White paper

How do you put a carbon emission value on a train journey?

Rail is one of the lowest-carbon ways to travel, and corporate travel programmes increasingly treat it as the alternative to flying. The emission value itself is where agreement breaks down. The Rail Emissions Model is a published method for calculating that value, and for showing what it rests on. It is global, free to use, and open to challenge.

Read the white paper Challenge the method

Version 2.7, August 2026. Library figures from the factor build of 18 August 2026.

What REM does with the record you have

Set what the booking data holds. The model resolves to the most specific factor it can defend and records how it got there.

What the record says about the train

Where the distance comes from

Tier 1 Operator-specific factor

Emissions
kgCO2e for the trip, well-to-wheel
Distance
user_provided
Method tier
Tier 1
Factor source
The operator's publication and its year
Indicative spread
±10 to 20%

187operators already in the model
150countries modelled from energy data
45countries with operator or national data
10service types, from tram to high speed
5tiers, operator data to global average

Why rail needs one method

Hundreds of operators sell tickets in dozens of formats, and the systems that track travel spending rarely record which train was taken, who ran it, or how it was powered.

The factors that do exist come from operators, national agencies and competing methodologies, and they disagree. They disagree on where the boundary falls, on how electricity is counted, on what counts as a regional train, and on how full the train is assumed to be. No single source works everywhere.

Market liberalisation keeps adding operators, Germany among the markets expecting new entrants from 2028. The strong local models, in the UK, France and Germany, each stop at their own border.

REM works with whatever the booking data holds. It takes the most specific credible factor it can defend, the operator's own published figure where one exists, and steps down a disclosed ladder from there. There is always an answer, so every trip stays in the inventory even when a field is missing, and the answer arrives with its reasoning attached.

The tier ladder

Five rungs, from the most specific defensible figure down to a global average. The tiers are source-selection steps. A lower number means a more specific source was found, which is a different claim from being closer to the truth.

0planned

Per-service operator data

A rung above the current top, for genuinely per-service or per-run operator data rather than a published annual average. The technology is ready. Operator data provision is the constraint, so treat Tier 0 as forthcoming rather than as part of the model.

next major release
1

The operator's own published figure

From the railway's own report, or a national authority's disclosure about that operator. Attributable to a named publication and year, and admitted only through a documented qualification sequence.

±10 to 20%
2

Built from the operator's traction mix

How much of its service runs on electricity, diesel, hydrogen or biofuel, combined with traction-specific factors and the country grid factor. Most profiles carry a shared energy intensity, so the mix is the operator-specific part.

±15 to 25%
3

The country's official rail figure

From a national ministry, statistical agency or recognised database. Sources include DEFRA, ORR, EPA, ADEME and SNCF, matched on country, service category and year.

±15 to 30%
4

Modelled from energy use

Energy per passenger-kilometre and how electrified the network is. It ranks services correctly and sits high on level, so use it for ordering and screening.

±30 to 50%, biased high
5

A global average

The last resort, computed from IEA rail data. Good for coverage, not for close comparison, and labelled as such in the output.

±50% or greater

These ranges characterise systematic variance for disclosure. They are not confidence intervals. Every tier returns secondary data as defined by ISO 14083, ranked by specificity rather than by class.

Every trip comes back with its reasoning

A bare kgCO2e figure cannot be audited. A figure that carries its distance, its tier and its factor source can be traced back to a published document and recomputed.

REM returns four things for every trip: the emissions, total CO2e in kilograms with the disclosed energy scope; the distance, and whether it was given, routed or estimated; the tier the figure resolved at; and the factor applied, with the publication it came from and its year.

The tier and the source travel with the value into the disclosure. Where two results are not comparable, on boundary, electricity basis, factor vintage or distance basis, REM surfaces the mismatch instead of resolving it quietly.

Where the value is used

Booking and travel management

Rail emissions shown as accurately as the data allows at the moment of choice, in the booking process, rather than months later in a report.

Corporate reporting

Scope 3.6 business travel under CSRD and ESRS and other reporting regimes, where the tier and the source travel with the value into the disclosure.

Modal comparison

The rail side of the air-versus-rail question, on the connections where the two compete for the same journey.

Review and governance

REM was published as a white paper because publication is the review mechanism. A framework meant to serve an industry has to be exposed to that industry before it is fixed. Five review layers apply, three of them independent of SQUAKE.

  1. Internal scientific review by SQUAKE's in-house team.
  2. Research institutes and university partners, including the Copernicus Institute of Sustainable Development at Utrecht. independent
  3. Rail operator representatives, testing the framework against how undertakings actually report. independent
  4. Travalyst's Independent Advisory Group, reviewing the model on an ongoing basis. independent
  5. Independent consultants and the open feedback form, which takes challenges straight into development.

Cadence and versions

The whole model gets a full review every quarter. Grid factors refresh each Q1, once the prior year's electricity statistics are published. A material mid-cycle correction can trigger an out-of-cycle patch release with an effective date.

Version numbering is the audit trail. Major versions change methodology or boundary, minor versions extend coverage, patch versions refresh data, and the applied factor vintage is disclosed in the output. The sequence from the first release through 2.7 records what review has changed.

Where a source is challenged and does not hold up, the affected records are removed from the library rather than adjusted, and the operator falls to the next applicable tier until a sound source is found. Partners are notified where a data refresh moves a portfolio by more than 2%, or where a new or corrected operator factor moves results by more than 5%.

