Estimated fuel legs: the supply chain before the burn

Most fuel publishers print one number per litre, kilogram or kilowatt-hour: the emissions of burning the fuel. Before it reached the burner, the fuel was extracted, processed and transported, and that supply chain, the well-to-tank leg, adds from about 4 % to about 35 % of the combustion value for the fossil fuels on this page, and 10 % or more for all of them but three US coal ranks. Where a publisher prints the leg, factor search serves it as published. Where a publisher prints combustion only, open-climate.ai adds the leg, computed by one stated rule from the publisher's own combustion value and a supply-chain ratio from a public source, sums a total beside it, and marks both as estimates. This page says how the leg is computed, what it is worth, and how to leave it out.

The leg and the total#

Both live on their own factors beside the base factor, with the same activity, region, unit and reference year, and the same licence as the base. Where the publisher prints the leg or a total, that printed number stays and nothing is estimated.

How the leg is computed#

The ratio. A donor is a source that prints both legs for the same fuel: a well-to-tank value U_d and a combustion value C_d, in the same unit, on the same heating value and for the same year. The ratio is r = U_d / C_d, and the leg is U = C x r. Because the donor's two values share their unit and heating value, both cancel in r: a ratio read per kilowatt-hour of net calorific value scales a base per litre, per kilogram or per kilowatt-hour of gross calorific value the same way. The leg also follows the base's own fuel: a fuel that emits more per litre when burned gets a proportionally larger leg.

Which year. The donor's year is the base's own year, else the nearest earlier, else the nearest later; rung W0 takes the same year only. The year used is printed in the leg's description. The UK Government's factors are loaded for 2026 only, so every W3 leg uses the 2026 ratio whatever the base's year.

Blends. Diesel and petrol with up to 10 % bio content (B5, B7, B10, E3, E5, E10) take the UK Government's average-blend ratios: diesel 0.23708, petrol 0.27974. A blend above 10 % is refused, as below.

A ceiling. A ratio of 1 or more is a mapping error, not a fossil supply chain, and no leg is written. The largest ratio served is LNG's, 0.353.

Where the ratio comes from#

Four rungs, and the first that resolves wins. The rung caps the quality tier the leg can reach, and names the reason it carries.

• New Zealand's ratio is built from its own inventory: production, processing, venting, flaring and the gas the energy industry and the pipeline compressors burn, over the gas burned in the economy. It works because New Zealand produces all its gas at home, so the inventory holds the whole chain. It leaves out the leaks of the gas network, which MfE publishes as a separate factor (see the MfE example).
• The United States' coal ratios cover coal delivered to a power plant. NETL has no anthracite basin, so anthracite, the other mixed-sector coals and petroleum coke take W3.
• JEC has no lignite pathway, so lignite in the EU takes W3. Finland's gas has come as LNG since 2022; the EU piped mix stands in for it, and the description says so.
• Coal types map to the UK Government's nearest: power-station coal to "Coal (electricity generation)", other hard coal and lignite to "Coal (industrial)", household coal to "Coal (domestic)", coking coal to "Coking coal". The description names the donor fuel.
• Judgement calls, each recorded in the fuel map: RVO's methane is mapped as natural gas, and its refinery gas and ethane as other petroleum gas; Statistics Finland's heating gas oils, about 8 % bio, as the diesel blend; MITERD's unlabelled diesel (Gasóleo A) and petrol (Gasolina) of 2007 to 2018 as the mineral fuels; Japan's city gas takes the LNG ratio, its small share from domestic wells included.

The candidates are the publishers' combustion factors for fuels, as factor search serves them; a value held back from search is not a candidate. Every ratio, with its source, edition and served identifiers, is on the parameters page.

What gets no leg#

A refused factor keeps its published value and gets no leg and no total.

