EFFECT OF GALLERIES ON THE WIND FLOW STRUCTURE AND POLLUTANT TRANSPORT WITHIN STREET CANYONS WITH OR WITHOUT FACADE ROUGHNESS ELEMENTS (BALCONIES)

© 2020 The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. 45 EFFECT OF GALLERIES ON THE WIND FLOW STRUCTURE AND POLLUTANT TRANSPORT WITHIN STREET CANYONS WITH OR WITHOUT FACADE ROUGHNESS ELEMENTS (BALCONIES)


INTRODUCTION
Air pollution in an urban canopy presents an important environmental problem and the study of pollutant dispersion in cities is not an easy matter. The interaction between the overhead atmospheric flow and the urban obstacles, such as buildings, generates a complex flow inside the street canyons that affects the pollutant dispersion.
Air pollution measured as particulate matter (PM) may lead to increased mortality rates (Dockery et al., 1993;Hales et al., 2010). The fine particulate matter, i.e. the particles with a diameter smaller than 2.5 mm (PM2. 5), has been estimated to cause about 3.3 million premature deaths per year worldwide (Lelieveld et al., 2015). The mass of The façade roughness elements such as balconies can significantly affect the near-façade airflow patterns (Chand et al., 1998;. A better understanding of the impact of façade elements on the near-façade airflow patterns and the pressure distributions on the façade is essential for the accurate evaluation of wind-induced natural ventilation (Gullbrekken et al., 2018;Ramponi et al., 2014), pollutant dispersion (Karkoulias et al., 2019) and convective heat transfer coefficient expressions for isolated buildings (Montazeri and Blocken, 2018).
Wind tunnel experiments and computational fluid dynamics (CFD) simulations have been employed to investigate the impact of building façade geometrical details on the near-façade airflow. The mean velocity and the Air change rate per hour (ACH) in 5-story building configuration (15m building height) with balconies using Steady RANS as a turbulence modeling approach was studied by Ai  The older studies were limited to a single aspect ratio, for which they studied the mean velocity, the mean surface pressure and the distribution of the carbon monoxide as a gaseous pollutant.
Our previous study (Karkoulias et al., 2019), presented results of the vertical distribution profile of aerosol traffic-generated particulate matter with respect to the different floor heights, in typical multistorey buildings forming deep street canyons (AR>2). It demonstrated the following: a) Inside the canyon three vortices were formed. One of them dominated the cross section and rotated in a counterclockwise direction. The smaller vortices rotated in the opposite direction and were forced near the extremities of the canyon. The smaller second vortex appeared inside the canyon top, while the slightly larger third appeared just above the street, both their centers lying closer to the leeward side of the canyon. Additional (much smaller) ones were observed inside the balcony cavities. b) The vortices (e.g. the main vortex) appeared to possess an inner core rotating as a "forced vortex" (i.e. solid body rotating air mass). In the leeward side they appeared to form "Rankine vortices", while in the windward side a shear layer zone was formed between the vortex and the façade. c) The central vortex rotated in the counter-clockwise direction with an angular velocity of ω1=+0,324 rad/s, the vortex near the leeward building and street corner rotated with an angular velocity of ω2=-0,094 rad/s. The smaller vortex near the leeward top corner rotated at a rate of ω3=-0,072 rad/s. d) The actual pollutant transport within a canyon cavity started from a dead zone near the street level, followed by an exponential reduction due to the flow diffusion imposed by the vortex structures. This was followed by a smaller dead zone near the roof while the final "wash out" was driven by the shear layer formed between the cavity flow and the outer wind.
In general, the balconies on the leeward building façade were the worst cases in the pollutant concentration levels, especially in the lower parts of the canyon. It was concluded that the balconies create vortices which trapped the air pollutants at lower heights. The multi-balcony configurations modified significantly the flow field and the relevant pollutant transport mechanisms.
The present study focused on a computational investigation of the effect of the galleries on the flow field structure and the vertical distribution of the concentration both near the building facades and on the centerline of the canyon. Numerical results are presented for various gallery and balcony combinations for an aspect ratio of h/w=2.33. They illustrate the formation of the vortices (i.e. number, nature, rotation rate) inside the canyon and the vertical concentration profiles of the aerosol particles which attributed to the complex wind flow structure and the physical layout of the street canyon. Finally, it is demonstrated that the presence of the galleries did not help the pollutant to escape from the canyon due to the lower dispersion of the pollutant and its accumulation inside the cavity.

