Download the open data

A whole library, in one file, in the shared schema every library here uses. Two ways in: the Open data page compares the libraries and hands you the file, and each library's own Open data section goes into detail on that one library.

Every library at once#

The Open data page is one card per library, with the edition, the count of factors and values, the audit result, the credit line behind Show the credit line, and Download.

Two ways to narrow it:

• What I need to do. Tick "Share with my customers", "Put in a product I sell" or "Modify and republish", and the list keeps only the libraries whose licence allows it. This is the fastest way to settle a licence question before you commit to a library, and the count above the list says how many are left.
• The filters: home region, library kind, licence, the language of the labels and the file format. The search box takes a library, a publisher or a region, and the sort switches between most factors, name and most recently loaded.

Each card carries the same five rights as the rights card, shortened to chips: Use, Publish, Share with customers, Put in a product, Modify. A right the licence does not grant is struck through rather than dropped, and a right that hangs on a condition says so, for example "if you credit the publisher". The chips are the short form of the card, read from the same facts.

Note. Looking for one factor rather than a whole library? Use factor search instead. The Open data page says so too, with Search factors.

One library's own section#

1. Go to Libraries and click a library.
2. Click Open data in the tab row, just after Overview.

The tabs are an index of one long page: you can also scroll down to the section. The section reads top to bottom in four parts: the files, what was standardised, the audit checks and your rights.

Several parts start folded, so the section stays short. Click a folded title to open it. Folded content is still part of the page, so your browser's find (Ctrl+F) and search engines find it.

Download a file#

The files block is titled with the library's name followed by "standardised by open-climate.ai", for example "ADEME Base Carbone, standardised by open-climate.ai". Under the title, one line says what that means: "The publisher's file, re-cut into the schema every library on open-climate.ai shares: one category axis, coded regions, canonical units, AR6 warming potentials, the source row kept beside each value."

So the file is not the publisher's original. It holds the same factors, in the columns and codes every library shares, with the publisher's own row kept for each value so you can check it against the source. What open-climate.ai standardised explains each change.

One row per file of the edition. The row names the file, badges its format, and then gives what the build recorded: the revision when there has been more than one, the number of rows, the date it was built, and the first characters of its sha256 checksum, so you can confirm the file you hold is the one published. A file that is announced but not built yet says "not built yet".

To download, from either place:

1. Click Download.
2. Without an account, the button asks you to create one, naming the file you asked for. The file itself is free; the account is what records who accepted the data terms, and it is how you find your downloads again.
3. Signed in, the download dialog opens. Pick the format, tick "I am aware I need to credit the publisher and open-climate.ai", and download.

The dialog is the same one wherever a file leaves, and it also holds the credit you owe: what to do as a consultant or a corporate, which is to cite the source in the reports and inventories you hand over; what to do as a software vendor, which is to keep the publisher's credit in your product and link back to open-climate.ai; the line itself, with Copy; and Read the full data terms. You accept the terms once and confirm the credit each time.

Note. The credit line is also written inside the file, on the references sheet, so it travels with the data when the file is passed on.

The file holds the whole edition. To take away only some factors, with the columns you choose, use Export to a spreadsheet instead.

When there is no download#

Two states look alike on the card, and only one of them is about the licence:

• "Searchable on the portal · values withheld", with a No download tag where the button would be. The publisher's terms do not allow redistribution, so the activities stay searchable, the values are withheld, and there is no file. Those rows are also left out of an export. A withheld value is unknown, not zero.
• "No file yet": the library is searchable on the portal and served through the API, and a downloadable workbook has not been built for it.

Either way the rights are shown, because what you may do with the publisher's own data does not depend on whether we host a copy of it. See licence tier and the four tiers.

What open-climate.ai standardised#

Under What open-climate.ai standardised, one tile per theme says what was done to the publisher's data so it can be searched and compared with every other library. Each tile counts what it covers in this library, in factor values, parts or parent factors.

The first four tiles are always shown: One category axis, Coded regions, Canonical units and Warming potentials. The fold "N more things standardised" holds the others. A theme appears only when the library has something for it: a library with no spend factors shows no Spend factors tile.

Each tile ends with two links:

• Learn more opens the docs page for that theme. What open-climate.ai standardised is the whole list in one table, with what each theme means.
• N checks behind this opens the audit checks at that theme's group, so you see the checks that test the tile's claim.

A tile counts what the library holds on its default slice: one value per factor, on the default indicator and the latest year. See top-level factor and sums to parent for how parts and parents relate.

How the data is checked#

The fold Audit checks lists the checks run on the library's data after it is loaded. It starts closed.

