Geometric effect of buildings on the dispersion of carbon dioxide cloud in idealized urban street canyons

Process Safety and Environmental Protection 122 (2019) 271–280

Contents lists available at ScienceDirect

Process Safety and Environmental Protection journal homepage: www.elsevier.com/locate/psep

Geometric effect of buildings on the dispersion of carbon dioxide cloud in idealized urban street canyons Wei Tan, Chaojie Li, Kang Wang, Guorui Zhu, Liyan Liu ∗ School of Chemical Engineering and Technology, Tianjin University, Tianjin, China

a r t i c l e

i n f o

Article history: Received 20 July 2018 Received in revised form 12 December 2018 Accepted 13 December 2018 Available online 23 December 2018 Keywords: Geometric effect Flows in street canyons CO2 dispersion Urban micrometeorology

a b s t r a c t This paper focused on the dispersion of carbon dioxide (CO2 ) cloud due to large accidental release in urban areas. A three-dimensional urban model was designed and applied to capture the circulation flow and channeling flow in street canyons. The effects of building height, aspect ratio and roof shape on CO2 cloud dispersion were studied by wind tunnel experiment and Computational Fluid Dynamics (CFD) method. Building Reynolds number was calculated to ensure Reynolds number independence and turbulent flow. The results of model evaluation indicate that 100% of the predictions were within a factor of two of the experimental measurements (FAC2), and the relative mean bias (MG) was about 7%. CO2 behaved as a dense gas, whose volume fraction was more than 30% near the ground and about 5% in the upper region, respectively. Only 0.5% of CO2 removed from the canyon was through the top opening. Increasing windward-side building height enhanced vertical recirculation, thus reduced CO2 volume fraction from 22.9% (equal to 422 g/m3 ) to 14.5% (267 g/m3 ). Increasing aspect ratio to 3/2 could take greater amount of air into the canyon through the larger area of the openings, and then reduced CO2 volume fraction to 15.9% (293 g/m3 ). © 2018 Published by Elsevier B.V. on behalf of Institution of Chemical Engineers.

1. Introduction There are increased concerns about large accidental releases of hazardous gas from pipelines or storage vessels in urban areas (Soulhac et al., 2016; Zhang et al., 2017). In the Viareggio liquefied petroleum gas (LPG) accident, about 46.7 t of LPG was released from a breach of the overturned tankcar and the flash fire resulted in 32 fatalities (Brambilla and Manca, 2010; Manca and Brambilla, 2010). At a Samsung facility in southern Seoul, safety system released carbon dioxide (CO2 ) after it wrongly detected fire and left one worker dead. An accidental release of hazardous gas can behave like a dense gas frequently, due to the high molecular weight of the substance, the low temperature, or the presence of aerosols (Pontiggia et al., 2009). Then the hazardous gas would keep close to the ground and present a great threat to human safety. A thorough understanding of dense gas dispersion in urban areas can be helpful in the emergency response to such accidents. Atmospheric dispersion models can be classified into three categories: Gaussian model, “Similarity-profile” model and Computational Fluid Dynamics (CFD) model (Liu et al., 2017b). Gaussian

∗ Corresponding author at: Tianjin University, 135 Yaguan Road, Haihe Education Park, Tianjin, 300350, China. E-mail address: [email protected] (L. Liu).

and “Similarity-profile” models are just mathematical models derived from physical observations of the field test and cannot capture the effects of obstacles. CFD model has an advantage over these two models as it allows the simulation of gas dispersion over complex terrains (Kumar et al., 2015; Liu et al., 2016; Toja-Silva et al., 2018). CFD model has been widely applied to the prediction of flow and dispersion in the simplified and idealized urban areas (Madalozzo et al., 2014; Carpentieri and Robins, 2015; Yuan et al., 2014). Scargiali et al. (2011) applied Computational Fluid Dynamics X (CFX) to investigate dense gas dispersion in an urban area, which was modeled as a simple network of straight roads. This approach could help setting up gas dispersion model in urban areas, but the utmost simplified urban model could not display the complex flow field well. Wingstedt et al. (2017) investigated dispersion of neutral and dense gas over a simplified urban area. Liu et al. (2017b) presented CFD models for the dispersion of CO2 over complex terrains. In the above mentioned two studies, the source installation was some distance from the urban model, which was slightly different from the leakage accident in street canyons. Tan et al. (2018a, 2018b) studied the dispersion of CO2 cloud in a long street with an intersection, and divided the whole process of gas dispersion into two stages. In the first stage, the motion of CO2 cloud was primarily determined by jet action and slightly affected by wind in near-source region. The next stage began with the collapse of

https://doi.org/10.1016/j.psep.2018.12.020 0957-5820/© 2018 Published by Elsevier B.V. on behalf of Institution of Chemical Engineers.

