Solar illumination angle optimisation for earthwork and enclosure shadow enhancement

Very-high-resolution satellite imagery acquired at low solar elevation angles can reveal earthworks, ditches and banks of only 20–50 cm relief by the shadows they cast. Systematic tasking geometry, not sensor resolution alone, determines what survives detection.

Sensors

Pleiades 1A/1B: 50 cm panchromatic resolution with agile off-nadir pointing up to 47 degrees; operator tasking parameters allow specification of solar elevation angle windows, making deliberate low-sun acquisitions practical. Revisit at a given point is roughly daily when both satellites are tasked together.
Pleiades Neo (3/4): 30 cm native panchromatic resolution, improving the minimum detectable shadow length for a given bank height. The same agile tasking model applies. At 30 cm ground sampling distance, a 1 m shadow is resolvable, which corresponds to a roughly 17 cm bank at a 10-degree solar elevation.
WorldView-2: 46 cm panchromatic resolution and an 8-band multispectral stack; agile enough for off-nadir tasking. The multispectral bands are less critical for shadow detection than for separating shadow tone from dark soil, reducing false positives in textured ground.
WorldView-3: 31 cm panchromatic resolution, the finest commercially available at the time of writing. SWIR bands can assist in discriminating shadow from moisture, though for earthwork shadow work the panchromatic channel carries most of the geometric signal.

The geometry that makes a 30 cm bank visible from 600 km

Shadow length is determined by a straightforward trigonometric relationship: length equals feature height divided by the tangent of the solar elevation angle. At a solar elevation of 10 degrees, the tangent is approximately 0.176, so a 30 cm bank casts a shadow roughly 1.7 m long. At 20 degrees elevation the same bank produces only 0.82 m of shadow; at 45 degrees, 30 cm. The implication is stark: halving the solar elevation angle more than doubles the shadow, and the gain accelerates as the sun approaches the horizon.

At 50 cm ground sampling distance, a 1.7 m shadow spans about three to four pixels, which is the practical minimum for confident linear detection. This means a 10-degree solar elevation is close to the useful floor for Pleiades imagery over features of 30 cm relief. Pleiades Neo at 30 cm pushes the detectable floor down to roughly 17–20 cm at the same sun angle, though atmospheric haze and surface texture introduce noise at those limits. The geometry is clean; the complications are radiometric.

Why low sun and good data quality pull in opposite directions

Low solar elevation is exactly the condition that degrades image radiometry. Longer atmospheric path length increases scatter and reduces contrast in shadowed areas. Adjacency effects from bright ground adjacent to shadow can compress the tonal range that distinguishes a shallow ditch shadow from ordinary variation in grass colour. In winter at British latitudes, a 10-degree solar elevation at midday is achievable, but the sun is also in the south, meaning east-west oriented earthworks cast shadows northward while north-south features may be entirely self-shadowed or shadow-free depending on aspect.

The practical approach used in published surveys over British and French field systems is to specify a solar elevation window of 8 to 15 degrees and to request imagery from two or more cardinal illumination directions across separate acquisitions. Features perpendicular to one illumination azimuth that are invisible in that pass become visible in a second pass from a different azimuth. Airborne survey programmes operated by Historic England have used exactly this logic for decades; satellite tasking now applies the same discipline, with the added constraint that the operator cannot fly a repeat pass at will and must negotiate acquisition windows with the satellite operator months in advance.

What the published record shows over British and French landscapes

Peer-reviewed work published in the Journal of Archaeological Science and in Remote Sensing (MDPI) has demonstrated detection of Romano-British field systems and Iron Age enclosures in Pleiades and WorldView imagery acquired under low-angle illumination, with feature relief confirmed against LiDAR-derived digital terrain models. The satellite detections are not independent discoveries in those studies; they are validations against known features. That is the correct scientific approach: establish what the sensor can and cannot see before claiming discovery.

French researchers working on Gallo-Roman cadastral systems in Provence have used Pleiades tasked at solar elevations below 15 degrees to trace field boundaries invisible in standard archive imagery. The key finding across these programmes is that archive imagery, acquired opportunistically under whatever sun angle happened to prevail at collection time, misses a substantial fraction of subtle earthworks that deliberate low-angle tasking recovers. The archive is not a substitute for systematic tasking; it is a starting point.

Processing: from raw pixel to mapped feature

Standard pre-processing applies: orthorectification using a high-quality DEM (ideally LiDAR-derived where available, or the satellite's own stereo DSM), atmospheric correction to surface reflectance, and pansharpening to merge the panchromatic and multispectral channels. After that, the analysis is largely geometric rather than spectral. Shadow edges are enhanced using directional gradient filters oriented perpendicular to the illumination azimuth. Morphological operators then extract linear and curvilinear features above a minimum length threshold.

Automatic extraction works well for long, straight features such as Roman field boundaries. Irregular curvilinear enclosures require either manual digitising over the enhanced imagery or more sophisticated object-based image analysis. False positives are common: field drains, tractor wheelings and modern fence lines all cast shadows of similar geometry. Filtering against known modern infrastructure layers removes most of these. Residual ambiguity is honest and should be reported; satellite shadow enhancement identifies candidate features, not confirmed archaeological sites.

