Latest DESI upgrades include AI models, new datasets, and more
GSL’s latest release of the powerful Dynamic Ensemble Scenarios for Impact-based decision support (DESI) is now live on both the GSL experimental product development server and the National Weather Service (NWS) cloud platform. Version 3.5, released September 2025, brings several new features and datasets to support research and operations across NOAA, as NWS meteorologists and hydrologists across the nation expand their use of DESI to analyze and communicate weather forecast information.

Data archive. Archiving of model data is now expanded to 14 days after initialization. In addition, it introduces a user-driven archive that allows NWS users to save data indefinitely for training and research purposes.
Wind Ninja. As part of a partnership with the US Department of Agriculture, the U.S. Forest Service’s WindNinja diagnostic model has been added to DESI as a new dataset. WindNinja is a high-resolution model that can resolve complex terrain to calculate wind speed and direction in difficult-to-predict areas. It was developed to be used by emergency responders within their typical operational constraints of fast simulation times (seconds), low CPU requirements (single processor laptops), and limited technical expertise. WindNinja now runs automatically on every NWS Spot Forecast request (a special, localized forecast requested by local officials), and custom runs can be submitted and generated in minutes.
New datasets. New model outputs added to DESI include Project EAGLE’s AI-powered GFS-AI and the ensemble GFS (GEFS-AI) and the experimental Rapid Refresh Forecast System (RRFS) and its ensemble (REFS). In addition, the research version of DESI includes the experimental HRRR-Cast, the AI version of HRRR, for testing and evaluation purposes.
New variables. The HRRR, HREF, RRFS, and REFS model output suite now includes maximum updraft and downdraft velocities, and air quality index. In addition, average recurrence intervals (ARI) were added for precipitation forecasts (QPF) at intervals 6, 24, 48, and 72-hours, depending on the selected dataset.
Climatology. Climatologies have been added for a number of variables. They are calculated on 31-day moving windows, computed on the 1st and 15th of every month from the European Centre for Medium-Range Weather Forecasts Reanalysis dataset, version 5 (ERA5). Departures from climatology can be displayed as percentiles.

New visualization features. DESI now provides a globe view with ESRI basemaps, which is the new default setting. A “zoom to dataset” function was added, allowing for quick pan and zoom to datasets in other domains. Another addition is the ability to save display settings using the page’s dynamic URL. Finally, airport locations were added as a new map overlay.
These latest enhancements continue to set the table for DESI’s formal transition to NWS operations as part of the upcoming, revolutionary system to be known as NWS Viz in 2026. A standalone DESI will remain as an internal research tool in NOAA, but NWS Viz, through co-development between GSL and the NWS, will serve NWS forecast offices and National Centers for rapid interpretation and evaluation of probabilistic data. Stay tuned for more news about the transition to NWS Viz in 2026!
Our Mission
Lead research and directed development through the transition of environmental data, models, products, tools, and services to support commerce, protect life and property, and promote a scientifically literate public.
Research Areas
Organizational Excellence, Earth System Prediction, Advanced Technologies, and Decision Support are the foundation to achieving the GSL Grand Challenge: Deliver actionable global storm-scale prediction and environmental information through advanced technologies to serve society.
Global Systems Laboratory