Landslide susceptibility mapping for buried fibre route risk assessment
Combining InSAR displacement history, high-resolution terrain, lithology proxies and rainfall climatology, slope-unit susceptibility scoring identifies buried fibre corridors at chronic risk of ground movement before a cable is ever trenched.
Sensors
- TanDEM-X WorldDEM (DLR/Airbus): Global digital elevation model at 12 m posting (0.4 m relative vertical accuracy over slopes), used to derive slope angle, aspect, curvature and topographic wetness index per slope unit. The 12 m resolution is sufficient for routing corridors but may miss small gully heads; lidar ground-truth is advisable on critical segments.
- Sentinel-1 C-band SAR (ESA): Repeat-pass InSAR at 6-day revisit (12-day for a single geometry) over most land surfaces, 5 x 20 m ground resolution in IW mode. Persistent-scatterer and SBAS processing detects line-of-sight displacement at millimetre scale over coherent targets. Dense vegetation and steep layover reduce coherence; C-band loses it faster than L-band in tropical forest.
- GPM IMERG (NASA/JAXA): Global Precipitation Measurement Integrated Multi-satellitE Retrievals, 0.1-degree grid, half-hourly to monthly. Provides rainfall intensity climatology and antecedent moisture proxies. IMERG is a statistical product; it underestimates orographic convective peaks, so treat intensity percentiles as indicative rather than precise triggering thresholds.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands, 20 m shortwave infrared, 5-day revisit at the equator. Clay and iron-oxide minerals diagnostic of weathered, slide-prone lithologies show distinct SWIR signatures. Cloud cover is the main constraint; tropical corridors may require 12-to-24-month composites to assemble a clean mosaic.
Why fibre planners need a slope-failure model, not just a slope map
Slope angle alone is a poor predictor of where a buried cable will be disrupted. A 15-degree clay-rich hillside with 2,000 mm of annual rainfall and a history of 3 mm per year of creep is far more dangerous to a direct-buried cable than a 35-degree granite face that has been stable for decades. The distinction matters because the failure modes differ: shallow translational slides shear a cable in minutes; deep-seated rotational failures deform a conduit over months before full rupture. Both require different mitigation strategies, and neither shows up on a simple gradient raster.
A susceptibility model integrates the physical controls: topographic form, material properties inferred from spectral indices, antecedent and peak rainfall, and any measurable pre-failure displacement. The result is a probabilistic score per slope unit, not a guarantee. It tells a route engineer which 10 percent of a 200 km corridor carries 60 percent of the ground-movement risk. That is enough to redirect alignment, specify armoured conduit, or flag segments for intrusive ground investigation.
What InSAR displacement history actually reveals
Sentinel-1 SBAS or persistent-scatterer processing over a multi-year archive can detect line-of-sight surface velocities down to roughly 1-2 mm per year over coherent ground. Slow-moving landslides, particularly deep-seated rotational failures, often exhibit this kind of precursory creep for years before catastrophic movement. A slope unit showing 4-8 mm per year of consistent downslope displacement in the InSAR record is already telling you something that no amount of desk-based geology review will reveal.
The honest limits are significant. C-band coherence degrades rapidly under dense vegetation, which is precisely where many tropical fibre routes run. Steep terrain introduces layover and shadow that blank out the very slopes of greatest concern. Where coherence is low, displacement absence is not the same as displacement absence in reality. L-band SAR (ALOS-2, or the forthcoming NISAR) maintains coherence better in vegetated terrain, and where the route justifies it, supplementing Sentinel-1 with archived L-band data is worth the additional cost.
Reading lithology from orbit: what SWIR bands can and cannot do
Sentinel-2 bands 11 and 12 (1.6 µm and 2.2 µm shortwave infrared) are sensitive to hydroxyl-bearing minerals: smectite, kaolinite and illite clays that dominate deeply weathered tropical regolith and are strongly associated with shallow slide susceptibility. The clay index (B11/B12 ratio) and iron-oxide index (B4/B3) can distinguish weathered from fresh bedrock exposures and flag zones where material properties are likely to be poor.
This is a proxy, not a geotechnical survey. Vegetation cover masks the signal on all but bare or sparsely vegetated slopes. Atmospheric correction errors introduce noise. The spectral signature of a smectite-rich soil and a kaolinite-rich soil can overlap in a single pixel. The output should be treated as a qualitative modifier to the susceptibility score, not as a substitute for a plasticity index from a laboratory sample. Used correctly, it narrows the list of segments warranting ground investigation; it does not replace that investigation.
Rainfall as a trigger: IMERG climatology and its ceiling
Most shallow landslides are triggered by rainfall exceeding a threshold intensity over a given duration. GPM IMERG provides a 20-plus-year climatological record at 0.1-degree resolution, from which return-period rainfall intensities can be estimated for any corridor segment. Overlaying the 10-year or 25-year 24-hour rainfall intensity against slope-unit susceptibility scores identifies where the combination of unstable ground and extreme rainfall is most likely to coincide.
