Construction crane census as a development pipeline indicator
Tower-crane counts derived from very-high-resolution satellite imagery give property economists a contemporaneous supply-pipeline signal that planning registers alone cannot provide, identifying active development volume and flagging unregistered construction activity across entire cities in a single pass.
Sensors
- Planet SkySat: 50 cm native resolution (product resampled to 50 cm pansharpened), sufficient to resolve crane masts, jibs and counterweights. Tasked on demand; revisit within hours over priority cities when constellation geometry permits.
- Maxar WorldView-3: 31 cm panchromatic, 1.24 m multispectral. The sharpest commercially available optical imagery in routine production. At this resolution, luffing-jib and flat-top crane types are distinguishable by boom profile. Revisit roughly 1–4.5 days depending on latitude and off-nadir tolerance.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, with a daily revisit capacity per site when both Neo 3 and Neo 4 are tasked together. Stereo and tri-stereo collection modes allow height estimation of crane tip above ground, which cross-checks building-core height at time of capture.
- Airbus Pléiades (legacy): 50 cm panchromatic. Slightly coarser than Neo but with a deep archive back to 2011, useful for constructing historical crane-count time series in cities where retrospective pipeline analysis is needed.
Why a crane is the most honest data point in property economics
A tower crane on a construction site is not a planning intention or a permit application. It is a capital commitment. The crane has been hired, erected and insured; the concrete core is rising. No other publicly observable signal sits so close to the boundary between pipeline and delivery.
Planning registers record approvals, not activity. A permission granted three years ago may still show as 'under construction' in a local authority database while the site has been mothballed. Conversely, some jurisdictions permit commencement before formal registration is complete, creating a gap between what the register shows and what is actually happening. A crane count from satellite imagery closes that gap. It measures the physical present, not the administrative record.
What a crane's shadow gives away
At 31 cm resolution, a WorldView-3 image resolves the mast, the jib, the counterweight arm and, in oblique or off-nadir collection, the hook block. Shadow geometry is equally informative. A crane casting a shadow across an adjacent rooftop allows straightforward trigonometric estimation of tip height given the solar elevation angle at acquisition time, a method published in multiple peer-reviewed remote sensing studies using Pléiades and WorldView imagery. That height figure, compared with the approved building height in the planning register, shows whether a structure is tracking to its permitted envelope.
Boom orientation matters too. A flat-top crane with its jib pointing consistently in one direction across sequential acquisitions may indicate a constrained urban site where the operator is managing airspace conflicts with neighbouring buildings. A luffing-jib crane signals a tight footprint. These distinctions are legible at sub-50 cm resolution and carry real information about site geometry, which in turn informs assumptions about construction programme duration and likely unit yield.
Building a city-wide census: method and honest limits
A crane census over a major city requires a cloud-free mosaic of sufficient resolution across the entire urban extent. For a city the size of London or Sydney, a single SkySat or WorldView-3 collect rarely covers the full area in one pass; a mosaic assembled from multiple strips collected within a short window (ideally the same day, at most within a week) is the practical approach. Temporal mismatch within the mosaic introduces a small counting error if cranes are erected or struck between strip dates.
Cloud cover is the principal operational constraint. Optical sensors cannot see through overcast skies, which is a genuine limitation in mid-latitude cities during winter months. A census attempted in December over northern European cities may require waiting days or weeks for a clear acquisition window. Radar sensors (Sentinel-1 SAR, for instance) can penetrate cloud but at 5 m resolution cannot reliably resolve individual crane structures; they are useful for detecting large-scale site activity but not for crane-level enumeration.
Detection is performed by a combination of object-based image analysis and human verification. Automated convolutional neural network classifiers trained on labelled VHR imagery can flag crane candidates with high recall, but false positives from tall masts, antenna towers and construction hoists require analyst review. The honest minimum detectable target is a mast of roughly 1.5 m cross-section at 50 cm resolution, which covers all standard tower cranes but may miss very small self-erecting units on low-rise sites.
Cross-referencing the planning register: where the signal gets interesting
The crane count becomes most valuable when it is joined to the planning register. Each crane location is geocoded and spatially matched against approved development applications. Three outcomes are possible: the site matches a live permission (expected), the site matches an expired or superseded permission (worth investigating), or the site has no matching permission within a reasonable search radius (potentially unregistered).
Unregistered activity is not always illegal. Permitted development rights, phased approvals and boundary ambiguities all produce apparent mismatches that resolve on closer inspection. But a persistent mismatch, confirmed across two or three sequential satellite acquisitions, is a legitimate flag for a planning authority or a property economist trying to understand true supply volume. In markets where informal construction is common, this cross-reference can reveal a meaningful fraction of actual development that official statistics simply do not capture.
