Volcanic edifice deformation monitoring with InSAR
Time-series InSAR turns repeated C- and L-band radar passes into surface displacement maps that reveal magma intrusion, chamber volume change, and flank creep weeks to months before a volcanic crisis becomes visible.
Sensors
- Sentinel-1 IW (C-band, ESA/Copernicus): 5.6 cm wavelength, 5 × 20 m ground range resolution in Interferometric Wide Swath mode, 250 km swath, 6-day repeat at the equator (12-day for a single satellite). Free and open archive from 2014. The operational workhorse for volcano monitoring globally, though C-band coherence degrades rapidly over dense tropical vegetation and on steep edifice flanks.
- ALOS-2 PALSAR-2 (L-band, JAXA): 23.6 cm wavelength penetrates vegetation canopy far better than C-band, preserving coherence on forested flanks where Sentinel-1 loses it. Stripmap mode: 3 m resolution; ScanSAR: up to 490 km swath at 60 m. Nominal 14-day repeat. Data access is licensed rather than open, but JAXA releases emergency observations through the Sentinel Asia and International Charter frameworks.
- RADARSAT Constellation Mission (C-band, CSA): Three satellites in the same orbital plane give a 4-day exact repeat and flexible incidence angles from 19° to 49°. Multiple look geometries from the same constellation help resolve the 1-D line-of-sight ambiguity that complicates single-track InSAR. Resolution ranges from 3 m (spotlight) to 100 m (ScanSAR). Commercially licensed.
- ICEYE X-band SAR constellation: Sub-metre spotlight resolution (0.5 × 0.25 m) with tasking latency of hours rather than days. Useful for detecting localised surface fractures, lava flow margins, or flank scarps that are below the resolution of operational C-band systems. Coherence windows are short at X-band over vegetated terrain, so ICEYE is best used for change detection between closely spaced acquisitions rather than long time-series.
- Sentinel-1 Persistent Scatterer and SBAS networks (derived product layer): Not a sensor, but the analytical layer that matters most operationally. Small Baseline Subset (SBAS) and Persistent Scatterer Interferometry (PSI) networks built from Sentinel-1 stacks can resolve line-of-sight displacement rates to 1–2 mm/year on stable surfaces. On active volcanic edifices, uncertainty is typically 5–20 mm per interferogram depending on atmospheric path-delay correction quality.
What the radar is actually measuring, and why that matters
InSAR measures the change in two-way travel time of a microwave pulse between two passes of the same satellite. Convert that time change to phase and you get a displacement map in the satellite's line-of-sight direction, accurate to a fraction of the wavelength used. For Sentinel-1 at C-band, that is roughly 2.8 cm per full phase cycle. Stack enough interferograms and time-series algorithms can resolve cumulative displacements of a few millimetres over months.
The physical signal of interest is ground uplift or subsidence caused by pressure change in a magma reservoir, a hydrothermal system, or a shallow intrusive body. Inflation typically produces a roughly concentric bulls-eye pattern of uplift centred above the pressure source. Deflation reverses it. Flank instability adds an asymmetric, downslope component. The pattern shape, magnitude, and spatial extent are what source models use to infer depth and volume.
From fringes to physics: Mogi, Okada, and their limits
The Mogi point-source model treats the magma reservoir as a pressurised sphere in a homogeneous elastic half-space. Given an observed surface displacement field, you can invert for four parameters: source depth, horizontal location, and volume change. It is computationally simple and works surprisingly well for reservoirs that are small relative to their depth. Published studies at Okmok, Uturuncu, and Campi Flegrei have used Mogi inversions to track volume changes of order 10 to 100 million cubic metres.
The Okada model is more general: it describes surface displacement from a rectangular dislocation in an elastic half-space, making it appropriate for dyke intrusions, sill emplacement, or fault-controlled flank movement. Both models assume a flat, uniform elastic medium, which volcanic edifices emphatically are not. Topographic stress concentration, heterogeneous rock properties, and hydrothermal alteration all introduce errors that are hard to quantify without independent constraints. Treat model outputs as hypotheses to test against seismic and geodetic ground truth, not as engineering specifications.
Coherence loss: the problem nobody in a brochure will mention
InSAR requires the ground surface to scatter radar energy in a statistically consistent way between two passes. Anything that changes the scatterer geometry, vegetation growth, rainfall, lava emplacement, or even wind-blown ash, decorrelates the phase and turns the interferogram into noise. On a densely forested tropical stratovolcano, C-band coherence can collapse entirely beyond a few weeks. L-band from ALOS-2 helps because the longer wavelength penetrates the canopy and reflects from the more stable ground beneath, but it is not a complete solution on steep, wet flanks.
Steep topography compounds the problem in two ways. First, foreshortening and layover in radar geometry mean that the upper flanks and crater walls are geometrically distorted or invisible. Second, the atmospheric path-delay correction, which is the largest single error source in volcanic InSAR, is poorly constrained over high-relief terrain where water vapour varies rapidly with altitude. Empirical corrections using ERA5 reanalysis or GACOS (Generic Atmospheric Correction Online Service) reduce the error but rarely eliminate it. Expect residual atmospheric noise of 1–3 cm in individual interferograms over high edifices, which is comparable to the deformation signal at quiescent volcanoes.
Operational monitoring: what the COMET Volcano Portal delivers
The Centre for Observation and Modelling of Earthquakes, Volcanoes and Tectonics (COMET), based at the Universities of Oxford and Leeds, runs an automated processing chain called LiCSAR that generates Sentinel-1 interferograms for over 900 volcanoes globally. Products are publicly available through the COMET Volcano Portal. Interferograms are typically produced within a few days of each Sentinel-1 acquisition, giving civil protection agencies a near-operational monitoring capability at no cost for the data or processing.
