Mechanism of Wind and Buoyancy Driving on Ventilation and Pollutant Transport in an Idealized Urban Street Canyon

Abstract

The mechanisms underlying the effects of wind and buoyancy on ventilation in urban street canyons are unclear. This study investigated the effects of facade heating on ventilation and pollutant transport in an idealized street canyon with a 1.67 aspect ratio through computational fluid dynamics simulations. The dispersion pattern of discharged hot pollutants was also studied. A primary recirculation was observed when facade heating was not applied; this recirculation was promoted in leeward-wall and ground heating cases. However, the recirculation was bifurcated into two recirculations in windward-wall heating cases, restricting ventilation. Enhanced recirculation increased the ventilation and decreased the pollution level; by contrast, air pollution increased considerably when the recirculation was bifurcated and ventilation was restricted. In the hot-pollutant case, similar results to those in the ground-heating case were observed. The hot discharged pollutant enhanced ventilation, reducing pollution. The pollutant transport mechanism was determined through an analysis of pollutant fluxes. For the one-recirculation pattern, air convection transported the pollutant from the ground level to the top boundary, and turbulent diffusion then caused pollutant removal. For the two-recirculation pattern, turbulent diffusion contributed substantially to pollutant transport both in the junction between the recirculations and through the top boundary of the street canyon.

1. Introduction

Urban areas have rapidly expanded and are characterized by densely packed high-rise buildings, which have resulted in a complex urban air environment. Building surfaces and the ground in cities are mostly constructed from materials with low reflectance, such as asphalt. These materials easily absorb heat generated by solar radiation, resulting in a rapid rise in surface temperature. The field measurements performed by Niachou et al. [1] in Athens, Greece, in 2002 revealed that the maximum surface temperature of an asphalt street could reach over than 60 °C at midday during the summer of that year. Chen et al. [2] employed a CitySim model (an urban energy model) to simulate the annual distribution of solar-induced wall temperatures inside idealized building arrays in five climate zones in China. They observed that the highest wall temperature exceeded 60 °C in Shanghai and 45 °C in other cities. For all of the cities, the highest wall temperature was substantially higher than the ambient air temperature at the hottest hour. The solar-radiation-induced heating of the walls of a street canyon is affected by the time of day and the orientation of sunshine as well as the building position (Figure 1). Chen et al. [3] carried out scaled outdoor field measurements to evaluate the influence of solar radiation on the wall temperature of two-dimensional (2D) street canyons. They determined that the maximum temperature difference between the east and west walls reached 12.7 °C. In addition to solar radiation, human activities, such as energy consumption in buildings and vehicle exhaust emissions, can exacerbate the local thermal environment and increase air pollution levels in urban areas.

Both the urban heat island and air pollution have become major problems in cities. Urban ventilation is considered a crucial countermeasure for improving the thermal environment and air quality in urban areas. Several factors can affect the ventilation performance of a street canyon, including avenue tree plantings [4,5,6], roof shapes [7,8], and canyon aspect ratio (height-to-width ratio). Cheng et al. [9] and Li et al. [10] have studied the effect of aspect ratio on airflow and pollutant dispersion inside 2D street canyons under isothermal conditions. At least two recirculations were generally observed for street canyons with an aspect ratio of >2, but only one recirculation was observed for street canyons with a unit aspect ratio. Liu et al. [11] proposed the concepts of air exchange rate and pollutant exchange rate for evaluating the ventilation and pollutant removal performance of street canyons; these metrics were extensively used in subsequent studies [9,12,13].

In addition to wind velocity, buoyancy effects may play a key role in ventilation under weak wind conditions. Studies have extensively examined the effect of buoyancy on ventilation and pollutant transport inside 2D street canyons by using computational fluid dynamics (CFD) models. These studies have focused on the thermal effect of ground heating alone [14,15,16,17,18], heating of different facades [19,20,21,22,23,24], and heating of all facades [19,22,24]. The adopted turbulence models included Reynolds-averaged Navier−Stokes (RANS) equations [14,15,16,19,21,22,24] and large-eddy simulation (LES) [17,18,20,23]. The results of these studies revealed that for a street canyon flow with only one recirculation, this recirculation weakened when the windward wall was heated. Leeward wall and ground heating usually strengthen the canyon’s recirculation. In particular, for a deep street canyon with an aspect ratio of four, a two-recirculation pattern can be changed to a single-recirculation pattern by the buoyancy effect caused by leeward wall and ground heating [21].