The output of this analysis is a georeferenced crane inventory with attributes: crane type where distinguishable, estimated jib radius, planning-register match status, and a time-series count if archive imagery is available. For property economists, the aggregate count by submarket is the headline figure; the site-level detail supports due diligence on specific assets.
Using crane counts as a leading indicator: what the property economics literature supports
Tower cranes appear on site roughly 12 to 24 months before practical completion on a typical high-rise residential or commercial scheme. This lag is well understood in property economics and makes crane counts a genuine leading indicator of future completions, not a coincident one. A rising crane count in a submarket today implies additional supply hitting that market in one to two years, with implications for rents, yields and land values in adjacent areas.
Several property consultancies publish periodic crane surveys based on ground observation, notably the twice-yearly surveys covering Australian and UK cities. Satellite-derived counts can replicate this methodology at higher frequency and lower cost per city, and can extend coverage to markets where no ground survey exists. The satellite approach does not replace local market knowledge, but it does give a consistent, repeatable, geographically exhaustive baseline that ground surveys struggle to match.
Satellize runs this type of analysis as part of its broader property-intelligence offering. The methodology is the same family of object detection and register cross-referencing used in the Tonga crop-estimation programme, adapted from agricultural parcel classification to urban structure detection. Clients receive a structured GIS layer rather than a PDF count, which means the data integrates directly into existing property analytics workflows.
Frequency, cost and the question of how often you actually need a census
A quarterly crane census is sufficient for most property economics applications. Cranes are not erected and struck in days; a three-month cadence captures meaningful change in active development volume without the cost of monthly tasking. In fast-moving markets during a construction boom, monthly may be justified for specific submarkets. City-wide annual counts work well for benchmarking and trend analysis but miss intra-year cycles.
The per-city cost of a census is dominated by imagery acquisition, not analysis. Archive imagery, where a cloud-free collect exists within the target window, is substantially cheaper than new tasking. Building an archive-first strategy, supplemented by fresh tasking only where the archive has gaps, is the practical approach for clients running multi-city programmes on a defined budget.
Typical figures
| Best available spatial resolution | 31 cm panchromatic (Maxar WorldView-3); 30 cm (Airbus Pléiades Neo) |
| Minimum detectable crane mast cross-section | Approximately 1.5 m at 50 cm resolution; smaller structures may be missed |
| Revisit (tasked) | Same-day to 4.5 days depending on sensor and latitude; SkySat can revisit within hours under favourable geometry |
| Cloud limitation | Optical only; overcast skies block collection entirely. SAR (Sentinel-1, 5 m) can detect site activity but cannot resolve individual cranes |
| Spectral bands used | Panchromatic for detection; multispectral (RGB + NIR) for contextual classification and shadow analysis |
| Archive depth | Pléiades legacy to 2011; WorldView constellation from 2007; SkySat from approximately 2014 |
| Mosaic window for city-wide census | Ideally same-day; practically within 7 days to limit temporal mismatch error |
| Typical census cadence | Quarterly for pipeline monitoring; monthly for high-activity submarket surveillance |
| Delivery format | Georeferenced GIS layer (GeoJSON or GeoPackage), attribute table with crane type, jib radius estimate, and planning-register match status |
| Positional accuracy | WorldView-3 CE90 approximately 3.5 m without ground control; better than 1 m with GCPs |
Analytics Satellize can run
| City-wide crane inventory | Object-based image analysis (OBIA) with CNN-assisted candidate detection and analyst verification on VHR optical mosaic | GeoJSON point layer with crane location, type classification, estimated jib radius and planning-register match flag |
| Submarket crane-count time series | Sequential census runs on archive and tasked imagery, aggregated to defined submarket polygons | Tabular time series (CSV) and chart-ready data showing crane count by submarket per period |
| Unregistered activity flag report | Spatial join of crane inventory against planning register polygons; mismatches flagged and classified by likely cause | Filtered GIS layer and summary report of sites with no matching live planning permission within 50 m |
| Crane-tip height estimate | Shadow-length trigonometry using solar elevation angle at acquisition time, applied to oblique or off-nadir VHR imagery | Attribute column in crane inventory GIS layer; cross-tabulated against approved building heights from planning register |
| Supply-pipeline leading indicator index | Aggregate crane count weighted by estimated jib radius (as proxy for site scale), normalised to submarket area or existing stock | Monthly or quarterly index value per submarket, delivered as structured data feed for integration into property economics models |
| Change alert: new crane erection or strike | Differencing between sequential census layers; new detections and disappearances flagged automatically | Alert feed (email or API) with location, date of change and link to before/after image 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.