Atmospheric pollutant dispersion near sources is typically simulated by Gaussian models because of their efficient compromise between reasonable accuracy and manageable com- putational time. However, the standard Gaussian dispersion formula applies downwind of a source under advective conditions with a well-defined wind direction and cannot calculate air pollutant concentrations under calm conditions with fluctuating wind direction and/or upwind of the emission source. Attempts have been made to address atmospheric disper- sion under such conditions. This work evaluates the performance of standard and modified Gaussian plume models using measurements of NO2, PM10, PM2.5, five inorganic ions and seven metals conducted near a freeway in Grenoble, France, during 11-27 September 2011. The formulation for calm conditions significantly improves model performance. However, it appears that atmospheric dispersion due to vehicle-induced turbulence is still underestimated. Furthermore, model performance is poor for particulate species unless road dust resuspension by traffic is explicitly taken into account.