Airport construction activity and airside expansion analysis
Very-high-resolution optical satellites can track runway extensions, apron expansions and terminal builds at civil and military airports with sub-metre clarity. This page covers sensor selection, resolution thresholds, object detection methods and the legal context for imaging sensitive airside facilities.
Sensors
- Maxar WorldView Legion: Panchromatic resolution of 29 cm and multispectral at approximately 1.2 m. A six-satellite constellation designed for revisit rates of 15 or more passes per day over mid-latitudes, making it the primary tool for detecting day-to-day changes in equipment position and earthwork extent.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, four-satellite constellation. Stereo and tri-stereo collection modes allow surface-model generation accurate to roughly 50 cm vertically, useful for measuring embankment volumes and grading progress on runway extensions.
- Planet SkySat: 50 cm panchromatic, 1 m multispectral. Fifteen satellites provide frequent revisit at lower per-image cost than the 30 cm class. Adequate for tracking large construction plant and gross earthwork boundaries; insufficient for distinguishing aircraft type or reading equipment markings.
- SPOT 7: 1.5 m panchromatic, 6 m multispectral. Useful for wide-area context and change detection across an entire airport campus, but below the resolution threshold for individual equipment classification. Best used as a screening layer before tasking finer sensors.
What a 30 cm pixel actually resolves on an airside
Resolution determines what questions you can answer. At 30 cm ground sample distance, a standard wide-body aircraft such as the Boeing 777 spans roughly 70 pixels in length and 20 in wingspan. That is enough to distinguish a narrow-body from a wide-body, identify a parked freighter by its nose-door geometry, and separate a mobile aircraft staircase from a jet bridge. Construction plant is similarly legible: a CAT 390 excavator is approximately 15 m long, covering 50 pixels at 30 cm, which is sufficient to classify it as a large excavator rather than a bulldozer or compactor.
At 50 cm, individual equipment classification becomes uncertain for smaller plant. At 1.5 m, you are counting machines, not typing them. Buyers should be clear about which question they are actually asking. Counting the number of active machines on a site is a reasonable activity proxy at 50 cm. Distinguishing a roller from a grader to infer which phase of runway sub-base work is underway requires 30 cm or better.
Earthworks, volumes and the geometry of a runway extension
A runway extension is not a single event. It proceeds through land clearance, drainage installation, sub-base compaction, base course laying, surface course paving and finally line marking, each phase leaving a distinct spectral and textural signature. Bare compacted gravel reflects differently from freshly laid asphalt, and asphalt reflects differently from painted threshold markings. Multispectral imagery in the near-infrared band is particularly useful for separating wet and dry earthworks from surrounding vegetation.
Stereo collection from Pléiades Neo or WorldView Legion allows construction of a digital surface model accurate to roughly 50 cm vertically in favourable conditions. Comparing sequential surface models gives volumetric estimates of cut-and-fill earthworks, which in turn allows inference of construction progress against a known design specification. This is not a substitute for ground survey, but it provides an independent check that is available within hours of tasking rather than weeks.
Equipment density as a proxy for construction tempo
The simplest and most defensible activity metric is equipment count per unit area per unit time. A site with twelve active machines on Monday and three on Friday is slowing down; the interpretation requires no specialist knowledge. Object detection models trained on labelled VHR imagery can automate this count across large airport campuses and multiple collection dates.
Published methods in the remote sensing literature use convolutional neural network architectures, particularly variants of YOLO and Faster R-CNN, applied to VHR optical imagery to detect and classify construction vehicles. Detection rates above 85 percent for large plant have been reported in peer-reviewed work at 30 cm resolution under good illumination. Performance degrades with shadow, partial occlusion by scaffolding or sheeting, and low solar elevation angles in winter at high latitudes. These are honest limits, not edge cases: a site heavily covered in safety netting or tarpaulins will defeat automated detection regardless of sensor resolution.
Legal and regulatory context for imaging airport facilities
Civil airports are generally imageable without restriction under the open-skies principles that govern commercial satellite operators. Most jurisdictions do not classify civil airport construction as protected infrastructure for the purposes of satellite imagery, and imagery of major international airports is freely available through open platforms. That said, several countries maintain national security restrictions on the release of sub-metre imagery of specific facilities, including military air bases and dual-use airports. Buyers should verify applicable national law and the terms of their imagery licence before distributing derived products.
Military airfields present a different picture. Imaging foreign military aviation facilities is legal under international law for satellites operating in compliance with their national licensing regimes, but some countries impose shutter-control orders on commercial operators, requiring that imagery of designated areas be withheld or degraded. The United States has historically applied such controls through the National Geospatial-Intelligence Agency, though the scope has narrowed over time as commercial resolution has become widely available. The practical consequence for a buyer is that archive availability over sensitive military sites may be patchy, and tasking approvals are not guaranteed.