In addition to applying 2D idealized street canyons, scholars have used three-dimensional (3D) building blocks to investigate the effect of wall heating on air flow and pollutant dispersion. Such scholars have adopted techniques such as wind tunnel experiments [25,26,27], field observations [27], and CFD simulations [28,29,30,31]. The results of the studies confirmed the enhancement of ventilation by ground heating in a 3D urban environment. Mei and Yuan [32] provided an exhaustive review of urban buoyancy-driven air flow and its modeling methods.

The urban heat island phenomenon and urban ventilation are closely interrelated processes. Accordingly, a comprehensive understanding of the ventilation mechanism under the combined effect of wind and heat is critical for comprehending the pollutant dispersion phenomena in urban areas. Although some studies have evaluated the buoyancy effect on ventilation, the driving mechanisms of wind and buoyancy (and their interactions) require further clarification. Moreover, the influence of ventilation on pollutant dispersion is still poorly understood. Flow and dispersion are more complex in 3D street canyons than in 2D canyons because these phenomena can be affected by side ventilation; this increases the difficulty of investigating flow and dispersion mechanisms. Additionally, ventilation is difficult to generalize from a 3D environment. A 2D street canyon is a fundamental unit of a city and is the best platform for deriving a basic understanding of the ventilation mechanisms at the neighborhood scale [12]. For an idealized street canyon, the upper boundary is the only route through which air and pollutant exchange between a street canyon and external space can occur. Studies on ventilation and dispersion inside this simple geometry can thus improve the understanding of basic flow and transport phenomena.

Considering the preceding discussion, this study conducted RANS simulations with a standard k-ε (SKE) model to examine the air flow and pollutant dispersion inside a full-scale 2D street canyon with a moderate aspect ratio. The pollutant was released from a line source located in the middle of the ground. The buoyancy effects generated by the heating of different façades (ground, windward wall, and leeward wall) were considered. The objectives of the current research were to study the coupled effect of wind and buoyancy driving on ventilation and pollutant transport, to elucidate the pollutant transport process through an analysis of the pollutant flux, and to reveal the key role of turbulent diffusion on pollutant removal. The dispersion patterns of the discharged hot pollutant were also compared to draw the final conclusions.

2. Validations of the CFD Model

OpenFOAM 5.0 was used in this study. Assuming an incompressible flow, this study conducted steady-state RANS simulations using an SKE model. The governing equations were the continuity, momentum, transport equations of temperature, turbulent kinetic energy k, and turbulent dissipation rate ε. The Boussinesq assumption was adopted to handle the buoyancy term in a momentum equation. Because the pollutant content is much smaller than the air content, the dispersion of pollutants negligibly affects the air density [33]. Thus, this study solved a passive scalar transport equation to simulate the pollutant diffusion in the air. The final equations are outlined as follows:

$${\displaystyle \frac{\partial U_j}{\partial x_j}}=0$$

(1)

$${\displaystyle \frac{\partial \left(U_i{U}_j\right)}{\partial x_j}}=-{\displaystyle \frac{\partial P}{\partial x_i}}+{\displaystyle \frac{\partial }{\partial x_j}}\left\{\left(\nu +v_t\right)\left({\displaystyle \frac{\partial U_i}{\partial x_j}}+{\displaystyle \frac{\partial U_j}{\partial x_i}}\right)\right\}-{\displaystyle \frac{2}{3}}{\displaystyle \frac{\partial k}{\partial x_i}}+\left[1-\beta \left(T-T_0\right)\right]g_i$$

(2)

$${\displaystyle \frac{\partial \left(T{U}_j\right)}{\partial x_j}}={\displaystyle \frac{\partial }{\partial x_j}}\left\{\left({\displaystyle \frac{\nu }{Pr}}+{\displaystyle \frac{v_t}{P{r}_t}}\right){\displaystyle \frac{\partial T}{\partial x_j}}\right\}$$

