InSAR permafrost thaw subsidence monitoring
Differential SAR interferometry detects millimetre-scale surface subsidence caused by permafrost active-layer thickening, using phase shifts between repeat passes. L-band systems maintain coherence where C-band fails, but separating frost heave from thaw settlement demands careful seasonal analysis.
Sensors
- ALOS-2 PALSAR-2: L-band (1.27 GHz), 3–10 m resolution in stripmap mode, 14-day repeat cycle. L-band penetrates vegetation canopy and maintains interferometric coherence over Arctic tundra through the growing season, making it the current workhorse for permafrost InSAR.
- NISAR (forthcoming, expected 2025): NASA-ISRO dual-frequency L+S band mission. L-band at 24 cm wavelength with 12-day repeat; S-band adds a second coherence window. Designed explicitly to support permafrost and ice-sheet deformation science at global scale.
- Sentinel-1 (C-band): 5.405 GHz, 5–20 m resolution, 6-day repeat over Arctic regions with both satellites. C-band loses coherence rapidly over vegetated tundra between summer passes, limiting reliable InSAR to winter or early-spring acquisitions when the surface is frozen and stable.
- TanDEM-X: X-band bistatic pair producing the TanDEM-X global DEM at 12 m posting. Used to establish a high-quality baseline digital elevation model against which multi-year subsidence trends are referenced, not for repeat-pass deformation directly.
What the phase shift is actually measuring
InSAR works by comparing the phase of radar echoes from two passes over the same ground. Where the surface has moved toward or away from the satellite between acquisitions, the round-trip path length changes, and that change appears as a phase difference in the interferogram. At L-band (wavelength roughly 24 cm), one full phase cycle corresponds to about 12 cm of range displacement. Subsidence of 5 mm is therefore a small but detectable fraction of a cycle, provided the signal is not buried in noise.
In permafrost terrain the displacement is vertical: the active layer above the permafrost table thaws each summer, the ice within it melts, and the ground surface drops. In winter the refreezing layer expands again, pushing the surface back up. The net annual signal is a superposition of seasonal heave and settlement on top of any long-term trend driven by climate-forced deepening of the active layer. Separating these components requires time series of interferograms spanning multiple seasons, not a single pair.
Why C-band loses the argument in summer
Coherence is the enemy of error in InSAR. If the scattering properties of the surface change between passes, the phase comparison becomes meaningless. Vegetation is the main coherence killer in Arctic tundra: grasses, sedges and low shrubs move with wind and grow between acquisitions. At C-band (5.4 GHz, wavelength about 5.6 cm), the radar interacts strongly with this surface layer. Over a 6-day Sentinel-1 repeat in July, coherence over vegetated tundra routinely falls below 0.3, the threshold below which phase measurements carry more noise than signal.
L-band at 24 cm wavelength penetrates the vegetation canopy and scatters preferentially from the soil surface and near-surface volume beneath. ALOS-2 PALSAR-2 studies over sites in Alaska and Siberia have demonstrated coherence values above 0.6 through the summer growing season, sufficient for reliable phase unwrapping. The practical consequence is that C-band InSAR over permafrost is largely restricted to winter acquisitions, which capture frost heave well but miss the summer thaw signal entirely. L-band captures both.
The annual cycle: heave, settlement and the ambiguity between them
A single interferogram spanning spring to autumn will show net subsidence in areas of active thaw. But the measured displacement is the algebraic sum of upward frost heave in spring, downward thaw settlement through summer, and any residual long-term trend. If the acquisition window does not bracket the full seasonal cycle, the analyst is looking at a partial integral and can misread it.
Published work using ALOS PALSAR time series over the North Slope of Alaska (notably the work associated with the Circumpolar Active Layer Monitoring network) has shown that peak seasonal amplitude in areas of ice-rich permafrost can reach 3–5 cm vertically, while multi-year linear trends from active-layer deepening are typically 1–10 mm per year. The trend signal is therefore smaller than the seasonal oscillation by roughly an order of magnitude, which means it only becomes detectable after several years of consistent acquisitions. Short archive gaps, changes in satellite look angle between missions, or atmospheric delay artefacts can all introduce apparent trends that are not real. Tropospheric water vapour is a particular problem at high latitudes where weather systems are rapid and spatially variable.
NISAR and what a dedicated mission changes
NISAR, the NASA-ISRO Synthetic Aperture Radar mission, is designed to address exactly these limitations. Its 12-day L-band repeat at global coverage, combined with a simultaneous S-band channel, will provide a consistent long-term archive that no single commercial or national mission has yet delivered for permafrost monitoring. The dual-frequency design means analysts can compare coherence and phase between bands to separate surface and subsurface scattering contributions.
The mission is also committed to open data access, which matters for the kind of multi-year baseline studies permafrost science requires. Gaps in the ALOS-1 to ALOS-2 transition (ALOS-1 ended in 2011, ALOS-2 launched in 2014) already complicate trend analysis for that decade. A continuous NISAR archive starting from its planned 2025 launch will eventually provide the decade-scale record needed to distinguish climate-driven active-layer deepening from interannual variability.