Revisit strategy and the problem of cloud
Airports in tropical and maritime climates spend a significant fraction of the year under cloud. A single tasking request is not a strategy. Effective monitoring requires a standing tasking agreement that attempts collection on every available pass and retains the best clear image within each reporting window. WorldView Legion's design target of 15 revisits per day over a given point means that even a 30 percent cloud-free probability yields multiple usable images per week. Pléiades Neo's four-satellite architecture offers similar frequency.
When cloud persists for weeks, optical monitoring fails. The correct response is not to wait but to switch to a complementary sensor. Synthetic aperture radar, covered in sibling pages on InSAR and SAR deformation monitoring, penetrates cloud and operates at night, though it cannot replicate the equipment-classification capability of VHR optical imagery. A monitoring programme that combines optical tasking with SAR coherence change detection is more resilient than one that relies on either alone. Satellize structures client tasking to include SAR fallback by default for sites in high-cloud-risk regions.
What the analytics actually produce
The output of a well-designed airport monitoring programme is not a folder of satellite images. It is a structured change record: which areas of the airside changed between collection dates, by how much, and at what rate. Useful deliverables include georeferenced change masks at the resolution of the source imagery, equipment count time series per designated zone, surface model differencing results expressed in cubic metres, and alert flags when activity in a defined area exceeds or falls below a threshold.
For project-finance or regulatory purposes, the chain of custody from raw image to derived metric matters. Imagery metadata, processing logs and method documentation need to be retained alongside the analytic outputs. This is not a technical afterthought. A lender or regulator who wants to verify a construction milestone needs to be able to trace the claim back to a specific image collected at a specific time, processed by a documented method.
Typical figures
| Best available panchromatic resolution | 29 cm (WorldView Legion), 30 cm (Pléiades Neo) |
| Multispectral resolution | 1.2 m (WorldView Legion and Pléiades Neo at 30 cm pan class) |
| Revisit frequency (30 cm class) | Up to 15 passes/day over a point (WorldView Legion design target); 4–6 passes/day (Pléiades Neo) |
| Stereo vertical accuracy | Approximately 50 cm CE90 under favourable conditions (Pléiades Neo tri-stereo) |
| Minimum detectable construction plant | Large plant (>10 m) reliably detected at 50 cm; classification of plant type requires 30 cm |
| Minimum detectable aircraft type distinction | Wide-body vs narrow-body distinguishable at 50 cm; fuselage detail and nose geometry at 30 cm |
| Spectral bands (Pléiades Neo) | Panchromatic, Blue, Green, Red, Red Edge, Near-Infrared (6 bands) |
| Tasking latency (commercial priority) | First collection attempt typically within 24–48 hours of order; delivery within hours of collection |
| Archive depth | WorldView archive from 2007 (WorldView-1); Pléiades archive from 2012; SkySat from 2014 |
| Standard delivery formats | GeoTIFF (orthorectified), NITF, cloud-optimised GeoTIFF; metadata in XML |
Analytics Satellize can run
| Construction activity index | Automated object detection (CNN-based, e.g. YOLO or Faster R-CNN variants) applied to VHR optical imagery to count and classify construction plant per zone | Time-series chart of equipment count per defined airside zone, delivered as CSV and GIS polygon layer per collection date |
| Earthwork change mask | Pixel-level change detection comparing sequential orthorectified images; spectral differencing in NIR and SWIR bands to separate bare soil, compacted gravel and asphalt | Georeferenced change raster with area statistics per phase category, updated each collection cycle |
| Surface volume estimate | Digital surface model differencing from stereo or tri-stereo pairs (Pléiades Neo or WorldView Legion); cut-and-fill volumes computed against a baseline DSM | Volume report in cubic metres per designated earthwork polygon, with uncertainty bounds based on stereo accuracy |
| Phase classification map | Supervised spectral classification using labelled training samples for clearance, sub-base, paving and marking phases; applied to multispectral composites | GIS layer with phase boundaries and estimated percentage completion per runway or apron segment |
| Activity threshold alert | Rule-based trigger on equipment count or changed-area statistics falling outside a defined envelope; designed for project-finance covenant monitoring | Automated alert (email or API push) with supporting image chip and metric summary when threshold is breached |
| Aircraft parking and gate utilisation snapshot | Object detection for parked aircraft; classification by size class using wingspan-to-fuselage ratio at 30 cm resolution | Tabular count of aircraft by size class per stand or apron zone, with georeferenced point layer |
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.