Standards and regulation

ISO 14083:2023

REM applies the standard's boundary, functional unit and disclosure principles, reporting well-to-wheel in gCO2e per passenger-kilometre. Certification against the standard is planned and has not happened yet.

CountEmissionsEU

Regulation (EU) 2026/1030 entered into force on 1 June 2026 and applies generally from 2 December 2030. It names EN ISO 14083:2023 and the same per-passenger-kilometre metric REM applies, so the boundary and functional unit already match.

CSRD and ESRS

Business travel emissions are disclosed under ESRS E1-6, with prior periods restated when a methodology change makes them non-comparable. REM's job is to make each change visible and quantifiable.

What REM is not

  • Not a claim to superior local accuracy. REM consumes local models and operator data, so it cannot exceed the best of them. Its function is to decide which source takes precedence and what applies where none exists.
  • Not primary data. Every current tier returns secondary data as defined by ISO 14083 and Regulation (EU) 2026/1030 Article 2(9). An operator's published annual average is a proxy for a specific journey.
  • Not yet peer reviewed and not yet certified. Three work pieces are heading for academic peer review: tier validation, electricity attribution and audit readiness.
  • Not silent about its limits. The white paper carries a limitations register covering load factors, seasonality, detour factors, seat-class double counting, traction-supply losses and the inherited global constants, each with the development priority attached.

Every figure is a published or modelled estimate, ranked by how specific it is, with its limitations written down. Read the full methodology

Who stands behind it

REM is developed and published by SQUAKE and is free to use. A monetised methodology gives its owner an incentive to steer it, and one standard the market aligns to is more useful than fifteen competing claims to accuracy. The precedent is the Travel Impact Model, which convened the sector around a single aviation method.

The Travalyst Coalition has aligned behind the model, with partners holding rail use cases including Skyscanner, Google, Trip.com Group, Amadeus, Sabre, BCD and Pitchup. Integration into the Travalyst Data Hub is under way. Pre-competitive alignment of that kind is the substantive test of whether a framework is credible.

Adoption is voluntary. Availability through SQUAKE's infrastructure is already in place, and implementation follows into partners' distribution channels.

Three ways to get REM running

Adoption is voluntary and the methodology is free in every case. What differs is who carries the maintenance.

Build it in yourself

The white paper carries the full methodology, the tier logic, the parameter tables and the limitations register, so nothing stops you implementing REM in your own stack. The maintenance then sits with you: the operator and country factor libraries, the grid factors and the service-type mappings have to be refreshed every quarter, and each release has to be versioned so your results stay reproducible and comparable with everyone else's.

Maintained by you

Connect to the Travalyst Data Hub

REM is being integrated into the coalition's Data Hub, which gives the industry open access to one maintained data set rather than a copy per partner. Register interest with Travalyst to be set up when access opens.

Maintained by Travalyst

Access it through SQUAKE

Availability through SQUAKE's infrastructure is already in place, in booking, travel management and mid- and back-office environments, with the audit output returned per trip and versioned datasets behind it.

Maintained by SQUAKE

Selected partners

Organisations working with REM, through the methodology itself, through distribution, or by supplying the operator and national data behind it.

SQUAKE
Travalyst
BTA Planet Plan
Rail Delivery Group
amongst others

Questions

What is REM?

A framework that assigns operators to data quality tiers, applies the corresponding calculation approach, and defines fallback logic where better data is unavailable. Open access and free to the industry.

What does it cost?

Nothing. The all rights reserved notice on the white paper is a publication formality and does not restrict industry use.

How are distances handled where an operator supplies a factor but no distance?

Straight-line distances are not applied. Deviation factors by train type are applied to city-to-city distance. Where large open water lies between two points, a polyline approach models a route around it before the deviation factors are applied. The same approach applies regardless of country of origin.

How can tier assignments be verified if the operator list is not published?

Every result names the tier it resolved at and the publication and year the factor came from, so the assignment is visible per trip without reference to a list. The methodology governing tier assignment is public, and the list itself is available on request. Publishing a static file without versioning and a maintenance concept would reproduce the problem REM exists to solve.

What is the path to assurance?

Sequenced: exposure, then operational exposure, then assurance. Certifying before stress-testing would mean reworking once feedback lands. The horizon is under two years. Peer review averages six to nine months and is outside SQUAKE's control.

How does feedback become a version?

Submissions are assessed, grouped by subject, and either incorporated into the next release or routed into the scientific work stream where they require primary research. The expanded service taxonomy began as a reviewer's objection that suburban services were being priced as regional ones. The correction to the detour-factor provenance came from a reader who opened the cited source and found it said something different.

Take part

The call to review is open to anyone: operators, buyers, suppliers, competitors and academics. Submissions are assessed on substance rather than source, and challenge is welcome from parties with a commercial interest in a different answer.

Challenge the method

Disagree with a parameter, have a better source, or think a limitation deserves different treatment. Points raised peer to peer get fixed faster than points raised indirectly.

Send feedback

Get an operator added or corrected

Operators, authorities and partners whose published factor is missing, out of date or mapped to the wrong service category. Send the publication and the place in it. The rest is ours to do.

Submit a source

Access the data

Register interest in using REM through the Travalyst Data Hub, or reach SQUAKE for implementation in booking, travel management and reporting systems.

Register interest