Surface settlement monitoring above tunnel boring operations
Persistent-scatterer InSAR can resolve millimetre-scale settlement above a tunnel boring machine before visible cracking begins, giving engineers an independent check on Peck-curve predictions and early warning of unexpected ground response.
Sensors
- Sentinel-1 (C-band SAR): 5.6 cm wavelength; Interferometric Wide Swath mode delivers 5 × 20 m ground resolution and a 6-day repeat at the equator (often 6-day effective revisit over European cities due to both satellites). Free and open. Adequate for slow, broad settlement troughs but coherence degrades over vegetated or disturbed surfaces.
- TerraSAR-X / TanDEM-X (X-band SAR): 3.1 cm wavelength; Stripmap mode gives ~3 m resolution, Spotlight up to ~1 m. Programmable 11-day repeat, reducible to ~2.5 days with tasking across ascending and descending orbits. Preferred for narrow urban corridors where the settlement trough width may be only 10–30 m.
- COSMO-SkyMed Second Generation (X-band SAR): Four-satellite constellation; Stripmap at ~3 m, Enhanced Spotlight at ~1 m. Revisit can reach 12 hours with two satellites tasked on the same geometry. Useful for fast-advancing TBMs where the settlement front moves several metres per day.
- ICEYE X-band SAR: Spotlight mode at ~0.5 m resolution. Small-satellite constellation allows flexible revisit scheduling; published revisit under 24 hours to a fixed point with multiple satellites. Coherence over hard urban surfaces is good, making it viable for persistent-scatterer analysis over short time series.
What the settlement trough tells you, and when
When a tunnel boring machine advances through saturated or cohesive ground, the soil above it relaxes into the void created by the excavation face and the tail void behind the shield. The resulting surface deformation follows a roughly Gaussian transverse profile, described by Peck's empirical formula. The trough width parameter i (the distance from the tunnel centreline to the inflection point) depends on tunnel depth and soil type. For a typical urban metro at 20 m depth in stiff clay, i might be 8–12 m; in looser ground it widens. Peak settlement directly above the TBM can range from a few millimetres to tens of millimetres depending on ground loss ratio, which is the volume of soil lost relative to the theoretical excavated volume.
The practical value of satellite InSAR here is not that it replaces levelling pins or automated total stations. It does not. Point sensors give higher precision at their exact locations. What InSAR adds is spatial continuity: a deformation map across an entire urban block, not just the handful of points where instruments were installed. It catches unexpected asymmetry in the trough, settlement extending further from the centreline than predicted, or a secondary zone of heave ahead of the TBM face that levelling arrays may miss entirely.
Persistent scatterers and why urban tunnelling is a good match
Persistent-scatterer InSAR (PS-InSAR) identifies pixels that maintain stable radar backscatter across a long stack of SAR acquisitions: building corners, lamp posts, exposed concrete, kerb stones. These become the measurement points. In a dense city, PS densities of hundreds to thousands per square kilometre are routine, which is exactly the environment above most metro tunnel drives. Each PS yields a displacement time series with precision in the range of 1–2 mm per measurement epoch under good conditions, though that figure assumes adequate coherence and a long enough archive to separate the deformation signal from atmospheric phase delay.
The method requires a minimum stack of roughly 15–20 acquisitions to begin producing reliable velocity estimates. For Sentinel-1 on a 6-day repeat, that is about three months of data before the first stable PS map emerges. TerraSAR-X on an 11-day repeat needs a similar number of scenes, so roughly five months. This temporal ramp-up is a genuine constraint: if monitoring is commissioned after the TBM has already been advancing for several months, the archive may already contain the most critical early data, but if the project starts without prior tasking, there is an unavoidable lag before PS-InSAR delivers its first reliable output.
Tracking a moving front: why temporal sampling matters more than resolution
A TBM advancing at 10–15 m per day moves the settlement influence zone by roughly one trough-width every few days. At 6-day revisit, Sentinel-1 captures the trough at perhaps two or three positions along its advance before it has moved well clear of a given building. That is usually sufficient to confirm whether the trough profile matches the Peck prediction and whether any structure is experiencing differential settlement across its footprint.
Faster TBMs, or situations where the ground response is unusually rapid, push the case toward X-band sensors with shorter revisit. COSMO-SkyMed Second Generation's sub-daily revisit capability is the most aggressive option currently available commercially, though the cost per acquisition rises sharply compared with Sentinel-1. The practical compromise for most urban metro projects is Sentinel-1 as the baseline archive, supplemented by TerraSAR-X Spotlight tasking over specific sensitive structures, such as listed buildings, hospitals, or bridge foundations, where higher spatial resolution and more frequent sampling are justified.
The signal-separation problem: tunnel, groundwater, or load?
Urban ground does not deform for only one reason at a time. A city block above a tunnel drive may simultaneously be experiencing seasonal groundwater drawdown, consolidation from a nearby building completed two years earlier, and the TBM-induced trough. PS-InSAR measures the sum of all these signals in the line-of-sight direction. Separating them requires either a long pre-construction baseline (to characterise background settlement rates before the TBM arrives) or independent data sources such as piezometer records or groundwater model outputs.
Ascending and descending orbit acquisitions help decompose the total displacement into vertical and east-west horizontal components, which can assist attribution. A TBM-induced trough is predominantly vertical with a small horizontal component directed toward the tunnel centreline; groundwater drawdown tends to be more uniformly vertical across a wider area. But the decomposition is imperfect, and north-south horizontal motion is invisible to side-looking SAR regardless of orbit geometry. Honest monitoring reports should quantify the residual ambiguity rather than present a single attributed settlement map as definitive.
