Skip to content

DWD RADKLIM / RADOLAN — usage#

The backend is reached through the EarthLens facade with the "radklim" key (or its "radolan" alias) and a dataset= product. It needs no credentials.

RADKLIM reprocessing (yearly NetCDF archives)#

from earthlens.core import EarthLens

lens = EarthLens(
    data_source="radklim",
    dataset="radklim-yw",          # 5-min reprocessed climatology
    start="2021-01-01",
    end="2021-12-31",
    lat_lim=[47.0, 55.0],           # must overlap Germany
    lon_lim=[6.0, 15.0],
    path="./radklim",
)
paths = lens.download()             # -> [Path('.../YW2017.002_2021_netcdf.tar.gz')]

radklim-yw is 5-min and radklim-rw is hourly. A [start, end] window maps to whole years: the reprocessing is served as one .tar.gz NetCDF archive per year, so a window spanning a year boundary downloads both years' archives. These are large files (YW ~13.5 GB/yr, RW ~836 MB/yr).

Operational RADOLAN (per-timestamp granules)#

lens = EarthLens(
    data_source="radklim",
    dataset="radolan-yw",           # 5-min operational, near-real-time
    start="2026-08-10T00:00",
    end="2026-08-10T06:00",
    fmt="%Y-%m-%dT%H:%M",
    lat_lim=[47.0, 55.0],
    lon_lim=[6.0, 15.0],
    path="./radolan",
)
paths = lens.download()             # per-5-min .hdf5 granules in the window

The operational stream keeps only a rolling ~2-day window. A request older than that returns an empty list and logs a warning pointing you at the RADKLIM archive.

To fetch the RADOLAN binary instead of HDF5 (needs a wradlib/pyramids decoder downstream), pass data_format="bin":

EarthLens(data_source="radklim", dataset="radolan-rw", data_format="bin", ...)

Reading a downloaded granule (pyramids)#

earthlens returns raw granule paths; read them with pyramids. The operational HDF5 opens directly:

from pyramids.dataset import Dataset

ds = Dataset.read_file(str(paths[0]))   # operational .hdf5
print(ds.shape)   # grid-dependent — e.g. (1, 1200, 1100) for the RW HDF5 above

The RADKLIM .tar.gz is a gzip-of-tar of per-timestamp .nc files; extract it (stdlib tarfile) and read the member .nc with pyramids.

Attribution#

RADKLIM / RADOLAN are DWD open geodata under CC-BY-4.0 / GeoNutzV. Credit Deutscher Wetterdienst (DWD) wherever you publish results derived from them.