AI weather forecasts still lag on rapid hurricane strengthening
AI weather models have made major gains in global forecasting, but they remain less reliable at regional scales. That limitation is especially important for predicting hurricanes that intensify quickly, such as Hurricane Polo in 2026 and Michael in 2018. The article explains that data availability and model design both contribute to the difficulty.
Global AI weather models now match some leading physics-based systems, helped by vast datasets, better algorithms and greater computing power. Decades of worldwide climate and weather records give these models many historical cases of atmospheric change, enabling pattern learning that was not feasible before.
Regional forecasting is harder, especially for hurricanes. Rapid intensification can turn weak storms into severe ones within hours. Hurricane Polo in 2026 reached Category 5 in 24 hours with 180 mph winds; Michael in 2018 became Category 5 before hitting Florida. Training data from stations, radar, buoys and satellites are uneven and sparse over remote ocean areas, while simulations show growing uncertainty at longer lead times.
Coastal residents, emergency planners and responders may be most affected. If AI tools remain less reliable for regional hurricane intensity, communities could still face short warning times when storms strengthen quickly, complicating evacuation and preparation. Better AI could eventually support earlier, more targeted alerts, but its current limits mean it may serve as a complement to, rather than a replacement for, established forecasting systems.