Docs / What was changed in each source

Changes to EF 3.1

Part of What was changed in each source. What is still open about EF 3.1 is on the issue tracker under the label ef-3.1.

EF 3.1 is the base list: its 93,993 rows are transformed into the consensus list rather than merged into it, and every other source is matched against the result. Its curated corrections live in ef-3.1-manual-fixes.json, which holds 210 entries; 74 of them alter a registry number and are listed in full below. The merge evidence — which rows were split and which were joined — is read off the build of 2026-08-29, commit 54b0f9d, and every uuid quoted is EF's own.

One name, two substances

EF 3.1 writes one name on two sets of thirteen rows, one set per compartment, and gives the two sets different registry numbers. The number decides the substance (see When a name and a CAS number disagree), so each set becomes its own flow object. Seven names split this way; an eighth, Water, splits on the compartment.

Each set holds the same thirteen compartments. The uuids, first set then second, in the order the table above uses:

The tridemorph row named for the commercial product has no indoor-air copy, so that set is twelve rows.

Water carries one CAS number, 7732-18-5, on all nine of its rows, and still splits, on the compartment: water released to air is water vapour, and water released to a water body is liquid water. The five air rows — 0342f5e5-b53e-4cec-9e67-fb197f24fff0 (unspecified), 2461cd99-2340-41b0-a110-1709835dea3f (long-term), 5cfda72a-b91b-4f7c-a8d4-1f8265d8cd4c (stratosphere and troposphere), 896d1d1e-2b87-4a83-b209-11a2ea997b49 (non-urban), 3960d174-95af-41ac-a7f3-a946cd237245 (urban) — are published as Water vapour; the four water rows — 5e50fc01-19c6-4377-a1cc-bc65a12498ea (fresh water), 631ecf13-0e51-4e35-8235-c6f80c60d72c (sea water), a3876d9b-a3e8-4680-861a-4e08642f1392 (unspecified), 4b3fe20b-eae3-4963-8326-d3d32fbfeeff (long-term) — as Water. All nine are in cubic metres.

An ionic charge was added or corrected

EF 3.1 names most of its dissolved metals with a Roman numeral — chromium (vi), arsenic (iii), iron (ii) — and the consensus list spells the same charge as Chromium(6+), Arsenic(3+), Iron(2+). Twenty-seven names were re-spelled that way on the 2026-08-29 build; that is notation, and none of them is listed here.

Six names say ion and give no charge, and in each case EF's own registry number on the row is a charged ion's. The published name now carries the charge the CAS number states. Each was approved as a ruling in preferred-label-decisions.json under #627, which applies one test: the rename must remove an ambiguity rather than add a claim. Sodium, lithium and potassium have exactly one ionic form, so the charge asserts nothing new. Tin is the case that mattered: the list already published Tin(2+), and Tin, ion sat beside it without saying that it was the other tin (#617).

The twenty metal-ion rows are all emissions to water: fresh water, sea water, unspecified and unspecified long-term. Methylmercury, CH3Hg⁺, carries a single positive charge in every case, so its rename asserts nothing the number did not already say. Titanium, ion (22541-75-9) keeps its generic name by a rejected ruling in the same file (#698): the list publishes no other titanium ion for it to be confused with.

Two charges were kept by correcting the registry number that contradicted them. Thirteen rows named vanadium (v) carry 15121-26-3, which is the CAS number of vanadium(2+); the name is right and the CAS number was wrong, so the CAS number was changed to 22537-31-1 and the rows are published as Vanadium(5+). Thirteen rows named cerium(3+) (3558989d-d552-499a-a235-54bf69879f7d in aircraft cruise height, ef6d6805-f188-4e31-8d34-a7ddc69aac48 in urban air, and eleven more) carry 7440-45-1, which is the CAS number of cerium the element — the number EF also writes on its cerium resource row (08a91e70-3ddc-11dd-925f-0050c2490048). On the 2026-08-29 build the number won, the thirteen joined the resource row on one substance published as Cerium, and the charge EF wrote on purpose was dropped along with the distinction between the ion's toxicity factors and the element's. The ion's own number, 18923-26-7, is now written on the thirteen rows, and the element's EC number 231-154-9 taken off them, so on the build of 2026-08-30 at 9fbb0d5 they are published as Cerium(3+) (#831). Both corrections are in the table under Registry numbers corrected.

