Summary
There's no way to pre-subset GHCN for a set of cases. Every run reads the full
parquet from GCS.
Ask
A pipeline that takes case metadata (an IndividualCase list or an events YAML)
and writes a per-case GHCN subset, which users then point GHCN(source=...) at.
What already exists
GHCN accepts a custom GHCN-format parquet via source=
(docs/recipes/near_real_time_point_observations.md)cache_dir caches aligned intermediates, but only mid-evaluation — it can't
be built ahead of time or shared
Notes
Same shape as the other case-catalog generators in data_prep/, so this should
be designed alongside #368 rather than separately.
Migrated from Linear EWB-27.
Summary
There's no way to pre-subset GHCN for a set of cases. Every run reads the full
parquet from GCS.
Ask
A pipeline that takes case metadata (an
IndividualCaselist or an events YAML)and writes a per-case GHCN subset, which users then point
GHCN(source=...)at.What already exists
GHCNaccepts a custom GHCN-format parquet viasource=(
docs/recipes/near_real_time_point_observations.md)cache_dircaches aligned intermediates, but only mid-evaluation — it can'tbe built ahead of time or shared
Notes
Same shape as the other case-catalog generators in
data_prep/, so this shouldbe designed alongside #368 rather than separately.
Migrated from Linear EWB-27.