ECMWF: Guide to Weather Datasets and Open Data

The European Centre for Medium-Range Weather Forecasts (ECMWF) is an independent intergovernmental organisation producing global numerical weather predictions (NWP) and maintaining one of the world’s largest meteorological archives. For researchers in atmospheric physics, climatology, oceanography, and environmental data science, ECMWF provides operational forecasts, historical reanalyses, and machine learning simulations. Accessing ECMWF data involves navigating three primary tiers: open real-time forecasts published under permissive open-access terms, multi-decade reanalyses managed through European Union Copernicus programmes, and the historical Meteorological Archival and Retrieval System (MARS) governed by institutional agreements.

What it is

ECMWF functions as both an international research institute and a 24/7 operational forecasting centre. Its core mission is generating global numerical weather predictions across medium-range (up to 15 days), extended-range (up to 46 days), and long-range (seasonal) horizons. The centre ingests billions of meteorological observations daily from satellites, radiosondes, radar networks, ocean buoys, and surface stations to initialize high-resolution atmospheric models.

In academic workflows, ECMWF serves as a major hub for Earth system research data, informing climate science, severe weather tracking, and hydrological forecasting. Beyond weather modelling, ECMWF supports World Meteorological Organization (WMO) programmes and collaborates with the European Space Agency (ESA) and EUMETSAT on the EU Destination Earth initiative to build digital twins of the Earth system.

Who runs it and what it covers

ECMWF was established in 1975 by intergovernmental convention and is supported by European Member States and Co-operating States. The organisation employs approximately 500 staff from over 30 countries across three principal sites:

Reading, United Kingdom: Headquarters and primary scientific research facilities.
Bologna, Italy: High-performance supercomputing facility and primary data archives.
Bonn, Germany: Operational centre managing EU-funded programmes, including Copernicus services.

ECMWF Data Collections and Scope

ECMWF curates several distinct tiers of operational and archived meteorological data:

Operational Forecasts (IFS and AIFS): High-resolution predictions from the physics-based Integrated Forecasting System (IFS) and the machine-learning Artificial Intelligence Forecasting System (AIFS), covering atmospheric dynamics, surface conditions, ocean waves, and ensemble spread.
ECMWF Open Data: Real-time forecast fields published under open licensing at 0.25-degree resolution, with native 0.1-degree (approximately 9 km) grids progressively phased in.
Atmospheric Reanalyses (ERA Series): Comprehensive climate reconstructions combining past observations with numerical models. ERA5 provides hourly global atmospheric records from 1940 onward, accessible via the Copernicus Climate Data Store (C3S) guide.
MARS Archive: The Meteorological Archival and Retrieval System, housing petabytes of operational model cycles and observation streams accumulated since the 1970s.

Licensing and Academic Access Tiers

Data access pathways depend on institutional affiliation and dataset tier:

Open Data (CC BY 4.0): ECMWF Open Data and Open Charts visualisations are released under the Creative Commons Attribution 4.0 International licence, allowing unrestricted non-commercial and commercial reuse with attribution.
Member State Access: National meteorological services and affiliated researchers access the full MARS archive through designated national Computing Representatives.
Research Project Fee Waivers: Non-commercial academic projects requiring data from the closed MARS archive can apply for a Research Project Fee Waiver under ECMWF Product Distribution Rules, waiving service charges upon execution of an ECMWF Service Agreement.

How to search or use it

For programmatic access to open forecast feeds, ECMWF publishes the official ecmwf-opendata Python client. Researchers can search, configure, and retrieve gridded forecast products through the following 6-step workflow.

Step 1: Install the Python Client

Install the official client from PyPI (Python 3.8+ required):

pip install ecmwf-opendata

Step 2: Initialize the Client and Select Data Mirror

Instantiate the client, specifying the model and resolution. Requests can default to ECMWF or route to public cloud mirrors (AWS, Azure, GCP) for enhanced throughput:

from ecmwf.opendata import Client

client = Client(source="ecmwf", model="ifs", resol="0p25")

Step 3: Define Request Coordinates and Meteorological Parameters

Configure parameters using standard WMO variable codes and model stream definitions:

time: Forecast base time in UTC (e.g., 0, 6, 12, 18).
type: Product classification, such as operational forecast (fc) or ensemble forecast (pf).
step: Lead time in hours (e.g., 24, 48, 72).
param: Meteorological variables, such as 2-metre temperature (2t) or mean sea level pressure (msl).

Step 4: Execute Retrieval to Local Storage

Run the retrieval request, targeting a local binary file:

request = {
    "date": 0,           # Latest model cycle
    "time": 0,
    "type": "fc",
    "step": [24, 48],
    "param": ["2t", "msl"],
}

client.retrieve(request, target="ecmwf_forecast.grib2")

Step 5: Ingest GRIB2 into Geospatial Packages

Open the downloaded GRIB2 file using xarray with the cfgrib engine for multi-dimensional array operations:

import xarray as xr

ds = xr.open_dataset("ecmwf_forecast.grib2", engine="cfgrib")
print(ds)

Step 6: Document Provenance and Attribution

Track model cycle, lead time, and parameters to ensure reproducibility across types of research data. For published outputs, apply the required CC BY 4.0 statement: “This product is based on data and products from the European Centre for Medium-Range Weather Forecasts (ECMWF).” For full dataset referencing, follow established data citation practices.

