Sensitivity of urban wind CFD predictions to geometric representation: Wind-tunnel validation around complex buildings

Accurate prediction of urban wind flow depends strongly on the geometric detail used in Computational Fluid Dynamics (CFD) models, yet simplified block representations remain common in urban studies. This work combines wind-tunnel measurements and CFD simulations to quantify how geometric simplification affects airflow, façade pressures, K-based comfort classification and low-velocity zones around two interconnected campus buildings with complex architectural features. A realistic 1:240 wind-tunnel model equipped with 194 pressure taps and 17 Irwin probes provided pressure-coefficient and near-ground velocity data for multiple wind directions. CFD simulations using the realizable k–ε model reproduced pressure coefficients with normalized RMSE values of 9–15% and near-ground velocities with mean relative errors of 12–17% in high-velocity regions. Comparisons between realistic and simplified geometries show that simplification systematically alters flow structures, particularly in recirculation and shear-layer regions. The area classified as low-velocity conditions (K < 0.2, where K is the local-to-reference wind-speed ratio) was underestimated by up to 66%, while intermediate/high-velocity comfort categories were overestimated by up to 53%, leading to significant shifts in comfort classification. Simulations including an upstream barrier further showed that simplified geometries alter the predicted aerodynamic response to the barrier by overestimating velocity reductions and failing to capture flow redistribution in semi-enclosed areas. Under controlled inflow conditions, these results demonstrate that geometric simplification introduces non-uniform errors in CFD-based urban wind assessment, affecting comfort classification and the predicted aerodynamic response to passive flow-control measures.