Résumé : Computational fluid dynamics (CFD) has become a reference tool for the investigation of pollutant emission and dispersion in urban areas, and for the assessment of the associated risk. In this framework, specific focus is given to the estimation of the downwind and ground level concentration of air pollutants coming from emission sources such as vehicular traffic, industrial plants or accidental events. Pollutant dispersion in the Atmospheric Boundary Layer (ABL) is strongly impacted by the turbulence. This leads to a complex coupled problem, considering the reciprocal influence of the two phenomena. A key role in pollutant dispersion is played by the turbulent Schmidt number Sc t which directly affects the turbulent dispersion coefficient and, consequently, the concentration field. No universally-accepted formulation for the turbulent Schmidt number exists in the literature, although its impact on the prediction of pollutant dispersion is recognised. Stemming from a brief review of the existing literature and knowledge on the topic, this paper aims to propose a novel approach for the optimal determination of Sc t , also through the use of uncertainty quantification. The proposed Sc t is based on the local turbulece level, and is validated on different idealized test cases, representative of typical urban configurations.