Parameters
Despite the valuable effort mentioned above, our understanding of the pollutant removal mechanism of a street canyon is rather limited. Liu et al. (2005), making use of their LES results, suggested a l留學(xué)生dissertation網(wǎng)different formulation using the air exchange rate (ACH), pollutant exchange rate(PCH), average pollutant concentration (Q) and pollutant retention time (s) to quantify the street canyon ventilation and pollutant removal performance. Since LES costs more computational resources, Li et al. (2005) estimated ACH using the more cost-effective RANS k _ 3 turbulence model by assuming isotropic turbulence. Favorable results were obtained comparing with the LES results (Liu et al. 2005). The difference between the computation time of RANS k _3 turbulence and LES models is tremendous. Therefore, it is worth to develop a formulation based on the k_ 3 turbulence model to quantify the ventilation and pollutant removal in street canyons. Cheung’s paper aims to modify the LES formulation of Liu et al. (2005) so that the whole formulation can be applied using RANS k _3 turbulence model to quantify the ventilation (ACH) and air pollutant removal (PCH) properties of street canyons. Instead of isotropic turbulence assumption, the eddy-viscosity and eddy-diffusivity models are used to calculate the turbulent components of ACH and PCH.

the ventilation rate (ACH) was decomposed to mean component and fluctuating component and so did the pollutant removal rate (PCH). This approach allowed us to examine the contribution of each component under different cases.

Where the " ACH'" was deduced from Reynolds stress tensor(equation 14)The net vertical flux across the roof-level of the street is zero, hence, Eq. 4 is equal to zero. To predict air ventilation of the street http://www.400190.com/dissertation_writing/MBA/canyons, we define the positive (ACH+) and negative (ACH-) air exchange rates based on the direction of the fluctuating vertical velocity. Thus, ACH+ represents removal of air from the street canyon to the outside place while ACH- represents entry of air from the outside place into the street canyon.

Analysis of the pollutant exchange rate (PCH) can improve our understanding of the pollutant removal mechanism in street canyons. The transient behavior of the PCH is obtained by integrating the product of the instantaneous fluctuating vertical velocity across the interface and the instantaneous resolved-scale pollutant concentration. Similar to the representation of ACH, positive PCH represents removal of air pollutant from the street canyon to the outside place, while negative PCH represents re-entry of air pollutant from the outside into the street canyon.
 

2.4.3wall treatment
One challenge in CFD turbulence model is how to treat the thin near-wall sublayer, where viscous effects become important. In flows with mass transfer, an accurate resolution of this layer is crucial – because the pollutant source was located on the ground and a huge amount of pollutant would transport across this layer. The most reliable way is to use a fine grid and a low-Re-number model. However, this can be very expensive, particularly in our 3-D simulations. The traditional solution has been to use wall-functions, and some research also shows in many cases, a proper chosen high-Re-number model (like RNG k-e mode) combined with wall function can even provide a better performance than the low- Re-number model.#p#分頁(yè)標(biāo)題#e#

The idea of wall function is to place the first computational node outside the viscous sublayer, and make suitable assumptions about how the near-wall velocity profile behaves, in order to obtain the wall shear stress. In a local equilibrium boundary layer, where the length scale grows linearly with distance from the wall, we obtain the log law. Based on this assumption, we can develop the wall function as follows:
2.4.5 The PISO Algorithm for Transient Flows
In this research, 2D simulations were calculated by SIMPLE method and 3D simulations were calculated by PISO method.

SIMPLE (Semi Implicit Method for Pressure Linked Equations) Patankar (1980) method is a widely-used scheme for solving the pressure-velocity coupling problems. It is not compulsory to completely resolve the linear pressure-velocity coupling, when a steady-state problem is going to be solved iteratively. The differences between consecutive solutions are no longer small. In one SIMPLE loop, firstly an approximation of the velocity field is obtained by solving the momentum equation where the pressure gradient term could be calculated by pressure distribution (an initial guess for the first step or previous iteration for the following step), then the pressure equation is formulated and solved to get the new pressure distribution, after that the velocity fields are corrected and the next set of conservative fluxes is calculated.
Numbers of variants on the SIMPLE scheme, like SIMPLEC, SIMPLER, were developed to improve the performance of calculation: the accuracy, stability and efficiency. One such variant is the PISO (pressure implicit solution by split operator method) scheme proposed by Issa(1982) which is very similar in structure with SIMPLE scheme but in some application, especially for transient flow problems, lead to better convergence. In the PISO scheme, the same decomposition of velocity and pressure corrections is made as described earlier in the SIMPLE scheme. However, a second corrector stage is added to account for neglected corrections of velocity and pressure fields in the first stage.
It is highly recommended to adopt PISO algorithm for all transient flow calculations, especially a relatively large time step was set. This is because PISO method can maintain a stable calculation with a larger time step which allows an under-relaxation factor of 1.0 for both momentum and pressure.