Read the summary#

Before you open the fold, its title line sums up the latest audit run:

• how many checks passed;
• how many known limitations were found;
• how many checks failed;
• how many checks are enforced when loaded;
• the date of the last run.

A check counts once in the summary, whichever group it sits in.

What each status means#

A known limitation tells you where the data is weaker. Read it before you rely on that part of the library: for example, a limitation on warming potentials matters if you compare values across GWP reports (why some rows stay unconverted), and one on units matters if you convert quantities (the conversions that are refused).

Failed does not mean anything was withheld. A check reports on the library, it does not gate it: no row leaves factor search, the API or a file because a check failed, and none is put back because one passed. The difference from a known limitation is what kind of rule the check states: a gating check states a hard rule, so a violation is a defect for us to fix, while an advisory one flags rows for review and expects to find some. The status that does keep data out of factor search is Enforced when loaded, and it acts before the load rather than after it.

Read that narrowly, in both directions. A flagged row is not necessarily a row you can see: some checks test parts of a library factor search does not serve, such as values from an edition a newer one superseded. And no audit result overrides a licence: where the terms forbid redistribution the values stay withheld and out of an export whatever the checks say, which is When there is no download.

The line gives you a count and not the rows, so the page cannot tell you whether the factor you are about to use is one of the violations. Read why the check matters for what a violation would do to a figure, and if it bears on your number, treat that part of the library with care. Rows that are missing rather than flagged are a different question: see See what was not imported, and why.

How the checks are grouped#

Inside the fold, the checks come in this order:

1. Needs attention: the failed checks first, then the known limitations. When nothing needs attention, this group is absent.
2. One group per theme, with the same names and in the same order as the tiles. The N checks behind this link on a tile opens the fold at its group.
3. Other checks: checks tied to no theme shown for this library.

Each line gives a status icon, the check's plain name (such as "Every region is a known code"), an info mark, the status, and the count in this library.

Why a check matters#

Hover over the info mark next to a check's name, give it keyboard focus, or tap it, to read why the check matters: what would go wrong in your figures if the check failed. On a narrow screen the text opens under the line.

In this library and across all libraries#

A line that found something shows "N in this library": the violations that concern this library. A passed check, a load guard and a check no run has recorded yet show the status alone.

Some checks run across all libraries at once, for example that a unit label means the same in every library. A line for such a check also shows, in grey, the same check's count across every library. A violation that involves several libraries counts on each library it touches, so the per-library counts of one check can add up to more than the count across all libraries.

Checks not listed here#

Two kinds of check are not listed:

• checks that only record a figure per library rather than pass or fail;
• checks that judge the shared vocabulary, not one library's data. See vocabularies are closed and documented.

The promises these checks back, such as nothing dropped and no silent rescaling, are set out in Guarantees you can rely on. See what was not imported, and why lists the rows a library left out.

What can I do with this data?#

What can I do with this data? is a question at the end of the section. Click it to open the answer. A link that points to it, such as the one on the library's Overview, opens it for you.

Inside, one Your rights card per licence answers five questions:

• "Use it to calculate my own footprint"
• "Publish my footprint results"
• "Share the file with my customers"
• "Put the factors in a product I sell"
• "Modify it or combine it with my own data"

Each answer is one of:

• Yes.
• No.
• Yes, if, followed by the conditions, such as "you credit the publisher" or "you share the result under the same licence".
• Ask followed by who holds the rights, as a link when a contact is known.
• Unclear: ask followed by who holds the rights, when no written grant settles the question.

Under the answers, the card says what the answer rests on (for example "Named licence" or "Publisher's published terms") and the date it was checked. Then come:

• the line "Not legal advice: the publisher's terms prevail.";
• Full terms, which opens the publisher's licence text;
• the Credit line, with Copy credit line. It credits the publisher first, then open-climate.ai with the library page's address. Print it with any values you reuse; Cite a factor shows where it goes.

For example, the ADEME Base Carbone card answers "Yes" to your own footprint and "Yes, if you credit the publisher" to the four other questions, on the basis of a named licence, etalab-2.0. The AIB card answers "Unclear" for sharing the file with customers and for putting the factors in a product you sell, and unclear is not permission.

To compare libraries on these answers instead of reading them one at a time, use the What I need to do filter on the Open data page. Five questions per licence explains how each answer is set, and Choosing a library puts the licence in the order you pick a library. See also licence tier and crediting the publisher.

Next#

• Cite a factor: the citation and the credit line.
• Export to a spreadsheet: only the factors you filtered.
• Choosing a library: the licence and the coverage, in the order you pick a library.
• Guarantees you can rely on: the promises the audit checks back.