SAR subsurface penetration for desert archaeological site detection
Long-wavelength SAR penetrates dry desert sediment to reveal buried structures, palaeochannels and ancient roads invisible at the optical surface. L-band and P-band backscatter from subsurface dielectric contrasts that optical sensors simply cannot reach.
Sensors
- ALOS-2 PALSAR-2: JAXA L-band (1.27 GHz) SAR with 3–10 m resolution in spotlight and stripmap modes; revisit approximately 14 days. L-band penetrates dry aeolian sand to depths of 1–2 m, occasionally deeper in very low-moisture conditions, backscattering from buried gravel lags, road surfaces and structural rubble.
- ESA BIOMASS: P-band (435 MHz) SAR launched 2024; designed for forest biomass but provides the longest wavelength spaceborne SAR yet flown. P-band penetration in dry, low-conductivity desert sediment can exceed 5 m in principle, though archaeological validation datasets are still accumulating. Resolution approximately 50 m in standard mode.
- Sentinel-1 C-band SAR: C-band (5.4 GHz) penetrates only a few centimetres into dry sand, making it largely surface-sensitive in desert conditions. Useful for surface roughness characterisation, sand dune mobility mapping and change detection that contextualises subsurface results, not for direct subsurface imaging. 5–20 m resolution, 6-day revisit on dual-satellite constellation.
- SRTM (Shuttle Radar Topography Mission): C-band and X-band interferometric SAR flown in 2000; 30 m global DEM. In hyper-arid regions the C-band signal partially penetrated thin sand sheets, producing elevation artefacts that researchers have repurposed as indirect evidence of near-surface buried topography. Archive only, no revisit.
Why dry sand is transparent to radio waves
The physics is straightforward. Radar penetration depth in any medium is governed by the dielectric constant and electrical conductivity of that medium. Dry quartz sand has a very low dielectric constant (roughly 2–4) and near-zero conductivity, which means L-band and P-band wavelengths pass through it with relatively little attenuation before striking a buried interface with higher dielectric contrast, such as compacted gravel, fired brick, mudbrick rubble or moist palaeochannel fill.
The critical threshold is moisture. Even a few percent volumetric water content raises the dielectric constant sharply, collapsing penetration depth from metres to centimetres. This is why the technique is essentially confined to hyper-arid environments: the central Sahara, the Rub' al Khali, the Dasht-e Kavir. Coastal desert margins, areas with seasonal fog or capillary rise from shallow water tables, all present ambiguity. A site that looks penetrable in one season may not be in another. Analysts must treat soil moisture as a variable, not a constant.
What the Sahara has already shown
The foundational result in this field came from the Shuttle Imaging Radar missions of the early 1980s. SIR-A and SIR-B L-band imagery over the Egyptian and Sudanese Sahara revealed buried palaeodrainage networks, later confirmed by ground survey, beneath sand sheets several metres thick. That work, published by McCauley and colleagues in Science in 1982, established the empirical basis for everything that followed.
ALOS PALSAR (the predecessor to PALSAR-2) extended the approach systematically. Studies over the Arabian Peninsula and North Africa used L-band backscatter anomalies to identify candidate buried road segments and settlement outlines, cross-checking against CORONA declassified optical imagery and ground-penetrating radar transects. The penetration depths documented in peer-reviewed literature cluster around 1–2 m for L-band in genuinely dry conditions, with occasional claims of deeper penetration where sand is exceptionally dry and coarse-grained.
P-band, which BIOMASS now provides from orbit for the first time, has a wavelength roughly four times longer than L-band. Airborne P-band campaigns over Saharan test sites, conducted as part of ESA's preparatory research, demonstrated backscatter returns from features at depths exceeding 4 m. Whether BIOMASS's 50 m ground resolution is fine enough to resolve individual archaeological features remains an open question. It is well suited to mapping palaeochannel networks and large buried landscape elements; it will not resolve a single room.
The ambiguities that will mislead an incautious analyst
Surface roughness produces backscatter that mimics subsurface returns. A lag gravel surface, a deflation pavement, or a zone of wind-rippled coarse sand can generate high backscatter values for purely surface reasons. Without ancillary data, distinguishing surface roughness anomalies from genuine subsurface structure is genuinely difficult. Multi-frequency comparison helps: if a feature is bright at L-band but absent at C-band, subsurface origin is more plausible. If it is bright at both, suspect the surface.
Subsurface geology is the other major confound. Buried wadi channels, basalt intrusions, calcrete horizons and gypsiferous layers all produce dielectric contrasts that SAR will dutifully image. These are not archaeological. Distinguishing them requires integration with geological maps, airborne geophysics where available, and ideally at least one ground-truth transect. SAR subsurface archaeology is a screening tool, not a ground-truth substitute.
Polarisation matters too. HH polarisation generally penetrates more deeply than VV in granular media, and cross-polarised returns (HV or VH) are sensitive to volumetric scatterers such as rubble or root channels. Quad-polarisation PALSAR-2 data, where available, gives the analyst more to work with than single-polarisation acquisitions, but quad-pol modes trade swath width for polarimetric completeness.
