AI struggles to predict real hurricane intensity

Experts indicate that while artificial intelligence has significantly advanced weather forecasting due to vast datasets, improved models, and increased computing power, it still struggles to accurately predict the true intensity of hurricanes.

The primary challenge lies in the transition from global to regional forecasting scales. While AI can learn patterns from decades of global meteorological records, hurricane intensity is often a regional prediction problem involving extreme, localized events like torrential rain and violent wind gusts.

Rapid intensification—where a storm jumps from a weak system to a destructive hurricane in hours—is particularly difficult to capture. For example, Hurricane Polo intensified from a tropical storm to a Category 5 hurricane with 290 km/h winds in just 24 hours off the Mexican Pacific coast. Detecting these patterns requires much higher-resolution data than current global datasets typically provide.