Land-use change detection for planning enforcement
Satellite change detection exposes unauthorised construction, green-belt encroachment and agricultural conversion before enforcement windows close. This page covers sensor choice, minimum detectable change size, and what courts and planning tribunals actually accept as evidence.
Sensors
- Sentinel-2 MSI: 10 m GSD in visible and near-infrared bands, 5-day revisit at mid-latitudes with both satellites. Sufficient to detect change events at the scale of a large outbuilding or field boundary shift, and free to access. Cloud cover is the primary constraint on usable revisit frequency.
- Maxar WorldView-3: 0.31 m panchromatic, 1.24 m multispectral GSD. Capable of resolving individual footings, vehicle tracks and material stockpiles. Tasked on demand; typical collect latency is 1 to 3 days subject to cloud and satellite geometry. Archive extends to 2014.
- Airbus Pléiades Neo: 0.30 m panchromatic, 1.2 m multispectral GSD. Daily revisit capacity over any point on Earth. The tri-stereo collect mode produces 3D surface models useful for estimating built volume, which matters when enforcement penalties are calculated on floor-area grounds.
- Planet SkySat: 0.50 m panchromatic, approximately 1 m multispectral. Up to 12 collects per day over a tasked area, making it the strongest option for near-real-time change monitoring of a specific parcel under active investigation.
What a changed parcel looks like in spectral space
Bare soil exposed by excavation has a distinct reflectance signature in the red-edge and shortwave-infrared bands that Sentinel-2 captures at 20 m GSD. Newly poured concrete reflects strongly across the visible spectrum and suppresses the near-infrared response that vegetation produces. These differences are measurable even before a structure reaches roof height, which is precisely when enforcement action is cheapest.
Change-vector analysis (CVA) compares two co-registered multispectral images pixel by pixel. The magnitude of the spectral change vector flags where something has altered; the direction of that vector in band space indicates what kind of change occurred. Agricultural-to-built conversion, vegetation clearance and hard-surface extension each produce distinguishable vector directions in a two-dimensional NIR-SWIR feature space. This is not a new idea: CVA was formalised in the remote-sensing literature in the 1980s and has been applied to land-cover monitoring at national scale by programmes including the European Urban Atlas.
The minimum detectable change is not a single number
At Sentinel-2's 10 m GSD, a single pixel covers 100 m². A change event needs to affect several contiguous pixels to be reliably distinguished from noise, so the practical detection floor is roughly 300 to 500 m² for a spectrally distinct surface type under good atmospheric conditions. That is large enough to catch a new barn, a significant hardstanding or a field boundary removal, but it will miss a garden extension or a small outbuilding.
WorldView-3 at 0.31 m panchromatic GSD reduces the detection floor to individual structural elements: a foundation trench, a retaining wall, a single shipping container used as an illicit dwelling. The trade-off is cost and coverage. Tasking a WorldView-3 collect over a single suspect parcel is economical; tasking it over a district to screen for unknown violations is not. The practical workflow pairs Sentinel-2 as the wide-area screening layer with commercial sub-metre imagery as the confirmation and adjudication layer.
Cloud cover is an honest constraint. In temperate maritime climates, persistent cloud can reduce the number of usable Sentinel-2 observations to fewer than four per month in winter. SAR imagery from Sentinel-1 (C-band, 10 m IW mode) can detect significant ground disturbance through cloud, though it does not carry the spectral information needed to classify change type with the same confidence as optical data.
Comparing current imagery against approved plans
Change detection only becomes enforcement-relevant when the detected change is cross-referenced against a legal baseline. That baseline is typically the approved land-use plan, the planning permission record, or a cadastral map showing permitted use. In practice, this means co-registering satellite-derived change polygons with a GIS layer of zoning designations. Any change polygon that falls within a green-belt zone, an agricultural protection area or a conservation boundary becomes a candidate violation.
The accuracy of this intersection depends heavily on the geometric accuracy of both datasets. WorldView-3 and Pléiades Neo carry on-board GPS and star-tracker attitude data that support positional accuracy of 3 to 5 m CE90 without ground control points, and better than 1 m CE90 with ground control. Cadastral datasets in many jurisdictions carry their own positional uncertainty, sometimes larger than the imagery error. That discrepancy needs to be quantified and disclosed, not hidden, when results are presented to a planning authority.
What planning tribunals and courts actually require
Satellite imagery has been admitted as evidence in planning enforcement proceedings in the United Kingdom, the European Union and several Commonwealth jurisdictions. The evidentiary bar is not primarily about image resolution. It centres on chain of custody, metadata integrity and expert interpretation.
A tribunal will want to know: when exactly was the image acquired, by what sensor, under what atmospheric conditions, and who processed it and how. The image metadata embedded in commercial products from Maxar and Airbus includes acquisition timestamp, satellite ID, sun angle and off-nadir angle. Preserving that metadata in an unmodified form, and having a qualified analyst sign a statement of interpretation, is more important to admissibility than achieving sub-metre resolution.
