Why Emission Factors Disagree
Ask five reputable sources for the carbon intensity of UK electricity and you get 0.131, 0.217, 0.281, 0.037 and 0.115 kg CO₂e per kWh. None of them is wrong. They are answers to five different questions, and almost nobody labels which question they answered. This page is a field guide to the four things that make published factors disagree — boundary, unit basis, GWP assessment and method — with the real numbers behind each, so you can tell which number you are holding.
Two factors for the same activity usually differ for one of four reasons: boundary (combustion only, upstream only, or full lifecycle), unit basis (gross vs net calorific value — an 11% gap on gas), GWP assessment (AR4, AR5 or AR6 — up to 20% on refrigerants), or method (location-based, market-based or residual mix). Before comparing two numbers, establish that all four match. If they don’t, you are not comparing like with like, and the difference tells you nothing about the underlying emissions.
Five right answers for one country’s electricity
These are all UK grid electricity, all per kilowatt-hour, all current, all from named publishers. They are held side by side in our corpus and you can pull them yourself with no account:
| kg CO₂e / kWh | What it measures | Publisher |
|---|---|---|
| 0.03682 | Well-to-tank only — the upstream fuel supply for generation, excluding the burn itself | DEFRA 2026 |
| 0.11507 | Biogenic CO₂ memo item — reported outside the scopes, never added to your total | DEFRA 2026 |
| 0.13096 | Location-based — direct combustion at the generator, grid average | DEFRA 2026 |
| 0.21741 | Full lifecycle — fuel supply chain and plant manufacture included | Ember 2025 |
| 0.28064 | Residual mix — what is left after contracted renewables are claimed by others | AIB 2025 |
The spread from lowest to highest usable figure is 2.1×. A company reporting on 0.131 and a company reporting on 0.281 are both compliant, both honest, and not comparable — and nothing on the face of either number says so.
curl "https://api.greencalculus.com/v1/factors?key_prefix=grid.gbr&limit=10"Axis 1 — boundary: what is counted
The commonest cause of disagreement, and the one with the largest spread. Three boundaries are in routine use and they are not interchangeable.
Combustion only counts what comes out of the stack or the tailpipe when the fuel burns. Well-to-tank counts everything before that — extraction, refining, transport — and excludes the burn. Full lifecycle counts both, and sometimes the plant that generated the electricity as well.
UK diesel makes it concrete. Combustion is 2.66155 kg CO₂e per litre; well-to-tank is 0.62409. Add them and you get the well-to-wheel figure of about 3.29 — 23% higher than the combustion number most people quote. DEFRA publishes both as separate rows precisely so that you choose deliberately, and the commonest error is to take the combustion factor because it appeared first and call it the footprint.
When a dataset mixes boundaries without labelling them, no amount of care downstream recovers the distinction. Ember’s grid figures are full-lifecycle and DEFRA’s are combustion-only, which is most of the gap between 0.217 and 0.131 above. Both publishers are clear about it in their own documentation; the problem arises when a third party republishes one alongside the other under a single column heading.
Axis 2 — unit basis: gross or net
Energy can be measured on a gross calorific value basis or a net one, and the same gas gets a different factor depending which you use. Not a rounding difference — a structural one:
| Factor key | kg CO₂e / kWh | Basis |
|---|---|---|
fuels.gbr.natural_gas.kwh_gcv |
0.18231 | Gross calorific value |
fuels.gbr.natural_gas.kwh_ncv |
0.20199 | Net calorific value |
That is an 11% difference with no warning attached to either number. UK meters and supplier bills report kWh on a gross basis, so a UK gas bill needs the GCV factor. Net appears in international energy statistics and some IPCC methods, which is why both exist. Choose from where your activity data came, not from which factor you found first.
Axis 3 — GWP assessment: AR4, AR5 or AR6
Converting methane, nitrous oxide or refrigerants to CO₂e requires a global warming potential, and those are revised with each IPCC assessment report. A factor is therefore stamped with an assessment it may not announce.
On CO₂-dominated fuels this barely moves the number. On refrigerants it moves it a lot, because the GWPs are in the thousands:
| Blend | AR4 | AR5 | AR6 | AR5→AR6 |
|---|---|---|---|---|
| R-404A | 3,922 | 3,943 | 4,728 | +19.9% |
| R-410A | 2,088 | 1,923 | 2,256 | +17.3% |
| R-507A | 3,985 | 3,985 | 4,775 | +19.8% |
The step is systematic: AR6 raises every common blend by roughly a fifth. Which basis you owe is set by your framework rather than by recency — UNFCCC decision 18/CMA.1 mandates AR5 for national inventory reporting, several corporate frameworks follow it, and older baselines often sit on AR4. The full picture, including the six newer blends that exist only on AR6, is on the refrigerant GWP page.
