Meteosat-12 (MTG-I1) first full-disk image, Flexible Combined Imager — © EUMETSAT, via Wikimedia Commons

EUMETSearch

Query and download data from the EUMETSAT DataStore and DataTailor — MTG FCI and MSG SEVIRI products — and export it to regular-grid or MajorTOM sparse Zarr stores.

Features

• Search EUMETSAT collections by time window and bounding box, with no download.
• One-off, single-window downloads (download_single) for event-based use cases.
• Batched, multi-day interval downloads (download_interval) with threaded I/O for larger jobs.
• Export downloaded data straight to a regular lat/lon Zarr store or a MajorTOM sparse Zarr store.
• Look up product metadata either by the package's short product key (e.g. "MTG-FCI-L2-AFM") or directly by EUMETSAT product code (e.g. "EO:EUM:DAT:0682").
• A tiled CLI pipeline (pipeline_data.py) for large-area, multi-tile batch downloads.

Installation

pip install eumetsearch

From source, for development:

git clone https://github.com/Riccardo7-DS/eumetsearch.git
cd eumetsearch
pip install -e .

Optional extras

pip install "eumetsearch[transform]"

Authentication

Downloading requires an EUMETSAT Data Store API key/secret — register at the EUMETSAT API key portal to get one. Copy .env.example to .env and fill in:

EUMETSAT_CONSUMER_KEY="..."
EUMETSAT_CONSUMER_SECRET="..."

EUMDownloader loads this automatically (python-dotenv), as long as the .env file is somewhere above your working directory.

Quickstart

1. Look up a product

Products are keyed by a short name in products_list (see Supported products), each mapping to metadata including the real EUMETSAT product code:

from eumetsearch import products_list, EUMDownloader

product_id = products_list["MTG-FCI-L2-AFM"]["product_id"]   # "EO:EUM:DAT:0682"

# Or go the other way — resolve metadata directly from a product code,
# without needing to know (or look up) the dictionary key:
meta = EUMDownloader.get_product_metadata("EO:EUM:DAT:0682")
print(meta["product_name"], meta["bands"])

2. Search without downloading

downloader = EUMDownloader(
    product_id=product_id,
    output_dir="./data/datastore_data",
)

results = downloader.search(
    start_time="2024-07-23T12:00:00",
    end_time="2024-07-23T12:10:00",
    bounding_box=[41.20, 41.02, -8.70, -8.48],  # [N, S, W, E]
)
print(results["count"], results["products"])

3. Download a single time window

For one-off, event-based downloads — no daily-interval splitting or thread-chunking, just one window in one shot:

file_list = downloader.download_single(
    start_time="2024-07-23T12:00:00",
    end_time="2024-07-23T12:10:00",
    bounding_box=[41.20, 41.02, -8.70, -8.48],  # [N, S, W, E]
    method="datatailor",  # or "datastore"
)

Files land in output_dir. See notebook/quickstart.ipynb for a full worked example, from search through to plotting the downloaded data.

4. Batch downloads over multiple days

downloader.download_interval(
    start_time="2025-06-01T08:00:00",
    end_time="2025-06-05T08:00:00",
    bounding_box=[41.20, 41.02, -8.70, -8.48],
    observations_per_day=6,
    jump_minutes=60,
)

download_interval sets up the daily-interval splitting; combine it with ZarrExport (regular lat/lon grid) or MajorTomZarrExport (sparse MajorTOM grid) to trigger the actual download and write results to Zarr. For large, multi-tile areas, the CLI pipeline wraps this end-to-end:

python -m eumetsearch.data_collection.pipeline_data \
    --product MTG-FCI-L1C-FDHSI \
    --channels vis_06 vis_08 \
    --region europe \
    --n_lat 4 --n_lon 4 \
    --start_date 2025-06-26T08:00:00 \
    --end_date   2025-06-26T13:00:00 \
    --observations_per_day 6 \
    --jump_minutes 60 \
    --majortom

Supported products

License

Apache License 2.0 — see LICENSE.txt.