Get Started
Home
Topics
Search
Library
Research questionHow can multimodal models improve tropical-cyclone track and intensity forecasts across weather foundation models without retraining?Tropical-cyclone forecasting must estimate both storm movement and intensity from atmospheric forecast fields and satellite imagery. Forecast fields differ across weather foundation models, so performance may not transfer when the model is applied to a new source.
AI
Machine Learning
Multimodal Models
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.TC-Next: Zero-Shot Multimodal Cyclone ForecastingThe evidence concerns TC-Next, trained on GraphCast forecasts for the Western Pacific using generic atmospheric variables and GridSat infrared imagery. It reports zero-shot application to Pangu-Weather, IFS HRES, and 2025 Western Pacific WeatherNext Cyclones, with comparisons against TempestExtremes and WeatherNext’s specialized direct tracker; the evidence does not establish performance beyond these settings.research paper · Sep 2, 2026
Related questions
How can we improve weather forecasts two to six weeks ahead despite rapidly degrading skill?How can neural ocean forecasts use cyclone tracks reliably?How can AI weather models produce high-resolution global forecasts without relying solely on analysis data?How can cloud-cover forecasts extend beyond nowcasting despite imperfect NWP initialization while preserving satellite spatial detail?