meteofrance/fr-radar-rainfall

📡 5 Minutes Radar Rainfall over mainland France

Short name: fr-radar-rainfall Source: Météo-France License: Etalab 2.0 (Open License 2.0)

🗂️ Dataset Summary

This dataset provides high-resolution radar-based rainfall accumulation data over mainland France. Each file contains the rainfall accumulation (in hundredths of millimeters) over the past 5 minutes, with a spatial resolution of 1 km.

The data is derived from the radar precipitation mosaic produced by Météo-France.

• —Temporal resolution: every 5 minutes
• —Spatial resolution: 1 km
• —Grid size: (1536, 1536)
• —Coverage: France mainland
• —Projection: Data is in a Stereographic projection and not in a regular Latitude/Longitude projection 🚨
• —Period covered: currently 2020–2025 (will be extended back to 2015 in future versions)
• —Data format: .npz (NumPy compressed archive)
• —Total size: approx. 15 GB/year (\~100,000 files/year)

📁 Dataset Structure

Files are grouped and archived by year, for ease of use. You can extract the tar archives with the following command:

tar -xvf YEAR.tar

Then, inside the yearly folders, each .npz file corresponds to a single 5-minute time step.

fr-radar-rainfall/
├── 2020/
│   ├── 202001010000.npz
│   ├── 202001010005.npz
│   └── ...
├── 2021/
│   └── ...
└── ...

Each .npz file contains a 2D NumPy array representing the rainfall accumulation over the French territory during that 5-minute interval. Units are hundredths of millimeters.

🧪 Example Usage

To open and visualize a single .npz file:

import numpy as np
import matplotlib.pyplot as plt

# Load the file
data = np.load("2020/202004201700.npz")  # Adjust path as needed
rain = data['arr_0']  # The array is stored under 'arr_0'
print(rain.shape)  # Shape = (1536, 1536)

# Negative values indicate no data, replace them with NaN:
rain = np.where(rain < 0, np.nan, rain)

# Visualize
plt.imshow(rain, cmap="Blues")
plt.colorbar(label="Rainfall (x0.01 mm / 5min)")
plt.title("Rainfall Accumulation – 2020-04-20 17:00 UTC")
plt.savefig("rainfall_20200420_1700.png")

[image]

The provided plots.py module contains some utilities to make nice maps in a regular lat/lon grid. To convert data to mm/h and plot a beautiful map:

import numpy as np
from plots import plot_map_rain

data = np.load("2020/202004201700.npz")
rain = data['arr_0']
rain = np.where(rain < 0, np.nan, rain)
rain = rain / 100  # Convert from mm10-2 to mm
rain = rain * 60 / 5  # Convert from mm in 5 minutes to mm/h

plot_map_rain(
    rain,
    title="Rainfall Rate – 2020-04-20 17:00 UTC",
    path="rainfall_20200420_1700_map.png"
)

[image]

🔍 Potential Use Cases

• —Precipitation nowcasting and short-term forecasting
• —Training datasets for machine learning or deep learning models in meteorology
• —Visualization and analysis of rainfall patterns
• —Research in hydrology, flood risk prediction, and climate science

📜 Licensing

The dataset is made available under the Etalab Open License 2.0, which permits free reuse, including for commercial purposes, provided proper attribution is given.

More information: https://www.etalab.gouv.fr/licence-ouverte-open-licence/

📦 Citation

If you use this dataset in your work, please cite it 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}
}

🙏 Acknowledgements

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