What’s new#

Stay up to date with the latest Earth Data Hub changes.

ERA5-Land daily data fixed and extended (2026-11-03)#

ERA5-Land daily data has been regenerated in place, at the same URL. It fixes seven accumulated variables that were one day late from 2022-04-13 to 2025-12-22. These wrong values were served from 2026-06-12 to 2026-11-03. It now covers the whole globe and ships with 60 variables instead of 14.

Breaking changes#

  • Latitudes extend to 90°S instead of 57.1°S, which adds Antarctica. The latitude dimension grows from 1472 to 1801 points.

Bug fixes#

  • e, pev, ro, ssr, ssrd, str and tp were one day late from 2022-04-13 to 2025-12-22, on all grid points: each day held the previous day’s total. So the total of 2022-04-12 appeared twice, and the total of 2025-12-22 was missing. The issue was present in the data served from 2026-06-12 to 2026-11-03.

  • str was missing (NaN) on all grid points from 2024-11-01 to 2024-11-27.

New variables#

  • 24 instantaneous variables, as daily means of the 24 hourly values: asn, fal, lai_hv, lai_lv, lblt, licd, lict, lmld, lmlt, lshf, ltlt, rsn, sd, sde, skt, snowc, src, stl1, stl2, stl3, stl4, swvl3, swvl4 and tsn.

  • 12 accumulated variables, as daily totals: es, evabs, evaow, evatc, evavt, sf, slhf, smlt, sro, sshf, ssro and strd.

  • 10 invariant fields, with no time dimension: cl, cvh, cvl, dl, glm, lsm, slt, tvh, tvl and z.

The 14 existing variables keep their names and units. The time range and the chunk shapes are unchanged.

Updating your analyses#

  1. Recompute results that use the seven accumulated variables for days from 2022-04-13 to 2025-12-22, if you read them between 2026-06-12 and 2026-11-03. Daily analyses are affected most. Totals over longer periods are also off, because each one includes the wrong day at both ends.

  2. Review code that relies on the size of the latitude dimension, or on its last row: the existing latitudes keep their positions, and the new ones are added after them, south of 57.1°S.

Copernicus DEM version 1 (2026-10-16)#

Version 1 of the global Copernicus DEM datasets, GLO-30 and GLO-90, is now available. The elevation values are unchanged, but the stores are now north-up Zarr v3 with a declared CRS, so they work directly with GDAL. Version 0 remains available to existing users until the end of 2026.

Breaking changes#

  • Latitudes run from north to south (north-up); version 0 was south-up.

  • The Zarr format is v3 instead of v2.

  • GLO-30 only: the duplicated 180° meridian column is gone, and longitudes span [-180, 180).

Bug fixes#

  • The EPSG:9518 compound CRS (WGS 84 + EGM2008 height) is declared in both CF and GDAL form; version 0 of GLO-90 declared none.

Performance#

  • GLO-30 chunks are 0.5° × 1° instead of 1° × 1°, so small-window reads download less data.

Migrating from version 0#

  1. Use zarr-python 3 (pip install "zarr>=3") or GDAL 3.8 or newer.

  2. Replace -v0.zarr with -v1.zarr in the store URLs.

  3. Reverse latitude slices: the bounds now go from north to south. A slice in the old order returns an empty selection, silently:

    rome = ds.dsm.sel(lat=slice(41.8, 42.0), lon=slice(12.4, 12.6))  # v0 south-up
rome = ds.dsm.sel(lat=slice(42.0, 41.8), lon=slice(12.4, 12.6))  # v1 north-up

  4. Review code that relies on array positions: row 0 is now the northern edge, and GLO-30 has one fewer longitude column.

ERA5 on Zarr v3 (2026-07-06)#

The hourly ERA5 datasets and ERA5 monthly data on single levels moved to new Zarr v3 stores. The data is unchanged, but the chunks are new. The previous Zarr v2 stores are no longer updated.

Breaking changes#

  • The Zarr format is v3 instead of v2.

  • The store URLs are new.

Performance#

  • The hourly datasets use spatio-temporal chunks that suit both regional and time-series analysis. See each dataset page for the chunk shapes.

Migrating from Zarr v2#

  1. Use zarr-python 3 (pip install "zarr>=3").

  2. Replace the store names in the https://data.earthdatahub.destine.eu/era5/ URLs:

    • reanalysis-era5-land-no-antartica-v0.zarr → era5-land-v0.zarr

    • reanalysis-era5-single-levels-v0.zarr → era5-single-levels-atmosphere-v0.zarr

    • reanalysis-era5-single-levels-ocean-v0.zarr → era5-single-levels-ocean-waves-v0.zarr

    • reanalysis-era5-pressure-levels-v0.zarr → era5-pressure-levels-v0.zarr

    • reanalysis-era5-single-levels-monthly-means-v0.zarr → era5-single-levels-atmosphere-monthly-v0.zarr

  3. If you pass explicit chunks to xr.open_dataset, align them with the new chunk shapes.

ERA5 ECV anomalies discontinued (2026-07-06)#

The ERA5 monthly anomalies of Essential Climate Variables are no longer updated. Compute up-to-date anomalies from ERA5 monthly data on single levels instead, with the reference period of your choice.

Migrating from the ECV dataset#

  1. Open era5-single-levels-atmosphere-monthly-v0.zarr instead of ecv-for-climate-change-1981-2010-anomaly-v0: t2m, tp and siconc keep their names.

  2. Subtract the monthly climatology of the reference period:

    t2m = ds.t2m
clima = t2m.sel(valid_time=slice("1991", "2020")).groupby("valid_time.month").mean()
t2m_anomaly = t2m.groupby("valid_time.month") - clima

    The discontinued dataset used 1981–2010: use slice("1981", "2010") to match it.