Why artificial intelligence still struggles to forecast rapid hurricane intensification

Recent advances have allowed AI to produce global forecasts comparable to leading physics-based systems, thanks to massive datasets, sophisticated models, and powerful computing. However, when the focus shifts to regional extremes such as rapid hurricane intensification, AI encounters two major hurdles. First, observational data are limited to coastal stations, radars, buoys and partial satellite coverage, leaving large oceanic gaps during critical development stages.

Second, even high-resolution model simulations lack the full three-dimensional detail needed for training. Adding to these challenges, hurricanes exhibit chaotic behavior where minute variations in initial conditions can cause large, unpredictable changes in intensity. Consequently, AI trained to minimize error may smooth out these fluctuations, capping how accurately it can forecast storm strength beyond a few days. Improving forecasts will require not only richer data and better hardware but also a deeper understanding of hurricane chaos and probabilistic prediction methods.