Shadow-length building height change detection for planning compliance
Cast shadows in very-high-resolution optical imagery encode building height with surprising precision. Comparing shadow lengths across survey epochs reveals storey-count changes without stereo imagery or LiDAR, flagging unpermitted vertical additions before enforcement teams arrive.
Sensors
- Maxar WorldView-3: 0.31 m panchromatic resolution; the finest commercially available optical imagery at scale. Shadow edges are resolved to sub-metre accuracy, making single-storey additions (roughly 3 m) detectable in favourable sun-angle conditions. Revisit approximately 1 day at mid-latitudes with off-nadir tasking.
- Airbus Pléiades Neo: 0.30 m panchromatic resolution, comparable to WorldView-3. The constellation of two satellites offers revisit of 1 to 2 days. Precise acquisition metadata including sun elevation and azimuth are supplied in the product header, which is essential for shadow-to-height conversion.
- Planet SkySat: 0.50 m resolution, slightly coarser than the Maxar and Airbus flagships. Adequate for detecting additions of two storeys or more in good sun-angle geometry. The fleet of 21 satellites enables frequent revisit for change-detection cadences of days to weeks.
- Airbus Pléiades (original): 0.50 m panchromatic resolution. Archive depth extends to 2012, which is valuable for establishing baseline height epochs against which recent imagery is compared. Metadata quality is consistent and well-documented.
What a shadow actually encodes
The geometry is straightforward trigonometry. When the sun is at a known elevation angle above the horizon, a vertical structure of height H casts a shadow of length L on flat ground such that H = L × tan(θ), where θ is the solar elevation angle. Every commercial very-high-resolution satellite records the solar elevation and azimuth at acquisition time in its metadata header. Those numbers are not estimates; they are computed from the satellite's precise orbital position and acquisition timestamp to sub-degree accuracy.
In practice, the analyst measures the shadow length in pixels, converts to metres using the ground sampling distance, applies the formula, and arrives at a height estimate. On flat terrain with a solar elevation between 30° and 60°, WorldView-3 or Pléiades Neo imagery at 0.3 m resolution can resolve height differences of roughly 1 to 2 metres, which corresponds to less than one standard storey. That is the method's ceiling. It does not require a stereo pair, a LiDAR pass, or any ground-control survey.
Where the method earns its keep in planning enforcement
Planning authorities in most jurisdictions receive no automatic notification when a building grows by one or two storeys. A homeowner adds a mansard conversion; a commercial landlord raises a parapet and installs plant rooms that push the effective ridge height past the consented limit. These additions rarely trigger a new planning application. Enforcement teams are small and reactive, responding to complaints rather than conducting systematic surveys.
Satellite change detection inverts that dynamic. By comparing shadow-derived height estimates from two acquisition epochs, an analyst can flag every structure in a target area that has grown beyond a threshold, say 2.5 metres, between a baseline image and a recent one. The output is a ranked list of candidate violations, not a confirmed enforcement case. Ground verification remains necessary. But the satellite pass converts an impossible manual audit into a tractable triage problem.
The honest limits: sun angle, latitude, and the resolution floor
At solar elevations below about 20°, shadow lengths become very long and the measurement uncertainty grows proportionally. A 0.3 m pixel error in shadow length translates to a larger height error when the sun is low. This is a significant constraint at latitudes above 55° in winter months, where solar elevation at midday can fall to 15° or less. Imagery acquired in those conditions is often unsuitable for quantitative shadow analysis, not because the sensor fails, but because the geometry amplifies small measurement errors into large height uncertainties.
The resolution floor is equally honest. A single-storey addition of approximately 3 metres produces a shadow extension of roughly 5 metres at a 30° solar elevation. At 0.3 m resolution that is about 17 pixels, which is detectable. At 0.5 m resolution it is 10 pixels, still workable. But if the addition is set back from the building edge, or if an adjacent structure occludes part of the shadow, the detectable increment rises. In dense urban canyons with closely packed buildings, shadow overlap can make individual structure attribution ambiguous. The method works best on semi-detached or detached structures with clear shadow fall zones.
Oblique sun azimuths introduce a further complication. If the sun is not perpendicular to the building façade, the shadow falls at an angle across adjacent rooftops or terrain that is not flat. Slope correction requires a digital terrain model. Without one, height estimates on sloping ground carry errors proportional to the terrain gradient.
Epoch comparison: choosing the right baseline
The method's value depends entirely on the quality of the baseline epoch. An archive image from Pléiades or WorldView taken at a similar solar elevation angle to the recent image minimises the geometric correction burden. Ideally, both images share solar elevations within five degrees of each other, so that systematic biases cancel in the differencing step.
