Geothermal field surface deformation and resource monitoring
InSAR tracks millimetre-scale uplift and subsidence at geothermal fields caused by injection, extraction, and thermally driven pressure changes. Published work at Reykjanes, Wairakei, and The Geysers shows how Sentinel-1 and TerraSAR-X time series reveal reservoir behaviour that wellhead sensors alone cannot capture.
Sensors
- Sentinel-1 (C-band SAR, ESA): 5.6 cm wavelength, IW mode ground range resolution approximately 5 x 20 m, 6-day repeat at mid-latitudes with both satellites active. Free archive from 2014 makes it the default choice for multi-year deformation time series at geothermal fields. Atmospheric noise is the dominant error source; tropospheric delay corrections are essential over volcanic terrain.
- TerraSAR-X (X-band SAR, DLR / Airbus): 3.1 cm wavelength, Stripmap mode at approximately 3 m resolution, Spotlight at approximately 1 m. Short wavelength gives higher phase sensitivity to small displacements but decorrelates faster over vegetated or hydrothermally altered ground. Revisit 11 days standard; tasked acquisitions can achieve shorter cycles. Published studies at Reykjanes and The Geysers used TerraSAR-X to resolve localised subsidence bowls near individual well clusters.
- COSMO-SkyMed (X-band SAR, ASI): Constellation of four satellites, Stripmap at approximately 3 m, Spotlight at approximately 1 m. Revisit potentially 1-4 days with constellation tasking. X-band coherence degrades rapidly on vegetated or wet ground, which limits its use at humid fields such as Wairakei; performs better on arid or lava-covered terrain.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength. Maintains coherence over vegetation and weathered ground where C- and X-band lose it. Approximately 3 m resolution in Fine mode, 14-day revisit. Particularly useful at geothermal fields in tropical or heavily vegetated settings where Sentinel-1 interferograms decorrelate. Longer wavelength reduces sensitivity per unit displacement but compensates by preserving phase signal over longer time gaps.
What the ground is actually telling you
Geothermal reservoirs deform their overlying rock in two main ways. Fluid extraction reduces pore pressure, causing compaction and surface subsidence. Reinjection of cooled water does the opposite, but imperfectly: the thermal contraction of rock around injection wells can produce localised subsidence even when net fluid volume is restored. The surface expression of these competing signals is a spatially complex pattern of uplift and drawdown that shifts as production schedules change.
At Wairakei in New Zealand, decades of production without reinjection caused cumulative subsidence exceeding 15 metres in the most affected area, well documented by levelling surveys. Satellite InSAR picks up the ongoing residual signal at rates of several centimetres per year, and can do so across the full field extent rather than at discrete benchmark locations. The spatial continuity matters: it reveals where the deformation front is migrating and whether new well clusters are opening fresh subsidence lobes.
Reykjanes and The Geysers: what published studies found
The Reykjanes peninsula in Iceland hosts both a high-temperature geothermal field and the plate boundary between the North American and Eurasian plates, which makes separating tectonic and reservoir signals a genuine analytical challenge. Published InSAR studies using Sentinel-1 and TerraSAR-X have identified inflation episodes linked to magmatic intrusion and deflation linked to production, sometimes superimposed on the same interferogram. Decomposing those signals requires time series long enough to distinguish episodic tectonic events from the smoother, production-correlated trend.
At The Geysers in California, the world's largest geothermal complex by installed capacity, published work using TerraSAR-X Persistent Scatterer InSAR (PSI) resolved subsidence rates of several centimetres per year near high-production zones and localised uplift near reinjection wells. The spatial offset between subsidence and uplift centres provided direct evidence of lateral fluid migration within the reservoir. That kind of spatial intelligence is not recoverable from wellhead pressure data alone.
X-band versus C-band: a genuine trade-off, not a ranking
X-band sensors such as TerraSAR-X and COSMO-SkyMed offer finer spatial resolution and higher phase sensitivity to small displacements. Near wellheads, where deformation gradients are steep and the features of interest are tens of metres across, that resolution advantage is real. A subsidence bowl centred on a single injection well may be entirely missed at Sentinel-1's effective resolution and smeared across several pixels even when detected.
The cost is coherence. X-band phase decorrelates quickly over hydrothermally altered ground, wet soil, or any surface that changes between acquisitions. Many geothermal fields sit in exactly those conditions. C-band Sentinel-1 typically sustains coherence over longer baselines and across a wider range of surface types, making it the more reliable choice for field-wide time series. The practical answer is usually to use both: Sentinel-1 for the spatial overview and long-term trend, TerraSAR-X for targeted high-resolution monitoring near critical infrastructure.
Separating atmosphere from ground: the time-series problem
A single interferogram over a geothermal field is almost uninterpretable without atmospheric correction. Tropospheric water vapour gradients over volcanic terrain can produce apparent range changes of several centimetres, comparable in magnitude to the genuine deformation signal. The standard mitigation is time-series analysis: SBAS (Small Baseline Subset) or PSI methods stack many interferograms and exploit the fact that atmospheric noise is largely uncorrelated in time while genuine deformation follows a physically plausible trend.