GWP values#

• A leg carries its donor's GWP set. AR5 from the UK Government, which has held its well-to-tank fuels at AR5 since 2023 (methodology paper, paragraph 1.18, p. 16). co2emissiefactoren.nl's own row for W0: AR5 for 2024, AR6 for 2025 and 2026. AR6 for NETL and EPA (CH4 29.8, N2O 273), AR5 for MfE's inventory, AR4 for JEC.
• A CO2-only base misses a small share. ISPRA, Japan's Ministry of the Environment, RVO and Statistics Finland print CO2 only. A CO2e ratio on W0, W1 or W3 then misses the base's own CH4 and N2O share, so the leg is low by it. On the UK Government's 2026 combustion values that share is 0.20 % for natural gas, CNG and LNG, 0.10 to 0.14 % for LPG, propane, butane and other petroleum gas, 0.31 to 2.07 % for other liquids, 0.52 to 9.40 % for coal and 0.28 % for petroleum coke, median 0.52 %. It is below every rung's median error and is not added to it.
• W2 runs the other way. JEC's ratio is CO2e over CO2, so a CO2-only base gets every gas of the supply chain, and a CO2e base such as MITERD's comes out higher by its own combustion CH4 and N2O share.
• Ratios are not converted between sets. The CH4 part of a leg would move 6.4 % from AR5 to AR6, and 19.2 % from AR4 to AR6; a whole leg moves less, by its CH4 share.
• Totals. A CO2-only base takes the leg's vintage, so its total is AR5, AR6 or AR4 as the leg is. An AR6 base with an AR5 leg, EPA and MITERD on W3, gives a mixed total (GWP values).

What the legs are worth#

A leg is served at the worst of three tiers: the cap of its rung, the tier of the base value it multiplies, and the tier its calibration error earns on the same scale as a publisher's stated uncertainty (how the tier is set): high below 25 %, medium below 50 %, low below 100 %, very_low above. The calibration error is measured against the publishers who print both legs: the rule's ratio against theirs, fuel by fuel. The combustion value cancels, so the error is s = max(r_rule, r_published) / min(r_rule, r_published) - 1, and a factor 2 over and a factor 2 under both score 100 %. It was measured once on 2026-09-25, no ratio was tuned to it, and the worst per rung and fuel family is served as the value's uncertainty_pct.

Watch out. No tier moves: W1's errors read medium and are capped low, W2's read low and are capped very_low, and every W3 worst reads very_low, its cap. What the error changes is the number you see. Median and worst tell different stories: a typical W3 gas or liquid leg is about 25 % off, a typical coal leg about 2 times off, and a few Australian rows set every W3 worst (without them, 66.8 % for gas, 124.1 % for liquids, 553.2 % for coal). The US gas figure depends on where the boundary sits: taking out leaks after the meter, which NETL's chain ends before, gives 52.7 %, and counting pipeline compressor fuel as upstream, as NETL does, gives 7.3 %. 42.1 % is served because it adjusts nothing. Treat a very_low leg as an order of magnitude.

The total#

Beside each leg, open-climate.ai sums a well_to_use total: the published combustion value plus the leg, one per base value. Its boundary_profile is {"upstream_fuel": "in", "fuel_combustion": "in"}, the profile the totals of publishers who print both legs carry. The total is is_estimated, takes its worst part's tier and the GWP rule above, and never replaces a published total. It carries its worst part's reason only when the parts at that tier agree on one: an RVO W0 total adds a medium published value to a medium leg, so its reason is empty and is_estimated is the mark to read. On a factor with several supply menus, each menu gets its own total (the Japan example).

Licences#

• Inputs. A source is used only where its licence grants all five rights, with no condition beyond credit, stating changes, no endorsement and notifying the publisher (five questions per licence). Seven pass: NETL's natural gas and coal studies (CC0 1.0), EPA's 40 CFR 98 Tables C-1 and C-2 (US Government work, public domain), JEC Well-to-Tank v5 (CC BY 4.0), MfE's greenhouse gas inventory and MBIE's energy balance tables (both CC BY 4.0), the UK Government's conversion factors (Open Government Licence v3.0) and co2emissiefactoren.nl (its terms: credit and notice to the publisher).
• Targets. A publisher's combustion value may be scaled unless its terms answer "no" or "ask" to modifying and combining the data. No published number is ever changed, so a term that only forbids altering the figures, such as the one under Spain's Law 37/2007 for MITERD, is met. The IPCC's emission factor database is held out: its terms ask permission to modify. The legs sit in the publisher's own release and carry its licence and its credit line, so they are redistributable exactly where the base value is, and never more. Every leg's source_record names its inputs and their licences.