DOMAIN MODELING, GRID CHARACTERISTICS AND BOUNDARY CONDITIONS
In the present study the street level galleries were added into the reference geometry (a narrow city canyon without balconies and an aspect ratio 2.33, as discussed in our recent paper (Karkoulias, et. al., 2019) to investigate how they affect the flow field inside the cavity and the distribution of the concentration on the building facades and the centreline of the canyon. The investigation was extended to other similar configurations by studying the combination between galleries and balconies. The results are presented in section 3 of this paper.
The canyon geometries employed in the present study are described in Table 1. The street canyon configuration named "Geometry A" with balconies upon the leeward façade and galleries on both facades is illustrated in Figure 1  The undisturbed flow inside the computational domain, the distribution of the vehicular exhaust, the local source strength and the corresponding average PM10 emission rate, were defined in our previous paper (Karkoulias et al., 2019). The geometric characteristics of the canyon and the facades elements (e.g. the geometry of the balconies, the geometry of the source and the characteristics of the traffic) were described in the same paper.
The galleries penetrated 2.5m into the interior of the building and they were 4.5m in height ( fig.1b). The prevailing wind direction was perpendicular to the long street canyon axis. As a result, the 3-D spatial domain was simplified into a two dimensional (2-D) one, while the wind speed above the roof-top level was set to be equal to 1.5 m/s (Fig. 1b). The building walls, the roofs and the street pavements were defined as "wall boundaries" (zero velocity and impermeability conditions), while the top boundary of the computational domain opposite the street was assigned as a symmetrical one.
The closure of the dynamic flow equations employed the steady state κ-ε RNG (Re-Normalized Group theory) method (Kim and Baik, 2004). In order to achieve a greater accuracy in a shorter computational time, the flow domain was divided into 2 regions, each with its own grid. The street canyon (height = 28 m, width = 12 m) incorporated a finer structured Quad (Cartesian) grid of uniform spacing whereas for the remaining domain (above the canyon) incorporated a coarser structured mesh of tetrahedral elements (Fig. 1c). The mesh of the computational domain had 182098 cells, 365114 faces and 183025 nodes.

MODEL VALIDATION
The validity of the numerical simulation employed in the present study is discussed in our previous publication (Karkoulias et al., 2019). It was assumed that the validation holds for the modified geometry as well. The fluent RANS code for complex architectural geometries was evaluated against wind tunnel experiments and field measurements in our previous paper. Comparisons demonstrate that the RANS results were overall in good agreement with the wind tunnel experiments and with the field measurements and could be used as basis for understanding the detailed flow dynamics.

