Evaluating Dust Particle Transport Performance within Urban Street Canyons with Different Building Heights

In developing cities, buildings with different heights obstruct the diffusion of pollutants. In this study, dust particles transported within urban street canyons were simulated using Reynolds-averaged Navier–Stokes (RANS) method, and airsolid two-phase flow fields were obtained on a regional scale. Four typical street building models were used in this study: (a) low-rise buildings (H/b = 1), (b) step-up building arrangements (H/b = 1, 2, 3, 4, 5), (c) step-down building arrangement (H/b = 5, 4, 3, 2, 1), and (d) high-rise buildings (H/b = 5). The particle volume fraction distribution in the four models reflected the basic properties of particle transportation in the canyons. Vortices were observed on the roofs of street canyons, which prevented particles from being transmitted into the canyons, and the vortex regions were characterized as low particle concentration. To evaluate the dust particle transport performance in the models, three indices, namely particle transport efficiency, suspension fraction and suspension density, were defined. These concepts were based on the particle number concentrations, which were obtained using the Lagrange approach. The high-rise building model (H/b = 5) demonstrated the lowest transport efficiency among all four models, and it also had the lowest suspension density in the street canyons. This implied that the high-rise buildings hindered the particles from being transporting further and that the particles could not enter the deep canyons easily. In the step-down building arrangement model (H/b = 5, 4, 3, 2, 1), the particle concentration level and suspension density in the canyons were lower than those observed in the step-up building arrangement model (H/b = 1, 2, 3, 4, 5), indicating that a step-down building height arrangement is appropriate for creating construction plans for developing cities. Finally, the variation of the streamwise air velocity, vertical velocity and fluctuating velocity on the roof of canyons along the x direction were separately examined. Notably, the fluctuating velocity was the dominant mechanism of the particle suspension in the canyons.


INTRODUCTION
Particular concerns have been raised about air pollution within urban areas during recent ten years.The importance of such investigations is associated with the increased levels of air pollutants in the atmosphere around urban areas.This increase has various causes, such as the continuing expansion of existing industries and the increased use of motor vehicles coupled with population growth, particularly in large urban areas (Bady et al., 2008).People who breathe in air pollutants, such as ozone, nitrogen oxides, or particulate matter (PM), may suffer from respiratory disease and even lung cancer (WHO, 2000).Microparticles from iron industries are inspired in the human respiratory tract, and the smaller the size of microparticles is, the more they invade human cells.The total surfaces of smaller particles are sufficiently large to absorb a high amount of pollutants (Kleinstreuer et al., 2008).
Two types of pollution sources exist: high-level sources, such as tall stacks from industry plants, and low-level sources, such as automobile exhausts.Regarding high-level sources, the Gaussian plume model (Kanda et al., 2006) is typically used to estimate pollutant concentrations; at such levels, obstacles (such as buildings) have little effect on the diffusion characteristics of pollutants.In addition to Gaussian type models, high-level sources can be simulated using computational fluid dynamics (CFD), which is more appropriate for complex situations.For example, Scungio et al. (2015) simulated the dispersion of ultrafine particles in emissions from incinerator plant stacks by reproducing the atmospheric boundary layer by using the k-epsilon (k-ε) Reynolds-averaged Navier-Stokes (RANS) turbulence model.However, in the CFD approach applied in this study, suspended PM was considered one of gaseous pollutants and the particle dynamics properties were ignored.Therefore, this approach could not be used to simulate the trajectories of particles and predict the particle concentration (Holmes and Morawska, 2006).
In practice, according to gas-solid two-phase flow, the application of CFD for predicting PM movement in closed flow fields has fully developed.For example, indoor particle sediment and concentration distribution can be obtained using the Detached Eddy Simulation model with a modified Lagrangian method (Wang et al., 2011).Mei (2012) predicted the erosion rate of fluid machinery induced by particle impacts.Furthermore, Farkas and Balashazy (2008) calculated particulate transportation in the human respiratory tract.The force function between particles and gas constituted their theoretical basis.Dust particulates migrate in the atmosphere, driven by wind; therefore, air and PM form gas-solid twophase flows in the atmosphere.The Lagrange method can be used to calculate particle movements in an urban space and estimate pollutant concentrations (Holmes and Morawska, 2006;Anfossi, 2010).
Regarding low-level sources, the structure of multibuilding obstacles in urban areas has a major effect on the diffusion of pollutants.Numerous researchers have focused on gaseous pollutant diffusion mechanisms and impact factors in ideal urban street canyons.Such impact factors include the height difference of buildings (Li et al., 2008a;Liu et al., 2011), street length and width (Hang et al., 2010), building density (Scargiali et al., 2011), and vegetation covering both sides of a street (Buccolieriet al., 2011).Several studies have focused on the influence factors of atmospheric pollutant diffusion within high-rise buildings (Li et al., 2010;Hang et al., 2012), and the findings of these studies have provided suggestions for city planning.Park et al. (2015) recently explored the dispersion of reactive pollutants in street canyons by using a coupled CFD-chemistry model.However, this study focused on evaluating particle transportation in a regional atmospheric environment by using the CFD Lagrange method and investigating the effect of urban building heights and arrangement on the airflow and dust particle transportation, which has been rarely studied.
This study aimed to evaluate the particle matter transport performance within street canyons.Determining only pollutant concentrations is insufficient for obtaining a complete picture of the transportation characteristics in city domains.Specifically, if the pollutant source changes, the concentration distributions also change.In this instance, understanding the removal capacity of pollutants by the wind within city domains is difficult.To thoroughly evaluate the removal performance of PM by the natural wind within such domains, other parameters must be considered in addition to the concentration distribution.Consequently, this study adopted a group of indices that describe the indoor particle deposition and diffusion for evaluating the outdoor domain, such as particle transport efficiency, suspension fraction, and suspension density.

