FR-MF - Rainfall rate over France mainland

Description

This notebook describes how to access and use the zarr-version of the French radar data provided by Météo-France. The dataset includes aproximately ~500.000 5 min timesteps covering a 1536 x 1536 resolution with 1km grid-size. The original radar.npz files are stored in (mm x 10^-2) and the .zarr dataset is in precipitation rate (mm/h).

Related repositories

• The conversion of the radar data was done by the code hosted on mlcast-community/mlcast-dataset-FR-MF.

• The radar dataset can be directly downloaded from Météo-France Hugging Face repository.

The v0.1.0 zarr dataset covers the temporal period of 2020-01-01 to 2024-12-31.

Citation

Data provided by Météo-France. Processed and distributed by Météo-France AI Lab for open research and development purposes. Please cite as:

@misc{radar_rainfall_france,
  title        = {5 Minutes Radar Rainfall over French Mainland Territory},
  author       = {Météo-France},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/datasets/meteofrance/radar-rainfall}},
  note         = {Distributed under Etalab 2.0 License}
}

Licence: CC-BY-4.0

import matplotlib.pyplot as plt
import xarray as xr

import cartopy.feature as cfeature
import cartopy.crs as ccrs
import mlcast_datasets

cat = mlcast_datasets.open_catalog()
list(cat.precipitation)

['radklim_hourly',
 'radklim_5_minutes',
 'dmi_10_minutes',
 'it_dpc_sri_5min',
 'uk_metoffice_5min',
 'be_rmi_radclim_mfb_5min',
 'fr_mf_prate_5min']

FR-MF - 5-min Rainfall rate

The French MF SRI dataset is available in the intake catalog as fr_mf_prate_5min.

ds = cat.precipitation.fr_mf_prate_5min.to_dask()
ds

<xarray.Dataset> Size: 5TB
Dimensions:        (time: 525787, y: 1536, x: 1536, missing_times: 389)
Coordinates:
  * time           (time) datetime64[ns] 4MB 2020-01-01 ... 2024-12-31T23:55:00
  * y              (y) float64 12kB -3.527e+06 -3.528e+06 ... -5.062e+06
  * x              (x) float64 12kB -6.197e+05 -6.187e+05 ... 9.153e+05
    lat            (y, x) float64 19MB dask.array<chunksize=(1536, 1536), meta=np.ndarray>
    lon            (y, x) float64 19MB dask.array<chunksize=(1536, 1536), meta=np.ndarray>
  * missing_times  (missing_times) datetime64[m] 3kB 2020-01-27T07:45:00 ... ...
Data variables:
    crs            float64 8B ...
    prate          (time, y, x) float32 5TB dask.array<chunksize=(1, 1536, 1536), meta=np.ndarray>
Attributes: (12/13)
    Author:                     Météo-France
    Copyright:                  Météo-France
    Processed by:               WebValley2026, Fondazione Bruno Kessler
    base_frequencies:           5min:2020-01-01T00:00/2024-12-31T23:55
    consistent_timestep_start:  2020-01-01T00:00
    history:                    Created at 2026-06-29T19:00:00+01:00
    ...                         ...
    mlcast_created_by:          WebValley2026, Fondazione Bruno Kessler, <web...
    mlcast_created_on:          2026-06-29T19:00:00+01:00
    mlcast_dataset_identifier:  FR-MF-prate
    mlcast_dataset_version:     0.1.0
    title:                      Météo-France Radar Rainfall Archive
    mlcast_created_with:        https://github.com/mlcast-community/mlcast-da...

xarray.Dataset

• Dimensions:
  • time: 525787
  • y: 1536
  • x: 1536
  • missing_times: 389
• Coordinates: (6)
  • time
    (time)
    datetime64[ns]
    2020-01-01 ... 2024-12-31T23:55:00

    dimension_names :

    ['time']

    calendar :

    proleptic_gregorian

    long_name :

    Time

    standard_name :

    time

    units :

    minutes since 2020-01-01

    array(['2020-01-01T00:00:00.000000000', '2020-01-01T00:05:00.000000000',
           '2020-01-01T00:10:00.000000000', ..., '2024-12-31T23:45:00.000000000',
           '2024-12-31T23:50:00.000000000', '2024-12-31T23:55:00.000000000'],
          shape=(525787,), dtype='datetime64[ns]')

  • y
    (y)
    float64
    -3.527e+06 ... -5.062e+06

    dimension_names :

    ['y']

    units :

    m

    standard_name :

    projection_y_coordinate

    long_name :

    y coordinate of projection

    axis :

    Y

    array([-3526818.459197, -3527818.459197, -3528818.459197, ..., -5059818.459197,
           -5060818.459197, -5061818.459197], shape=(1536,))

  • x
    (x)
    float64
    -6.197e+05 -6.187e+05 ... 9.153e+05

    dimension_names :

    ['x']

    units :

    m

    standard_name :

    projection_x_coordinate

    long_name :

    x coordinate of projection

    axis :

    X

    array([-619652.095362, -618652.095362, -617652.095362, ...,  913347.904638,
            914347.904638,  915347.904638], shape=(1536,))

  • lat
    (y, x)
    float64
    dask.array<chunksize=(1536, 1536), meta=np.ndarray>

    dimension_names :

    ['y', 'x']

    grid_mapping :

    crs

    long_name :

    Latitude

    standard_name :

    latitude

    units :

    degrees_north

    Array Chunk Bytes 18.00 MiB 18.00 MiB Shape (1536, 1536) (1536, 1536) Dask graph 1 chunks in 2 graph layers Data type float64 numpy.ndarray

  • lon
    (y, x)
    float64
    dask.array<chunksize=(1536, 1536), meta=np.ndarray>

    dimension_names :

