Rock glacier creep rate and permafrost ice content estimation
Rock glaciers move centimetres to metres per year, and that velocity encodes permafrost temperature and ice content. InSAR and SAR offset tracking on Sentinel-1 and ALOS-2 PALSAR-2 turn surface motion into a structural and thermal proxy.
Sensors
- Sentinel-1 A/B (C-band SAR, ESA): 6- to 12-day repeat at mid-latitudes; 5 × 20 m IW mode resolution. C-band coherence degrades over vegetated or snow-covered surfaces, but bare debris maintains coherence well enough for differential InSAR detecting millimetre-scale displacement per cycle. Free and open archive from 2014.
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): L-band (23.6 cm wavelength) penetrates thin snow cover and maintains coherence over longer baselines than C-band, making it better suited to slow-moving or partially vegetated rock glaciers. Stripmap mode delivers 3 × 3 m resolution; 14-day nominal repeat, though tasking is commercial.
- TerraSAR-X / TanDEM-X (X-band SAR, Airbus/DLR): 11-day repeat, Stripmap at 3 m or Spotlight at ~1 m resolution. X-band loses coherence quickly over rough debris in wet conditions, but its fine resolution resolves individual flow units and shear margins that C-band blurs. Useful for fast-moving fronts exceeding ~50 cm per year.
- Copernicus DEM (GLO-30 and GLO-90): Global 30 m and 90 m DEMs derived from TanDEM-X interferometry, publicly released. Used as the reference surface for InSAR processing and for multi-epoch DEM differencing to detect volume change and snout advance. Vertical accuracy typically better than 4 m at 90th percentile over mountainous terrain.
What a rock glacier's surface velocity actually tells you
A rock glacier is not simply a pile of boulders sliding on ice. It is a creeping permafrost body whose internal deformation follows a power-law rheology sensitive to ice temperature, ice content, and pore-water pressure. Velocity is therefore a proxy variable. A rock glacier accelerating from 20 cm per year to 60 cm per year over a decade is not just moving faster; it is telling you that ground temperatures have risen, ice has softened, or both. Published work in the Swiss Alps has linked interannual velocity anomalies to summer air-temperature anomalies with a lag of one to three years, consistent with thermal diffusion depths of several metres.
The velocity field also encodes internal structure. A rock glacier with a single fast-moving central tongue and slow lateral margins has a different ice distribution from one with uniform creep across its width. Differential velocity across the surface, measured by InSAR or pixel-offset tracking, maps shear zones, compression ridges and extensional crevasse-like features that are otherwise invisible from the surface. That structural information is what allows ice-content estimation: high-velocity zones in cold permafrost imply high ice fractions, often 40–70 % by volume in alpine rock glaciers according to borehole studies in the Alps and Andes.
Two measurement techniques, one honest comparison
Differential InSAR (D-InSAR) measures the phase difference between two SAR acquisitions to detect line-of-sight displacement with sub-centimetre precision per interferogram. For a rock glacier moving at 5–50 cm per year, a 12-day Sentinel-1 pair captures 1.6–16 mm of motion, which sits comfortably above the noise floor on coherent debris. The technique works best on slow-moving, coherent surfaces. It fails when displacement between passes exceeds roughly a quarter of the radar wavelength (about 1.4 cm for C-band over 6 days), causing phase wrapping that is difficult to unwrap reliably on complex terrain.
SAR offset tracking (also called pixel-tracking or intensity cross-correlation) does not use phase at all. It cross-correlates intensity patches between two images to find the sub-pixel shift. Precision is typically 1/20th to 1/30th of the pixel size, so about 15–30 cm for Sentinel-1 IW mode. That makes it useless for slow creep but essential for fast-moving fronts or surge events where InSAR loses coherence entirely. ALOS-2's L-band and finer pixel size push offset-tracking precision to a few centimetres. The two methods are complementary: InSAR for the slow interior, offset tracking for the fast snout and for validation.
Limits the physics imposes, and limits the archive imposes
Snow cover is the most persistent problem. C-band coherence collapses under even a few centimetres of wet snow, and most alpine rock glaciers are snow-covered for four to seven months per year. Practical measurement windows are typically July through September at high elevations. L-band is more tolerant of dry snow but not immune. Seasonal gaps mean annual velocity estimates rest on a handful of summer interferograms, and a single anomalous pair corrupted by late snowfall can bias the result substantially.
Geometric distortions in steep terrain are a second hard limit. SAR images in mountainous areas suffer foreshortening, layover and shadow. Slopes facing away from the satellite are often invisible, and the ascending and descending geometries together still leave some aspects unobserved. Decomposing line-of-sight displacement into true three-dimensional velocity vectors requires at least two look directions, and even then the east-west and vertical components are recovered more reliably than north-south motion. Rock glaciers oriented north-south relative to the satellite track present the worst geometry.
