extract_UM_met#
Extract Numerical Weather Prediction (NWP) meteorology from the Met Office Unified Model (UM) — .pp files archived on JASMIN and CEDA — into analysis-ready yearly Zarr stores.
The archive is not laid out for analysis: it is thousands of gzipped .pp files,
split across time blocks and geographic tiles, on a grid whose resolution changes
partway through the record. extract_um_met is one command that pulls the tiles
your region needs, joins them, regrids them onto a target you define, and appends
the result to a single store per year.
python -m extract_um_met run --domain SA --date 2016
# → {zarr_save_directory}/SOUTHAMERICA/SOUTHAMERICA_Met_2016.zarr
Accessing the data#
The raw archive is available via JASMIN. If you have access to the NAME workspace, you can access the full record, from 2011 onwards. Otherwise, you can access data for 2015–2024 only by linking your CEDA and JASMIN accounts or by downloading the data directly from CEDA.
Three ideas shape everything else#
Mk blocks split the archive in time. From 2011 the UM is released in numbered
blocks, each at a fixed native resolution — and the resolution changes between
them, which is why regridding onto a stable target is part of the job. You never
choose the Mk: give a --date and it is resolved in code.
World regions split it in space. Each timestep of the global archive arrives as 14 tiles; a domain declares which ones it overlaps, and the pipeline extracts each and joins them.
Grid modes decide what you land on — footprint (default), regular, or
native — set per domain in config.yaml. Adding a domain is config only, no
code.
All three are covered in Concepts.
Documentation#
Environment (jaspy), TMPDIR, writing config.yaml, native grids.
Start here.
The CLI: run / extract / make-native-grid, options, resume, SLURM,
adding a domain.
The data types, Mk blocks & resolutions, world regions, grid modes.
Variables, model levels, output format and chunking, raw UM inventory.
Error messages, what causes them, and what to do about each.
Silent defects, file quirks, and whether the store you already have is affected.
Module map — what each module does and how the pipeline flows.
Remaining work, the open questions behind current behaviour, how to open an issue, and who wrote this.
Common tasks#
I want to… |
Go to |
|---|---|
Get set up on JASMIN for the first time |
|
Extract a year of an existing domain |
|
Add a new domain |
|
Work out which world regions my area needs |
|
Change which variables or levels are extracted |
|
Change how the output store is chunked |
|
Find out what is in a store I already have |
|
Submit a big run to SLURM |
|
Work out why a run failed |
|
Report a bug or request a feature |
|
Add a data type or grid mode |