Highway cut-slope stability monitoring during road construction
Freshly excavated cut slopes are the leading cause of construction-phase landslide fatalities on mountain roads. Short-interval SAR coherence and offset-tracking methods can detect millimetre-scale pre-failure creep weeks before collapse, but geometry and shadow zones impose real limits every project must plan around.
Sensors
- Sentinel-1 (ESA, C-band, 5.6 cm wavelength): Interferometric Wide Swath mode delivers 5 × 20 m ground range resolution across a 250 km swath. Repeat pass every 6 days over most land areas (12 days from a single satellite). Free and open archive from 2014. C-band coherence degrades rapidly on vegetated or freshly disturbed surfaces, which is actually diagnostic: a coherence drop on a previously stable slope segment is itself an early-warning signal.
- ALOS-2 PALSAR-2 (JAXA, L-band, 23.6 cm wavelength): L-band penetrates sparse vegetation and maintains coherence over longer intervals than C-band, making it better suited to slopes with residual scrub cover. Stripmap mode achieves 3 × 3 m resolution; High-Sensitivity mode reaches 6 m. Nominal 14-day repeat, though off-nadir tasking flexibility allows shorter revisit by combining ascending and descending acquisitions. Archive extends to 2014.
- COSMO-SkyMed (ASI, X-band, 3.1 cm wavelength): X-band delivers the finest spatial resolution of the operational SAR fleet: Spotlight mode reaches 1 m, Stripmap 3–15 m depending on mode. Very short coherence windows (1–3 days) are achievable through the four-satellite constellation, which matters on rapidly evolving construction slopes. Commercial tasking required; archive depth and pricing are mission-specific.
- TerraSAR-X / TanDEM-X (Airbus, X-band, 3.1 cm wavelength): Staring Spotlight mode achieves better than 0.25 m azimuth resolution, enabling sub-centimetre line-of-sight displacement measurement via persistent scatterer InSAR on engineered slope features such as drainage channels and rock bolts. Revisit of 11 days from a single satellite; commercial tasking allows more frequent acquisitions. TanDEM-X bistatic pairs can generate a baseline DEM at 12 m posting for geometric correction.
Why cut slopes fail during construction, not after
When a road contractor blasts and excavates into a hillside, the slope that remains is not in equilibrium. Stress redistribution begins immediately. Water infiltrates freshly exposed joints. Clay-rich layers that were previously confined begin to creep. The critical window is the first wet season after excavation, before drainage works are complete and before any surface treatment has been applied. This is precisely when monitoring is least likely to be in place.
The failure mode is rarely sudden. Most cut-slope collapses are preceded by days or weeks of measurable displacement, typically in the range of a few millimetres per day accelerating to centimetres per day. That precursor signal is detectable by radar interferometry, provided the geometry cooperates and acquisitions are frequent enough to catch the acceleration before phase wrapping makes the data uninterpretable.
What a radar satellite actually measures on a slope
SAR interferometry measures the change in round-trip travel time between the satellite and the ground surface between two passes. Each 2π phase cycle corresponds to half the radar wavelength of displacement in the satellite's line of sight (LOS). For Sentinel-1 C-band, one full fringe represents 2.8 cm of LOS movement. For ALOS-2 L-band, the equivalent is 11.8 cm. That larger ambiguity interval makes L-band more forgiving on fast-moving slopes but less sensitive to very slow creep.
Offset tracking, the alternative to phase-based InSAR, cross-correlates amplitude patches between two SAR images to estimate displacement. It does not require phase coherence, so it works on slopes where vegetation or surface disturbance has destroyed the interferometric signal. The trade-off is sensitivity: offset tracking typically resolves displacements of one-tenth to one-twentieth of a pixel, which means roughly 0.5–2 m for Sentinel-1 Stripmap and around 0.1 m for COSMO-SkyMed Spotlight. It catches large, fast movements that InSAR cannot unwrap, not the millimetre-scale precursor creep.
Persistent Scatterer InSAR (PS-InSAR) and Small Baseline Subset (SBAS) methods build time series from stacks of interferograms to separate displacement signal from atmospheric noise. On construction sites, the number of stable scatterers changes week by week as machinery moves and surfaces are regraded. This limits the density of reliable PS points compared to a settled urban environment, but engineered features such as concrete drainage channels, rock-bolt plates and culvert headwalls provide consistent reflectors.
The geometry problem: shadow, layover and the slopes you cannot see
This is the most important caveat to state clearly. SAR sensors image from the side. A slope facing toward the satellite is compressed in the image (foreshortening) or, if steep enough, overlaid on the terrain behind it (layover). A slope facing away from the satellite falls in radar shadow and returns no signal at all. For a typical Sentinel-1 descending pass over a north-south mountain road in the northern hemisphere, slopes facing roughly west are well illuminated; slopes facing east may be in shadow.
The practical mitigation is to combine ascending and descending track acquisitions. Ascending and descending geometries illuminate opposite slope aspects, so a slope invisible on one geometry is often measurable on the other. Even so, very steep slopes (greater than roughly 45 degrees from vertical in the range direction) may be unmonitorable by any current orbital SAR system. On such slopes, corner reflectors or GNSS targets installed during construction provide the only reliable deformation reference. Any monitoring design that does not map shadow and layover zones against the actual slope inventory is incomplete.
