The flow characteristics inside urban street canyons were studied in a laboratory water channel. The approaching flow direction was horizontal and perpendicular to the street axis. The street width was adjusted to form street canyons of aspect ratios 0.5, 1.0, and 2.0. The velocity field and turbulent intensity were measured with a laser Doppler anemometer at various locations within the street canyons, which were used to elucidate the flow pattern inside the street canyons. It was found that the previous numerical modeling results are in good agreement with the current experimental results at most locations. For the street canyon of aspect ratio 0.5, which belongs to the wake interference flow regime, the mean and fluctuating velocity components were more difficult to measure as compared with the other two cases because of its more complicated flow pattern. Some guidelines for numerical modeling were developed based on the measurement results. The data presented in this paper can also be used as a comprehensive database for the validation of numerical models.
With rapid urbanization and city expansion, public concern on urban environmental issues has been increasing over the past decades. In response to the demand for better air quality, many studies of airflow and pollutant dispersion in built environments have been carried out through field observations (Louka et al. 2000; Xie et al. 2003; Qin and Kot 1993), laboratory experiments (Pavageau and Schatzmann 1999; Gerdes and Olivari 1999), and numerical models (Li et al. 2006). Street canyons are the basic element of urban areas, which have a distinct climate where microscale meteorological processes dominate (Oke 1988) and air ventilation and pollutant removal occur solely through the roof level. The most important feature of street-canyon microclimate is the wind-induced flow patterns inside the canyons. These unique microscale meteorological processes affect not only the local air quality but also the comfort of city inhabitants (Bottema 1993). Hence, there is an increasing interest in the study of airflow and pollutant dispersion in urban street canyons. The extensive studies performed so far have improved our understanding of the pollutant dispersion mechanism and helped us to minimize the damage caused by harmful air pollutants in urban areas.
Physical modeling is a useful approach to street-canyon pollution research. It has the advantage of having fully controllable street-canyon geometries and upwind boundary conditions for airflow. Moreover, it provides an opportunity to examine the linear and nonlinear effects of various parameters individually and/or in combination (Meroney et al. 1996). These modeling techniques and their results are often used to validate numerical models (Baik et al. 2000; Sagrado et al. 2002) and have provided useful information on wind flow patterns and pollutant distribution (Gayev and Savory 1999; Pavageau and Schatzmann 1999; Kastner-Klein et al. 2001). The physical modeling studies of street-canyon flow and pollutant dispersion were conducted almost exclusively in wind tunnels. The popularity of wind tunnels is mainly due to their capability of simulating various environments that are similar to the atmospheric boundary layer. Recently, Ahmad et al. (2005) extensively reviewed the various wind-tunnel studies on pollutant dispersion at urban street canyons and intersections.
Another useful facility for laboratory physical modeling is water channels. Water channels have been utilized widely in geophysical applications, although their application to urban flow and dispersion research is in its early stage. Baik et al. (2000) successfully studied the flow in street canyons using a recirculating water channel. Recently, they also used the same technique to investigate the effects of street bottom heating and inflow turbulence on street-canyon flow (Kim and Baik 2005). Liu et al. (2003) carried out a water tank experiment to investigate the convective flow induced by bottom heating and the effects of the ambient wind on the flow in nonsymmetrical urban street canyons using the particle image velocimetry (PIV) technique. Caton et al. (2003) investigated the dispersion mechanisms of a passive tracer in a two-dimensional model of a street canyon in a recirculating water channel. Their principal concern was the pollutant transfer between the street and the external flow. All of these studies revealed that the water channel can reproduce the results obtained in wind-tunnel experiments and can be employed to study the flow and pollutant transport inside urban street canyons.
In this study, a systematic physical modeling study of the flow in idealized street canyons was carried out in a water channel in which velocity and turbulence quantities were measured with a laser Doppler anemometer (LDA). The objective of this study is to explore the flow characteristics within street canyons of different aspect ratios. The data obtained would also be useful as validation datasets for various numerical studies related to this research area.
