Influence of geometry modeling approaches in a urban microscale large eddy simulation with realistic inflow conditions

Geometry preparation is a major challenge in computational fluid dynamics (CFD) simulations of urban wind flow due to large domains and complex structures. The lack of high-quality 3D city models, due to limitations in building reconstruction algorithms and the lack of high-quality data, often leads to the use of oversimplified cubic geometries to represent urban environments. This study examines the necessity of detailed building models in accurately simulating wind flow using large eddy simulation (LES) with realistic inflow conditions.

We compare three levels of detail (LoDs) of building geometries using a case study of a passing cold front through the city of Dallas, TX, USA. The study involves simulating two distinct atmospheric conditions – pre-frontal and post-frontal – over a 4.5 x 4.5 km area of downtown Dallas. We used FastEddy®, a GPU-accelerated LES code, to conduct our simulations. Inflow conditions were derived from a weather research and forecasting (WRF) simulation, and building models were automatically reconstructed from two open datasets.

We investigate the differences in building geometries among different LoDs and their effects on changing flow patterns. Our preliminary results indicate that higher LoDs must be considered, especially for applications that are relevant at larger heights above ground level.