How to Cite an Emission Factor
There is no agreed way to cite an emission factor. Reports say “DEFRA 2026” and stop, which sounds like a citation and is not one — DEFRA published two different 2026 files seven weeks apart, and that string cannot tell you which. A citation has one job: let a reader recover the exact number you used. This page proposes five fields that do that, for any publisher, with copy-paste formats for reports, footnotes, BibTeX and code.
Cite five things: the value and unit, the publisher and dataset, the location inside it (sheet and cell, or a stable key), the edition or version, and the date you accessed it. Publisher-and-year alone is not enough, because publishers revise within a year. If your source offers a resolvable permalink to the exact value at the exact version, include that too and the other five become checkable rather than merely stated.
What a citation is actually for
Academic citation conventions exist to credit an author. An emission factor citation has a different and harder job: reproduction. Eighteen months after publication, somebody — an assurer, a regulator, your successor — needs to take your report, find the number you used, and confirm it was the number you said it was.
That is a higher bar than attribution, and it is why the usual short form fails. “DEFRA 2026” credits the publisher perfectly well. It just does not let anyone find the value, because a publisher can issue several editions under one year, a dataset can contain several rows for what sounds like one fuel, and both of those happened in 2026.
So the test for any citation format is simple: could a stranger, with only this string, arrive at the same number?
The five fields
These are publisher-neutral. They work for DEFRA, the EPA, ecoinvent, a national inventory or a vendor API, because they describe what any factor has rather than how any one publisher formats it.
| Field | Example | Why it is needed |
|---|---|---|
| 1. Value and unit | 0.18231 kg CO₂e/kWh | So a mismatch is visible immediately, without a lookup. |
| 2. Publisher and dataset | DESNZ, UK Government GHG Conversion Factors 2026 | Whose number it is. The dataset name matters as much as the organisation. |
| 3. Location within it | Fuels sheet, cell D42 — or a stable key | Datasets hold thousands of rows and several plausible candidates per fuel. |
| 4. Edition or version | 2026 edition, July revision | The field that distinguishes two files published under one year. |
| 5. Date accessed | 8 September 2026 | Fixes which revision you could have seen, when the publisher does not version explicitly. |
A sixth, when available: a resolvable permalink to that value at that version. Not every publisher offers one. Where it exists it converts the citation from a claim into something a reader can check in one click, which is the difference between being trusted and being verifiable.
Copy-paste formats
Same five fields, rendered for different destinations.
In report prose — when the number appears once and the reader needs it inline:
Natural gas: 0.18231 kg CO2e/kWh (gross CV), UK Government GHG
Conversion Factors 2026, Fuels sheet cell D42, accessed 8 September 2026.As a footnote — when the body text must stay readable:
Emissions from gas calculated at 0.18231 kg CO2e/kWh.[1]
[1] Department for Energy Security and Net Zero, UK Government GHG
Conversion Factors 2026 (July revision), 'Fuels' sheet, cell D42.
Gross calorific value basis. Accessed 8 September 2026.As BibTeX — for academic and technical writing:
@misc{desnz2026gas,
author = {{Department for Energy Security and Net Zero}},
title = {UK Government GHG Conversion Factors 2026},
year = {2026},
note = {Fuels sheet, cell D42; natural gas, kWh gross CV;
0.18231 kg CO2e/kWh; accessed 8 September 2026},
url = {https://www.gov.uk/government/collections/government-conversion-factors-for-company-reporting}
}As a code comment — where the constant lives, if you must hard-code:
# 0.18231 kg CO2e/kWh - UK natural gas, gross CV
# DESNZ, UK GHG Conversion Factors 2026, 'Fuels'!D42
# Retrieved 2026-09-08. Re-check on the next June release.
GAS_KWH_GCV = 0.18231That third line matters more than it looks. A constant with a citation but no expiry is a constant nobody revisits, and the citation ages into a record of when the number stopped being checked.
What most citations get wrong
Publisher and year only. The commonest form, and insufficient for the reason at the top of this page: publishers revise within a year. The UK 2026 factors were published on 11 June and republished on 31 July, correcting values that had shipped as zero. “DEFRA 2026” names both files.
