open-climate.aiDocs

Report a wrong value or a newer edition

Numbers checked against the live data on

Every library page carries a report form at its foot, and that is the route for both of the things a reader usually finds: a served value that looks wrong, and an edition the publisher has released that we do not serve yet. It needs no account, it takes a minute, and a reply comes back to the address you leave.

What is worth reporting#

Three things, and they all go through the same form.

  • A value you believe is wrong. It differs from the figure the publisher prints, or its unit, region, year or boundary does not describe what the publisher published.
  • An edition newer than the one served. You have seen a release on the publisher's own site that the library page does not list as Served.
  • A library that is missing altogether. That one is the Libraries page rather than a library page: Missing a library? says where.

Check these six first#

Most surprises in factor search are deliberate, and each of them is written down. Reading the right one first saves you a report and tells you what to do instead.

What you are looking at What it means
An empty value The publisher gives no figure, or the value is waiting for the publisher's approval. It is not zero, and it is never filled in from somewhere else. See the value on a factor page
A value withheld, with a dagger The publisher's terms grant no right to republish the number. See when there is no download
A unit factor search will not convert into yours A conversion that is not a fixed ratio is refused rather than guessed. See units
An activity that is not here at all A row can be left out on purpose, with the reason recorded. See what was not imported
A served number that differs from the publisher's own Usually a recorded normalisation, not an error. See how factors are cleaned and trace a value to its source
A number that moved since you used it A new edition was loaded, even though the Dataset date on the page did not move. See what a refresh changes and what the Dataset date is

One more is worth its own line, because it is the report we receive most often: a newer edition the library page already knows about. The line above its Versions table carries the verdict of the last update check, and how a newer edition is looked for reads all four. One library sits on "something newer, which we are reviewing" permanently and by design, for a reason that is not a lag: how often the data changes names it and says why.

If the page already says what you were about to tell us, there is nothing to send.

Where the form is#

On a library page, the Last loaded card in the stat row at the top carries a link, Report a newer version. It opens the card at the foot of the same page, headed Know a newer edition?, with a button Report it. On the Libraries page the same card reads Missing a library? instead, and the link to it sits in the page header as Ask for a new library.

The form is anonymous and short.

Field What to put in it
Library Prefilled on a library page, which also states the edition we serve. On the Libraries page, name the library or its publisher
Version you saw Library pages only: the publisher's own label for it, such as v26.0 or 2026. The Libraries page form does not carry this field
Where you saw it A web address starting with http:// or https://
Your email Optional, and the only way to get a reply
Anything else The note. It stops at 4,000 characters, and each of the other boxes at 300, so you cannot type or paste past either

Naming the library is required, and so is at least one of the rest. On a library page that means the version, the address or the note; on the Libraries page, where there is no version field, the address or the note. Everything else is optional.

For a wrong value, the note is where the report actually lives. Put three things in it, and it can be checked in one sitting:

  1. The factor's address, which is the page address you are reading it on.
  2. The value you read here, with its unit.
  3. The publisher's own figure, and what you are comparing against. Put that source in Where you saw it: a report with a link is checkable, one without it is a search.

What happens next#

The form answers Report received with a reference, "report" and an id. Keep it if you expect to follow up, because it names your report exactly.

The report is stored and reaches the people who maintain the library. If you left an email, the reply goes there.

A correction does not get patched into the served value in place. It arrives the way every other change to a library arrives, as a load of a new or corrected edition, which is why what a refresh changes is the page that says what will move for you when it lands. No turnaround is promised, here or anywhere else on this site: publishers set the rhythm, and so does the work of standardising what they release.

If the form refuses#

Nothing you typed is lost when it does. The message says which rule broke.

Four are checked in the page, before anything is sent, and in this order:

  • "Enter a valid email address, or leave it empty." The email box holds something that is not an address. Leaving it blank is fine; half an address is not.
  • "Name the library you are reporting." The library field is empty on the Libraries page form.
  • "Tell us the version you saw, where you saw it, or add a note." Nothing but the library was filled in, so there is nothing to act on.
  • "The source must be a web address starting with http:// or https://." Paste the full address, not the site name.

Two more come back from the service, after the form has been sent:

  • "Too many reports from this network in the last hour." Reports are limited per network, ten an hour, which is a defence against automated submissions rather than a judgement about you. Wait, then send it again.
  • "We couldn't store the report" and "The report service did not respond" are ours, not yours. Try again.

When to write instead#

The form is built around one library. For anything wider, a question about the model, the licences, an API key or the data terms, write to hello@open-climate.ai. The same address is published in the contact block of the API reference, and open-climate.ai/contact is the form version of it.

How you hear about corrections#

Nothing is pushed to you, and no one writes to tell you that your report landed in an edition. Where a change is written down says which record carries what, and it is the page to watch.

If a number you have already published must stay reproducible, watching is the wrong tool for it: keep your own copy at the moment you use the value, as pin a factor value to a release sets out.