A new analysis from Carbon Brief finds that traditional numerical weather prediction models still outperform artificial intelligence (AI) when forecasting record-breaking extreme weather events. The assessment challenges the narrative that AI is rapidly replacing conventional techniques in meteorology.
Emissions impact is indirect but significant: inaccurate extreme weather forecasts can lead to inadequate preparation, increasing human and economic losses. Better predictions help reduce the carbon footprint of emergency responses and improve resilience planning, though no specific emissions figures were cited in the source.
The finding highlights a gap in investment dynamics. While billions of dollars have poured into AI-driven weather startups like Tomorrow.io and Atmo AI, the traditional models remain more reliable for the most dangerous storms, heatwaves, and floods. The cost of upgrading these conventional systems is often lower than deploying AI at scale.
Geopolitically, the work underscores the importance of maintaining and sharing global meteorological data, as extreme events respect no borders. The Paris Agreement's loss and damage provisions depend on accurate forecasts for vulnerable nations, many of which rely on Western modeling centers.
Some researchers argue the gap is narrowing. AI models trained on massive datasets are improving rapidly, and hybrid approaches that blend machine learning with physics may eventually surpass current methods. However, for now, the old guard holds the edge when it matters most.