Temporal bracketing matters enormously. If an enforcement notice requires proof that a structure did not exist before a specific date, you need imagery from before that date showing absence, and imagery from after showing presence. Archive depth is therefore a material factor. Maxar's WorldView archive extends to 2014; Sentinel-2 archive begins in 2015. For disputes about changes that predate those archives, historical aerial photography held by national mapping agencies is often the only option, and its integration with current satellite data requires careful co-registration.
Running the workflow at scale
A local planning authority monitoring hundreds of parcels cannot task commercial imagery over all of them continuously. The scalable approach is a two-stage triage: automated Sentinel-2 change detection runs across the entire jurisdiction on every cloud-free acquisition, flagging parcels where spectral change exceeds a calibrated threshold. Flagged parcels then trigger commercial tasking for confirmation. This keeps commercial imagery costs proportionate to actual alert volume.
Satellize runs this kind of multi-sensor change pipeline for clients who need sovereign-grade data custody and audit trails. The analytics architecture is the same one that underpins the Tonga crop-estimation programme, adapted for built-environment change rather than agricultural yield. The Overhead column has covered several public-record cases where satellite change detection has been cited in enforcement proceedings, and those analyses are freely available.
One practical note on latency: Sentinel-2 data is typically available in the Copernicus Data Space within 24 hours of acquisition. Commercial tasking products from Maxar and Airbus are usually delivered within 4 to 8 hours of the collect completing. For most planning enforcement purposes, same-day or next-day data is adequate. The exception is a case where an operator is racing to demolish evidence before an inspector arrives, in which case SkySat's intra-day revisit capacity is the appropriate tool.
Typical figures
| Screening spatial resolution | 10 m (Sentinel-2 visible/NIR); 20 m (Sentinel-2 red-edge/SWIR) |
| Adjudication spatial resolution | 0.30 to 0.50 m panchromatic (Pléiades Neo, WorldView-3, SkySat) |
| Screening revisit | 5 days at mid-latitudes (Sentinel-2 two-satellite constellation); cloud-limited in practice |
| Commercial tasking latency | 1 to 3 days for first collect; 4 to 8 hours from collect to delivery (Maxar, Airbus) |
| Minimum detectable change area (optical screening) | 300 to 500 m² (Sentinel-2, spectrally distinct surface); ~5 m² (sub-metre commercial) |
| Key spectral bands for change classification | NIR (842 nm), red-edge (705 nm, 740 nm), SWIR (1610 nm, 2190 nm) on Sentinel-2 MSI |
| Positional accuracy (commercial, without GCP) | 3 to 5 m CE90 (WorldView-3, Pléiades Neo); better than 1 m CE90 with ground control |
| Archive depth | Sentinel-2 from 2015; WorldView from 2014; Landsat from 1972 (coarser resolution) |
| Delivery formats | GeoTIFF, COG, GeoPackage, shapefile; change polygons as GeoJSON or KML for GIS ingestion |
| SAR cloud-penetration option | Sentinel-1 C-band IW mode, 10 m GSD, 6-day revisit; detects ground disturbance, not spectral class |
Analytics Satellize can run
| Jurisdiction-wide change alert layer | Automated change-vector analysis on Sentinel-2 bi-temporal image pairs, thresholded by spectral magnitude and direction in NIR-SWIR feature space | Weekly GeoJSON alert feed of candidate-violation polygons, attributed with change magnitude, date range and zone classification |
| Parcel-level confirmation report | Sub-metre commercial imagery (WorldView-3 or Pléiades Neo) co-registered to cadastral boundary; before/after composite with analyst annotation | PDF enforcement dossier with timestamped imagery, metadata extract, positional accuracy statement and analyst interpretation declaration |
| Temporal change timeline | Dense time-series analysis across all available Sentinel-2 and archive commercial collects over a named parcel; change onset date estimated by breakpoint detection | Chronological image strip with change onset date range and confidence interval, formatted for tribunal submission |
| Built-volume estimate | Pléiades Neo tri-stereo DSM differencing against pre-construction baseline surface model; volume computed from height delta over change footprint | GIS polygon with estimated floor area and above-ground volume in cubic metres, relevant to penalty calculation |
| Green-belt and zoning intersection layer | Spatial join of detected change polygons against client-supplied approved land-use plan GIS layers; flags violations by zone category | Attributed shapefile of intersecting violations, ranked by zone sensitivity, for prioritisation by enforcement officers |
| SAR-optical fusion alert (cloud-persistent monitoring) | Sentinel-1 coherence change detection combined with Sentinel-2 optical flags; SAR triggers alert when optical data is unavailable due to cloud | Unified alert feed with source-sensor flag, enabling continuous monitoring regardless of weather conditions |
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.