Ground subsidence monitoring for property structural risk
Satellite radar interferometry measures ground deformation to sub-centimetre precision over cities, clay belts and former landfill sites, giving property owners and lenders an objective record of structural risk before cracks appear in the walls.
Sensors
- Sentinel-1A/B (ESA): C-band SAR at 5.6 cm wavelength, 6-day repeat at mid-latitudes when both satellites are operational (12-day with one). Ground range resolution 5 x 20 m in Interferometric Wide Swath mode. Free and open archive from 2014. Coherence loss over vegetation limits PS-InSAR to built surfaces and bare ground.
- COSMO-SkyMed (ASI): X-band SAR constellation of four satellites; Stripmap mode delivers 3 m resolution, SpotLight as fine as 1 m. Sub-daily revisit possible via multi-satellite tasking. X-band is more sensitive to surface moisture and vegetation than C-band, but delivers finer spatial detail on individual structures.
- TerraSAR-X / TanDEM-X (DLR / Airbus): X-band, SpotLight mode to 1 m resolution. The TanDEM-X pair enables single-pass interferometry for baseline DEM generation. Staring SpotLight mode achieves better than 0.3 m azimuth resolution, useful for monitoring individual buildings or infrastructure assets.
- RADARSAT-2 (MDA): C-band, 24-day exact repeat orbit, Ultra-Fine mode to 3 m resolution. Left- and right-looking capability provides additional viewing geometries, which helps decompose line-of-sight displacement into vertical and east-west horizontal components when combined with ascending and descending passes.
What a radar satellite actually measures, and what it does not
Synthetic aperture radar interferometry compares the phase of microwave pulses returned from the same patch of ground on two separate passes. A shift in phase corresponds to a change in the distance between satellite and surface. The technique is sensitive to displacements along the satellite's line of sight, which runs at roughly 30 to 46 degrees from vertical depending on the orbit geometry. That geometry matters enormously: a purely vertical settlement of 10 mm appears as roughly 8 to 9 mm of line-of-sight change, but a horizontal motion of the same magnitude in the east-west direction appears as a similar signal. The two are ambiguous unless you combine ascending and descending passes, which look at the ground from opposite sides and allow a partial decomposition into vertical and horizontal components. North-south motion is nearly invisible to polar-orbiting SAR regardless of pass direction.
The practical consequence is that InSAR results should always state which pass geometry was used and whether decomposition has been attempted. A single ascending-pass time-series reported as 'vertical subsidence' is technically line-of-sight displacement. That distinction matters when the question is whether a building is sinking or being pushed sideways by an adjacent excavation.
PS-InSAR and SBAS: two ways to extract a signal from noise
Two processing families dominate operational subsidence monitoring. Persistent Scatterer InSAR (PS-InSAR) identifies individual pixels, typically corner reflectors, metal structures or exposed rock faces, that return a stable radar signal across a long stack of images. Displacement histories are estimated for each persistent scatterer relative to a reference point assumed to be stable. The method can achieve measurement precision of 1 to 2 mm per year in the line-of-sight direction over a dense urban area with a multi-year Sentinel-1 archive. Small Baseline Subset (SBAS) processing instead selects image pairs with short spatial and temporal baselines, averages over distributed targets rather than point scatterers, and produces spatially denser displacement maps at the cost of some precision. SBAS is often preferable over industrial estates or peri-urban areas where point scatterers are sparse.
Both methods require a stable reference network. If the chosen reference point is itself moving, every measurement in the network is biased by that motion. GNSS benchmarks, when available, provide an independent check. Without them, the analyst must make a geological argument that the reference area is stable, which is not always possible over a city underlain by variable geology.
Where coherence fails: the vegetated area problem
Coherence is the statistical measure of how similar two radar returns are between passes. Vegetation, with its leaves and branches moving between acquisitions, produces near-zero coherence at C-band over intervals of more than a few days. This is not a processing limitation that better algorithms can overcome; it is a physical property of the scattering mechanism. The practical result is that InSAR cannot monitor subsidence beneath parkland, gardens, agricultural fields or tree-lined streets. Measurements stop at the edge of the tarmac.
X-band systems such as COSMO-SkyMed and TerraSAR-X lose coherence even faster over vegetation than C-band Sentinel-1, though their finer resolution compensates by resolving individual building elements. In practice, a city with generous green space will have significant gaps in its PS-InSAR coverage. Those gaps often coincide with the clay-rich soils most prone to seasonal shrink-swell, which is an uncomfortable irony. Ground-truth surveys or extensometers remain necessary in those locations.
Clay soils, landfill and groundwater: the three main drivers
Shrink-swell clay is the dominant cause of subsidence damage to residential property in the United Kingdom, with the Building Research Establishment estimating that clay-related movement costs insurers hundreds of millions of pounds annually. InSAR time-series over London's clay belt, published by the British Geological Survey using Sentinel-1 data, have documented seasonal displacement cycles of 5 to 15 mm, with net downward trends in areas of prolonged dry summers. The satellite record, going back to 2014 for Sentinel-1 and earlier for ERS and Envisat, provides a baseline against which future climate-driven drying can be assessed.
Former landfill sites present a different problem. Organic waste decomposes and compresses over decades, producing irregular, differential settlement that is highly damaging to structures built on or adjacent to the site. InSAR is well suited here because the settlement rates, often 10 to 50 mm per year in active phases, are large relative to the measurement noise, and the surfaces are typically sealed and coherent. Groundwater extraction creates a third pattern: broad, bowl-shaped subsidence over aquifer systems, documented in cities from Mexico City to Jakarta, where rates can exceed 100 mm per year in extreme cases. InSAR at Sentinel-1 resolution maps these bowls clearly, though distinguishing extraction-driven compaction from natural consolidation requires hydrogeological context.
