InSAR ground deformation monitoring at heritage sites and monuments
Interferometric SAR detects ground displacement at millimetre scale, revealing subsidence and differential settlement that threaten monuments and buried stratigraphy before visible damage appears. This page explains how Sentinel-1, COSMO-SkyMed and TerraSAR-X are applied at documented heritage sites, and where the technique fails.
Sensors
- Sentinel-1A/B (ESA): C-band SAR at 5.6 cm wavelength. Interferometric Wide Swath mode delivers 5 x 20 m resolution across a 250 km swath with a 6-day repeat (12-day with single satellite after Sentinel-1B loss in 2021). Free and open archive from 2014. Coherence degrades over vegetated or disturbed surfaces within one to two repeat cycles.
- COSMO-SkyMed (ASI): X-band SAR at 3.1 cm wavelength. Spotlight mode achieves 1 m resolution; StripMap reaches 3-15 m. Constellation of four satellites allows revisit as short as 12 hours for priority tasking. X-band coherence is high over masonry and bare ground but very sensitive to moisture changes in soil.
- TerraSAR-X / TanDEM-X (DLR): X-band, 3.1 cm. High-Resolution Spotlight mode reaches 1 m resolution. Staring Spotlight achieves sub-metre resolution and is particularly suited to persistent-scatterer analysis over dense built fabric. Revisit approximately 11 days, reducible by tasking.
- RADARSAT-2 (MDA): C-band, 5.6 cm wavelength. Ultra-Fine mode delivers 3 m resolution. Right- and left-looking capability allows ascending and descending geometries to be combined for 2.5D displacement decomposition. Useful archive depth from 2007.
Why a monument's slow sink is harder to catch than its sudden collapse
Structural failure at heritage sites rarely announces itself. The Colosseum's southern sector has been subsiding at rates measured by published InSAR studies at roughly 1-2 mm per year, driven by differential compaction of the fill beneath the arena floor and by groundwater drawdown from Rome's urban abstraction. Those rates are imperceptible to visual inspection over any reasonable inspection interval. They are, however, well within the detection floor of persistent-scatterer InSAR, which on stable urban reflectors routinely resolves displacement to 0.5-1 mm per year.
The physics is straightforward. Two SAR acquisitions of the same area from the same orbital geometry produce phase images. Where the ground has moved in the line-of-sight direction between passes, the phase difference encodes that displacement. One full phase cycle (2π) corresponds to half the radar wavelength: roughly 2.8 cm for C-band Sentinel-1, roughly 1.55 cm for X-band COSMO-SkyMed. Fractions of a cycle are measurable, which is why sub-centimetre sensitivity is achievable in practice, not just in theory.
Persistent scatterers versus SBAS: choosing the right processing chain
Two processing families dominate published heritage applications. Persistent-scatterer interferometry (PSI) identifies individual pixels, typically corner reflectors in masonry, metal fixtures or exposed rock, that maintain phase coherence across a long stack of interferograms. Over the Colosseum, published studies using COSMO-SkyMed data identified thousands of persistent scatterers across the monument's fabric, allowing deformation to be mapped at the scale of individual structural bays. PSI requires a dense archive, usually 20 or more acquisitions, and performs best over built fabric with stable point-like reflectors.
Small-baseline subset (SBAS) processing uses interferogram pairs with short spatial and temporal baselines to maximise coherence, then inverts the network to recover a displacement time series. SBAS tolerates distributed scatterers and is better suited to partially vegetated sites or earthwork contexts where no single pixel dominates. At Angkor Wat, published research applied SBAS to Sentinel-1 data to track seasonal and long-term settlement of the hydraulic infrastructure, particularly the embankments of the West Baray reservoir, where differential compaction threatens the geometry of a system that has managed water for nearly a millennium. The two approaches are not mutually exclusive: hybrid methods exist, and the choice depends on site character, archive depth and the spatial scale of the deformation signal being sought.
Venice: where the city is the instrument
Venice's historic centre is arguably the most intensively InSAR-monitored heritage landscape on Earth. The combination of dense masonry, minimal vegetation and a well-documented subsidence history going back to 20th-century industrial groundwater extraction makes it an ideal persistent-scatterer environment. Published studies using ERS, Envisat, COSMO-SkyMed and Sentinel-1 data have tracked the transition from net subsidence during the extraction era to near-stabilisation and localised uplift following aquifer recovery, with residual differential settlement continuing at rates of 1-3 mm per year across different sestieri.
The practical output for site managers is a deformation velocity map overlaid on cadastral data, allowing individual buildings or bridge abutments showing anomalous rates to be prioritised for structural survey. Seasonal signals driven by thermal expansion of masonry and groundwater fluctuation are separable from secular trends when the time series is long enough, typically several years. Sentinel-1's free archive from 2014 makes that baseline accessible without procurement cost.