272

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

CO2 cloud, and gravity and wind together influenced CO2 dispersion. But the variability of building geometry was not taken into consideration. Urban areas are so varied that the law of dispersion in a specific building layout cannot be appropriate for the full range of building shapes and layouts. In the current paper, a three-dimensional urban model was designed and applied. This model was simple but could describe urban structure more realistically, and then, could capture the complex flow field in street canyons well. Based on the model, this paper investigated CO2 dispersion in the street canyon with building height variation, aspect ratio (W/H, where W was the street width and H was the building height (Chang and Meroney, 2003; Scungio et al., 2015)) variation and roof shape variation respectively. Meanwhile, CO2 behaved as a dense gas and the results can be used as a reference for the accidental releases of other hazardous gases such as chlorine or liquefied natural gas (LNG). The discussion of dense gas behavior in the street canyon, coupled with the comparison between CO2 and nitric oxide (NO), were also presented. 2. Methodology 2.1. Wind tunnel experiment 2.1.1. Experimental apparatus Wind tunnel experiment was conducted in an open-circuit wind tunnel with a cross-section of 0.75 m × 0.5 m, as shown in Fig. 1. It included both fog experiment and CO2 dispersion experiment. Fan system was made up of fan device, fairing and flow deflector. The hexagonal cellular device and tangled net followed the entrance, to decrease the separation of air flow and vortex. The wind tunnel continued with a zone with spires (with width of 0.054 m and height of 0.36 m) and roughness elements (cubes with sides of 0.018 m and with sides of 0.036 m), where the boundary layer for the appropriate urban roughness was developed. CO2 cylinder, flow meter, pressure regulator and three-way valve composed the release system. The CO2 inlet direction was oriented vertically upwards and could eject CO2 with purity 99.9%. Four infrared sensors were attached on the outer surface of buildings to obtain CO2 volume fraction, as labeled with upsidedown red triangle. Measuring range of sensors was 0–100% and precision was 0.1% by volume. Temperature and relative humidity were measured by hygrothermograph. The flow behavior was analyzed qualitatively through fog experiment. The source was positioned in the middle of the street canyon, and fog was released by the suction force induced by the

Fig. 2. Velocity at the height of building.

incoming air flow and buoyance. An ordinary video camcorder was used to record the gas flow pattern. The wind direction was perpendicular to the target street canyon (marked with blue lines). A hot-wire probe at the height of building was used to measure the upstream velocity. Wind velocity was measured at a frequency of 1 Hz, as shown in Fig. 2. Then mean wind velocity, turbulent kinetic energy and its specific dissipation rate were gained through measured data processing. Turbulent kinetic energy and its specific dissipation rate were calculated from the measured turbulence intensity and turbulence integral scale. Note that turbulence integral scale was measured based on the Taylor’s frozen field hypothesis, which was utilized to convert time-resolved hot-wire data to spatial data (Tutkun, 2017). Then calculated by Lu =



U  2u

∞

Ru () d

(1)

0

where, U was the mean velocity at different height,  u 2 was the variance of fluctuating velocity u(t) and Ru () was the autocorrelation function, which defined as R () =

E[u (t) u (t + )]  2u

(2)

Building Reynolds number was adopted to ensure Reynolds number independence and turbulent flow. ReH = UH · H/v

(3)

where, UH was the mean velocity at the height of building, 0.6594 m/s in this study; v was kinematic viscosity of air, 1.48 × 10−5 m2 /s. ReH was 2.23 × 103 , larger than the value of Recrit which was 2.1 × 103 (Uehara et al., 2003; Cui et al., 2017).

Fig. 1. Wind tunnel and measuring points.

2.1.2. Physical model configuration Street canyons, as the basic elements in urban areas, are jointed through the intersection, and further form the urban fabric (O’Neill et al., 2016; Mo and Liu, 2018). This can be shown by taking the place where the 1983 Stockholm hydrogen accident occurred for example (Venetsanos et al., 2003). An important prerequisite for this study is the development of urban model more realistic than the usual array of cubic buildings. The three-dimension urban model with building height variation (Case 2 and Case 3), aspect ratio variation (Case 4 and Case 5) and roof shape variation (Case 6) was applied, as shown in Fig. 3. The top view shows that the model consisted of five streets along x axis, connected through a streamwise street along y axis. Through setting up the intersection, simulation

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

273

Fig. 4. CO2 volume fraction of each measuring point for Case 2.

tion and z positive upwards. Measuring points’ coordinates were respectively Point 1 (0, −0.025, 0.025) near CO2 inlet, Point 2 (0.125, −0.025, 0.025) located on the leeward side, Point 3 (0.125, 0.025, 0.025) located on the windward side, and Point 4 (0, 0.05, 0.05) above the windward building. The possible source of an accidental release of hazardous gas is usually a device that contains a larger amount of harmful gas. Due to the high pressure in pressured vessel or pipeline (Witlox et al., 2014; Liu et al., 2015; Joshi et al., 2016), the leakage rate is always high. In this study, CO2 was released continuously at a velocity of 0.5628 m/s, corresponding to the leakage rate of about 0.2688 m3 /h. The release velocity can not only reflect the initial momentum of jet on dispersion but also keep most of CO2 cloud in the street canyon. CO2 volume fraction was measured at 1 Hz, as shown in Fig. 4. In the experiment, the duration of the release was 240 s. CO2 volume fraction increased rapidly in the first 40 s, and then tended to be stable. However, there were still minor fluctuations at this stage. After stopping release, the volume fraction tended to decrease rapidly. Each experiment was repeated three times, and the average volume fraction between 120 s and 240 s was calculated for model evaluation. 2.2. CFD setup