3 Model Validation

The computational simulations of flow are basically numerical experiments to study the various flow properties in a domain concerned. Like calibration of experimental equipment, validation of model by using some standard benchmark tests is also crucial for computational analysis. To justify the reliability of the computational code and the quality of its performance, the validation cases are simulated and the results are compared with literature publications. The validation in this research includes two aspect, the comparison with the flow and scalar transport.

3.1 Validation of wind profile.#p#分頁(yè)標(biāo)題#e#

The experiments were conducted in a 12.6 m long hydraulic flume which has a1.2 m×1.2 m cross section with a 2.4 m long test section enclosed in tempered glass. A combination of turbulence spires and upstream surface roughness was used to generate a turbulent atmospheric boundary layer (ABL) flow with a mean surface roughness. The mean profile could also be fitted with a simple power law as summarized in equation (21).

where H = 50 mm was the height of the typical obstacle studied. The profile exponent of 0.26 is typical of flow over rough, urban-like, terrain. Detailed specifications of the flow field, including turbulence information, are found in the original report (Macdonald et al., 2000).

Regular arrays of obstacles, consisting of up to 24 rows, were placed in the test section of the hydraulic flume to simulate built-up areas. The surface obstacles consisted of 50 mm aluminum blocks or 50 mm sections of aluminum angle that were arranged in simple staggered and square (in-line) arrays. Cube arrays at three packing densities were tested, covering the range of isolated roughness flow (λf= 0.0625), wake interference flow (λf = 0.16) and skimming flow (λf = 0.44). The definitions of the packing density and other geometrical parameters used are shown schematically in Figure 1.

Each velocity profile in the arrays consisted of point measurements made at 24 heights between z = 0 and 400 mm using a SonTekTM acoustic Doppler velocimeter (ADV) operating at 20 Hz. This instrument uses the Doppler shift of a 16 MHz acoustic pulse to measure the three components of the fluid velocity. The acoustic pulse is reflected off of tiny particles transported by the flow and it is sampled by an array of three receivers focused on a small control volume (Figure 3). A two-minute sampling period was found to give stable, representative values of the mean velocity and turbulence variances.

Fluid velocities measured at a point in an obstacle array are very sensitive to proximity to the obstacles and show much spatial variability. In order to remove the large spatial variability in the measured velocity profiles, and to provide a representative mean profile, the results from the experiments were spatially averaged. In this procedure, the profiles measured at several cross-stream locations behind a given row in the obstacle array are averaged to give the mean profile. Because of the regularity of the obstacle arrays, this could be done with relatively few individual profiles (typically five). Figure 4 shows five separate profiles measured behind the ninth row in the λf = 0.16 cube array. The presence of a recirculation cavity is indicated by the negative mean velocities in the lower part of the profile measured directly behind the cube. However, after the spatial averaging, this type of information is ‘smeared’ out. The resulting data set discussed below includes spatially averaged profiles of mean velocity.

