Evaluating the practicality of a Prandtl mixing-length model for urban airflow: Predictive fidelity vs. computational economy

Source abstract

Neutral urban mean-wind simulations are essential for pedestrian-level wind assessment and early-stage urban planning, but the computational cost of conventional turbulence closures often limits their use in large-scale analyses. This study evaluates the practical utility of a Prandtl mixing-length model (MLM) within a time-dependent Reynolds-averaged Navier–Stokes (RANS) framework. MLM’s performance is validated against wind-tunnel measurements across five benchmark cases, ranging from isolated buildings to a realistic urban district (Niigata City), and compared with the standard k–ε (SKE), RNG k–ε (RNG), and large-eddy simulation (LES) models. Model performance is assessed for key mean-flow features, including windward stagnation, wake recirculation, canyon vortices, and pedestrian-level strong-wind zones. A vector accuracy score (VAS), an extension of the conventional hit-rate concept, quantifies local mean-flow agreement, while an accuracy–cost trade-off metric (TOM) summarizes computational efficiency. Results demonstrate that MLM reproduces major flow structures, including several pedestrian-level strong-wind zones, with competitive accuracy. While showing limitations in resolving fine-scale rooftop vortices, MLM yielded favorable TOM values in selected mean-flow-dominant regions. In terms of computational cost, MLM required only 2,345 core-min, representing just 15% of the cost of LES (15,746 core-min) and 70% of two-equation RANS models. These findings position MLM as a cost-effective first-pass screening option for neutral, mechanically driven urban mean-flow assessment within the tested validation envelope, rather than as a replacement for higher-fidelity RANS or LES.