Axis 4 — method: whose electricity is it
Unique to electricity, and the reason two of the five opening numbers exist. Location-based uses the grid average where you consumed. Market-based uses what you contracted for, so a renewable tariff can take it toward zero. Residual mix is what remains once everyone else’s contracted renewables are stripped out — which is why the UK residual figure, 0.281, is more than double the location-based one.
The GHG Protocol Scope 2 Guidance requires dual reporting: both a location-based and a market-based total. A single electricity number, unlabelled, does not satisfy it, and the two are not alternatives to choose between. The Scope 2 methodology page sets out which applies when.
Telling which one you have
Four questions, in this order. They resolve almost every disagreement without needing to contact the publisher.
- Read the unit string, not the number. “kg CO₂e per kWh” and “kg CO₂ per kWh” are different quantities — the second omits methane and nitrous oxide entirely. A source that reports CO₂ where you assumed CO₂e is understating, silently.
- Look for the words well-to-tank, WTT, lifecycle, or combustion. If none appears anywhere near the figure, assume combustion-only and verify before adding anything to it.
- Find the GWP basis. If the source does not state AR4, AR5 or AR6, and the factor covers anything other than pure CO₂, you cannot restate it later against a different basis.
- For electricity, find the method word. Location, market or residual. If it is absent, it is almost always location-based — but almost always is not good enough for a disclosure.
If a source answers none of the four, the honest conclusion is that you have a number without a definition. It may still be the right number; you just cannot demonstrate that it is.
What good labelling looks like
Our position, since this page is on our site and you should discount it accordingly: we think every factor should carry its boundary, its unit basis, its GWP set and its method in the payload, not in a PDF footnote. That is why the five UK figures at the top of this page sit as five separate keys with five separate basis strings rather than as one number with a caveat.
It is also why we hold values that make us look worse. The residual-mix figure of 0.281 is the least flattering number we publish for UK electricity, and it is there because for a company without a renewable contract it is the correct market-based input.
If you want to check any factor’s labelling — ours or anyone’s — the boundary checker walks the four axes for a single figure. And if you are choosing a data source rather than diagnosing one, how to evaluate a carbon data API turns the same questions into five runnable tests.
Frequently asked questions
The one your reporting framework names, matched to the boundary your activity data describes. For a UK company reporting under SECR that is normally the DEFRA location-based figure for electricity and the DEFRA combustion factors for fuels, with well-to-tank reported separately where required. “Which is most accurate” is the wrong question — they measure different things, so accuracy is not what separates them.
For fuels, generally yes, and DEFRA publishes the two components separately so you can add them. What you must not do is add a well-to-tank figure from one publisher to a combustion figure from another without checking they share a unit basis and a GWP set — that is two of the four axes, and mismatching either makes the sum meaningless.
Because contracted renewable generation has already been claimed by the companies that bought it. Removing it from the pool leaves a dirtier remainder, which is what an uncontracted consumer actually draws on in market-based terms. A residual mix that equalled the grid average would be a contradiction in terms — it would mean no renewable energy had been contracted by anyone.
Probably not, and it is likely a vintage difference rather than a structural one — grid factors move a few per cent a year on generation mix alone. The differences worth investigating are the structural ones on this page, which are 11% and upward and do not shrink over time because they are not measurement noise. If two sources differ by 100%, you have a boundary or method mismatch, not a data-quality problem.
Between GWP assessments, yes, if the source publishes the gases separately — recombine the CO₂, CH₄ and N₂O components with the GWP set you need. Between gross and net calorific value, yes, with the calorific values. Between boundaries, no: a combustion factor does not contain the information needed to derive its upstream half. That is why keeping the gases separate in your own store is worth the effort.
They are usually clear in their own documentation; the failures cluster in republication. We have had to correct our own labelling too — our Ember-sourced grid rows were re-annotated in August 2026 once it was clear they were full-lifecycle rather than location-based, and 212 rows moved to a correctly named key. That is the failure mode to watch for in any aggregated source, including this one.