GNSS data from the Pacific Northwest Geodetic Array documented recurring SSEs from the early 2000s, and subsequent Sentinel-1 SBAS analysis confirmed surface displacement fields consistent with slip on the plate interface at 25–40 km depth. The spatial extent of individual SSEs, typically 100–300 km along-strike, is resolvable from Sentinel-1 interferogram sequences but requires careful masking of the Olympic Peninsula's dense forest cover where coherence is low.

Hikurangi has become arguably the best-instrumented subduction SSE laboratory on Earth, partly because shallow offshore SSEs produce unusually large onshore signals. Published ALOS-2 and Sentinel-1 studies have imaged individual events with slip patches resolved at roughly 10–20 km spatial scale. The GEONET GNSS network provides dense onshore constraint. Offshore, ocean-bottom pressure gauges and seafloor geodesy add data that InSAR cannot supply, a reminder that satellite geodesy has a hard coastline boundary.

Guerrero's large, slow events (sometimes called 'silent earthquakes') were documented primarily by GNSS before Sentinel-1 became available. Subsequent InSAR time-series analysis has added spatial detail to slip models, showing that the Guerrero gap does not slip uniformly but in patches. The long duration of Guerrero SSEs (months rather than days) actually helps InSAR: longer accumulation time means larger surface signal, and atmospheric noise averages down over many acquisitions.

Depth ambiguity and the limits of surface data

InSAR measures surface displacement in the satellite line-of-sight direction. Converting that to a fault slip model requires assumptions about fault geometry, elastic structure of the crust, and the depth of the slipping patch. For subduction zones with well-characterised geometry (Hikurangi's slab geometry is constrained by seismic tomography to within a few kilometres), the inversion is reasonably well-posed. For less-studied zones, the depth ambiguity between a small shallow slip patch and a larger deep one can be substantial.

Ascending and descending orbit combinations help decompose line-of-sight displacement into approximate horizontal and vertical components, which constrains dip-slip versus strike-slip partitioning. ALOS-2's left-looking capability adds a third geometry on some targets. Even with multiple geometries, the near-vertical sensitivity of InSAR means that horizontal displacements from deep SSEs are better constrained than vertical ones, which is the opposite of what GNSS delivers. The two methods are genuinely complementary rather than redundant.

Building an operational SSE monitoring product

Operational SSE detection differs from academic event studies in one important respect: latency. Sentinel-1 data typically reaches the Copernicus Open Access Hub within 24 hours of acquisition. A near-real-time SBAS pipeline using rolling 6–12 day interferogram pairs can flag anomalous line-of-sight velocity changes within 2–3 days of the acquisition that captures them. The challenge is distinguishing a genuine SSE onset from a single-acquisition atmospheric artefact. Minimum viable confirmation requires at least two consecutive acquisitions showing consistent spatial pattern and sign.

Satellize structures SSE analytics as a time-series velocity monitoring service over named subduction segments, using LiCSAR interferograms as the primary input, GACOS-corrected where tropospheric noise is elevated, with GNSS time-series ingested for joint validation. The approach is the same one applied to the Tonga crop-estimation programme in a different physical domain: open-constellation data, rigorous processing, and explicit uncertainty quantification rather than a single authoritative number.

For government clients with civil defence mandates, the relevant output is not a slip model. It is a time-stamped alert that a surface displacement anomaly consistent with SSE activity has begun, with a spatial footprint and an uncertainty envelope, delivered before the event has finished. That framing shapes the processing choices: speed and false-alarm rate matter more than centimetre-scale slip resolution. Speak to the analytics team about the specific subduction segment and the GNSS network density available.

Typical figures

Primary SAR wavelength (Sentinel-1)	C-band, 5.6 cm; line-of-sight phase sensitivity ~2.8 mm per fringe
Primary SAR wavelength (ALOS-2)	L-band, 23.6 cm; better coherence in vegetated terrain
Sentinel-1 spatial resolution (IW mode)	5 m range × 20 m azimuth; typically multi-looked to ~80–100 m for time-series
Sentinel-1 repeat cycle	6 days (A+B combined over Europe/priority zones); 12 days elsewhere
ALOS-2 repeat cycle	14 days; tasking not open-access
Minimum detectable surface displacement (SSE)	~5 mm in time-series after atmospheric correction; higher (~10–15 mm) in wet forested terrain
Atmospheric noise (single interferogram, coastal mountains)	5–10 cm RMS; reduced to 1–2 cm RMS with GACOS correction
Sentinel-1 archive depth	From April 2014 (Sentinel-1A launch); LiCSAR processed interferograms available for tectonic zones
Alert latency (near-real-time pipeline)	2–3 days after SAR acquisition, requiring two consecutive confirming passes
Deliverable formats	GeoTIFF displacement maps, NetCDF time-series, GIS-ready velocity grids, PDF event reports

Analytics Satellize can run

InSAR displacement time-series over named subduction segment	SBAS or PS-InSAR stacking of Sentinel-1 IW interferograms; GACOS atmospheric correction; MintPy or π-RATE inversion	Monthly-updated GeoTIFF velocity maps and NetCDF time-series per pixel, covering the defined segment
SSE onset alert	Anomaly detection on rolling 6–12 day interferogram pairs; spatial pattern matching against expected SSE footprint for the target zone	Time-stamped alert with spatial footprint polygon, estimated displacement magnitude and uncertainty, delivered within 3 days of confirming acquisition
Joint InSAR/GNSS slip model	Elastic half-space inversion (Okada or layered-Earth equivalent) combining LiCSAR line-of-sight displacements and GEONET/PBO GNSS 3-D vectors	Fault slip distribution map (GeoTIFF + PDF) with depth, along-strike extent and equivalent moment magnitude
Atmospheric noise characterisation report	GACOS ERA5-based tropospheric delay modelling; comparison of pre- and post-correction interferogram RMS over stable reference pixels	Per-acquisition noise budget table, flagging acquisitions where atmospheric artefacts exceed the expected SSE signal; included in each monthly report
Historical SSE catalogue for a subduction segment	Retrospective SBAS analysis of full Sentinel-1 archive (2014–present) over the target zone; event identification by temporal velocity anomaly thresholding	Event catalogue (CSV + GIS layer) listing date, duration, peak displacement, spatial footprint and estimated equivalent slip for each detected episode
Coherence and data-quality assessment	Per-interferogram coherence mapping; identification of decorrelation zones (dense vegetation, water, snow) that limit detection capability	Coverage confidence map showing where SSE signals below a user-specified threshold cannot be reliably detected; informs GNSS densification priorities

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.