Commercial and industrial land-use zoning compliance audit
Satellite imagery combined with SAR activity signatures and object-based classification can identify parcels operating outside their gazetted zoning category, flagging candidates for ground inspection before enforcement action.
Sensors
- Planet SkySat: 50 cm native resolution (resampled to 50 cm product), revisit up to once or twice daily over tasked areas. Resolves individual vehicles, rooftop plant and storage yard extent. Suitable for detecting large-vehicle presence and yard activity that distinguishes industrial use from residential.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral, 8 VNIR bands plus 8 SWIR bands. SWIR bands discriminate roofing materials (metal, asphalt, concrete) and surface coatings that correlate with land-use category. Tasked revisit typically 1 to 4.5 days depending on latitude and cloud. The highest-resolution commercial option for parcel-level object identification.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, 4 spectral bands. Constellation of two satellites gives same-day stereo capability and sub-daily revisit over priority areas. Comparable to WorldView-3 for object detection; stereo pairs allow 3-D rooftop structure characterisation.
- Sentinel-1 SAR (C-band): IW mode delivers 10 m resolution, 6-day repeat at equator (3 days with both satellites). SAR backscatter is sensitive to metallic surfaces, large vehicles and industrial structures regardless of cloud cover or time of day. Temporal coherence change detection over 6- to 12-month stacks reveals persistent heavy-vehicle activity or new hard-surface yards that optical imagery alone may miss.
- Planet PlanetScope: 3 m resolution, near-daily global revisit, 8-band SuperDove. Insufficient resolution for individual vehicle identification, but the dense time series supports land-use activity classification via temporal spectral signatures across a full calendar year, separating residential from commercial patterns by surface reflectance variance.
What a zoning violation actually looks like from 400 kilometres up
A residential parcel operating as an unlicensed light-industrial site leaves a specific set of observable signatures. Rooftop mechanical plant (air-handling units, condensers, extraction stacks) appears as raised metallic structures with distinct SWIR reflectance. Storage yards show compacted bare soil or concrete with low NDVI and high albedo. Heavy goods vehicles, if present during imaging, are identifiable at 50 cm or better as objects 8 to 14 metres long with characteristic aspect ratios. None of these features are definitive alone, but in combination they form a signature that diverges sharply from the expected residential or light-commercial pattern.
The practical detection floor is set by parcel size and sensor resolution. A WorldView-3 or SkySat collect can resolve a single articulated lorry parked in a rear yard. Sentinel-1 cannot resolve individual vehicles at 10 m, but a time series of 20 to 30 SAR acquisitions over six months will reveal whether a site shows persistently high C-band backscatter consistent with metallic surfaces and vehicle movement, a pattern that residential parcels do not produce. The combination of high-resolution optical for object identification and SAR for temporal activity confirmation is more reliable than either sensor alone.
The classification pipeline: from pixel to parcel verdict
The workflow begins with spectral classification of the imagery into broad land-cover classes: built impervious, vegetated, bare soil, water. Object-based image analysis (OBIA) then segments the scene into meaningful objects at the parcel scale, grouping pixels by spectral similarity and shape rather than treating each pixel independently. This matters because a residential rooftop and an industrial rooftop may share similar reflectance in a single band but differ in size, texture, adjacency to yard space and the presence of ancillary structures.
Against the OBIA output, a set of discriminating features is computed: rooftop material class from SWIR ratios, impervious yard area as a fraction of parcel footprint, vehicle object count and dimensions, presence of loading bays or ramps detectable from shadow geometry, and NDVI of the parcel interior. These features feed a supervised classifier trained on labelled examples of compliant and non-compliant parcels from the same or similar urban contexts. The output is a per-parcel compliance probability score, not a binary verdict. Parcels above a defined threshold are flagged for follow-up; those in an ambiguous middle band are noted for lower-priority review.
Temporal activity signatures add a second layer. A parcel that shows vehicle presence on weekday morning collects but not on weekends, or that shows consistent SAR backscatter elevation between 07:00 and 18:00 local time, is exhibiting an operational pattern inconsistent with residential use. Planet's near-daily archive makes this kind of temporal profiling feasible over a full year, capturing seasonal variation that a single collect would miss.
What the method cannot see, and why that matters for enforcement
The most important limit is one of geometry. Satellite sensors observe surfaces, not interiors. A residential building converted to a warehouse, a ground-floor flat operating as an unlicensed food-production facility, or a basement used for storage will produce no detectable external signature if the operator is careful about what appears on the roof and in the yard. Indoor land-use violations are invisible to this method.
Cloud cover is a persistent constraint in tropical and temperate climates. A single optical collect over a cloud-affected city may take weeks to acquire. SAR mitigates this for activity detection but cannot substitute for the high-resolution optical imagery needed for object-level identification. Revisit frequency also matters: a site that operates illegally only on certain days, or that clears its yard before a predictable satellite pass, may evade detection in a sparse time series. Operators with awareness of satellite schedules can, in principle, time their activities accordingly, though this is easier said than done when commercial constellations offer sub-daily revisit.