2. Experimental apparatus and design
The experiments were carried out in a 10-m-long, 0.3-m-wide, and 0.5-m-high laboratory flume located at the Croucher Laboratory of Environmental Hydraulics, at The University of Hong Kong. The schematic diagram of the laboratory flume is shown in Fig. 1a. Recirculating water flow could be maintained in the flume at a flow rate up to 28 L s−1. The working test section was located in the middle of the flume and all experiments were performed there for the best flow uniformity. The sidewalls of the test section were made of toughened transparent glass. The ambient flow speed inside the test section was adjusted by varying the flow rate and the water depth through an adjustable overflow weir at the downstream end of the test section. The two-component flow velocities and turbulent intensities (in the streamwise and vertical directions) were measured with a two-color LDA (Fig. 1b). The LDA was a fiber-optic system (model 55X, supplied by Dantec Dynamics). It used a 4-W argon ion laser for the laser source and a correlation-based processor for the computation of flow velocities from the Doppler bursts. The LDA worked in the backscatter mode with a frequency shift to measure the reverse velocities. It is a nonintrusive measurement method, except for the presence of seeding particles; hence, the original flow field will not be disturbed by the measurement devices. In this study, the flow was seeded with the polycrystalline powders of nominal diameter at 30 μm (supplied by Optimage). This type of seeding particle was selected because of its advantage of having neutral buoyancy, extrafine grain, and good light-scattering properties. No density stratification was considered in this study.
Six to 10 identical model buildings (29.8 cm × 10 cm × 10 cm in size) were aligned perpendicular to the prevailing flow direction in the working test section to form the street canyons, with the target canyon located at the center (Fig. 2a). The upstream canyons can serve to make the inflow turbulence fully developed before it reaches the target canyon. The enlarged view of the target canyon is shown in Fig. 2b. The height of the model buildings was fixed at h = 10 cm and the width of the canyon b could be varied according to the characteristic aspect ratio (street height:canyon width, h/b). The depth of water used in all of the experiments was about 40 cm, and the reference velocity (freestream velocity) U was taken at z = 30 cm measured from the flume bottom. The Reynolds number based on the free-stream velocity and building height, Uh/ν, was about 12 000, which is the same as that in our previous numerical studies (Liu et al. 2004; Li et al. 2005). This value of Reynolds number also complies with the criteria suggested by Hoydysh et al. (1974) to ensure that the flow pattern in the street canyon is independent of viscous effects (Meroney et al. 1996).
Three sets of experiments were carried out in this study, that is, street canyons of aspect ratios 0.5, 1.0, and 2.0 (corresponding to b = 20, 10, and 5 cm, respectively). For each case, 6, 8, and 10 buildings were used to form the street canyons, respectively. For each measurement point, mean and fluctuating velocities were measured for a time period of 180 s, which included 1000–3000 valid velocity data (Doppler bursts), depending on the measurement locations. The data presented in this paper were measured on the vertical plane in the middle of the spanwise direction (indicated by y). All of the data acquisition and the subsequent velocity conversion and averaging processes were controlled automatically by the LDA system.
3. Results and discussions
In this section, the measurement results of the velocities and their fluctuations at several locations are presented, along with the results obtained from our previous numerical modeling. The flow features suggested by these results are also discussed. It is our intention that these results will lead to some guidelines for future numerical modeling of airflow inside street canyons, as well as comprehensive validation databases for the development of these numerical models.
The numerical results shown in this section come from our previously published work (Liu et al. 2004; Li et al. 2005). The 3D large-eddy simulation (LES) study of Liu et al. (2004) employed a computational domain of 0.5h for the free surface layer in the vertical direction, while the 2D k–ɛ study of Li et al. (2005) adopted 3h to simulate the free surface layer. The inlet and outlet conditions were set to be periodic in the LES study and the free stream was simulated by a pressure-driven background flow. This boundary condition gives a configuration of infinitely repeating street canyons in the streamwise direction. In the spanwise direction, the LES model also prescribed a periodic boundary condition, representing an infinitely long street-canyon configuration. On the other hand, the k–ɛ model prescribed the inflow velocity profile and turbulent properties at the inlet. Several street canyons were set upstream and downstream of the target street canyon to simulate an urban environment, which is the same as the current experimental study.