Citing the landing page, not the location. A link to a collection page tells a reader where to start looking, not what to look at. With thousands of rows and several plausible candidates per fuel, the location is the part that does the work.
Omitting the unit basis. “0.18231 kg CO₂e/kWh” is incomplete without gross-or-net, because the net figure for the same gas is 0.20199 — an 11% difference. Any qualifier that changes the number belongs in the citation: gross or net, combustion or well-to-tank, which GWP assessment.
Citing a tool instead of a source. If a calculator gave you the number, the calculator is not the source; it is how you reached the source. Cite the publisher whose dataset the value came from, and name the tool separately if the method matters.
Before publishing, read your own citation as though you had never seen the number and ask whether you could find it. If the answer needs a follow-up email, the citation is not finished — and the person sending that email in 2028 will not have anyone to send it to.
Citing at scale: the factor register
Prose citation stops working past about a dozen factors. Beyond that the workable pattern is a factor register: one table, appended to the report or held as a companion file, with a row per distinct factor used and the five fields as columns.
The body of the report then cites the register rather than the publisher — “factors as listed in Appendix B” — and every number traces through one hop. It reads better, it survives review, and it turns the citation problem into a table you generate rather than prose you maintain.
It also makes the annual update tractable: when a publisher ships a new edition, the register is the diff surface. You compare last year’s register against this year’s factors and the changes are the list of things to re-examine, which is the practice set out in emission factor version control.
If your source can do the work
Assembling five fields by hand for every factor is exactly the sort of task that gets skipped under deadline. Some sources return the citation with the value, which removes the discipline problem by removing the work.
Ours does — every response carries an attribution block, the publisher and dataset in provenance, the exact source cell, the version that answered, and a permalink to that value at that version:
curl "https://api.greencalculus.com/v1/factors?search=natural+gas&limit=1"The permalink is the part worth borrowing whoever you buy from. A citation that resolves to a page showing the value, its source and its edition is checkable by the person reading your report, not merely by you — and that is the whole point of the exercise.
If you are evaluating sources partly on this, provenance depth is one of the five tests in how to evaluate a carbon data API. And if two sources give you different numbers for what looks like the same thing, that is usually a citation problem before it is a data problem — why emission factors disagree covers the four axes.
And if the reason you are here is an approaching assurance engagement, the citation format is one of three things a verifier tests on your factor data. The assurance readiness checker asks the other two — whether you retained the evidence, and whether you can rebuild last year’s figure — and tells you which of the three is your weakest.
Frequently asked questions
No, and 2026 is the year that proves it. The UK factors were published on 11 June and republished on 31 July, correcting values that had shipped as zero rather than blank. That string names both files, so a reader cannot tell which numbers you used. Add the edition or revision and the date you accessed it.
The data source, always — that is where the number came from. Name the calculator separately if its method affects the result, in the same way you would name a statistical package alongside the dataset. A reader needs to find the factor; the tool is how you reached it, not what it is.
Then the date you accessed it does that job, and you should keep a copy of the file with a checksum. Many publishers version implicitly by revising in place, so an access date plus your own archived copy is the only way to demonstrate later what you actually saw.
Cite the inputs with the five fields each, then state the derivation explicitly — the formula, and any assumption you supplied. A derived factor is a small piece of methodology, and the reader needs to be able to redo the arithmetic, not just locate the ingredients. Blend GWPs are the everyday example: composition, constituent values, and the weighting.
Both, split by volume. Up to about a dozen factors, inline or footnoted reads fine. Past that, use a factor register in an appendix and reference it from the body. What does not work is naming the publisher in the body and nothing anywhere else — that is the form that looks cited and is not.
Not that we have found. Reporting standards say figures must be traceable and verifiable without specifying a citation format, so practice varies from a full source cell to a bare organisation name. The five fields here are our proposal for a minimum, offered for adoption rather than as a proprietary format — if a body publishes something better, we will use theirs.