Regional aquifer drawdown from construction dewatering detected by InSAR
Large construction dewatering schemes can depress piezometric levels across an entire urban district, causing subsidence that damages third-party buildings and infrastructure far outside the site fence. Sentinel-1 time-series InSAR maps these bowls at millimetre scale, but separating dewatering from other abstraction requires ancillary piezometer data.
Sensors
- Sentinel-1 (ESA, C-band, 5.6 cm wavelength): Interferometric Wide swath mode delivers 5 x 20 m ground resolution with 6-day repeat at mid-latitudes when both satellites are operational. The free, global archive from 2014 onward enables retrospective baseline establishment before a dewatering scheme begins. Phase sensitivity to line-of-sight displacement is approximately 3-5 mm per interferogram under good coherence conditions.
- TerraSAR-X / TanDEM-X (Airbus, X-band, 3.1 cm wavelength): Stripmap mode achieves 3 m resolution with an 11-day revisit; Spotlight mode reaches 1 m. Shorter wavelength improves sensitivity to small displacements but decorrelates faster over vegetated or disturbed surfaces. Tasked commercially; useful for resolving fine spatial structure within a subsidence bowl that Sentinel-1 blurs.
- COSMO-SkyMed (ASI, X-band): Constellation of four satellites achieves revisit intervals as short as 1-4 days in enhanced mode, which is valuable for tracking rapid dewatering-induced subsidence during active pumping phases. Stripmap resolution is approximately 3 m; Spotlight reaches sub-metre. Commercially tasked.
- Sentinel-1 Persistent Scatterer / SBAS processing (derived product): Multi-temporal InSAR algorithms such as Persistent Scatterer Interferometry (PSI) and Small Baseline Subset (SBAS) extract displacement time series from stacks of 20 or more interferograms, reducing atmospheric noise and improving velocity precision to roughly 1-2 mm/year over urban areas with dense persistent scatterers.
Why a basement excavation can crack buildings three streets away
Deep foundation pits for metro stations, underground car parks and large basements routinely penetrate confined or semi-confined aquifers. To keep the excavation dry, contractors run dewatering wells continuously, sometimes pumping thousands of cubic metres per day. That extraction does not stay local. Drawdown propagates radially through the aquifer, depressing piezometric head across a zone that can extend hundreds of metres to several kilometres from the pit, depending on aquifer transmissivity and pumping duration.
Where clays overlie the depressurised aquifer, effective stress increases and the clay consolidates. The surface settles. Buildings, pipelines and tunnels founded in or on that clay settle with it, unevenly if they straddle the edge of the drawdown cone. The settlement may be millimetric but enough to crack masonry, misalign drainage, or induce differential movement in adjacent metro tunnels. The site contractor's monitoring prisms cover the immediate vicinity; they rarely cover the full extent of the influence zone. InSAR does.
What the published record shows about scale and rate
Shanghai has been studied more thoroughly than almost any other city for dewatering-induced subsidence, partly because its soft Quaternary clays are highly compressible and partly because its metro expansion has been relentless. Published InSAR studies using Envisat and Sentinel-1 data have documented localised subsidence bowls of 20-60 mm centred on major excavation sites, with the bowl edge detectable at distances of 300-800 m from the pit boundary. Rates during active pumping have reached 10-30 mm per year in the most affected zones, tapering to background urban settlement rates of 5-10 mm per year at the bowl margin.
London's Crossrail programme, now the Elizabeth line, provided a well-instrumented case. The project combined conventional levelling, extensometers and, retrospectively, InSAR analysis using ERS and Envisat archives. Subsidence bowls above station box excavations at sites such as Farringdon and Paddington were measurable from space, consistent with ground-truth levelling to within a few millimetres. The key published finding across multiple cities is that InSAR detects the spatial footprint of dewatering influence earlier and more completely than any ground-sensor network that a contractor would realistically deploy.
Separating dewatering signal from the urban noise floor
Urban areas subside for many reasons simultaneously: consolidation of recent fill, loading from new construction, leaking water mains, seasonal groundwater fluctuation and long-term aquifer depletion from municipal abstraction. A raw InSAR velocity map cannot tell you which mechanism is responsible for any given millimetre of movement. Separating dewatering-induced subsidence requires a combination of spatial reasoning and ancillary data.
The spatial signature of a dewatering bowl is distinctive: a roughly elliptical or circular depression, centred near the active excavation, with a gradient that steepens toward the pit and a temporal evolution that tracks the pumping programme. Background urban settlement tends to be more diffuse and correlated with older compressible deposits rather than with active construction sites. Superimposing the InSAR displacement field on a hydrogeological model, calibrated with available piezometer records, allows the dewatering component to be isolated. Without at least some piezometer data, the attribution remains probabilistic. That is an honest limit of the method and any analyst who does not state it clearly is overselling.
Processing chain from SAR archive to deformation map
A typical Sentinel-1 time-series analysis for a construction dewatering case begins with co-registering a stack of single-look complex (SLC) images to a common reference geometry. Interferograms are formed for pairs within a defined temporal and perpendicular baseline threshold. After topographic phase removal using the Copernicus DEM or TanDEM-X DEM, atmospheric delay is estimated and subtracted using either a weather model (ERA5 is commonly used) or a spatial-temporal filtering approach. The residual phase is then decomposed into a linear velocity and a time series of displacement epochs.
PSI methods identify pixels with stable phase history, typically man-made structures, and achieve point densities of hundreds to thousands of scatterers per square kilometre in dense urban areas. SBAS methods use distributed scatterers and are better suited to less urbanised fringes of the bowl. The two approaches are often combined. Output is a georeferenced displacement time series at each scatterer, projected into line-of-sight displacement and, where ascending and descending geometries are available, decomposed into vertical and horizontal components. Vertical displacement is the primary quantity of interest for dewatering assessment.
