Enhancing geometric fidelity and wind-field prediction accuracy in complex urban environments using a Cartesian cut-cell CFD model

Accurate prediction of pedestrian-level winds in dense urban environments is essential for ventilation assessment and pollutant-dispersion studies. However, conventional Cartesian-grid computational fluid dynamics (CFD) models often represent oblique or curved building surfaces using a stair-step approximation, which distorts urban morphology unless the grid is globally refined. In this study, we integrated a Cartesian cut-cell method (CCM) into a RANS-based CFD model to mitigate boundary distortion by modifying control-volume fluid fractions and face areas in cells intersected by buildings, while maintaining base grid resolution. The CCM was evaluated using the Architectural Institute of Japan (AIJ) Niigata urban-area benchmark, comparing simulations against wind-tunnel measurements at 80 locations across 16 inflow directions. Performance was benchmarked against the conventional stair-step method (SSM) using the index of agreement (IOA), root-mean-square error (RMSE), and mean bias (MB), alongside detailed flow analyses in representative districts. Relative to the SSM, the CCM increased domain-averaged IOA by 18% and reduced the RMSE and MB by 18% and 55%, respectively. These improvements are primarily attributed to a more realistic representation of building corners and road–building interfaces, which alleviates spurious flow blockage and systematic underestimation of wind speeds inherent in the SSM. While performance gains were substantial, the CCM tended to displace reattachment points further downstream in certain wake regions—a reflection of RANS-based turbulence modeling limitations becoming more apparent with higher geometric fidelity. Nevertheless, the CCM offers a robust improvement in urban wind prediction without additional grid refinement, supporting reliable applications in wind-corridor planning and urban air-quality modeling.