What happened
Rice University researchers, led by Avantika Gori and Yanmo Weng, published a study in the Journal of Geophysical Research: Atmospheres evaluating AI global weather models Pangu-Weather and Aurora. Analysing approximately 200 tropical cyclones from 2020-2025, the study found these AI models accurately predict storm tracks but struggle with physical structure. Specifically, models exhibited deviations in gradient wind balance and overestimated inner core size in stronger storms, impacting wind pattern realism, per Gori.
Why it matters
AI weather models' inability to accurately reproduce storm physical structures introduces significant risk for hazard modelling and impact assessment. While fast, these models' deviations in gradient wind balance and inner core size mean outputs require expert interpretation and bias correction. Procurement teams and security architects evaluating AI forecasting tools must account for these physical limitations, integrating human expertise to validate outputs for high-consequence events like tropical cyclones.
Subscribe for Weekly Updates
Stay ahead with our weekly AI and tech briefings, delivered every Tuesday.
