On This Page:
Welcome
Welcome to the Get Code page for Project EAGLE (Experimental AI Global and Limited-area Ensemble forecast system), specifically for the nested-EAGLE application.
Nested-EAGLE is a joint effort between NOAA’s Office of Oceanic and Atmospheric Research (OAR) Research Laboratories, the Weather Program Office (WPO)’s Earth Prediction Innovation Center (EPIC) and the National Weather Service (NWS). Project EAGLE provides NOAA and the broader weather enterprise (including government, industry, and academia) with the ability to rapidly test, develop, and demonstrate Artificial Intelligence (AI) models for global and limited-area ensemble forecasting.
Description
This page provides information on how to access the nested-EAGLE code base, to set up a nested-EAGLE environment, to run the application, and to find documentation and supporting resources.
For any publications based on work using nested-EAGLE, please include the following citation:
The Earth Prediction Innovation Center (EPIC) and the NOAA AI for Numerical Weather Prediction (AI4NWP) Working Group. (2026). The nested-EAGLE (Experimental AI Global and Limited-area Ensemble forecast system) (v1.1.0). Zenodo. https://doi.org/10.5281/zenodo.19672025
Getting Started
Nested-EAGLE provides a publicly available AI for numerical weather prediction (NWP) framework for the broader weather enterprise to enable rapid testing, development, and demonstration for near-real time (NRT) forecasts. It is supported on NOAA Research and Development High Performance Computing System (RDHPCS) Ursa and for non-NOAA developers or community members, the native cloud environment provided by Microsoft Azure. Both platforms provide the Graphics Processing Unit (GPU) resources required for the timely execution of the current iteration of the EAGLE machine learning (ML) model training workflow. Users satisfying the following two prerequisites will find the nested-EAGLE pipeline preconfigured and ready to run; see note below for other users:
- Access to NOAA’s RDHPCS machine Ursa, a similar High Performance Computing (HPC) system, or the native Azure cloud
- Access to a project with GPU node availability
Note that while the current release is supported on Ursa and the native Azure, advanced users may replicate nested-EAGLE on other HPC systems. Subsequent releases will broaden the support of the nested-EAGLE application to other suitable cloud environments and HPC systems.
Once these prerequisites are met, users can navigate to the nested-EAGLE User’s Guide and follow the links to the Quick Start Guide. The guide provides step-by-step instructions for setting up a nested-EAGLE environment on Ursa and running the nested-EAGLE pipeline.
Documentation & User Support
If you encounter a problem using the nested-EAGLE pipeline that appears to be a bug, please report it through GitHub Issues. For idea sharing, tips, and discussion, please use GitHub Discussions. For general questions regarding troubleshooting, please use GitHub Discussion Q&A. These discussions and Q&A topics may eventually populate EPIC’s Technical FAQ page for EAGLE. The latest nested-EAGLE User’s Guide provides the most comprehensive information about the application, including links to more detailed technical documentation for:
- preprocessing
- training
- forecasting
- postprocessing
- verification
- visualization
To get started with the nested‑EAGLE application, refer to this nested-EAGLE-based Video Tutorial, which uses a Google Colab notebook to train and run an AI model.
Preprocessing is supported by the NOAA Physical Sciences Laboratory’s (PSL) ufs2arco for NOAA data ingestion, reformatting, and creation. Training and forecasting are computed with the European Centre for Medium-Range Weather Forecast (ECMWF) and partner agencies’ Anemoi-training and Anemoi-inference. Forecast output postprocessing for verification and visualization (V&V) is accomplished via PSL’s eagle-tools. Verification statistics are computed via NOAA Global Systems Laboratory’s (GSL) wxvx package.
Releases
Version
Description
The latest release of the nested-EAGLE is v1.1.0.
To learn more about the release, visit Nested-EAGLE Application v1.1.0 Release
Release Date: 07/31/2026
Release Description: The nested-EAGLE v1.1.0 expands the v1.0.0 release to include the addition of a global configuration into the nested-EAGLE framework, integration of NRT forecast capability for both global and global-nested configurations into the nested-EAGLE framework, introduction of a retrospective scorecard to compare various AI- and physics-based forecast systems, the addition of a NRT forecast verification page for global-nested configurations, and extended the platform support to native Azure cloud for non-NOAA developers and community members. Users can now run the NRT global EAGLE application with the NOAA Office of Model Development (OMD) checkpoint file used by Artificial Intelligence for Global Forecast System (AIGFS) and the global-nested EAGLE application with the PSL generated global-nested EAGLE checkpoint file.
Release Date: 04/22/2026
Release Description: The nested-EAGLE v1.0.0 is designed for training and evaluating a global-nested configuration of the ECMWF Anemoi ML forecast system. This release covers the end-to-end workflow, from preprocessing training data through forecast generation, verification, and visualization. The nested-EAGLE Application is designed for training an ML model on both a global domain (a choice of ~25 km or 100 km resolution) and a nested-CONUS domain (a choice of ~6 km or 15 km resolution) to produce atmosphere-only forecasts for a range of variables. The target audience for nested-EAGLE v1.0.0 is NOAA affiliates with access to RDHPCS Ursa.