THE EFFECT OF THE GALLERIES IN THE WIND FLOW STRUCTURE
The four cavity geometries (Table 1) studied in our previous publication (Karkoulias et al., 2019) which highlighted the influence of the balconies on the flow field shown in Figure 2. The narrow cavity without balconies with aspect ratio h/w=2.33 Geometry A The Reference case with balconies on the leeward building façade Geometry B The Reference case with balconies on the windward building façade Geometry C The Reference case with balconies on both building facades These same cavity geometries were employed in the present study to investigate the effect of galleries, in combination with the balconies, on the flow field and on the distribution of the concentration on both buildings facades and the canyon cavity centerline. Figure 3: Flow field streamlines inside the canyon cavity with galleries and aspect ratio 2.33, (a) without balconies, (b) balconies on leeward side, (c) balconies on windward side, (d) balconies on both sides.
The climate of street canyons is primarily controlled by the micro-meteorological effects of the urban geometry rather than the mesoscale forces controlling the climate of the boundary layer (Hunter et al;1992). The outer wind and the flow structures inside the canyon cavity do not correlate in a simple manner. The pollutant transfer requires the detailed knowledge of both the outer wind and the inner cavity flows to be analyzed.
The number of vortices, the position of their center and the angular velocity (ω) (to be discussed below) for the vortices inside the cavities without and with galleries are shown in Table 2: ωupper = -0.20 rad/s ωmiddle = +0.34 rad/s ωlower = -0.06 rad/s * The x coordinate was measured from the leeward side while the y coordinate was measured from the street level.
As it will become apparent, the presence of the galleries modifies the vortex structures and the associated particulate mass convection mechanism quite drastically. The nature of the vortices may be studied as follows: (FIG.3A) The addition of galleries into the reference Geometry formed three vortices ( fig.3a). Fig. 4 illustrates the variation of the magnitude of the vertical component of the flow velocity along the horizontal line that pass through the centres of the vortices. This figure implies that:

THE REFERENCE GEOMETRY WITH GALLERIES ON BOTH SIDES
The main vortices appeared to have an inner core rotating as a "forced vortex", i.e. the vertical velocity component (uy) increased with distance (r) from the centre. In other words, the vortex contained a nearly solid mass that rotated with a nearly constant angular velocity (ω= The main middle clockwise rotating vortex extended down from the level of the road up to the roof of the canyon and occupied most of its cross section. This vortex rotated with an angular velocity of ωmain= -0.75 rad/s which implies a period of rotation of the order of 8.37s. Fig.4 demonstrates that in near the leeward side "Rankine" vortices were formed while near the windward side a shear layer zone was formed between the vortex and the façade. The small counterclockwise rotating vortices inside the galleries rotated with an angular velocity of ωleeward gallery =+0.13 rad/s and ωwindward gallery =+0.31 rad/s. The leeward gallery vortex had a period of rotation of the order of 48.3s while the windward gallery vortex had a period of the order of 20.2 seconds, i.e. at a much lower pace than the main one.

Figure 4:
The y-velocity along horizontal line passing through the center of the main middle vortex, leeward and windward gallery vortex located at 17.72m, 1.45m, and 2.08m respectively away from the street level