(3)

$${\displaystyle \frac{\partial \left(C{U}_j\right)}{\partial x_j}}={\displaystyle \frac{\partial }{\partial x_j}}\left\{\left({\displaystyle \frac{\nu }{Sc}}+{\displaystyle \frac{v_t}{S{c}_t}}\right){\displaystyle \frac{\partial C}{\partial x_j}}\right\}$$

(4)

$${\displaystyle \frac{\partial \left(k{U}_j\right)}{\partial x_j}}={\displaystyle \frac{\partial }{\partial x_j}}\left\{\left(\nu +{\displaystyle \frac{v_t}{{\sigma }_k}}\right){\displaystyle \frac{\partial k}{\partial x_j}}\right\}+P_k+G_b-\epsilon$$

(5)

$${\displaystyle \frac{\partial \left(\epsilon U_j\right)}{\partial x_j}}={\displaystyle \frac{\partial }{\partial x_j}}\left\{\left(\nu +{\displaystyle \frac{v_t}{{\sigma }_{\epsilon }}}\right){\displaystyle \frac{\partial \epsilon }{\partial x_j}}\right\}+{\displaystyle \frac{\epsilon }{k}}\left(C_{\epsilon 1}{P}_k+C_{\epsilon 3}{G}_b-C_{\epsilon 2}\epsilon \right)$$

(6)

where i = 1, 2, 3 and j = 1, 2, 3 represent the three coordinate directions; U₁, U₂, and U₃ represent the mean velocities in the x, y, and z directions, respectively; P represents the mean pressure; T represents the mean temperature; T₀ represents the reference temperature; C represents the concentration in volume percentage; β represents the thermal expansion rate; and g_i represents the acceleration of gravity, which was set to (0, 0, –9.81) m/s². Moreover, Pr denotes the Prandtl number; Pr_t denotes the turbulent Prandtl number, which was set to 0.85 in this study; Sc denotes the Schmidt number; Sc_t denotes turbulent Schmidt number, which was set to 0.5 in this study; and ν and ν_t denote laminar and turbulent kinematic viscosity, respectively. P_k and G_b are the production terms caused by shear and buoyancy, respectively, and σ_k (=1.0), σ_ε (=1.3), C_ε1 (=1.44), C_ε2 (=1.92), and C_ε3 (=1.44) are the model constants.

The second-order Upwind scheme commonly used in wind engineering was adopted for all the convection terms. The SIMPLE algorithm was used for the pressure–velocity calculations. The turbulence model and numerical algorithms used in the current study were initially evaluated by comparison with the results of a wind tunnel experiment [30]. In the experiment, an urban street model including cubical building blocks (size h = 60 mm) of 14 rows × 9 columns was arranged in an unstable boundary layer. The floor temperature was controlled at T_f = 45 °C. Both the mean inflow velocity and temperature profiles satisfied logarithmic laws; U_h = 1.34 m/s was the mean velocity in the wind direction at the building height. The reference temperature difference ∆T was 30 °C. A line source with a 37-mm width was arranged at the ground between the seventh and eighth rows of the building blocks. A tracer gas (ethylene; 100% C₂H₄, 35 °C) was released to air at a volume flow rate q₀ of 8.33 × 10⁻⁶ m³/s. The reference concentration C₀ was q₀/(U_hh²).