Practical limits any buyer should understand
InSAR measures displacement in the satellite's line-of-sight direction, not purely vertical. Converting to vertical requires knowing the local incidence angle and assuming horizontal displacement is negligible, which is generally reasonable for permafrost subsidence but less so near slopes or thermokarst margins where lateral movement occurs. Phase unwrapping, the process of resolving the integer-cycle ambiguity in the phase signal, can fail in areas of rapid spatial gradient or low coherence, producing unwrapping errors that propagate across the scene.
Atmospheric correction is non-trivial. The GNSS-based tropospheric delay products used in tectonic InSAR are sparse at high latitudes where GNSS networks are thin. ERA5 reanalysis atmospheric models are commonly used instead, but their spatial resolution (roughly 31 km) is coarser than the deformation features of interest. Expect residual atmospheric noise of 1–2 cm in individual interferograms; it is only through time-series averaging (small baseline subset or persistent scatterer methods) that the noise floor approaches the millimetre level. Dense urban infrastructure produces persistent scatterers readily; open tundra does not, so the PS approach is less effective there than small-baseline stacking.
Satellize runs permafrost subsidence analytics on ALOS-2 and Sentinel-1 archives and will integrate NISAR data under its open-data licence on launch. The Tonga crop-estimation programme is a different domain, but the underlying time-series processing pipeline is shared infrastructure.
From interferogram to infrastructure decision
The end product for most government clients is not a stack of interferograms but a displacement velocity map: a raster showing millimetres per year of vertical motion, with uncertainty bounds, delivered as a GeoTIFF or vector polygon layer that planners can overlay on infrastructure footprints. Roads, pipelines, airstrips and building foundations in permafrost regions are all sensitive to differential settlement at the centimetre scale over years.
A velocity map updated annually, with seasonal decomposition showing the summer thaw amplitude separately from the long-term trend, gives infrastructure managers a quantitative basis for prioritising inspection and maintenance. The honest caveat is that InSAR shows where the ground is moving, not why. Ground-truth from active-layer thickness measurements or soil-ice content surveys is needed to convert a displacement signal into a geotechnical risk rating. Space data narrows the search area; it does not replace the borehole.
Typical figures
| Spatial resolution (ALOS-2 stripmap) | 3 m range × 3 m azimuth (fine mode); 10 m in standard stripmap |
| Spatial resolution (Sentinel-1 IW) | 5 m range × 20 m azimuth |
| Repeat interval | 14 days (ALOS-2); 6 days (Sentinel-1, dual satellite); 12 days (NISAR, planned) |
| Radar frequency / wavelength | L-band 1.27 GHz / 24 cm (ALOS-2, NISAR); C-band 5.4 GHz / 5.6 cm (Sentinel-1) |
| Minimum detectable displacement (time series) | ~1–3 mm per year after multi-interferogram stacking; single-pair noise floor ~1–2 cm |
| Seasonal amplitude range (published Arctic sites) | 1–5 cm vertical, ice-rich tundra; <1 cm, drained or sandy sites |
| Archive depth | ALOS-1 from 2006; ALOS-2 from 2014; Sentinel-1 from 2014; gap between ALOS-1 and ALOS-2 complicates decade-scale trends |
| Delivery formats | GeoTIFF displacement velocity raster, vector polygon risk zones, CSV time-series per AOI |
| Atmospheric correction approach | ERA5 tropospheric delay model (31 km resolution); residual error 1–2 cm per interferogram |
| Coverage | Global for Sentinel-1 and NISAR; ALOS-2 tasked on request, Arctic coverage prioritised by JAXA science programme |
Analytics Satellize can run
| Annual displacement velocity map | Small baseline subset (SBAS) time-series inversion across multi-year interferogram stack | GeoTIFF raster, millimetres per year vertical, with per-pixel uncertainty, updated annually |
| Seasonal amplitude decomposition | Sinusoidal model fit to displacement time series separating annual heave-settlement cycle from linear trend | Two-layer GeoTIFF: summer thaw amplitude and long-term trend velocity, with confidence intervals |
| Infrastructure settlement risk layer | Overlay of velocity raster on client-supplied infrastructure footprints; threshold classification by settlement rate | Vector polygon GIS layer with risk-tier attribution (low / moderate / high), suitable for GIS or asset-management platforms |
| Coherence quality map | Per-pixel mean coherence across seasonal stack, flagging areas where phase measurements are unreliable | GeoTIFF masking low-coherence zones; advisory note on which areas require ground validation |
| Change alert: anomalous subsidence onset | Statistical threshold detection on rolling 3-interferogram displacement increment against site baseline | Email or API alert with coordinates, magnitude estimate and flagged interferogram thumbnail |
| Multi-mission trend reconciliation report | Cross-calibration of ALOS-1, ALOS-2 and Sentinel-1 time series using overlapping acquisition windows and stable reference targets | PDF technical report with decade-scale displacement history per AOI and documented uncertainty budget |
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.