A library download credits every input beside the publisher, on its References sheet: each source's citation, its licence and the change open-climate.ai made to its figures, which CC BY 4.0 and the Open Government Licence both ask to be stated. An XLSX export from factor search lists them on its notes sheet. A CSV file or a SQL query carries the publisher's credit line only, so where is_estimated is true, add these credits yourself (keep only rows you may redistribute).

Which libraries get legs#

Publishers that print the leg keep their own: the UK Government, ADEME, co2emissiefactoren.nl and .be, the Austrian and German environment agencies and Australia's NGA factors. A few of their fuels print combustion only; those are left for later, since the natural donor there is the publisher itself.

Worked examples#

Each value is per the base's own unit, in kg CO2e (kg CO2 where the base is CO2 only). The combustion value is the publisher's; the ratio, the leg and the total are open-climate.ai's.

W0: RVO natural gas#

co2emissiefactoren.nl prints both legs of natural gas for 2024: 0.355 over 1.779 kg CO2e per normal cubic metre, served as 0.000355 over 0.001779 per litre, so r = 0.19955, AR5. The unit cancels, and the ratio scales RVO's per-kilowatt-hour value: 0.20232 x 0.19955 = 0.040373. RVO's 2025 and 2026 values take the list's own year, whose ratio is the same but AR6, so those totals are AR6. The leg is medium, the W0 cap, with no uncertainty_pct. RVO prints CO2 only, so the leg is low by natural gas's CH4 and N2O share, about 0.2 %.

W1: EPA and MfE natural gas#

EPA. NETL's 2020 profile, the nearest earlier year, puts 8.728114 g CO2e upstream of each megajoule (gross calorific value) of gas delivered through the distribution network; EPA's own 40 CFR 98 combustion factor for the same megajoule is 50.3453 g: r = 0.173365. EPA's base is high and AR6; the leg is low by its cap, with an uncertainty_pct of 42.1. Browse the base in factor search.

MfE. New Zealand's inventory for 2024 gives r = 0.131901, AR5, so the leg is 0.19543032 x 0.131901 = 0.025777 and the total 0.221208, low by the cap, with no uncertainty_pct. MfE also publishes the leaks of its gas transmission and distribution network as a separate factor, 0.005823648 kg CO2e per kWh gross in 2024. The ratio leaves those leaks out so they are not counted twice, and that factor stays outside the total: add it beside the total if you report the network.

W2: Statistics Finland and ISPRA natural gas#

JEC's EU-mix piped gas, 2020 data (the nearest earlier year for both): production, long-distance pipeline and high-pressure distribution, 9.00842 g CO2e over 56.1169 g CO2 per megajoule (net calorific value), r = 0.160530, AR4. Both bases print CO2 only, and the ratio is CO2e over CO2, so the leg carries every gas of the chain. Finland's gas has come as LNG since 2022, so the EU mix stands in for a supply chain it no longer describes. Both legs are very_low, with an uncertainty_pct of 72.2, and both totals are AR4.

W3: EPA diesel, MITERD B7 and Japan's city gas#

EPA diesel. The UK Government's 100 % mineral diesel, 2026: 0.06291 over 0.26806 per kWh net, r = 0.234686, AR5. EPA's base is per litre and AR6, so the total's vintage is mixed. uncertainty_pct 179.7, the W3 liquids worst.

MITERD B7. A blend with 7 % bio takes the UK Government's average-blend diesel: 0.06181 over 0.26071, r = 0.237083. MITERD prints CO2e under AR6; the leg is AR5, so the total is mixed. Browse the base in factor search.