Figure 2 illustrates a comparison of the SKE model’s predictions with experimental measurements regarding the normalized mean velocity in wind direction, temperature, and concentration along a vertical centerline located between the eighth and ninth rows of the building blocks. The mean streamwise velocity, mean temperature, and mean concentration predicted by the SKE model were in favorable agreement with the experimental measurements. The selection of an appropriate turbulent Schmidt number is essential for accurately simulating a concentration field with a RANS model. Jiang et al. [31] examined the influence of Sc_t on the calculated concentration in an SKE model. They observed that a higher Sc_t value led to an overestimation of the concentration field, and Sc_t values of 0.4–0.6 provided the best prediction of the concentration inside urban blocks. Accordingly, the Sc_t value was set to 0.5 in the current research. Figure 3 displays a comparison of the SKE model’s predictions with predictions of the LES model conducted by Jiang and Yoshie [30] regarding the mean concentration and velocity vectors inside a street canyon including the source line. In the results of both SKE and LES models, a large recirculation was obtained, and the recirculation core was located near the upper boundary. Hence, the pollutant tended to be blown to the leeward side of the street canyon. The distribution patterns of mean concentration and velocity vector predicted by the SKE model were noted to be similar to those in the LES results. Overall, the prediction accuracy of the SKE model was satisfactory, indicating that it can be used to study ventilation and pollutant dispersion in a heated environment. Cooperative projects for evaluations of turbulence models for the prediction of pedestrian wind environments have been conducted by the Architectural Institute of Japan [34,35,36]. According to the above studies, the revised k-ε models, such as the renormalization group (RNG) k-ε model and realizable k-ε model, well predict wind velocities around strong-separation regions, while the calculated wind velocities by the standard k-ε model shows better agreement with the experimental results in weak wind regions behind a building and within street canyons. One advantage of using a RANS model is that it enables full-scale calculations because a wall function can be used near the wall to reduce the mesh number.

3. Simulation Settings and Boundary Conditions

This study conducted 2D full-scale simulations to investigate the flow and pollutant dispersion phenomena. Ten rows of building blocks with a height H of 10 m were arranged in the domain (Figure 4a). The distance between the buildings L was 6 m. The aspect ratio of the street canyon H/L was thus 1.67. For simulating the 2D flow, the height of the simulation domain was 10H, and the width of the domain was 0.1H. The inlet boundary was located 4H ahead of the first row of building blocks, and the outflow boundary was located 12H downstream of the last row of building blocks. The space between the fifth and sixth rows was selected as the target street canyon to investigate the flow and dispersion phenomena. The facade definitions (leeward wall, ground, and windward wall) of the target street canyon are presented in Figure 4b. A tracer gas was discharged from a line source located in the middle of the ground of the target street canyon; the gas was released at a volume flow rate q of 2 × 10⁻⁴ m³/s. C* = q/(U_HH²) was the reference concentration, where U_H is the inlet wind velocity at the building height position and was set to 2 m/s.

Table 1. Simulation cases.

Cases	Wall Heating Position	Type of Released Pollutant
Case 1	No heating	Cold pollutant (25 °C)
Case 2	Leeward wall	Cold pollutant (25 °C)
Case 3	Ground	Cold pollutant (25 °C)
Case 4	Windward wall	Cold pollutant (25 °C)
Case 5	No heating	Hot pollutant (60 °C)

presents a summary of the simulation cases. A total of five simulation cases were considered. For all simulations, the incoming air temperature T_a was controlled to 25 °C. For Cases 1–4, a cold pollutant with a temperature of 25 °C was released from the line source. Case 1 was the base case without any facade heating (neutral condition) and was used for comparison with the other cases. For Cases 2–4, different walls of the target street canyon were heated to 50 °C to analyze the facade heating effect on flow and pollutant dispersion. For Case 5, the pollutant was released at a high temperature of 60 °C to study the buoyancy effect caused by the hot pollutant. A hexahedral mesh system with 2.2 million cells was used for all of these simulations. The distance from the center of the first cell to the wall was controlled to 40 mm to satisfy the standard for using a wall function. The mesh growth rates were controlled to <1.25. Finer meshes were used for the target street canyon to accurately capture the flow structures (Figure 4c).

The inflow velocity profile and turbulent profiles recommended by the Architectural Institute of Japan for wind engineering applications [37] were adopted as the inflow conditions. These are described as follows:

$$U_z=U_H{\left({\displaystyle \frac{z}{H}}\right)}^{\alpha }$$

(7)

$$k={\left(I_z\cdot U_z\right)}^2$$

(8)

$$I_z=-0.1\times {\left({\displaystyle \frac{z}{z_G}}\right)}^{-\alpha -0.05}$$

(9)

$$\epsilon =C_{\mu }^{1/2}\cdot k\cdot {\displaystyle \frac{d{U}_z}{dz}}$$

(10)

where H is the building height; U_H is the mean streamwise velocity at height H, which was set to 2 m/s in this study; and α is the power-law exponent determined by terrain category, which was set to 0.15 in this study to simulate B-type terrain roughness in Chinese code [38]. Moreover, I_z is the turbulence intensity at height z, and z_G is the height of the boundary layer, which was set to 350 m for B-type roughness.

