Airport aircraft parking count as aviation-demand indicator
Very-high-resolution optical imagery lets analysts count and classify parked commercial aircraft at major airports daily, producing a ground-truth proxy for capacity utilisation that balance sheets and load-factor disclosures cannot fake.
Sensors
- Planet SkySat: 50 cm native resolution, up to 12 revisits per day at mid-latitudes when tasked. Panchromatic and pan-sharpened RGB. Sufficient to resolve narrowbody fuselage widths (3-4 m) and widebody fuselages (5-7 m) with confidence.
- Maxar WorldView-2 / WorldView-3: WorldView-3 delivers 31 cm panchromatic and 1.24 m 8-band multispectral. Revisit at a given hub is roughly 1-4.5 days depending on latitude and tasking priority. The multispectral bands support airline livery colour discrimination as a secondary classification aid.
- Airbus Pleiades Neo: 30 cm panchromatic, 1.2 m multispectral, four-satellite constellation offering daily revisit at most major airports when tasked. Stereo and tri-stereo collection modes allow rudimentary height estimation, useful for distinguishing parked aircraft from large ground equipment.
- Airbus OneAtlas archive: Historical SPOT and Pleiades archive extending back to the 1980s and 2011 respectively. Enables pre-crisis baseline counts and long-run capacity-utilisation trend analysis at hub airports.
- Sentinel-2 (ESA / Copernicus): 10 m multispectral, 5-day revisit globally, free. Insufficient to count individual aircraft but useful for detecting gross parking-area occupancy changes at very large airports and for cross-checking tasked imagery acquisition windows.
What a parked aircraft actually reveals
When an airline parks a widebody on a remote stand for more than a day or two, it is not resting the aircraft. It is signalling that the route network cannot absorb it. During the COVID-19 groundings of 2020, satellite images of Victorville, Alice Springs and Teruel became primary evidence for the scale of the demand collapse weeks before quarterly filings arrived. The mechanism is simple: aircraft on the ground are not generating revenue, and large fleets of them cluster visibly at airports with available apron space.
The geometry is unambiguous at sub-metre resolution. A Boeing 737-800 fuselage is 3.76 m wide; an Airbus A380 is 7.14 m. At 30-50 cm ground sample distance, those dimensions are resolved across 7-24 pixels respectively. Automated object detection trained on known airframe silhouettes can classify narrowbody, widebody and regional turboprop categories with high confidence in clear conditions. Livery colour, visible at WorldView-3's 8-band multispectral resolution, provides a secondary check for airline attribution at hub airports where carriers operate dedicated terminal aprons.
How the count is built: from pixels to a capacity signal
The analytical pipeline has three stages. First, candidate aircraft regions are identified by template-matching or convolutional object-detection models trained on labelled VHR imagery. Published studies using SkySat and WorldView imagery have demonstrated detection rates above 90 percent for aircraft on open aprons in clear conditions, with false positives concentrated around ground-service vehicles and jet bridges of similar aspect ratio.
Second, each detection is classified by size category using fuselage-width estimation from the image. The pixel footprint of the fuselage perpendicular to the aircraft axis, corrected for the known off-nadir angle of the acquisition, gives an estimated physical width. Widths below roughly 4 m are classed as narrowbody or regional; above 5.5 m as widebody. The gap between those ranges is narrow enough that misclassification is rare except at high off-nadir angles (beyond about 25 degrees), where foreshortening introduces uncertainty.
Third, the count at each airport is expressed as a fraction of that airport's documented stand capacity, derived from published aerodrome charts and historical baseline imagery. That fraction is the utilisation signal: the share of available parking that is occupied by grounded aircraft. A rising fraction at a carrier's primary hub, sustained across multiple acquisition days, is a direct operational stress indicator.
Revisit arithmetic and what it means for timeliness
Daily revisit at a single airport is achievable with tasked SkySat or Pleiades Neo, but it costs. A more practical regime for most financial-intelligence applications is every-other-day tasking at a curated list of 20-40 hub airports, which covers the majority of global seat capacity. Planet's published SkySat constellation capacity supports this at major hubs in North America, Europe and East Asia without difficulty.
Latency from acquisition to delivered count is typically 4-12 hours for tasked commercial imagery, depending on downlink scheduling and processing pipeline. That is fast enough for same-day equity research or credit-monitoring alerts, though it is not real-time in the air-traffic-control sense. The honest comparison is against airline operational disclosures, which arrive quarterly, and against flight-tracking aggregators, which show scheduled versus operated flights but not the physical location of grounded assets.