a. Street canyon of aspect ratio 1.0
To evaluate the current experimental results, two sets of wind-tunnel experimental data from literature are selected for comparison—one is the measurement collected by Kovar-Panskus et al. (2002) in an isolated street canyon at a Reynolds number of 5.6 × 104, and the other is Hamburg University’s Compilation of Experimental Data for Validation of Microscale Dispersion Models (CEDVAL; see information available online at http://www.mi.uni-hamburg.de/introduction.433.0.html) result, which is taken from a 7 × 3 cubic array (the B-1-1 case). As shown in Fig. 3a, the streamwise velocity along the vertical centerline measured by Kovar-Panskus et al. (2002) is generally in good agreement with the current experimental result in the core region of the cavity, whereas, over the street canyon, it is smaller when compared with the current experimental data. The smaller streamwise wind speed measured in CEDVAL is mainly due to the 3D effect. In the CEDVAL experiment, the street length was finite and thus the air also entered the street canyon parallel to the street axis in the spanwise direction. This spanwise air movement shared part of the air exchange of the street canyon, which in turn reduced the streamwise wind speed relative to that in a 2D street canyon. The discrepancy observed in the streamwise and vertical velocity fluctuations (Figs. 3c,d) can also be attributed to the aforementioned 3D effect. The streamwise velocity fluctuation was smaller in the 3D cubic array case than its 2D street-canyon counterpart (Fig. 3c). The turbulent kinetic energy mainly contained originally in the streamwise fluctuation was redistributed to the other two components, that is, spanwise and vertical fluctuations. Therefore, the vertical velocity fluctuation in 3D cubic array shows a large discrepancy from the current 2D street-canyon data (Fig. 3d).
Figures 3 –7 show the velocity and turbulence intensity profiles for the street canyon of aspect ratio 1. According to the classifications of Oke (1988), the flow pattern inside this street-canyon configuration should be a skimming flow regime, which is characterized by both a stable and isolated recirculation at the street-canyon center and an ambient flow that is decoupled from the flow in the cavity. Along the vertical centerline (x/b = 0.5), upstream vertical line (x/b = 0.25), and downstream vertical line (x/b = 0.75), the profiles of normalized streamwise velocities show some similarities (Figs. 3a, 4a, 5a). Also shown are the results of previous numerical studies, which agree well with the current experimental data within the street cavity (z/h < 1). Above the cavity (z/h > 1), the LES results show large discrepancy when compared with the k–ɛ model and the experimental results. As described above, this large discrepancy is mainly caused by the small vertical extent of the free surface layer employed by the LES study. As a result, the streamwise velocity above the cavity in the LES study evolved quickly to the free-stream velocity, making the results above the cavity differ very much from those of the other two studies. Our experimental results (not shown here) suggest that, from z = 2h to z = 3h, the measured streamwise velocities vary slightly (less than 2%). This in turn suggests that the height of the free surface layer should be at least h in numerical modeling for more reliable results. The LES study by Cui et al. (2004) also noted this problem of vertical extent of the computational domain. They argued that larger vertical extent allowed larger eddies to develop above the rooftop, and thus the flow there was more turbulent.
Figures 3b, 4b and 5b depict the vertical profiles of the vertical velocities along the center, upstream, and downstream lines, respectively. Upstream (Fig. 4b), the flow clearly exhibits an upward motion while a downward motion can be observed downstream (Fig. 5b). This observation is in line with the fact that the flow lies within the skimming flow regime and a single recirculation is formed within the cavity. At the center (Fig. 3b), there is a weak upward motion resulting from a slight downstream shift of the vortex center, as identified in our previous numerical study (Li et al. 2005).
One of the advantages of this study is that turbulence intensities can also be obtained, which have not been covered in the previous water channel experimental studies (Baik et al. 2000; Liu et al. 2003; Kim and Baik 2005). Here, the primes (′) represent the deviation of the turbulence properties from their statistical means u and w. The turbulence intensities given by the k–ɛ model were estimated from the calculated turbulence kinetic energy (TKE) by assuming isotropic turbulence (Li et al. 2005).
