Transformer-based Neural Operators for 3D Wind Field Prediction over Complex Mountainous Terrain

This paper presents a Transformer-based neural operator framework specifically designed for predicting three-dimensional wind fields over complex mountainous terrain. The approach addresses two fundamental bottlenecks of traditional CFD simulations: the expert-intensive mesh generation required for irregular topography, and the computationally expensive iterative solvers that can take hours to days even on high-performance clusters. By leveraging neural operator techniques, the method accelerates inference while maintaining prediction accuracy essential for renewable energy deployment and regional weather modeling applications.