A zero-gradient condition was set for the outlet boundary, and a symmetry condition was set for the top boundary. The domain width was 0.1H, and symmetry conditions were given for the two side boundaries to imply that the geometry was infinitely long in the spanwise direction (simulating a 2D flow). Wall conditions were used for the building surfaces and ground, and wall functions were adopted to model the near-wall flow and heat transfer; this is the best opinion when using high-Reynolds-number RANS models in an engineering application. Steady-state simulations were performed using the SKE model. The solved governing equations and adopted numerical algorithms were introduced in Section 2.

4. Results and Discussion

4.1. Effects of Facade Heating on Ventilation and Pollutant Transport

Figure 5 displays the temperature difference ∆T = T − T_a between the target street canyon and external space for the three facade heating cases (Cases 2–4). Case 2 (leeward wall heating) had a higher temperature near the side of the leeward wall, especially in the region near the top boundary. In Case 3 (ground heating), the air temperature was higher in the region near both the ground and leeward wall. In Case 4 (windward-wall heating), the entire target street canyon was filled with hot air. Figure 6 displays the distributions of streamlines and the normalized velocity magnitude U_mag/U_H for Cases 1–4. For Case 1 (no facade heating), the wind entered the street canyon from the top corner of the windward wall and formed a large primary recirculation.

Table 2. Number of recirculations inside 2D street canyons with aspect ratios of 1–2 under isothermal conditions.

References	Aspect Ratio H/L	Scale	Turbulence Models	Recirculation Number
This study	1.67	Full	Standard k-ε	1
Xie et al. [7]	1	Reduced	Standard k-ε	1
Huang et al. [8]	1	Reduced	Standard k-ε	1
Cheng et al. [9]	1 2	Reduced	RNG k-ε	1 2
Cheng et al. [16]	1	Reduced	RNG k-ε	1
Cheng and Liu [17]	1	Reduced	LES	1
Li et al. [18]	1 2	Reduced	LES	1 2
Xie et al. [19]	1 2	Reduced	RNG k-ε	1 2
Chen et al. [21]	1.1	Reduced	RNG k-ε	1
Cintolesi et al. [23]	1	Reduced	LES	1

presents a comparison of this study’s results with recirculation numbers from the literature for 2D street canyons with aspect ratios of 1–2 under isothermal conditions. For most cases involving an aspect ratio of <2, one recirculation was observed; two recirculations were observed in cases involving an aspect ratio of 2. In this study, the aspect ratio of the street canyon was 1.67; consistent with the literature, only one recirculation was observed. The present study investigated the impacts of buoyancy on ventilation and pollutant transport inside a 2D street canyon with a moderate aspect ratio. Thus, the aspect ratio of the street canyon was selected to be large to increase the buoyancy effect while being sufficiently small to ensure that the street canyon had only one recirculation under isothermal conditions. Compared with that in Case 1, the velocity magnitude in Cases 2 and 3 was clearly larger near all of the walls, indicating that the canyon recirculation was stronger. In Case 4, the primary recirculation was bifurcated into a top clockwise recirculation and a bottom counterclockwise recirculation.