THE GEOMETRY A WITH GALLERIES ON BOTH SIDES (FIG.3B)
In Geometry Α with galleries, five vortices were formed ( fig.3b). Three main vortices (i.e. the upper, the middle and the lower vortex) and two secondary vortices (i.e. the leeward and the windward gallery vortex) were formed inside the cavity. Fig. 5 illustrates the profile of the uy along the horizontal line that pass through the centres of the vortices. The results indicated that the lower main vortex rotated in the clockwise direction with an angular velocity of ωlower =-0.23 rad/s which implies a period of rotation of the order of 27.30s. In other words, the centre of the vortex rotated at a constant rate of about 13.2 degrees per second. Fig.5 demonstrated that in the leeward side the nature of the vortex structures formed "Rankine vortices" (i.e. it appeared to be composed of an inner solid body rotating air mass that diffuses at the outer limits). On the other hand, near the windward side a shear layer zone was formed between the vortex and the façade. The angular momentum of the rotating air masses dissipates within the shear zones, leading to a weakening of the linear vortices inside the canyon. Eventually the canyon vortices loose coherence as the core radius increases and the vortices expand to fill the entire region (Kingdon R., 2008). The inner part of the flow corresponds to a rigidly rotating core while the outer region becomes a free vortex. This prevents the velocity from becoming infinite at the center of rotation. At the same time, for a radial distance r >R (where R is the core radius), the model reverts to a free vortex, which allows for the velocity to decay at large distances. A Rankine vortex constitutes an amalgamation of the forced and free vortices profiles (Katopodes, N., 2019).
The results indicated that something similar occurs with the middle main vortex rotated in the counter clockwise direction with an angular velocity of ωmiddle =+0.14 rad/s (i.e. a period of rotation equal to 44.85s). The centre of this vortex rotated at a constant rate of about 8.04 degrees per second.
Τhe upper main vortex rotated in the clockwise direction with an angular velocity of ωupper =-0.76rad/s while the central one rotated at a constant rate of about 43.6 degrees per second. Apparently, the uppermost vortex rotated very fast (when compared to the other two) and this implies the presence of very strong shear phenomena all around its periphery. The most intensive upper main vortex transported the pollutants from the middle part of the canyon to the higher levels. The counter-clockwise vortices inside the galleries rotated with an angular velocity of ωleeward gallery =+0,13rad/s and ωwindward gallery =+0,38rad/s. The centre of the leeward gallery vortex rotated at a constant rate of 7.46 degrees per second so it rotated at a similar rate with the lower main vortex. The centre of the windward gallery vortex rotated at a rate of 21.8 degrees per second which is much faster than the corresponding magnitude of the lower vortex.    Table 2 it is obvious that the galleries affect the flow structure (i.e. number, position, and size of the vortices) inside the cavity. These changes are discussed below: Inside the reference geometry with galleries the number of vortices and their position changed (three vortices versus two). Inside the cavity there was only one main vortex that occupied the cross section of the cavity. Its angular velocity was about six times higher than the velocity of the leeward gallery vortex and twice than the velocity of the windward gallery vortex. In the corresponding geometry without galleries the lower vortex had a higher angular velocity than the upper one which means that it had the potential to transfer mass to higher levels. Hence, it may be deduced that the lower vortices have a reduced potential to convect mass towards the main vortex.
In Geometry A with galleries, five vortices appeared (i.e. three inside the main section of the cavity and two in the galleries) against the three inside the respective geometry without galleries. The presence of the galleries did not influence the flow structure of the three main vortices significantly but only changed the size and their angular velocities. The angular velocity of the upper and lower vortex was higher than the middle one. In the configuration without galleries the opposite had been observed.
The galleries in Geometry B modified the number of the vortices (three versus two) creating a vortex inside the leeward building gallery and shifting the center of the lower vortex downwards and closer to the leeward facade.
The angular velocity of the lower vortex was somewhat greater than the upper while in the case without galleries the lower vortex had the half angular velocity than the upper one.
The galleries in the Geometry C were also modified the flow field structure (i.e. three vortices versus five). The lower vortex extended from the leeward gallery to the entrance of the windward gallery, while the two middle vortices were replaced by a larger one. The angular velocity of the middle vortex was higher than the upper and the lower vortex. Especially the angular velocity of the lower vortex was about five times smaller than the middle one.
In the corresponding geometry without galleries the upper and the lower vortex had the same angular velocity which was much higher than the angular velocity of the middle ones.