Integrating SAR subsurface data into an archaeological workflow
No responsible archaeological programme treats SAR subsurface anomalies as sites. They are hypotheses. The practical workflow runs from SAR anomaly detection, through comparison with optical archives (CORONA, Landsat, Sentinel-2), through high-resolution DEM analysis for micro-topographic corroboration, and only then to prioritised field investigation or ground-penetrating radar survey.
Multitemporal stacking improves confidence. A buried feature with stable dielectric contrast will appear consistently across acquisitions taken under varying surface conditions. A surface artefact, such as a fresh sand sheet deposited by a recent storm, will not. Comparing ALOS-1 PALSAR data from the mid-2000s with current PALSAR-2 acquisitions gives a baseline spanning nearly two decades, long enough to separate stable subsurface signals from ephemeral surface noise.
Satellize can run this multi-stage screening workflow against client-specified areas of interest, delivering prioritised anomaly maps as GIS layers with associated confidence scores derived from multi-frequency and multi-temporal consistency checks. The underlying method is the same open-literature approach used in academic programmes; the value is systematic coverage at scale rather than site-by-site manual interpretation.
What BIOMASS changes, and what it does not
BIOMASS is the first P-band SAR in orbit with global coverage ambitions. For desert archaeology, its significance is the wavelength. At 69 cm, P-band interacts with buried features at depths that L-band simply cannot reliably reach. Early results from airborne precursor campaigns are encouraging for large-scale landscape archaeology: buried palaeodrainage systems spanning tens of kilometres, ancient road corridors, and the outlines of major buried settlement complexes are plausible targets.
What BIOMASS will not do is replace excavation-scale survey. At 50 m resolution, a buried room is invisible. A buried city quarter is marginal. A buried river system or a major ancient road corridor is detectable. The mission also carries ionospheric correction challenges at P-band that do not affect L-band, and its primary science objective is forest biomass, meaning archaeological tasking depends on data access agreements and acquisition scheduling that were not designed with heritage in mind. Analysts should treat BIOMASS archaeological data as opportunistic for now, and PALSAR-2 as the workhorse.
Typical figures
| Primary frequency (PALSAR-2) | L-band, 1.27 GHz (23.6 cm wavelength) |
| Primary frequency (BIOMASS) | P-band, 435 MHz (69 cm wavelength) |
| Spatial resolution (PALSAR-2 stripmap) | 3–10 m depending on mode; spotlight achieves ~1 m |
| Spatial resolution (BIOMASS standard mode) | Approximately 50 m |
| Revisit (ALOS-2) | 14 days exact repeat; off-track pointing can reduce effective revisit |
| Estimated subsurface penetration depth (L-band, dry sand) | 1–2 m in documented hyper-arid conditions; deeper penetration claimed but less consistently validated |
| Estimated subsurface penetration depth (P-band, dry sand) | Up to 4–5 m in airborne campaign data; orbital validation ongoing |
| Minimum detectable buried feature (L-band) | Features with lateral extent greater than ~10 m and meaningful dielectric contrast; smaller features below reliable detection threshold |
| Archive depth (PALSAR-2 predecessor PALSAR) | ALOS PALSAR data from 2006–2011 available; PALSAR-2 from 2014 to present |
| Key confounds | Surface roughness, subsurface geology (wadi fill, calcrete, basalt), soil moisture above ~2–3% volumetric water content |
Analytics Satellize can run
| Subsurface anomaly map | L-band backscatter intensity analysis with surface roughness correction using C-band reference; multi-polarisation decomposition where quad-pol data available | GIS polygon layer of candidate subsurface features with backscatter intensity values and polarimetric class labels |
| Multi-frequency penetration confidence score | Comparison of L-band and C-band backscatter to separate surface-dominated from subsurface-dominated returns; features bright at L-band and absent at C-band scored as higher-confidence subsurface candidates | Raster confidence layer (0–1 scale) overlaid on anomaly polygons, exported as GeoTIFF |
| Multitemporal stability assessment | Time-series consistency check across PALSAR and PALSAR-2 archive acquisitions; stable anomalies across seasons and years flagged as priority targets | Ranked candidate site list with temporal stability metrics, delivered as CSV and shapefile |
| Palaeochannel network extraction | SAR backscatter thresholding combined with SRTM DEM flow-routing analysis; channel network topology mapped and compared against known ancient settlement distributions | Vector palaeochannel network layer with estimated channel width and relative backscatter magnitude |
| Change detection report (sand mobility vs. stable subsurface signal) | Sentinel-1 C-band multitemporal coherence and intensity change detection to map active aeolian sand movement; used to mask dynamic surface areas from subsurface anomaly maps | Sand mobility mask and updated anomaly map excluding recently mobile sand zones, as GeoTIFF and PDF summary |
| Field investigation priority ranking | Weighted scoring combining multi-frequency confidence, temporal stability, proximity to known sites from published literature, and micro-topographic corroboration from DEM | Prioritised site list with coordinates, confidence tier and supporting evidence summary, formatted for field team briefing |
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.