Archive depth matters. Pléiades imagery goes back to 2012, WorldView-2 to 2009. For a planning authority investigating whether a structure predates a given consent date, that archive can be decisive. The analyst selects the earliest clean image that post-dates the original planning consent, establishes the consented height from shadow measurement, then compares against the most recent acquisition. The delta is the candidate violation magnitude.
Turning candidate flags into a usable enforcement product
Raw shadow measurements are not an enforcement product. The deliverable that matters to a planning officer is a georeferenced layer of flagged parcels, each annotated with the measured height in both epochs, the derived change in metres, the solar elevation at each acquisition, and a confidence band reflecting the geometric uncertainty. A flag with a 95% confidence interval of ±1.5 m straddling the permitted height limit is a low-priority candidate. One showing a 4 m addition with a ±0.8 m uncertainty band is worth a site visit.
Satellize structures this kind of tiered output for planning clients, ranking flags by confidence and by the magnitude of the apparent violation. The analytics pipeline is the same class of photogrammetric shadow analysis used in published academic remote sensing literature, applied systematically across a defined area of interest. The Tonga crop-estimation programme uses a different analytic method, but the underlying principle of extracting quantitative physical measurements from publicly documented sensor geometry is consistent across the work.
For authorities covering large boroughs or districts, the practical cadence is one survey pass per quarter, with a rapid-response tasking option when a complaint triggers a specific parcel check. Quarterly cadence is sufficient to catch most unpermitted construction before it is completed and rendered harder to enforce against.
What this method cannot replace
Shadow-length analysis is a screening tool, not a legal instrument. Height measurements derived from it carry geometric uncertainty that no amount of processing eliminates entirely. A confirmed enforcement notice requires a physical survey or a LiDAR-derived point cloud with documented accuracy. The satellite method's job is to direct that expensive ground resource to the right addresses.
It also cannot detect horizontal footprint expansion, basement excavation, or internal storey additions that do not change the external roofline. Those require complementary methods. Used for what it is actually good at, detecting vertical growth of roughly one storey or more on structures with clear shadow fall zones, it is a genuinely useful and under-exploited tool in the planning compliance toolkit.
Typical figures
| Best available spatial resolution | 0.30 m panchromatic (WorldView-3, Pléiades Neo) |
| Minimum detectable height change (favourable geometry) | Approximately 1.5 to 3 m (0.5 to 1 storey) at 0.3 m resolution, 30°–60° solar elevation |
| Solar elevation range for reliable analysis | 20° to 70°; below 20° uncertainty grows sharply; above 70° shadows become very short |
| Revisit cadence (commercial tasking) | 1 to 2 days (WorldView-3, Pléiades Neo); 2 to 7 days (SkySat fleet) |
| Archive depth | Pléiades from 2012; WorldView-2 from 2009; WorldView-3 from 2014 |
| Spectral bands used | Panchromatic only for shadow measurement; multispectral used for land-cover context |
| Latency from tasking to delivery | Typically 24 to 72 hours for priority tasking; archive orders same day |
| Terrain correction requirement | Digital terrain model required for slopes exceeding approximately 5° |
| Delivery formats | GeoTIFF orthoimage, GeoPackage or Shapefile of flagged parcels, PDF summary report |
| Coverage per acquisition | WorldView-3 strip up to 13.1 km wide; Pléiades Neo strip up to 14 km wide |
Analytics Satellize can run
| Baseline height map | Shadow-length photogrammetry using recorded solar elevation angle from image metadata; trigonometric inversion H = L × tan(θ) | GeoPackage layer of per-structure height estimates with uncertainty bands, keyed to parcel IDs |
| Change-epoch height delta map | Pixel-level shadow measurement in recent image minus baseline estimate; difference thresholded at client-specified minimum change (e.g. 2.5 m) | Ranked list of flagged parcels with height-change magnitude, confidence interval, and acquisition metadata for both epochs |
| Planning-consent exceedance flags | Derived height compared against consented height from planning register; flags raised where measured height exceeds consent plus uncertainty margin | Enforcement-triage GIS layer with exceedance magnitude and confidence classification (high / medium / low) |
| Oblique-geometry quality mask | Per-image solar elevation and azimuth assessment; pixels where shadow falls onto adjacent rooftops or sloped terrain flagged as unreliable | Confidence mask delivered alongside height layer; low-confidence zones excluded from enforcement flags |
| Quarterly change-monitoring report | Repeat shadow analysis on a defined area of interest at agreed cadence; automated comparison against rolling baseline | Quarterly PDF report with map of new flags, dismissed flags, and persistent candidates awaiting ground verification |
| Rapid single-parcel height check | On-demand tasking of a specific address following a complaint; shadow measurement against archive baseline | Single-parcel report within 72 hours of tasking, including annotated image crop and height estimate with stated uncertainty |
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.