Docs / What was changed in each source

Changes to BAFU 2026 v1

Part of What was changed in each source. What is still open about BAFU is on the issue tracker under the label bafu.

BAFU 2026 v1 arrives as an ecoSpold v1 archive, from which 2,679 elementary flows are extracted, and it is merged third — after ecoinvent 3.12 and ecoinvent 3.8 — on the build of 2026-08-29 from commit 54b0f9d. It ships no identifier of its own: a row is a name, a compartment, a sub-compartment and a unit, and the same substance is written once per sub-compartment. The uuids this project gives BAFU rows are derived from those four fields (and a country suffix where one is present), so this page identifies every row by name, original compartment and unit rather than by uuid.

Corrections to what BAFU shipped are in bafu-2026-v1-manual-fixes.json (71 entries: 47 registry numbers, 9 synonyms, 7 units, 6 names, 2 compartments) and, because BAFU inherits ecoinvent 2's spellings and habits through SimaPro, in simapro-lineage-manual-fixes.json (8 entries, applied to BAFU and to Stepwise 2006 alike). Where a row belongs is stated by hand for 29 rows in bafu-2026-v1-match-overrides.json. On the build, 2,564 rows matched by rule, 23 by a curated target, and 92 became new substances.

One name, two substances

On the 2026-08-29 build of 54b0f9d five BAFU names ended up on two different substances. Two of those were defects, and both were corrected the next day (#824, #830); on the build of 2026-08-30 at 9fbb0d5 three names remain split, and in each the split is forced by BAFU's own data — a registry number on one row that the others lack, or a compartment that changes what the substance physically is. All five are described here, the two corrected ones in the past tense, so that a reader of either build finds the row.

Aluminium — BAFU writes the name on 14 rows. The one in emissions to water / unspecified carries 22537-23-1, which is the aluminium ion's number, so it is published as Aluminium(3+). The resources / in ground row carries 7429-90-5, the metal's number, and the other twelve carry none and are matched on the name; all thirteen are Aluminium. The number outranks the name (When a name and a CAS number disagree), and it was BAFU's own CAS number.

Manganese — the same shape. The emissions to air / unspecified row carries 16397-91-4, the manganese(2+) ion's number, and is published as Manganese(2+); the other thirteen rows carry the metal's 7439-96-5 or nothing, and are Manganese.

Copper ion — three rows, and a split that has been corrected. The one in emissions to air / high. pop. carried 7440-50-8, which is the CAS number of copper the metal, and on the 2026-08-29 build it landed on Copper beside BAFU's own plain Copper row in that compartment, while the two rows without a CAS number — emissions to air / unspecified and emissions to water / unspecified — were matched on the name and landed on Copper, Ion, the substance for copper dissolved with its charge unstated. The metal's number is BAFU's own: the archive writes it on all 38 exchanges of that row. The name is the half that is right — BAFU already lists plain Copper under 7440-50-8 in this very compartment — so the number is removed from the row in bafu-2026-v1-manual-fixes.json, and on the build of 2026-08-30 at 9fbb0d5 all three Copper ion rows land on Copper, Ion (#830). Worth reporting to BAFU.

Water — 169 rows, more than any other BAFU name (the count includes the country-suffixed copies described under Rows merged into one flow). Water released to air is water vapour, and the 55 rows in the four air sub-compartments are published on Water vapour; the 114 rows released to water — river, ocean, groundwater, fossil water and unspecified — are published on Water. Same name, two physical states, and the compartment is what says which.

Carbon monoxide, biogenic — three rows, and a defect that has been corrected. The high. pop. and unspecified air rows carry 630-08-0 and land on Carbon Monoxide (biogenic), the substance EF 3.1 characterises. The low. pop. row ships with no CAS number at all — none of the 117 exchanges of it in BAFU's archive writes one — so on the 2026-08-29 build its name alone reached no existing substance and it minted a second one, Carbon Monoxide, Biogenic, with one flow and no characterisation factors. It was the same pattern Carbon dioxide, biogenic had until its missing number was supplied, and it got the same fix: 630-08-0 is written on the name in bafu-2026-v1-manual-fixes.json, and on the build of 2026-08-30 at 9fbb0d5 all three rows land together (#824).

