Water-channel study of flow and turbulence past a two-dimensional array of obstacles

TL;DR

This study investigates how flow and turbulence behave over a two-dimensional obstacle array simulating urban canopies, revealing the influence of aspect ratio on flow characteristics and turbulence profiles.

Contribution

It provides detailed turbulence parameter fields for various aspect ratios and compares viscous dissipation estimates, enhancing understanding of urban canopy flow dynamics.

Findings

Flow depends strongly on aspect ratio within the canyon.

Vertical momentum flux is highly affected by aspect ratio.

Different turbulence regimes influence the extent of the roughness sub-layer.

Abstract

A neutral boundary layer was generated in the laboratory to analyze the mean velocity field and the turbulence field within and above an array of two-dimensional obstacles simulating an urban canopy. Different geometrical configurations were considered in order to investigate the main characteristics of the flow as a function of the aspect ratio (AR) of the canopy. To this end, a summary of the two-dimensional fields of the fundamental turbulence parameters is given for AR ranging from 1 to 2. The results show that the flow field depends strongly on AR only within the canyon, while the outer flow seems to be less sensitive to this parameter. This is not true for the vertical momentum flux, which is one of the parameters most affected by AR, both within and outside the canyon. The experiments also indicate that, when (i.e. the skimming flow regime), the roughness sub-layer extends up to a height equal to 1.25 times the height of the obstacles (H), surmounted by an inertial sub-layer that extends up to 2.7 H. In contrast, for (i.e. the wake-interference regime) the inertial sub-layer is not present. This has significant implications when using similarity laws for deriving wind and turbulence profiles in canopy flows. Furthermore, two estimations of the viscous dissipation rate of turbulent kinetic energy of the flow are given. The first one is based on the fluctuating strain rate tensor, while the second is related to the mean strain rate tensor. It is shown that the two expressions give similar results, but the former is more complicated, suggesting that the latter might be used in numerical models with a certain degree of reliability. Finally, the data presented can also be used as a dataset for the validation of numerical models.