WHY DID SOME k-ε CLOSURES PERFORM BETTER IN AN URBAN CFD BENCHMARK?

This paper examines why some k-ε closures performed better in the AIJ Case D urban-flow benchmark. The pedestrian-level flow around sharp-edged blocks involves separation, wake recirculation, local acceleration, wall blocking, and non-equilibrium turbulence response. Under a fixed numerical setup, four retained cases formed a near-baseline group, with root mean square deviation (RMSD) from 0.053 to 0.066 against 0.049 for the validated Lag-EB k-ε baseline. Better agreement was associated with closures addressing near-wall anisotropy and delayed stress response in separated regions. Lag-EB k-ε performed best because it combined elliptic blending with stress-strain lag.