Fig. 3. Configuration of the urban street canyons (Red point represents CO2 inlet position, green point represents Point 1, black point represents Point 2, yellow point represents Point 3 and blue point represents Point 4) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

of the helical vortex in street canyon could be achieved. Two of the streets along x axis were located upstream of the target street canyon to explicit develop an urban boundary layer, while two were located downstream. The width of all the buildings (B) was 0.05 m, and other sizes were tagged in Fig. 3. CO2 inlet was a circle with a diameter of 0.013 m. The coordinate system considered CO2 inlet as the origin with y measured as positive in the downwind direc-

Through CFD numerical simulation, flow characteristics and cloud dispersion features can be precisely described. The CFD software ANSYS-Fluent was adopted in the simulation. The simulations were conducted in two steps: the flow field in the street canyon was computed before CO2 release by a steady simulation, and then used as the initial condition to study CO2 cloud dispersion in transient condition. As CO2 was released at ambient pressure, flow was defined as incompressible gas. Boussinesq approximation (Scargiali et al., 2005) was adopted for the buoyancy treatment. 2.2.1. Turbulence model Particular attention has to be paid to turbulence model as it plays an important role in the simulation of dispersion (Pontiggia et al., 2010). The SST k–ω turbulence model is a two-equation eddyviscosity model. In the boundary layer, the k–ω formulation makes the model applicable through the viscous sub-layer down to the wall. And in the free-stream, it can switch to a k-ε form and thereby avoid the problem that the k–ω is too sensitive to the inlet freestream turbulence properties (Tauseef et al., 2011; Xing et al., 2013; Wen et al., 2016; Yu and The, 2016).

274

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280 Table 1 Summary of the parameters used in the simulations.

Fig. 5. Mesh on part of vertical centerplane (Red segment represents CO2 inlet position) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

2.2.2. Domain and mesh The dimension of computational domain was the same as that of wind tunnel. The domain had an upstream length of 10 H, a downstream length of 29 H, a lateral length of 4.5 H, and a height of 10 H, which met the practice guidelines suggested by the literature (Franke et al., 2011). Structured mesh was used, and appropriate refinement close to CO2 inlet, walls and ground was performed, as shown in Fig. 5, which made y+ value (the value of normalized distance to the wall) beyond 1. To ensure a mesh independent solution, CO2 volume fraction at Point 2 and Point 3 was compared, as shown in Fig. 6. The results from the mesh amount of around 3 million deviated less than 1% as compared to those of around 6 million. Therefore, the mesh of around 3 million elements was selected keeping in view both the accuracy and the computational effort. 2.2.3. Boundary conditions Appropriate boundary conditions are the key to cloud dispersion simulation (Garcia-Sanchez et al., 2017; Liu et al., 2017a). Briggs et al. (2001) described three wind tunnel experiments which measured the rate of vertical growth of dense gas cloud, and gave criteria for the proper wind and turbulence profiles. In the current paper, boundary conditions at wind inlet were also described by profile of mean wind velocity and turbulent structure functions. Profile of mean wind velocity was obtained through measured data fitting, and expressed by the power law according to AIJ (Architectural Institute of Japan) guidelines (Tominaga et al., 2008). ␣ = 0.308 in the power law profile was appropriate for urban areas (CCPS, 1999). The turbulent kinetic energy and its specific dissipation

H/m

UH /m·s−1

IH /%

LH /m

0.05

0.6594

6.53

0.328

rate profiles could represent the influence of spires and roughness elements, which were not included in the domain. And then, fluctuating velocity components could be derived to predict turbulent dispersion (Kim and Baik, 2003; Gorle et al., 2009). A user-defined subroutine for including the profiles into the FLUENT code was developed. The vertical profiles of mean wind velocity, turbulent kinetic energy and its specific dissipation rate are given below: U (z) = UH

 z 0.308

(4)

H

where U(z) was the mean velocity at z. 2 · k (z) = 1.5 · UH2 · IH

 H 0.566

(5)

z

where I was the turbulence intensity. ω (z) =

2.236 · UH · IH ·

 H  0.595 z

(6)

LH

where L was turbulence integral scale. The numerical values of the involved parameters used in the simulations are shown in Table 1. At the top surface and two lateral sides of the domain, symmetry condition was specified. Velocity inlet condition was for wind inlet and CO2 inlet. And the outflow condition was assumed at the downstream boundary, as presented in Fig. 7. 2.2.4. Model evaluation CO2 volume fraction was one significant concern, and was selected as the parameter for model evaluation in this research. CO2 volume fraction at the monitoring points, corresponding to the measuring points in the wind experiment, had reached a relative stable value at t = 15 s. The calculated volume fraction of CO2 for Case 2 and Case 4 were compared with experimental data, as shown in Table 2. For Case 4, Point 4 was covered by the roof, so the value was not in view. The simulated value was very consistent with the experimental data at point 1. However, the coincidence degrees at Point 2, Point 3 and Point 4 were lower because of the predominant function of flow field for dispersion (Nosek et al., 2017; Park et al., 2015; Karra et al., 2017). The prediction of the flow field was of primary importance to gas dispersion, thus the comparation of flow velocity would also be useful for model evaluation. The fog experiment was conducted to seek a qualitative analysis for the flow behaviour. The numerical simulations agreed well with the flow behaviours observed in the experiment, which can be seen in Section 3.1. Statistical performance indicators, including the geometric mean bias (MG), the geometric mean variance (VG), and the fraction of the predictions within a factor of two of experimental measurements (FAC2) were used to evaluate the validity of the model (Chang and Hanna, 2005), as shown below:



MG = exp 1n CO − 1n CP



(7)

VG = exp (1n Co − 1n CP )2 

Fig. 6. Mesh-independence examination.

FAC2 = fraction of data that satisfy

(8) 0.5 ≤

Cp ≤ 2.0 Co

(9)

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

275

Fig. 7. Computational domain and boundary conditions.

Table 2 Volume fraction of CO2 at measuring points. Case

2 4

Point 1

Point 2

Point 3

Point 4

Experimental

Simulated

Experimental

Simulated

Experimental

Simulated

Experimental

Simulated

0.292 0.337

0.294 0.319

0.176 0.088

0.222 0.112

0.101 0.081

0.201 0.049

0.026 ----

0.033 ----

Table 3 Summary of the statistical performance. Case

MG

VG

FAC2

2 4 Median

0.747 1.111 0.929

1.157 1.110 1.134

1.000 1.000 1.000

where Cp was the prediction of the volume fraction, Co was the observation of the volume fraction, and was the average over the data set. A perfect model has MG, VG, and FAC2 = 1.0. Due to the influence of complex flow field with numerous feature structures (Oke, 1988; Blocken et al., 2011; Merlier et al., 2018), 0.7 < MG < 1.3, VG < 1.6 and 0.5 < FAC2 < 1 represent acceptable ranges (Hanna et al., 2004). As shown in Table 3, the statistical performances of experimental and calculated volume fractions were all within acceptable ranges. All the predictions were within a factor of two of the experimental measurements, and MG was about 7%. The model was satisfactory for CO2 dispersion in urban street canyons. 3. Results and discussion 3.1. Flow characteristics The vertical plane at x = 0 m, vertical plane at y = 0 m, and horizontal plane at z = 0.025 m, as shown in Fig. 8, were the research areas in this study. For better analysis of CO2 dispersion in the target street canyon, the flow field and CO2 volume fraction in larger regions were demonstrated. The contours of velocity magnitude and streamlines obtained by the steady simulation are shown in Fig. 9.

Fig. 8. The vertical plane at x = 0 m (blue), the vertical plane at y = 0 m (black) and the horizontal plane at z = 0.025 m (red) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

The flow structures in different cases displayed large differences. For Case 1, one vertical clockwise vortex caused the downward flow on the windward side and the upward flow on the leeward side, the same as that reported in the literature (Li et al., 2016). The downward velocity near the windward side was larger than 0.2 m/s, while the upward velocity near the leeward side showed a low value. Horizontally, airflow entered the street canyon at the windward side, forming three vortices. The wind velocity in the streamwise street along y axis was high, ranging from 0.2 m/s to 0.5 m/s, so the outward channeling flow was dominant. Under the action of the vortex near the intersection, the velocity the at the leeward side became larger than that at the windward side. Increas-

276

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

zone with low relative velocity. The horizontal flow structure was also characterized by two counter-rotating vortices, and the velocity was less than 0.1 m/s. When aspect ratio was 3/2, a lager vertical vortex was created. There was a stagnant zone in the lower region near the leeward wall. As the area of the top and lateral openings became larger, greater amount of air could flow into the canyon. The flow structure in the canyon with slanted-shaped roof was found to be significantly different from that with flat-shaped roof. The form of the vortex in the upper region was influenced by the roof shape. On the horizontal dimension, the whole street canyon was influenced by two large-scale counter-rotating vortices. The flow field in the canyon contained various types of flow patterns, including vertical circulation flow, horizontal vortex and the outward channeling flow. The vertical circulation flow can carry cloud towards the roof level and lead to high concentration at the leeward side. The horizontal vortex near the intersection can cause the removal of cloud through the left lateral opening. The outward channeling flow and the horizontal vortex are the major cause for the removal through the right lateral opening. Fog experiment was conducted to seek a qualitative analysis for the flow behaviour. With respect to Fig. 10, fog rose up to the roof level due to the suction force induced by the incoming air flow and buoyance. Most of fog would disperse to the downwind areas with the flow aloft. The other part re-entrained back into the canyon with the vertical circulation flow. When increased leeward-side building height, almost all the fog dispersed in the downward wind direction due to disappearance of the vertical vortex. For a high windwardside building, the vertical recirculation took almost all the fog back into the street canyon at the windward side. The flow feature of fog in Case 4 was similar to that in Case 2. Most of fog rose up to the top, and this was because the vertical vortex played a less important role than the force pushing upwards did. In Case 5, fog accumulated near the leeward wall in the lower layer, corresponding to the stagnant zone. In the canyon with slanted-shaped roof, fog had obvious trend to the leeward roof.

