High-resolution DEM analysis for landscape and earthwork archaeology
Spaceborne and airborne lidar DEMs reveal earthworks, hollow ways and field systems as micro-topographic anomalies of tens of centimetres. Processing methods such as local relief modelling and sky-view factor visualisation make these features legible where optical imagery shows nothing.
Sensors
- TanDEM-X WorldDEM: Bistatic X-band SAR interferometry. WorldDEM delivers 12 m posting with absolute vertical accuracy better than 10 m and relative vertical accuracy of approximately 2 m (1 sigma) over open terrain; the commercial WorldDEM4Ortho product reaches 0.4 m posting over selected areas. Ideal for regional landscape surveys in open, low-vegetation terrain.
- Copernicus DEM GLO-30: Derived from TanDEM-X, globally available at 30 m posting, freely distributed. Relative vertical accuracy is broadly comparable to WorldDEM at this resolution. Useful as a baseline for large-area steppe and dryland surveys where sub-metre posting is not required.
- GEDI (Global Ecosystem Dynamics Investigation, ISS): Full-waveform lidar operating at 1064 nm from the International Space Station. Footprint diameter approximately 25 m, shot spacing approximately 60 m along-track, non-contiguous sampling. Ground elevation accuracy under closed canopy is around 3 cm RMSE in published validation studies; not a raster DEM, but provides canopy-penetrating ground returns where optical stereo fails.
- SRTM (Shuttle Radar Topography Mission): C-band interferometric SAR, 30 m global posting. Collected in 2000; vertical accuracy approximately 16 m absolute, 6 m relative (90th percentile). Useful as a long-baseline comparison layer and for regional context, though too coarse for sub-metre earthwork detection.
- Airborne lidar (ALS, client-commissioned): Point densities of 4 to 50 points per square metre are achievable commercially, yielding bare-earth DEMs at 0.25 to 0.5 m posting. This is the resolution floor at which features such as 20 to 30 cm high linear banks become reliably detectable. Spaceborne systems cannot currently match this; the gap is real and should be stated plainly.
What a floating roof gives away
A ploughed-down Iron Age field boundary may stand no more than 20 to 30 cm above its surroundings. That is too little for a shadow in an optical image and too little for the human eye walking the field. It is, however, a measurable height difference, and elevation data do not care whether a feature is visible.
The key insight behind landscape DEM archaeology is that ancient earthworks survive as coherent micro-topographic signals long after their surface expression has been reduced by agriculture, erosion or vegetation. Hollow ways, the sunken tracks worn by millennia of foot and cart traffic, can persist as negative anomalies of 0.3 to 1 m. Burial mounds in the Syrian and Jordanian steppe, documented in programmes such as the APAAME aerial survey, appear in TanDEM-X derivatives as positive anomalies of similar magnitude. The signal exists; the question is whether the chosen DEM has sufficient vertical precision to separate it from noise.
Resolution floors and the honest arithmetic
The free Copernicus DEM GLO-30 at 30 m posting will not resolve a 10 m wide linear bank. WorldDEM at 12 m posting begins to show larger mound complexes and broad hollow-way systems but misses narrow features. Airborne lidar at 0.25 to 0.5 m posting is, at present, the only spaceborne-or-airborne technology that reliably detects the full range of earthwork scales that archaeologists care about. This is not a criticism of spaceborne systems; it is physics.
GEDI is a partial exception for forested terrain. Its waveform decomposition can return ground elevation beneath closed canopy where X-band SAR and optical stereo both fail, and published work from Mesoamerican sites (including comparisons with airborne lidar over Maya lowland terrain) has shown that GEDI ground returns correlate well with ALS bare-earth elevations. The limitation is sampling geometry: GEDI's non-contiguous 25 m footprints do not form a continuous raster, so interpolation artefacts are significant. It is a reconnaissance tool, not a mapping tool.
SRTM remains useful precisely because it is old. A 2000-vintage baseline compared against a current WorldDEM or commercial stereo DSM can reveal volumetric change at tells and mound complexes, a method covered separately in the sibling page on tell volume loss quantification.
Processing steps that separate signal from terrain noise
Raw DEMs are dominated by regional slope. A hillside trending at 2 degrees swamps a 0.3 m earthwork unless that regional trend is removed. Local relief modelling (LRM) subtracts a smoothed version of the surface, computed over a neighbourhood radius chosen to match the scale of features of interest, typically 50 to 200 m for earthwork archaeology. What remains is the residual micro-topography.
Sky-view factor (SVF) visualisation, developed by Žiga Kokalj and colleagues at the Research Centre of the Slovenian Academy of Sciences and Arts and widely adopted since, computes for each cell the proportion of the sky hemisphere visible from that point. Concave features (ditches, hollow ways) have low SVF; convex features (banks, mounds) have high SVF. The resulting greyscale image is perceptually intuitive and has become a standard deliverable in European landscape archaeology. It is particularly effective on airborne lidar DEMs but produces useful results on WorldDEM at 12 m posting for larger features.
Openness (positive and negative), slope, and multi-directional hillshade composites each emphasise different feature geometries. No single visualisation is universal. A linear bank running parallel to the illumination azimuth of a hillshade will disappear; rotating the azimuth or using SVF recovers it. Systematic processing should apply at least three visualisation methods and compare results.
