Emissions factors
These figures are computed. Nobody measured a vehicle. Every number on this site comes from a published emission factor multiplied by a modelled distance, and this page states which factor, which publication, which table and which row.
What this is not
It is not a measurement of your journey. No sensor, no fuel log and no carrier report goes into any figure here. It is an average for a mode, applied to a distance we work out from two sets of coordinates. Treat every number on the site as an estimate of the right order of magnitude, not as the emissions of the service you are about to board.
How the figure is worked out
Three steps, and the whole of it fits on one line. Take the great-circle distance between the two cities. Multiply it by a detour factor, because track and road are longer than a straight line. Multiply that by the factor for the mode.
grams = round(factor for the mode x round(great-circle km x detour))The detour factor comes from the corridor's own record in the network, and is 1.03 for a flight, which is the allowance for departure and arrival routing rather than a road detour. Distance is rounded to whole kilometres before the factor is applied, so a journey of 174.4 km is charged as 174.
There is exactly one distance test in the whole path: a flight over 3,000 km uses the long-haul factor instead of the short-haul one. Everything else uses one factor per mode at every distance. The service classes shown on the mode pages are bands for reporting, not bands for emissions.
The factors, and where each one comes from
This is the whole table. Every row now names the document, the table or flat file row, the year and the accounting basis it was read from. In version 1.0.0 three rows named nothing at all, because an invented citation is worse than an admitted gap; those three have been replaced with published figures rather than dressed up.
| Factor | Applies to | gCO2e per pkm | Scope | Nearest published | Source |
|---|---|---|---|---|---|
| rail | Train, any distance | 33 | Well to wheel | 33.35 | Matches the source |
| rail-electric-hs | Declared in the model, applied to no journey | 16 | Well to wheel | 15.8 | Matches the source |
| coach | Coach, any distance | 34 | Well to wheel | 33.73 | Matches the source |
| air | Flight, up to 3,000 km | 209 | Well to wheel | 208.78 | Matches the source |
| air-long | Flight, over 3,000 km | 293 | Well to wheel | 293.41 | Matches the source |
| ferry | Ferry, any distance | 23 | Well to wheel | 22.95 | Matches the source |
| transfer | Declared in the model, applied to no journey | 143 | Well to wheel | 143 | Matches the source |
| walk | Declared in the model, applied to no journey | 0 | Well to wheel | 0 | Matches the source |
Three rows in this table are applied to nothing. The high-speed rail figure is declared and never selected: no operator record distinguishes a high speed service from any other train, so every train is charged the mixed-traction average. The car transfer and walking figures are declared and never selected because connections between legs contribute nothing to a journey total: the ride to the station counts in the door-to-door duration and not in the door-to-door emissions.
The assumptions, stated
- Occupancy
- Each factor carries its publisher's own occupancy assumption and nothing here changes it. The rail and car figures are fleet averages implied by observed passenger kilometres, and the car figure works out at 1.6 people per car. The coach figure assumes 17.56 passengers, a figure its publisher notes is drawn from bus and coach data to 2007 and is probably low for a coach. The ferry figure assumes a foot passenger and luggage and no share of the vehicle deck. The aviation figures use the load factor the airlines flew.
- One basis, well to wheel
- Every factor here counts both the fuel burned in the vehicle and the energy spent producing and delivering that fuel. This is the only basis on which the two publishers can be quoted side by side: the European inventory publishes a well-to-wheel total and does not let you take the upstream half back out, while the UK set publishes the two halves separately, so its well-to-tank row is added to its direct row to reach the same basis. Version 1.0.0 mixed the two, which understated coach and ferry by about a quarter against rail and made every cross-mode comparison on this site quietly wrong.
- Radiative forcing
- Aviation warms the climate by more than its carbon dioxide alone, through contrails and nitrogen oxides, and a multiplier is the usual way of carrying that. Both air factors here include one, and it is the same multiplier on both: 1.7, the central estimate the UK set recommends and applies to the carbon dioxide component only. Version 1.0.0 claimed 1.9 on both and in fact had one figure with a multiplier and one without, so it reported long-haul flying as cleaner per kilometre than short-haul, which is the wrong way round.
