WindINR uses implicit neural representations for local wind queries

WindINR implements a latent-state implicit neural representation framework for high-resolution, continuous wind queries in complex terrain, enabling fast local corrections from sparse observations.