Authors and acknowledgements

This documentation and codebase are currently maintained by Philip Brohan (Met Office).

All errors are the responsibility of the maintainer; credit is widely shared.

While the human maintainer is entirely responsible for the project, almost all the work was done by AI: GitHub Copilot wrote all the code and almost all the documentation. The transcriptions were done by five small VLMs:

• Huggingface SmolVLM2 2.2B

• IBM Granite vision 4.1-4b

• Google Gemma 3 4b

• Google Gemma 4 E4B

• Mistral-Small-3.1-24B

All of which were downloaded from Huggingface.

• This project was created by Philip Brohan (philip.brohan @ metoffice.gov.uk). He was funded by the Met Office Hadley Centre Climate Programme funded by BEIS, and by the Met Office Weather and Climate Science for Service Partnership (WCSSP) South Africa as part of the Newton Fund.

• This project is a follow-on to Rainfall Rescue, which established the general method.

• This project’s outputs are used by Auto daily Rainfall QC which turns the raw transcriptions into ready-to-use observations.

• The document images used were provided by the Met Office Digital Library and Archive imaged from the collection of the National Meteorological Library and Archive. The original documents exist because of the efforts of thousands of observers, scientists, archivists and administrators over many years.

• The validation sample of 75 images+transcriptions (from Irish Daily Rainfall records) was provided by Ciara Ryan.

• Documentation is built with Sphinx and MyST Markdown, and hosted through GitHub and GitHub Pages.

If you spot a bug or have a suggestion, please use the issue tracker: Raise an issue.

Appearance on this page does not imply endorsement by any person or organization.

This document is distributed under the terms of the Open Government Licence. Source code included is distributed under the terms of the BSD license.