Fast localized urban microclimate emulator: wind and temperatures in unseen urban morphologies
FLUME-FNO is a data-efficient framework that predicts 3D wind and temperature fields from building geometry alone. It uses a Multi-Directional Distance Feature to encode visible free space, followed by a Local FNO model enhanced with gated MLPs. This allows for rapid, accurate microclimate predictions in unseen urban areas using only geometric input.