3.2. CO2 dispersion characteristics

Fig. 9. Contours of velocity magnitude and streamlines in the vertical plane at x = 0 m (left) and in the horizontal plane at z = 0.025 m (right). White arrows represent wind direction.

ing the leeward-side building height led to the disappearance of the main vertical vortex. This was because the diameter of the theoretical vortex was larger than the street width and the windward-side building would prevent it from forming. On the horizontal dimension, only two vortices were observed. However, the presence of a high windward-side building halted air aloft traveling downstream and took them into the canyon. Even at the leeward side, the upward velocity could be up to 0.2 m/s. Horizontally, the peak velocity at the leeward side was larger than 0.3 m/s and that at the windward side was larger than 0.2 m/s. In the case of aspect ratio equal to 2/3, two small counter-rotating vortices appeared in the

The time step was 1 × 10−3 s in the transient simulation to assure Courant number was less than one, which means that the simulation time step was smaller than the fluid flight time over the cell. The calculation was carried out up to t = 15 s. To conduct the comparison between dense gas and neutral gas, the simulation with NO was performed. The leakage rate of NO was the same as that of CO2 by volume. The volume fraction and concentration of CO2 and NO in the vertical plane at y = 0 m are shown in Fig. 11. The overall view shows that both of the volume fraction and mass concentration of CO2 were larger than those of NO. CO2 volume fraction was over 50% in the near-source zone, and decreased with the distance from CO2 inlet. CO2 cloud appeared a stratification pattern in the street canyon, with the volume fraction more than 30% near the ground and about 5% in the upper region, respectively. Unlike CO2 , most of NO dispersed through the top opening. The volume fraction of NO remained about 10% from 0.2 H to H, and was negligibly small near the ground. The volume flux ratio dispersed from the street canyon through each canyon opening were also computed respectively, as shown in Fig. 12. The total volume flux consisted of the volume flux through the top opening, the left lateral opening and the right lateral opening. CO2 behaved as a dense gas in the street canyon and only 0.5% was removed through the top opening. The flux ratio of CO2 removed through the left lateral opening was 56.8%, and that through the right lateral opening was 42.7%. For NO, 95.6% dispersed through the top opening.

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

277

Fig. 10. Flow visualization with fog releasing from the inlet (looking toward the negative x axis). White arrows represent wind direction.

Fig. 11. Volume fraction and mass concentration of CO2 (left) and NO (right) on vertical dimension at y = 0 m.

The volume fraction and mass concentration of CO2 at the leeward side and the windward side are shown in Fig. 13. CO2 accumulated along with the leeward side of building, resulting in the high volume fraction and concentration. Taking CO2 volume fraction at the ground level for instance, it ranged from 30% to 50% at the

leeward side and was less than 30% at the windward side. Besides, CO2 volume fraction decreased with the distance from CO2 inlet increasing at the leeward side, while the peak at the windward side appeared on the area where the outward channeling flow mixed with the horizontal vortex near the right lateral opening.

278

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

Fig. 12. The volume flux ratio dispersed from the street canyon through each canyon opening.

Fig. 13. Volume fraction and mass concentration of CO2 at the leeward side (left) and at the windward side (right).

Fig. 14. Contours of velocity magnitude and streamlines in the vertical plane at x = 0 m (left) and in the horizontal plane at z = 0.025 m (right) for case 1 at t = 15 s. White arrows represent wind direction.

For dense gas, not only did the flow field affect its dispersion, but the dense gas also affected the evolution of the flow field. Contours of velocity magnitude and streamlines for case 1 at t = 15 s are shown in Fig. 14. The effect mainly displayed in two aspects: the vertical momentum of the source and density effect. The former inhibited the formation of large vortex near the source. In the vertical plane, an elliptic vortex was generated at the left side, and a small vortex was located near the jet in the upper layer at the right side. The magnitude distribution of velocity on each side showed the similar characteristics, with the peak of 0.3 m/s near the wall. In the far-source region, the horizontal vortex also disappeared under density effect. The channeling flow was dominant, and the velocity at the windward side was larger than that at the leeward side. Contours of CO2 volume fraction in the vertical plane at x = 0 m and in the horizontal plane at z = 0.025 m are shown in Fig. 15.

