Accelerating Innovation Through Community Modeling

The Relationship Between the Earth Prediction Innovation Center (EPIC) and the United Forecast System (UFS)

Diagram of NOAA's Simplified Operational Forecast Suite, illustrating the relationship between the Unified Forecast System (U F S), U F S Applications, and the U F S’s 21 stand-alone systems

The Earth Prediction Innovation Center (EPIC) was created in response to the congressional charge to accelerate the development of the U.S. capacity to make early, accurate, and effectively communicated environmental predictions. It is designed to support, sustain, and foster the collaborative work being done by the participating members of the Unified Forecast System (UFS). Put more concisely, EPIC fuels and enables the UFS.

EPIC is designed to be the catalyst for community research and modeling advances that continually inform and accelerate advances in our nation’s operational forecast modeling systems. Additionally, EPIC ensures that NOAA’s operational needs and the research and development community are supported by an effective research to operations to research (R2O2R) process resulting in improved numerical weather prediction (NWP) capacity and a more streamlined R2O2R process. EPIC will meet this goal by improving accessibility and usability of the UFS for community members. 

The UFS is a community-based, coupled, comprehensive Earth modeling system, and it is also the community of professionals focused on numerical weather prediction across the Weather Enterprise. The UFS numerical applications span local to global domains—everything from your area’s daily weather forecast to global El Niño/La Niña patterns. It also includes predictive time scales from sub-hourly analyses (for example, when hurricane winds will reach a destination) to seasonal predictions (for example, what summer will be like). The UFS is designed to support the entire Weather Enterprise and to be the source system for NOAA‘s operational numerical weather prediction applications (e.g., the Global Forecast System).

Numerical weather prediction is complex since it attempts to simplify environmental interactions into mathematical formulas packaged as lines of computer code that are then run through supercomputers to generate predictions. In order to improve the numerical weather prediction process, EPIC will focus on the UFS Weather Model, which includes the physics, data assimilation, and post-processing systems that support it. The UFS community will benefit from this accelerated development process and will be integral to improving the science and innovations within the weather model. 

In the future, EPIC’s work will be expanded to extend infrastructure and user support for the UFS to fully coupled Earth system prediction. This will transform the operational suite of models, such as convective-allowing models (e.g., High Resolution Rapid Refresh) and fully coupled forecast systems (e.g., the Seasonal Forecast System, National Water Model, and Ocean Forecast Systems), into one modeling system.

Future User Scenario

A user will navigate to an integrated development environment (IDE) in the Cloud, where they can find cloud resources, models, and data available in a single location. IDE’s are applications for writing code (similar to a word processor for writing text documents), and they come with tools, such as a debugger, used to troubleshoot the code. This IDE will combine high-performance parallel computing resources with dedicated user support to speed up and ease the user’s modeling work and contributions. Fully-coupled cloud-ready Earth system models that expand upon prior weather applications will be available alongside the data needed for testing and benchmarking. Standard testing protocols and procedures will clarify and facilitate the update process, which will also take place in the development environment. Updates to the modeling system will be proposed through these standardized protocols and followed by a series of automated tests linked to continuous integration and continuous deployment pipelines supported by peer reviewers and the accountable code manager(s). The process will be made transparent through open meeting minutes, thorough documentation, and community-based acceptance criteria. Accepted updates will be included in the next modeling system release with contributors acknowledged in the leaderboard.