- What is left out
- Building and maintaining track, road, runway and vessel. Manufacturing and scrapping the vehicle. Getting to the station and away from it at the other end. None of it is in any figure on this site.
What it comes to, by mode
Measured over the 8,994 corridors where more than one of our modes actually runs, so both halves of every comparison come from the same set of city pairs rather than from two unrelated networks.
- 6,831Train
- 7,208Coach
- 5,382Ferry
- 113,069Flight
| Mode | Corridors | Median km | Median gCO2e |
|---|---|---|---|
| Train | 8,160 | 207 | 6,831 |
| Coach | 8,151 | 212 | 7,208 |
| Ferry | 371 | 234 | 5,382 |
| Flight | 1,855 | 541 | 113,069 |
On the 1,052 corridors where both a train and a flight run, the flight is a median 6.52 times the train.
A median is not a promise about any one journey. Half the corridors are above it. The ratio moves with distance, because the same two factors are applied over the same kilometres, so it tells you about the factors as much as about the corridors.
How wrong this could be
Four sources of error, in the order they matter. The 1.7 aviation multiplier is called significantly uncertain by the publisher that recommends it, and it is the largest single lever on any flight figure here: leaving it out would cut both air factors by about forty percent. Three of the eight factors are UK figures applied to a European network, because no European publisher separates an intercity coach from an urban bus or a foot passenger from a car passenger. The distance is a great circle with a flat detour factor rather than a route, so a corridor through mountains is understated. And an average for a mode says nothing about a particular service, whose real figure moves with the train, the aircraft, the load on the day and the electricity behind the wire.
Where a number is an estimate, this site says estimated. Every figure derived from this dataset is an estimate.
What is wrong with this today
Stated here rather than in a changelog nobody reads, because a journalist quoting the flight number should know this first.
- Applying a radiative forcing multiplier at all is a judgement, not a measurement. Both air factors carry 1.7, and the publisher recommending it calls that value subject to significant uncertainty. A comparison site that omits it shows flights about forty percent lower than this one does, and neither of us is reading a different aircraft.
- Rail and coach cannot be separated on these numbers. The published figures are 33.35 and 33.73 grams, a gap of about one percent, from two different publishers, two different years and two different fleets. Rounding to whole grams widens that gap to three percent on screen and in the table above. Nothing here supports calling either mode greener than the other, and a ranking that puts one above the other on this margin is reading noise.
- Three factors are UK figures applied to a European network: the coach, the ferry and both aviation bands. The European inventory publishes a bus category that mixes intercity coaches with urban buses at fourteen passengers a vehicle, and a ro-pax figure that does not separate the foot passengers from the cars on the deck. The UK set answers exactly the question this product asks, and it answers it about a different fleet.
- The high-speed rail factor is applied to nothing. The registry models speed per mode rather than per operator, so every rail operator carries a nominal 140 km/h and nothing in the data can tell a high speed service from any other train. The factor is kept and sourced so that a feed carrying real operator speeds needs a flag rather than a number.
- Connections contribute zero grams. A connection carries minutes and walking minutes and no distance, and every factor here is per passenger kilometre, so there is nothing to multiply. Assuming a transfer distance would put an invented number beside sourced ones with no way to tell them apart, so the zero stays and is stated. The car factor exists for the day a distance does.
Take the data
Two CSV files, a Frictionless descriptor, Table Schema per file and SHA-256 checksums. Every row carries its source, the URL, the date it was read and whether it is filed or derived.
- factors.csv, the table above with the full provenance note per row
- corridor_emissions.csv, 18,537 rows, one per mode per corridor where more than one mode runs
- datapackage.json, the Frictionless descriptor, with a Table Schema per file
- CC BY 4.0. Cite it and reuse it, including commercially. The EU database right is waived on the same terms.
Orlero. Emission factors and corridor emissions, version 2.0.0. 18 September 2026.