Induced seismicity risk from wastewater injection and InSAR precursors
Deep disposal of oilfield wastewater raises pore pressure on pre-existing faults, triggering earthquakes. Sentinel-1 InSAR time-series can detect millimetre-scale surface deformation that precedes or accompanies these events, constraining fault geometry and causal injection volumes.
Sensors
- Sentinel-1 A/B (C-band SAR, ESA): Primary workhorse for InSAR time-series. Ground range resolution approximately 5 × 20 m in Interferometric Wide Swath mode; 250 km swath; 6-day repeat per satellite, 12-day with a single satellite. Phase sensitivity to line-of-sight displacement is theoretically sub-centimetre but practically limited to roughly 5–10 mm over agricultural or scrubland surfaces by decorrelation and atmospheric noise.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): L-band (1.27 GHz) penetrates vegetation and dry soil better than C-band, improving coherence over the vegetated margins of Oklahoma and Texas injection fields. Stripmap mode offers 3 m resolution; ScanSAR covers 350 km at 100 m. Revisit 14 days. Useful for detecting deeper or broader deformation lobes where C-band decorrelates.
- COSMO-SkyMed (X-band SAR, ASI): X-band at 3 cm wavelength and spotlight modes down to 1 m resolution give fine spatial detail on localised surface ruptures or fault scarps associated with M > 4 induced events. Revisit can be tasked to sub-daily in emergency mode. Less suited to wide-area time-series due to cost and narrower swath.
- USGS ShakeMap and National Earthquake Information Center catalogue: Provides ground-truth seismic event locations, depths, and magnitudes. Publicly accessible catalogue records allow injection-volume records from EPA Underground Injection Control (UIC) databases to be correlated spatially and temporally with InSAR-detected deformation, closing the causal chain.
- EPA Underground Injection Control well records: Publicly reported monthly injection volumes and wellhead pressures for Class II disposal wells. These are the independent variable in pore-pressure diffusion modelling; InSAR displacement is the dependent observable. Without this pairing, InSAR deformation alone cannot distinguish injection-induced from natural or extraction-induced signals.
Why disposing of water underground moves the ground above it
Oilfield produced water, which can exceed ten barrels for every barrel of oil recovered in mature basins, is typically re-injected into deep disposal formations under pressure. The injected fluid diffuses outward through porous rock, raising pore pressure along the path. When that pressure front reaches a pre-existing fault that is already close to its frictional failure threshold, the effective normal stress across the fault drops, and slip becomes possible at stress levels that would otherwise be safe. This is the mechanism behind the well-documented seismicity surge in Oklahoma between roughly 2009 and 2016, when disposal volumes rose sharply and the state briefly recorded more M ≥ 3 earthquakes per year than California.
The physics is described by pore-pressure diffusion, where the pressure front advances roughly as the square root of time multiplied by hydraulic diffusivity. Typical crystalline basement diffusivities of 0.1 to 1 m² per second mean a pressure perturbation can reach a fault 5 km from the injection point within weeks to months. This time lag between injection and seismicity is both a diagnostic signature and, in principle, a window for intervention.
What InSAR actually sees, and what it misses
Interferometric SAR measures the change in radar travel time between two satellite passes, converting phase difference to line-of-sight displacement with a theoretical precision of a few millimetres per interferogram. In practice, atmospheric water vapour introduces path-delay noise of 5–20 mm per scene, which time-series methods such as Persistent Scatterer InSAR (PS-InSAR) or Small Baseline Subset (SBAS) processing reduce by stacking many interferograms. Over the flat, dry terrain of the Permian Basin or central Oklahoma, coherence is generally good and annual displacement rates below 10 mm can be resolved with confidence.
The critical limitation for induced seismicity is depth. Most disposal wells inject into formations 1.5–4 km deep, and the induced earthquakes nucleate at comparable or greater depths. Surface deformation from a small to moderate induced earthquake (M 3 to 5) at 5 km depth is spread over a broad, low-amplitude lobe. A published analysis of the 2016 M 5.8 Pawnee, Oklahoma earthquake detected coseismic surface displacement of a few centimetres using Sentinel-1, sufficient to constrain fault strike and dip. But aseismic precursory slip at depth, if it occurs at all before induced events, would produce surface signals well below 5 mm and may be indistinguishable from atmospheric artefacts without long time-series averaging. Claiming routine precursor detection for small induced events would overstate what the current public record supports.
Correlating injection records with deformation time-series
The analytical value of InSAR for induced seismicity is strongest when displacement time-series are compared directly against EPA UIC injection-volume records for nearby Class II wells. A number of published studies, including work on the Permian Basin and the SCOOP and STACK plays in Oklahoma, have shown that cumulative surface subsidence or uplift patterns correlate spatially with high-volume disposal clusters, and that the onset of detectable deformation lags injection start by months consistent with diffusion timescales.
Fault geometry is a second output. Where coseismic fringes are clear enough, elastic dislocation modelling (Okada-type forward models) fitted to the InSAR displacement field constrains fault strike, dip, slip magnitude and depth. This is directly useful to operators seeking to understand which mapped or unmapped faults are activated, and to regulators setting traffic-light protocol thresholds. The honest caveat is that for events below roughly M 4, the surface signal is often too weak and spatially broad to constrain fault geometry independently of seismic catalogue data.
Resolution floors and the geometry problem for shallow events
Induced earthquakes in Oklahoma and Texas are often shallow, with hypocentral depths of 3–8 km. Shallower events produce more concentrated surface deformation, which is easier to detect but also more likely to be masked by local land-use change, vegetation growth or infrastructure movement. Sentinel-1's 5 × 20 m resolution in IW mode is sufficient to map deformation lobes from M ≥ 4.5 events at these depths, but the lobe diameter for a M 3.5 event at 5 km depth is only a few kilometres and the peak displacement may be 1–3 mm, sitting at or below the atmospheric noise floor of a single interferogram.
