Groundwater depletion detection in irrigated aquifer systems using InSAR
Interferometric SAR detects millimetre-to-centimetre annual subsidence caused by aquifer compaction in over-pumped irrigation districts, providing spatially continuous evidence of unsustainable extraction that sparse piezometer networks routinely miss.
Sensors
- Sentinel-1 (C-band SAR, ESA): 5.6 cm wavelength; Interferometric Wide Swath mode at 5 × 20 m ground resolution, 250 km swath. Repeat pass every 6 days over most land areas with two satellites. Free archive from 2014. Workhorse for persistent-scatterer and SBAS time-series over agricultural plains.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength penetrates vegetation canopy better than C-band, reducing decorrelation over crop-covered fields. Stripmap mode at 3 m resolution; ScanSAR at 60 m, 350 km swath. 14-day repeat. Particularly effective over the Indus Basin and North China Plain where dense crops degrade C-band coherence.
- COSMO-SkyMed (X-band SAR, ASI/Italian MoD): 3.1 cm wavelength; Stripmap at 3 m resolution, Spotlight at 1 m. Constellation of four satellites gives revisit of 1–4 days at mid-latitudes. X-band is sensitive to very small displacements but loses coherence faster over vegetated surfaces. Best suited to urban or bare-soil scatterers within an irrigation district.
- NISAR (L- and S-band SAR, NASA/ISRO, launch expected 2025): Dual-frequency design will deliver 12-day global repeat at 6–10 m resolution in both L- and S-band simultaneously. Designed explicitly for deformation science; expected to become the reference dataset for aquifer subsidence monitoring once operational.
What the ground is actually telling you
When groundwater is extracted faster than recharge replaces it, pore pressure in the aquifer matrix drops. The fine-grained clay layers interbedded with sand and gravel compact under the weight of overlying sediment. That compaction manifests at the surface as subsidence. In unconsolidated alluvial basins under heavy irrigation, the rate can reach 10–30 cm per year in the most stressed zones, as documented in peer-reviewed studies of the Central Valley of California, the North China Plain, and the Indus Basin.
The critical point for water managers is that this signal is irreversible in its inelastic component. Elastic subsidence, caused by seasonal drawdown, partly recovers when water levels rise. Inelastic compaction of clay layers does not. A time series that shows accelerating, non-recovering subsidence is therefore not just a proxy for extraction rate; it is evidence of permanent aquifer storage loss. InSAR cannot tell you directly how many cubic metres of water have been removed, but it tells you where permanent damage is occurring and at what pace, which is often the more actionable fact.
How interferometric SAR turns phase noise into displacement maps
SAR satellites illuminate the ground with microwave pulses and record the phase of the returning signal. When two acquisitions from nearly identical orbital positions are combined into an interferogram, the phase difference encodes surface displacement along the satellite's line of sight to sub-centimetre precision. One fringe in a Sentinel-1 interferogram represents 2.8 cm of line-of-sight displacement. Over a flat irrigated basin, where vertical motion dominates, that converts to roughly the same vertical movement.
Single interferograms are noisy. The operational approach is to stack dozens or hundreds of acquisitions using either Persistent Scatterer InSAR (PS-InSAR), which anchors the analysis to stable point reflectors such as buildings, roads and irrigation infrastructure, or Small Baseline Subset (SBAS) methods, which average coherent patches across the landscape. Both approaches produce mean velocity maps and displacement time series at each scatterer or pixel. Over a well-instrumented basin, PS-InSAR velocity maps from Sentinel-1 have been validated against GPS benchmarks to within 1–2 mm per year in published studies.
Atmospheric water vapour is the main noise source. Tropospheric delay can mimic or mask a centimetre of apparent displacement in a single interferogram. Time-series methods suppress this by averaging over many acquisitions, and corrections using ERA5 reanalysis or GACOS atmospheric models reduce residual errors further. Over large basins, a residual uncertainty of 2–5 mm per year in mean velocity is a realistic honest estimate after atmospheric correction.
What the published record shows in the major irrigation basins
The Central Valley of California is the most extensively studied case. InSAR studies using ENVISAT, Sentinel-1 and COSMO-SkyMed have documented subsidence bowls exceeding 25 cm per year in the Tulare Basin during drought years, with cumulative subsidence in some locations surpassing 8 metres since the mid-twentieth century. The USGS and NASA have both published results from this area. The spatial pattern reveals that the worst compaction follows the distribution of deep wells tapping confined aquifer layers, not simply the density of irrigated acreage.
On the North China Plain, L-band ALOS and C-band Sentinel-1 studies have mapped subsidence funnels beneath Beijing and Hebei Province reaching 10 cm per year, driven by decades of groundwater extraction for wheat and maize irrigation. The Indus Basin in Pakistan shows similar patterns, with subsidence rates of several centimetres per year documented in Balochistan and Punjab irrigation districts. In each case, the InSAR spatial pattern reveals heterogeneity that a network of even a few dozen piezometers would not resolve.
Honest limits of the method
InSAR measures displacement along the satellite's line of sight, not purely vertical movement. Decomposing line-of-sight displacement into vertical and horizontal components requires acquisitions from both ascending and descending orbits. Where only one geometry is available, an assumption of purely vertical motion introduces error, typically small over flat alluvial basins but non-negligible near faults or steep terrain.
