Docs / What was changed in each source

What was changed in each source

Every list that goes into the consensus flow list is taken as its publisher shipped it, and then some of its rows are altered before, or while, they are matched. Most of those alterations are housekeeping — a name is title-cased, a unit symbol is spelled the standard way, a compartment is translated into the shared vocabulary — and none of that is recorded here. These pages record the changes that alter what a row says: the ones a reader who knows the source list would want to check.

One page per input list, the five ecoinvent releases by version number. Which of them the standard build merges, and in what order, is the Role column — 3.12 is merged before 3.8, and the three releases between them are curated but not merged:

The four kinds of change these pages show

One name, two substances. The source uses one name for rows that the consensus list places on two different substances. Usually a registry number forces the split — EF 3.1 writes 1,3-benzenediamine on twenty-six rows, thirteen with the CAS number of the free base and thirteen with the CAS number of its hydrochloride salt; ecoinvent 3.8 writes Sodium on eleven rows that its own later releases rename, eight of them to the sodium ion and three left as the metal — and sometimes the compartment does: Water released to air is water vapour, and Water released to a river is not. The page says which rows went where, and why.

An ionic charge was added or corrected. A dissolved metal is an ion, and the source either named it as the bare element, wrote , ion without saying which one, or wrote a charge that its own registry number contradicts. The page lists the rows whose published name carries a charge the source's name did not, and where the charge came from. Writing Chromium VI as Chromium(6+) is a change of notation only, and is not listed.

Rows merged into one flow. Several rows of the source land on one consensus flow — two spellings of a name, an ore-grade suffix on a resource, a [Deleted] tombstone beside the live row, a place name on the end of a BAFU flow — so a reader following one of them back to the source will find more than one row there. The page groups them by substance and lists every source row.

A registry number was corrected. The source ships a CAS or EC number that names the wrong substance, or a wrong number in the sense that it belongs to a mixture rather than its constituent, or no CAS number at all where the substance has one; the correction is written down before anything reads the row, with the reason. Normalising a CAS number's formatting — stripping leading zeros, checking the check digit — is not a correction and is not listed.

What is deliberately not on these pages

• Case and spelling. nitrogen dioxide becomes Nitrogen Dioxide; Naphtalene lands on Naphthalene. A misspelling that reaches the right substance is listed under rows merged if it sat beside a correctly spelled row, and nowhere otherwise.
• Units. kilogram and kg are one unit. A row whose unit was actually changed — BAFU's Energy, from coal in MJ landing on Hard Coal, which EF 3.1 also accounts in MJ — appears under rows merged because the rows merged, and the conversion is stated there.
• Contexts. Every source's compartment vocabulary is mapped onto the consensus one, and that mapping is described in Flow contexts rather than here.
• Names that lost the contest for the published label. When ecoinvent's Ethene and EF's Ethylene are one substance, one name is the label and the other is a synonym. That is a naming decision, described in Names that belong to something else, and it does not alter what the source row says.

Where the evidence on these pages comes from

Two places, and each page says which.

The curated corrections are in the source's own fix file, bundled with the package under src/brightway_flows/data/ — ef-3.1-manual-fixes.json, ecoinvent-3.8-manual-fixes.json, bafu-2026-v1-manual-fixes.json, and so on. Every entry there carries the reason in full; the pages here condense it. A test (tests/test_documentation.py) checks that every registry-number correction in a fix file is on that source's page, so the two cannot drift apart silently.

The merge outcomes — which rows landed together, and which rows of one name were split — are read off a build, and a build's numbers change when a rule changes. Unless a page says otherwise, the build is the one made on 2026-08-29 from commit 54b0f9d, which merged ecoinvent 3.12, ecoinvent 3.8, BAFU 2026 v1 and Stepwise 2006, in that order, into the transformed EF 3.1. Writing these pages against that build turned up eight inconsistencies — duplicate substances minted for one BAFU and two ecoinvent names, an ion published as its element on three lists, one mistaken reading — and they were filed as #824 to #831 and fixed the next day. Where a page describes one of them it gives both states, and the corrected state is the build of 2026-08-30 at 9fbb0d5, on which 43 rows are decided differently from 54b0f9d, every one of them a row those issues name.

What is still open

These pages record what was changed. What has not been settled for a list is on the issue tracker, where every open issue carries the list it is about: EF 3.1, ecoinvent, BAFU, Stepwise 2006, AGRIBALYSE 3.2, GreenDelta's EF 3.1 for openLCA. The full set of labels, by topic and by who has to answer, is listed under Known limitations.

Finding the original row

• EF 3.1 and ecoinvent ship a uuid on every row, and the pages quote it. An ecoinvent uuid is the same across releases, so a row quoted on the 3.8 page can be looked up in 3.12's ElementaryExchanges.xml as well.
• BAFU 2026 v1 and Stepwise 2006 ship no identifier. The uuids this project gives their rows are derived from the name, the compartment, the sub-compartment and the unit as the vendor wrote them — and, for BAFU, the country suffix — so they are stable across builds but are not the vendor's. The pages quote the name, the original compartment and the unit instead, which is what a reader needs to find the row in the ecoSpold archive or the SimaPro method file.