Tower structural deformation monitoring using persistent-scatterer InSAR
Persistent-scatterer InSAR tracks millimetre-scale displacement of steel lattice towers across multi-year SAR archives, flagging subsidence or tilt before it becomes a structural or regulatory problem.
Sensors
- Sentinel-1 (ESA): C-band SAR (5.405 GHz), 6-day repeat at equator with two satellites, IW mode ground range resolution approximately 5 m x 20 m. Free archive from 2014. Preferred for long time-series PS-InSAR owing to archive depth and consistent acquisition geometry.
- COSMO-SkyMed Second Generation (ASI): X-band SAR (9.6 GHz), Stripmap mode at 3 m resolution, spotlight modes to sub-metre. Shorter wavelength improves sensitivity to small metallic scatterers but reduces atmospheric coherence window. Revisit 1-4 days with constellation of four satellites.
- TerraSAR-X / TanDEM-X (DLR/Airbus): X-band, spotlight mode to 0.25 m resolution. Excellent phase stability on discrete metallic targets. Archive from 2007. Typically used for high-confidence PS analysis on individual priority towers rather than fleet-wide screening.
- ICEYE X-band SAR constellation: X-band, stripmap at approximately 3 m and spotlight to sub-metre. Growing constellation offers flexible tasking and sub-daily revisit in some latitudes. Shorter archive than Sentinel-1 limits multi-year trend confidence.
Why steel towers make reliable radar targets
Persistent-scatterer InSAR works by identifying pixels that maintain a stable, high-amplitude radar return across dozens or hundreds of SAR acquisitions. Steel lattice towers are near-ideal candidates. The dihedral and trihedral geometries formed by angle-iron members produce strong, repeatable corner-reflector responses at C- and X-band wavelengths. A single tower may generate several independent persistent scatterers (PS) at different heights on the structure, which is useful: if the upper PS displaces relative to the lower, you have evidence of tilt rather than uniform settlement.
The phase of each PS encodes range change to sub-wavelength precision. At C-band (Sentinel-1, wavelength approximately 5.6 cm), the theoretical displacement sensitivity approaches 1 mm per year when coherence is high and atmospheric phase screens are well-modelled across a stack of 30 or more acquisitions. At X-band the wavelength is roughly 3.1 cm, which tightens the ambiguity interval but makes atmospheric correction more demanding. Neither band removes the fundamental ambiguity: PS-InSAR measures line-of-sight displacement, not vertical or horizontal displacement directly. Decomposing into vertical and east-west components requires ascending and descending passes, and north-south motion remains largely unobservable from polar orbits.
Ground conditions that make this worth running
Not every tower network needs a PS-InSAR survey. The method earns its cost when towers sit on ground that moves: active mining subsidence, shrink-swell clay, poorly compacted fill over old landfill, karst dissolution, or waterlogged peat. Expansive clay is a particular problem in temperate climates where seasonal moisture variation drives annual heave-and-settlement cycles of 10-30 mm, masking any secular trend unless the processing explicitly separates seasonal and linear components.
Undermined ground is the highest-risk scenario. Longwall coal extraction can produce surface subsidence bowls of several metres over months, with differential settlement across a tower footprint reaching tens of millimetres. A PS time series will show the onset and acceleration of that bowl long before a visual inspection would flag anything. The practical value is not just structural: regulators in several jurisdictions require telecoms operators to demonstrate ongoing structural compliance for towers above a certain height, and a documented monitoring record changes the conversation with both insurers and planning authorities.
What the processing chain actually does
A PS-InSAR workflow begins with co-registration of all acquisitions to a single master scene, then forms interferometric pairs. Pixels are selected as persistent scatterers using amplitude dispersion thresholds (typically below 0.25) or phase stability criteria across the stack. A network of PS is then used to estimate and remove the atmospheric phase screen, which is the dominant noise source at both C- and X-band. What remains is modelled as a linear displacement rate plus a residual topographic error term.
Tower-specific processing adds a step: each PS is tagged to a structure from an asset register or from automatic detection of high-coherence point clusters in the SAR amplitude image. Displacement time series are then extracted per tower and per scatterer height. Velocity maps give the fleet-wide view; individual time series give the forensic detail needed to decide whether a specific tower requires a ground inspection. Typical deliverables are velocity rasters in mm per year, time-series plots per PS cluster, and a ranked alert list of towers exceeding a defined displacement threshold.
Honest limits: what can go wrong
Atmospheric phase screens are the method's principal adversary. Tropospheric water vapour varies spatially at scales of kilometres, introducing apparent displacement signals of several millimetres per acquisition. Over a long stack this largely averages out, but in humid tropical or coastal environments the residual atmospheric noise can exceed the signal of interest. Ionospheric effects at C-band are minor at mid-latitudes but non-negligible near the equator. External atmospheric correction using ERA5 reanalysis or GACOS products reduces but does not eliminate this noise.