Where the method fails and what to do about it
Temporal decorrelation is the most common processing failure. If the area around the pit is stripped of vegetation, covered in loose spoil or flooded with groundwater seepage, coherence drops and the interferogram becomes noisy precisely where displacement is largest. X-band sensors such as TerraSAR-X decorrelate faster than C-band Sentinel-1 in disturbed terrain, though their higher resolution can partially compensate by finding stable scatterers within the clutter.
Atmospheric delay is the dominant error source for individual interferograms. A single tropospheric anomaly can introduce apparent displacement of 10-20 mm that mimics a real signal. Time-series methods suppress this by averaging over many acquisitions, but a short monitoring window, say fewer than 15 scenes, leaves residual atmospheric noise that can mask slow dewatering signals in the 2-5 mm/year range. Cloud cover is irrelevant for SAR, which is one of the reasons SAR is preferred over optical for this application. The remaining hard limit is that InSAR measures surface displacement, not piezometric head directly. The relationship between head change and surface settlement depends on aquifer storativity and clay compressibility, parameters that must come from site investigation data, not from the satellite.
Satellize integrates Sentinel-1 time-series InSAR with client-supplied piezometer records and published hydrogeological parameters to produce dewatering influence assessments. The Tonga crop-estimation programme demonstrated the organisation's approach of combining open-constellation data with domain-specific modelling; the same principle applies here, with geomechanics replacing agronomy.
What the deliverable looks like in practice
A dewatering influence assessment typically comprises three layers. First, a baseline velocity map derived from pre-construction Sentinel-1 archives, establishing the natural settlement rate for each persistent scatterer in the study area. Second, a monitoring time series updated at each satellite pass during the active dewatering period, with an anomaly flag triggered when displacement at any scatterer exceeds the baseline rate by a defined threshold, commonly 5 mm above background over a 30-day window. Third, a spatial extent map of the inferred drawdown cone, derived by fitting a Theis or numerical groundwater model to the displacement field.
These outputs are most useful to three audiences. Project insurers need to establish whether damage claims from neighbouring property owners are plausibly attributable to the dewatering scheme. Regulators need to verify that the contractor's mitigation measures, recharge wells, slurry walls, reduced pumping rates, are having the claimed effect. And the contractor's own geotechnical team needs early warning that the influence zone is expanding beyond the predicted boundary, before a third-party structure reaches a damage threshold. Delivering results as georeferenced GIS layers, with time-stamped displacement tables per structure, makes the data directly usable in legal and regulatory proceedings.
Typical figures
| Primary sensor | Sentinel-1 C-band SAR, 5.6 cm wavelength |
| Ground resolution (Sentinel-1 IW mode) | 5 m range x 20 m azimuth (multi-looked to ~14 x 14 m in standard processing) |
| Revisit interval | 6 days at mid-latitudes with Sentinel-1A and 1B both operational; 12 days with single satellite |
| Line-of-sight displacement sensitivity | Approximately 3-5 mm per interferogram; 1-2 mm/year velocity precision in PSI time series over dense urban areas |
| Minimum detectable dewatering signal | Approximately 5-10 mm cumulative settlement above background, depending on coherence and atmospheric noise level |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); enables pre-construction baseline of several years |
| Commercial supplement resolution | TerraSAR-X Stripmap 3 m; Spotlight ~1 m. COSMO-SkyMed Stripmap ~3 m, revisit 1-4 days |
| Atmospheric correction | ERA5 reanalysis or spatial-temporal filtering; residual error typically 5-15 mm per interferogram before time-series averaging |
| Delivery formats | GeoTIFF displacement maps, GeoPackage or Shapefile scatterer time series, CSV displacement tables, PDF assessment report |
| Processing latency (monitoring mode) | Typically 2-5 days after satellite acquisition for updated displacement map |
Analytics Satellize can run
| Pre-construction baseline velocity map | PSI or SBAS time-series InSAR on 2-4 years of pre-project Sentinel-1 archive | GeoTIFF and scatterer shapefile showing mean annual displacement rate and seasonal signal per pixel, delivered before dewatering commences |
| Dewatering influence zone delineation | Spatial fitting of Theis radial flow model or numerical groundwater model to InSAR displacement field, calibrated against available piezometer records | Polygon GIS layer showing inferred drawdown cone extent at defined piezometric head contours, updated monthly |
| Structure-level displacement time series | Persistent scatterer extraction at identified buildings, walls and infrastructure within the study area | Per-structure CSV time series with epoch dates, cumulative displacement and velocity, flagged against user-defined alert thresholds |
| Anomaly alert when displacement exceeds threshold | Statistical change detection on rolling 30-day displacement window against pre-construction baseline distribution | Email or API alert with scatterer ID, location, exceedance magnitude and time-series plot attachment |
| Vertical and horizontal displacement decomposition | Combination of ascending and descending Sentinel-1 geometries to separate vertical settlement from lateral movement toward the pit | Dual-component GeoTIFF maps; lateral displacement layer is particularly relevant for assessing risk to adjacent tunnels and retaining structures |
| Mitigation effectiveness assessment | Before-and-after time-series comparison bracketing installation of recharge wells or cut-off walls, with spatial correlation to mitigation geometry | Comparative report quantifying change in subsidence rate and spatial extent following mitigation, suitable for regulatory submission |
| Third-party damage attribution report | Spatial and temporal correlation of InSAR displacement onset with pumping records; comparison of affected-structure displacement against background settlement cohort | PDF technical report with figures, methodology, uncertainty bounds and data provenance, formatted for legal or insurance use |
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.