One more correction belongs beside these, although no BAFU name was split by it: BAFU's five Hydrocarbons, unspecified rows (emissions to air / unspecified; emissions to water / unspecified, ocean, lake and river) carry no number and differ from EF 3.1's hydrocarbons (unspecified) by a comma and a pair of brackets, and on the 2026-08-29 build they landed on a duplicate bucket two ecoinvent rows had minted. Five curated targets in bafu-2026-v1-match-overrides.json now send them to EF's own substance; the lake and river rows keep their water body, which EF does not have, and the bucket gets a flow there (#825).

An ionic charge was added or corrected

BAFU writes a charge as a roman numeral — Chromium VI, Cadmium II, Tin (II) — and the list writes the same charge as Chromium(6+), Cadmium(2+), Tin(2+). That is notation, and on the build it covers 11 names and 26 rows; none of it is listed here. Three names did gain a charge their spelling did not state.

Two names lose the word ion on publication without any change of meaning: Ammonium, ion (six water rows) is published as Ammonium and Perchlorate, ion (one river row) as Perchlorate, because ammonium and perchlorate are ions by definition and the word adds nothing.

Six of BAFU's roman-numeral rows shipped with the wrong number or none, and their charge was confirmed rather than added by writing the ion's number on the row; those are under Registry numbers corrected below.

Rows merged into one flow

On the build, 237 consensus flows carry more than one BAFU row. They fall into a small number of patterns.

Country suffixes

BAFU appends a country or region code to the name of a flow when a dataset reports it for a place: Water, AT, Water, FR, Water, unspecified natural origin, CN. The suffix says where the water was used, which the list records nowhere in a flow — a flow is a substance and a compartment — so the suffix is stripped and the row lands with its unsuffixed sibling. 261 rows carry a suffix, and 87 distinct codes are seen: the ISO-2 countries from AE to ZA, and the regions Europe, GLO, OECD, RAF and RER. The names it happens to:

The plain Water rows, by where they land:

The single kilogram row in each water compartment was converted to cubic metres at 1 kg = 0.001 m³ (bafu-2026-v1-manual-fixes.json, #409), so that it lands beside its 53 suffixed siblings rather than on a flow of its own; the air rows stay in kilograms, which is how EF 3.1 states water vapour.

One nuclide, two units

BAFU reports the same radionuclide in the same compartment twice, once in Bq and once in kBq. Both rows land on one consensus flow, and the mapping back to the kBq row carries the factor 1 kBq = 1,000 Bq. This happens 159 times.

Silver-110 is published as Silver-110m: the 24-second ground state is never what an effluent inventory reports, and the 250-day isomer is (see the EF 3.1 page). BAFU's two Uranium-238 rows in Bq carry 7440-61-1, uranium the element's number, and reach the isotope only because that CAS number is corrected — see Registry numbers corrected.

Resource corrections

BAFU carries eleven flows whose name ends , resource correction. A dataset built with recycled material books a negative amount of the primary resource charged upstream — BAFU's own note on one dataset reads "Resource correction added for share of primary aluminium" — so the correction and the extraction are one quantity stated in two places, and only their sum is the extraction that happened. Nine of the eleven carry the base substance's own registry number and land with it on that alone; Gravel and Energy, gross calorific value, in biomass carry none and were given the base name as a synonym so that they land the same way (#386).

Ore-grade names

Like ecoinvent 3.8, BAFU names a metal resource once per deposit — Copper, 0.52% in sulfide, Cu 0.27% and Mo 8.2E-3% in crude ore. The percentages are a property of the ore body and the metal is the same, so all of them land on the metal. All rows are resources / in ground, kg.

Energy carriers by mass and by energy

BAFU ships each fossil and nuclear fuel twice: as a mass (Coal, hard, kg) and as an energy content (Energy, from coal, MJ). EF 3.1 accounts these resources by energy, so both rows land on the MJ flow, the energy row as it is and the mass row with a stated heating value. Measured over all 11,947 datasets in the release, the mass row and the energy row of a fuel never appear in the same dataset — Energy, from coal in 14 datasets, Coal, hard in 139, none carrying both — so nothing is counted twice by sharing a flow (#70). All are resources / in ground unless stated.

