Announcement: The nested‑EAGLE (Experimental AI Global and Limited‑area Ensemble) Application v1.0.0 Released on 4/22/2026

[priyarpillai]

The Earth Prediction Innovation Center (EPIC) and the NOAA Artificial Intelligence for Numerical Weather Prediction (AI4NWP) Working Group proudly announce the public release of the nested‑EAGLE (Experimental AI Global and Limited‑area Ensemble) Application v1.0.0. This release complements Environmental Modeling Center’s (EMC)’s GraphCast‑based global‑EAGLE‑solo and global‑EAGLE-ensemble forecast systems currently operational within NOAA. The nested‑EAGLE v1.0.0 delivers a machine learning (ML) based, end‑to‑end (E2E) workflow that integrates with European Centre for Medium-Range Weather Forecasts (ECMWF) and partner agencies’ Anemoi framework, supporting data preprocessing, model training, inference, postprocessing, verification, and visualization.

The EAGLE framework is designed to support machine-learning (ML) model training on global domain (a choice of ~25 km or 100 km resolution) and nested-CONUS domain (a choice of ~6 km or 15 km resolution), producing atmosphere‑only forecasts for multiple variables. This initial nested-EAGLE release is supported on the NOAA Research and Development High Performance Computing System (RDHPCS) Ursa system, with future releases planned to expand to native Microsoft Azure cloud environments.

The workflow includes environment setup, preprocessing using NOAA Physical Sciences Laboratory (PSL)’s ufs2arco utility to convert NOAA and non‑NOAA datasets into Anemoi‑ready ARCO formats, model training via Anemoi-core, and forecast generation through Anemoi‑inference. Postprocessing converts outputs into verification‑ready formats.

Verification is performed using NOAA Global Systems Laboratory (GSL)’s wxvx package, which leverages Developmental Testbed Center (DTC)’s Model Evaluation Tools (MET) to compute statistics and generate diagnostic plots for standard 2‑D and 3‑D variables. The eagle‑tools Python package provides additional utilities for postprocessing and visualization. Repository testing includes optimized E2E workflow tests and static code analysis.

User Support Resources for the nested‑EAGLE Application v1.0.0

• GitHub Repository: https://github.com/NOAA-EPIC/EAGLE/tree/release/public-v1.0.0

• User’s Guide: https://epic-eagle.readthedocs.io/en/v1.0.0/ (This comprehensive nested-EAGLE documentation includes setup, execution, and testing instructions)

• Required datasets publicly accessible on AWS S3
  (Important Note: Some of these data buckets may initially appear empty, as they can take some time to load)

  • GFS analysis dataset: https://noaa-gfs-bdp-pds.s3.amazonaws.com/index.html
  • HRRR analysis dataset: https://noaa-hrrr-bdp-pds.s3.amazonaws.com/index.html
  • PrepBufr observations: https://noaa-reanalyses-pds.s3.amazonaws.com/index.html#observations/reanalysis/conv/prepbufr/

Additional User Support available through NOAA/EPIC/EAGLE GitHub Discussions Forum

• Submit questions through GitHub Discussions' Q&A page
• Report issues or bugs through GitHub Issues
• Contact the team directly at support.epic@noaa.gov.

For more information about this release, including contributors, see the nested-EAGLE Application v1.0.0 release article on the EPIC Community Portal!