As described in Case 1, the vertical vortex resulted in the reverse flow in the lower layer, and CO2 cloud near the ground moved towards the leeward wall. The removal of cloud by the lateral opening resulted from circulation flow and channeling flow was the main pattern. The maximum volume fraction at the windward side was up to 30%, appearing on the area where the outward channeling flow mixed with the horizontal vortex near the right lateral opening. This went beyond the simple feature that the volume fraction on the leeward side was higher than that on the windward side. Increasing leeward-side building height caused the disappearance of the main vertical vortex. Most of CO2 cloud came back, leading to the increase of volume fraction. CO2 volume fractions at the leeward side and the windward side were not so much different, both ranging from 20% to 30%. Whereas, increasing windward-side building height resulted in the enhancement of vertical circulation

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

279

Fig. 16. Average volume fraction and mass concentration of CO2 in the street canyon at t = 15 s.

amount of CO2 moved towards the leeward roof in the upper recirculation zone. The horizontal distribution was governed by the two recirculation regions inside the canyon. CO2 volume fraction on the windward side was higher than that on the leeward side. 3.3. Average volume fraction and mass concentration of CO2 In order to fully understand geometric effect of buildings on CO2 dispersion, average volume fraction and mass concentration of CO2 in the target street canyon were also investigated, as shown in Fig. 16. Note that the street heights for Cases 2, 3, 4 and 6 were set to 0.05 m. The results show that the average volume fraction decreased from 22.9% (equal to 422 g/m3 ) to 14.5% (267 g/m3 ) as the windward-side building height increased. This was mainly because the enhanced vertical circulation flow took more air into the street canyon to dilute CO2 to a low level. The volume fraction in the case of W/H = 3/2 was 15.9% (293 g/m3 ), and the reason could be that the larger area of the top and lateral openings could also take greater amount of air into the canyon. The average volume fraction increased to 27.9% (514 g/m3 ) with leeward-side building height increasing, mainly due to the disappearance of the vortex on vertical dimension. CO2 volume fraction in Case 4 showed a slight decrease of 1.7%. However, with the slanted-shaped roof, the average volume fraction increased to 25.3% (467 g/m3 ). 4. Conclusions

Fig. 15. Contours of CO2 volume fraction in the vertical plane at x = 0 m (left) and the horizontal plane at z = 0.025 m (right) at t = 15 s. White arrows represent wind direction.

flow, thus leading to a higher air exchange rate and a lower CO2 volume fraction. Also, it caused CO2 volume fraction negligibly small on the windward side. In W/H = 2/3 case, CO2 volume fraction in the lower region of windward side was large due to the counterclockwise vortex. A small amount of CO2 could reach the upper region and was slanted to the leeward wall. In the case of W/H = 3/2, the vertical vortices carried CO2 towards the leeward wall, with volume fraction ranging from 10% to 30%. And CO2 volume fraction on the windward side was less than 10%. Contours of volume fraction of CO2 in the canyon with slanted-shaped roof showed that some

This paper focused on the effect of building height, aspect ratio and roof shape on CO2 dispersion within a street canyon. A threedimensional urban model was designed to capture the circulation flow and channeling flow. The results of model evaluation indicate that all the predictions were within a factor of two of the experimental measurements, and the MG was about 7%. CO2 behaved as a dense gas. The volume fraction was more than 30% near the ground and about 5% in the upper region, respectively. The volume fraction of NO remained about 10% in the vertical plane. The volume flux ratio of CO2 dispersed through the top opening was 0.5%, while that of NO was 95.6%. Average volume fraction of CO2 cloud increased from 22.9% (equal to 422 g/m3 ) to 27.9% (514 g/m3 ) with leeward-side building height increasing, mainly due to the disappearance of vertical vortex. Increasing windward-side building height could enhance vertical vortex, thus reduced CO2 volume fraction to 14.5% (267 g/m3 ). Increasing aspect ratio to 3/2 could also take greater amount of air into the canyon, and then reduced CO2 volume frac-

