Influence of Street Trees on Turbulent Fluctuations and Transport Processes in an Urban Canyon: A Wind Tunnel Study

Abstract

The presence of vegetation within urban canyons leads to non-trivial patterns of the concentration of airborne pollutants, as a result of the complex structure of the velocity field. To investigate the relationship between concentration, velocity fields and vegetation density, we have performed wind-tunnel experiments in a reduced-scale street canyon, oriented perpendicular to the external wind flow, within which we placed a steady ground-level line source of a passive tracer. The aerodynamic behavior of vegetation was reproduced by inserting plastic miniatures of trees along the two long sides of the canyon, according to three different densities. The canyon ventilation was investigated by acquiring one-point simultaneous statistics of concentration and velocity over a dense grid of points within the canyon. The results show that the presence of trees hinders the upward mean vertical velocity at the rooftop, causes a reduction of the turbulent kinetic energy inside the canyon, and reduces the energy content of the large scales. The scalar concentration is conversely characterized by an enhanced level of turbulent fluctuations, whose magnitude is not dampened increasing the tree density. Within the canyon, high tree density inhibits turbulent mass fluxes, which are instead enhanced at roof level, where the mean component of the scalar flux is however hindered. A statistical analysis of concentration time series reveals that the lognormal distribution is suitable to model concentration fluctuations and extreme events, in dispersing plumes emitted by a linear source.

Appendix 1: Expressions of the Third and Fourth Moments of the Gamma, Lognormal and Weibull Distributions

The skewness and kurtosis of the gamma, lognormal and Weibull 2p distributions are reported as a function of the only parameter $C_v$:

• (1) Gamma distribution:

  $$\begin{aligned} \begin{aligned} S_k(\vartheta )&=\frac{2}{\sqrt{\vartheta }} \\ K_u(\vartheta )&=\frac{6}{\vartheta }+3, \end{aligned} \end{aligned}$$

  (8)

  where $\vartheta$ is defined as $\vartheta =C_v^{-2}$.

• (2) Lognormal distribution:

  $$\begin{aligned} \begin{aligned} S_k(C_v)&=[\exp (\ln (1+C_v)+2]\sqrt{\exp [\ln (1+C_v^2)]-1}\\ K_u(C_v)&=\exp [4\ln (1+C_v^2)]+2\exp [3\ln (1+C_v^2)]+3\exp [2\ln (1+C_v^2)]-3. \end{aligned} \end{aligned}$$

  (9)

• (3) Weibull 2p distribution:

  $$\begin{aligned} \begin{aligned} S_k(C_v)&=\varGamma \Bigl (1+\frac{3}{C_v^{-1.086}}\Bigr )C_v^{-3} a_w^3-3 C_v^{-1}-C_v^{-3}\\ K_u(C_v)&=\varGamma \Bigl (1+\frac{4}{C_v^{-1.086}}\Bigr )C_v^{-4} a_w^4-4S_k C_v^{-1}-6C_v^{-2}-C_v^{-4}, \end{aligned} \end{aligned}$$

  (10)

  where

  $$\begin{aligned} a_w=\frac{1}{\varGamma (1+C_v^{1.086})}. \end{aligned}$$

  (11)