An Innovative Decoupled Spatio-Vertical–Temporal Learning Model Based on Topography Guidance for Wind Vector Forecasting
Highlights What are the main findings? TG-SVTFormer improves multi-step wind speed forecasting by decoupling spatiotemporal features and preserving fine spatial details. Integrating topography and upper-air states adds physical constraints, reducing long-horizon prediction errors. What are the implications of the main...