AirCast-SR: A Foundation Model for Kilometer-Scale Atmospheric Super-Resolution via Latent Consistency Diffusion

AirCast-SR introduces a foundation model for atmospheric super-resolution, downscaling global AI weather forecasts from 28 km to 1 km resolution for 67-hour predictions of eight surface variables. Utilizing a 3D U-Net within a Latent Consistency Model diffusion framework, it addresses computational limits of traditional NWP models.