280

W. Tan et al. / Process Safety and Environmental Protection 122 (2019) 271–280

tion to 15.9% (293 g/m3 ). Average volume fraction showed a slight decrease when aspect ratio equal to 2/3. However, CO2 volume fraction increased to 25.3% (467 g/m3 ) with the slanted-shaped roof, and cloud moved towards the leeward roof. The results can also be extended to the real atmospheric boundary layer based on similarity theory. However, the limitation of the experimental results has to be mentioned. Scaling issue of buoyancy force has not been matched between the scaled model and the real scenario. In other words, the results of CO2 in the scaled model may not be applied to the CO2 release in the real cases. Clearly, employing LDV (Laser Doppler Velocimetry) techniques to measure the velocity in wind tunnel experiment and making the velocity validation are necessary in a future research. Meanwhile, more concrete data that quantifies the effect of building geometry is needed to provide meaningful reference for urban planners and emergency responders. Acknowledgements This work was supported by The National Key Research and Development Program of China (Program No. 2016YFC0801900), and Tianjin Science and Technology Planning Project (Project No. 15ZCZDSF00550). Field support from National Center for Fire Engineering Technology was gratefully acknowledged. References Blocken, B., Stathopoulos, T., Carmeliet, J., Hensen, J.L.M., 2011. Application of computational fluid dynamics in building performance simulation for the outdoor environment: an overview. J. Build. Perform. Simu. 4 (2), 157–184. Brambilla, S., Manca, D., 2010. The Viareggio LPG railway accident: event reconstruction and modeling. J. Hazard. Mater. 182 (1–3), 346–357. Briggs, G.A., Britter, R.E., Hanna, S.R., Havens, J.A., Robins, A.G., Snyder, W.H., 2001. Dense gas vertical diffusion over rough surfaces: results of wind-tunnel studies. Atmos. Environ. 35 (13), 2265–2284. Carpentieri, M., Robins, A.G., 2015. Influence of urban morphology on air flow over building arrays. J. Wind Eng. Ind. Aerodyn. 145, 61–74. CCPS, Center for Chemical Process Safety, 1999. Guidelines for Consequence Analysis of Chemical Releases. AIChE, New York. Chang, J.C., Hanna, S.R., 2005. Technical Description and User’s Guide for the BOOT Statistical Model Evaluation Software Package, Version 2.0. Chang, C.H., Meroney, R.N., 2003. Concentration and flow distributions in urban street canyons: wind tunnel and computational data. J. Wind Eng. Ind. Aerodyn. 91 (9), 1141–1154. Cui, P.Y., Li, Z., Tao, W.Q., 2017. Numerical investigations on Re-independence for the turbulent flow and pollutant dispersion under the urban boundary layer with some experimental validations. Int. J. Heat Mass Transf. 106, 422–436. Franke, J., Hellsten, A., Schlunzen, H., Carissimo, B., 2011. The COST732 best practice guideline for CFD simulation of flows in the urban environment a summary. Int. J. Environ. Pollut. 44, 419–427. Garcia-Sanchez, C., Van Tendeloo, G., Gorle, C., 2017. Quantifying inflow uncertainties in RANS simulations of urban pollutant dispersion. Atmos. Environ. 161, 263–273. Gorle, C., van Beeck, J., Rarnbaud, P., Van Tendeloo, G., 2009. CFD modelling of small particle dispersion: the influence of the turbulence kinetic energy in the atmospheric boundary layer. Atmos. Environ. 43 (3), 673–681. Hanna, S.R., Hansen, O.R., Dharmavaram, S., 2004. FLACS CFD air quality model performance evaluation with Kit Fox, MUST, prairie grass, and EMU observations. Atmos. Environ. 38, 4675–4687. Joshi, P., Bikkina, P., Wang, Q.S., 2016. Consequence analysis of accidental release of supercritical carbon dioxide from high pressure pipelines. Int. J. Greenh. Gas Control. 55, 166–176. Karra, S., Malki-Epshtein, L., Neophytou, M.K.A., 2017. Air flow and pollution in a real, heterogeneous urban street canyon: a field and laboratory study. Atmos. Environ. 165, 370–384. Kim, J.J., Baik, J.J., 2003. Effects of inflow turbulence intensity on flow and pollutant dispersion in an urban street canyon. J. Wind Eng. Ind. Aerodyn. 91 (3), 309–329. Kumar, P., Feiz, A.A., Ngae, P., Singh, S.K., Issartel, J.P., 2015. CFD simulation of shortrange plume dispersion from a point release in an urban like environment. Atmos. Environ. 122, 645–656. Li, X.-X., Britter, R., Norford, L.K., 2016. Effect of stable stratification on dispersion within urban street canyons: a large-eddy simulation. Atmos. Environ. 144, 47–59. Liu, X., Godbole, A., Lu, C., Michal, G., Venton, P., 2015. Study of the consequences of CO2 released from high-pressure pipelines. Atmos. Environ. 116, 51–64. Liu, B., Liu, X., Lu, C., Godbole, A., Michal, G., Tieu, A.K., 2016. Computational fluid dynamics simulation of carbon dioxide dispersion in a complex environment. J. Loss Prevent. Proc. Ind. 40, 419–432.

