Land subsidence from groundwater over-extraction detected by InSAR
Excessive groundwater pumping compacts aquifer sediments and causes surface subsidence measurable to millimetres per year by satellite radar interferometry. Sentinel-1 time-series reveal subsidence bowls across major urban aquifers, but translating deformation rates into extraction volumes requires hydrogeological modelling.
Sensors
- Sentinel-1 (ESA, C-band SAR, 5.6 cm wavelength): Interferometric Wide Swath mode gives 250 km swath at 5 x 20 m resolution with 6-day repeat over most land areas (12-day with one satellite). The workhorse for PS-InSAR and SBAS time-series over urban aquifers. Free archive from 2014 onward.
- ALOS-2 PALSAR-2 (JAXA, L-band SAR, 23.6 cm wavelength): Longer wavelength penetrates vegetation and maintains coherence over agricultural land where C-band decorrelates. Stripmap mode reaches 3 m resolution; ScanSAR covers 350 km at 100 m. Revisit 14 days. Useful for peri-urban and irrigated-field subsidence.
- COSMO-SkyMed (ASI, X-band SAR, 3.1 cm wavelength): Spotlight mode achieves 1 m resolution, enabling detection of differential subsidence across individual structures and infrastructure. Revisit can be as short as 1 day with the four-satellite constellation. Commercial tasking; archive from 2007.
- TerraSAR-X / TanDEM-X (DLR, X-band SAR): Staring Spotlight mode reaches sub-metre resolution. Particularly valued for monitoring critical infrastructure (pipelines, bridges, levees) within a broader subsidence bowl. Revisit 11 days; commercial tasking required.
What compaction looks like from 700 km up
When groundwater is withdrawn faster than natural recharge replaces it, pore pressure in the aquifer sediments drops. Clay and silt layers compress. Some of that compression is elastic and partly recoverable when water levels recover; a larger fraction is inelastic, meaning the pore structure collapses permanently. The surface above subsides, sometimes by centimetres per year, sometimes by metres over decades.
Interferometric SAR measures the change in the two-way travel time of radar pulses between successive satellite passes. A displacement of half the radar wavelength shifts the phase by a full cycle. At Sentinel-1's C-band wavelength of 5.6 cm, the theoretical sensitivity to line-of-sight displacement is a few millimetres, though atmospheric water vapour noise is the practical limit in most settings, typically around 5 to 10 mm for a single interferogram. Time-series methods (PS-InSAR, which tracks stable point scatterers, and SBAS, which uses short-baseline image pairs) average down that noise over many acquisitions and routinely resolve seasonal and secular trends at the 1 to 3 mm per year level over urban areas.
Documented rates give the scale of the problem
Mexico City is the most studied case. Parts of the city overlie a drained lake bed of compressible lacustrine clays up to 300 m thick. Published InSAR studies using ERS, Envisat and Sentinel-1 data have recorded subsidence exceeding 30 cm per year in the most affected districts, with cumulative settlement of more than 10 m over the twentieth century. Tehran shows rates above 25 cm per year in the southern suburbs, documented with Envisat and Sentinel-1 time-series. Jakarta, which has since restricted deep-well pumping in part because of satellite evidence, recorded rates above 20 cm per year in the north of the city. These are not anomalies; they represent a pattern across any city that relies heavily on unconsolidated alluvial or lacustrine aquifers.
The spatial pattern of a subsidence bowl is itself informative. A roughly circular bowl centred on a dense well field is consistent with local extraction. An elongated trough aligned with a river valley or fault suggests structural controls on compaction. Differential subsidence across a single city block can indicate a transition between aquifer units or a buried fault acting as a flow barrier.
Elastic versus inelastic compaction: the distinction that matters for policy
Not all subsidence is equally bad. Elastic compaction occurs in sandy, coarser-grained layers and reverses when groundwater levels recover. Inelastic compaction, concentrated in interbedded clays, is essentially permanent at human timescales. The two can be separated in a time-series by correlating deformation with seasonal groundwater-level fluctuations measured in monitoring wells. Where subsidence tracks the seasonal pumping cycle and partially recovers each wet season, the aquifer is still in the elastic regime. Where the long-term trend continues downward even when seasonal recovery is visible, inelastic compaction is underway.
This distinction matters enormously for infrastructure planning and for regulation. A city whose aquifer is still in the elastic regime has time to act. One where inelastic compaction dominates is accumulating permanent structural damage to foundations, sewers, flood-defence walls and metro tunnels regardless of future pumping policy. InSAR alone cannot make this determination with certainty; it requires paired groundwater-level data from piezometers to close the interpretation.
What InSAR cannot tell you without a model
Subsidence rate is not extraction volume. The relationship between pumping rate and surface deformation depends on aquifer geometry, sediment compressibility, the thickness of clay interbeds, and the depth of the pumped interval. Two cities with identical subsidence rates may have very different extraction volumes if their aquifer sediment columns differ. Translating InSAR deformation maps into volumetric estimates of groundwater loss requires a hydrogeological model calibrated with borehole lithology, pumping records and groundwater-level data. InSAR provides the surface boundary condition; the model does the inversion.