The limitation is that LiCSAR produces interferograms, not interpreted alerts. Translating a fringe pattern into a statement about eruption probability requires volcanological judgement. Automated change-detection algorithms exist and are improving, but false positives from atmospheric artefacts remain a real operational problem. Several national volcano observatories now combine COMET products with their own seismic and GPS networks to provide the multi-parameter context needed for credible hazard assessments.
Where commercial tasking adds something open data cannot
Sentinel-1's 6-day repeat is adequate for slowly evolving unrest but too slow to track a rapidly inflating edifice in the days before an eruption. RADARSAT Constellation's 4-day repeat and flexible incidence angles help, and ICEYE can be tasked within hours of an escalating alert. The value of commercial SAR in a volcanic crisis is not higher resolution for its own sake but acquisition on demand, at a specific incidence angle, on a specific date. That flexibility matters when you need to capture a displacement increment before the next lava flow buries the reference surface.
Satellize integrates open-constellation time-series with commercial tasking on client licence, applying SBAS processing and source modelling to produce displacement rate maps and volume-change estimates. The analytics workflow is the same one used in our crop-estimation programme for the Kingdom of Tonga, adapted to a very different geophysical signal. For a volcano observatory or a national civil protection agency, the practical deliverable is a GIS layer updated on each satellite pass, with flagged anomalies and a model-derived depth estimate when the signal is strong enough to invert reliably.
What to expect, and what to plan for when it fails
InSAR is most reliable at volcanoes with low vegetation cover, low rainfall, and moderate topographic relief. Dry stratovolcanoes in the Andes, the Aleutians, and the East African Rift are well-served. Tropical island volcanoes, including much of Indonesia, the Philippines, and the Caribbean arc, are harder. There, L-band is not optional; it is the only viable approach for flank-scale deformation, and even then, coherence is intermittent.
Plan for gaps. A major eruption will bury reference scatterers under lava or ash, resetting the time-series. A long rainy season can produce weeks of unusable interferograms. The monitoring strategy should therefore include GPS benchmarks and tiltmeters as independent ground truth, with InSAR providing the spatial context that point sensors cannot. The combination is more informative than either alone, and that is the honest case for satellite deformation monitoring: not a replacement for ground networks, but a way to see the whole edifice at once.
Typical figures
| Typical spatial resolution (Sentinel-1 IW) | 5 × 20 m (ground range × azimuth); multilooked to ~40–80 m for interferometric products |
| Typical spatial resolution (ALOS-2 PALSAR-2 Stripmap) | 3 m single-look; typically multilooked to 10–30 m for coherent processing |
| Repeat interval | 6 days (Sentinel-1, two satellites); 14 days (ALOS-2); 4 days (RADARSAT Constellation); hours to days (ICEYE, tasked) |
| Line-of-sight displacement sensitivity | 1–5 mm per interferogram on stable, coherent surfaces; 5–20 mm uncertainty over vegetated volcanic terrain after atmospheric correction |
| Minimum detectable volume change (Mogi inversion) | Approximately 10 million m³ for a source at 5 km depth, depending on edifice coherence and atmospheric noise |
| Radar frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1, RCM); L-band 1.236 GHz / 23.6 cm (ALOS-2); X-band ~9.65 GHz / 3.1 cm (ICEYE) |
| Archive depth | Sentinel-1: from October 2014; ALOS-2: from 2014; ALOS-1 PALSAR extends L-band archive to 2006 |
| Latency (operational products) | COMET LiCSAR interferograms typically available 2–5 days after Sentinel-1 acquisition; commercial tasking products 12–48 hours after acquisition |
| Coverage | Sentinel-1 IW: 250 km swath; global coverage of land at approximately 12-day intervals per track |
| Delivery formats | GeoTIFF displacement rasters, GeoJSON anomaly polygons, NetCDF time-series stacks, PDF summary reports |
Analytics Satellize can run
| Wrapped and unwrapped interferogram stack | Two-pass differential InSAR with DEM subtraction (SRTM or Copernicus DEM 30 m); GACOS or ERA5 atmospheric correction | GeoTIFF interferogram series per satellite track, updated each acquisition cycle |
| Line-of-sight displacement rate map | Small Baseline Subset (SBAS) time-series inversion of coherent interferogram network | Annual velocity GeoTIFF with per-pixel uncertainty estimate; refreshed monthly or on request |
| Mogi point-source inversion | Non-linear least-squares inversion of unwrapped displacement field against Mogi elastic half-space model | PDF report stating best-fit source depth (km), horizontal location (decimal degrees), and volume change (m³) with 95% confidence bounds |
| Okada dislocation model fit | Rectangular dislocation inversion (Okada 1985 formulation) for dyke or sill geometry | GIS layer showing modelled dislocation plane geometry; PDF report with opening/slip magnitude and uncertainty |
| Coherence anomaly alert | Automated coherence-change detection against rolling 90-day baseline; threshold exceedance flagging | Near-real-time alert (email or API push) with GeoJSON polygon of anomalous zone and candidate cause classification |
| Flank instability time-series | SBAS decomposition isolating downslope displacement component using ascending and descending track combination | Monthly GeoTIFF of downslope velocity; time-series CSV for selected benchmark pixels |
| Multi-sensor deformation composite | Fusion of Sentinel-1 (C-band) and ALOS-2 (L-band) displacement fields to extend coherent coverage onto vegetated flanks | Merged displacement GeoTIFF with sensor provenance mask; delivered after each ALOS-2 acquisition |
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.