Two names carry , ion that the published name drops, and in each case EF's own CAS number is what decided it:

• Acrylate, ion carries 79-10-7, acrylic acid, and is published as Acrylic Acid (three rows, including 94966655-8318-414b-9fcf-40be777f62de to the ocean and 0eed1672-eb5d-4324-95f4-3ca787797e45 to surface water).
• Perchlorate, ion carries 14797-73-0, which Common Chemistry names Perchlorate, and is published under that name (four rows, including 429408a8-1a0a-44b0-90d0-7616a0e7e1cb and 23e5a0b6-7cde-4f47-baec-87ab213cbbb8).

Rows merged into one flow

EF 3.1 publishes some substances twice in one compartment, usually once under an industry code and once under the chemical name, and often with the characterisation factors on only one of the two. Once both rows carry the same registry number they are one substance, and a curator ruling in elementary-flow-collision-decisions.json names which row survives and which is retired onto it. There are 48 such rulings, all of them merges, over sixteen pairs of names; the retired row keeps its uuid and points at the survivor, so an inventory written against it still lands (#304, #400, #530, #885).

In every case where one row holds the characterisation factors, that row survives. Where both are characterised identically, the row carrying the registry number, the synonyms and EF's own comment survives. Where neither carries a factor at all — the water vapour pair — the row every other list's correspondence and override names survives.

The four primary energy from … rows come from a block of legacy ILCD identifiers (04202046 to 0420204a) that EF inherited beside its newer names. Most of the other pairs exist because the registry number on the code row is this project's — supplied in the table below — and the moment the two rows carried one CAS number they became one substance.

Registry numbers corrected

Every entry in ef-3.1-manual-fixes.json that changes a CAS or EC number is here, in four groups. The fix file holds the full reasoning for each.

Numbers that named the wrong thing

Numbers removed because they belong to something else

Numbers supplied where EF 3.1 shipped none

Ten ordinary substances, each shipped without any identifier:

Fifty-five refrigerants, blowing agents and heat-transfer fluids that EF 3.1 names only by an industry designation — HFC-245cb, PFC-218, HG-02 — and ships with no formula, structure or number (#202). A designation is a complete specification, but only a table a reader can audit turns it into a substance. Two kinds of evidence were accepted. For thirty-five, PubChem's record for the compound lists the designation among its own synonyms, and where the designation's digits encode a formula, the formula was checked against the record. For the other twenty, PubChem holds no record carrying the code, and the link is EF's own characterisation factor: EF's Climate change factors are a transcription of IPCC AR6 WGI Table 7.SM.7, which prints the designation, the structure and the GWP100 on one line, so a flow characterised at 616 kg CO2-eq is the substance whose row reads 616.

One designation was deliberately left without a CAS number: i-HFE-7200 is 163702-06-5, and EF 3.1 already carries that CAS number on HFE-7200, the commercial product, which is a mixture of the n- and i- ethers. Asserting the number would merge a mixture with one of its components.

Numbers the build itself filled in

Besides the curated entries, the build adds a registry number to an EF substance that shipped none where a single, checkable source states it. Eighteen substances received one on the 2026-08-29 build:

A name lookup is accepted only where the flow had no CAS number at all. It never replaces a CAS number EF shipped, for the reason When a name and a CAS number disagree gives: a generic name looked up returns a generic number, and the specific one EF wrote would be lost.

Name changes, for completeness

The remaining 137 entries in the fix file change a name, a synonym list, a compartment or a unit rather than a CAS number, and are outside the four kinds of change this page is about. Seventy-four rename a substance from its industry part number to its chemical name — CFC-12 to Dichlorodifluoromethane — keeping the part number as a synonym (#521). The rest repair EF's own text: sixteen labels broken by a line wrap in the source (eico- safluoro for eicosafluoro, and four more of the kind), Palladium-234m for protactinium-234m (palladium has no isotope anywhere near mass 234), Praseodym-147 for Praseodymium-147, silver-110 for silver-110m (the long-lived isomer a reactor inventory reports), HFC-1234yf for HFO-1234yf (the substance is an olefin), turpentine for Gum turpentine, dmpa for Zytron, polyhaloalkene for 2,3,3,3-Tetrafluoropropene, 2-hydroxypropanoic acid for D-Lactic acid, mecoprop-p for Mecoprop, and silicium tetrafluoride for silicon tetrafluoride — Silicium is the Latin name of the element, EF is the only list here that uses it for SiF₄, and CAS, ecoinvent, BAFU and AGRIBALYSE all write Silicon tetrafluoride (#853) — and a handful of others. Each entry in the fix file states its evidence.

One unit was changed: actinium (08a91e70-3ddc-11dd-91b3-0050c2490048) ships in MJ because EF points it at the calorific-value property by a one-digit typo in the property's identifier, and is published in kg, the unit ecoinvent gives the same resource; nothing characterises the flow, so no score moves.