Surface settlement monitoring above tunnel and metro construction

Mechanised tunnelling creates predictable Gaussian settlement troughs at surface. X-band InSAR tracks these troughs at millimetre precision, flagging deviations from Peck-curve predictions before structural damage reaches buildings or utilities.

Sensors

TerraSAR-X / TanDEM-X: X-band (9.65 GHz), stripmap mode delivers 3 m resolution, spotlight mode reaches 1 m. Eleven-day exact repeat orbit; coherence in dense urban areas is typically high enough for PS-InSAR over stacks of 20+ acquisitions. The short wavelength (3.1 cm) means one fringe equals ~1.55 mm of line-of-sight displacement, giving sub-millimetre sensitivity after phase unwrapping.
COSMO-SkyMed (first and second generation): X-band (9.6 GHz) Italian constellation of four satellites. Revisit can reach daily for priority tasking; standard stripmap gives 3–15 m resolution, enhanced spotlight 1 m. Used in published Istanbul metro monitoring studies precisely because its short revisit reduces temporal decorrelation in active construction zones.
Sentinel-1 A/B: C-band (5.405 GHz), 12-day repeat per satellite (6-day with both active). IW mode gives 5 × 20 m resolution. Longer wavelength (5.6 cm) reduces sensitivity to small displacements relative to X-band and suffers more decorrelation over disturbed soil, but the free, dense archive since 2014 makes it valuable for establishing pre-construction baselines and for wide-area context around a project.
ICEYE X-band SAR constellation: Commercial X-band constellation with sub-metre spotlight capability and, for priority clients, sub-daily revisit at a given point. Useful for filling gaps between TerraSAR-X or COSMO-SkyMed passes during critical tunnelling phases. Archive depth is shorter than legacy systems, limiting long-baseline PS-InSAR stacks.

What the Peck trough tells you, and where it goes wrong

In 1969, Ralph Peck compiled field data from soft-ground tunnelling and showed that surface settlement above a driven tunnel follows a Gaussian curve transverse to the tunnel axis. The trough width parameter i (the distance from the tunnel centreline to the inflection point) scales predictably with tunnel depth and soil type. Volume loss, typically 0.5–2% of the excavated face area in well-controlled EPB or slurry TBM drives, sets the trough amplitude. This gives engineers a forward prediction: before the TBM arrives beneath a building, they know where the settlement will peak and roughly how deep it will be.

The model breaks down when ground conditions are heterogeneous, when the TBM face pressure drifts, when grout injection behind the shield is uneven, or when a building's foundations redistribute load in ways the free-field model ignores. Those deviations are exactly what monitoring must catch. A trough that is narrower than predicted, or offset laterally, often signals a localised void or a zone of unexpectedly stiff ground that is concentrating strain. A trough that is wider and shallower than predicted can mean the TBM is losing face pressure. Neither scenario is visible to a site engineer walking the surface; both are visible to InSAR.

Why X-band coherence survives a construction site

InSAR depends on the phase relationship between two radar acquisitions remaining stable, a property called coherence. Urban construction sites are hostile to coherence: plant movement, stockpiled materials, scaffolding and fresh concrete all scatter radar energy unpredictably. C-band (Sentinel-1) struggles here because its longer wavelength is sensitive to centimetre-scale surface changes across the whole resolution cell. X-band's shorter wavelength means its resolution cells are smaller in physical terms at equivalent geometry, and persistent scatterers (PS), such as building facades, lamp posts, kerb edges and manhole covers, dominate the return even when the surrounding ground is disturbed.

Persistent Scatterer InSAR (PS-InSAR) and the related Small Baseline Subset (SBAS) approach exploit this. Over a stack of 20–30 X-band acquisitions, algorithms identify pixels whose phase behaviour is consistent across all interferograms, fit a linear plus seasonal displacement model, and recover displacement time-series at those pixels with standard deviations routinely cited in the literature at 0.3–1 mm per acquisition. Published work on the Crossrail project used TerraSAR-X PS-InSAR to monitor settlement along the Elizabeth line tunnels beneath central London, detecting millimetre-scale precursor motion in buildings along the Paddington to Farringdon corridor before any visible damage occurred. The Istanbul metro extensions, monitored with COSMO-SkyMed, produced comparable results in a geologically more variable setting, with alluvial deposits beneath the Bosphorus shoreline requiring careful atmospheric correction.

From phase to decision: the monitoring workflow

Raw interferograms are not a deliverable. The operational workflow runs from SAR acquisition through co-registration, interferogram formation, atmospheric phase screen estimation and removal, PS or SBAS network solution, and finally geocoding to a ground coordinate system. The output is a displacement time-series at each PS point, typically expressed in the satellite line-of-sight direction. Converting line-of-sight to vertical and horizontal components requires either ascending and descending geometry combinations or assumptions about the likely displacement direction (vertical, for pure settlement).

Against that time-series, the monitoring team overlays the predicted Peck trough at each TBM chainage position and date. Automated threshold alerts fire when observed settlement at a PS point exceeds the predicted value by a defined margin, or when the rate of settlement acceleration crosses a trigger level. These thresholds are typically set in a monitoring specification agreed with the structural engineer before tunnelling begins, following guidance such as that in the UK's CIRIA Report C760 on tunnelling-induced ground movement. The alert latency from SAR acquisition to notification depends on processing pipeline design; with commercial X-band satellites and cloud processing, same-day delivery is achievable.