    ['y', 'x']

    grid_mapping :

    crs

    long_name :

    Longitude

    standard_name :

    longitude

    units :

    degrees_east

    Array Chunk Bytes 18.00 MiB 18.00 MiB Shape (1536, 1536) (1536, 1536) Dask graph 1 chunks in 2 graph layers Data type float64 numpy.ndarray

  • missing_times
    (missing_times)
    datetime64[m]
    2020-01-27T07:45:00 ... 2024-12-...

    dimension_names :

    ['missing_times']

    calendar :

    proleptic_gregorian

    units :

    minutes since 2020-01-01

    array(['2020-01-27T07:45', '2020-01-27T07:50', '2020-01-27T07:55', ...,
           '2024-11-22T07:10', '2024-11-26T08:55', '2024-12-23T08:50'],
          shape=(389,), dtype='datetime64[m]')

• Data variables: (2)
  • crs
    ()
    float64
    ...

    proj4 :

    +proj=stere +lat_0=90 +lat_ts=45 +lon_0=0 +x_0=0 +y_0=0 +a=6378137 +rf=298.252840776245 +units=m +no_defs +type=crs

    crs_wkt :

    PROJCS["unknown", GEOGCS["unknown", DATUM["unknown", SPHEROID["unknown",6378137,298.252840776245]], PRIMEM["Greenwich",0], UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]]], PROJECTION["Polar_Stereographic"], PARAMETER["latitude_of_origin",45], PARAMETER["central_meridian",0], PARAMETER["false_easting",0], PARAMETER["false_northing",0], UNIT["metre",1], AXIS["Easting",SOUTH], AXIS["Northing",SOUTH], USAGE[ SCOPE["Engineering survey, topographic mapping."], AREA["Europe - between 39.48°N and 54.18°N; -9.96°E and 14.55°E."], BBOX[39.477880507122855, -9.964999999999996, 54.18403098825441, 14.54948729204824] ] ]

    spatial_ref :

    PROJCS["unknown", GEOGCS["unknown", DATUM["unknown", SPHEROID["unknown",6378137,298.252840776245]], PRIMEM["Greenwich",0], UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]]], PROJECTION["Polar_Stereographic"], PARAMETER["latitude_of_origin",45], PARAMETER["central_meridian",0], PARAMETER["false_easting",0], PARAMETER["false_northing",0], UNIT["metre",1], AXIS["Easting",SOUTH], AXIS["Northing",SOUTH], USAGE[ SCOPE["Engineering survey, topographic mapping."], AREA["Europe - between 39.48°N and 54.18°N; -9.96°E and 14.55°E."], BBOX[39.477880507122855, -9.964999999999996, 54.18403098825441, 14.54948729204824] ] ]

    semi_major_axis :

    6378137.0

    semi_minor_axis :

    6356752.0

    inverse_flattening :

    298.252840776245

    reference_ellipsoid_name :

    unknown

    longitude_of_prime_meridian :

    0.0

    prime_meridian_name :

    Greenwich

    geographic_crs_name :

    unknown

    horizontal_datum_name :

    unknown

    projected_crs_name :

    unknown

    grid_mapping_name :

    polar_stereographic

    standard_parallel :

    45.0

    straight_vertical_longitude_from_pole :

    0.0

    false_easting :

    0.0

    false_northing :

    0.0

    latitude_of_projection_origin :

    90.0

    [1 values with dtype=float64]

  • prate
    (time, y, x)
    float32
    dask.array<chunksize=(1, 1536, 1536), meta=np.ndarray>

    dimension_names :

    ['time', 'y', 'x']

    grid_mapping :

    crs

    long_name :

    Total precipitation rate

    standard_name :

    precipitation_flux

    units :

    kg m-2 h-1

    Array Chunk Bytes 4.51 TiB 9.00 MiB Shape (525787, 1536, 1536) (1, 1536, 1536) Dask graph 525787 chunks in 2 graph layers Data type float32 numpy.ndarray

• Attributes: (13)

  Author :

  Météo-France

  Copyright :

  Météo-France

  Processed by :

  WebValley2026, Fondazione Bruno Kessler

  base_frequencies :

  5min:2020-01-01T00:00/2024-12-31T23:55

  consistent_timestep_start :

  2020-01-01T00:00

  history :

  Created at 2026-06-29T19:00:00+01:00

  license :

  CC-BY-4.0

  mlcast_created_by :

  WebValley2026, Fondazione Bruno Kessler, <webvalley@fbk.eu>

  mlcast_created_on :

  2026-06-29T19:00:00+01:00

  mlcast_dataset_identifier :

  FR-MF-prate

  mlcast_dataset_version :

  0.1.0

  title :

  Météo-France Radar Rainfall Archive

  mlcast_created_with :

  https://github.com/mlcast-community/mlcast-dataset-FR-MF@0.1.1

var_name = "prate"
crs_name = ds[var_name].grid_mapping
data_crs = ccrs.Projection(ds[crs_name].crs_wkt)

g = (
    ds[var_name]
    .sel(time="2023-11-17")
    .isel(time=slice(None, 3))
    .plot(
        transform=data_crs,
        cmap="viridis",
        add_colorbar=True,
        col="time",
        robust=True,
        subplot_kws=dict(projection=data_crs),
        vmin=0,
    )
)

for ax in g.axes.flat:
    ax.coastlines()
    ax.gridlines(draw_labels=["top", "left"])

/tmp/ipykernel_2512/3845991549.py:20: FutureWarning: self.axes is deprecated since 2022.11 in order to align with matplotlibs plt.subplots, use self.axs instead.
  for ax in g.axes.flat: