Tornado over landscape.
The National Oceanic and Atmospheric Administration (NOAA) will start using a new suite of weather forecasting models driven by artificial intelligence.
The models are supposed to deliver faster and more accurate predictions than their non-AI predecessors. However, they won’t replace the traditional models, which use complex mathematical equations instead of machine learning. Instead, the AI programs will be pulling information, at least in part, from the traditional models.
"NOAA's strategic application of AI is a significant leap forward in American weather model innovation," said Neil Jacobs, NOAA's administrator, in a statement. "These AI models reflect a new paradigm for NOAA in providing improved accuracy for large-scale weather and tropical tracks, and faster delivery of forecast products to meteorologists and the public at a lower cost through drastically reduced computational expenses."
The agency says it estimates the AI programs will required 91% to 99% less computing power than the traditional models, but those percentages don’t account for the cost of training the models, which is an energy-intensive process.
AI modeling will be used in three areas. The Artificial Intelligence Global Forecast System, or AIGFS, will use AI to deliver weather forecasts. A 16-day forecast will take about 40 minutes to complete, getting data to forecasters more quickly than traditional models, which can take hours.
The Artificial Intelligence Global Ensemble Forecast System (AIGEFS) will provide a range of forecast possibilities. And a Hybrid-GEFS will blend AI with traditional modeling to develop predictions.
The new models were developed in partnership with academia and private industry, using Google DeepMind’s GraphCast technology. They are still relatively untested in the real world and will need to be further refined, the agency said.
Photo credit: Arshad876/Shutterstock.com
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.