Japan's city gas. Japan's Ministry of the Environment publishes city gas per litre of gas (0.00205 kg CO2 per litre is 2.05 kg per cubic metre), CO2 only, for 2025. Almost all of it is imported LNG, so it takes the UK Government's LNG ratio, 0.07214 over 0.20434 = 0.353039, and the leg is 0.000724, very_low, with an uncertainty_pct of 113.5. The menu rule: a supplier's gas and its supply chain are the same whatever contract it is sold under, so every supply menu of every supplier carries this one leg, computed from the default value (代替値), a menu sold at zero emissions included. Each menu gets its own total: that menu's own value plus 0.000724.

Conformance with the GHG Protocol#

The Scope 3 Standard, category 3 (Table 5.5, p. 41) and its Technical Guidance (Table 3.1, p. 39) place the extraction, production and transportation of the fuels a company burns, their well-to-tank emissions, in scope 3 category 3, reported apart from the combustion itself in scope 1. So report the leg in category 3 and the published combustion value in scope 1. The well_to_use total adds the two for the whole chain, the form ISO 14083 uses for transport, where well-to-wheel is well-to-tank plus tank-to-wheel.

• Representativeness. W3 applies the UK's supply chain to another country's fuel, and W2 the EU average to one member state: a representativeness deviation of the kind the Technical Guidance (p. 44) and the Scope 3 Standard (Table 7.6) ask you to disclose.
• CO2-only bases give a leg low by the base's CH4 and N2O share, as above.
• No double counting with a network factor. Where a publisher prints network leaks separately, as MfE does, the ratio leaves them out.

Whatever you report, disclose the method: a reader of your inventory should learn that the well-to-tank leg was estimated, from which source, and at which tier. None of this is published or endorsed by the GHG Protocol or by any of the publishers named.

How a leg is marked, and how to leave it out#

• On the factor, is_estimated is true, and so is is_derived, which it implies; the total is is_estimated too. In factor search the row carries the badge "Added by open-climate.ai" and its description is open-climate.ai's: who added it and from whose data, the method with the actual numbers, the donor year, the assumptions, the cap and its reason, the sources with their licences, and a not-endorsed line. In the API the field is estimated.
• The description of the leg and of the total is written in English, French and German, each from its own template with the same numbers, not machine translated.
• On the value, co2e_origin is estimated_from_reference. It is never anything else on a leg.
• system_boundary is upstream_fuel on the leg and well_to_use on the total, and the leg's data_quality_reason_normalized is one of the four rung tokens above. estimated_from_proxy_region is used by fuel legs only.
• The leg's key is the base factor's key with _est and the boundary appended (//_est//upstream_fuel), and the total's key is the base factor's key with _sum and the target boundary appended (//_sum//well_to_use). The keys are stable, so when the base or a donor is re-imported, the refresh that follows replaces the leg and the total in place rather than adding new ones. The value's source_record names the base value, the rung, the fuel, whether a blend ratio was used, the donor values or ratio row, the licences, the formula and the donor year.

To use published numbers only, filter is_estimated = false in the open-data files or SQL (the files carry is_estimated and co2e_origin as columns), or estimated in the API. On a top-level query, keep is_top_parent out of that filter: the publisher's combustion factor sits one level down, beneath the estimated total, so take it in the total's place with NOT is_estimated AND (is_top_parent OR parent_id IN (SELECT factor_id FROM open_ef.factors_flat WHERE is_estimated)). The estimated electricity legs share upstream_fuel and three of the reasons; to drop the fuel legs alone, drop is_estimated rows whose category_code is fuels. is_derived = false is stricter: it also drops the totals summed from a publisher's own printed parts (parents, children, double counting).

Related#

• The parameters and sources behind the fuel legs
• Estimated electricity legs
• System boundaries
• Data quality ratings and the normalised tier
• GWP values and IPCC report versions