Fault scarp morphology and palaeoseismic rupture mapping from high-resolution DEMs
High-resolution DEMs from spaceborne lidar and bistatic SAR expose fault scarps, sag ponds and pressure ridges that record prehistoric ruptures, letting geologists estimate slip-per-event and recurrence intervals without a trench.
Sensors
- GEDI (Global Ecosystem Dynamics Investigation, ISS): Full-waveform lidar at 25 m footprint spacing along sparse parallel tracks (~600 m cross-track separation at the equator). Vertical accuracy ~2.5 cm RMSE over flat ground; degrades under dense canopy. Useful for extracting scarp height in wooded terrain where photogrammetric DEMs are blinded by the canopy surface. Not a wall-to-wall DEM; gap-filling with another source is always required.
- TanDEM-X (X-band bistatic SAR, DLR/Airbus): 12 m global DEM (TanDEM-X 12 m) with absolute vertical accuracy ~2 m and relative accuracy ~0.2 m over smooth terrain. Performs well in arid, sparsely vegetated settings where X-band penetrates little biomass. Layover and foreshortening on steep scarps facing the sensor can suppress or distort the very feature being measured; acquisition geometry must be checked.
- ALOS World 3D (AW3D30): 30 m global DEM from JAXA PRISM stereo imagery. Vertical accuracy ~5 m globally, better in open terrain. Adequate for identifying major fault traces and large-displacement scarps (>5 m throw), but insufficient for diffusion-age modelling of subtle scarps below ~1 m height.
- Maxar WorldDEM (commercial stereo photogrammetry): Sub-2 m posting DEM from WorldView stereo pairs. Vertical accuracy ~1 m CE90 in open terrain. Resolves metre-scale scarps and pressure ridges. Cloud cover at acquisition and canopy remain limiting factors; unlike lidar, it captures the vegetation surface, not the ground.
- Airborne lidar derivatives (third-party point clouds): Ground-classified point clouds at 1–8 points per square metre yield 0.25–1 m bare-earth DEMs, the current practical gold standard for scarp morphology. Satellize ingests client-supplied or open-archive lidar (e.g. USGS 3DEP) rather than flying its own sensors.
What a scarp's shape encodes
A fault scarp is not just a step in the landscape. Its cross-sectional profile is a clock. Fresh ruptures produce near-vertical free faces that degrade through time as colluvial wedges accumulate and hillslope diffusion smooths the crest. The relationship between scarp height, maximum slope angle and elapsed time since rupture was formalised by Hanks, Bucknam and colleagues in the 1980s and has been refined repeatedly since. The key quantity is the diffusivity constant, typically in the range of 0.5 to 5 m² per thousand years for semi-arid soils, which converts observed curvature into an age estimate.
The method only works if the DEM resolves the inflection points at the scarp crest and toe. A rule of thumb from published sensitivity analyses: the DEM posting should be no coarser than one-fifth of the scarp height. A 1 m scarp therefore demands a 20 cm or better DEM, which means airborne lidar or very-high-resolution stereo. TanDEM-X at 12 m is adequate for scarps with throws of several metres or more. GEDI fills a specific niche: it can pierce sparse woodland that defeats photogrammetry, recovering ground elevation at the scarp face where a stereo DEM would return the canopy top.
Reading the landforms: scarps, sag ponds and pressure ridges
Palaeoseismologists look for a family of landforms, not a single feature. On strike-slip faults, the diagnostic suite includes linear scarps, shutter ridges, offset stream channels, sag ponds in releasing bends and pressure ridges in restraining bends. On normal faults, the key indicators are the scarp itself, the colluvial wedge at its base and the back-tilted facets above. Each landform has a characteristic topographic signature that a good DEM makes unambiguous.
Sag ponds are particularly useful because they are closed depressions that can be dated independently through sediment cores. Pressure ridges appear as elongate swells a few metres high and tens to hundreds of metres long, easily confused with fluvial levées or aeolian features at coarse resolution. At 1 m DEM resolution they become morphologically distinct. At 12 m they are detectable only if their relief exceeds roughly 2 to 3 m. This is an honest constraint: TanDEM-X will not resolve every feature that a palaeoseismologist needs, and the spec table below reflects that.
GEDI over woodland: promise and gap problem
The Wasatch Front, the Apennines, parts of the Anatolian fault system and many other seismically active regions carry enough tree cover to render photogrammetric bare-earth models unreliable. GEDI's 1064 nm full-waveform lidar separates ground returns from canopy returns, recovering scarp height even under 50 to 70 per cent canopy cover, though accuracy degrades as cover increases. Published comparisons against airborne lidar ground truth show GEDI ground elevation errors of roughly 0.3 to 1 m RMS under moderate woodland, rising sharply under closed-canopy forest.
The fundamental problem is spatial sampling. GEDI's tracks are separated by roughly 600 m at the equator, and within a track the 25 m footprints are spaced 60 m apart along-track. A fault scarp running obliquely to the track may be sampled by only a handful of shots. The workflow is therefore to use GEDI shots to constrain the vertical datum and scarp height at sampled points, then interpolate using a coarser continuous DEM such as TanDEM-X or AW3D30 as the morphological backbone. Neither source alone is sufficient.
TanDEM-X in arid settings: where it earns its place
Desert and semi-arid alluvial fans are the best terrain for TanDEM-X palaeoseismic work. Sparse vegetation, stable surfaces and strong radar contrast between the scarp face and the fan surface combine to give relative vertical accuracies approaching 0.2 m over smooth ground. The 12 m posting resolves scarps with throws above roughly 2 to 3 m reliably. Several published studies of the Dead Sea fault system and the Basin and Range province have used TanDEM-X to map fault traces continuously over hundreds of kilometres, far beyond what airborne lidar campaigns can cover economically.