3.2 Validation of scalar transport

The wind tunnel experiments were carried out in the Dispersion Modeling Wind Tunnel at the BRE Ltd. Cardington Laboratory in Bedfordshire. This facility is designed and specially equipped for pollutant dispersion modeling using tracer gases, and has a working section 1.5 m high, 4.3 m wide, 22 m long. The upwind part of the working section is used to generate a suitable scaled model of the atmospheric boundary layer using Counihan’s (1969) system of a crenellated fence, vortices generators and a long fetch of roughness tiles. This produces a rough wall boundary layer. The boundary layer at the leading edge of the arrays was approximately 0.8 m deep, suggesting a scale factor of about 1 : 750 (Counihan, 1975). However, if it is assumed that the obstacles themselves (0.10m cubes) dictate the scales of turbulence within the array, and they are regarded as models of prototype buildings about 10 m high, then the scale factor is 1 : 100. The equivalent full scale value of z 0 would then be about 0.25 m, typical of suburban terrain with scattered low buildings. Mean wind speed and longitudinal turbulence intensity profiles were measured at the position immediately upwind of the obstacle arrays and are shown in the report by Hall et al. (1996a). Although logarithmic, the mean wind speed profile could also be expressed as Equation(21). The turbulence intensity near the ground in the simulated boundary layer was in the range between 0.2 and 0.3, which also agrees with Counihan’s recommendations for urban boundary layers. Dispersion experiments were carried out within the boundary layer in the downwind part of the working section. In the present work the fan speed was set to produce a reference wind speed of 1.5 ms-1 at the height of the cube obstacles (i.e. at 0.1 m). This yielded a reference Reynolds number of about 1.0×104, which is above the lower limit (4×103) for Reynolds number independence of shear flow around surface-mounted cubes (Snyder, 1981; Castro and Robins, 1975). Because of the low wind tunnel operating speed, conventional hot-wire anemometry could not be used. The mean wind speed and longitudinal turbulence were measured with a type LBJ pulsed-wire anemometer from MALVERN Precision Devices, England, which is capable of operating reliably down to 0.2 ms-1 in highly turbulent or recirculating flows. The pulsed-wire probe was mounted on an automated traverse gear. A theoretical and experimental investigation of the operation of the pulsed-wire is given by Bradbury and Castro (1971). When measuring the problems in the wind tunnel, the pulsed wire was operated at a “shot rate” of 4 pulses per second, for a total of 250 pulses (62.5 s) for each point measurement. This was sufficient to obtain stable mean and turbulent velocities. Concentration measurements were made by detecting a neutrally buoyant gas tracer (47% CH4 + 53% Ar) discharged from the source which consisted of a small discussing plug of about 10 mm diameter and which released the tracer gas with low discharge momentum. Gas samples were drawn from a sampling array in the wind tunnel through small bore tubing into one of three 20-port Scan valves and three Flame Ionization Detectors (FIDs) which sequentially sampled the measurement points. A fourth FID continuously sampled the background concentration of the tracer near the wind tunnel entrance and this was automatically subtracted from the measured concentrations. The concentration signals from the FIDs were sampled at a rate of 2 Hz and the total sampling time for each point measurement was two minutes, which was su¦cient to obtain a stable mean concentration. The source flow rate was adjusted to produce concentrations well within the linear range of the FIDs. The range and linearity of the FIDs were checked automatically at the start of each experiment against certified gas samples in the range 0-10,000 ppm of methane in air. It took approximately 24 min to sample all the measurement points for one set of problems. In order to ensure the integrity of the system, the first sampling location of all three FID systems was collocated, so that any differences between the FIDs due to calibration or other errors could be detected easily at the beginning of each test run. The agreement was usually better than 5%. The whole concentration sampling procedure was computer controlled and analyzed to produce printouts of the mean concentration in the nondimensional form: #p#分頁(yè)標(biāo)題#e#
C_K=(CUH^2)⁄Q (7)
In the wind tunnel, the basic experimental layout was as shown in Fig. 1, where the x-axis is aligned with the mean wind direction and the y-axis is across the wind. After the development of a suitable boundary layer, the models (consisting of cubes of height H"0.1 m) were laid out on a smooth surface of length 2.5 m along the x-axis. The plume dispersion was measured by lateral and vertical sampling arrays that were Þxed at the downwind edge of the smooth surface. The lateral sampling array was at ground level and spanned the width of the obstacle array (2 m or 20H), while the vertical sampling array rose to 0.5 m (5H) above the ßoor. An extra row of cubes was always added behind the sampling arrays to ensure representative ßow around the sampling positions. The sampling array was set at 0.5 H behind the next row upwind; or for arrays with spacing less than 0.5H, the sampling stations were located at the upwind faces of the dummy row. Dispersion measurements was made for the source upwind of the array (in front of a cube), but with the sampling array at varying numbers of rows downwind. This type of measurement set-up is shown in Fig. 1 and is representative of plumes passing from relatively smooth terrain before impinging on the obstacle arrays

As shown in Fig. 1, the plume source was located on the ground on the centerline of the array, except in some cases of acute wind angles where the source had to be laterally offset to ensure the centre of the plume fell on the centre of the concentration sampling array. The source was placed a distance (S-0.5H) upwind of the first row.

5 3D Case
5.1 Space-average pollutant concentration inside canyons
Figure 1-Figure 4 presents the unsteady behaviours of the space average pollutant concentration inside the street canyon for different geometric settings. Canyon 0 represents the canyon with pollutant source, Canyon 1 represents the first canyon next to Canyon 0 in the downwind direction and Canyon 2 represents the second one. The sampling period was between 3200s and 4000s.