Coherence loss is the other hard limit. Vegetation growing around a tower base, or structural modification such as antenna additions, can disrupt the PS phase history and introduce a gap or bias in the time series. Towers in dense forest, or in regions with heavy seasonal snow cover, may simply not yield usable PS. Sentinel-1's 6-day revisit is excellent for time-series density, but its 5 m x 20 m IW mode pixel footprint means that closely spaced towers or towers in urban clutter may have multiple structures contributing to a single pixel. X-band spotlight modes resolve this, at higher cost and shorter archive depth. There is no configuration that is simultaneously cheap, high-resolution, and atmospherically immune.
From velocity map to maintenance decision
The output of a PS-InSAR run is a velocity field, not a maintenance schedule. Turning one into the other requires thresholds agreed with structural engineers. A commonly cited alert threshold in infrastructure monitoring literature is 2-3 mm per year of anomalous differential displacement between foundation and mid-tower PS, but the appropriate value depends on foundation type, soil category, and the operator's risk appetite. A tower on rock with a 2 mm per year signal warrants a different response than a guyed mast on fill with the same number.
Satellize structures this as a tiered output: a fleet-wide velocity layer updated on each new Sentinel-1 cycle, a ranked watch-list of towers exceeding operator-defined thresholds, and on-demand deep-dive reports using higher-resolution X-band data for towers that appear on the watch-list. The Sentinel-1 layer is the screening tool; X-band is the confirmation instrument. A ground inspection is still required before any engineering decision, but the InSAR output tells the field team exactly which tower to visit and what to look for at the foundation.
Archive depth and the value of retrospective analysis
Sentinel-1A has been acquiring data since April 2014. That is a decade-long archive of C-band phase history over most land surfaces, available at no cost through ESA's Copernicus Data Space. For a telecoms operator inheriting a tower estate with incomplete maintenance records, a retrospective PS-InSAR analysis over this archive can reveal which towers have been moving for years and which are stable. This is qualitatively different from a single structural inspection: it provides a continuous displacement record rather than a point-in-time snapshot.
TerraSAR-X extends the X-band archive to 2007 for priority sites, though data access requires a commercial licence. The combination of a long Sentinel-1 baseline for fleet screening and targeted TerraSAR-X or COSMO-SkyMed analysis for high-risk sites is the most cost-effective approach for operators managing estates of hundreds or thousands of towers.
Typical figures
| Spatial resolution (Sentinel-1 IW) | ~5 m range x 20 m azimuth per pixel; PS density depends on target coherence |
| Spatial resolution (COSMO-SkyMed / TerraSAR-X spotlight) | 0.25–1 m; resolves individual tower members |
| Revisit interval | 6 days (Sentinel-1 two-satellite); 1–4 days (COSMO-SkyMed); flexible (ICEYE) |
| Displacement sensitivity | ~1 mm per year line-of-sight under favourable coherence; 3–5 mm per year typical in humid or low-coherence conditions |
| Minimum stack size for reliable PS | 30 acquisitions recommended; 20 minimum for preliminary screening |
| SAR frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1); X-band ~9.6 GHz / 3.1 cm (COSMO-SkyMed, TerraSAR-X, ICEYE) |
| Archive depth | Sentinel-1: from April 2014; TerraSAR-X: from 2007; ICEYE: from 2018 |
| Displacement components observable | Line-of-sight; vertical and east-west by ascending/descending combination; north-south not reliably recoverable |
| Delivery formats | GeoTIFF velocity rasters, CSV/shapefile PS point layers, PDF time-series reports, GIS-ready alert layers |
| Processing latency (operational monitoring) | 3–7 days after new SAR acquisition, depending on atmospheric correction data availability |
Analytics Satellize can run
| Fleet-wide displacement velocity map | Persistent-scatterer InSAR (PS-InSAR) over Sentinel-1 IW stack; atmospheric correction via ERA5 or GACOS | GeoTIFF raster and shapefile of PS points with mm/year velocity, updated per Sentinel-1 cycle |
| Per-tower displacement time series | PS cluster extraction and phase unwrapping per registered tower asset; linear and seasonal component separation | CSV time series and PDF chart per tower, showing cumulative displacement and velocity trend |
| Threshold-based structural alert list | Operator-defined velocity and acceleration thresholds applied to PS time series; anomaly flagging with confidence score | Ranked watch-list GIS layer and email alert, updated per processing cycle |
| Tilt estimation from multi-height PS | Differential displacement between low and high PS on the same structure; geometric conversion to angular tilt using known tower height | Per-tower tilt magnitude and direction report for priority structures |
| Retrospective baseline analysis | Full Sentinel-1 archive stack from 2014; SBAS or PS-InSAR depending on coherence; identifies pre-existing displacement trends | Historical displacement summary report per tower estate or geographic cluster |
| High-resolution confirmation survey | COSMO-SkyMed or TerraSAR-X spotlight PS-InSAR for towers flagged by Sentinel-1 screening | Sub-metre PS point cloud with velocity and time series, formatted for structural engineering review |
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.