Acquisition geometry is the main operational hazard. X-band SAR images from a fixed look direction; a scarp face oriented toward the sensor will be compressed in range (foreshortening) and may appear artificially steep or, in extreme cases, disappear into layover. Ordering TanDEM-X data from both ascending and descending geometries and comparing the results is standard practice for this reason. Where the two geometries disagree on scarp height, the lower value is usually the more reliable one, and field checking is warranted.
From DEM to recurrence interval: the analytical chain
Slip-per-event estimates come from measuring scarp height and subtracting the component attributable to long-term fault dip, leaving the net vertical displacement for the most recent rupture. Where multiple scarps are stacked, each terrace riser records a separate event. Recurrence intervals require at least two independently dated events, typically from radiocarbon or optically stimulated luminescence dating of offset features or trench stratigraphy. The DEM analysis constrains the geometry; the geochronology constrains the timing.
Diffusion-age modelling adds a third line of evidence. By fitting the observed scarp profile to the diffusion equation, and using a locally calibrated diffusivity constant, analysts can estimate the time since rupture without a trench. The uncertainty is large, often a factor of two or three, because diffusivity varies with lithology, climate and aspect. The method is most useful for ranking scarps by relative age and for flagging features that appear geologically young and therefore warrant closer investigation.
Satellize can run the full DEM-processing and morphometric-extraction chain on client-supplied or publicly archived elevation data, delivering georeferenced scarp maps, height profiles and preliminary slip estimates as GIS layers.
Where the method reaches its limits
Dense forest is the most common show-stopper. Airborne lidar at adequate point density remains the only reliable solution under closed canopy; no spaceborne sensor currently available provides both wall-to-wall coverage and the ground-point density needed for sub-metre bare-earth DEMs in forested terrain. GEDI is a partial answer, not a complete one.
Scarp diffusion ages become unreliable below about 1 m of relief, because DEM noise at that scale is comparable to the morphological signal. They also become unreliable above roughly 100,000 years, because the scarp profile approaches a steady state that is insensitive to further diffusion. The method works best for Holocene and latest Pleistocene ruptures, which is also where the seismic hazard interest is greatest. For older features, cosmogenic nuclide dating of offset surfaces is a better geochronological tool, though it still benefits from the same high-resolution DEM input for geometric control.
Typical figures
| Best available spatial resolution (airborne lidar) | 0.25–1 m bare-earth DEM posting; depends on point density and classification quality |
| TanDEM-X DEM posting | 12 m; relative vertical accuracy ~0.2 m over smooth arid terrain, ~2 m absolute |
| GEDI footprint and track spacing | 25 m footprint, ~60 m along-track spacing, ~600 m cross-track separation at equator |
| GEDI vertical accuracy (open ground) | ~2.5 cm RMSE; degrades to 0.3–1 m RMS under moderate woodland canopy |
| AW3D30 posting and accuracy | 30 m; ~5 m vertical RMSE globally, better in open terrain |
| Minimum detectable scarp height (TanDEM-X) | ~2–3 m throw in arid settings; smaller scarps require sub-metre DEM |
| Minimum detectable scarp height (1 m airborne lidar DEM) | ~0.2–0.5 m; below this, DEM noise dominates the morphological signal |
| Diffusion-age model reliable range | Approximately 500 to 100,000 years BP; poorly constrained outside this window |
| Archive depth (TanDEM-X global DEM) | Global mosaic acquired 2010–2015; scene-level data available on request from DLR |
| Delivery formats | GeoTIFF DEM, shapefiles or GeoPackage for scarp traces, CSV height profiles, PDF morphometric report |
Analytics Satellize can run
| Fault trace and scarp inventory map | Semi-automated ridge and break-of-slope detection on hillshade and slope-derivative rasters; manual QA against multiple illumination azimuths | GeoPackage of fault trace polylines with attributed scarp height, facing direction and confidence class |
| Scarp height profiles and slip-per-event estimates | Swath-averaged topographic profiles perpendicular to fault strike; geometric decomposition into dip-slip and long-term tilt components following published normal-fault scarp conventions | CSV of profile data plus PDF report with interpreted slip magnitudes and uncertainty ranges |
| Scarp diffusion age estimates | Nonlinear least-squares fitting of observed scarp profiles to the 1D hillslope diffusion equation; sensitivity analysis over published diffusivity ranges for the target lithology and climate | Table of estimated ages with 1-sigma uncertainty bounds; flagged where profile shape is ambiguous |
| Sag pond and closed-depression inventory | Hydrological sink extraction from pit-filled and raw DEMs; morphometric filtering to separate tectonic sags from fluvial and aeolian depressions by elongation ratio and orientation relative to fault trace | GIS polygon layer of candidate sag ponds with morphometric attributes |
| GEDI-constrained bare-earth surface in wooded terrain | Ground-return extraction from GEDI L2A data; kriging or spline interpolation anchored to GEDI shots, with TanDEM-X as the morphological trend surface | GeoTIFF hybrid DEM with per-pixel confidence layer indicating GEDI shot density |
| Multi-geometry TanDEM-X scarp consistency check | Co-registration and differencing of ascending and descending TanDEM-X DEMs; flagging of cells where height discrepancy exceeds 1 m as probable layover or foreshortening artefacts | Artefact mask GeoTIFF and annotated scarp inventory distinguishing reliable from suspect measurements |
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.