Turning a displacement map into a property risk product
A raw InSAR velocity map is not a property risk product. Converting millimetres-per-year into something a lender, insurer or local authority can act on requires several additional steps. First, displacement rates must be classified against damage thresholds. The widely cited Boscardin and Cording framework, and subsequent refinements by Burland and colleagues, relate differential settlement and angular distortion to damage categories ranging from cosmetic cracking to structural failure. Applying those thresholds to InSAR-derived settlement gradients across a building footprint produces a damage-likelihood score.
Second, the result must be presented with its uncertainty. PS-InSAR precision degrades with distance from the reference point, with atmospheric phase delay (water vapour introduces apparent range changes of up to 10 mm that must be modelled or filtered), and with the density of scatterers. A responsible output states the estimated measurement uncertainty alongside the displacement value. Satellize applies this processing chain to Sentinel-1 open-archive data and, where finer resolution is warranted, to commercially tasked X-band acquisitions, delivering outputs as georeferenced GIS layers with per-point velocity, uncertainty and damage-class attribution. The same analytical rigour applied to Tonga's crop programme applies here: the number means what it says, or it says so.
One practical note for buyers: archive depth matters. A single interferogram shows one displacement episode. A five-year time-series distinguishes seasonal reversible motion from irreversible long-term settlement. Commissioning a short baseline study to assess a single transaction is far less informative than accessing the full Sentinel-1 archive, which runs from 2014 and is free.
Honest limits before you commission a study
InSAR does not work everywhere. Dense vegetation, steep topography and rapid surface change all degrade or destroy coherence. Sentinel-1's 5 x 20 m pixel cannot resolve subsidence affecting a single narrow building; COSMO-SkyMed or TerraSAR-X are needed for that. Atmospheric correction is imperfect, and residual tropospheric noise sets a practical floor on detectable rates of around 1 to 2 mm per year for a well-processed multi-year Sentinel-1 stack, and somewhat worse for shorter time-series. The technique measures surface displacement, not the cause; attributing movement to clay shrinkage versus a leaking drain versus a nearby tunnel requires ground investigation. The satellite tells you where to look and how urgently. It does not replace the structural engineer.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | 5 m range x 20 m azimuth; PS points spaced at roughly one per 20-100 m in dense urban areas |
| Spatial resolution (COSMO-SkyMed / TerraSAR-X SpotLight) | 1-3 m; individual building elements resolvable |
| Revisit interval (Sentinel-1, single satellite) | 12 days; 6 days when both A and B satellites operational |
| Line-of-sight displacement precision (PS-InSAR, multi-year stack) | 1-2 mm per year velocity; single-epoch precision approximately 3-5 mm |
| Minimum detectable subsidence rate (practical floor) | Approximately 1-2 mm per year for a 3+ year Sentinel-1 time-series after atmospheric correction |
| Archive depth (Sentinel-1) | 2014 to present; ERS and Envisat archives extend coverage to 1992 for some regions |
| Radar frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1, RADARSAT-2); X-band ~9.6 GHz / 3.1 cm (COSMO-SkyMed, TerraSAR-X) |
| Coverage per acquisition (Sentinel-1 IW) | 250 km swath width; continental-scale coverage in a single pass |
| Coherence limitation | Vegetated surfaces incoherent at C-band over 6-12 day intervals; technique restricted to built surfaces and bare ground |
| Delivery formats | GeoTIFF velocity maps, point-cloud shapefiles or GeoPackage with per-point velocity, uncertainty and damage class; CSV time-series per asset |
Analytics Satellize can run
| City-scale subsidence velocity map | SBAS processing of Sentinel-1 IW stack; atmospheric correction via ERA5 reanalysis or spatial filtering | GeoTIFF raster and vector point layer showing mean line-of-sight velocity and uncertainty per pixel, classed by severity |
| Persistent scatterer time-series per building footprint | PS-InSAR with building-footprint overlay; displacement aggregated to median per footprint | CSV and GeoPackage with monthly displacement values per asset, flagged where angular distortion exceeds Burland damage thresholds |
| Differential settlement gradient map | Spatial gradient of PS velocity field across building footprints; angular distortion computed from adjacent scatterer pairs | GIS polygon layer with damage-class score (cosmetic to structural) per building, suitable for portfolio screening |
| Seasonal vs. long-term trend decomposition | Harmonic regression on PS/SBAS time-series to separate annual shrink-swell cycle from irreversible downward trend | Per-point report of seasonal amplitude and net secular rate; flags assets where secular rate exceeds defined threshold |
| Landfill and brownfield site settlement assessment | SBAS over sealed landfill surfaces; rate and spatial pattern compared against site boundary and adjacent property footprints | Site report with settlement rate map, estimated remaining settlement trajectory and buffer-zone risk classification |
| Ascending / descending pass displacement decomposition | Combination of ascending and descending PS-InSAR velocity fields to separate vertical and east-west horizontal components | Dual-component velocity map (vertical and horizontal) with uncertainty bounds; narrative interpretation of dominant motion mechanism |
| Change-point alert for accelerating subsidence | Sequential statistical testing on rolling PS time-series to detect step changes or acceleration in displacement rate | Automated alert layer updated on each new Sentinel-1 acquisition cycle; threshold configurable per client risk appetite |
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.