Figure 7 illustrates diagrams of the influence of wind and buoyancy on recirculation inside the street canyon. In Case 1 (Figure 7a), the oncoming wind entered the deep street canyon after impinging the top corner of the windward wall and formed a large clockwise recirculation. The whole street canyon was under weak wind conditions, as observed from the magnitude of the wind velocity displayed in Figure 6a. Thus, buoyancy could affect ventilation. In a gravitational field, any hot air could move upward because of its low density; this can be seen from the ideal gas law. Accordingly, the influence directions were the same for both wind and buoyancy when the leeward wall was heated (Figure 7b); this enhanced the recirculation compared with that under the neutral condition, where the wind was the only driving force. The influence directions of wind and buoyancy were opposite when the windward wall was heated (Figure 7d); hence, the primary recirculation caused by the wind was weakened by the buoyancy effect, resulting in bifurcation into two recirculations. This resulted in poor ventilation because the fresh air coming from the top boundary could not reach the bottom of the street canyon. This driving mechanism was also identified by Mei and Yuan [32]. A strong interaction between wind and buoyancy driving was clearly observed from the ground-heating case (Figure 7c). In this case, the recirculation first blew the heated air near the ground to the leeward wall, and the buoyancy effect then facilitated the ventilation near the leeward wall. In the second stage, the driving mechanism was similar to that in the leeward-wall heating case (Figure 7b).

Ventilation can strongly affect pollutant dispersion inside a street canyon. Figure 8 displays the normalized concentration distributions for Cases 1–4. In Case 1 (Figure 8a), the discharged pollutant from the ground was blown to the leeward side by the recirculation. Consequently, the pollution level was higher near the leeward wall of the street canyon. The pollutant was partly transported out of the street canyon. In Cases 2 and 3, the pollutant concentration was lower because these cases had better ventilation. In Case 4, the pollutant had a high concentration throughout the street canyon (Figure 8d) because two recirculations were formed in this case. Some amount of the pollutant discharged from the ground accumulated in the lower region of the street canyon because of the bottom recirculation shown in Figure 6d.

Figure 9 presents the distributions of normalized velocity magnitude and the pollutant concentration near the wall of the street canyon for Cases 1–3. To evaluate the mesh sensitivity, Case 1 was again simulated using a coarser mesh (1 million cells). The results were only slightly affected by the mesh system and were insensitive to the mesh density for the RANS turbulence model. Compared with those in Case 1, the velocity magnitude clearly increased and the pollutant concentration decreased in Cases 2 and 3.

4.2. Analysis of the Pollutant Fluxes

This study explored the ventilation and pollutant transport mechanism by analyzing the vertical velocity and pollutant fluxes. In turbulence theory, the total pollutant flux can be decomposed into its convective and turbulent components as below:

$$〈u_{j}c〉=〈u_j〉〈c〉+〈{u^{\prime }}_j{c}^{\prime }〉$$

(11)

where u_j (= u, v, w) is the instantaneous velocity in the x, y, and z directions; c is the instantaneous concentration; and u′_j and c′ are the velocity and concentration fluctuations, respectively. The brackets “< >” indicate a time average. For a 2D street canyon, the upper boundary is the only route through which air and pollutants can be exchanged between the canyon and the external environment. Accordingly, the mean vertical velocity <w> (or W) and total pollutant flux in the vertical direction <wc> are key metrics that reveal the ventilation and pollutant transport mechanisms in a canyon. For each mesh in the top boundary, multiplying W by the face area can yield the air flow rate in this mesh; similarly, multiplying <wc> by the face area can yield the pollutant flow rate through the mesh. In an LES model, all three terms in Equation (11) (total pollutant flux, convective flux, and turbulent flux) can be easily obtained by implementing time averaging during a simulation. However, only the mean flow variables can be obtained in a steady-state RANS simulation; obtaining the instantaneous and fluctuating values of a variable is difficult. In a RANS model, the following eddy viscosity model (gradient diffusion hypothesis) [39] is usually adopted to calculate the turbulent flux:

$$〈{u^{\prime }}_j{c}^{\prime }〉=-{\displaystyle \frac{{\nu }_t}{S{c}_t}}\cdot {\displaystyle \frac{\partial C}{\partial x_j}}$$

(12)