The output of a satellite-based audit is a ranked list of candidate parcels for inspection, not a legal finding. Ground truth is mandatory before any enforcement action. The satellite analysis reduces the inspection burden by concentrating field officers on high-probability sites rather than requiring blanket surveys, but it does not replace the inspector.
Integrating with cadastral and zoning data
The analysis is only as useful as the reference data it is checked against. Gazetted zoning maps must be in a georeferenced vector format, with parcel boundaries that are accurate enough to assign each classified object to a specific parcel. Where cadastral boundaries are imprecise or outdated, a sub-metre imagery overlay can help adjudicators determine which parcel a detected structure or vehicle belongs to. Cadastral boundary verification is a separate analytical task, covered elsewhere in this library.
Zoning categories vary considerably between jurisdictions. A category that permits 'light commercial' in one municipality may prohibit vehicle repair in another. The classifier must be configured against the specific permitted-use definitions of the jurisdiction being audited, not a generic urban typology. This configuration step is the most labour-intensive part of a new deployment and typically requires input from the planning authority's legal team.
Practical deployment: what a city authority should expect
A city-scale audit covering, say, 200 to 500 square kilometres of urban area at WorldView-3 or SkySat resolution requires coordinated tasking across multiple collects, typically acquired over two to six weeks depending on cloud probability. Processing time for OBIA classification and feature extraction at parcel scale runs to days rather than hours for a city of that size, assuming automated pipelines rather than manual digitising.
The output delivered to a planning authority is typically a GIS layer of flagged parcels with per-parcel confidence scores, a summary report ranking the highest-probability violations by category, and a set of image chips showing the detected features for each flagged site. This is the format that Satellize structures its analytics deliverables around, drawing on the same pipeline approach used in its Tonga crop-estimation programme, adapted for urban classification rather than agricultural spectral indices.
Repeat audits at six- or twelve-month intervals are more valuable than a single snapshot. Violations that appear between audits indicate new non-compliance; sites that resolve between audits may indicate self-correction following warning letters. The temporal comparison is itself an enforcement tool, demonstrating to landowners that the planning authority has a persistent observation capability rather than a one-off survey.
Typical figures
| Best spatial resolution (optical) | 30 cm (Pléiades Neo panchromatic); 31 cm (WorldView-3 panchromatic) |
| Best spatial resolution (SAR) | 10 m (Sentinel-1 IW mode); 1 m Spotlight modes available on commercial SAR constellations |
| Revisit (optical, tasked) | Sub-daily (SkySat over priority areas); 1 to 4.5 days (WorldView-3, latitude-dependent) |
| Revisit (SAR, free) | 6-day repeat per satellite; 3-day combined for Sentinel-1A and 1B when both operational |
| Minimum detectable object | Individual heavy goods vehicle (approx. 8 m length) at 50 cm resolution; rooftop plant units >2 m diameter at 30 cm |
| Spectral coverage | VNIR + 8-band SWIR (WorldView-3); 4-band VNIR (Pléiades Neo, SkySat); C-band radar (Sentinel-1, 5.4 GHz) |
| Cloud limitation | Optical sensors fully blocked by cloud; SAR unaffected. Tropical cities may require 2 to 6 weeks to acquire cloud-free optical cover |
| Archive depth | Sentinel-1 from 2014; WorldView-3 from 2014; SkySat from 2016; PlanetScope near-daily archive from 2016 |
| Indoor violations detectable | No. Method is limited to externally observable surface signatures |
| Typical deliverable format | GeoPackage or Shapefile of flagged parcels with confidence scores; image chips; PDF summary report |
Analytics Satellize can run
| Per-parcel compliance probability score | Object-based image analysis with supervised classification (Random Forest or SVM) on VNIR/SWIR feature stack | GIS vector layer with score attribute, filterable by zone category and confidence band |
| Industrial activity temporal profile | Multi-temporal SAR backscatter stack analysis (Sentinel-1 IW, 20+ acquisitions) and PlanetScope NDVI/albedo variance time series | Per-parcel activity chart and flag for sites showing persistent non-residential backscatter signature |
| Vehicle presence count and classification | Object detection on sub-metre optical imagery using template matching or CNN-based detector trained on labelled vehicle objects | Point layer of detected vehicles with length and width attributes; aggregated count per parcel |
| Rooftop material classification | SWIR band-ratio indices applied to WorldView-3 8-band SWIR product, discriminating metal, asphalt, concrete and membrane roofing | Raster classification layer and per-parcel dominant material attribute |
| Change detection between audit epochs | Bi-temporal or multi-temporal image differencing and OBIA re-classification at 6- or 12-month intervals | Change polygon layer highlighting parcels with new non-compliant features since previous audit; trend report |
| Priority inspection shortlist | Multi-criteria scoring combining optical classification confidence, SAR activity flag and parcel-zoning mismatch severity | Ranked CSV and PDF report of top-N parcels recommended for ground inspection, with image evidence chips |
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.