Figures 3c,d, 4c,d and 5c,d show the vertical profiles of streamwise and vertical velocity fluctuations along the center, upstream, and downstream lines, respectively. It is obvious that the results from both the numerical models and the current experimental study generally agree well with each other, with the local maxima located slightly above the roof level and nearly constant values in the cavity core region. This observation confirms the conclusion drawn by Kastner-Klein et al. (2001), which showed that the velocity fluctuations were rather uniformly distributed in the canyon without traffic. The slight increase of velocity fluctuations immediately above the canyon is mainly due to the flow shear there. This shear may cause some sheltering effect on the vertical turbulent transport across the canyon top and thus suppresses the pollutant removal from the canyon. It is also noteworthy that the magnitudes of both fluctuating velocity components are nearly the same, which indicates the approximate isotropy at the cavity’s core region. This finding can be further confirmed by the fact that the velocity fluctuations obtained from the k–ɛ model using the isotropy assumption are in line with those from the current experimental study.
The horizontal profiles of normalized velocities and fluctuations along the roof level (z/h = 1) and half-height level (z/h = 0.5) are shown in Figs. 6 and 7, respectively. The streamwise velocity at roof level (Fig. 6a) predicted by LES were much larger than the results obtained from the current experiment and the k–ɛ model because of the smaller computational domain for the free surface layer, as mentioned above. At the half-height level, the streamwise velocity (Fig. 7a) is nearly zero, while the vertical velocity (Fig. 7b) changes direction quickly, with an upward motion upstream and downward motion downstream. All of these features illustrate the characteristics of an isolated recirculation at the cavity center.
b. Street canyon of aspect ratio 2.0
The flow in the street canyon of aspect ratio 2.0 also belongs to a skimming flow regime, but the flow pattern is more complicated than that in the street canyon of aspect ratio 1.0. In this situation, there are two vertically aligned primary counterrotating recirculations in the cavity. The profiles of the normalized streamwise velocity at the center (Fig. 8a), upstream (Fig. 9a), and downstream (Fig. 10a) vertical lines demonstrate this complication. The streamwise velocities change directions twice from the roof level down to the ground level, while in the street canyon of aspect ratio 1.0 the streamwise velocities only change direction once. The LES results also show some discrepancy from the other two above the street cavity (z/h > 1), resulting from the reason described above. The vertical velocity along the vertical centerline (Fig. 8b) also exhibits a complicated pattern—at the upper part of the cavity, the flow is upward; while at lower part, the flow changes to downward. Upstream (Fig. 9b), the flow is mostly upward because of the upper clockwise recirculation. Only a very weak downward motion is observed at the lower part, which suggests that the lower recirculation formed in the narrow cavity is much weaker than the upper one. This result is within our expectation because the strong shear induced by the free stream creates the upper recirculation and the latter then forms the lower recirculation through a relatively weaker shear. Downstream (Fig. 10b), similar phenomenon can be identified, and the only difference from its upstream counterpart is that most of the motion within the cavity is downward.
The profiles of normalized streamwise and vertical velocity fluctuations (Figs. 8c,d, 9c,d and 10c,d) in the cavity show some differences from the street canyon of aspect ratio 1.0. Rather than distributing uniformly, these velocity fluctuations vary much in the core region, especially along the downstream vertical line, where the fluctuations change sharply. These features can be attributed to the complexity of the flow pattern described above. However, there is one finding in common between the two street canyons: the velocity fluctuations immediately above the canyon are slightly increased and may thus cause some sheltering effect on the pollutant dilution from the cavity. Moreover, because of the complex flow pattern inside the cavity, the case of the aspect ratio of 2.0 should be more unfavorable for pollutants released at the pedestrian level (e.g., from vehicular sources) to be removed from the street canyon through the roof level.
