Improvement of a stochastic backscatter model and application to large‐eddy simulation of street canyon flow

Abstract

A stochastic backscatter (SB) subgrid‐scale (SGS) model is applied, for the first time, to large‐eddy simulation (LES) of street canyon flow. We model a ‘skimming flow’ regime under a neutrally stratified atmosphere, in which the approaching wind is perpendicular to the along‐street axis of a street canyon of unity aspect ratio. Previous LESs of this type have shown an under‐prediction of the intensity of the primary eddy (PE) that forms within the street canyon, indicating a lack of momentum transfer across the roof‐level shear layer. The SB model, however, acts to increase this momentum transfer, bringing the simulated PE intensity significantly closer towards that observed in a corresponding wind‐tunnel experiment. A metric for the PE intensity, ωPE, based on the two‐dimensional vorticity field, is increased from around 70% of the wind‐tunnel ωPE value (with the Smagorinsky SGS model) to as much as 90% (with the SB model). Calculation of the air exchange rate at roof‐level confirms that the rate of entrainment into the street canyon is increased with the inclusion of backscatter.We also outline an improvement to the SB model prior to its application. In its previous version, a constraint on the magnitude of the backscatter acceleration variances ensured a theoretically appropriate level of additional grid‐scale (backscattered) energy. Here, a further constraint on the magnitude of the main covariance term also facilitates a better representation of grid‐scale vertical momentum flux. This new constraint alone can help to increase the simulated ωPE value by as much as 10% of the wind‐tunnel ωPE value, and requires almost no additional computational effort. The effect of varying the magnitude and length‐scale of the imposed backscatter (via the backscatter coefficient and length of the filter used to generate the backscatter acceleration fields, respectively) is also investigated.