Strengths and limits

Strengths

Predictive Skill: The Integrated Forecasting System consistently ranks among the most accurate global atmospheric models for medium-range forecasts.
Machine Learning Integration: The AIFS framework offers data-driven machine learning forecasts alongside traditional numerical physics.
Decadal Reanalysis: ERA5 provides an unbroken, highly validated historical baseline for climate impact studies.
Permissive Open Data: CC BY 4.0 terms for real-time streams eliminate licensing hurdles for academic researchers.
Interoperable Standards: Conformance with WMO GRIB2 and NetCDF conventions ensures broad scientific software support.

Limits

Specialized Binary Encoding: GRIB2 files require specialized software libraries (GDAL, eccodes) and cannot be opened in tabular software.
Archive Access Controls: Deep MARS queries outside Open Data and Copernicus require institutional authorization or formal fee waivers.
Bandwidth Overhead: High-resolution global atmospheric grids produce large files requiring deliberate storage planning.

Alternatives

Researchers evaluating atmospheric platforms often compare ECMWF with other national and global data repositories, including NOAA NCEI environmental data access services. The table below outlines major alternatives.

Platform / Repository	Operating Entity	Primary Models and Scope	Open Access Licensing	Standard Data Formats
ECMWF	ECMWF (Intergovernmental: UK, Italy, Germany)	IFS, AIFS; Medium, extended, seasonal global forecasts; MARS archive	CC BY 4.0 for Open Data; Fee Waiver for MARS archive	GRIB2, NetCDF, BUFR
Copernicus Climate Data Store (CDS)	ECMWF on behalf of the EU	ERA5 global reanalysis, regional reanalyses, CMIP climate projections	Free and open public access (Copernicus Licence)	GRIB, NetCDF
NOAA NCEI	NOAA (United States)	GFS forecasts, CFS reanalyses, global observational archives	Public domain (US Government work, open without restrictions)	NetCDF, GRIB2, CSV, GeoTIFF
DWD Open Data	Deutscher Wetterdienst (Germany)	ICON global and regional models, COSMO reanalyses, radar mosaics	GeoNutzV / CC BY 4.0 open data	GRIB2, NetCDF, BZ2
UK Met Office DataHub	Met Office (United Kingdom)	Unified Model (UM) atmospheric forecasts, UK regional grids	Open Government Licence (OGL) for open tiers	JSON, GRIB2, NetCDF

FAQ

Is ECMWF data free for academic researchers?

Yes. Real-time forecast products published under ECMWF Open Data are available at no charge under a Creative Commons Attribution 4.0 International (CC BY 4.0) licence. In addition, historical climate reanalyses such as ERA5 are freely accessible through the Copernicus Climate Data Store. Non-commercial research projects needing specialized historical data from the MARS archive can apply for a Research Project Fee Waiver under ECMWF Product Distribution Rules.

What is the difference between ECMWF Open Data and the MARS archive?

ECMWF Open Data is a curated subset of real-time forecasts from IFS and AIFS models, released at 0.25-degree resolution via open APIs and cloud mirrors. The MARS repository is ECMWF’s master archive containing decades of operational model cycles and observations. Full MARS querying requires Member State authorization or an approved research fee waiver.

What software is needed to open ECMWF GRIB2 files?

GRIB2 files require specialized geospatial and meteorological decoders. In Python, workflows typically combine xarray with the cfgrib backend and ECMWF’s eccodes library. Command-line users inspect files using grib_ls, grib_dump, or wgrib2. Desktop Geographic Information Systems such as QGIS can display GRIB2 layers when compiled with GDAL.

How should researchers cite ECMWF datasets in scholarly publications?

For ECMWF Open Data, authors must include the formal attribution notice: “This product is based on data and products from the European Centre for Medium-Range Weather Forecasts (ECMWF).” For reanalysis products like ERA5 retrieved via Copernicus, cite the primary reanalysis journal reference, specify the dataset version, extract dates, and include the persistent digital object identifier (DOI) provided by the Copernicus portal.

Sources

European Centre for Medium-Range Weather Forecasts (ECMWF). “About ECMWF.” Governance and organisational structure. URL: https://www.ecmwf.int/en/about (Checked 2026-11-07).
European Centre for Medium-Range Weather Forecasts (ECMWF). “ECMWF Datasets Catalogue.” Real-time and archived data catalogue. URL: https://www.ecmwf.int/en/forecasts/datasets (Checked 2026-11-07).
European Centre for Medium-Range Weather Forecasts (ECMWF). “Service Agreements and Use Cases: Licences Available.” Licensing terms, CC BY 4.0 policy, and Research Project Fee Waivers. URL: https://www.ecmwf.int/en/forecasts/accessing-forecasts/licences-available (Checked 2026-11-07).
ECMWF Open Data Project. “ecmwf-opendata: Client library to download ECMWF open data.” GitHub repository and PyPI distribution documentation. URL: https://github.com/ecmwf/ecmwf-opendata (Checked 2026-11-07).