NOAA Launches Next-Generation AI-Powered Global Weather Models
NOAA has introduced an unprecedented new operational, artificial intelligence AI-powered global weather prediction model, and it is a big step in the speed, efficiency, and accuracy of the forecasts. The models will also offer forecasters quicker delivery of more precise directives, consuming a tiny portion of the computation capacity.
NOAA’s strategic use of AI is a major step into the innovation of weather models in America, according to Neil Jacobs, Ph.D., the administrator of NOAA. These AI models indicate a new paradigm of NOAA to deliver better accuracy of the large-scale weather and tropical paths, and also to deliver forecast products more quickly to the meteorologists and the public at a lower cost due to drastically lowering the computational expenditures.
The new AI weather models have three different applications:
AIGFS (Artificial Intelligence Global Forecast System): It is a weather forecast model that employs AI to make better weather forecasts in a faster and more efficient way (requiring as few as 99.7% fewer computing resources) compared to its conventional counterpart.
AIGEFS (Artificial Intelligence Global Ensemble Forecast System): This is an ensemble model using AI that offers a list of likely outcomes of predictions to the meteorologists and decision-makers. Initial findings indicate better results as compared to the traditional GEFS, which extends the forecast ability by 18-24 hours.
HGEFS (Hybrid-GEFS): A groundbreaking hybrid (i.e., a hybrid) between the new AI-based AIGEFS (above) and the flagship ensemble model of NOAA, the Global Ensemble Forecast System. Preliminary results indicate that this novel model, a first-of-its-kind strategy of an operational weather centre, is always more successful than the AI-only and physics-only ensemble systems.
Further on, new AI models of operation.
AIGFS is a novel AI-based system that utilises numerous data streams to create weather predictions similar to predictions made by conventional weather prediction systems, including GFS.
Performance: demonstrates better forecast skill compared to the traditional GFS on numerous large-scale features. It is worth noting that it shows that errors in tropical cyclone tracks significantly decrease with increased lead time.
Efficiency: The most radical aspect of AIGFS. A 16-day forecast consuming one forecast only costs the GFS operation 0.3% of computing resources, and it takes around 40 minutes to complete. This decreased latency implies that the forecasters receive important data at a faster rate as compared to the conventional GFS.
Opportunity to improve on: v1.0 has a weakening in tropical cyclone intensity forecasts, but future versions will overcome this challenge by providing better track forecasts.
This is a forecast by AIGFS in the form of a map, which is dated December 10, 2025, showing the heavy precipitation caused by an atmospheric river hitting the U.S. Pacific Northwest. Such AI weather models will save lives and properties by enhancing the accuracy and timeliness of the forecast of phenomena like the disastrous flooding that hit the Northwest. For this reason, the US government ought to prioritise the conservation of forests nationwide. This is why the US government should focus on the preservation of forests in the country.
AIGEFS is a 31-member AI-based ensemble, like the GEFS, which offers a set of opportunities to weather forecasters and decision-makers instead of a single forecast model solution.
Performance: The performance of the forecast skill is similar to that of the operational GEFS.
Efficiency: can be run with only 9% of the computing power of the running GEFS.
Future improvement area: Developers keep on improving the capability of the ensemble to generate a variety of forecast outcomes.
HGEFS – the most creative application in the new suite. The HGEFS is a 62-member “grand ensemble,” which is the result of adding the 31 members of the physical GEFS to the 31 members of the artificial intelligence-based AIGEFS.
Performance: the HGEFS enables the formation of a larger, more robust ensemble by joining two different modelling systems (one physics-based, one AI-based), thus modeling forecast uncertainty better. Consequently, the HGEFS has been constantly performing better than the GEFS and the AIGEFS in most of the key measures of verification.
The first of its kind at NOAA: to the best of our knowledge, such an ensemble system of physical-AI hybrids is the first of its kind in the world.
Future enhancement:
NOAA is still trying to enhance the hurricane intensity predictions provided by HGEFS.
This is an extension of Project EAGLE, a collaborative effort of the Office of Meteorology in the National Weather Service at NOAA, the Oceanic and Atmospheric Research laboratories at NOAA, the Environmental Modelling Center at the National Centers of Environmental Prediction at NOAA, and the Earth Prediction Innovation Center.
Through Project EAGLE and the Earth Prediction Innovation Center, Jacobs added that NOAA scientists are still busy with the members of academia and the private industry in making further advancements in forecasting technology.
The team used the Google DeepMind GraphCast model as a starting point and trained the model on the analysis of NOAA by its own system, the Global Data Assimilation System. This further training on the data provided by NOAA enhanced the performance of the Google model, especially with GFS-based initial conditions.
