Cadastral boundary verification using sub-metre imagery
Sub-metre commercial imagery lets land registries check registered parcel boundaries against physical reality at scale, flagging encroachments and area discrepancies before they become disputes. Accuracy requirements are strict and the legal status of satellite-derived boundaries is not equivalent to a ground survey.
Sensors
- Airbus Pléiades Neo: 30 cm native panchromatic resolution, 70 cm multispectral. Stereo and tri-stereo collection modes support DSM generation. CE90 geolocation accuracy without ground control points is published at better than 0.5 m, which is the threshold for cadastral-grade work. Revisit roughly twice daily at mid-latitudes.
- Maxar WorldView Legion: 30 cm panchromatic. Constellation of six satellites targets revisit of 15 passes per day over priority areas. Stereo capability and Maxar's published CE90 of less than 0.4 m with refined attitude data makes it directly competitive with Pléiades Neo for boundary work.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral, plus SWIR and CAVIS bands. Extensive archive depth from 2014 onwards is useful for retrospective encroachment dating. CE90 with vendor-supplied RPCs and a DEM is typically cited at 1 m or better, marginally above the 0.5 m threshold without GCPs.
- Planet SkySat: 50 cm panchromatic, 1 m multispectral. Rapid revisit and competitive pricing make it attractive for wide-area screening, but its published CE90 without GCPs sits around 1 m, which is insufficient for definitive boundary measurement. Useful as a triage layer to prioritise where sub-metre tasking is needed.
What the 0.5 m CE90 requirement actually means in practice
CE90 is the radius of a circle within which 90 per cent of measured points fall relative to their true ground position. A CE90 of 0.5 m means that nine out of ten pixels are placed within half a metre of where they belong on the ground. For cadastral work this matters because a boundary dispute often turns on encroachments of 20 to 50 cm. An image with a 1.5 m CE90 cannot reliably resolve that question; it can only confirm gross discrepancies.
Pléiades Neo and WorldView Legion both publish CE90 figures below 0.5 m under nominal collection conditions, without ground control points, relying on onboard star-tracker attitude determination and precise orbit data. That is a significant engineering achievement. It does not mean every collect achieves it. Off-nadir angles above roughly 25 degrees, atmospheric refraction at low sun angles, and residual DEM errors all degrade real-world accuracy. Any serious cadastral programme should budget for at least a sparse network of ground control points to validate and, where necessary, refine the vendor-supplied rational polynomial coefficients.
Orthorectification: where the DEM matters more than the sensor
A raw satellite image is a perspective projection. Buildings lean outward, hillsides are compressed, and every elevation change shifts pixels laterally. Orthorectification removes that distortion by projecting each pixel onto a flat datum using a digital elevation model. For flat terrain the process is forgiving. For hilly cities or areas with tall buildings, DEM quality becomes the binding constraint on final positional accuracy.
The Copernicus DEM (derived from TanDEM-X) is available globally at 30 m and 10 m posting and is the practical baseline for most programmes. In dense urban areas, a 30 m DEM cannot resolve individual building heights, so building lean in very-high-resolution imagery is only partially corrected. A lidar-derived DSM at 0.5 m posting can reduce orthorectification residuals to centimetre level, but acquiring one defeats the purpose of a low-cost satellite screening programme. The practical compromise is to use the best available DEM, flag parcels in high-relief or tall-building zones as requiring field verification, and document the expected residual error in the methodology.
One underappreciated issue: cadastral registers often reference a legal datum (OSGB36 in the UK, for example) that differs from the WGS84 ellipsoid used by satellite imagery. Datum transformation errors of 1 to 3 m are common if this step is skipped. Any workflow that compares satellite-derived geometry to registered coordinates must apply the correct datum shift before drawing conclusions.
What the imagery can and cannot detect
At 30 cm resolution, physical boundary markers are often directly visible: walls, fences, hedgerows, kerb lines, drainage channels and changes in surface material all appear as distinct linear features. Automated edge-detection and segmentation algorithms can extract these features at scale, producing a vector layer of observed boundaries that can be differenced against the registered cadastre.
The limits are real. A boundary that runs through open grassland with no physical marker leaves no spectral signal. A shared wall sitting precisely on the legal line is indistinguishable from one that has been built 30 cm into a neighbour's parcel. Underground boundaries, such as the footprint of a basement extending beyond the above-ground structure, are invisible to optical sensors entirely. Vegetation overhanging a boundary obscures the ground beneath it. Cloud cover over tropical and equatorial cities can make consistent annual coverage difficult even with high-revisit constellations. None of this disqualifies satellite imagery as a screening tool, but it does mean that a flagged discrepancy is a referral for investigation, not a finding of encroachment.
