Résumé : Atmospheric Boundary Layer (ABL) is the region below the stratosphere that is significantly affected by Earth's surface. In this article, an improved ABL model has been tested and validated with field measurements, while simultaneously highlighting the consequences of using an incorrectly formulated ABL model when performing pedestrian wind comfort and safety analysis using Computational Fluid Dynamics (CFD) in a full-scale, realistic setting. The improved ABL model was successfully validated on wind tunnel scale test cases previously, and this article provides evidence that the validation can be extended to realistic, urban-scale test cases as well. Steady, incompressible, Reynolds Averaged Navier–Stokes (RANS) simulations were performed and the results obtained from the incorrect ABL model are compared with those from an improved, consistent ABL model, and the main differences between the two formulations are pointed out in terms of wind velocities and turbulent kinetic energies at the pedestrian height level, and wind comfort maps. The analyses show that using an incorrectly formulated ABL model results in an over-prediction of near-ground velocities due to an artificial wind acceleration, a significant under-prediction of turbulent kinetic energy values, and consequently resulting in a significantly worse pedestrian wind comfort map that differs up to 3 comfort classes in some regions.