Computational Model
In this study, Ansys CFX commercial code was used to execute the CFD calculation.Turbulence models are the foundation of numerical simulations, and researchers have compared the performance of the renormalization group (RNG) k-ε model (Cheng et al., 2009), standard k-ε model, Reynolds stress model (Salim et al., 2011), and one-equation Spalart-Allmaras model (Scungio, 2013) in evaluating atmospheric pollutant dispersion in urban streets.Each turbulence model has advantages and limitations.In the current study, the incompressible RANS RNG k-ε turbulence model (Yakhot et al., 1986) was used, because Cheng (2008Cheng ( , 2009) ) validated its applicability to evaluating street canyons.In the following model equations, the overbars on the variables represent the ensemble averaged quantity that is commonly employed in RANS modeling.The governing equations comprise the continuity and momentum conservation equations, as shown in Eqs. ( 1) and (2), respectively: where i u is the fluid velocity tensor, x i is the spatial coordinate, p is the kinematic pressure, and ν is the kinematic viscosity.Eqs.(1) and (2) are expressed in tensor notation, and the typical summing convention on repeated indices (i, j = 1, 2) is employed.The Reynolds stress tensor R ij is modeled using the eddy-viscosity model: where k is the turbulent kinetic energy (TKE), ν t is the turbulent kinematic viscosity, and δ ij is the Kronecker delta function.For the closure of the RNG k-ε turbulence model, the transport equations for TKE (Eq.( 4)) and the dissipation rate of TKE (Eq.( 5)) are evaluated: where ν eff = ν + v t is the effective kinematic viscosity, and α k and α ε are the inverse effective Prandtl numbers.Moreover, v t = c v k 2 /ε is assumed using the RNG theory.
The TKE production P k in the second term on the righthand sides of Eqs. ( 4) and ( 5) can be expressed as follows: The modeling constants for the RNG k-ε turbulence model are σ k = 1, c ν = 0.09, P r = 0.85, c 1ε = 1.42, c 2ε = 1.68, For microparticles (d p ≥ 1 µm), a Lagrange frame of reference can be used to analyze the trajectory of particles in the atmosphere, and the effects of diverse forces such as Brownian, Saffman, drag, and gravity can be assessed.When the flow domain is relatively complicated, this code can be used in conjunction with fluid flow solvers.Hence, at each fluid flow time step, this code interacts with flow properties computed by the solver, and the particle positions advance with time (Saidi et al., 2014).Because of the high particleto-air density ratio, dilute particle suspensions, negligible Brownian motion force, and thermophoretic forces, drag and gravity are considered the dominant forces away from the wall (Kleinstreuer et al., 2008).The dynamic equations of particulate transport can be expressed as follows: 3 ( ) 6 ( ) where u i P is the particle velocity, ρ is the fluid density, ρ P is the particle density, d P is the particle diameter, S is the density ratio between a particle and adjacent fluid, ν is the kinematic viscosity, δ is the unit delta function, τ is the particle relaxation time, and d ij = (u ij + u ji )/2 is the deformation rate tensor (Kao et al., 2009).