The legal gap between a satellite indicator and a surveyed boundary
This distinction is not a technicality. In most jurisdictions, a cadastral boundary has legal force only when established by a licensed surveyor using methods prescribed in statute, typically total station or GNSS survey with centimetre-level accuracy and a formal field record. A satellite-derived boundary indicator has no equivalent legal standing. It is evidence, not determination.
Several national land registries, including those in the Netherlands and parts of Scandinavia, have published frameworks for using very-high-resolution imagery as a first-pass consistency check, with discrepancies above a defined threshold triggering mandatory field survey. That is the appropriate model. Satellite imagery reduces the cost and time of identifying which parcels need attention; it does not replace the surveyor who resolves them. Programmes that present satellite outputs as definitive boundary determinations create legal liability and, in disputed cases, tend to be challenged successfully in court.
The practical deliverable from a satellite-based cadastral screening programme is therefore a risk-ranked list: parcels where the observed physical boundary diverges from the registered boundary by more than a defined tolerance (say, 0.5 m after accounting for imagery CE90), with the magnitude and direction of divergence recorded. That list then drives a prioritised field survey programme, which is far cheaper than resurveying an entire jurisdiction from scratch.
Running the analysis at national scale
A national cadastral screening programme typically works in three phases. First, archive imagery is used for broad coverage where recency matters less than cost. WorldView-3's archive from 2014 onwards covers most urban areas of interest. Second, fresh tasking of Pléiades Neo or WorldView Legion is commissioned for high-priority zones, dense urban cores, or areas with recent development pressure. Third, change detection between archive and current imagery identifies parcels where the physical boundary has shifted since the last registration event, which is often a stronger signal of encroachment than a static comparison alone.
Satellize structures this kind of multi-source workflow for government clients, combining open-data orthorectification inputs (Copernicus DEM, Sentinel-2 for context) with commercial tasking on client licence. The analytic output is a GIS layer and ranked discrepancy report, not raw imagery. For reference on the analytics approach, the Tonga crop-estimation programme uses a comparable multi-source fusion model, though the application domain is entirely different.
Processing at national scale requires tiling strategies, consistent radiometric normalisation across different collect dates, and careful handling of the seam lines where adjacent images were acquired under different sun angles. These are solved problems in photogrammetric practice, but they require explicit quality-control steps that are often underspecified in procurement documents. Buyers should ask vendors to document their orthorectification workflow, the DEM source used, and the GCP validation procedure before accepting deliverables.
Typical figures
| Best available panchromatic resolution | 30 cm (Pléiades Neo, WorldView Legion, WorldView-3) |
| CE90 geolocation accuracy (no GCPs) | < 0.5 m (Pléiades Neo, WorldView Legion); ~1 m (WorldView-3, SkySat) |
| CE90 with GCP refinement | 0.1–0.3 m achievable with well-distributed GCPs and accurate DEM |
| Revisit rate | Up to 15 passes/day (WorldView Legion priority areas); ~2/day (Pléiades Neo mid-latitudes) |
| Minimum detectable boundary feature | Linear features ~0.5 m wide (wall, fence) reliably visible at 30 cm resolution |
| DEM input (standard) | Copernicus DEM GLO-30 (30 m posting) or GLO-10 (10 m posting), TanDEM-X derived |
| Archive depth | WorldView-3 from 2014; Pléiades from 2012; SkySat from 2016 |
| Delivery formats | GeoTIFF (orthorectified), GeoPackage or Shapefile (boundary vectors), PDF discrepancy report |
| Cloud cover constraint | Optical only; persistent cloud cover in tropical cities may require multi-date compositing over weeks |
Analytics Satellize can run
| Boundary discrepancy layer | Automated edge detection and segmentation of physical boundary features (walls, fences, hedgerows) in orthorectified imagery, differenced against registered cadastral vectors | GIS vector layer (GeoPackage/Shapefile) with per-parcel divergence magnitude and direction |
| Encroachment risk ranking | Statistical ranking of parcels by divergence magnitude relative to CE90 uncertainty budget; parcels above threshold flagged for field referral | Ranked CSV and PDF report with parcel IDs, divergence values and recommended action |
| Area discrepancy analysis | Photogrammetric footprint measurement from orthorectified imagery compared to registered parcel area; percentage and absolute deviation computed per parcel | Tabular report with registered vs. measured area, percentage deviation, and confidence band |
| Retrospective encroachment dating | Multi-date change detection using archive imagery (WorldView-3 from 2014, Pléiades from 2012) to identify when a physical boundary feature first appeared or shifted | Timeline annotation per flagged parcel indicating earliest detectable date of discrepancy |
| Orthorectification quality report | GCP-based residual analysis and DEM-source documentation; per-tile CE90 estimate reported alongside imagery deliverables | Accuracy metadata file (XML/PDF) accompanying each imagery tile |
| Field survey prioritisation map | Spatial clustering of high-risk parcels combined with access-route analysis to optimise field survey routing | Prioritised survey schedule and map for ground-truth teams |
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.