Conditional Flow Matching for Generation of 3D Multi-variable Instantaneous Urban Microclimate Fields
This research proposes Conditional Flow Matching (CFM), a novel generative framework to rapidly create 3D instantaneous urban wind and temperature fields for microclimate design. It uses building geometry and mean flow as guidance to generate realistic, turbulent snapshots in seconds, overcoming the high computational cost of traditional Large Eddy Simulation (LES). The method successfully captures complex turbulence features like wakes and temperature gradients.