Regular oscillating behaviours were observed in Figure 5.1. In this case, all the geometry parameters, including building height, building width in streamwise and spanwise direction and street width are all set to 1 meters. This case is treated as a reference when we study and compare other cases. The curve of space-average pollutant concentration in Canyon 0 resembles sinusoid. The period is about 200 seconds and the swing of concentration is about 0.004. The space-average pollutant concentration of Canyon 1 and 2 is approximately 18.5% and 12.5% of the Canyon 0, and also follow very simple and regular patterns: the pollutant in Canyon 1 is always more than that in Canyon 2 and the oscillating periods of these two canyons are as same as that of Canyon 0.

Figure 5.2 depicts the space-average pollutant concentration when the canyon width was doubled. The oscillating behaviours are still relatively regular, though the curve of concentration in Canyon 0 is not sinusoid any more. The period of oscillation is also doubled as the street width, which imply the pollutant transport and accumulation mechanisms follow the similar patterns in both WR=1 and WR=2 cases. The pollutant concentration in Canyon 0 are approximately 1.5 times larger than that in case of WR=1, and the pollutant amount compare to Canyon 0 in Canyon 1 and 2 is much less. The result demonstrates that the wider canyon would embarrass the pollutant transport in streamwise direction. Figure 2 also shows an interesting phenomenon, the space-average pollutant concentration in Canyon 1 and 2 are very close, and in some time point, the pollutant concentration is even higher in Canyon2, which means in some situation, far away from pollutant source may not be lead to a better environment.#p#分頁(yè)標(biāo)題#e#

When the canyon width becomes even larger, in Figure 3, where the WR=3, the irregular behaviours are hard to summarise as oscillation. It is possible the cycle of this case is too long to be allocated in the current sampling period, however, under such assumption, the cycle period increase much larger than its expected value under the same rule of WR =1 and 2. Therefore, the pollutant transport mechanisms have to be changed in the settings.

Figure 5.4 reveal the change caused by different building height. In this case, WR was remained to 1 while the building height was set to 2 (AR=2). The area source amount is as same as case 1 but the canyon volume is doubled, thus the general pace average pollutant concentration dropped to about 0.15, half of the value in case 1. The oscillating period increases to about 600s and the curve become much rougher. The most interesting findings in the case is the non-synchronization of pollutant concentration in Canyon 1 and 2. The all other case which only have difference in WR but with the same AR(=1), though the swings or periods might changes, the concentration curves of Canyon 1 and 2 are always in phase. In Figure 4, we can observe that the crest of concentration curve in Canyon 1 is always corresponded to the trough of Canyon 2. In considerable long period, the pollutant concentration in further location (Canyon 2) is higher than that of closer location (Canyon1).

Figure 5.5-5.8 depicts the pollutant transport situation to the upwind direction. Normally, the pollutant should be blow by the wind and transported to the downwind region. In case of WR=1,3 and AR=4, the space average pollutant concentration are near to 0, while in case of WR=2, this amount cannot be neglected. The pollutant concentration inside the canyon -1 is comparable to that in the downwind canyon 0 and 1.

4.2 WR=1

The data was sampled at time point 3000s, when the free flow had already travelled the whole computational domain over 20 times, and lasted long enough to obtain about 5 oscillating circle. In this case, the ACH has little difference between each canyon (<10%) as shown in Table 1, which assure the air ventilation performance would not be affected by the boundary and the location of the computational domain, and provide a suitable surrounding to study the pollutant transport. Both in upward and sideward interface, "ACH'" possessed the majority, but not definitive. In upward interface, the amount of air ventilation lead by turbulence is about 3 times huger than that lead by mean flow, while about 1.5 times in the sideward interface.

For the downwind neighbouring street canyons (1 and 2), we can learn that the pollutant is entering the canyon from the side and leaving from the top. A universal character for pollutant transport in the case is: In upward interfaces, turbulence component dominates the pollutant transport, while in sideward interfaces mean component dominates, no matter removal or entrainment.