Liu, S.M., Pan, W.X., Zhang, H., Cheng, X.L., Long, Z.W., Chen, Q.Y., 2017a. CFD simulations of wind distribution in an urban community with a full-scale geometrical model. Build. Environ. 117, 11–23. Liu, X., Godbole, A., Lu, C., Michal, G., 2017b. Investigation of terrain effects on the consequence distance of CO2 released from high-pressure pipelines. Int. J. Greenh. Gas Control 66, 264–275. Madalozzo, D.M.S., Braun, A.L., Awruch, A.M., Morsch, I.B., 2014. Numerical simulation of pollutant dispersion in street canyons: geometric and thermal effects. Appl. Math. Model. 38 (24), 5883–5909. Manca, D., Brambilla, S., 2010. Complexity and uncertainty in the assessment of the Viareggio LPG railway accident. J. Loss Prevent. Proc. Ind. 23 (5), 668–679. Merlier, L., Kuznik, F., Rusaouen, G., Salat, S., 2018. Derivation of generic typologies for microscale urban airflow studies. Sustain. Cities Soc. 36, 71–80. Mo, Z.W., Liu, C.-H., 2018. Wind tunnel measurements of pollutant plume dispersion over hypothetical urban areas. Build. Environ. 132, 357–366. Nosek, S., Kukacka, L., Jurcakova, K., Kellnerova, R., Janour, Z., 2017. Impact of roof height non-uniformity on pollutant transport between a street canyon and intersections. Atmos. Environ. 227, 125–138. O’Neill, J.J., Cai, X.-M., Kinnersley, R., 2016. Stochastic backscatter modelling for the prediction of pollutant removal from an urban street canyon: a large-eddy simulation. Atmos. Environ. 142, 9–18. Oke, T.R., 1988. Street design and urban canopy layer climate. Energy Build. 11 (1–3), 103–113. Park, S.J., Kim, J.J., Kim, M.J., Park, R.J., Cheong, H.B., 2015. Characteristics of flow and reactive pollutant dispersion in urban street canyons. Atmos. Environ. 108, 20–31. Pontiggia, M., Derudi, M., Busini, V., Rota, R., 2009. Hazardous gas dispersion: a CFD model accounting for atmospheric stability classes. J. Hazard. Mater. 171 (1–3), 739–747. Pontiggia, M., Derudi, M., Alba, M., Scaioni, M., Rota, R., 2010. Hazardous gas release in urban areas: assessment of consequences through CFD modelling. J. Hazard. Mater. 176 (1–3), 589–596. Scargiali, F., Di Rienzo, E., Ciofalo, M., Grisafi, F., Brucato, A., 2005. Heavy gas dispersion modelling over a topographically complex mesoscale - A CFD based approach. Process Saf. Environ. 83, 242–256. Scargiali, F., Grisafi, F., Busciglio, A., Brucato, A., 2011. Modeling and simulation of dense cloud dispersion in urban areas by means of computational fluid dynamics. J. Hazard. Mater. 197, 285–293. Scungio, M., Arpino, F., Cortellessa, G., Buonanno, G., 2015. Detached eddy simulation of turbulent flow in isolated street canyons of different aspect ratios. Atmos. Pollut. Res. 6 (2), 351–364. Soulhac, L., Lamaison, G., Cierco, F.-X., Ben Salem, N., Salizzoni, P., Mejean, P., Armand, P., Patryl, L., 2016. SIRANERISK: modelling dispersion of steady and unsteady pollutant releases in the urban canopy. Atmos. Environ. 140, 242–260. Tan, W., Li, C.J., Wang, K., Zhu, G.R., Wang, Y., Liu, L.Y., 2018a. Dispersion of carbon dioxide plume in street canyons. Process Saf. Environ. 116, 235–242. Tan, W., Wang, K., Li, C.J., Liu, L.Y., Wang, Y., Zhu, G.R., 2018b. Experimental and numerical study on the dispersion of heavy gases in urban environments. Process Saf. Environ. 116, 640–653. Tauseef, S.M., Rashtchain, D., Abbasi, S.A., 2011. CFD-based simulation of dense gas dispersion in presence of obstacles. J. Loss Prevent. Proc. Ind. 24 (4), 371–376. Toja-Silva, F., Pregel-Hoderlein, C., Chen, J., 2018. On the urban geometry generalization for CFD simulation of gas dispersion from chimneys: comparison with Gaussian plume model. J. Wind Eng. Ind. Aerodyn. 177, 1–18. Tominaga, Y., Mochida, A., Yoshie, R., Kataoka, H., Nozu, T., Yoshikawa, M., Shirasawa, T., 2008. AIJ guidelines for practical applications of CFD to pedestrian wind environment around buildings. J. Wind Eng. Ind. Aerodyn. 96 (10–11), 1749–1761. Tutkun, M., 2017. Markovian properties of velocity increments in boundary layer turbulence. Physica D 351, 53–61. Uehara, K., Wakamatsu, S., Ooka, R., 2003. Studies on critical Reynolds number indices for wind-tunnel experiments on flow within urban areas. Bound.-Layer Meteorol. 107 (2), 353–370. Venetsanos, A.G., Huld, T., Adams, P., Bartzis, J.G., 2003. Source, dispersion and combustion modelling of an accidental release of hydrogen in an urban environment. J. Hazard. Mater. A105, 1–25. Wen, J.X., Le Fur, P., Jie, H., Rao, V.C.M., 2016. Further development and validation of CO2 FOAM for the atmospheric dispersion of accidental releases from carbon dioxide pipelines. Int. J. Greenh. Gas Control 52, 293–304. Wingstedt, E.M.M., Osnes, A.N., Akervik, E., Eriksson, D., Reif, B.A.P., 2017. Large-eddy simulation of dense gas dispersion over a simplified urban area. Atmos. Environ. 152, 605–616. Witlox, H.W.M., Harper, M., Oke, A., Stene, J., 2014. Validation of discharge and atmospheric dispersion for unpressurised and pressurised carbon dioxide releases. Process Saf. Environ. 92 (1), 3–16. Xing, J., Liu, Z.Y., Huang, P., Feng, C.G., Zhou, Y., Zhang, D.P., Wang, F., 2013. Experimental and numerical study of the dispersion of carbon dioxide plume. J. Hazard. Mater. 256, 40–48. Yu, H.S., The, J., 2016. Validation and optimization of SST k-omega turbulence model for pollutant dispersion within a building array. Atmos. Environ. 145, 225–238. Yuan, C., Ng, E., Norford, L.K., 2014. Improving air quality in high-density cities by understanding the relationship between air pollutant dispersion and urban morphologies. Build. Environ. 71, 245–258. Zhang, N., Ni, X.Y., Huang, H., Duarte, M., 2017. Risk-based personal emergency response plan under hazardous gas leakage: optimal information dissemination and regional evacuation in metropolises. Physica A 473, 237–250.