Honest limits of the method

PS-InSAR over active construction is not infallible. Where the surface is genuinely incoherent, such as freshly excavated soil, standing water in a compound, or a site covered in temporary structures, there are simply no PS points and the method is blind. Gaps in coverage of 20–50 m around an active shaft are common. Atmospheric water vapour introduces apparent displacement signals of several millimetres per acquisition; in humid coastal cities this is a serious noise source that requires either external tropospheric models (ERA5, GACOS) or spatial filtering, both of which introduce their own uncertainties.

Revisit rate is a genuine constraint during fast TBM drives. A TBM in London clay can advance 15–20 m per day. With an 11-day TerraSAR-X repeat, the trough can develop and partially recover between acquisitions, making it harder to reconstruct the peak settlement. Combining ascending and descending passes, or adding a second constellation such as COSMO-SkyMed, effectively halves the revisit gap. ICEYE can fill further. Even so, the method complements rather than replaces conventional levelling pins and tiltmeters at the highest-risk structures; it adds spatial coverage and archive depth that point sensors cannot match.

What published projects actually showed

The Crossrail (Elizabeth line) project in London is probably the best-documented case in the public literature. Researchers processing TerraSAR-X data over the central section reported PS point densities of several hundred per square kilometre in the built-up corridor, with displacement measurement precision of approximately 0.5 mm per epoch. Settlement troughs matching Peck predictions to within 1–2 mm were recovered for tunnel drives beneath Tottenham Court Road and Farringdon. Critically, buildings with deep piled foundations showed near-zero settlement while adjacent shallow-founded Victorian terraces showed 3–6 mm of settlement, a spatial contrast that only dense PS coverage reveals.

Istanbul metro monitoring with COSMO-SkyMed over soft alluvial ground near the Bosphorus demonstrated that X-band PS-InSAR could detect differential settlement between adjacent buildings of 2–3 mm, sufficient to flag structures approaching the angular distortion thresholds used in Turkish structural codes. Both programmes confirmed that InSAR-derived settlement maps, when delivered within days of acquisition, gave project teams enough lead time to adjust TBM face pressure or grout volumes before damage thresholds were breached.

Satellize's processing pipeline can ingest TerraSAR-X, COSMO-SkyMed, Sentinel-1 and ICEYE data against a client's tunnelling programme, delivering geocoded PS time-series and automated Peck-deviation alerts through a GIS feed or the Overhead analysis channel.

Typical figures

Spatial resolution (X-band spotlight)	1 m (TerraSAR-X, COSMO-SkyMed); sub-metre (ICEYE)
Spatial resolution (X-band stripmap)	3 m (TerraSAR-X, COSMO-SkyMed first gen)
Revisit interval	11 days (TerraSAR-X single pass); 1–4 days with COSMO-SkyMed constellation priority; sub-daily possible with ICEYE
Minimum detectable displacement (PS-InSAR)	0.3–1 mm per epoch line-of-sight, after atmospheric correction, over stacks of 20+ acquisitions
PS point density (dense urban)	Typically 100–500 PS/km² depending on building stock and construction disturbance
Radar frequency / wavelength	X-band: ~9.6 GHz / 3.1 cm (TerraSAR-X, COSMO-SkyMed, ICEYE); C-band: 5.405 GHz / 5.6 cm (Sentinel-1)
Archive depth	TerraSAR-X from 2007; COSMO-SkyMed from 2007; Sentinel-1 from 2014 (free); ICEYE from ~2018
Alert latency (operational pipeline)	Same-day to next-day from acquisition, depending on downlink and processing schedule
Coverage per pass	TerraSAR-X stripmap: ~30 × 50 km swath; COSMO-SkyMed Himage: ~40 × 40 km; Sentinel-1 IW: ~250 km swath
Delivery formats	GeoTIFF displacement maps, CSV or GeoJSON PS time-series, shapefile alert polygons, GIS WMS/WFS feed

Analytics Satellize can run

PS-InSAR displacement time-series	Persistent Scatterer InSAR (Ferretti et al. 2001 method class) over X-band SAR stack	GeoJSON or CSV point layer with displacement history per PS point, updated each acquisition cycle
Peck trough deviation map	Comparison of observed PS settlement profile against Gaussian trough predicted from TBM chainage, depth, and geotechnical volume-loss estimate	Per-chainage deviation report (PDF + GIS layer) showing observed vs predicted trough width and amplitude
Settlement rate acceleration alert	Rolling linear regression on PS time-series; alert fires when rate exceeds client-defined threshold (e.g. 2 mm/week) or acceleration exceeds trigger	Automated alert feed (email, webhook or GIS push) with PS point ID, current rate, and exceedance magnitude
Building differential settlement assessment	Angular distortion calculation from PS points on adjacent facade corners; compared against published damage category thresholds (e.g. Burland 1995 classification)	Building-level risk tier map (GIS polygon layer) updated per acquisition, exportable to structural engineer's monitoring log
Pre-construction baseline displacement map	SBAS or PS-InSAR over Sentinel-1 archive (2014 to project start) to characterise pre-existing settlement trends	Static GeoTIFF and summary report distinguishing tunnelling-induced motion from background subsidence
Multi-geometry vertical/horizontal decomposition	Combination of ascending and descending line-of-sight displacement vectors to separate vertical settlement from lateral ground movement toward the tunnel	Two-component (vertical + east-west) displacement GeoTIFF per epoch, with uncertainty bounds

Sources

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.