Curated ECMWF static dataset covering the June 2026 European heatwave. Contains the Extreme Forecast Index (EFI) for 2 m temperature and four Thermofeel human-thermal-comfort parameters, computed from the operational ENS. Cited as 10.5281/zenodo.21224650.

About this dataset

This static dataset provides a curated set of probabilistic extreme-weather indicators and human-thermal-comfort parameters covering the June 2026 European heatwave (1 June 2026 - 5 July 2026). All fields are computed from the operational ECMWF Ensemble Prediction System (ENS) and the Thermofeel post-processing library.

It is intended to support post-event analysis, verification studies, communication and media work, and downstream research on early-warning signals for heat extremes over Europe.

DOI: 10.5281/zenodo.21224650 Version: 1.0.0 Licence: CC-BY 4.0 Hosted at: ECMWF public S3 bucket (anonymous access)

Contents

The dataset contains two families of fields, initialised twice daily (00 and 12 UTC) for every forecast cycle from 2026-06-01 to 2026-06-30 and 2026-07-05.

Extreme Forecast Index (EFI) - 2 m temperature

• Ensemble-derived probabilistic indicator of how anomalous the forecast 2 m temperature distribution is compared to the model climate (M-climate) for the same location and calendar week.

• Dimensionless, bounded on [−1, +1]; positive values indicate anomalously warm conditions.

Thermofeel human-thermal-comfort parameters

Computed from ENS 2 m temperature, 2 m dewpoint, 10 m wind and radiation fields using the ECMWF Thermofeel Python library:

Temporal coverage

• Initialisation cycles: 2026-06-01 → 2026-06-30 + 2026-07-05 | all 12-hourly (00 and 12 UTC).

• Forecast steps: 0-144h at 3-hourly resolution; 150-360h at 6-hourly resolution.

Spatial coverage

Global domain.

File format

GRIB2. We recommend ecCodes ≥ 2.42.0.

File-naming convention

Files in this dataset use the standard ECMWF real-time file-naming convention. Please see the following documentation page for more details: File naming convention and format for real-time data

• Filename example for extreme forecast index (EFI) - 2 m temperature
  ecm_wf_ifs-ens_od_enfo_efi_20260613T000000Z_20260623T000000Z_240h
• Filename example for thermofeel human-thermal-comfort parameters:
  ecm_wf_ifs-ens-cf_od_oper_fc_20260613T120000Z_20260623T120000Z_240h
  Note that type=oper has been used instead of type=pproc since a non-thermofeel parameter (2t) is inside the files

Where to find the data

The data are hosted openly on the ECMWF public S3 bucket: 

s3://ecmwf-static-datasets/heatwave/june-2026/

There are no credentials needed, pass --no-sign-request to the AWS CLI (or the equivalent anonymous flag in any S3-compatible client).

Examples: accessing the data

Browse the archive

aws s3 ls --no-sign-request s3://ecmwf-static-datasets/heatwave/june-2026/

Download a single file

aws s3 cp --no-sign-request s3://ecmwf-static-datasets/heatwave/june-2026/20260613/ecm_wf_ifs-ens-cf_od_oper_fc_20260613T120000Z_20260623T000000Z_228h .

Mirror the full dataset

aws s3 sync --no-sign-request s3://ecmwf-static-datasets/heatwave/june-2026/ ./june-2026-heatwave/

Programmatic access - Python / boto3

import boto3
from botocore import UNSIGNED
from botocore.config import Config

# Create an unsigned S3 client
s3 = boto3.client(
    "s3",
    config=Config(signature_version=UNSIGNED)
)

# List objects
resp = s3.list_objects_v2(
    Bucket="ecmwf-static-datasets",
    Prefix="heatwave/june-2026/",
)

for obj in resp.get("Contents", []):
    print(obj["Key"], obj["Size"])

# Download one file
s3.download_file(
    "ecmwf-static-datasets",
    "heatwave/june-2026/20260613/ecm_wf_ifs-ens-cf_od_oper_fc_20260613T120000Z_20260623T000000Z_228h",
    "local.grib2",
)

Reading a GRIB2 file - Python / xarray + cfgrib

import xarray as xr

ds = xr.open_dataset(
    "s3://ecmwf-static-datasets/heatwave/june-2026/20260613/ecm_wf_ifs-ens-cf_od_oper_fc_20260613T120000Z_20260623T000000Z_228h",
    engine="cfgrib",
)

print(ds)

Provenance

• Forecasts: ECMWF Integrated Forecasting System (IFS), ENS configuration, operational cycle 50r1, 51 members (control + 50 perturbed).

• EFI: computed against the operational M-climate (model re-forecast climate) for the corresponding calendar week and location.

• Thermofeel outputs: computed member-by-member from ENS 2 m temperature, 2 m dewpoint, 10 m wind and radiation fields using Thermofeel. Ensemble mean and selected quantiles are archived.

Limitations

• This dataset captures the operational forecast view of the event, it is not a reanalysis. Values reflect ECMWF's forecasts at issue time, not observed conditions.

• EFI is a signal of unusualness relative to model climate, not a magnitude. It should be read alongside the underlying ensemble distribution and, where possible, the Shift of Tails (SOT) diagnostic.

• Thermofeel indicators are diagnostic combinations of forecast variables; their accuracy inherits the underlying forecast skill, particularly for wind and radiation at short lead times.

Citation

Please cite the dataset via its DOI:

Pidduck, E., et al. (2026). ECMWF EFI and Thermofeel heat-stress indicators for the June 2026 European heatwave (Version 1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21224650

BibTeX

@dataset{pidduck_2026_heatwave_efi_thermofeel,
  author    = {Pidduck, Emma and others},
  title     = {ECMWF EFI and Thermofeel heat-stress indicators for the June 2026 European heatwave},
  year      = {2026},
  publisher = {Zenodo},
  version   = {1.0.0},
  doi       = {10.5281/zenodo.21224650},
  url       = {https://doi.org/10.5281/zenodo.21224650}
}