Limits the procurement team should know before signing off
C-band coherence degrades quickly over disturbed ground surfaces. Active construction zones, freshly excavated trenches, and stockpiles of spoil can decorrelate the very pixels closest to the works, leaving a gap in the PS network precisely where ground movement is largest. X-band is less susceptible but not immune. This is not a reason to abandon the method; it is a reason to plan the monitoring geometry carefully, using descending orbits when ascending geometry looks toward a noisy construction face, and to accept that the innermost few metres of the trough may need ground-based instruments regardless.
Atmospheric phase delay is the other persistent nuisance. Tropospheric water vapour introduces apparent range changes of up to 10–15 mm in a single interferogram. PS-InSAR mitigates this by estimating the atmospheric contribution statistically across the PS network, but over short time series or in areas with few PS points, residual atmospheric noise can mask real deformation signals at the millimetre level. Independent atmospheric correction using ERA5 reanalysis or GACOS (Generic Atmospheric Correction Online Service) data reduces this, but adds processing complexity.
Satellize integrates Sentinel-1 archive processing with commercial tasking on TerraSAR-X and COSMO-SkyMed Second Generation, and can configure a settlement-monitoring workflow that includes pre-construction baseline extraction, fortnightly PS updates during the drive, and automated threshold alerts when displacement rates exceed agreed limits. The Tonga crop-estimation programme is a different domain entirely, but the underlying principle of combining open-archive data with targeted commercial acquisitions applies equally here.
From deformation map to engineering decision
The deliverable that matters to a project engineer is not a colour-coded displacement raster. It is a comparison of observed settlement contours against the predicted Peck trough at each TBM chainage, flagged when the two diverge by more than a defined threshold. That requires integrating the InSAR output with the TBM advance log, the geotechnical baseline report, and the structural sensitivity classification of each building above the route.
A well-configured monitoring programme produces a weekly settlement report keyed to TBM position, a GIS layer showing current and cumulative displacement per PS point, and an alert feed that triggers when any PS cluster exceeds the green-amber-red thresholds set in the environmental monitoring plan. The alert latency from SAR acquisition to engineer notification is typically 24–48 hours for commercial sensors with direct downlink to regional ground stations, and 48–72 hours for Sentinel-1 processed through Copernicus Data Space. That is fast enough to inform decisions about grouting, TBM face pressure adjustment, or precautionary evacuation of a sensitive structure.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 5 × 20 m ground range; PS point spacing in dense urban areas typically 5–20 m |
| Spatial resolution (TerraSAR-X Spotlight) | ~1 m; effective PS spacing in urban fabric 2–10 m |
| Revisit period | Sentinel-1: 6 days (dual-satellite, mid-latitudes); TerraSAR-X: 11 days standard, ~2.5 days multi-geometry tasking; COSMO-SkyMed SG: sub-daily possible with tasking |
| Minimum detectable displacement (PS-InSAR) | 1–2 mm per epoch under good coherence conditions; 3–5 mm in practice over short time series or noisy urban environments |
| Radar frequency / wavelength | C-band (Sentinel-1): 5.405 GHz / 5.6 cm; X-band (TerraSAR-X, COSMO-SkyMed SG, ICEYE): ~9.6 GHz / 3.1 cm |
| Minimum stack for stable PS solution | 15–20 acquisitions; approximately 3 months at 6-day revisit |
| Archive depth (Sentinel-1) | From 2014 (Sentinel-1A launch); pre-construction baselines available for most urban areas globally |
| Alert latency | 24–48 hours (commercial X-band with regional ground station); 48–72 hours (Sentinel-1 via Copernicus Data Space) |
| Atmospheric correction uncertainty | Residual tropospheric noise 2–5 mm per interferogram after GACOS or ERA5 correction |
| Delivery formats | GeoTIFF displacement rasters, GeoJSON PS point layers, CSV time series per PS cluster, PDF engineering report, threshold-alert feed (email or API) |
Analytics Satellize can run
| Pre-construction background settlement baseline | PS-InSAR time-series analysis on Sentinel-1 archive (StaMPS or SqueeSAR class algorithms) | GIS layer of pre-existing settlement rates per PS point along the tunnel corridor, delivered before TBM launch |
| Fortnightly settlement trough map keyed to TBM chainage | Incremental PS-InSAR update with TBM advance log overlay; Peck-curve comparison at each chainage | PDF engineering report with observed versus predicted trough profiles and residual anomaly map |
| Sensitive-structure differential settlement assessment | PS cluster analysis across individual building footprints using TerraSAR-X Spotlight; angular distortion calculation per structure | Building-by-building differential settlement table with green-amber-red status against Burland and Wroth damage category thresholds |
| Real-time threshold alert | Automated displacement rate monitoring against project environmental monitoring plan limits; triggered when any PS cluster exceeds defined mm/day rate | Alert notification via email or API within 48 hours of SAR acquisition; includes PS cluster map and time-series plot |
| Signal decomposition: tunnel versus groundwater contribution | Ascending and descending orbit combination for vertical and east-west displacement separation; comparison against piezometer or groundwater model inputs | Attribution report quantifying likely tunnel-induced component and residual ambiguity, with confidence bounds |
| Post-drive consolidation monitoring | Continued PS-InSAR time series for 12–24 months after TBM passage to track secondary consolidation and confirm stabilisation | Quarterly GIS update and annual settlement closure report for submission to planning authority or insurer |
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.