Cargo airport freighter-to-passenger mix analysis from aircraft classification
Very-high-resolution optical satellites measure fuselage length and wingspan on airport aprons to distinguish freighters from passenger aircraft, giving a repeatable proxy for dedicated air-cargo capacity at major hubs.
Sensors
- Maxar WorldView-3: 0.31 m native panchromatic resolution, enabling fuselage width discrimination and door-configuration reading on wide-body aircraft. Revisit roughly 1 day at mid-latitudes under tasking. The primary sensor for this use case.
- Airbus Pléiades Neo: 0.30 m panchromatic resolution, 50 cm multispectral. Stereo pairs available for height estimation, which helps confirm aircraft type by tail height. Revisit under 24 hours with the four-satellite constellation.
- BlackSky Global: 0.5 m resolution, with a revisit cadence of up to 15 passes per day over selected targets. Lower resolution than WorldView-3 limits reliable wingspan measurement to wide-body aircraft only, but the high revisit suits time-series monitoring.
- Planet SkySat: 0.5 m resolution, tasked on demand. Useful for change-detection time series and confirming aircraft presence or absence at stands, though fuselage detail is marginal for narrow-body classification.
What a parked aircraft reveals from 500 km up
A Boeing 747-8F has a fuselage length of 76.3 m and a wingspan of 68.4 m. A 747-400 passenger variant is 70.6 m long with the same wingspan. An Airbus A320 passenger jet is 37.6 m long. At 0.30 to 0.31 m ground sample distance, those differences are resolved by hundreds of pixels. Aircraft classification from satellite imagery is therefore not a speculative exercise. It is a measurement problem with a known answer most of the time.
The classification pipeline works in three steps. First, aircraft are detected on the apron using object-detection models trained on labelled satellite imagery. Second, fuselage length and wingspan are measured in pixels and converted to metres using the image's ground sample distance and any available look-angle correction. Third, measurements are matched against published aircraft dimension databases to assign a type or type family. The output is a per-stand record: aircraft present, estimated type, freighter or passenger designation.
Freighters, combis and the converted middle ground
Dedicated freighters, combination aircraft (combis) and passenger jets with belly-hold capacity represent three structurally different contributions to air-cargo supply. Tracking the mix at a hub matters because belly-hold capacity is a residual of passenger scheduling. When passenger demand falls, belly capacity disappears faster than dedicated freighter capacity. The 2020 collapse of passenger flying made this visible to anyone watching cargo rates.
Dedicated freighters are the most straightforward to identify. Main-deck cargo doors, either nose-loading or side-loading, are visible at 0.30 m resolution when open and sometimes inferrable from door-surround geometry when closed. Wide-body freighters also lack cabin windows, which appear as a regular row of small features on passenger aircraft. At 0.31 m GSD, window rows are detectable on clear imagery at low look angles.
Converted passenger freighters are the hard case. A Boeing 737-800BCF has nearly identical external dimensions to a 737-800 passenger aircraft. The conversion adds a cargo door but does not change fuselage length or wingspan. Livery is often the only discriminant, and livery is not always conclusive from a single satellite image. Honest accounting of this method requires flagging converted narrow-bodies as ambiguous unless a cargo door is visible or the operator is known.
Building a capacity index from apron counts
A single image gives a snapshot. A time series gives an index. Collecting imagery of the same airport at regular intervals, say weekly or monthly, and running the classification pipeline on each image produces a count of freighter stands occupied, passenger stands occupied and unoccupied stands. Weighting each aircraft type by its published maximum payload gives an approximate instantaneous cargo-capacity figure for the apron.
That figure is not a throughput number. Aircraft on stands are not necessarily loading or unloading, and turnaround times vary. What the index captures is fleet deployment: which operators are basing freighters at which airports, and whether that mix is shifting. A sustained increase in wide-body freighter presence at a hub that historically relied on belly-hold capacity is a signal worth examining. A sudden drop in freighter stands occupied, without a corresponding rise in passenger aircraft, may indicate demand softening or fleet redeployment to another corridor.
Published research on aircraft detection and classification from very-high-resolution imagery, including work using WorldView and Pléiades data, has demonstrated type-level accuracy above 90 percent for wide-body aircraft under good imaging conditions. Narrow-body accuracy is lower, particularly for converted freighters. Any capacity index derived from this method should carry an explicit uncertainty band for the converted-aircraft fraction.
