Effects of source inhomogeneity on pollutant dispersion from a cubic building array are investigated as a function of the external wind direction. Using building-resolving large-eddy simulation, it is found that the results depend strongly on the source location and source uniformity inside a near-field region defined by a radius of homogenisation (RAD) based on the spatial autocorrelation of the pollutant concentration. The sensitivity of the RAD to the source location changes abruptly around 30° and is greatly reduced for wind angles between 30 and 45°, in agreement with velocity statistics and the mean horizontal streamlines. The optimal source allocation, which is a proxy for emissions from time-dependent traffic, also changes around 30°. This work clarifies the relationship between inhomogeneous velocity and pollutant statistics and may be applied to the formulation of traffic control policy.

This article deals with the state-of-the-art of experimental and numerical studies carried out regarding air pollutant dispersion in urban environments. Since the simulation of the dispersion field around buildings depends strongly on the correct simulation of the wind-flow structure, the studies performed during the past years on the wind-flow field around buildings are reviewed. This work also identifies errors that can produce poor results when numerically modelling wind flow and dispersion fields around buildings in urban environments. Finally, particular attention is paid to the practical guidelines developed by researchers to establish a common methodology for verification and validation of numerical simulations and/or to assist and support the users for a better implementation of the computational fluid dynamics (CFD) approach.

Flow and dispersion of traffic pollutants in a generic urban neighborhood with avenue-trees were investigated with Computational Fluid Dynamics (CFD). In Part I of this two-part contribution, quality assessment and assurance for CFD simulations in urban and vegetation configurations were addressed, before in Part II flow and dispersion in a generic urban neighborhood with multiple layouts of avenue-trees were studied. In a first step, a grid sensitivity study was performed that inferred that a cell count of 20 per building height and 12 per canyon width is sufficient for reasonable grid insensitive solutions. Next, the performance of the realizable k-ε turbulence model in simulating urban flows and of the applied vegetation model in simulating flow and turbulence in trees was validated. Finally, based on simulations of street canyons with and without avenue-trees, an appropriate turbulent Schmidt number for modeling dispersion in the urban neighborhood was determined as Sct = 0.5.