Most cities suffer from elevated concentrations of pollution as a result of road traffic exhaust emissions. Our picture of personal exposure is improving and measurements indicate that exposure to ambient air pollution in cities is highly variable in time and space. This talk will discuss methods to estimate vehicle emissions using real-world emissions data and machine learning approaches. Secondly, the talk will describe recent research comparing the results from two air quality models. The large-eddy simulation (LES) model DALES-Urban is used to evaluate the predictive skill of the operational air quality model SIRANE. The use of LES in this study presents a novel approach to air quality model evaluation, avoiding the uncertainty associated with field and experimental validations and providing numerical control that permits systematic analysis of targeted parametrisations and assumptions. A case study is conducted over South Kensington, London under steady-state, neutral conditions, simulating the dispersion of both inert (NOx) and reactive (NO, NO2 and O3) pollutants. SIRANE is shown to successfully capture the dominant trends with respect to canyon-averaged concentrations and along-canyon velocities. The prediction of along-canyon velocities is shown to exhibit sources of systematic error dependant on the angle of incidence of the mean wind. The validity of the uniform in-canyon concentration is assessed by analysing the pedestrian, leeward and windward concentrations resolved in DALES-Urban. Vertical variations in concentrations within the urban canopy are shown to be more significant than directional asymmetry, particularly for primary pollutants. A correction factor of ~1.66 is proposed in order to avoid significant underestimations of pedestrian level exposure. The prevalence of intersections and advective nature of the shear layer are highlighted as important differences between modelling real heterogeneous urban topology and idealised infinite canyons.