External atmospheric corrections using ERA5 reanalysis data or GACOS (Generic Atmospheric Correction Online Service) reduce residual tropospheric noise further, typically by 30 to 60 percent in published comparisons, though the improvement is scene-dependent. Ionospheric delay is a secondary concern at C- and X-band over mid-latitudes but becomes significant for L-band ALOS-2 data, particularly at high latitudes or during periods of elevated solar activity. Honest interpretation requires reporting the residual atmospheric uncertainty alongside the deformation estimate, not just the deformation itself.
What InSAR cannot do here
InSAR measures displacement in the satellite's line of sight, not in three-dimensional space. Separating vertical from horizontal motion requires at least two viewing geometries, ascending and descending passes, and even then the north-south component is poorly constrained by C- or X-band systems. At fields near active faults, unresolved horizontal motion can be misread as vertical deformation.
Temporal resolution is a genuine constraint. Sentinel-1's 6-day repeat means that rapid deformation events, such as the subsidence pulses that can follow induced seismicity, may be aliased or missed entirely between acquisitions. Dense time series from commercial constellations with shorter revisit can help, but at higher cost and with the coherence caveats already noted. Cloud cover does not affect SAR, which is one of its clearest advantages over optical geodetic methods in humid volcanic environments.
From interferogram to operational decision
The analytic chain from raw SAR to an operational decision for a geothermal operator has several steps: co-registration, interferogram generation, phase unwrapping, atmospheric correction, time-series inversion, and finally interpretation against the production log. Each step introduces uncertainty that compounds. A displacement map delivered without uncertainty bounds is not useful for engineering decisions.
Satellize structures geothermal deformation analytics as time-series GIS layers updated on each new satellite pass, with flagged exceedances when deformation rates cross operator-defined thresholds. The method class is standard PSI or SBAS applied to open Sentinel-1 data, supplemented by commercial tasking on client licence where X-band resolution is needed near wellheads. The Tonga crop-estimation programme is a different domain entirely, but the underlying pipeline logic, ingesting open constellation data, applying a validated processing chain, and delivering a decision-ready output, is the same.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | Approximately 5 x 20 m ground range; effective PSI point density depends on surface coherence, typically 50-500 points per km² over stable ground |
| Spatial resolution (TerraSAR-X Spotlight) | Approximately 1 m; enables detection of deformation gradients across individual wellhead pads |
| Minimum detectable displacement (line of sight) | Approximately 2-5 mm per year in PSI time series over coherent targets; single-interferogram noise floor typically 5-20 mm depending on atmospheric conditions |
| Revisit period | Sentinel-1: 6 days (dual satellite); TerraSAR-X: 11 days standard; COSMO-SkyMed: 1-4 days with constellation tasking; ALOS-2: 14 days |
| SAR frequency / wavelength | C-band (Sentinel-1): 5.6 cm; X-band (TerraSAR-X, COSMO-SkyMed): 3.1 cm; L-band (ALOS-2): 23.6 cm |
| Archive depth | Sentinel-1: from October 2014; ERS/Envisat C-band archive extends to 1992 for some sites; TerraSAR-X from 2007 |
| Latency (open data) | Sentinel-1 NRT products available within approximately 1 hour of acquisition; standard SLC products within 24 hours via Copernicus Data Space |
| Cloud penetration | SAR is unaffected by cloud cover; all-weather acquisition is a primary advantage over optical geodetic methods in humid volcanic settings |
| Atmospheric correction uncertainty | Residual tropospheric noise typically 3-15 mm per interferogram after ERA5 or GACOS correction; improvement 30-60% over uncorrected data in published comparisons |
| Delivery formats | GeoTIFF displacement rasters, GeoPackage or Shapefile PSI point clouds, CSV time-series tables, threshold-alert feeds |
Analytics Satellize can run
| Field-wide deformation velocity map | SBAS or PSI time-series inversion on Sentinel-1 SLC stack; atmospheric correction via ERA5 or GACOS | Annual-average line-of-sight velocity GeoTIFF with uncertainty layer; updated quarterly or on request |
| Vertical and horizontal displacement decomposition | Ascending and descending pass combination; least-squares decomposition assuming negligible north-south component | Separate vertical and east-west velocity GIS layers with propagated uncertainty bounds |
| Production-correlated deformation time series | PSI time series cross-referenced against operator-supplied injection/extraction logs; linear regression and residual analysis | Time-series plots per well cluster with annotated production events; PDF report and CSV export |
| High-resolution wellhead subsidence mapping | TerraSAR-X or COSMO-SkyMed Spotlight PSI; co-registration to Sentinel-1 reference frame for multi-sensor consistency | 1-3 m resolution displacement map around individual wellhead pads; GeoTIFF and annotated PDF |
| Deformation rate exceedance alert | Automated threshold monitoring on rolling 30-day displacement accumulation from Sentinel-1 time series | Email or API alert with flagged location, current rate, and historical context; triggered within 48 hours of threshold crossing |
| Atmospheric noise audit | Comparison of uncorrected and ERA5/GACOS-corrected interferogram stacks; spatial power spectrum analysis to quantify residual tropospheric signal | One-page uncertainty report accompanying each velocity product; flags periods of elevated atmospheric noise |
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.