On This Page:
Welcome
The Earth Prediction Innovation Center (EPIC) and the NOAA Artificial Intelligence for Numerical Weather Prediction Applications (AI4NWP) Working Group support a community-based, comprehensive Earth modeling system. NOAA’s operational model suite for numerical weather prediction (NWP) is quickly broadening to include artificial intelligence-based models. The community enables research, development, and contribution opportunities within the broader Weather Enterprise (including government, industry, and academia).
Description
Global medium-range weather forecasting is critical to decision-making across many social and economic domains. Traditional numerical weather prediction uses increased computing resources to improve forecast accuracy, but historical weather data cannot be directly used to improve the underlying model. We introduce a machine learning-based method called “GraphCast”, which can be trained directly from reanalysis data. It predicts hundreds of weather variables over ten days at 0.25-degree resolution globally in under one minute. We show that GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets, and its forecasts support better severe event prediction, including tropical cyclones, atmospheric rivers, and extreme temperatures. GraphCast is a key advancement in accurate and efficient weather forecasting, and it helps realize the promise of machine learning for modeling complex dynamic systems.
GraphCast: Learning skillful medium-range global weather forecasting. (2023, Aug. 4). Lam, Sanchez-Gonzalez, Willson, Wirnsberger, Fortunato, Alet, Ravuri, Ewalds, Eaton-Rosen, Hu, Merose, Hoyer, Holland, Vinyals, Stott, Pritzel, Mohamed, Battaglia. GraphCast: Learning skillful medium-range global weather forecasting
Getting Started
To get started with GraphCast, users can walk through a Jupyter Notebook as a demonstration: UICFW Workshop GraphCast Training. This will provide a quick start on understanding how to install, initialize, load data, run the model, and train the model. Then, the baseline resides in NOAA-PSL/graphcast: GraphCast: Learning skillful medium-range global weather forecasting.
Documentation & User Support
The Land DA User’s Guide has the most comprehensive information on the Land DA System. Users may need different versions of the User’s Guide depending on their goals:
Version
Description
Documentation for the main branch. This includes Jupyter notebooks.
Users can also get expert help through the GitHub Discussions Q&A.
Developer Support
The Land DA Developer Information page provides information on the Land DA hierarchical repository structure, developer support, and testing.
Releases
The latest release of the Land DA System is v1.2.0. See the Releases page for more information on current and past releases.
Release Date: 12/11/2023
Release Description: The Land Data Assimilation (DA) System combines the Noah-MP land surface model with data assimilation capabilities into a user-friendly workflow. The Land DA workflow code base is charting a path forward to unify the Noah-MP forecast model with the UFS Weather Model (WM). Updates for this release include:
- Integration of the UFS Noah-MP land component into the Land DA System for use as an alternative to the Common Community Physics Package (CCPP) Noah-MP LSM land driver. The coupling layer of the land component is developed using the Earth System Modeling Framework (ESMF) and the National Unified Operational Prediction Capability (NUOPC) interoperability layer.
- Updates to model forcing options for use of the UFS land component
- Provided a new analysis option in the cubed-sphere native grid using GSWP3 forcing
- Established global land grid-point consistency with the head of the UFS WM baseline test cases
- Added a new sample configuration file (settings_DA_cycle_gswp3)
- Included an additional ECMWF ERA5 reanalysis forcing option in the existing vector-to-tile conversion analysis process
- CTest suite upgrades—the ERA5 CTests now test the operability of seven major components of Land DA: vector2tile, create_ens, letkfoi_snowda, apply_jediincr, tile2vector, land_driver, and ufs_datm_land
- Upgrade of the JEDI DA system to JEDI Skylab v4.0
- Updates to sample datasets for the release (see the Land DA data bucket)
- Singularity container (ubuntu20.04-intel-landda-release-public-v1.2.0.img) updates to support the changes described above
- Documentation updates to reflect the changes for this release
Known Issues:
- For the GSWP3 forcing option, an artificial GHCN snow depth observation file is provided for a single-cycle analysis test for 2000-01-03. The GHCN observation database will be extended in the near future.
Documentation:
Release Date: 05/25/2023
Release Description: The Land Data Assimilation (DA) System combines the Noah-MP land surface model with data assimilation capabilities into a user-friendly workflow. The Land DA workflow code base is charting a path forward to unify the Noah-MP forecast model with the UFS Weather Model (WM). Updates for this release include the migration of the Land DA System to the ufs-community GitHub space, the addition of a UFS WM build option, modulefile updates to use the spack-stack Unified Environment on Level 1 systems, and DA upgrades to utilize JEDI’s Skylab v3.0 release of jedi-bundle.
Known Issues:
- The GitHub Actions workflow YAML for the Docker-based build and CTest was turned off due to the limited disk space provided in the GitHub Actions free runner.
Documentation:
Other Links
- The spack-stack repository contains the prerequisite software needed to run the Land DA System and other UFS software. This software is preinstalled on Level 1 systems (e.g., Hera, Orion). Users can view the spack-stack documentation to install it on other systems.
- JEDI Documentation (Skylab v4.0): The Land DA System uses JEDI code for DA. The documentation contains information about the JEDI code and how to work with it.
- Land DA System Data Bucket: Get downloadable data to run supported Land DA cases.
- NOAA EPIC Docker Hub: A repository of container images (including Land DA container images) that users can download. Land DA containers package the Land DA System together with all its software dependencies for an easier experience building and running Land DA.
- UFS Land DA System README.md file
Get Started with the Land DA System
Resources
Public Release Code
v1.2.0 (latest)
For Developers
Land DA System Repository Structure
Useful Links
spack-stack code and spack-stack documentation
JEDI code and JEDI documentation (Skylab v4.0)
