Docs / What was changed in each source

Changes to ecoinvent 3.8

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

ecoinvent 3.8 ships 4,421 elementary exchanges. On the 2026-08-29 build of commit 54b0f9d it was merged second, after ecoinvent 3.12: 4,406 of its rows landed on a consensus flow that already existed (196 of them through ecoinvent's own correspondence table), 13 added a new compartment under a substance the list already had, and 2 minted a new substance. The curated corrections to the list are in ecoinvent-3.8-manual-fixes.json (27 entries), the curated targets in ecoinvent-match-overrides.json, and the identities of its dissolved-metal rows come from ecoinvent-canonical-identities.json.

Most of what is on this page comes from one habit of the 3.8 release: it names a metal in water, air or soil as the bare element — Lead, Aluminium, Copper — and gives it the element's registry number, while every release from 3.9.1 onwards names the same row, under the same uuid, Lead II, Aluminium III or Copper ion. ecoinvent means one flow by one uuid, so the newest release's name is taken as the vendor's current opinion of what the row is, and 3.8's row is published under that identity. Nothing 3.8 wrote is altered on the row itself; the newer name and number are what the matching reads, and the row's provenance keeps 3.8's own words.

One name, two substances

On the 2026-08-29 build of 54b0f9d seven names in 3.8 landed on more than one substance. Two of those — Molybdenum and Carbon — were defects and have been corrected (#828, #827); on the build of 2026-08-30 at 9fbb0d5 five remain, and in each either the registry numbers 3.8 wrote differ, or the newest release names some rows and no longer carries the others.

Sodium, eleven rows. Eight — the four air compartments and the four soil compartments, 235b2f02-ff98-4da9-82e2-e4246c7b7990 (air, unspecified) among them — are Sodium I with the ion's number 17341-25-2 in 3.12 and are published as Sodium(1+). The other three, 460716f9-f5b8-4ee0-8198-545b24822124 (air, lower stratosphere + upper troposphere), 73a1b292-e31d-4100-a413-304795a3656f (water, ground-) and 723f56f9-c68e-5692-a969-0fb2a70bce20 (water, unspecified), are absent from 3.12, so nothing overrides 3.8's element number 7440-23-5, and they are published as Sodium, the metal.

Iron, ten rows. Nine air and soil rows are Iron ion in 3.12, which gives them no registry number, and they are published as Iron, Ion — an iron ion of unstated charge. The tenth, e3043a7f-5347-4c7b-89ee-93f11b2f6d9b (water, ground-), is absent from 3.12, keeps 7439-89-6 and is published as Iron, the element.

Calcium, nine rows. Eight are Calcium II (14127-61-8) in 3.12 and are published as Calcium(2+). b24b66e6-633c-49bf-a6ce-1e81da0715ce (air, lower stratosphere + upper troposphere) is absent from 3.12 and stays Calcium.

Molybdenum, fourteen rows — split on the 2026-08-29 build, and since corrected. Thirteen rows are Molybdenum VI in 3.12, but 3.12 left the element's number 7439-98-7 on them (3.9.1 had the ion's, 16065-87-5), and a CAS number outranks a name (see When a name and a CAS number disagree), so all thirteen were published as Molybdenum. The one row 3.12 no longer carries, 697e8ecf-bdbe-4857-92c9-e01960c91db7 (air, lower stratosphere + upper troposphere), was decided by 3.9.1's number and alone was published as Molybdenum(6+). The ion's number is now written on the name in the 3.10.1, 3.11 and 3.12 fixes files, and the newest-release identities are derived from each release with its fixes applied — otherwise 3.12's raw number would have been substituted back over the correction — so on the build of 2026-08-30 at 9fbb0d5 all fourteen 3.8 rows are published as Molybdenum(6+) (#828).

Carbon, four rows in soil — split on the 2026-08-29 build, and since corrected, all carrying 7440-44-0. ecoinvent's own correspondence table sends 14ea575b-5caa-4958-acf7-0bcc47f9cadf (soil, unspecified) to EF 3.1's elemental carbon, where it joins 3.8's own Elemental carbon row. The agricultural, forestry and industrial rows (62859da4-f3c5-417b-a575-8b00d8d658b1, b72e46aa-9034-4dd2-856d-a43241d3b1f4, 7f8fd1ca-0412-4b2e-90fd-a9d294d947a3) had no table row and were matched on the CAS number and the name, which reached the consensus substance Carbon. The curated target now covers those three uuids too — the same ones 3.12 ships — so they are published as Elemental Carbon, on EF's soil, unspecified flow, the only soil flow EF publishes for it (#827).

Water, eleven rows. The five air rows (075e433b-4be4-448e-9510-9a5029c1ce94, air unspecified, and four others) are published as Water vapour; the six water rows (2404b41a-2eed-4e9d-8ab6-783946fdf5d6, water unspecified, and five others) as Water. The compartment decides.

Water, unspecified natural origin, three resource rows. The sub-compartment decides: 831f249e-53f2-49cf-a93c-7cee105f048e (in water) is Water, 478e8437-1c21-4032-8438-872a6b5ddcdf (in ground) is Groundwater, and 2caa889e-8187-459d-963a-fa47a79c5378 (fossil well) is Fossil groundwater.