IMERG's resolution is approximately 11 km at the equator. It systematically underestimates convective peaks in complex terrain, where orographic enhancement can push actual intensities 30-50 percent above the gridded value. For corridors crossing high mountain ranges, supplementing IMERG with any available rain-gauge records or higher-resolution regional reanalysis products improves the trigger-rainfall estimates materially. The susceptibility map should carry an explicit note on this uncertainty.
Building the slope-unit model: method and honest output
Slope units, delineated by dividing a watershed into hillslope facets bounded by ridges and channels, are the natural mapping unit for landslide susceptibility. They behave more physically than arbitrary grid cells because a slope unit approximates the area that would move as a single body. From the TanDEM-X WorldDEM, slope units are extracted using standard GIS hydrological tools; at 12 m posting, units as small as 0.5-1 hectare can be resolved, which is adequate for corridor-scale routing decisions.
Each slope unit receives a score aggregated from the input layers: slope angle and curvature from the DEM, displacement velocity from InSAR, clay and iron-oxide indices from Sentinel-2, and rainfall intensity percentile from IMERG. Statistical methods such as logistic regression or frequency-ratio analysis, calibrated against a regional landslide inventory where one exists, convert these inputs into a relative susceptibility class. Where no inventory exists, the model is uncalibrated and the output should be labelled as a relative hazard index rather than a validated probability. That distinction matters for how a client presents the map to regulators or insurers.
Satellize runs this workflow on open-constellation data for corridor assessments, with commercial DEM licensing handled on the client's behalf. The Tonga crop-estimation programme demonstrated that multi-sensor fusion on island terrain is tractable even where ground data are sparse; the same data-fusion discipline applies here.
From susceptibility score to route decision
A susceptibility map has three practical uses in a fibre routing project. First, it informs alignment: high-susceptibility slope units can often be avoided by moving the route 50-200 m laterally onto a more stable facet, at modest additional civil cost. Second, it specifies where enhanced installation standards apply: armoured conduit, increased burial depth, or concrete encasement in conduit trenches crossing class-4 or class-5 slope units. Third, it prioritises the ground investigation programme. Intrusive investigation is expensive; a susceptibility map that concentrates trial pits and borehole locations on the highest-risk 15 percent of a route can halve the investigation budget while improving confidence where it matters most.
The map does not eliminate the need for ground investigation before route finalisation. A slope unit scoring high on all remote-sensing indicators may turn out, on inspection, to have a competent rock mass beneath a thin colluvial veneer. The opposite is also true. The probabilistic output is a decision-support tool for a geotechnical engineer, not a replacement for one.
Typical figures
| DEM spatial resolution | 12 m (TanDEM-X WorldDEM); 0.4 m relative vertical accuracy on slopes |
| InSAR displacement sensitivity | ~1-2 mm/year line-of-sight over coherent targets; Sentinel-1 IW mode, 5 x 20 m ground resolution |
| InSAR revisit | 6-day (dual-geometry combined); 12-day single geometry; archive from 2014 |
| Multispectral resolution | 10 m (VIS/NIR), 20 m (SWIR); Sentinel-2 archive from 2015 |
| Rainfall grid resolution | ~11 km (0.1 degree); GPM IMERG half-hourly to monthly; record from 2000 |
| Slope-unit minimum size | ~0.5-1 ha at 12 m DEM posting; smaller units not reliably resolved |
| Susceptibility output classes | Typically 4-5 relative classes (very low to very high); probabilistic if calibrated against an inventory |
| Delivery formats | GeoTIFF raster, GeoPackage vector (slope units with attribute scores), PDF report, optional WMS/WMTS tile service |
| Typical corridor archive latency | Displacement time series processed from existing archive; 3-6 weeks from commission to delivery for a 200 km corridor |
Analytics Satellize can run
| Slope-unit susceptibility map | Frequency-ratio or logistic-regression fusion of DEM derivatives, InSAR velocity, spectral indices and rainfall percentiles per slope unit | GeoTIFF and GeoPackage with per-unit susceptibility class and contributing-factor scores; PDF technical report |
| InSAR displacement time series | SBAS or persistent-scatterer processing of Sentinel-1 SLC archive; line-of-sight velocity and displacement history per coherent target | Point-cloud GeoPackage with velocity, displacement time series and coherence values; flagged anomalies along corridor buffer |
| Lithology proxy raster | Sentinel-2 SWIR clay index (B11/B12) and iron-oxide index (B4/B3) computed from cloud-free annual composite | GeoTIFF rasters per index; classified overlay indicating zones of likely clay-rich or weathered material |
| Rainfall intensity climatology | GPM IMERG monthly and event-scale statistics; return-period intensity estimation per slope unit using L-moment fitting | CSV and GeoTIFF of 5-, 10- and 25-year 24-hour rainfall intensity per 0.1-degree cell intersecting corridor |
| High-risk segment register | Threshold overlay: slope units scoring class 4-5 susceptibility within a user-defined corridor buffer | Ranked table of segments by susceptibility class, length, and dominant risk driver; formatted for geotechnical investigation planning |
| Change detection for active slide identification | Sentinel-2 NDVI and bare-soil index differencing between two epochs to identify new or enlarged slide scars | GeoTIFF change layer and alert report flagging new disturbances intersecting the corridor buffer since a specified baseline date |
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.