Heritage and historic district change detection for conservation
Very-high-resolution optical satellites can detect single-storey additions, roof replacements and demolitions on registered historic buildings before enforcement windows close. But nadir geometry has real limits on narrow streetscapes, and cloud cover imposes latency that planners must plan around.
Sensors
- Maxar WorldView Legion: 30 cm panchromatic, 1.2 m multispectral. Up to 15 revisits per day over a target city once the full constellation is operational, making it the most capable option for detecting incremental changes on individual buildings.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, with a stereo and tri-stereo collection mode that produces surface models useful for detecting height additions. Revisit roughly once per day per satellite; the two-satellite pair improves coverage.
- Planet SkySat: 50 cm panchromatic, 1 m multispectral. Revisit can be tasked to sub-daily cadence. Resolution sits at the practical floor for detecting single-storey additions; facade material discrimination is marginal at this scale.
- Archival IKONOS and QuickBird: 82 cm and 61 cm panchromatic respectively. Both constellations are now retired, but their archives extend from 1999 and 2001 onwards, providing the longest commercially available baseline for historic district comparison.
What a roof gives away, and what it hides
From directly overhead, a satellite sees rooftops with exceptional clarity. A new zinc-clad dormer on a Victorian terrace, a concrete block addition to a Georgian rear elevation, or the replacement of original clay pantiles with modern concrete interlocking tiles: all of these are detectable at 30 cm resolution, provided the footprint is large enough to span several pixels. The practical minimum detectable structure at 30 cm is roughly 1 to 2 metres across, which covers most single-storey rear additions and all full-storey extensions.
The problem is geometry. Nadir imagery looks straight down. Facades facing the street, the very surfaces most regulated by conservation orders, are seen only as thin slivers or not at all, depending on building height and street width. In a medieval city centre where streets may be 4 to 6 metres wide and buildings 10 to 15 metres tall, a satellite at nadir sees almost no facade. Stereo collection from Pléiades Neo partially compensates by allowing off-nadir views, but the angular range is limited and shadow occlusion remains a persistent constraint. Satellite imagery is a powerful first filter for heritage enforcement; it is not a substitute for a ground inspection on facade-critical cases.
Building a long baseline: IKONOS to WorldView Legion
The commercial very-high-resolution archive stretches back further than most planners realise. IKONOS launched in September 1999; QuickBird followed in 2001. Both satellites collected imagery over most major historic cities within a few years of launch, and that archive is still accessible through Maxar's historical catalogue. A conservation authority seeking to establish the pre-development condition of a disputed site can often find imagery predating any planning application by fifteen to twenty years.
Comparing imagery across that span requires careful co-registration. Sensor geometry, collection angle and atmospheric correction all vary across decades. The standard approach is to orthorectify all images to a common digital surface model, apply a consistent pansharpening method (typically Gram-Schmidt or Brovey), and then run pixel-level or object-based change detection. Change detection algorithms such as image differencing, principal component analysis of multi-date stacks, or deep-learning change detection trained on urban datasets all have published track records in peer-reviewed remote sensing literature. The honest caveat: a 20-year archive comparison will flag many legitimate changes alongside violations. Screening false positives requires either a human analyst or a well-trained classifier with local training data.
Resolution floors and what falls below them
At 30 cm, WorldView Legion and Pléiades Neo can resolve objects roughly the size of a roof tile cluster, a chimney stack, or a small skylight. They cannot reliably distinguish original lime mortar repointing from cement repointing, identify the species of timber used in a replacement window frame, or detect the removal of original ironmongery. These are precisely the alterations that heritage officers often care most about.
Colour information helps, but only modestly. The multispectral bands on these sensors cover visible and near-infrared wavelengths. A new slate roof will show a spectrally distinct signature from original clay tiles in many cases, and fresh concrete reads differently from weathered stone. However, spectral discrimination of surface materials at this scale is probabilistic, not definitive. A 1.2 m multispectral pixel averages reflectance across an area larger than most individual building materials. Hyperspectral sensors with the spectral resolution to distinguish lime render from Portland cement do not yet operate at sub-metre spatial resolution from orbit. That gap is real, and any system claiming to enforce material-level compliance from space alone is overstating what physics currently allows.
Revisit cadence and the enforcement window
Heritage enforcement has a time dimension that distinguishes it from most land-use monitoring. An unauthorised structure built over a weekend may trigger a time-limited enforcement notice; in England, for example, the four-year rule for operational development means that an unlawful structure completed and concealed for four years may become immune from enforcement action. Satellite revisit frequency therefore matters legally, not just analytically.