INVESTIGATION OF THE AIR POLLUTANT TRANSPORT MECHANISM AND COMPARISON BETWEEN THE GEOMETRY A WITH AND WITHOUT GALLERIES
A further discussion of Geometry A with galleries is given below because the corresponding geometry without galleries had been studied numerically and experimentally in our previous work. Also, the addition of the galleries did not significantly change the structure of the cavity flow. That is, the number of the main vortices in the cavity (except the vortices created in the galleries) was the same. Also, the centers of the three vortices as shown in Table  2 are approximately in the same position. Therefore, the main effect comes from the relative sizes of the vortices.
In the Geometry A with galleries ( fig.3b) the results indicated that the middle vortex became smaller in size while the size of the other two was increased. The lower vortex transported the pollutant to the leeward side from the street level up to the 8m level. The middle vortex transported the pollutant to the windward side from 8m level up to the 25m level and the upper one transported the pollutant out of the canyon. In general, in the geometry A with galleries the mechanism of the vortices makes the ventilation of the canyon difficult. The effect of the galleries was immediately apparent by comparing the angular velocities of the vortices presented in the above paragraph. Therefore, the vortex in the gallery of the leeward building had a lower angular velocity than the lower main vortex. Thus, the vortex had a little potential and the pollutant could not be transported out of the gallery. The result was the accumulation of the pollutants in this area. Also, the vortex in the gallery of the windward building had a higher angular velocity than the lower main vortex. The pollutant was therefore transported from the windward gallery to the bottom of the cavity. The lower main vortex was more intensive than the middle one and transported the pollutant to the leeward side and to the upper levels. Finally, the pollutant was concentrated on the windward side because the middle vortex had lower angular velocity than the upper one. So, the middle vortex did not have the potential to transport the pollutant to the upper vortex. Due to the high intensity of the upper vortex the concentration near the top of the cavity decreased because the pollutant escaped from this area. The leeward gallery vortex generated a maximum in the pollutant distribution of the order of 15.74μgr/m 3 near the entrance of the gallery. The minimum mass concentration near the leeward wall was 15.66μgr/m 3 . Inside the opposite gallery (i.e. the windward gallery), the vortex generated a maximum mass concentration near the windward wall of the order of 12.53μgr/m 3 and a minimum near the entrance of the gallery of the order of 12.5μgr/m 3 . These data imply that pollutant accumulated inside both galleries. In other words, the galleries do not help the pollutants to escape. On the contrary, the mass concentration is increased within the bottom of the canyon.  Table 3 records the percentage reduction of the vortex extreme concentrations from vortex to vortex, resulted from a comparison of the data presented in Figure 8 for Geometry A with galleries and Geometry A studied in our previous paper (Karkoulias et. al., 2019).  (fig. 9a). The concentration exhibited to remain constant up to the height of 7m above the street level and then decreased up to the height of 18m. The uy profile justifies this reduction. From the height of 18 m up to the 23 m, the concentration remained almost constant while the uy was decrease. Then the concentration was followed by a reduction while in the last meter, a sharp decrease was observed due to the upward flow and the external flow.
The PM10 Mass Concentration vertical profile near the leeward building façade ( fig. 9b) was interpreted as follows: The increase of the uy ( fig. 9a) along the positive direction led to the reduction of the concentration up to 3m. The distribution of the concentration up to 6 m was almost constant while then the increase in velocity along the negative direction led to a further decrease in concentration up to 12 m. From there the concentration did not change due to the uy which was constant. The concentration continued to remain almost constant with a slight decrease up to 27 m. A sharp decrease in the concentration was observed over the uppermost meter due to the upward flow -external flow interaction. In review, it is concluded that the air pollutant transport-dispersion mechanism consisted of three steps: 1) A dead zone was generated at the bottom of the canyon and the concentration of the pollutant in this region was increased by the presence of the galleries. This presence affected the flow field in such a way that made more difficult the upward movement of the pollutants to higher levels.
2) The enhanced mixing in the middle of the canyon due to the rotation of the vortices constituted the second step. The solid mass rotation of the vortices generated very strong shearing among their outer boundaries that not only diffused pollutant mass from vortex to vortex but in addition convected the pollutants forcing them to attach on the buildings facades. 3) A shear layer zone was created at the top of the canyon between the roof top level of the canyon and the upper wind external flow.