For the upwind canyons, the pollutant entrainment is too small to be noticed, therefore in this case no analysis is processed for upwind pollutant transport.#p#分頁(yè)標(biāo)題#e#

In Canyon 0 where located the area pollutant source, both the upward and sideward interface have their contribution to pollutant removal. However, the component which contributes more in these two directions is totally different. Learning from Table1, at upward interface, the turbulence component ("PCH'" ) dominates the pollutant removal, and the mean flow could only carry a very slight scale of pollutant back into the street canyon. In sideward direction, the pollutant removed by turbulence is in the same scale as the upward interface, however as this interface is directly linked to the ground pollutant source, the pollutant removed by the mean flow becomes the majority, which have a significant larger level. In Canyon 0 where located the area pollutant source, both the upward and sideward interface have their contribution to pollutant removal. However, the component which contributes more in these two directions is totally different. Learning from Table1, at upward interface, the turbulence component ("PCH'" ) dominates the pollutant removal, and the mean flow could only carry a very slight scale of pollutant back into the street canyon. In sideward direction, the pollutant removed by turbulence is in the same scale as the upward interface, however as this interface is directly linked to the ground pollutant source, the pollutant removed by the mean flow becomes the majority, which have a significant larger level.

Figure 10 and 11 depict the PCH distribution in the upward and sideward interface. From the Figure 9 and 11, we can find the overall PCH contour is very close to ¯("PCH" ), which implies that in most individual position, ¯("PCH" ) dominates the local pollutant transport. However, those transport processes perform not only removal but also entrainment. At canyon top, these two contrast effects finally present the ¯("PCH" ) as a subordinate component for pollutant removal. For the sideward direction, a strong spanwise rotation inside the street canyon carry the pollutant out of the canyon at the ground level and take in at the upper level. As the mean flow near the ground could carry a large amount of pollutant directly from the source canyon, the overall ¯("PCH" ) perform a very large positive value compare to its upward counterpart in canyon 0. (Figure 10b)

The figures also show that the peaks of "PCH'" (which take main contribution to the fluctuating component of pollutant removal) of upward and sideward are all near to leeward region of the canyon, where k is relatively weak. Equation() indicates the "PCH'" depends on both the pollutant concentration and turbulence. In canyon 0, the peaks of "PCH'" (Figure 11) is located in the contrast direction of the peak of k (Figure 12). When pollutant emits from the source canyon, as the concentration is higher, especially in the leeward facade where the pollutant is accumulated by the streamwise circulation, the distribution of pollutant decides the "PCH’ " in the both upward and sideward interface.#p#分頁(yè)標(biāo)題#e#

In the neighbouring canyons, as the ventilation effect is still significant, and the pollutant from source will move upward by both ventilation and diffusion, the pollutant entrainment lead by the spanwise rotation in upper of sideward interface is significant, while the ground without source has much less pollutant to travel out, the overall ¯("PCH" ) performs a large negative value

Figure 15 show that the "PCH'" peaks of upward and sideward in Canyon 1are all near to windward region of the canyon, where are peaks of k too. Based on Equation(), in this case "PCH'" mainly depends on turbulence. Unlike canyon 0 which contains the pollutant source, canyon 1 has a much less pollutant concentration, but the same level of ventilation. It is the ventilation properties, not the pollutant distribution, decide the pollutant removal mechanism in the canyon.

In canyon 2, the distance to source canyon increases and the pollutant concentration inside canyon becomes even smaller, while the ventilation properties remain. Thus the pollutant transport mechanism of canyon 2 follows the same pattern of canyon 1, as shown in Figure 17.

4.2 WR2

Figure 2 presents the average concentration of the canyon 0, 1 and 2 when building’s width equals to 2. The period of the curve increases from near 200s in previous case to about 400s, while the amplitude increase also in the same scale. In canyon 0, both the pollutant source area and canyon volume increase by 2 times, while the average concentration increase, this result shows the wider street canyon is less efficient for pollutant removal. The amplitude of concentration curve also increased, which implies a more drastic change of environment inside the street canyon.

Table 2 posts its great difference to the previous case: Though the upward interface area of the street canyon was enlarged about 2 times, the ACH increase little. The comparison between Table 1 and 2 shows that ¯("ACH" ) was actually increased (by about 1.5 times) while "ACH'" had nearly no improvement. Since "ACH'" hold the majority of the ACH, this results in the poor amelioration of air ventilation.

However, the changing of ACH structure has much more influence to the PCH composing than to itself. In the upward face of canyon 0, the negative region of PCH decreased, and this change agrees with the conclusion that the overall performance of pollutant removal here is improved. In the sideward interface, pollutant removal only occurred in the leeward corner.

