How to reproduce and extend¶
This project is designed to be reproduced and extended. Everything — code, notebooks, environment specification, and this documentation — lives in a single Git repository: github.com/Philip-Brohan-MO/Auto-Daily-Rainfall-QC-MO.
If you are familiar with GitHub, fork or clone the repository. If you’d rather not, you can download the whole thing as a zip file.
Software environment¶
Everything runs in the ADRQ Conda environment specified in
environments/ADRQ.yml. See Installation for the one-time
setup, then activate it before doing anything else:
conda activate ADRQ
Compute¶
The local machine is used only for orchestration and the small in-notebook demonstrations. The compute-intensive work — matching every transcription against every Rainfall Rescue station-year, quality control over hundreds of millions of station-days, the SEF export, and the animation renders — runs on the SPICE HPC cluster via SLURM.
The processing code itself is not cluster-specific and will run in any suitable
Python environment; the submission scripts and job specs under scripts/slurm/
are written for a SLURM scheduler and configured through
scripts/slurm/config.sh.
Running the workflow¶
The workflow is driven by the notebooks under notebooks/, run in order. Start
with the workflow overview, which lays out the stages and
links to each notebook.
The documentation¶
These web pages are built with Sphinx from the
Markdown sources in the docs/ directory, and published to
GitHub Pages automatically on every push to main
by the workflow in .github/workflows/docs.yml. To build them locally:
pip install sphinx myst-parser
sphinx-build -b html docs docs/_build/html
or, from inside the docs/ directory with the ADRQ environment active:
make html
Credits and acknowledgements¶
This is the quality-control follow-on to Auto Daily Rainfall, which established the small-VLM ensemble approach that produced the daily transcriptions quality-controlled here.
The monthly rainfall records that anchor every station’s location and metadata come from Ed Hawkins’s Rainfall Rescue project and its army of volunteer transcribers.
Contact¶
Contact Philip Brohan
This document is distributed under the terms of the Open Government Licence. Source code is distributed under the terms of the BSD licence.