Permafrost and ground-stability assessment for tower foundation siting

InSAR-derived displacement time series reveal subsidence, frost heave and slow landslide creep at candidate tower sites before a foundation is poured. Satellite radar detects surface motion; it does not replace the geotechnical bore.

Sensors

Sentinel-1 (C-band SAR, ESA): 5.6 cm wavelength, 12-day exact repeat in Interferometric Wide Swath mode, 250 km swath, 5 × 20 m ground range resolution. Standard workhorse for permafrost InSAR: archive from 2014, free access, global coverage. Coherence degrades over vegetated or snow-covered surfaces.
ALOS PALSAR-2 (L-band SAR, JAXA): 23.6 cm wavelength penetrates vegetation canopy and maintains coherence over forested or snow-covered terrain where C-band loses phase. 14-day repeat, 3–10 m resolution in spotlight mode, 70 m in ScanSAR. Better suited to boreal and alpine sites with dense ground cover.
TanDEM-X (X-band SAR, DLR/Airbus): Bistatic interferometry from two satellites flying in close formation. Primary use here is generating a high-quality baseline DEM (12 m posting, ~2 m vertical accuracy) for removing topographic phase from displacement stacks. Not a repeat-pass displacement sensor in routine commercial operation.
ICEYE SAR constellation (X-band): Sub-metre spotlight resolution, flexible tasking, revisit of 1–3 days over a target area by combining multiple satellites. Useful for urgent site checks or for resolving ambiguities flagged by Sentinel-1 analysis. Commercial acquisition; costs apply per tasking.

Why tower foundations fail in cold regions

Permafrost underlies roughly 25 percent of the Earth's land surface. It is not a monolith. Active-layer thickness, ice content, taliks and thermokarst features vary over distances of tens of metres, which is precisely the scale that matters when siting a tower foundation. A pad that sits on ice-rich permafrost will subside as ground temperatures rise; one that straddles a talik may tilt asymmetrically. In mountainous terrain, slow-moving debris flows and solifluction lobes add a lateral component to the problem.

Conventional site surveys sample a handful of bore locations, which is economical but spatially sparse. A tower site covering perhaps 20 × 20 metres can conceal a frost polygon boundary or the toe of a creeping lobe that a single bore misses entirely. Satellite radar interferometry does not replace the bore, but it maps the spatial pattern of motion across a candidate site and its surroundings before the drill rig is mobilised, directing it to the right spots.

What differential phase actually measures, and what it does not

Repeat-pass InSAR works by comparing the phase of radar echoes from two or more satellite passes over the same area. Any change in the distance between sensor and ground between passes shifts the phase. For Sentinel-1 at C-band, one full phase cycle corresponds to half the wavelength, roughly 2.8 cm of line-of-sight displacement. With careful processing, sub-centimetre and in favourable conditions millimetre-scale displacement rates become detectable over time series of months to years.

The critical limitation is that InSAR measures surface displacement in the radar line-of-sight direction. It cannot see subsurface ice content, soil composition or pore-water pressure. A site showing zero surface motion could still have high ice content that will mobilise under future thermal loading. Conversely, a site showing seasonal frost heave of 2–3 cm per winter cycle may be behaving entirely predictably and engineering solutions may be straightforward. The displacement map is evidence; the interpretation requires a geotechnical engineer who understands the local geology.

Coherence is the other honest caveat. Dense vegetation, fresh snow and rapidly changing soil moisture all decorrelate the radar signal, destroying the phase relationship between passes. L-band (ALOS PALSAR-2) is substantially more coherent than C-band over forested and snow-covered terrain, but even L-band loses coherence under deep snowpack. Processing windows must be chosen to avoid periods of rapid surface change, and sites where coherence is persistently low simply cannot be assessed by InSAR alone.

From raw interferograms to a displacement time series

A single interferogram is rarely sufficient. Atmospheric water vapour introduces apparent path-length changes of up to several centimetres, easily mistaken for ground motion. The standard mitigation is to stack many interferograms, either through Small Baseline Subset (SBAS) processing or through Persistent Scatterer InSAR (PS-InSAR). SBAS averages over many image pairs, suppressing atmospheric noise. PS-InSAR identifies individual coherent point scatterers, such as boulders, exposed rock faces or man-made objects, and tracks their phase history with high precision. In open tundra, natural PS targets are sparse; SBAS is typically preferred.

For a tower-siting study, the output is a displacement velocity map (millimetres per year) and, where the time series is long enough, a seasonal decomposition separating annual frost heave and thaw subsidence cycles from any secular trend. A site with a clear downward secular trend, say 5–15 mm per year of subsidence, warrants immediate geotechnical investigation. A site with symmetric seasonal motion and no trend is a different risk profile entirely.

Matching sensor choice to site conditions

For open tundra above the tree line, Sentinel-1 is usually the first choice: free archive back to 2014, 12-day repeat, and reasonable coherence over bare or low-vegetation surfaces. Nine years of archive means a displacement time series can be extracted before a single field visit, giving the network planner a ground-motion history that predates the project by years.