There are further technical limits worth stating plainly. InSAR measures displacement in the radar line-of-sight direction, which for typical satellite geometries is roughly 30 to 45 degrees from vertical. Purely vertical subsidence is underestimated by a factor of approximately 1/cos(incidence angle). Horizontal motions, which can occur near the edges of a subsidence bowl or along faults, require combining ascending and descending orbit geometries to decompose. Dense vegetation, agricultural fields with rapidly changing surface conditions, and sand dunes all cause temporal decorrelation that breaks the interferometric signal, particularly at C-band. L-band (ALOS-2) is more tolerant of vegetation but has coarser resolution in standard modes.
From deformation map to regulatory evidence
A time-series product for a major urban aquifer typically begins with a stack of 50 to 200 Sentinel-1 acquisitions spanning three to six years. PS-InSAR identifies thousands to millions of stable scatterers (building rooftops, road surfaces, metal structures) and estimates a velocity and deformation history for each. SBAS uses distributed scatterers and is better suited to peri-urban or agricultural zones. The outputs are geocoded point clouds or gridded velocity maps, usually delivered as GeoTIFF or vector layers, showing mean annual displacement rate and cumulative deformation.
For a water authority or urban planning ministry, the actionable products are: a spatial map of subsidence rate with statistical uncertainty bounds; a time-series for selected critical infrastructure locations; a classification of the bowl into elastic and inelastic zones where piezometric data allow; and change detection alerts when rates accelerate beyond a defined threshold. Satellize processes Sentinel-1 stacks for clients requiring sovereign-grade analysis, applying the same open-method SBAS and PS-InSAR pipelines used in the published literature. The Tonga crop-estimation programme is a different application domain, but the underlying time-series processing infrastructure is shared.
The honest value proposition is this: InSAR provides spatial coverage that no ground-sensor network can match at comparable cost, and it does so retrospectively using the archive. A city that installed no piezometers in 2015 can still reconstruct its subsidence history from Sentinel-1 data back to that year. That is a genuinely rare capability.
Typical figures
| Sentinel-1 spatial resolution (IW mode) | 5 m range x 20 m azimuth (ground range); typically resampled to 10–20 m for time-series |
| Sentinel-1 revisit | 6 days (two-satellite constellation over Europe and high-priority areas); 12 days elsewhere |
| Minimum detectable velocity (PS-InSAR, urban, multi-year stack) | 1–3 mm per year under good coherence conditions; ~5–10 mm per year in noisier settings |
| COSMO-SkyMed / TerraSAR-X resolution (Spotlight) | 1–3 m; enables per-structure deformation monitoring |
| Archive depth (Sentinel-1) | April 2014 onward (globally variable); ERS/Envisat heritage extends to 1992 for some regions |
| Radar frequency / wavelength | C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band 3.1 cm (COSMO-SkyMed, TerraSAR-X) |
| Line-of-sight incidence angle (typical) | 30–45°; vertical displacement underestimated by factor cos(θ) unless ascending+descending combined |
| Swath width | 250 km (Sentinel-1 IW); 350 km (ALOS-2 ScanSAR); 40 km (COSMO-SkyMed Stripmap) |
| Typical processing latency (time-series update) | Days to weeks after new acquisition, depending on pipeline and stack size |
| Delivery formats | GeoTIFF velocity maps, CSV/Shapefile point clouds, time-series CSV per scatterer, GIS-ready vector layers |
Analytics Satellize can run
| Mean annual velocity map | PS-InSAR or SBAS time-series on Sentinel-1 stack (50+ scenes) | GeoTIFF raster showing mm/year displacement rate with per-pixel uncertainty; suitable for GIS overlay with cadastral data |
| Cumulative deformation time-series per scatterer | PS-InSAR stable-target tracking | CSV time-series for selected infrastructure points (bridges, levees, buildings), updated with each new satellite pass |
| Elastic vs inelastic compaction classification | Seasonal decomposition of InSAR time-series correlated with available piezometric records | Zonal polygon layer classifying the aquifer footprint; requires client-supplied groundwater-level data |
| Subsidence bowl delineation and volume estimate | Spatial integration of velocity field over coherent area; volume is surface proxy only, not extraction volume | Shapefile of bowl boundary at defined rate thresholds (e.g. >5 mm/yr, >20 mm/yr) with summary statistics report |
| Acceleration alert | Sequential change-point detection on rolling PS-InSAR time-series | Automated alert (email or API) when a monitored location exceeds a client-defined rate threshold between consecutive epochs |
| Multi-sensor displacement decomposition | Combination of ascending and descending orbit geometries to separate vertical and east-west horizontal motion | Two-component displacement raster (vertical + horizontal) for areas with both orbit geometries available |
| Historical reconstruction from archive | SBAS processing of full Sentinel-1 archive (2014 onward) or ERS/Envisat for pre-2014 periods | Long-record deformation history report covering up to 30 years for priority areas, formatted for regulatory submission |
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.