[glwagner]

Adds six variables to the ERA5 single-level dataset (reanalysis-era5-single-levels) so that a global single-level ERA5 predownload for the LESbrary can be run on our clusters through NumericalEarth.

NumericalEarth name	CDS request name	netCDF name	Units	Type
:boundary_layer_height	boundary_layer_height	blh	m	instantaneous
:convective_available_potential_energy	convective_available_potential_energy	cape	J kg⁻¹	instantaneous
:convective_precipitation	convective_precipitation	cp	m (hourly accumulation)	accumulated
:land_sea_mask	land_sea_mask	lsm	0–1	invariant
:sea_ice_concentration	sea_ice_cover	siconc (GRIB shortName ci)	0–1	instantaneous
:forecast_albedo	forecast_albedo	fal	0–1	instantaneous

Request names, units, and types come from the ERA5 parameter tables. I checked the netCDF names, units, and long names against a real one-hour CDS download of each variable, and each one loads into a Field.

• cp is handled the same way as tp: MetersPerHour conversion to kg m⁻² s⁻¹ at load time, and it is in ERA5_window_averaged_variables, so its window covers the hour ending at the stamp.
• lsm is time-invariant. It is requested like an instantaneous variable, the same way surface geopotential/topography already are, because the CDS returns the same field for every date. A comment in the variable map now says this.
• :sea_ice_concentration uses the name ECCO already uses in NumericalEarth.
• ERA5-Land counterparts are left out. Only fal exists in ERA5-Land, and it can be added later.

Tests: test_cds.jl checks the request and netCDF names for each variable and the cp unit conversion. test_metadata.jl checks that cp has a backward averaging window and that the other five are instantaneous.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!