Computational Domain and Boundary Conditions
This study investigated the flow and particle transport properties of a two-dimensional (2D) idealized street canyon with various building heights.The computational domain (Fig. 1) measured 200 (H) × 1000 (W) m and comprised four identical street canyons under a free surface layer.The widths of both the street canyon and building were b = 20 m, and the building height was H b = n × b (n = 1, 2, 3, 4, 5).The extended length behind the final building was set to Le = 720 m to ensure a fully developed turbulence at the outlet.Four typical street models were used in this study: (a) lowrise buildings (H/b = 1), (b) step-up building arrangement (H/b = 1, 2, 3, 4, 5), (c) step-down building arrangement (H/b = 5, 4, 3, 2, 1), and (d) high-rise buildings (H/b = 5).These geometric models represent an undeveloped city with ranch houses, developing city with varying building heights, and developed city with a cluster of high-rise buildings, respectively.
The free-stream flow over the street canyons is driven by an inflow boundary with a constant stream-wise velocity  An industry source with a constant particle emission rate (Q = 17.5 kg s -1 ) was placed at the inlet of the domain, and the particle size ranged from 2.5 × 10 -6 to 10 × 10 -6 m.Because the main objective of this study was to evaluate the naturalwind-driven transport performance of dust particles from high-level sources and the effect of building groups on particle movement, no other pollutant sources (e.g., from vehicles emissions) were considered in the street canyons.
The free-slip boundary condition was used at the top of the free surface layer.This height is sufficient for the development of a free-stream wind layer at a constant inflow wind speed.The no-slip boundary condition was applied to all the walls, roofs, and ground of the street canyons.This study generated a hexagonal mesh by using commercial ICEM software.The grid sensitivity test was conducted from coarse mesh composed of 85800 nodes to fine mesh composed of 607026 nodes.The air velocities along the line of x = 20 m, y = 0-20 m (in the centre of the first canyon) were used to examine the grid-independence, and the results were illustrated in Fig. 2. From Fig. 2, when mesh scale exceeded 331384 nodes, the calculated results satisfied the grid independent demand.More than 330000 nodes were used to discretize the computational domain.A finer mesh, measuring 2.5 × 10 -5 of the minimum elements, was applied to the vicinity of the solid walls to accurately capture the velocity boundary layer, whereas a coarser mesh was used near the outlet and upper boundaries.All the calculation were performed on computational grids obtained from a sensitivity.

Particle Transport Parameters
The concepts of deposition efficiency, deposition fraction, and deposition density were originally developed for quantifying particle deposition in human tracheobronchial models (Farkas and Balashazy, 2008;Kleinstreuer et al., 2008).Kao et al. (2009) used the concept of removal efficiency to evaluate PM transport behavior in multiple rooms with closed system boundaries.Although the atmospheric environment in urban areas has an open boundary, it was considered a closed area with an elastic boundary in this study.Fine particles have a low inertia, and they always diffuse along with the airflow and are suspended in the atmosphere, instead of being deposited on walls (Lai and Chen, 2007;Kleinstreueret al., 2008;Kumar et al., 2011).Therefore, three indices, namely transport efficiency, suspension fraction, and suspension density, were used to evaluate the particle transport performance and the influence of building height and arrangement.The definitions of these indices are presented as follows:  Particle transport efficiency = the number of particles escaping from the computational domain/the number of particles injected into the computational domain. Particle suspension fraction = the number of suspended particles in the assessed region/the number of particles entering the region. Particle suspension density = the number of particles suspended in a defined region (e.g., one street canyon)/ the surface area of that region.
Particle transport efficiency quantifies the inhibitory effects of building groups on particle movement.The suspension fraction value can be used to compare the relative number of particles in the each street canyon.The suspension density characterizes the number of particle in a unit area for excluding the effect of different areas of street canyons.