Tasking constraints buyers rarely anticipate

Solar elevation at a given location is fixed by date, time and latitude. For a site at 52 degrees north (central England), solar elevation at local noon reaches a minimum of roughly 14 degrees at the winter solstice and a maximum of about 61 degrees at midsummer. The window in which elevation falls below 15 degrees at any time of day is narrow in autumn and spring, and essentially absent in summer. Buyers planning a shadow-enhancement campaign over a British site need to schedule tasking between October and February.

Cloud cover compounds the problem. The UK averages cloud cover above 60 percent in winter months, and commercial tasking contracts typically require multiple attempts to achieve a single clear acquisition in a specified sun-angle window. Realistic planning should budget for three to five tasking attempts per target area per season. Pleiades and WorldView operators offer best-effort and guaranteed-delivery contract tiers at different price points; the guaranteed tier is worth serious consideration for time-critical heritage survey work where the low-sun window closes for another year if the attempt fails.

Satellize structures its commercial tasking engagements to include explicit sun-angle constraints in the tasking brief, drawing on the same geometric planning approach used in its Tonga crop-estimation programme, where acquisition timing relative to seasonal conditions was equally non-negotiable.

Honest limits of the method

Shadow enhancement detects relief. It does not detect buried features that have no surface expression whatsoever; those require thermal, spectral or radar methods covered in sibling pages of this library. Features beneath dense vegetation canopy are invisible regardless of sun angle. Ploughed land destroys the micro-relief that generates shadows, so the method is most productive over permanent pasture, rough grazing and unimproved grassland, precisely the land types that have preserved earthworks in the first place.

Resolution is not the binding constraint in most cases. A 50 cm sensor is sufficient for features with relief above 25–30 cm at a 10-degree sun angle. The binding constraints are cloud probability, the narrowness of the low-sun window at the target latitude, and the availability of a suitable illumination azimuth to cross the dominant orientation of the earthworks being sought. Plan around those three factors first, then choose your sensor.

Typical figures

Spatial resolution (panchromatic)	30 cm (WorldView-3, Pleiades Neo) to 50 cm (Pleiades 1A/1B, WorldView-2)
Minimum detectable shadow length	Approximately 1.0–1.5 m (3–4 pixels at 30–50 cm GSD)
Minimum detectable bank height at 10° solar elevation	~17 cm (30 cm sensor) to ~26 cm (50 cm sensor); theoretical floor, not guaranteed in practice
Optimal solar elevation window	8–15 degrees; below 8 degrees atmospheric haze degrades contrast; above 20 degrees shadow length drops sharply for sub-50 cm features
Useful tasking season (52° N)	October to February for sub-15° solar elevation at midday
Revisit (tasked, single satellite)	1–3 days depending on off-nadir tolerance; daily with dual Pleiades constellation
Archive depth	Pleiades archive from 2012; WorldView-2 from 2009; low-sun-angle archive imagery sparse and largely opportunistic
Spectral bands used	Panchromatic primary; multispectral (4–8 bands) for shadow/soil discrimination and false-positive reduction
Delivery formats	Orthorectified GeoTIFF, pansharpened bundle; vector feature layers in GeoPackage or Shapefile
Cloud cover constraint	Less than 10% over target area required; UK winter average cloud cover 60–70%, requiring multiple tasking attempts

Analytics Satellize can run

Sun-angle acquisition plan	Solar geometry calculation (azimuth, elevation) for target coordinates across a defined date range, cross-referenced against historical cloud-probability climatology	PDF tasking brief specifying optimal acquisition windows, illumination azimuths and fallback dates for the target site
Orthorectified low-angle image mosaic	Rigorous sensor model orthorectification using LiDAR or stereo DSM; atmospheric correction to surface reflectance; pansharpening (Gram-Schmidt or IHS)	GeoTIFF mosaic at native resolution, projected to national grid of client's jurisdiction
Directional shadow-enhancement layer	Directional gradient filtering (Sobel, Prewitt or anisotropic diffusion) oriented perpendicular to illumination azimuth; contrast-limited adaptive histogram equalisation in shadow zones	Enhanced GeoTIFF suitable for manual interpretation and digitising, with filter parameters documented
Candidate earthwork feature vectors	Object-based image analysis (OBIA) with morphological filtering; length, linearity and orientation thresholds applied; modern infrastructure mask subtracted	GeoPackage of candidate linear and curvilinear features with confidence scores and orientation attributes
Multi-azimuth composite detection map	Logical union of feature vectors extracted from two or more acquisitions at contrasting illumination azimuths; features confirmed in one direction and absent in another flagged for review	GIS layer combining detections from multiple passes, with per-feature illumination-direction provenance recorded
Shadow-length relief estimate	Measured shadow length converted to height estimate via tan(solar elevation); compared against available LiDAR DTM where present to assess agreement and sensor performance	Attribute table appended to feature vectors with estimated relief range and LiDAR validation flag

Sources

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.