Sentinel-1 coherence as a change detector, not just a displacement estimator
On a stable, dry, sparsely vegetated slope, the 6-day coherence of Sentinel-1 C-band is typically 0.4–0.7. When surface material begins to move, even at rates below the displacement sensitivity threshold, coherence drops toward zero because the scatterer geometry within each resolution cell changes between passes. A coherence map time series can therefore flag active zones before a displacement measurement is possible. This is particularly useful in the first weeks after excavation, when the surface is still settling and phase-based InSAR is unreliable.
The limitation is ambiguity. Rainfall, vegetation growth and construction machinery all also reduce coherence. Distinguishing slope instability from a passing excavator requires either very high temporal resolution (daily, which no free-access SAR constellation currently provides) or contextual reasoning from optical imagery and rainfall records. Neither is trivial in a mountain construction environment.
Designing a monitoring programme that acknowledges what satellites cannot do
Satellite SAR is most valuable as a wide-area screening tool that identifies which slope segments warrant ground investigation, not as a replacement for inclinometers, piezometers or extensometers on known problem slopes. A credible monitoring design layers the two: satellite coherence and displacement maps flag anomalies across the entire road corridor at intervals of 6–12 days; ground instruments provide continuous, high-sensitivity data on the two or three slopes that the satellite analysis identifies as highest priority.
Latency matters. The gap between SAR acquisition and a usable displacement product is typically 12–48 hours for near-real-time processing pipelines, depending on data access arrangements and processing infrastructure. For a slope showing centimetre-per-day acceleration, 48 hours is acceptable. For a slope already at the margin, it is not. The monitoring plan should specify the trigger thresholds at which satellite data alone is insufficient and ground evacuation protocols must activate.
Satellize runs coherence time-series and SBAS displacement processing on Sentinel-1 and ALOS-2 open archives, with commercial tasking added for COSMO-SkyMed or TerraSAR-X on corridors where geometry or revisit requirements demand it. The approach is the same one applied in the Tonga crop-estimation programme: open data where it is sufficient, commercial data where it is not, and honest reporting of what the geometry will and will not show.
Typical figures
| Best available SAR spatial resolution (commercial) | 0.25 m azimuth (TerraSAR-X Staring Spotlight); 1 m (COSMO-SkyMed Spotlight) |
| Free-access SAR resolution (Sentinel-1 IW) | 5 × 20 m ground range (range × azimuth) |
| Minimum detectable LOS displacement (PS-InSAR) | 1–3 mm per epoch under good coherence conditions; degrades to ~10 mm on low-coherence construction surfaces |
| Revisit interval | 6 days (Sentinel-1 two-satellite constellation over most land); 14 days (ALOS-2 nominal); 1–3 days (COSMO-SkyMed four-satellite constellation) |
| Radar frequency and wavelength | C-band 5.6 cm (Sentinel-1); L-band 23.6 cm (ALOS-2); X-band 3.1 cm (COSMO-SkyMed, TerraSAR-X) |
| Phase ambiguity interval (one fringe = half wavelength) | 2.8 cm LOS (C-band); 11.8 cm LOS (L-band); 1.55 cm LOS (X-band) |
| Offset-tracking displacement sensitivity | 0.1–2 m depending on pixel size and cross-correlation window; suited to fast-moving failures only |
| Archive depth | Sentinel-1 from April 2014; ALOS-2 from 2014; COSMO-SkyMed and TerraSAR-X commercial archives from 2007–2010 |
| Processing latency (near-real-time pipeline) | 12–48 hours post-acquisition for coherence and displacement products |
| Geometric blind zones | Slopes steeper than ~45° from vertical in range direction; shadow and layover zones mapped per track geometry before deployment |
Analytics Satellize can run
| Slope shadow and layover map | Radar geometry simulation using a project DEM (TanDEM-X 12 m or SRTM 30 m) against satellite incidence angle and track azimuth | GIS polygon layer showing monitored, foreshortened and blind zones per ascending and descending track, delivered before first acquisition |
| 6-day coherence time series | Interferometric coherence estimation from Sentinel-1 IW SLC pairs; pixel-wise temporal stack | Raster time series (GeoTIFF) and anomaly alert when coherence drops below project-defined threshold on any slope segment |
| LOS displacement time series (SBAS) | Small Baseline Subset InSAR using multi-look interferograms, atmospheric correction via ERA5 reanalysis, unwrapping by minimum cost flow | GIS point layer of displacement velocities (mm/day) updated each acquisition cycle, with uncertainty estimates per point |
| Acceleration alert | Velocity change detection on SBAS or PS time series; threshold set per slope segment based on baseline creep rate established in first 30 days | Automated alert (email or API webhook) with slope ID, current velocity, 7-day trend and recommended ground-inspection priority |
| Offset-tracking displacement map (fast-movement events) | Normalised cross-correlation of SAR amplitude patches between pre- and post-event acquisitions; applicable when phase coherence is lost | Vector displacement field (GeoTIFF and shapefile) showing magnitude and direction of movement for any slope segment with >0.5 m displacement |
| Corridor-wide stability ranking | Composite scoring of coherence trend, SBAS velocity and slope geometry (aspect, gradient, lithology proxy from optical band ratios) across all cut slopes in the project corridor | Ranked slope inventory table with monitoring priority tier, updated monthly, formatted for integration into project risk register |
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.