MODEL VALIDATION
The wind field of the RANS RNG k-ε turbulence model used in the current study was validated using water channel experiments (Baik and Kim, 2002) and simulations (Baik and Kim, 2002;Cheng et al., 2008).For the water channel experiment, the street canyon was constructed using two building models measuring 0.1 m (L) × 0.1 m (W) × 0.4 m (H), which were positioned 0.1 m apart to form an isolated street canyon with h/b = 1.According to the model validation procedure of Baik and Kim (2002), the measured and simulated vertical velocity profiles on the upwind (0.15 b measured from the leeward wall) and downwind (0.15 b measured from the windward wall) sides inside a street canyon were compared (Figs.2(a) and 2(b)).Fig. 3 shows the numerical results of Cheng (2008) and the present study.The variation trends of the vertical wind profiles on the upwind and downwind sides of the four results were almost identical.Therefore, in the current study, the CFD results were considerably close to those of the water channel experiment.
The roof-level mean vertical velocity calculated using the 2D k-ε turbulence model was also compared with those derived in a previous numerical simulation (Cheng et al., 2008) and water channel experiment (Li et al., 2008b) (Fig 4).The comparison suggested that the mathematical and physical models, the calculation approach and mesh scale used in the current study were scientifically suitable for evaluating street canyons.

Air Flow Streamline and Velocity Vector in the Street Canyon
During the process of rebuilding old cities, a large number of new high-rise buildings and old ranch houses constitute the uneven underlying surface.Fig. 5 shows a streamline of air flow in the region at 0 < X < 500 m at a constant inlet velocity of u 0 = 1 m s -1 .
The interaction between an approaching atmospheric flow and a city results in complicated flow patterns between buildings, along streets.An air mass approaching a city can either enter the streets, flow above buildings, or around  buildings (Buccolieri et al., 2010).As shown in Fig. 5, a reversed flow occurred and a wake was formed behind the final building when the air flowed through the building group.PM was transported around the vortex and could not penetrate the core of the cell in the area.
When air encountered obstacles, the flow speed and direction changed and a vortex was formed on top of the buildings (Figs.5(a), 5(c), and 5(d)).The recirculation cell on the roof of buildings was a stagnation zone and did not engage in mass exchange with the upper air because the air cushion prevented the particles from penetrating the canyons.As illustrated in Fig. 5(b), the airstream rose along the building roofs instead of forming eddies, and the dustloaded air flowed into the canyons; therefore, every canyon had the chance to "obtain" air with particles.
Fig. 6 illustrates air velocity contours and vectors in the street canyons.In model a, the airflow in the canyon produced a vortex, and the flow velocity direction on the roof was opposite to the inflow direction.This airflow structure leads to the suspension of particles in the canyon atmosphere and making it difficult for them to diffuse.Therefore, PM enters rooms through natural ventilation and affects indoor air quality (Poupard et al., 2005;Yang et al., 2015).
In model b, the incremental building height accelerated the airflow on the roof of the canyons along the x direction; by contrast, it slowed down the airflow at the bottom of the canyon, decreased along the x direction.This led to the difficulty of the diffusion of PM that has entered the canyons before.Moreover, the airflow variation in model c was contrary to that in model b, meaning that the decreasing building group plays a role in blocking air transmission and reducing its speed.
As shown in model d, H/b = 5 is a typical deep canyon.In this deep canyon, the velocity is lower than 0.07u 0 at the bottom of the canyon.If H/b exceeds unity, the flow velocity in the canyon decreases; therefore, once PM enters the canyon, it is difficult to discharge and is consequently suspended in the air (Li et al., 2008;Scungio et al., 2013).

Evaluate Particle Transport Performance Particle Matter Volume Fraction
Fig. 7 illustrates the distribution of particle volume fraction in the computational domain.Under the inlet condition, which involved setting particles uniform injection with the air, particles were mainly distributed in the upper space of the computational domain.Because of the obstruction caused by buildings, the particulate volume fraction behind the buildings was extremely low.In addition, the particle volume fraction on the roofs of the buildings in models a, c, and d was low.Compared with the results in Fig. 5, these areas represented the location of eddy presence.Specifically, the vortex drove the horizontal transportation of the PM and hindered its entry into the core of the cells.
As shown in Fig. 7(c), a high particle fraction zone was observed above buildings 3-5.A comparison of this zone with the streamlines in Fig. 3(c) revealed that the high particle fraction zone appeared between the two reverse eddies.This thus confirmed that the two reverse eddies drove the transportation of particles and increased the particle volume fraction at the location.

