Content area
Full text
ABSTRACT
A theoretical and statistical comparison of a three-dimensional computational fluid dynamics (CFD) model with two Gaussian plume models is proposed on the Prairie Grass data field experiment for neutral conditions, using both maximum arcwise concentrations and spatially paired observations. In theory, it is impossible to have the same near-source behavior with the Eulerian CFD code as with the Gaussian plume models. The former presents the inability to account for the dependence of the turbulent diffusivity to the distance from the source, contrary to plume models for which this dependence is fitted to observations. The study described herein looks at the practical implications of these theoretical differences by comparing the two different types of models on a flat-terrain case, a situation favoring Gaussian models. The results herein show that the Eulerian CFD model gives acceptable results, both for arc-maximum concentrations and spatially paired observations. Indeed, the statistical performances are above the criteria of "good performance" commonly defined in literature. In general, the results for Eulerian code fall between those of a Gaussian model that has been fitted using the Prairie Grass dataset and those of one fitted with different datasets.
(ProQuest: ... denotes formulae omitted.)
1. Introduction
In support of air-quality studies, computational fluids dynamics (CFD) techniques are now well known and the usage of CFD atmospheric models is increasing, caused by recent advances in computational power. They are particularly useful for predicting the dispersion of gases or aerosols in urban areas or around industrial sites, where flow and turbulence are significantly affected by the presence of groups of buildings and thus are widely inhomogeneous and unsteady. For this purpose, CFD models have shown reasonable agreement with available observations, through extensive field experiments. For example, Kit Fox (Hanna and Chang 2001) and the Mock Urban Setting Test (MUST; Yee and Biltoft 2004) are recent experiments used for the evaluation of simulations of flow and dispersion in idealized city configurations (Hanna et al. 2004; Milliez and Carissimo 2007). For such complex geometries, classical analytical models such as puff or plume Gaussian models do not appear to be appropriate.
However, for very simple geometries and conditions, such as the flat terrain found in the Prairie Grass experiment (Barad 1958), there have been reports on the difficulty...