2026
PHYS-DIFF: A PHYSICS-INSPIRED LATENT DIFFUSION MODEL FOR TROPICAL CYCLONE FORECASTING
ICASSP 2026poster
Tropical cyclone (TC) forecasting is critical for disaster warning and emergency response. Deep learning methods address computational challenges but often neglect physical relationships between TC attributes, resulting in predictions lacking physical consistency. To address this, we propose Phys-Di…