Particle Transport Efficiency, Suspension Fraction and Suspension Density
Varying building heights from low to high in developing cities causes positive or negative effects on PM transmission.First, to evaluate the effects of building groups on particulate suspension and dispersion in street canyons, this study focused on the particle concentration distribution in the location.In this paper, the number of particles in the computational domain is an essential parameter for evaluating air quality because although microparticles have a low mass concentration, they have a high number concentration.The concepts of particle transport efficiency, suspension fraction, and suspension density are based on the number concentration.
This study defined the extent of the evaluated domain as 0 < X < 280 m and 0 < Y < 200 m (Fig. 1, dotted lines).As  shown in Table 1, the minimal particulate transport efficiency in the evaluated domain was observed in model d (H/b = 5) and the maximum value was observed in model a (H/b = 1).The building arrangement of H/b = 1 was favorable for the efficient dispersion of particles.This is because as the building height increased, the particle transmission was hindered.However, the air quality in the street canyon of the highrise buildings did not deteriorate.As illustrated in Fig. 7, the particle concentration distribution was not uniform in the evaluated domain; the particle concentrations in the atmosphere above the buildings were high, but those in the canyons, which are people's actual living environment, were low.Therefore, a PM suspension fraction could be used to represent the suspension and residence of particles in different areas.
Fig. 8 shows a comparison of particle suspension fractions in the canyons of various street models, where I-IV represent four street canyons in sequence and V represents the atmospheric zone above the buildings (Fig. 1).As shown in Fig. 8, although model d demonstrated the lowest particulate transport efficiency among all the models, 90% of the particles in this model resided in area V, only 10% of the particles were suspend in the canyons; therefore, this model did not exhibit the lowest air quality in people's living environment.In model a, although the transport efficiency was high, the canyons constituted 32% of the particle suspension fraction.In particular, in model b, the particle suspension fraction in the canyons was 68%, indicating that the structure of model b was hindered for particle diffusion and that it severely affected the air quality in the living environment.For model c, the transport efficiency was 35.4%, but canyons I and II had a low level of particulates.Therefore, urban building groups of decreasing heights are beneficial for providing high air quality in people's living environment.
Fig. 9 illustrates a comparison of suspension density of PM in the canyons and above the buildings for different street structures.Because the street canyon area in model a (H/b = 1) was small, the suspension density was high.Although model d demonstrated the largest canyon area among all the models, it had a low particulate suspension density.The canyons in model b had a higher particulate suspension density compared with those in model c.These results clearly proved that step-up building arrangements do not facilitate particle transportation.According to the comparison of the particulate suspension densities, for models a and b, the particle concentration at the top of the buildings was high, but the local area was also large; therefore, the number of particles per unit area was lower than that in the canyons.Certainly, in models a and b, people living in the streets may experience more particle dust in the atmosphere compared with those living at the top of the buildings.

Mechanism of Particle Transport
In the atmosphere, atmospheric wind speed and turbulence mainly restrict the movement of PM emissions from industrial stacks.The horizontal component of the atmospheric wind velocity determines the horizontal momentum of PM, which is used to derive the particulate transmission distance.The vertical component of air velocity along the roof of the canyons determines the transmission of particles into or out of the canyon.
Figs. 10-12 illustrate the variation of the streamwise air velocity u, vertical air velocity v, and fluctuating velocity v on the roof of the canyons in the x direction, respectively.The horizontal axis represents a dimensionless length X/b.Because the velocity at the walls is zero, the width of the buildings was omitted from the x-axis.Therefore, X/b = 1-2 represents the first canyon, X/b = 2-3 represents the second canyon, and so on.
This study determined that the horizontal airflow velocity u Table 1.Particle transport efficiency in the building region.increased along the x direction in model b (Fig. 10).Therefore, an ascending building arrangement increases the airflow acceleration, which accelerates the horizontal movement of PM.Overall, Fig. 10 shows two types of recirculation zones: weak and strong recirculation zones (Buccolieri et al., 2010).The weak recirculation zones, which had a positive normalized horizontal velocity rate, in models b and c did not change the flow direction.The strong recirculation zones in model d changed the flow direction.This resulted in the retention of local PM, as explained in Section 4.2.2.Fig. 11 shows the vertical component of mean velocity along the roof of the street canyons.In canyons II, III, and IV in model b, the vertical velocity first decreased and then increased, but the velocity values in all these canyons were positive.Therefore, the mean velocity was not sufficient for transporting the PM into the canyon.Models a and d exhibited the same variations in the vertical air velocity.In model c, the vertical velocity along the x direction decreased, and that on the roof of canyon IV was negative, indicating that the PM purged into the canyons with the airflow.In addition, in canyons I and II, the vertical flow was so weak that the vertical velocity was approximatively equal to zero.
Fig. 12 shows the profiles of the vertical velocity fluctuations.Air exchange rate represents the rate of air removal from a street canyon, and it is defined as follows:   (Cheng et al.,   2008; Liu et al., 2011)   where v is the velocity component in the vertical direction, Γ roof is the roof area of the street canyon, and λ is the average period.Therefore, the vertical velocity fluctuation is the key factor affecting the air exchange rate, particularly when the mean vertical velocity v  is low.