Figure 10 displays the normalized mean vertical velocity and total pollutant flux along the top boundary of the target street canyon for Cases 1–3. A negative vertical velocity was observed on the right side of the top boundary, signifying that oncoming wind entered the street canyon; a positive vertical velocity was observed on the left side of the top boundary, indicating that air flowed out of the street canyon. For steady flow, the amount of air entering the street canyon must equal the amount exiting it. Heating of the leeward wall (Case 2) and the ground (Case 3) both clearly increased the magnitude of the vertical velocity at the top boundary, signifying that recirculation was enhanced for these two cases, leading to better ventilation and a reduction in pollution inside the street canyon (Figure 8 and Figure 9). The total vertical pollutant flux was positive along the entire top boundary for all three cases, implying that the pollutant was leaving the street canyon. For steady flow, the total amount of pollutants leaving the street canyon from the top boundary must equal the number of pollutants discharged from the line source in the ground. Because these amounts are the same, the number of pollutants leaving the top boundary must be the same for all the cases. Accordingly, in Cases 2 and 3, the increase in pollutant outflow on the left side led to a reduction in the pollutant outflow on the right side.

Figure 11 presents the distributions of vertical convective and turbulent fluxes along the top boundary for Cases 1–3. The distributions of the convective fluxes were similar to the distributions of vertical velocity in Figure 10a. A positive convective flux was observed on the left side of the top boundary, signifying that the pollutant was transported out of the street canyon through air convection, and a negative convective flux was observed on the right side of the top boundary, signifying that some of the pollutant was brought back into the street canyon through air convection. The enhancement of ventilation in Cases 2 and 3 increased the effect of air convection on pollutant exchange. The vertical turbulent flux was positive along the whole top boundary, and its distribution was similar to that of the total pollutant flux shown in Figure 10b. The largest vertical turbulent flux was observed near the leeward side, and its magnitude gradually decreased along the positive x direction. This positive turbulent flux indicates that the pollutant was transported out of the street canyon by turbulent diffusion throughout the top boundary. A comparison of the results in Figure 11a,b indicated that the magnitude of turbulent flux was much larger than the magnitude of convective flux, implying that turbulent diffusion dominated pollutant transport in the top boundary. Although the convective flux was negative on the right side of the top boundary, its magnitude was much smaller than the positive turbulent flux. Hence, the pollutant was not brought back into the street canyon at any location on the top boundary. The large magnitude of turbulent flux is attributable to the difference in the pollution level between the street canyon and external space. Compared with that in Case 1, the turbulent flux was slightly smaller in Cases 2 and 3 because these cases had lower pollution levels inside the street canyon.

Figure 12 illustrates the normalized pollutant fluxes inside the street canyon for Case 1 with only one recirculation. The convective flux was positive near the leeward wall and negative near the windward wall. The magnitude of the convective flux was much larger along the leeward wall because the pollution level was higher in this region. Compared with the convective flux magnitude, the magnitude of turbulent flux was much larger in the top boundary and was much smaller throughout the street canyon. On the basis of this finding, this study proposed the following pollutant transport process: Air convection first transported the pollutant upward to the top boundary along the leeward wall, and turbulent diffusion then caused pollutant removal at the top boundary. The LES simulation conducted by Cai et al. [20] also revealed this transport process.

Figure 13 presents the normalized pollutant flux inside the street canyon for Case 4 with double recirculation. Because the signs of the convective flux and vertical velocity were the same, a large positive convective flux occurred near the lower region of the windward wall and the upper region of the leeward wall. The turbulent flux was positive throughout most of the street canyon, and its peak value was at the top boundary and the junction of the two recirculations shown in Figure 6d. The pollutant transport process inside the street canyon with two recirculations must therefore include four stages. In the first stage, the bottom counterclockwise recirculation transported the discharged pollutant upward to the middle of the street canyon along the windward wall. In the second stage, the pollutant could not be transported further upward by air convection because of the bottom recirculation; thus, turbulent diffusion instead further transported the pollutant upward at the junction of the two recirculations. In the third stage, the top clockwise recirculation then moved the pollutant upward to the top boundary along the leeward wall. In the final stage, turbulent diffusion transported the pollutant out of the street canyon at the top boundary. The gradient diffusion hypothesis presented in Equation (12) indicates that the effect of turbulent diffusion on pollutant transport is strongly related to the gradient of mean pollutant concentration. Figure 14 displays the distributions of the negative normalized concentration gradient along three vertical lines inside the street canyon. In Case 1, peak values of the concentration gradient were clearly observed near the top boundary; the peak value was larger near the top of the leeward side; this is consistent with the distributions of turbulent flux shown in Figure 12b. Two peak values of the concentration gradient were observed near the top boundary and in the region between z/H = 0.5–0.75 in Case 4, implying that turbulent diffusion contributed substantially to pollutant transport in the vertical direction in these regions.