Where the method breaks down
Look angle is the first problem. Satellites do not always image from directly overhead. Off-nadir angles beyond roughly 20 to 25 degrees introduce foreshortening that compresses apparent fuselage length and can cause wingspan to be underestimated. High-latitude airports, where off-nadir tasking is more common due to orbital geometry, are most affected. A correction using the published look angle and aircraft height is possible but adds uncertainty.
Cloud cover is the second problem. Optical imagery is useless through thick cloud, and many major cargo hubs sit in regions with persistent overcast seasons. Frequent revisit constellations like BlackSky and SkySat improve the probability of a clear acquisition within a given window, but they cannot guarantee it. A monthly index built from weekly tasking attempts will have missing data points.
Aircraft on remote stands or inside maintenance hangars are invisible. Apron coverage is only as complete as the image footprint and the operator's stand-use practices. Some airports also have multiple apron areas separated by taxiways, requiring careful mosaic handling to avoid double-counting or gaps.
From airport apron to air-cargo market signal
The value of this analysis is not in any single airport image. It is in consistent, comparable data across a set of hubs over time. Cargo airports worth monitoring for global air-freight signal include Memphis, Louisville, Hong Kong, Leipzig, Dubai, Anchorage and a handful of others that handle disproportionate shares of global dedicated freighter movements. Tracking freighter-stand occupancy at those airports simultaneously, with consistent methodology, produces something closer to a market index than a series of unrelated observations.
Satellize runs this kind of multi-site, time-series optical analysis on client licence, adding commercial tasking on top of open-data baselines where archive depth is needed. The Tonga crop-estimation programme is a different domain, but the underlying pattern is the same: repeated structured measurement from satellite, converted into an economic signal.
For a freight forwarder, airline or commodity trader, the question is not whether satellites can see aircraft on aprons. They clearly can. The question is whether a consistent, auditable methodology applied to the right set of airports produces a signal that is earlier or more granular than what trade associations and airline capacity reports provide. For dedicated freighter fleets, the answer is yes. For belly-hold capacity, the answer depends on how much of the passenger schedule is already visible from other sources.
Typical figures
| Best available spatial resolution | 0.30 m GSD (Pléiades Neo, WorldView-3 panchromatic) |
| Minimum reliably classifiable aircraft | Wide-body jets (fuselage width above ~5 m) at 0.30 m GSD; narrow-bodies classifiable by length but type ambiguity increases |
| Revisit under tasking | Sub-daily at mid-latitudes with BlackSky or Pléiades Neo; 1 day typical for WorldView-3 |
| Look-angle constraint | Off-nadir angles beyond 20 to 25 degrees degrade fuselage-length measurement; high-latitude airports most affected |
| Cloud limitation | Optical only; no data through thick cloud. Persistent overcast regions require multi-attempt tasking strategies |
| Archive depth | WorldView constellation archive from 2007; Pléiades from 2011; SkySat from 2014 |
| Spectral bands used | Panchromatic primary for classification; multispectral (4-band or 8-band) for livery colour and door-configuration disambiguation |
| Typical classification accuracy (wide-body) | Above 90 percent reported in published literature under low off-nadir, clear conditions; lower for converted narrow-bodies |
| Delivery formats | GeoJSON stand-level records, CSV time-series index, georeferenced image chips, periodic PDF summary |
Analytics Satellize can run
| Per-stand aircraft classification | Object detection followed by bounding-box dimension extraction and lookup against published aircraft dimension databases | GeoJSON layer with stand ID, detected aircraft dimensions, assigned type family and freighter/passenger flag per image acquisition |
| Freighter-to-passenger ratio time series | Repeated classification runs over a defined airport apron, aggregated by acquisition date, with converted-aircraft uncertainty band applied | CSV time-series table and chart showing weekly or monthly freighter share of occupied stands |
| Estimated instantaneous cargo-payload capacity index | Aircraft type counts weighted by published maximum structural payload per type, summed per apron snapshot | Indexed capacity figure per acquisition, normalised to a baseline period, delivered as a data feed or periodic report |
| Operator fleet deployment map | Livery and tail-marking reading combined with type classification to assign aircraft to known operators where imagery resolution permits | Per-hub operator-presence table updated on each tasking cycle |
| Multi-hub cargo-capacity dashboard | Parallel tasking and classification across a defined set of cargo airports, with consistent methodology applied to each | Comparative dashboard showing freighter-stand occupancy and capacity index across hubs, updated weekly or monthly |
| Converted-freighter ambiguity flag | Dimension-match scoring that identifies aircraft falling within the overlap zone between passenger and converted-freighter type families, flagged for manual review or supplementary data | Uncertainty annotation on each classified stand record, with a summary ambiguity rate per airport per period |
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.