Compare Figure 10-b and Figure 19-b, in both settings, the negative ¯("PCH" ) is allocated near the windward face end, and the positive ¯("PCH" ) is concentrated at the leeward face central. This PCH distribution is the result of the air ventilation. As the air in the artery is directly driven by the free stream and has a higher velocity, the flow inside the street canyon is blocked inside and forms the spanwise sub-recirculation like the streamwise main-recirculation inside the canyon driven by the free stream, but weaker because the channelling flows in the artery are weaker than the free stream. As a result, the pollutant would be accumulated in the leeward central facade, and removed by main-recirculation. (Figure 9)The wider streets allow the full development of this sub recirculation and therefore the ¯("PCH" ) improved.#p#分頁(yè)標(biāo)題#e#

For the sideward interface, the two cases follow the similar pattern, and the each components of PCH just simply increased by about 1.5 times, the same level of average pollutant concentration increased.

From Figure 11,12,20,21, we can find the width of the street canyon increased, the basic pattern of PCH’and k remain similar with previous case. The k achieves maximum in the windward facade and minimum in the leeward facade. The PCH’ peaks near the sideward interface and close to the leeward facade and in the different position of k peak, showing the dominance of pollutant concentration to PCH’.

Based on Table 1 and 2, we can find the overall PCH on upward interface decreased when street canyons become wider. However, the dominant component of these two cases is different. In the upward interface, when WR=1, PCH’ take the majority of the overall PCH while the total amount of ¯("PCH" ) can be neglected. When WR=2, only ¯("PCH" ) contributes to the pollutant removal, while PCH’ contributes to pollutant entrainment.

The pattern of the sideward pollutant removal changes obviously. In the WR=1, the situation is very ideally. PCH of one side is about half of that of upward interface but with different direction, which indicates a conversation inside the domain. However, the magnitude of PCH of one side is much larger than the upward PCH, which implies an unstable situation inside the canyon.

From Table 2 and Figure 25, 26, we can find the upward pollutant removal and its component of Canyon2 just simply decreased and follows the similar pattern with Canyon 1, however, the sideward PCH performs totally different. The sideward interface has a strong positive PCH here which is in contrast with the Canyon 1. The data shows the PCH’ just simply decreased as in the upward interface, the difference is because of (PCH) ?. The magnitude of PCH in sidewards interface is much larger than that in upwards interface, which implies that less pollutant is transported to the canopy and the majority of pollutant is still move among t the street level.

4.3 AR=2

Figure 4 presents the average concentration of the canyon 0, 1 and 2 when building’s height equals to 2. The pollutant transport patterns have great difference with other case. Because the roof height doubled, most of pollutant prefer to leave canyon0 through the sideward interface. Compare to the same case in 2D situation, as the pollutant can only transport through the roof, though the building height doubled, the PCH in the roof only reduce in half. In the 3D case, we can find from Table 1 and Table 3 that the PCH through upward interface decreased greatly and neglectable when compare the sideward PCH

In Figure 4, the <C> curve of canyon 1 and 2 present a increasing pattern, which implies in the sampling period, the pollutant haven’t achieve a stable statues inside the canyon 1 and 2. Thus in this section the discussion would only focus on the Canyon0.#p#分頁(yè)標(biāo)題#e#

In the sideward interface of canyon 0 of this case, (PCH) ? also dominates the pollutant transfer. However, the amount decreased about half. Figure 28 shows the difference of (PCH) ? pattern with case I. Compare to Figure 9 and 10, we can find in case I, (PCH) ? peaks at the leeward corner of the ground while at the contrast direction in case III. From the previous analysis, we know the situation of case I is lead by the recirculation inside the canyon. And this circulation in case I is the only one while when buildings becomes higher, the bottom of the canyon will generate a sub recirculation with opposite rotating direction of the upper recirculation. This sub recirculation is generally weaker than the upper one, and leads to a relatively smaller PCH.

The sideward PCH’ just decreased with a vary small scale. From Figure 28c and d, we cannot find a tight relation between the PCH’ peak and TKE peak, the PCH’ just simplt peaks at where the pollutant accumulate. This pattern follows the previous two cases.

From the figure, we can find the PCH across the upward interface of Canyon 0 decrease about 10 times, However, as the upper side of the canyon is still dominated by the main (upper) recirculation, the 留學(xué)生dissertation網(wǎng)patter of (PCH) ? and k haven’t changed too much.
However, PCH’ peaks asymmetrically, which shows an obviously oscillating behaviours.

As Canyon II and III haven’t achieve stable status, here just lists figures of their pollutant removal properties as references.