For boreal forest or alpine sites with significant snow cover, ALOS PALSAR-2 L-band is the better starting point, though it carries acquisition costs and the archive is less dense than Sentinel-1. Where a candidate site shows ambiguous or borderline motion in the Sentinel-1 stack, ICEYE tasking can provide high-resolution imagery within days, useful for checking whether an anomaly is a genuine ground feature or an artefact of a single decorrelated pass.

TanDEM-X contributes the baseline DEM. Topographic phase removal is not trivial in steep terrain: errors in the reference DEM propagate directly into the displacement estimate. Using a TanDEM-X DEM at 12 m posting rather than a coarser global DEM reduces this error source materially in mountainous sites.

Practical outputs for a network planning team

The deliverable a network planner actually needs is not a stack of interferograms. It is a ranked site-risk assessment: which candidate locations show active motion, at what rate, in what direction, and over what spatial extent. A GIS layer showing mean displacement velocity, overlaid on a high-resolution optical basemap, lets a site-acquisition team make a first cut before any field visit. Sites with velocity magnitudes below a defined threshold, say 2 mm per year, and no coherent seasonal anomaly, can be provisionally cleared for standard geotechnical investigation. Sites above threshold are flagged for detailed bore programmes or, in some cases, eliminated from the candidate list.

Satellize runs SBAS and PS-InSAR processing pipelines on Sentinel-1 and ALOS PALSAR-2 archives and can deliver displacement velocity maps, seasonal decomposition layers and site-ranked risk reports as GIS-compatible outputs. The Tonga crop-estimation programme demonstrated the team's ability to run time-series analytics on open-constellation data in island and remote-territory contexts where field access is constrained. The same operational model applies to high-latitude tower siting.

What InSAR cannot resolve, and what must follow

A clean InSAR result, showing no detectable motion at a candidate site, is not a foundation clearance certificate. It means the surface has not moved detectably during the observation window, at the spatial resolution of the analysis. Ice-rich permafrost that has been stable for decades can mobilise rapidly under a point thermal load such as a heated equipment shelter or a piled foundation that conducts heat into the ground. Thermal modelling of the active layer, informed by ground-temperature measurements, is a separate and mandatory step.

Regulatory requirements in most permafrost jurisdictions, including Canada, Russia, Norway and Alaska, mandate geotechnical investigation before construction regardless of remote-sensing results. The satellite analysis is best framed as a pre-screening tool that makes the mandatory geotechnical programme more targeted and therefore more cost-effective, not as a substitute for it. That framing is also the honest one.

Typical figures

Sentinel-1 ground range resolution (IW mode)	5 m × 20 m (range × azimuth)
Sentinel-1 repeat interval	12 days (single satellite); 6 days with two satellites where both are operational
ALOS PALSAR-2 resolution (Stripmap mode)	3–10 m, depending on mode; 14-day repeat
TanDEM-X DEM posting and vertical accuracy	12 m posting; ~2 m absolute vertical accuracy (published DLR specification)
Minimum detectable displacement rate (favourable conditions)	~1–2 mm per year over multi-year SBAS or PS-InSAR stacks; degrades with coherence loss
Sentinel-1 archive depth	From April 2014 (Sentinel-1A); continuous to present
ALOS PALSAR-2 archive depth	From 2014; predecessor ALOS PALSAR from 2006
Radar frequency / wavelength	C-band: 5.405 GHz / 5.6 cm (Sentinel-1); L-band: 1.27 GHz / 23.6 cm (PALSAR-2); X-band: ~9.6 GHz / 3.1 cm (ICEYE, TanDEM-X)
Typical processing latency for site assessment	2–4 weeks for archive-based SBAS time series; faster with pre-processed Sentinel-1 products
Delivery formats	GeoTIFF displacement velocity maps, CSV time-series per site, GeoPackage or Shapefile risk-ranked site layer, PDF site report

Analytics Satellize can run

Mean displacement velocity map	SBAS InSAR time-series processing on Sentinel-1 IW stack; atmospheric correction via spatial filtering or ERA5 tropospheric model	GeoTIFF raster, mm/year line-of-sight velocity, covering candidate site cluster and 2 km buffer
Seasonal frost heave / thaw subsidence decomposition	Harmonic regression on displacement time series to separate annual sinusoidal component from secular trend	Per-site time-series CSV with amplitude, phase and linear trend parameters; PDF summary chart
Site-ranked ground-motion risk table	Threshold classification of velocity magnitude and trend significance; spatial clustering of anomalous pixels within each candidate site polygon	GeoPackage attribute table with risk tier (low / moderate / high) per candidate site; exportable to standard site-acquisition databases
L-band coherence and displacement stack (forested / snow-covered sites)	ALOS PALSAR-2 SBAS processing; coherence map used as data-quality mask before velocity estimation	GeoTIFF velocity and coherence layers; flagging of sites where coherence is insufficient for reliable displacement estimation
High-resolution anomaly confirmation imagery	ICEYE spotlight SAR tasking over flagged sites; amplitude change detection and single-pair interferogram	Sub-metre SAR amplitude image with annotated displacement anomaly, delivered within agreed tasking window
Baseline DEM for topographic phase removal	TanDEM-X global DEM ingestion; used as reference surface in all interferometric processing chains	Clipped and reprojected DEM GeoTIFF for project area, with vertical datum documentation

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.