Unlicensed construction and land-use violation detection
Satellite change-detection compares current very-high-resolution imagery against cadastral baselines to flag structures built without permission. New concrete and metal roofing have distinct spectral and textural signatures that algorithms can isolate at scale, long before an inspector arrives on site.
Sensors
- Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral (8 bands including SWIR), tasked on demand. At this resolution individual roof panels, building footprints down to roughly 5 m² and construction-material type are distinguishable. Revisit at mid-latitudes is typically 1 to 4.5 days depending on off-nadir tolerance.
- Planet SuperDove: 3 m multispectral (8 bands, 400–900 nm), near-daily global revisit. Ideal for frequent change screening across large jurisdictions; footprint detection is reliable for structures above roughly 15–20 m² in plan area. The dense time-series is the main asset: construction that takes weeks is visible as a step-change in reflectance.
- Planet SkySat: 50 cm panchromatic, 0.8 m colour, tasked. Sits between SuperDove and WorldView-3 in cost and resolution; useful for confirming flagged sites before committing to a WorldView-3 collect or a field visit.
- Sentinel-2 MSI: 10 m (visible and NIR), 20 m (red-edge and SWIR), 5-day revisit at the equator, free and open. Practical for monitoring large rural plots or tracking construction progress on major unlicensed infrastructure. Individual residential extensions are below the detection floor; aggregate land-use change is not.
What new concrete tells a spectrometer
Freshly poured concrete and galvanised metal roofing share a high broadband reflectance across visible and near-infrared wavelengths, typically 40–60% in the 0.5–0.9 µm range, compared with the 5–15% reflectance of bare soil or established vegetation at the same bands. That contrast is not subtle. An 8-band sensor such as WorldView-3 or SuperDove can resolve it even when the structure is small, because the spectral signature of the material differs from its surroundings rather than blending with them.
SWIR bands (WorldView-3 carries eight SWIR channels centred between 1.195 µm and 2.365 µm) add a second discriminant: concrete and metal have characteristic absorption features that distinguish them from clay tiles, thatch or asphalt. This matters in dense urban settings where everything looks grey in a panchromatic image. Texture metrics, specifically the local variance in pixel values across a small kernel, further separate smooth new slab roofs from the rougher texture of older built fabric or disturbed earth.
The baseline problem: what you are comparing against
Change detection is only as good as the reference it compares against. The standard workflow anchors on one of three baselines: a georectified cadastral polygon layer showing permitted structures, an earlier satellite or aerial image predating the alleged violation, or a combination of both. Cadastral records vary enormously in quality. In countries with well-maintained land registries, polygon boundaries are accurate to 0.5–1 m; in others, the cadastre is decades out of date and itself requires correction before it can serve as a legal reference.
Where the cadastre is unreliable, a bitemporal image pair, typically one image from before a planning application deadline and one current collect, becomes the primary evidence. Automated change-detection algorithms (image differencing, principal-component analysis of the change vector, or supervised classifiers trained on labelled construction examples) flag pixels or objects that have shifted beyond a defined threshold. The output is a candidate list, not a verdict. Every flagged polygon still requires human review, and ideally field confirmation, before enforcement action. Algorithms produce false positives from seasonal vegetation change, shadow shifts between acquisitions, and sensor-angle differences.
Resolution floors and what falls below them
The practical detection floor depends on the sensor and the structure type. At 3 m (SuperDove), a single-storey residential extension of 20 m² or less is unlikely to be reliably detected as a distinct object; it may shift a pixel's spectral value without producing a clean new polygon. At 30 cm (WorldView-3), extensions down to roughly 4–6 m² are detectable, though the confidence drops sharply for structures partially occluded by existing roof overhangs or mature tree canopy.
Cloud cover is a persistent operational constraint. Optical sensors cannot see through cloud, which in humid tropical and subtropical regions means that some months-long construction projects complete before a clear-sky collect is available. Radar (Sentinel-1 C-band SAR) can penetrate cloud and has been used to detect building footprints via coherence change, though at 5 m IW mode resolution it is better suited to detecting that construction has occurred than to characterising what was built. Combining SAR for timing and optical for characterisation is the most reliable approach in cloud-prone jurisdictions.
How national land registries are using this in practice
Several European national mapping and cadastral agencies, including those operating under EU INSPIRE Directive obligations, have published pilot results using Sentinel-2 and commercial VHR imagery for automated building-change detection. The standard workflow produces a GIS layer of candidate new or modified footprints, ranked by confidence score, which planning enforcement officers use to prioritise site visits. The satellite does not issue the enforcement notice; it compresses the triage step from months of aerial-survey cycles to days.
The honest limitation of the satellite layer is temporal resolution at the high end. A WorldView-3 tasked collect costs money and is not run daily over every municipality. Practical programmes typically use Planet SuperDove as a continuous screening layer (near-daily, low cost per km²) and trigger a SkySat or WorldView-3 collect only when the screening layer flags a candidate. This two-tier architecture keeps costs proportionate to the enforcement caseload.
