CFD modelling of dispersion in neutral and stable atmospheric boundary layers : results for prairie grass and Thorney Island
Résumé
The purpose of this paper is to assess the impact on dispersion model predictions of errors introduced by Computational Fluid Dynamic (CFD) models of Atmospheric Boundary Layers (ABLs). It is a known problem that CFD models using the standard k-ε turbulence model struggle to maintain the correct ABL profiles along the length of a flat, unobstructed computational domain. Appropriate ABL profiles may be imposed at the inlet but they are often progressively modified downwind by the CFD model until they no longer represent the specified stability class and/or wind speed. Various solutions have been proposed in the literature to address this issue, although many of them are complex and difficult to implement in commercial CFD software. Also, little is known about the impact of the ABL profiles on dispersion model predictions. To examine this issue, CFD simulations are presented for two sets of field scale experiments: Prairie Grass and Thorney Island. The Prairie Grass cases considered involve continuous releases of a passive tracer in both neutral and stable ABLs. One of the reasons for studying these experiments is to compare dispersion predictions from a standard CFD solution to results obtained from fixing the ABL with prescribed profiles throughout the flow domain. This approach is only possible for passive releases, where the flow field is unaffected by the presence of the tracer gas. Simulations are then presented for a Thorney Island experiment which involved a continuous release of dense gas in a stable ABL. The results show that the modified ABL profiles produced by the CFD models affect the predicted concentrations in most cases. The maximum differences range from 50% to a factor-of-two in the two Prairie Grass cases, although the differences are minimised in the neutral ABL by using a short upwind fetch in the CFD model. For the Thorney Island case, attempts were made to use a modified k-ε turbulence model to maintain the correct stable ABL profiles but the solution was numerically unstable and it failed to produce results. Results are presented for the standard k-ε turbulence model with two different roughness lengths, which produce different results. The inherent difficulties in resolving dense-gas flows over rough surfaces using CFD models are discussed.
Origine : Publication financée par une institution
Loading...