Google's WeatherNext Cyclones lands Nature cover, fixing AI weather models' intensity blind spot
jiqizhixin · x · 2026-09-25
Google unveiled WeatherNext Cyclones (WN-C), featured on the cover of Nature, achieving state-of-the-art tropical cyclone track forecasting.
- The problem: existing AI weather models run at 28 km resolution and inherit a low-intensity bias from 0.25° global analysis data — they can predict where a cyclone goes but not how strong it gets, long considered a major limitation.
- The method: WN-C trains end-to-end on global analysis data plus the historical IBTrACS cyclone database, using a deterministic mapping to convert sparse tabular cyclone track data into gridded tensors at 0.25°, letting the model learn intensity directly from real cyclone records.
- Capability: it generates 15-day ensemble forecasts of global weather and cyclone scenarios, delivering the most advanced ensemble predictions for both track and intensity.
More from Research
- 320Gb/s unamplified transmission demoed with 100GHz Ge PD and TFLN MZM on foundry SiPh — jwt0625 · 2026-09-26
- Comparing Ghent and imec silicon photonics PDs: 320 Gb/s links but 5 dB grating coupler loss — jwt0625 · 2026-09-26
- NSF FRR robotics meeting workshop: four talks on skill learning and physical intelligence — YuXiang_IRVL · 2026-09-26
- Fiora Starlight proposes letting models write singularity scenarios to ease OoD generalization anxiety — repligate · 2026-09-26
- MIT study: aging brains keep robust language networks despite cognitive decline — DrKavner · 2026-09-26
- Déjà View, a NeurIPS Oral: one looped transformer block matches 3D reconstruction models 8-10x its size — ZGojcic · 2026-09-26