What the published record shows in the Middle East and Mesoamerica
In the basalt desert (harrat) of southern Syria and northern Jordan, researchers using WorldDEM and SRTM derivatives have mapped networks of kite-shaped hunting structures, wheel-shaped geoglyphs (the so-called desert kites and Nazca-analogue forms), and hollow-way systems radiating from Bronze Age tells. The basalt surface is essentially vegetation-free, which makes it one of the most favourable environments on earth for spaceborne DEM archaeology: no canopy, minimal soil moisture variation, and high surface roughness contrast between disturbed and undisturbed ground.
In the Maya lowlands, the situation is the opposite. Closed tropical canopy defeated optical and radar approaches for decades. The Caracol lidar survey (2010, airborne) and subsequent work at sites including Tikal demonstrated that ALS at sufficient point density could map an entire urban landscape beneath forest cover, revealing causeways, reservoirs, agricultural terraces and residential platforms invisible in any other data type. GEDI has since been used to extend this kind of ground-return analysis across broader Mesoamerican regions, though the sampling gaps mean that individual site mapping still requires dedicated airborne acquisition.
The practical lesson from both regions is the same: spaceborne DEMs are powerful for regional prospection and prioritisation, but site-level mapping for heritage management or excavation planning almost always requires airborne lidar or, at minimum, commercial sub-metre stereo imagery with structure-from-motion processing.
Where Satellize fits into a DEM archaeology workflow
Satellize runs DEM-based landscape analysis on open datasets including WorldDEM, Copernicus DEM GLO-30 and GEDI, applying local relief modelling, SVF and openness visualisations as standard processing steps. For clients commissioning airborne lidar or commercial stereo acquisitions, the analytics pipeline ingests those datasets directly and returns georeferenced GIS layers of anomaly candidates ranked by confidence and morphological class.
The approach is the same one that underpins the Tonga crop-estimation programme: take the best available elevation or spectral data, apply documented quantitative methods, and return outputs that a non-specialist can act on. For a national heritage authority with a large survey area and a limited airborne budget, a WorldDEM-based regional scan is a rational first step to identify priority zones before committing to expensive ALS acquisition. That is the honest use case for spaceborne DEMs in this discipline.
The cloud problem does not apply, but other limits do
One advantage DEM analysis has over optical and multispectral methods is that it is insensitive to cloud cover after acquisition. A WorldDEM tile collected under any atmospheric condition is equally useful once the interferometric processing is complete. This matters in regions with persistent cloud, though the Middle East steppe and Mesoamerican highland sites have quite different cloud climatologies.
The limits that do apply are vertical precision, spatial resolution and canopy penetration, as described above. There is also an interpretive limit: a circular positive anomaly in a DEM might be a burial mound, a natural knoll, a spoil heap from modern construction, or an artefact of the interferometric processing. Ground-truthing or cross-referencing with historical imagery remains essential. DEM analysis narrows the search area; it does not replace fieldwork.
Typical figures
| WorldDEM spatial posting | 12 m standard; 0.4 m (WorldDEM4Ortho, selected areas) |
| Copernicus DEM GLO-30 posting | 30 m, global, freely available |
| WorldDEM relative vertical accuracy | ~2 m (1 sigma) open terrain; better over flat bare surfaces |
| GEDI footprint and ground accuracy | ~25 m footprint; ~3 cm RMSE ground elevation under canopy (published validation) |
| Minimum detectable earthwork height (spaceborne) | ~0.5 to 1 m with WorldDEM 12 m; sub-0.3 m requires airborne lidar at ≥4 pts/m² |
| Airborne lidar point density (commercial) | 4 to 50 pts/m²; bare-earth DEM at 0.25 to 0.5 m posting |
| SRTM archive date | February 2000; 30 m global posting |
| Delivery formats | GeoTIFF (LRM, SVF, openness rasters), GeoPackage (anomaly polygons), PDF report |
| Processing latency (spaceborne DEM) | Typically 2 to 4 weeks from data order to interpreted GIS output |
| Archive depth | TanDEM-X global acquisition 2010 to present; SRTM 2000; GEDI 2019 to present |
Analytics Satellize can run
| Regional earthwork anomaly map | Local relief modelling (LRM) on WorldDEM or Copernicus DEM GLO-30; neighbourhood radius tuned to feature scale | GeoTIFF residual relief layer and GeoPackage polygon layer of candidate anomalies, ranked by relief amplitude |
| Sky-view factor visualisation | SVF algorithm (Kokalj et al. method) applied to bare-earth DEM; highlights concave and convex micro-topography simultaneously | Greyscale GeoTIFF suitable for direct overlay in GIS or CAD, with accompanying PDF interpretation guide |
| Multi-direction hillshade composite | Eight-azimuth hillshade stack merged to single image; reduces directional shadow bias that suppresses linear features | GeoTIFF composite raster; optionally animated as a rotating-sun video for presentation use |
| GEDI ground-return interpolation for forested sites | GEDI L2A ground elevation shots filtered by quality flag, kriging or natural-neighbour interpolation to produce a sparse bare-earth surface | Interpolated GeoTIFF with uncertainty band raster; flagged as reconnaissance-grade, not mapping-grade |
| Morphological classification of anomaly candidates | Shape metrics (circularity, elongation, orientation) computed on LRM anomaly polygons; compared against published earthwork typologies for the region | Attribute table in GeoPackage with morphological class, confidence score and recommended ground-truth priority |
| Priority zone report for airborne lidar scoping | Spatial clustering of high-confidence anomalies to define minimum bounding polygons; area and flight-line estimates computed for ALS tender preparation | PDF report with priority polygon shapefile and indicative ALS acquisition parameters for client's procurement team |
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.