Archive depth is an asset that is often underused. Sentinel-1 data from 2014 onward, combined with ERS and Envisat archives extending to the 1990s, allow velocity time series spanning 25–30 years on some targets. That temporal baseline is long enough to detect acceleration trends attributable to climate forcing rather than interannual variability. The ERS/Envisat data require careful orbit and atmospheric correction, but the signal-to-noise ratio is acceptable on the largest, fastest rock glaciers.
From velocity map to ice-content estimate
Converting a surface velocity field into an ice-content estimate requires a model. The most widely applied approach uses Glen's flow law for ice-rich permafrost, fitting measured surface velocities to a depth-integrated deformation model to back-calculate the effective ice viscosity and, from that, the ice fraction. The method was applied to rock glaciers in the Andes by Monnier and Kinnard and to Alpine sites in published studies using ALOS-1 and Sentinel-1 data. Results are sensitive to assumed basal boundary conditions and ice temperature, so outputs carry uncertainty ranges of ±10–20 percentage points in ice content.
DEM differencing adds a complementary volume-change signal. Comparing the Copernicus GLO-30 DEM against older photogrammetric or SRTM surfaces reveals surface lowering or advance at the snout. Combined with velocity, this constrains mass flux and, indirectly, ice loss rates. The approach is honest about its limits: DEM vertical errors of 2–5 m over rough terrain mean that volume-change signals smaller than a few tens of thousands of cubic metres are not reliably detectable.
Practical outputs and where Satellize fits
A monitoring programme for a single rock glacier typically produces annual velocity maps at 20–50 m posting, a time series of mean and peak velocity with uncertainty bounds, a structural map of shear zones and compression ridges, and a qualitative ice-content zonation derived from the velocity field. For hazard applications, the critical output is an acceleration alert: a statistically significant departure from the multi-year trend that may indicate destabilisation of the snout or a proglacial hazard.
Satellize runs this processing chain on open Sentinel-1 data and adds commercial ALOS-2 or TerraSAR-X tasking where the geometry or coherence demands it. The analytics follow published InSAR and offset-tracking methods, with atmospheric correction using ERA5 reanalysis or GACOS products. Deliverables are GIS-ready rasters and time-series tables, not black-box scores. The same infrastructure that supports the Tonga crop-estimation programme handles multi-site mountain monitoring; the sensor stack and processing logic differ, but the principle of running sovereign analytics on open data does not.
Typical figures
| Spatial resolution (InSAR, Sentinel-1 IW) | 5 × 20 m single-look; typically multilooked to 20–40 m for phase quality |
| Spatial resolution (ALOS-2 Stripmap) | 3 × 3 m; multilooked to 10–20 m for interferometry |
| Minimum detectable displacement (D-InSAR, C-band) | ~3–5 mm line-of-sight per 6-day pair on coherent debris; ~1 cm practical threshold with atmospheric noise |
| Minimum detectable displacement (offset tracking, Sentinel-1) | ~15–30 cm; suited to velocities above ~50 cm per year |
| Revisit period | 6 days (Sentinel-1 A+B combined, mid-latitudes); 14 days (ALOS-2); 11 days (TerraSAR-X) |
| Measurement window (alpine) | Typically July–September for coherent C-band; L-band extends window by 4–6 weeks |
| Archive depth | Sentinel-1 from 2014; ERS/Envisat from ~1992 (selected targets); ALOS-1 from 2006 |
| DEM vertical accuracy (Copernicus GLO-30) | Better than 4 m at 90th percentile over mountainous terrain |
| Velocity uncertainty (annual composite) | ±1–3 cm per year for InSAR time series; ±5–15 cm per year for offset tracking |
| Delivery formats | GeoTIFF velocity rasters, CSV/NetCDF time series, GeoPackage structural maps, PDF technical report |
Analytics Satellize can run
| Annual surface velocity map | Multi-temporal D-InSAR or SBAS time-series analysis on Sentinel-1 or ALOS-2 interferometric pairs | GeoTIFF raster of mean annual line-of-sight and decomposed horizontal velocity, with uncertainty layer |
| Velocity time series and trend detection | SBAS or PS-InSAR stacking over multi-year archive; Mann-Kendall trend test on annual composites | CSV/NetCDF time series per rock glacier polygon; PDF chart with trend line and confidence interval |
| Acceleration alert | Statistical threshold on rolling 3-year velocity anomaly relative to baseline; triggered when departure exceeds 2-sigma | Email or API alert with map attachment and supporting interferogram mosaic |
| Structural zonation map | Differential velocity gradient analysis to delineate shear margins, compression ridges and extensional zones | GeoPackage polygon layer with zone classification and velocity statistics per zone |
| Qualitative ice-content zonation | Velocity-to-ice-fraction inversion using Glen's flow law for ice-rich permafrost, following published Alpine and Andean parameterisations | GeoTIFF ice-content index raster with documented uncertainty range (±10–20 percentage points) |
| Snout advance and volume-change estimate | Multi-epoch DEM differencing (Copernicus GLO-30 vs. SRTM or archival photogrammetry) combined with flux-gate velocity | Tabular volume-change estimate with detection-limit caveat; snout polygon time series in GeoPackage |
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.