Where coherence fails: an honest account of the method's limits
InSAR is not universally applicable. Coherence, the statistical similarity of phase between two acquisitions, collapses wherever the surface changes between passes. Vegetation is the primary culprit: even a light wind during acquisition is enough to randomise the phase return from a canopy, making interferometry impossible without very short repeat intervals. Tropical sites with dense tree cover, including much of Angkor's surrounding forest, produce coherent results only over cleared ground, masonry surfaces and water bodies under specific conditions.
Soil disturbance, including ploughing, erosion and construction, destroys coherence equally effectively. Sites in active agricultural landscapes or undergoing consolidation works will have gaps in the deformation record precisely when activity is highest. Atmospheric water vapour introduces phase delays that can mimic displacement signals of several centimetres; correcting for this requires either external weather data or statistical filtering across a large interferogram stack. Finally, InSAR measures displacement only in the satellite's line-of-sight direction. Decomposing that into vertical and horizontal components requires combining ascending and descending pass data, which doubles the data requirement and is not always geometrically feasible at sites with steep topography.
From deformation map to conservation decision
A deformation velocity map is not itself a conservation plan. The analytic value comes from integrating InSAR output with ground truth: borehole logs, geotechnical surveys, structural inspection records and historical damage documentation. Where InSAR flags an anomalous subsidence rate, the next question is always whether the cause is groundwater, load redistribution, foundation failure or something else entirely. Satellite data narrows the search area and sets the timeline; it does not replace the engineer on site.
Satellize can run PSI and SBAS processing chains over Sentinel-1 open-archive data and commission commercial tasking from COSMO-SkyMed or TerraSAR-X where higher spatial resolution or more frequent revisit is needed. Delivered outputs are georeferenced deformation velocity rasters and time-series point datasets in standard GIS formats, accompanied by a written interpretation note. The approach is the same whether the site is a single standing monument or a landscape-scale heritage zone. For organisations wanting to understand what a first analysis would show before committing to a monitoring programme, a single-site feasibility assessment using the existing Sentinel-1 archive is a practical starting point.
Typical figures
| Spatial resolution (Sentinel-1 IW mode) | 5 x 20 m ground range; PS analysis localises to individual pixel scatterers within that footprint |
| Spatial resolution (COSMO-SkyMed Spotlight) | ~1 m; TerraSAR-X Staring Spotlight sub-metre |
| Revisit interval | 6 days (Sentinel-1, two-satellite); 12 hours minimum (COSMO-SkyMed tasked); ~11 days (TerraSAR-X) |
| Minimum detectable displacement rate (PSI, stable urban surface) | 0.5-1 mm per year under good coherence conditions; ~2-3 mm per year for SBAS over distributed scatterers |
| Radar frequency / wavelength | C-band 5.6 cm (Sentinel-1, RADARSAT-2); X-band 3.1 cm (COSMO-SkyMed, TerraSAR-X) |
| Archive depth | Sentinel-1 from April 2014; ERS/Envisat legacy archive extends to 1992 for some geometries |
| Atmospheric correction | ERA5 reanalysis or GACOS service used to reduce tropospheric phase delay artefacts; residual error typically 2-5 mm per interferogram |
| Displacement measurement geometry | Line-of-sight; vertical/horizontal decomposition requires ascending + descending combination |
| Minimum interferogram stack for PSI | Typically 20-30 acquisitions; SBAS operable from ~10 with careful baseline selection |
| Delivery formats | GeoTIFF deformation velocity rasters, CSV/shapefile PS point datasets, NetCDF time series, PDF interpretation report |
Analytics Satellize can run
| Deformation velocity map | Persistent-scatterer interferometry (PSI) over multi-year Sentinel-1 or COSMO-SkyMed stack | GeoTIFF raster and PS point shapefile showing mean annual displacement rate per scatterer, with uncertainty estimate |
| Displacement time series per structure | SBAS network inversion or PS time-series extraction | CSV and chart showing cumulative displacement per monitoring point across the full archive period, exportable to structural monitoring dashboards |
| Differential settlement risk ranking | Spatial clustering of PS velocity anomalies against monument footprint polygons | Ranked site list with anomaly magnitude and location, formatted for conservation priority review |
| Seasonal versus secular signal separation | Harmonic decomposition of PS time series to isolate annual thermal and hydrological cycles from long-term trend | Two-layer GIS output: secular velocity map and seasonal amplitude map, with written interpretation note |
| Coherence loss change detection | Multitemporal coherence map differencing to flag surface disturbance events between acquisitions | Alert layer identifying areas of sudden coherence loss, useful for detecting unplanned ground disturbance within a heritage buffer zone |
| Feasibility assessment for new sites | Retrospective coherence and PS density analysis over existing Sentinel-1 archive | Written report stating whether site character supports PSI, SBAS or neither, with archive-based sample interferogram and preliminary velocity estimate |
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.