A breakthrough artificial intelligence system developed by researchers at the University of Cambridge is set to redefine weather forecasting. The AI model, named Aardvark Weather, delivers highly accurate predictions tens of times faster than current forecasting methods while using significantly less computing power.
Supported by the Alan Turing Institute, Microsoft Research, and the European Centre for Medium-Range Weather Forecasts (ECMWF), Aardvark has the potential to transform global weather prediction, researchers say. The findings were recently published in Nature.
“Aardvark reimagines current weather prediction methods, offering faster, cheaper, more flexible, and more accurate forecasts than ever before,” said Professor Richard Turner from Cambridge’s Department of Engineering, who led the research. “It is thousands of times faster than all previous weather forecasting methods.”
Traditional weather forecasting relies on complex models running on supercomputers, requiring large teams and hours of processing time. Recent advances by companies like Huawei, Google, and Microsoft have integrated AI into parts of this process, improving speed and accuracy. However, Aardvark takes a radical step forward by replacing the entire forecasting pipeline with a single machine learning model.
Using data from satellites, weather stations and sensors Aardvark produces both global and local forecasts within minutes on a standard desktop computer. Even with just 10% of the input data used by conventional systems, it already outperforms the U.S. Global Forecast System (GFS) in key variables and competes with forecasts from expert meteorologists.
Anna Allen from Cambridge’s Department of Computer Science and Technology says “This is just the beginning of Aardvark’s potential Its end-to-end AI approach can be applied to hurricanes, wildfires, tornadoesand even broader Earth system forecasting, including air quality and ocean dynamics.”
Beyond its speed and accuracy Aardvark’s flexibility allows for customized forecasts tailored to specific industries and locations, something that traditional forecasting models take years to develop. This capability is particularly valuable for developing nations, where access to high-powered computing and expert teams is limited.
Dr. Scott Hosking from The Alan Turing Institute. “By shifting weather prediction from supercomputers to desktop computers says “Aardvark’s breakthrough is not just about speed it’s about access we can democratize forecasting and bring powerful weather predictions to data-sparse regions worldwide.”
The research team is now setting up a new division within the Alan Turing Institute to deploy Aardvark in developing countries, with a focus on high-precision environmental forecasting for weather, oceans, and sea ice.