X-band sensors such as COSMO-SkyMed improve spatial resolution but worsen temporal coherence over agricultural land because the shorter wavelength is more sensitive to surface change between passes. L-band ALOS-2 offers the best coherence on vegetated sites but has a 14-day revisit and limited tasking flexibility. No single sensor solves all constraints simultaneously. A practical monitoring programme combines Sentinel-1 time-series for wide-area trend detection with targeted ALOS-2 acquisitions over high-risk injection clusters and COSMO-SkyMed tasking after any M ≥ 4 event.
What a monitoring programme looks like in practice
A credible InSAR-based induced seismicity monitoring programme starts with a baseline displacement velocity map derived from at least 18–24 months of Sentinel-1 acquisitions processed through SBAS or PS-InSAR. This map is overlaid with active Class II disposal well locations, reported monthly volumes and the USGS seismic catalogue. Spatial clusters where deformation rate exceeds a site-specific threshold (typically 5–15 mm per year, depending on atmospheric noise level) and where injection volumes are high flag priority areas for closer scrutiny.
Satellize applies this workflow operationally, running open-constellation InSAR analytics on client-defined areas of interest and integrating public well records. The approach is methodologically similar to published academic studies on Oklahoma and the Permian Basin, not a proprietary invention. Updates are delivered as GIS-ready displacement velocity grids, time-series plots per well cluster, and a written interpretation note flagging any anomalies against the seismic catalogue. The honest limitation is that this workflow identifies correlation and constrains plausible causal geometry; it does not predict individual earthquakes or their timing.
Typical figures
| Primary sensor spatial resolution | Sentinel-1 IW mode: ~5 × 20 m range × azimuth; ALOS-2 Stripmap: ~3 m; COSMO-SkyMed Spotlight: ~1 m |
| Swath width | Sentinel-1 IW: 250 km; ALOS-2 ScanSAR: up to 350 km; COSMO-SkyMed Spotlight: ~10 km |
| Revisit interval | Sentinel-1: 6 days (two satellites); ALOS-2: 14 days; COSMO-SkyMed: tasked, potentially sub-daily in emergency mode |
| Line-of-sight displacement sensitivity (time-series) | Practically ~5–10 mm per year over coherent surfaces after atmospheric correction; single-interferogram noise floor ~5–20 mm |
| Minimum detectable coseismic event (surface signal) | Approximately M 4.0–4.5 at 5 km depth for Sentinel-1; smaller events produce signals below atmospheric noise floor of individual interferograms |
| Radar frequency / wavelength | Sentinel-1: C-band, 5.6 cm; ALOS-2: L-band, 23.6 cm; COSMO-SkyMed: X-band, 3.1 cm |
| Archive depth | Sentinel-1: from 2014 (Sentinel-1A launch); ALOS-2: from 2014; ALOS-1 PALSAR extends L-band archive to 2006 |
| Latency (open data) | Sentinel-1 NRT products available within ~1–3 hours of acquisition via Copernicus Data Space; processed InSAR products typically 24–72 hours depending on pipeline |
| Injection well data source | EPA Underground Injection Control (UIC) Class II well records; monthly volumes publicly reported; spatial coordinates available via EPA ECHO database |
| Delivery formats | GeoTIFF displacement velocity grids, NetCDF time-series stacks, Shapefile or GeoJSON fault-trace overlays, PDF interpretation reports |
Analytics Satellize can run
| Baseline displacement velocity map | SBAS or PS-InSAR time-series processing of Sentinel-1 IW SLC stack; atmospheric correction via ERA5 or Generic Atmospheric Correction Online Service (GACOS) | GeoTIFF mean line-of-sight velocity grid (mm/year) covering client area of interest, updated quarterly or on request |
| Injection-cluster deformation correlation report | Spatial join of displacement velocity anomalies with EPA UIC Class II well locations and reported monthly volumes; Pearson or Spearman correlation of volume time-series against displacement rate | PDF report with ranked well clusters by deformation-injection correlation coefficient, flagging statistically significant associations |
| Coseismic displacement map and fault-geometry estimate | Pre- and post-event interferogram differencing; Okada elastic dislocation forward-model inversion fitted to observed fringe pattern to estimate fault strike, dip, rake and slip | GeoTIFF wrapped and unwrapped interferogram, GeoJSON fault-plane solution, PDF model summary with uncertainty bounds |
| Pore-pressure diffusion front estimate | Simplified radial diffusion model (pressure front distance proportional to square root of hydraulic diffusivity × time) calibrated against injection start date and seismicity onset from USGS catalogue | GIS polygon showing estimated pressure-front radius at monthly intervals; tabular comparison with observed seismicity locations |
| Traffic-light protocol input layer | Threshold-based alert system cross-referencing real-time USGS ShakeMap magnitudes against InSAR-derived deformation anomaly zones and injection volumes | Automated alert feed (GeoJSON or email) triggered when seismic magnitude within a deformation anomaly zone exceeds operator-defined threshold |
| Multi-sensor coherence and deformation quality assessment | Comparison of Sentinel-1 C-band, ALOS-2 L-band and (where available) COSMO-SkyMed X-band coherence maps over the area of interest to identify optimal sensor for site conditions | One-page sensor-selection memo with coherence statistics per land-cover class, informing tasking decisions for follow-on monitoring |
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.