Landslide Detection and Monitoring with Satellite Data
Quick Answer: Satellites detect landslides through two complementary approaches: (1) Post-event optical/SAR change detection identifies landslide scars — bare soil exposed where vegetation was removed, with characteristic elongated or lobate shapes. (2) Pre-event InSAR monitoring detects slow ground displacement (mm/year) on unstable slopes, potentially providing early warning before catastrophic failure. Sentinel-1 InSAR achieves mm/year displacement precision on bare rock and urban slopes. Optical detection works best for large landslides (>0.1 ha) that remove vegetation. Rainfall-triggered shallow landslides may be too numerous and small for systematic satellite detection.
In 2017, satellite radar data revealed that a slope above the village of Xinmo in Sichuan, China, had been creeping at rates of several centimeters per year for at least two years before it catastrophically collapsed, burying over 60 homes. The displacement was visible in retrospective InSAR analysis — but nobody had been monitoring that particular slope.
This illustrates both the power and the challenge of satellite landslide monitoring: the technology to detect pre-failure displacement exists, but applying it systematically to the millions of potentially unstable slopes worldwide is an enormous undertaking.
Post-Event Detection: Finding Landslides That Already Happened
Optical Detection
Fresh landslides create distinctive features in optical imagery:
Spectral change: Vegetation is removed, exposing bare soil or rock. NDVI drops from 0.6-0.8 (forest) to 0.1-0.3 (bare soil). This spectral contrast is strong and easily detected through image differencing.
Morphological features: Landslide scars have characteristic shapes — elongated along the slope direction, with a concave source area (scarp) at the top and a lobate deposit zone at the bottom.
Size range: Satellite detection is most reliable for landslides larger than about 0.01-0.1 hectares, depending on sensor resolution:
- Sentinel-2 (10m): Detects landslides > ~0.1 ha
- Commercial satellites (1-3m): Detects landslides > ~0.01 ha
- Very high resolution (<1m): Individual debris flows detectable
SAR-Based Detection
SAR detects landslides through:
- Backscatter change: Removal of vegetation changes the scattering mechanism, altering both VV and VH backscatter
- Coherence loss: Surface disruption destroys coherence
- Advantage: Works through clouds — critical because landslides are typically triggered by heavy rainfall events, which produce cloud cover that prevents optical observation
Inventory Mapping After Major Events
After rainfall events, earthquakes, or typhoons that trigger hundreds or thousands of landslides simultaneously, satellite-based inventory mapping provides:
- Total number and area of landslides triggered
- Spatial distribution relative to triggering factors (rainfall, ground shaking)
- Road and infrastructure blockages
- Areas at risk of secondary hazards (landslide-dammed lakes)
The 2015 Nepal earthquake triggered over 25,000 landslides mapped from satellite imagery — a catalogue that would have taken years to compile through field survey alone.
Pre-Event Monitoring: Catching Slopes Before They Fail
InSAR Displacement Monitoring
Many large landslides don't fail suddenly. They creep — moving millimeters to centimeters per year — before accelerating to catastrophic failure. InSAR detects this pre-failure creep:
PS-InSAR: Identifies stable radar targets (buildings, rock outcrops) on or near slopes and tracks their displacement over months to years. Typical precision: 1-2 mm/year.
SBAS: Provides broader spatial coverage than PS-InSAR, including natural terrain, but with lower precision per pixel.
What to look for: Persistent downslope displacement that accelerates over time is the classic precursor to catastrophic failure. The Fukuzono/Voight failure forecast model predicts failure timing based on the acceleration of displacement.
Practical Implementation
Systematic slope monitoring requires:
- Identify target slopes: From landslide susceptibility maps, geological surveys, or known problem areas
- Process InSAR time series: Using Sentinel-1 archive (typically 2-5 years of data)
- Identify anomalous displacement: Slopes moving faster than their geological context suggests
- Establish monitoring: Regular processing to track acceleration
Several countries have implemented or are implementing national-scale InSAR monitoring:
- Italy: The COSMO-SkyMed-based monitoring covers known landslide-prone areas
- Norway: Sentinel-1-based monitoring of unstable rock slopes above fjords
- Japan: Monitoring of slopes destabilized by earthquakes and heavy rainfall
Limitations of InSAR for Landslide Monitoring
Vegetation: Dense vegetation causes rapid decorrelation at C-band, reducing InSAR measurement quality. Many landslide-prone slopes in tropical and temperate regions are heavily vegetated, limiting the effectiveness of Sentinel-1 InSAR. L-band systems (ALOS-2, future NISAR) penetrate vegetation better.
Steep terrain: Slopes facing away from the satellite may fall in radar shadow; slopes facing toward the satellite may experience severe foreshortening or layover. Both reduce or prevent InSAR measurement.
Movement direction: InSAR measures displacement in the satellite's line-of-sight direction. A slope moving perpendicular to the look direction may show minimal InSAR signal despite significant actual displacement.
Rapid movement: InSAR loses coherence when displacement exceeds half a wavelength between acquisitions (~2.8 cm for Sentinel-1's 6-12 day revisit). Fast-moving landslides that displace more than this between passes are unmeasurable by standard InSAR.
Susceptibility and Hazard Assessment
Satellites contribute to landslide hazard assessment even without detecting active landslides:
DEM-derived terrain variables: Slope angle, curvature, aspect, and flow accumulation from satellite-derived DEMs (SRTM, ALOS World 3D) are primary inputs to susceptibility models.
Land cover: Satellite-derived land cover maps identify destabilizing factors — deforestation, road cuts, urbanization on slopes.
Soil moisture: SAR-derived soil moisture indicates when slopes are approaching saturation — the condition that triggers many rainfall-induced landslides.
Historical inventory: Satellite-based landslide inventories provide the training data for statistical susceptibility models (logistic regression, Random Forest, neural networks).
Landslide Dam Monitoring
When landslides block river valleys, the resulting landslide-dammed lakes pose extreme flood hazards. Satellite monitoring provides:
Lake detection: SAR or optical imagery detects the growing lake behind the dam Lake volume estimation: From lake area (satellite) and valley geometry (DEM) Dam stability assessment: InSAR monitors deformation of the dam itself Outburst flood warning: Rapid lake level rise detected by satellite altimetry or area expansion
The 2018 Baige landslide in Tibet dammed the Jinsha River, creating a lake that threatened millions downstream. Satellite monitoring tracked the lake's growth hourly, providing critical data for evacuation decisions.
From Research to Operations
Satellite landslide detection has matured significantly but faces a fundamental scaling challenge: there are millions of potentially unstable slopes worldwide, and systematic monitoring requires processing vast quantities of SAR data continuously.
Current operational status:
- Post-event detection: Operational at national and international agencies (Copernicus EMS, NASA, JAXA)
- Regional InSAR monitoring: Operational in several countries for known high-risk areas
- Global systematic monitoring: Still largely a research aspiration; computational requirements are being addressed through cloud computing and automated processing
The technology gap is closing. As SAR data volumes increase (Sentinel-1, NISAR, commercial SAR constellations) and processing becomes more automated, the vision of comprehensive global landslide monitoring moves closer to reality. For now, the most effective approach combines satellite-based susceptibility mapping to identify where monitoring is needed, InSAR for continuous displacement tracking of the most critical slopes, and rapid post-event detection to document triggered landslides and assess secondary hazards.