The streamwise and vertical velocities at roof level (Figs. 11a,b) are similar to those of the street canyon of aspect ratio 1.0, because the upper recirculation is, to a certain extent, like the single isolated recirculation in the street canyon of aspect ratio 1.0, driven by the free stream shear above the canyon. Both numerical results of streamwise velocity are larger than the measurements from the current experimental study. At the half-height level (Figs. 12a,b), the streamwise velocities are all negative, while the vertical velocities are positive upstream and negative downstream. This indicates that the half-height level is within the upper recirculation, and the upper recirculation is larger than the lower one. This observation agrees well with the numerically calculated streamfunction plots (Liu et al. 2004; Li et al. 2005).
c. Street canyon of aspect ratio 0.5
Unlike the previous two cases, the flow in the street canyon of aspect ratio 0.5 belongs to the wake interference flow regime, in which the wake behind the leeward building is disturbed by the recirculation created in front of the windward building. This wake interference makes the flow pattern inside the street canyon much more complicated because the flow is no longer decoupled from the free stream. On the other hand, the interaction between the flow inside the cavity and the free stream helps to dilute the pollutants in the canyon.
Figures 13a, 14a and 15a show the vertical profiles of normalized streamwise velocity along the center, upstream, and downstream vertical lines. Obviously, the results from both the current experiment and previous numerical models differ from each other, suggesting the complexity of the flow pattern in this case. In general, the streamwise velocity profiles resemble those of the street canyon of aspect ratio 1.0, but exhibit a relatively larger magnitude of negative velocities. The vertical profile of the normalized vertical velocity along the vertical centerline (Fig. 13b) shows that the motion is upward and the local maximum occurs at the core of the canyon. The upstream vertical velocities (Fig. 14b) exhibit the same feature as the velocities along the vertical centerline, while the downstream vertical velocities (Fig. 15b) are downward with a very small magnitude. These observations suggest that there is a large primary recirculation located at the core region of the canyon, whose center shifts downstream (near the location of x/b = 0.75).
The vertical profiles of the normalized velocity fluctuations along the center, upstream, and downstream lines are shown in Figs. 13c,d, 14c,d, 15c,d. Apparently, these fluctuations vary dramatically in the cavity, which are totally different from those observed in the street canyons of aspect ratios 1.0 and 2.0. Moreover, the discrepancy between the fluctuations obtained from the k–ɛ model and the other two is noticeable. This may indicate that the isotropic assumption underlying the k–ɛ model is not valid in this case. Similar to the situation occurring in the street canyons of aspect ratios 1.0 and 2.0, the local maxima of these fluctuations appear slightly above the roof level.
The horizontal profiles of the normalized velocity and fluctuation at the roof level are depicted in Fig. 16. At the roof level, the velocity profiles show some resemblance to those measured in the street canyon of aspect ratio 1.0. The vertical velocity (Fig. 16b) is nearly zero at most places except at the windward corner, where the vertical velocity increases quickly. This suggests that there is a separation point near the roof-level windward corner.
The flow fields in the street canyons of aspect ratios 0.5, 1.0, and 2.0 were physically modeled with a laboratory water channel. The velocity and turbulent intensity field were measured with a laser Doppler anemometer at various locations within the street canyons. The main objective of this study is to investigate the characteristics of the street-canyon flows. The data obtained would also be useful as a validation database for numerical modeling. It was seen that our previous numerical results were in good agreement with the measurements in most locations.
The flow pattern inside the street canyons was revealed through analysis of the water-channel measurements. It was revealed that, for more reliable numerical results, a free surface layer height of at least h should be used in numerical modeling. Among the three cases measured in this study, it was shown that the street canyon of aspect ratio 0.5 produced a greater difference when compared with the numerical results because of its nature of wake interference flow regime, which is different from the other two skimming-flow cases (aspect ratio 1.0 and 2.0).
As part of an ongoing project on physical modeling of flow field and pollutant dispersion in urban street canyons, this article only dealt with the flow-field aspect in urban street canyons. The other part of great interest is pollutant dispersion. We would make use of the laser-induced fluorescence (LIF) technique to measure the pollutant concentration and its fluctuation in the same experimental facility.
This work is supported by the Hong Kong Research Grant Council under Grants HKU 7196/03E and HKU 7111/04E. The authors extend their sincere thanks to Mr. C. H. Tong at the Croucher Laboratory of Environmental Hydraulics for his extensive technical assistance during this experiment.
Corresponding author address: Dennis Y. C. Leung, Department of Mechanical Engineering, The University of Hong Kong, 7/F, Haking Wong Building, Pokfulam Road, Hong Kong, China. Email: email@example.com