Status
Follow-up from the ILAMB obs4REF migration.
lai-AVH15C1 -> obs4MIPs (DONE)
Migrated to the obs4MIPs NOAA-NCEI-LAI-AVHRR-5-0 reference (ESGF) and re-minted. Solve migration is green (intake-esgf fetch into the fixture, dict-form yaml, ilamb.txt prune, regenerated snapshot). Native baselines re-minted on gus (cmip6 + cmip7), applied at diagnostic_version: 3; full ILAMB unit suite passes.
Notable: the reference is ~0.05 deg (5 km, 4.8 GB), far finer than any CMIP model, which hung ilamb3. Added a reusable climate_ref_coarsen_spatial transform (conservatively coarsens the reference to 0.5 deg before comparison; models pass through). Details and the mint gotchas (--force fetch, dropping the v-prefixed version facet, dask .compute() before coarsening) are in plans/ilamb-lai-obs4mips-migration.md.
mrsos-WangMao -> blocked: file is not obs4MIPs-compliant
The CEDA obs_for_ref_v2 WangMao file (ORNL/WangMao/mon/mrsos/50km/v20250321/..., 9.3 GB) carries no obs4MIPs global attributes -- no activity_id, source_id, grid_label, or nominal_resolution; only contact, created on, and note = "OLC ORS". parse_obs4mips rejects it, so it cannot be ingested or produce a solve-fixture row. Distinct from the FLUXNET naming split (#784): there the metadata is inconsistent, here it is absent. Blocked until the file is re-formatted into obs4MIPs upstream.
emp-GLEAMGPCP2.3 -> blocked: no obs4REF equivalent
No obs4REF equivalent for GLEAM evapotranspiration (et/evspsbl) yet. The ilamb/evspsbl/GLEAMv3.3a/et.nc registry entry and string-form source remain until one is published.
pr/GPCPv2.3 and tas/CRU4.02 stay in ilamb.txt regardless -- they back gpp-WECANN relationship analyses.
Tooling added
scripts/fetch_obs4mips_fixture.py (make fetch-obs-fixture) uses intake-esgf to grab obs4MIPs reference data from ESGF and merge full-fidelity rows into the solve fixture. climate_ref_coarsen_spatial transform for very-fine references.
Status
Follow-up from the ILAMB obs4REF migration.
lai-AVH15C1-> obs4MIPs (DONE)Migrated to the obs4MIPs
NOAA-NCEI-LAI-AVHRR-5-0reference (ESGF) and re-minted. Solve migration is green (intake-esgf fetch into the fixture, dict-form yaml,ilamb.txtprune, regenerated snapshot). Native baselines re-minted on gus (cmip6 + cmip7), applied atdiagnostic_version: 3; full ILAMB unit suite passes.Notable: the reference is ~0.05 deg (5 km, 4.8 GB), far finer than any CMIP model, which hung ilamb3. Added a reusable
climate_ref_coarsen_spatialtransform (conservatively coarsens the reference to 0.5 deg before comparison; models pass through). Details and the mint gotchas (--forcefetch, dropping the v-prefixed version facet, dask.compute()before coarsening) are inplans/ilamb-lai-obs4mips-migration.md.mrsos-WangMao-> blocked: file is not obs4MIPs-compliantThe CEDA
obs_for_ref_v2WangMao file (ORNL/WangMao/mon/mrsos/50km/v20250321/..., 9.3 GB) carries no obs4MIPs global attributes -- noactivity_id,source_id,grid_label, ornominal_resolution; onlycontact,created on, andnote = "OLC ORS".parse_obs4mipsrejects it, so it cannot be ingested or produce a solve-fixture row. Distinct from the FLUXNET naming split (#784): there the metadata is inconsistent, here it is absent. Blocked until the file is re-formatted into obs4MIPs upstream.emp-GLEAMGPCP2.3-> blocked: no obs4REF equivalentNo obs4REF equivalent for GLEAM evapotranspiration (
et/evspsbl) yet. Theilamb/evspsbl/GLEAMv3.3a/et.ncregistry entry and string-form source remain until one is published.pr/GPCPv2.3andtas/CRU4.02stay inilamb.txtregardless -- they backgpp-WECANNrelationship analyses.Tooling added
scripts/fetch_obs4mips_fixture.py(make fetch-obs-fixture) uses intake-esgf to grab obs4MIPs reference data from ESGF and merge full-fidelity rows into the solve fixture.climate_ref_coarsen_spatialtransform for very-fine references.