Coherence is the other hard constraint. Sentinel-1 C-band loses coherence rapidly over dense, tall vegetation. In rice-paddy dominated systems or areas with frequent crop rotation, large fractions of the landscape may yield no usable signal between sowing and harvest. L-band ALOS-2 mitigates this substantially but does not eliminate it. The practical consequence is that subsidence maps over heavily cropped areas show gaps or increased noise during the growing season, and annual velocity estimates carry larger uncertainty than those over bare soil or urban scatterers.
Finally, the method detects compaction but cannot attribute it unambiguously to irrigation extraction without supporting data. Industrial groundwater use, urban pumping, and natural consolidation of young sediments can produce similar signals. Correlation with pumping records, well logs, or GRACE satellite gravity data is necessary to close the attribution argument.
Combining InSAR with GRACE and field data
GRACE and GRACE-FO measure total terrestrial water storage anomalies at roughly 300 km resolution by detecting changes in Earth's gravity field. InSAR provides the spatial detail GRACE cannot. Used together, GRACE constrains the basin-scale water volume change while InSAR maps where within the basin the compaction is concentrated. The combination is more persuasive to a regulator or water authority than either dataset alone.
Piezometer networks, where they exist, remain the ground truth for hydraulic head. InSAR time series can be calibrated against piezometer records to convert displacement rates into approximate storage-change estimates using published specific storage coefficients for the local geology. Satellize structures its aquifer-monitoring analytics around this three-layer integration: InSAR velocity maps as the spatial backbone, GRACE anomalies for volume context, and available well records for calibration, with explicit uncertainty bounds on every derived estimate.
Delivery cadence is typically quarterly for velocity updates, with annual reports that separate elastic seasonal signal from the inelastic trend that indicates permanent storage loss. That distinction is the number that matters most to a ministry of agriculture or a basin water authority.
What a monitoring programme actually produces
The output of a sustained InSAR monitoring programme is a time series of subsidence velocity maps, updated as new SAR acquisitions arrive, with anomaly alerts when a zone accelerates beyond a defined threshold. Spatial resolution of the velocity maps is typically 20–100 m depending on the processing approach and the density of coherent scatterers. At that resolution, individual well fields, canal commands, and farm clusters become distinguishable.
For a water authority, the most useful derivative product is a ranked list of sub-basin zones by inelastic subsidence rate, cross-referenced with licensed extraction volumes where those records exist. That ranking is a direct input to enforcement prioritisation. It is also the kind of evidence that survives scrutiny in water-rights adjudication, because the displacement record is physically grounded, spatially explicit, and archived in a publicly accessible satellite dataset that neither party to a dispute controls.
Satellize can run this analytics stack on Sentinel-1 open-archive data for any basin with adequate coherence, adding ALOS-2 or COSMO-SkyMed tasking where C-band coherence is insufficient. The Tonga crop-estimation programme demonstrates the organisation's approach to sustained agricultural analytics in data-sparse environments; the aquifer monitoring workflow follows the same principle of anchoring every derived number to a physically interpretable measurement.
Typical figures
| Typical spatial resolution (velocity maps) | 20–100 m depending on PS or SBAS approach and scatterer density |
| Minimum detectable velocity (annual mean) | 2–5 mm per year after atmospheric correction; single-interferogram noise 5–20 mm |
| Revisit period (Sentinel-1) | 6 days (two-satellite constellation); 12 days with one satellite |
| Revisit period (ALOS-2 PALSAR-2) | 14 days |
| SAR frequency / wavelength | C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band 3.1 cm (COSMO-SkyMed) |
| Swath width | 250 km (Sentinel-1 IW); 350 km (ALOS-2 ScanSAR); 40 km (COSMO-SkyMed Stripmap) |
| Archive depth | Sentinel-1 from 2014; ALOS-2 from 2014; ENVISAT ASAR from 2002 (for long baselines) |
| Processing latency (operational update) | Typically 2–4 weeks after acquisition, depending on atmospheric correction pipeline |
| Delivery formats | GeoTIFF velocity and displacement rasters, GeoPackage / Shapefile scatterer point clouds, PDF quarterly report, optional WMS/WMTS tile service |
| Vertical vs. horizontal decomposition | Requires ascending + descending passes; single-geometry products assume vertical-dominant motion |
Analytics Satellize can run
| Basin-wide mean subsidence velocity map | PS-InSAR or SBAS time-series inversion over multi-year Sentinel-1 stack | Annual GeoTIFF raster at 20–50 m resolution with uncertainty layer; PDF summary report |
| Inelastic vs. elastic displacement separation | Seasonal decomposition of displacement time series correlated with known irrigation calendar | Per-zone time-series charts distinguishing recoverable seasonal signal from permanent trend; GIS polygon layer |
| Subsidence acceleration alerts | Change-point detection on rolling 12-month velocity window; threshold defined per basin | Automated alert report flagging zones where velocity has increased beyond agreed threshold, delivered within 4 weeks of triggering acquisition |
| Sub-basin extraction risk ranking | Spatial clustering of inelastic velocity hotspots cross-referenced with irrigation command boundaries | Ranked table of command zones by subsidence severity; GeoPackage with zone polygons and velocity statistics |
| GRACE-InSAR integrated storage-change estimate | Downscaling GRACE terrestrial water storage anomalies using InSAR spatial weights; specific storage coefficients from published literature | Quarterly basin water balance summary with volume-change estimates and explicit uncertainty ranges; tabular data for import into water-accounting systems |
| Long-baseline historical subsidence reconstruction | SBAS processing of ENVISAT ASAR archive (2002–2012) spliced to Sentinel-1 series | 20-year cumulative displacement map where archive coherence permits; PDF interpretive report |
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.