CONCLUSION
On the basis of gas-solid two-phase flow theory and the Lagrange method, this study calculated the air flow and PM transportation on an atmospheric region scale by using numerical simulation.The concepts of particle transport efficiency, suspension fraction, and suspension density were adopted to evaluate the effect of building height and arrangement on the particulate transmission and concentration distribution.The following conclusions were drawn: 1.A vortex was observed behind the final building when the air flowed through the buildings.The flow eddies on the roofs of the street canyons in building heights arranged of descending order (model c, H/b = 5, 4, 3, 2, 1) and highrise buildings (model d, H/b = 5) hindered the transportation of particles into the canyon.The vortex formation region demonstrated low particle volume fraction.
2. The high-rise building model (H/b = 5) had the lowest particle transport efficiency among all four models, and the suspension density in street canyons of this model was also the lowest.By contrast, the low-rise building model (H/b = 1) exhibited the highest particle transport efficiency among all models, but the suspension density in street canyons in this model was high.Therefore, high-rise buildings hinder the diffusion of particles, and make it difficult for particles to enter deep canyons.
3. The particle number concentration and suspension density in the canyons of the step-down building model (H/b = 5, 4, 3, 2, 1), were lower than those observed in the canyons of the step-up building model (H/b = 1, 2, 3, 4, 5).Therefore, street models involving step-down building height arrangements are the appropriate models for city development.
4. Observation of the airflow velocity on the street canyon roofs revealed that the step-up building arrangement increased the airflow speed and accelerated particulate transportation, whereas the step-down building arrangement reduced the airflow speed, with the vertical speed being almost zero, and weakened the velocity fluctuation, minimizing the number of particles in the canyons.The horizontal component of the airflow velocity determines the particle transmission efficiency and distance, and the vertical velocity fluctuation determines the quantity of particle transported into the canyons.Notably, the velocity fluctuation was the dominant cause of particle suspension in the canyon.
The 2D idealized street canyon models have some limitations.For examples, buildings and streets are assumed to have an infinite length, and the natural wind at the inlet is considered to be perpendicular to the street centerline.However, realistic urban areas are three-dimensional (3D).The presence of street with finite lengths and street crossings enhances local turbulence intensities and produces momentum exchange between neighboring streets.When the approaching wind is not perpendicular to the street centerline, a helical flow is formed along the street.Therefore, deriving the relevant results by using 3D urban models is necessary in future studies.

Fig. 1 .
Fig. 1.Computational domain for particle transportation in the urban area.

Ⅴ
1~5 buildings, Ⅰ~Ⅳ street canyons, Ⅴ region above buildings 1 m s -1 ), and an intermediate turbulence intensity value (I = 0.005) was imposed at the inlet.A pressure outlet boundary was adopted at the downstream outflow.

Fig. 2 .
Fig. 2. Comparison of air velocity profiles in the center street canyon by different mesh nodes.
Fig. 3. Mean vertical velocity at the upwind and downwind position inside the center street canyon.

Fig. 10 .
Fig. 10.Profiles of dimensionless streamwise velocity along the roofs of the street canyons.

Fig. 11 .
Fig. 11.Profiles of dimensionless vertical mean velocity along the roofs of the street canyons.
Fig. 12  indicates that the vertical velocity fluctuation is higher than the mean value for model d (H/b = 5), implying that turbulent transport plays a dominant role in the transmission of particles into the canyon.As shown in Figs.7 and 12, the vertical velocity fluctuation and the mean velocity in model c were low; therefore, fewer particles were transmitted into the canyon in model c, leading to a low particle concentration in this canyon.