Evaluation of a turbulent flow and dispersion model in a typical street canyon in York, UK

It is likely that the use of computational fluid dynamics (CFD) modelling approaches to represent the dispersion of pollutants within urban areas will become increasingly common in the future. If such models are to be used within a decision-making context, their evaluation for a range of urban geometries should be undertaken. In this work a k-epsilon flow model coupled with a Lagrangian stochastic dispersion model are used to model the flow and dispersion of a traffic-related pollutant in a complex urban street canyon within the city of York, UK. The results are compared with measurements of mean wind, turbulence and carbon monoxide concentrations from a field experiment. For comparison, measured concentrations were normalised by the background wind speed and an estimate of the traffic emissions based on traffic flows. The normalisation was found to be inapplicable at background wind speeds below 2 m s(-1) and therefore comparisons are restricted to higher wind speeds. The mean flow is found to be well predicted by the model and helps to interpret complex flow structures in the street, which contained vertical recirculation, corner vortices, channelling and convergence. The modelled worst case concentrations are generally higher than the measurements which could potentially be due in part to the reduced vertical velocities in the model and lower levels of predicted turbulence for certain background wind conditions when compared to the field data. Sensitivity analysis suggests that the modelled concentrations are fairly robust with respect to changes in grid resolution and upwind velocity profiles. The variability of the normalised measured concentrations was high, which is thought to be due to variability within the traffic emissions due to the presence of different traffic regimes, as well as large-scale turbulence within the background flow. (c) 2005 Elsevier Ltd. All rights reserved.