Two spellings of one thing

One flow filed in several compartments

BAFU files some resources under more than one sub-compartment, and the consensus context vocabulary sends them to one place, so the rows land together. Nothing about the substance changes.

Water withdrawals

92 BAFU rows land on one flow, Water in Resource → Water. Four names are involved and two of them are curated.

The other water withdrawals merge only their country copies: Water, cooling, unspecified natural origin (8 rows → Cooling water; one RER row is filed under resources / in ground), Water, well (6 rows in three compartments → Groundwater), Water, river (4 → River water), Water, lake (2 → Lake water).

Registry numbers corrected

Every registry-number entry in bafu-2026-v1-manual-fixes.json and simapro-lineage-manual-fixes.json is listed here.

Numbers BAFU wrote that name the wrong substance

Rows that carried two CAS numbers, one of them the ion's

BAFU's archive does not number a flow once; every dataset writes its own CASNumber, and the extraction keeps all of them. On ten rows the datasets disagree — most write the metal's number and a few write the ion's — and the merge reads the first number, which sorts the ion's five-digit number ahead of the metal's. Those rows reached the ion while the dozen sibling rows of the same name reached the metal, so one emission of cadmium to air was two substances depending on the sub-compartment. The ion's number was removed, leaving what most of BAFU's datasets wrote (#792).

Numbers removed

Numbers supplied where BAFU shipped none

Nineteen rows name, under an ordinary second name, a substance the list already publishes — Arsine for hydrogen arsenide, Methyl mercaptan for methanethiol — and no rule can connect two spellings that share no pattern. Without a CAS number each minted a second substance with no characterisation factors (#512).

Six more rows spell a charge or a mineral that the list already publishes under a CAS number, and shipped without one (#617, #597). No BAFU row reaches a charged form on its spelling alone, so each of these minted a duplicate beside the substance it names.

And three biogenic carbon rows were missing the CAS number their same-named siblings carry (#661, #663, #824):

The other corrections in the fix files

Names. Energy, waste heat, air (emissions to air / unspecified, MJ) is renamed Heat, waste, BAFU's own name for the same release on fourteen other rows. Gas, natural/m3 and Gas, mine, off-gas, process, coal mining/m3, both with the unit field reading Nm3, lose the /m3 from the name: the unit field is believed. Wood, unspecified, standing/kg and Water, process, unspecified natural origin/kg lose their suffix after the unit conversions below, and Water, process, unspecified natural origin/m3 in resources / in water becomes Water, unspecified natural origin, which is BAFU's own name for the same water on 78 other rows. From the lineage file, Uranium, 451 GJ per kg, Uranium, 560 GJ per kg and Uranium, 2291 GJ per kg become Uranium, and Carbon dioxide, in air with 124-38-9 becomes Carbon dioxide, biogenic, in air, so that the CAS number is read beside an origin. Every renamed row keeps its shipped spelling as a synonym for matching.

Units. Seven kg rows of liquid water become m³ at 1 kg = 0.001 m³: Water (four water compartments), Waste water (ocean, river), Chemically polluted water (river), Water, process, surface, Water, embodied in product and Water, process, unspecified natural origin (three compartments). Wood, unspecified, standing/kg becomes m³ at 0.00204 m³/kg. The three Uranium, N GJ per kg rows become MJ at the factor their own name states — 451,000, 560,000 and 2,291,000 MJ/kg.

Compartments. The two Carbon dioxide, in air rows filed under resources / biotic and resources / unspecified move to resources / in air: the name says where the carbon comes from and the compartment contradicted it.

Synonyms. COD, Chemical Oxygen Demand, BOD5, Biological Oxygen Demand and TOC, Total Organic Carbon are given EF 3.1's shorter names as synonyms so that they land on EF's flows, exactly as ecoinvent's identical strings already do through its correspondence table; Energy, gross calorific value, in biomass and its resource correction are given Biomass, and Gravel, resource correction is given Gravel, for the reasons above.