An ionic charge was added or corrected

Twenty-three bare element names published with a charge, 243 rows. In each case 3.8 wrote the element's name and the element's number, and the newest release that carries the row writes the ion. The consensus list follows the newest release and publishes the ion; the charge comes from the newer release's registry number. For strontium and caesium the newest release states the charge in the name but still writes the element's number, which ecoinvent-3.12-manual-fixes.json corrects (#934); the corrected identity is what 3.8's rows follow.

The rows are the air and soil compartments, and for the elements with fewer than 13 rows the missing compartments are ones 3.8 files under a different name (Cadmium, ion in water) or ones 3.12 no longer carries. The 14th row, where there is one, is air / lower stratosphere + upper troposphere, which 3.10.1 dropped and 3.9.1 decides. Chromium VI (18540-29-9, 14 rows) is published as Chromium(6+), which is the same statement in a different notation and is not a change.

Ten X, ion names in water given a charge, 50 rows. 3.8 names the dissolved form in the five water compartments Cadmium, ion, Calcium, ion, Chromium, ion, Lithium, ion, Nickel, ion, Potassium, ion, Silver, ion, Sodium, ion, Vanadium, ion and Zinc, ion, and on every one of them 3.8 itself wrote the specific ion's registry number — 22537-48-0 for cadmium, 14127-61-8 for calcium, and so on down the table above. The number states the charge the name left out, and the rows are published as Cadmium(2+), Calcium(2+), Chromium(3+), Lithium(1+), Nickel(2+), Potassium(1+), Silver(1+), Sodium(1+), Vanadium(5+) and Zinc(2+). Vanadium, ion is the one where 3.8's own CAS number (22541-77-1) differs from the newest release's (22537-31-1, Vanadium V); both are vanadium ions and the newer one is followed.

Six bare element names published as an ion of unstated charge, 59 rows. Antimony, Arsenic, Copper, Iron, Tin and Titanium in the air and soil compartments carry the element's number in 3.8. From 3.9.1 onwards ecoinvent names them Antimony ion, Arsenic ion, Copper ion, Iron ion, Tin ion and Titanium ion and deletes the CAS number. The vendor deliberately declined to say which ion, so the consensus list does the same and publishes Antimony, Ion, Arsenic, Ion, Copper, Ion, Iron, Ion, Tin, Ion and Titanium, ion; picking a charge would be a reinterpretation the vendor refused.

Six X, ion names whose registry number is a neutral or polyatomic species, 19 rows. These are not charge corrections. The number 3.8 wrote settles what the row is:

The last row is worth knowing about: an inventory that reports acrylate in water under 3.8's name will find it on the acid's flow, because that is the substance 3.8's registry number designates.

Rows merged into one flow

77 consensus flows carry more than one 3.8 row. They fall into five groups.

Ore-grade resource names

ecoinvent 3.8 names a mined metal once plainly (Copper, in ground) and once per deposit it has data for (Copper, 0.52% in sulfide, Cu 0.27% and Mo 8.2E-3% in crude ore, in ground). An ore grade describes the deposit, and a resource flow is the metal taken out of it, so every one of these lands on the plain row's substance and the grade is kept as nothing more than the source row's own name. 42 substances, 146 rows in all:

The titanium group deserves a second look. Titanium, in ground (2f033407-6060-4e1e-868c-9f362d10fdb2) carries the element's number 7440-32-6; the three rows merged with it are titanium dioxide ores — TiO2, 54% in ilmenite, 18% in crude ore, in ground (90a94ea5-bca4-483d-a591-2e886c0ff47f) and TiO2, 54% in ilmenite, 2.6% in crude ore, in ground (78cd4852-e7b9-4301-adf7-51e730b0356a), both 13463-67-7, and TiO2, 95% in rutile, 0.40% in crude ore, in ground (ec0fa5ce-51b4-4792-a8e8-c4ee668eddc3), 1317-80-2, which Common Chemistry names Rutile. They land on Titanium because ecoinvent's own correspondence table sends them there — and they are converted on the way: a kilogram of the oxide is 0.599 kg of titanium, and ecoinvent-match-overrides.json has carried that factor on all three uuids since the ruling in #596. The published source reference carries it as the conversion multiplier. BAFU's two rows of the same ores take the same factor, and an expectation now holds all five rows to it (#829, which was opened on a misreading and closed against that check).

[Deleted] tombstones beside the live row

3.8 carries three rows whose name begins [Deleted] and which sit beside a live row in the same compartment, with the same unit and the same registry number. ecoinvent's own chain of correspondence tables (3.8 to 3.9 to EF 3.1) deletes them outright; a curated target in ecoinvent-match-overrides.json sends each one where its sibling goes instead, so an inventory still using the old row is not dropped.

Two spellings of one substance in one compartment

All in air / unspecified unless stated; both rows carry the same registry number except where noted.

An ion beside its own element or its own , ion row

In water / ground- and water / unspecified, 3.8 sometimes carries a row under the bare name and another under X, ion, with the same ion registry number on both, so they were one substance in the vendor's own terms:

Everything else

Registry numbers corrected

Fourteen entries in ecoinvent-3.8-manual-fixes.json change a registry number, and they are applied before any matching runs. Nine of them make 3.8 say what a later release already says about the same uuid: the vendor corrected its own CAS number, and the older export is read as meaning what the newer one states.

Other corrections in the same file

These change something other than a registry number, and are listed because a reader looking for the row would otherwise find a different name or unit.