WorldView Legion's target of up to 15 revisits per day over priority cities is the highest available from a commercial optical constellation. In practice, cloud cover, tasking conflicts and orbital geometry reduce effective clear-sky revisit. Over northern European cities, cloud-free revisit in winter can fall to once per week or less. Tasking a constellation at high frequency over a specific district is possible but expensive; the practical approach for most conservation authorities is quarterly or bi-annual baseline surveys supplemented by triggered tasking when a planning application is received or a complaint is filed. That cadence catches most significant structural changes while keeping data costs proportionate.
From pixel change to enforcement-ready evidence
A change-detection flag is the beginning of an enforcement workflow, not the end. The analytic output needs to be presented in a form that a planning inspector or legal team can use. That means georeferenced change polygons with confidence scores, side-by-side image comparisons with metadata showing collection date and sensor, and a clear statement of what the imagery can and cannot confirm.
Object-based image analysis, in which the analyst works with spectrally and spatially coherent segments rather than individual pixels, reduces noise and produces cleaner change polygons. Combining optical change detection with a surface model difference (generated from stereo pairs collected at two dates) adds a height-change layer that can quantify the approximate volume of an addition. Satellize structures this kind of multi-date, multi-product workflow for clients who need outputs that hold up under scrutiny, drawing on the same analytical discipline applied in the Tonga crop-estimation programme, where the evidentiary standard for government reporting is equally unforgiving. The deliverable is not a heatmap; it is a documented, auditable change record.
Practical limits to state clearly before procurement
Cloud cover is the dominant operational risk for optical heritage monitoring. It cannot be mitigated by sensor choice alone; synthetic aperture radar can penetrate cloud, but SAR at Sentinel-1's 5 to 20 m resolution is far too coarse for individual building change detection, and commercial SAR at 25 to 50 cm (Capella Space, ICEYE) is still maturing in terms of archive depth and change-detection workflows for urban fabric.
Shadow is the second constraint. In high-latitude cities in winter, low solar elevation angles cast long shadows across narrow streets, obscuring the very rooftop and rear-elevation details that matter most. Collection scheduling to maximise solar elevation is standard practice but not always possible within tasking windows. Finally, pansharpening, the process of merging 30 cm panchromatic detail with 1.2 m colour information, introduces spectral artefacts that can confuse material-change detection. Analysts working on heritage applications should validate pansharpened outputs against the native multispectral bands before drawing spectral conclusions.
Typical figures
| Best available panchromatic resolution | 30 cm (WorldView Legion, Pléiades Neo) |
| Best available multispectral resolution | 1.2 m (WorldView Legion, Pléiades Neo); 1 m (SkySat) |
| Maximum tasked revisit (optical, clear sky) | Up to 15 times per day (WorldView Legion, full constellation); ~1 per day per satellite (Pléiades Neo) |
| Practical cloud-free revisit (northern Europe, winter) | As low as once per week; planning around seasonal cloud is essential |
| Archive depth | From 1999 (IKONOS); from 2001 (QuickBird); Pléiades from 2012; SkySat from ~2014 |
| Minimum detectable structure (nadir, 30 cm sensor) | Approximately 1 to 2 m across; single-storey rear additions generally detectable |
| Spectral bands (Pléiades Neo example) | Panchromatic, Blue, Green, Red, Red Edge, Near-Infrared |
| Stereo surface model vertical accuracy | Typically 0.5 to 1 m CE90 from Pléiades Neo stereo pairs over urban areas |
| Delivery formats | GeoTIFF (orthorectified), NITF, change polygon shapefiles or GeoPackage, PDF annotated reports |
| Facade detection from nadir | Severely limited on streets narrower than building height; off-nadir stereo partially compensates |
Analytics Satellize can run
| Rooftop change polygon layer | Object-based image analysis (OBIA) with multi-date image differencing; co-registered orthorectified inputs | GeoPackage of flagged change polygons with confidence score, date pair and change type classification |
| Height-addition detection | Digital surface model differencing from stereo Pléiades Neo collections at two dates | Raster difference layer and vector polygons showing statistically significant height increases, with approximate volume estimate |
| Long-baseline archive comparison report | Multi-date pansharpened image stack from IKONOS/QuickBird archive to current sensor; principal component or deep-learning change detection | Illustrated PDF report with side-by-side image comparisons, georeferenced change inventory and metadata table suitable for planning evidence |
| Roof material change classification | Spectral index analysis on native multispectral bands; supervised classification using local training samples | Classified raster and summary table of roof material types by parcel, with change flags between survey dates |
| Triggered alert on new construction activity | Automated change detection on tasked imagery following planning application receipt; threshold-based flagging | Email or API alert with image chip, coordinates and preliminary change assessment within 48 hours of clear-sky collection |
| Conservation area condition baseline | Full-area OBIA mapping of building footprints, roof types and visible alterations from a single high-resolution collect | GIS layer set with attributed building polygons, usable as a legal baseline record for future comparison |
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.