Evidence quality and legal admissibility
Satellite imagery is increasingly accepted as supporting evidence in planning enforcement proceedings, but its admissibility depends on chain-of-custody documentation, sensor calibration records and the provenance of the analysis. Maxar and Planet both provide sensor metadata, acquisition geometry and radiometric calibration coefficients with commercial delivers. These records matter: a defence lawyer will ask whether the apparent new structure could be an artefact of a different sun angle between the two acquisitions, or a shadow from a neighbouring building not present in the baseline image.
Orthorectification accuracy is a related concern. WorldView-3 achieves CE90 positional accuracy of roughly 3.5 m without ground control points, improving to sub-metre with GCPs. At cadastral boundaries measured in tens of centimetres, that residual error can be legally significant. Programmes that intend to use imagery as primary evidence, rather than as a triage tool, should budget for GCP collection or tie-point registration against a surveyed reference layer.
Satellize structures analytics workflows to produce audit-ready outputs: timestamped imagery, documented processing steps, and a confidence classification that explicitly flags ambiguous cases for field resolution rather than forcing a binary determination from the algorithm alone. The same discipline applied to the Tonga crop-estimation programme, where output uncertainty bounds are reported alongside the headline figures, applies here.
Archive depth as an enforcement tool
Planet's archive extends to 2016 for SuperDove-class imagery over most of the globe; Maxar's archive for WorldView-2 and WorldView-3 goes back to 2009 and 2014 respectively. Sentinel-2 coverage begins in 2015. This means that for a structure claimed to predate a particular planning regime, satellite imagery can often provide a dated record of when the land surface changed. That is a materially different capability from a single current collect: it allows investigators to establish not just that an unlicensed structure exists, but approximately when it appeared, which is often the decisive question in enforcement proceedings where a statute of limitations applies.
Typical figures
| Finest spatial resolution (optical) | 30 cm panchromatic (Maxar WorldView-3) |
| Multispectral resolution | 1.24 m (WorldView-3), 3 m (Planet SuperDove), 10 m (Sentinel-2 visible/NIR) |
| Revisit cadence | Near-daily (Planet SuperDove); 1–4.5 days tasked (WorldView-3/SkySat); 5 days (Sentinel-2) |
| Minimum detectable new footprint | ~4–6 m² at 30 cm; ~15–20 m² at 3 m; not reliable below ~200 m² at 10 m |
| Spectral bands relevant to material discrimination | Visible, NIR, red-edge, SWIR (WorldView-3 8-band SWIR 1.195–2.365 µm) |
| Cloud penetration | None (optical); Sentinel-1 C-band SAR (5 m IW) used for cloud-proof timing layer |
| Positional accuracy (WorldView-3, no GCP) | CE90 ~3.5 m; sub-metre with ground control points |
| Archive depth | 2009 (WorldView-2), 2014 (WorldView-3), 2015 (Sentinel-2), 2016 (Planet SuperDove) |
| Coverage scalability | Sentinel-2 and Planet provide global or national-scale screening; VHR tasked to flagged sites only |
| Delivery formats | GeoTIFF imagery, GeoJSON/Shapefile change polygons, confidence-scored CSV candidate lists |
Analytics Satellize can run
| Baseline construction footprint map | Object-based image analysis (OBIA) applied to VHR optical imagery; buildings segmented by spectral and textural properties | GIS polygon layer of existing permitted structures, georeferenced to cadastral datum, with material-class attribute |
| Automated change-detection screening layer | Bitemporal image differencing and change-vector analysis on Planet SuperDove time-series; flagged pixels aggregated to candidate polygons above a minimum area threshold | Weekly or monthly GeoJSON alert feed of candidate new footprints, ranked by confidence score, covering a defined administrative area |
| VHR confirmation collect and characterisation | Tasked WorldView-3 or SkySat acquisition over flagged candidates; OBIA and SWIR-based material classification | Per-site report with annotated imagery, footprint area, estimated construction date range and material classification |
| Retrospective construction-date estimation | Time-series analysis of Planet and Sentinel-2 archive to identify the image epoch in which the spectral change first exceeds detection threshold | Dated timeline chart and supporting imagery strips for use in enforcement proceedings |
| SAR coherence change layer (cloud-prone regions) | Sentinel-1 interferometric coherence change detection between two C-band acquisitions; loss of coherence indicates surface disturbance consistent with construction | GeoTIFF coherence-difference layer flagging disturbance events, merged with optical screening output |
| Enforcement prioritisation score | Spatial join of candidate footprints against zoning, protected-area and flood-risk layers; scoring by violation severity class | Ranked enforcement queue in tabular and GIS format, with supporting evidence package per site |
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.