Evaluating modelling uncertainty in CFD predictions of building affected dispersion

This paper summarises an investigation of the uncertainties in Computational Fluid Dynamics (CFD) modelling of gas dispersion in the vicinity of buildings and complex terrain. The main objectives were to evaluate the variability of results due to the way in which a CFD code may be applied and to evaluate the accuracy of predictions in realistic scenarios. A group of four organisations undertook simulations with the same CFD code of twelve cases of continuous or instantaneous emissions in stable or neutral atmospheric conditions; geometry ranged from a single building to an industrial site. Substantial variations were observed in the sets of results for each case. This resulted from user errors, differences in boundary condition specification, source modelling and, most commonly, numerical accuracy. The accuracy of the CFD solutions was assessed by comparing predictions with wind-tunnel measurements (for continuous releases).