Original Broadcast Date: 03/15/2026
Presented by Maximus
Artificial intelligence is helping the National Oceanic and Atmospheric Administration take a major step forward in weather forecasting. Vijay Tallapragada, senior scientist at NOAA’s Environmental Modeling Center, says the agency is introducing a new generation of forecasting applications designed to dramatically speed up predictions while improving forecasting capability.
The new applications use artificial intelligence to enhance NOAA’s existing forecasting systems rather than replace them. Tallapragada explains that these AI models build on the agency’s current operational framework, particularly the Global Forecast System, often referred to as the GFS or the “American model.”
“These are not replacing any models,” Tallapragada says. “These are augmenting our existing capabilities that we have in our current operational production suite.”
The Global Forecast System serves as the foundation for much of NOAA’s operational guidance. By integrating AI models with this system, forecasters are able to produce improved predictions more quickly while also reducing the computing resources required to generate forecasts.
One of the most significant benefits of the new technology is speed. Traditional weather models take considerable time and computing power to produce forecasts. Tallapragada explains that generating a 10-day forecast with existing operational models typically takes about three hours.
The new AI models reduce that process dramatically.
“The AI models can produce them in less than a minute,” he says.
The efficiency gains go beyond speed alone. The AI models also require far fewer computational resources to run. Tallapragada notes that they use roughly three-tenths of one percent of the computing power needed by traditional forecasting systems.
That improvement allows NOAA to deliver forecasts to its stakeholders far more quickly. Emergency managers, decision makers, and the general public all benefit from receiving weather information sooner.
“We are providing these forecasts much sooner to our stakeholders and emergency managers and the general public,” Tallapragada says.
The additional time can make a meaningful difference when preparing for hazardous weather events. Faster forecasts allow organizations to make earlier decisions and better prepare for potential risks.
Although the technology dramatically accelerates the forecasting process, the focus of the current AI models remains on medium-range weather predictions. Tallapragada explains that these models currently support forecasts out to about ten days.
However, their speed and efficiency allow forecasters to run many more variations of a forecast than previously possible. These variations, known as ensembles, help scientists estimate the probability of different weather outcomes.
Because the AI models require far less computing power, NOAA can generate large numbers of ensemble forecasts.
“That’s the beauty of these AI models,” Tallapragada says. “They can produce hundreds of members of ensembles in predicting the weather.”
Producing more ensemble forecasts improves probabilistic predictions, giving forecasters a clearer understanding of potential weather scenarios. While the models are not yet extending forecasts into seasonal or longer-range outlooks, they significantly improve the medium-range predictions that are often the most actionable.
Tallapragada describes the development of the new AI forecasting tools as part of a broader technological shift in the field. Around 2022, private-sector companies began rapidly developing AI-based weather models, driven in part by advances in graphical processing units, or GPUs, which can perform complex computations quickly.
As these technologies gained momentum, NOAA sought to ensure it could take advantage of the same capabilities.
“We wanted to be not behind in adopting those technologies,” Tallapragada says.
The agency launched an initiative known as Project Eagle to explore and implement AI-based forecasting systems. The name stands for Experimental AI-Based Global and Limited Area Ensemble.
The project moved quickly. Tallapragada says the effort began in the spring of 2025, and within six months the team had developed a system capable of running in real time and demonstrating the speed and value of AI-driven forecasts.
By December 2025, NOAA had already implemented three versions of the AI models into operational use.
“This kind of a fast-track implementation was never heard of before,” Tallapragada says.
He attributes the rapid progress to several strategic decisions. One of the most important factors was adopting modern software development practices, including DevOps approaches that help accelerate the transition from research to operational deployment.
Another key element was collaboration with partners in the private sector. Technologies developed by organizations such as Google DeepMind contributed to the AI modeling capabilities, while NOAA provided the extensive datasets needed to train the models.
By combining those resources, the team was able to test the systems quickly and bring them into operations efficiently.
Tallapragada says the process also demonstrates how government agencies can successfully integrate artificial intelligence into existing technologies rather than replacing systems entirely.
Looking ahead, the experience is shaping how NOAA approaches future technology deployments. The success of the AI forecasting models provides a roadmap for integrating research and operations more seamlessly using modern software development methods such as DevSecOps and agile practices.
“This is providing the pathfinder for our future implementations,” Tallapragada says.
Beyond improving routine forecasts, the technology could support a broader range of weather-related applications in the future. NOAA is exploring ways to apply the models to high-impact weather events, on-demand forecasting, and other specialized uses.
The AI models introduced through Project Eagle represent an important step in modernizing NOAA’s forecasting capabilities. By combining artificial intelligence, advanced computing technologies, and collaboration with private partners, the agency is demonstrating how innovation can improve both the speed and effectiveness of critical government services.
For forecasters and the public alike, the result is simple: faster predictions, better insight into potential weather scenarios, and more time to prepare for whatever the atmosphere may bring.