THE VERTICAL DISTRIBUTION OF THE CONCENTRATION ON BOTH BUILDING FACADES AND THE CENTERLINE OF THE CANYON
In the Reference geometry with galleries the middle-height vortex initially transported the pollutant towards the leeward facade, to be followed by a similar convection on the windward facade. The concentration of the PM10 near the latter remained almost constant from the ground level up to the height of about 24 meters. Above this height, the concentration decreased sharply up to the 28 meters, because the pollutants diffused to the outer atmosphere, as the vortex interacted with the outer free wind flow through the free shear layer. In the leeward building façade, the concentration decreased from the street level up to the 3 m one, remaining constant above it up to the height of 15 m. Finally, it increased up to 23m to be followed by a sharp drop up to the top of the canyon due to the vortex contact with the upper boundary layer (fig.10a).
The Geometry A with galleries data were discussed in detail in section 3.2 and the distribution of the PM10 mass concentration near the leeward and windward façades as well as along the canyon centerline are illustrated in figure  9b.
The Geometry B with galleries flow field was also modified so that the vertical distribution of concentration on the building facades was as follows: near the windward façade the concentration decreased between the street level and the 16 m height, followed by a sharp decrease up to the 22 m height. Finally, it remained constant from the 22 m height up to the height of 26 m. Along the last two meters appeared a sharp decrease in concentration up to the roof of the canyon. At the leeward facade the concentration evolved as follows: initially it increased from a low magnitude on the street level up to approximately the 8 m height, followed by a sharp decrease up to the height of 16 m, above which it remained was almost constant up to a height of 24 m. Finally, along the last four meters the concentration increased steadily up to the top of the canyon (fig. 10b). The leeward gallery vortex contributed to the increased particle concentration within the low heights of the building. On the opposite side, the upper main vortex contributed to an increased particle concentration along the upper floors of the same building.
In the most complex scenario (i.e. Geometry C with galleries), the mechanism of the vortices does not help the pollutant to escape from the canyon.  From the results above, the deviation of the mass concentration between the street level and the roof of the canyon in the leeward and windward side and in the centerline was calculated. Table 4 below presents the results for the geometries studied above with galleries and those without galleries studied in a previous work (Karkoulias et al., 2019). The major findings may be reviewed as follows: 1) In general, the galleries proved to be another important factor affecting the wind flow structures. In contrast to the simple geometry without roughness, the multi-balcony configurations with galleries significantly modified the flow field and the relevant pollutant transport mechanisms. The presence of the galleries did not help the pollutant to escape from the canyon. 2) In many of the above figures ( fig. 9b, 10) illustrated the same tendency as the Geometry A in which the field measurements were made. More specifically, the vertical distribution of the concentration did not decrease exponentially with the height but there were intervals in which it increased and burdened the adjacent area.
3) The addition of the galleries changed the number, the position and the rotation rate of the vortices. In Geometry A with galleries the middle vortex acquired less intensity than the corresponding one in the Geometry A. This means less dispersion of the pollutant and its accumulation in the middle of the cavity. 4) In Geometry A with galleries the concentration decreased by 8.4% at the height of the center of the middle vortex on the leeward side, compared to the 80% observed in the Geometry A with no galleries. On the windward side, at the same height, the concentration decreased by 12.6% compared to the 25% observed in the corresponding geometry without galleries. 5) The deviation of the mass concentration between the street level and the roof of the canyon in the leeward and windward side as well as the centerline of the canyon, in geometries with galleries, was smaller than that of the corresponding geometries without galleries (Table 4). 6) The pollutant transport within a canyon cavity started from a dead zone near the street level, followed by an exponential reduction due to the flow diffusion imposed by the vortex structures. This was followed by a smaller dead zone near the roof while the final "wash out" was driven by the shear layer between the cavity flow and the outer wind.

CONCLUSIONS
Galleries incorporated on both building facades of a street canyon influence the air quality in it. The presence of galleries modified the flow field inside the canyon and indicates a significant influence on the mass concentration distribution of the polluting particles. The most relevant effect on the mass transfer rate from the street level to the roof level was the presence, induced by galleries and balconies, of several vortices in the street canyon that reduced the overhaul mass transfer.
The surface roughness of the building facades affect the flow features over and within the urban street canyons and as a consequence influence the mean and turbulent exchanges at the pedestrian level. The façade elements such as galleries produced lower mixing and so, they made the escape of pollutants difficult. Therefore, it appeared reasonable to suggest that through a formal exploration of buildings geometries, the ventilation potentional of urban canyons could be increased leading to an improvement of the air quality within the street canyon.

SOURCES OF FUNDING
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.