Cambridge researchers have developed a fully AI-driven weather forecasting system that could transform how Britain predicts storms, floods and heatwaves, offering faster and more accurate warnings than traditional methods.
The breakthrough system, published in Nature journal, learns directly from historical weather data rather than relying on the complex physics equations and supercomputer simulations used by conventional forecasting. Professor Shankar Somasundaram, who led the research at Cambridge's Department of Applied Mathematics and Theoretical Physics, said the AI identifies intricate patterns in atmospheric data to predict future weather without solving physical equations.
The speed advantage could prove crucial for British communities facing rapidly developing extreme weather. Traditional numerical weather prediction models require enormous computational power and time to run, but the AI system delivers significantly faster forecasting cycles - vital for issuing timely warnings about severe storms or sudden temperature drops.
Early testing shows the Cambridge AI system performs comparably to traditional models, and in some cases surpasses them for certain forecasts. The technology shows particular promise for short-term predictions and identifying the onset of extreme weather - where even small improvements in accuracy can save lives and property.
The research represents the most radical departure yet from conventional meteorology, building on growing global efforts to integrate machine learning into weather forecasting. Several organisations already use AI components alongside existing systems, but Cambridge's fully AI-driven approach marks a significant leap forward.
The system will likely complement national services like the Met Office initially, rather than replacing them outright. However, as AI technology advances and accumulates more training data, its role in delivering reliable weather predictions for British households and businesses is expected to expand dramatically.