As mentioned, the pollution level was lower when the leeward wall and ground were heated because ventilation was enhanced. However, this enhancement of ventilation did not cause more pollutants to be transported out of the street canyon; this is because, for steady-state flow, the pollution leaving the top boundary must equal the amount of pollutant discharged from the line source in all cases. The enhanced ventilation increased pollutant removal near the leeward side but decreased the pollutant removal near the windward side of the top boundary. The reduction in the pollution level is likely because the increase in ventilation caused more fresh air to entire the street canyon; thus, the pollutant inside the street canyon was diluted by the fresh air. Accordingly, the pollutant transport process inside a street canyon is determined by both air convection and turbulent diffusion and turbulent diffusion plays an essential role in pollutant removal. Although the ventilation increased when the leeward wall and ground were heated in this study, the contribution of turbulent diffusion to pollutant removal decreased when the pollution level decreased inside the street canyon.

4.3. Dispersion of Hot Pollutant

The flow and distribution pattern for the discharged hot pollutant (60 °C; Case 5) were assessed, and the results were compared with those obtained for Case 1. Figure 15 presents the streamlines and normalized velocity magnitude, temperature difference, and normalized concentration for Case 5. Only one recirculation was observed, and this recirculation was enhanced compared with that for Case 1. The hot pollutant mixed with the cold air when it was discharged from the line source, and this hot mixture with a high pollutant concentration was blown to the leeward side of the street canyon by the primary recirculation. This drove the recirculation on the leeward side, enhancing ventilation and reducing the pollution inside the street canyon. This situation was similar to that observed in the case of ground heating. Both the concentration and temperature were high on the leeward side of the street canyon because of the clockwise recirculation. The distribution patterns of concentration and temperature were similar; this is consistent with the use of regions with a high air temperature to detect smoke propagation in some fire simulations.

5. Conclusions

This study investigated the airflow and pollution diffusion inside a 2D street canyon with an aspect ratio of 1.67 under wind and thermal coupling conditions. The CFD technique adopted in this study was validated by comparison with experimental results. The impacts of different facade heating schemes on ventilation and pollutant transport were investigated. The driving mechanisms of wind and buoyancy were clarified. The pollutant transport processes for one and two circulation patterns were compared, and the critical effect of turbulent diffusion on pollutant removal was revealed. The main conclusions of this study are outlined as follows:

A primary recirculation was observed inside the street canyon in Case 1 (no facade heating). The heating position affected whether thermal buoyancy promoted or inhibited the recirculation flow. Leeward wall heating (Case 2) and ground heating (Case 3) could accelerate the recirculation, but the recirculation was restricted and bifurcated into two recirculations when the windward wall was heated (Case 4). Compared with those in Case 1, the ventilation was enhanced and the pollution level decreased in Cases 2 and 3. In Case 4, the ventilation was poor, which led to a high pollution level.

The pollutant transport process was determined through an analysis of pollutant fluxes. For single-recirculation patterns (Cases 1–3), air convection first transported the pollutant upward along the leeward wall, and turbulent diffusion then caused pollutant removal through the top boundary. For the two-recirculation pattern (Case 4), the bottom counterclockwise recirculation transported the pollutant to the middle of the street canyon, and turbulent diffusion transported the pollutant further upward. The pollutant was then moved to the top boundary by the top clockwise recirculation, where it was removed through turbulent diffusion. Thus, the turbulent diffusion contributed significantly to pollutant removal and dominated pollutant transport at the top boundary.

The results obtained in Case 5 (hot pollutant case) were similar to those obtained in Case 3; the recirculation inside the street canyon was strengthened, and the ventilation was enhanced. Therefore, the pollution level inside the street canyon decreased.