The limits you need to know before relying on this
Cloud cover is the most significant operational constraint. Major hub airports in monsoon-affected regions, the Pacific Northwest and northern Europe can experience multi-day cloud-obscured periods that break the count series. Optical imagery simply does not penetrate cloud. Synthetic aperture radar (SAR) from Sentinel-1 or ICEYE can image through cloud, but at 5-20 m resolution it cannot reliably distinguish a parked narrowbody from a fuel bowser or a taxiing aircraft from a stationary one. SAR fills temporal gaps but cannot substitute for optical classification.
Aircraft in hangars are invisible to any spaceborne sensor. At airports with extensive maintenance infrastructure, a significant fraction of the fleet may be undergoing scheduled C-checks or heavy maintenance and will not appear in the count. This introduces a systematic undercount that varies by airport and season. Analysts should cross-reference with published maintenance-facility capacity at each hub.
Night acquisitions by optical sensors are not operationally useful for aircraft counting. Airports are lit, but the contrast and shadow geometry that enable classification in daylight are absent. This means counts are effectively limited to daylight passes, which at high latitudes in winter can be as few as one per day even with aggressive tasking.
Finally, the count measures parking occupancy, not the reason for parking. A carrier deliberately parking aircraft for a scheduled fleet rotation looks identical, in the imagery, to one parking aircraft because demand has collapsed. Interpretation requires context: flight-schedule data, published capacity guidance and, where available, ADS-B flight-tracking records to distinguish aircraft that have not moved in 48 hours from those cycling through normal turnarounds.
Where the signal fits in a credit or equity workflow
For aviation-sector credit analysts, the parking count serves as an out-of-cycle operational check between quarterly filings. A sustained increase in parked widebodies at a carrier's hub, appearing in imagery before any public disclosure, is a leading indicator of revenue-per-available-seat-kilometre pressure. It does not replace financial modelling, but it grounds the model in physical reality rather than management guidance.
For equity research, the most useful application is comparative: tracking the ratio of parked aircraft across competing carriers at the same hub over the same period. If one airline is consistently parking more of its fleet than a peer operating similar routes, the imagery provides a quantitative basis for the divergence, independent of either carrier's public statements.
Satellize runs structured airport-count programmes on tasked VHR imagery, delivering periodic count reports and time-series feeds to financial clients. The methodology is the same family of object-detection and geometric-classification approaches used in published remote-sensing research on airport monitoring. Clients specify the airport list and the reporting cadence; the output is a structured dataset, not a narrative opinion.
Typical figures
| Spatial resolution (optical) | 30-50 cm panchromatic (WorldView-3, Pleiades Neo, SkySat); 1.2-1.24 m multispectral |
| Revisit at a tasked hub | Daily to sub-daily (SkySat, Pleiades Neo); 1-4.5 days (WorldView-2/3) depending on latitude and tasking priority |
| Delivery latency | 4-12 hours from acquisition to processed count, depending on downlink and pipeline |
| Minimum detectable aircraft | Regional turboprop (~3 m fuselage width) at 50 cm GSD; narrowbody and widebody with high confidence at 30-50 cm |
| Classification categories | Narrowbody (3-4 m fuselage), widebody (5-7 m fuselage), regional turboprop; airline attribution where livery is visible |
| Cloud penetration | None (optical); SAR provides cloud-immune imagery but cannot classify aircraft type reliably |
| Archive depth | Pleiades from 2011, WorldView-2 from 2009, SPOT archive from 1986 (via OneAtlas); SkySat from 2014 |
| Coverage | Any airport globally that can be tasked; open-sky aprons only; hangared aircraft not detectable |
| Delivery formats | Structured count dataset (CSV/JSON), GIS point layer with aircraft locations, time-series chart, periodic PDF report |
Analytics Satellize can run
| Daily parked-aircraft count per airport | Convolutional object detection on VHR panchromatic imagery, trained on labelled airframe silhouettes | Structured daily feed (JSON/CSV) with count by aircraft size category and stand location |
| Fleet utilisation ratio | Parked count divided by documented stand capacity from aerodrome charts and baseline imagery | Time-series chart and alert when ratio exceeds analyst-defined threshold |
| Widebody versus narrowbody split | Fuselage-width estimation from pixel footprint corrected for off-nadir angle | Count table by aircraft category, updated each acquisition; GIS point layer with classification attribute |
| Carrier attribution at hub airports | Livery colour classification using WorldView-3 8-band multispectral; stand-assignment cross-reference with published terminal maps | Carrier-level count table where attribution confidence exceeds defined threshold |
| Multi-airport comparative dashboard | Parallel count series across a client-specified list of hubs, normalised to stand capacity | Periodic comparative report and live dashboard feed for up to 40 airports |
| Historical baseline and anomaly detection | Archive imagery processed through the same detection pipeline; statistical deviation from rolling 90-day baseline | Alert report flagging airports where current parking count deviates significantly from seasonal norm |
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.